Reliability assessment of the EN 1992 restrained shrinkage

crack model as applied to liquid retaining structures in South

Africa

by
Esther Mwamba

Submitted in fulfilment of the academic requirements of

Master of Science
In Civil engineering
College of Agriculture, Engineering and Science
University of KwaZulu-Natal
Durban
South Africa
2nd December 2016
Supervisor: Mrs C.H. McLeod
 i

Declaration

The research contained in this dissertation was completed by the candidate while based
in the discipline of civil engineering, of the College of Agriculture, Engineering and
Science, University of KwaZulu-Natal, Howard College, South Africa. The contents of
this work have not been submitted in any form to another university and, except where
the work of others is acknowledged in the text, the results reported are due to
investigations by the candidate.

_________________________

Signed: (Candidate: Esther Mwamba. Student Number: 211505003)

_________________________

Signed: (Supervisor)

Date:
 ii

Abstract

The South African design of liquid retaining structures (LRS) has traditionally involved the use
of the British codes of practice, namely BS 8007:1987 and BS 8110-2:1985, due to South Africa
not yet having developed its own equivalent code. BS 8007:1987 and BS 8110-2:1985 have since
been replaced by EN 1992-3:2006 and EN 1992-1-1:2004 respectively. South African engineers
are presented with the option of adopting the Eurocode 2 (EN 1992) design code for the design
of LRS in place of the superseded corresponding British design codes; however, in the case of
adoption, the issue of the code’s suitability for use under local conditions and thus its reliability
requires investigation. Hence, an investigation into the reliability performance of the EN 1992
crack model as applied in the South African context will be undertaken. Cracking, a serviceability
limit state, takes precedence over the effects of the ultimate limit state where the infringement of
crack limits in liquid retaining structures may result in the loss of structural integrity.

The First Order Reliability Method (FORM) of analysis was the probabilistic method of choice
in this investigation. This research focussed on cracking due to restrained deformation with edge
and end restraint conditions both being considered. The influence of significant parameters of the
crack model was assessed in probabilistic terms. Model uncertainty and the restraint factor were
both found to have borne the most influence on the reliability performance of the crack model.
This research aimed to improve the reliability of the EN 1992 crack model for use in the South
African context. This was achieved through attaining an understanding of the influence held by
respective design variables on the crack model, thus bringing to light where within the crack
models sensitivities lay. This then indicated the potentially most effective ways in which
reliability compliance could be brought about in the case of code calibration. Future research must
be conducted on the stochastic nature of the restraint factor and other basic variables. Research
must also be conducted into the model uncertainty for crack formation.
 iii

Acknowledgement

This dissertation could not have been completed without the help and continued support of my
supervisor Mrs C. H. McLeod, your tireless work and knowledge have been greatly appreciated.
A huge thank you is also extended towards my colleagues who were always there to lend an ear
and help out where they could. Lastly, I’d like to extend my sincerest thanks to my family and
friends- you all have played a part in making this experience all the more enjoyable and rewarding.
 iv

Table of Contents

Declaration ..................................................................................................................................... i

Abstract ......................................................................................................................................... ii

Acknowledgement ....................................................................................................................... iii

Table of Contents ......................................................................................................................... iv

List of Figures ............................................................................................................................ viii

List of Tables ............................................................................................................................. xiii

List of Symbols ........................................................................................................................... xv

Chapter 1: Introduction ................................................................................................................. 1

1.1 Introduction ................................................................................................................... 1

1.2 Aim................................................................................................................................ 2

1.3 Objectives...................................................................................................................... 2

1.4 Outline of Thesis ........................................................................................................... 3

Chapter 2: Review of Restrained Cracking in Liquid Retaining Structure Design ....................... 4

2.1 Liquid Retaining Structures in South Africa ................................................................. 4

2.1.1 Introduction ................................................................................................................. 4

2.1.2 Design of Liquid Retaining Structures in South Africa ........................................ 5

2.2 Restrained Cracking ...................................................................................................... 6

2.2.2 Restraint Conditions .............................................................................................. 7

2.2.3 Restraint Degree .................................................................................................. 11

2.3 Review of Design Codes ............................................................................................. 14

2.3.2 BS cracking model .............................................................................................. 16

2.3.3 Eurocode cracking model .................................................................................... 23

2.3.4 Issues Surrounding Crack Width Estimation ...................................................... 28

2.4 Autogenous healing ..................................................................................................... 33

2.5 Conclusion .................................................................................................................. 35

Chapter 3: Structural Reliability ................................................................................................. 37

 v

3.1 Introduction ................................................................................................................. 37

3.2 Limit State ................................................................................................................... 37

3.3 Basic Reliability Theory ............................................................................................. 38

3.4 Reliability Index and Target Reliability ...................................................................... 41

3.5 The First Order Reliability Method ............................................................................. 46

3.6 Statistical Parameters of the EN 1992 Restrained Cracking Serviceability Limit State
48

3.6.1 Introduction ......................................................................................................... 48

3.6.2 Model uncertainty (θ) .......................................................................................... 49

3.6.3 Concrete cover (c) ............................................................................................... 53

3.6.4 Limiting crack width (wlim) ................................................................................. 54

3.6.5 Concrete tensile strength (fct,eff) ........................................................................... 55

3.6.6 Restraint degree/factor (R) .................................................................................. 56

3.7 Partial Safety Factors .................................................................................................. 57

3.8 Previous Research on the Reliability of the Cracking Serviceability Limit State ....... 60

3.9 Concluding Remarks ................................................................................................... 62

Chapter 4: Parametric Study of the EN 1992-1-1 & 3 and Corresponding Codes BS 8007 & BS
8110-2 Restrained Shrinkage Crack Models ............................................................................... 63

4.1 Introduction ................................................................................................................. 63

4.2 Design Parameters:...................................................................................................... 63

4.2.1 Material parameters ............................................................................................. 64

4.2.2 Physical parameters ............................................................................................. 66

4.3 Methodology for Crack Width Estimation .................................................................. 67

4.3.1 Crack Estimation Following EN 1992 ................................................................ 67

4.3.2 Crack Estimation Following BS 8007 and BS 8110-2 ........................................ 68

4.4 Influence of cover versus φ/ρp, eff .............................................................................. 68

4.4.1 Influence of Cover on Crack Spacing ................................................................. 69

4.4.2 Influence of the φ/ρp, eff Ratio on Cracking....................................................... 71

4.5 Depth of effective area ................................................................................................ 73

 vi

4.5.1 Influence of Section Thickness and Concrete Cover on Effective Tension Depth
74

4.5.2 Influence on Reinforcement Bar Diameter on Effective Tension Area .............. 75

4.6 Comparison of BS 8007 and EN 1992 Edge Restraint Estimation on Crack Width ... 76

4.6.1 Influence of Section Thickness on the EN 1992 Edge Restraint Crack Model ... 76

4.6.2 Influence of Restraint Factor on EN 1992 Edge Restraint Crack Model ............ 78

4.6.3 Influence of Restraint Factor on BS 8007 Edge Restraint Crack Model ............ 79

4.7 Comparison of BS 8007 and EN 1992 End Restraint Estimation of Crack Width ..... 80

4.7.1 Influence of Section Thickness on the EN 1992 End Restraint Crack Model .... 81

4.7.2 Influence of Section Thickness on the BS 8007 End Restraint Crack Model ..... 83

4.8 Parameter Sensitivities in Crack Model: Summary and Concluding Remarks ........... 83

Chapter 5: FORM Analysis of EN 1992 Crack Model: Methodology, Results and Discussion . 87

5.1 Introduction: ................................................................................................................ 87

5.2 Methodology of Reliability Analysis:............................................................................... 88

5.2.1 Reliability Analysis Formation.................................................................................. 88

5.2.2 Representative Liquid Retaining Structure ......................................................... 89

5.2.3 Probabilistic Theoretical Models of Basic Variables .......................................... 89

5.3 Results and Discussion ..................................................................................................... 96

5.3.1 Influence of cover and φ/ρp, eff ................................................................................... 96

5.3.2 Influence of Effective Tension Area ................................................................. 101

5.3.3 Influence of Section Thickness ......................................................................... 104

5.3.4 Influence of Restraint Factor ............................................................................. 108

5.3.5 Influence of Model Uncertainty ........................................................................ 110

5.4 Summary ................................................................................................................... 113

Chapter 6: Sensitivity Analysis of EN 1992 Crack Model: Methodology, Results and Discussion
................................................................................................................................................... 115

6.1 Methodology of the Reverse FORM Analysis: ............................................................ 116

6.1.1 Reliability Analysis Formation................................................................................ 116

6.2 Results and Discussion .............................................................................................. 119

 vii

6.2.1 Sensitivity Factors at Varying Model Uncertainty ............................................ 119

6.2.2 Theoretical Partial Safety Factors ..................................................................... 132

6.2.3 Potential Partial Factors for Code Calibration (Edge vs. End Restraint): ......... 144

6.2.4 Influence of the Choice of Reliability Index (β) ............................................... 146

6.3 Comparison of Results for Deterministic and Probabilistic Analysis ....................... 149

6.4 Conclusion ................................................................................................................ 150

Chapter 7: Final Conclusions .................................................................................................... 153

7.1 Introduction ............................................................................................................... 153

7.2 Literature Review ...................................................................................................... 153

7.3 Parametric study ........................................................................................................ 153

7.4 FORM analysis of EN 1992 ...................................................................................... 154

7.5 Sensitivity Analysis................................................................................................... 154

7.6 Deterministic Versus Probabilistic Approach ........................................................... 155

7.7 Recommendations ..................................................................................................... 155

References ................................................................................................................................. 157

Appendix ................................................................................................................................... 165

Appendix A: Deterministic Parametric Study........................................................................... 166

Appendix B: Data for FORM Analysis ..................................................................................... 175

Appendix C: Sensitivity Analysis of EN 1992 ......................................................................... 183

 viii

List of Figures

Figure 2.1: Relationship between Stress and Strain from Change in Temperature due to Concrete
Hydration (Greensmith, 2005) ...................................................................................................... 7
Figure 2.2: Difference between End and Edge Restraint (Bamforth, 2010). ................................ 8
Figure 2.3: Crack Pattern of Concrete Member Subjected to Edge Restraint (Highways England,
1987) ............................................................................................................................................. 9
Figure 2.4: Crack Pattern of Concrete Member Subjected to End Restraint (Highways England,
1987) ............................................................................................................................................. 9
Figure 2.5: Difference in Cracking between End and Edge Restraint (Figure M.2 of EN 1992-
3:2006) ........................................................................................................................................ 10
Figure 2.6: Crack Pattern of Concrete Member Subjected to Edge and End Restraint (Highways
England, 1987) ............................................................................................................................ 10
Figure 2.7: Change in Degree Of Freedom (Kamali, Svedholm and Johansson, 2013) ............. 11
Figure 2.8: Restraint Level at Centre of Section (ACI, 2002) .................................................... 12
Figure 2.9 Effective Concrete Area (BS 8007:1987) .................................................................. 19
Figure 2.10: Restraint Factors (Figure A.3 of BS 8007:1987) .................................................... 21
Figure 2.11: Typical Cases of Effective Concrete Area Following (Figure 7.1 of EN 1992-1-
1:2004) ........................................................................................................................................ 27
Figure 2.12: Comparison of the Cover and Bar Slip Terms of the EN 1992 Crack Spacing Formula
with Experimental Data (Kaethner, 2011) .................................................................................. 31
Figure 3.1: Space of Reduced Variates E’ and R’ (as adapted from Ang and Tang (1984)) ...... 39
Figure 3.2: Comparison of Measure Crack Widths against the EN 1992-3 Predicted Crack Widths
for a Concrete Member Restrained Along its Base (Kamali, Svedholm and Johansson, 2013). 51
Figure 3.3: Comparison of Measured Crack Widths to Predicted Crack Widths of BS 8007:1987
and EN 1992-3:2006 (Bamforth, Shave & Denton, 2011) .......................................................... 52
Figure 3.4: Example of Observed Crack Pattern and their Correlating Restraint Factor (Kamali et
al., 2013). .................................................................................................................................... 54
Figure 3.5: Probability Distribution of the Early Age In-Situ Tensile Strength of C30/37 Concrete
(Bamforth, 2010). ........................................................................................................................ 55
Figure 3.6: Variation of ACI Calculated Restraint Degree with Change in Concrete Elastic
Modulus at Early Age (A0/An = 1), (Bamforth et al., 2010). ...................................................... 57
Figure 4.1: Influence of Cover on Crack Spacing for Both EN 1992 and BS 8007 ................... 71
Figure 4.2: The influence of φ/ρp, eff (or φ/ρ) Ratio on Crack Spacing for BS 8007 and EN 1992-
1-1 (40 mm Cover and 250 mm Section Thickness). .................................................................. 73
 ix

Figure 4.3: Influence of Section Thickness on Crack Width for Edge Restrained Crack Model
(EN 1992).................................................................................................................................... 77
Figure 4.4: Influence of Reinforcement Area on Crack Width for Edge Restrained Crack Model
(BS 8007) .................................................................................................................................... 78
Figure 4.5: Influence of Restraint on the Edge Restrained Crack Model (EN 1992) ................. 79
Figure 4.6: Influence of Restraint on the Edge Restrained Crack Model (BS 8007) .................. 80
Figure 4.7: Comparing EN 1992 and BS 8007 End Restraint Equation (40 mm cover, 75 mm
reinforcement spacing) ................................................................................................................ 81
Figure 4.8: Influence of Section Thickness on Crack Width for End Restraint (EN 1992) ........ 82
Figure 5.1: Summary of Statistical Parameters ........................................................................... 91
Figure 5.2: First and Second Iteration of FORM Analysis of EN 1992 Restrained Shrinkage Crack
Model (Edge Restraint, hc,eff = 2.5(c + ϕ/2)) ............................................................................... 94
Figure 5.3: Example of Convergence Achieved After Eight Iterations (Edge Restraint, hc,eff =
2.5(c + ϕ/2)) ................................................................................................................................ 95
Figure 5.4: Influence of Cover and φ/ρp, eff (Edge Restraint) ....................................................... 98
Figure 5.5: Influence of φ/ρp, eff Ratio on Reliability Index (Edge Restraint) .............................. 99
Figure 5.6: Influence of Cover and φ/ρp, eff (End Restraint) ...................................................... 100
Figure 5.7: Influence of φ/ρp, eff Ratio on Reliability Index (End Restraint) ............................. 101
Figure 5.8 Influence of Effective Tension Area (Edge Restraint) ............................................ 103
Figure 5.9: Influence of Effective Tension Area (End Restraint) ............................................. 104
Figure 5.10: Influence of Section Thickness (Edge Restraint) ................................................. 106
Figure 5.11: Influence of Section Thickness (End Restraint) ................................................... 108
Figure 5.12: Influence of Restraint Factor (Edge Restraint) ..................................................... 110
Figure 5.13: Influence of Model Uncertainty (Edge Restraint) ................................................ 111
Figure 5.14: Influence of Model Uncertainty (End Restraint) .................................................. 112
Figure 6.1: First and Second Iteration of the Reverse FORM Analysis of EN 1992 Crack Model
................................................................................................................................................... 117
Figure 6.2: Example of Convergence Achieved After Seven Iterations (Edge Restraint, h c,eff =
2.5(c + ϕ/2)) .............................................................................................................................. 118
Figure 6.3: Edge Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ........................................................................... 121
Figure 6.4: Edge Restraint Sensitivity of Model Uncertainty (θ) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ........................................................................... 121
Figure 6.5: Edge Restraint Sensitivity of Section Thickness (h) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = h/2) ........................................................................................ 123
 x

Figure 6.6: Edge Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = h/2) ......................................................................................... 123
Figure 6.7: Edge Restraint Sensitivity of Model Uncertainty (θ) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = h/2) ........................................................................................ 124
Figure 6.8: End Restraint Sensitivity of Section Thickness (h) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ........................................................................... 127
Figure 6.9: End Restraint Sensitivity of Concrete Cover (c) with Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ........................................................................... 127
Figure 6.10: End Restraint Sensitivity of the Effective Concrete Tensile Strength (f ct,eff) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2)) .............................. 128
Figure 6.11: End Restraint Sensitivity of Model Uncertainty (θ) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2))............................................................................. 128
Figure 6.12: End Restraint Sensitivity of Section Thickness (h) for Varying Model Uncertainty
Coefficient of Variance (hc,eff = h/2) ......................................................................................... 130
Figure 6.13: End Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc,eff = h/2) ......................................................................................... 130
Figure 6.14: End Restraint Sensitivity of the Effective Concrete Tensile Strength (fct,eff) for
Varying Model Uncertainty Coefficient of Variance (hc,eff = h/2) ............................................ 131
Figure 6.15: End Restraint Sensitivity of Model Uncertainty (θ) for Varying Model Uncertainty
Coefficient of Variance (hc,eff = h/2) ......................................................................................... 131
Figure 6.16: Edge Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for Varying
Model Uncertainty Coefficient of Variance (hc, eff =2.5(c +φ/2)) .............................................. 134
Figure 6.17: Edge Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =2.5(c +φ/2)) ................................ 134
Figure 6.18: Edge Restraint Theoretical Partial Safety Factors of Section Thickness (h) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2) ........................................... 136
Figure 6.19: Edge Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for Varying
Model Uncertainty Coefficient of Variance (hc, eff = h/2) .......................................................... 136
Figure 6.20: Edge Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2) ........................................... 137
Figure 6.21: End Restraint Theoretical Partial Safety Factors of Section Thickness (h) for Varying
Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ............................................. 139
Figure 6.22: End Restraint Partial Safety Factors of Concrete Cover (c) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ........................................................ 139
Figure 6.23: End Restraint Theoretical Partial Safety Factors of the Effective Concrete Tensile
Strength (fct,eff) for Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2)) 140
 xi

Figure 6.24: End Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2)) ............................... 140
Figure 6.25: End Restraint Theoretical Partial Safety Factors of Section Thickness (h) for Varying
Model Uncertainty Coefficient of Variance (hc, eff =h/2)........................................................... 142
Figure 6.26: End Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for Varying
Model Uncertainty Coefficient of Variance (hc, eff =h/2) .......................................................... 143
Figure 6.27: End Restraint Theoretical Partial Safety Factors of the Effective Concrete Tensile
Strength (fct,eff) for Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2) ............. 143
Figure 6.28: End Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =h/2) ............................................. 144
Figure A.1: Edge Restraint Crack Model Inputs for Deterministic Analysis ........................... 167
Figure A.2: End Restraint Crack Model Inputs for Deterministic Analysis ............................. 168
Figure A.3: BS 8007 and EN 1992 Data for Varying Concrete Cover Value .......................... 169
Figure A.4: BS 8007 and EN 1992 Data for Varying ϕ/ρp,eff ratio ........................................... 169
Figure A.5: EN 1992 Data for Edge Restraint with Varying Section Thickness ..................... 170
Figure A.6: EN 1992 Data for End Restraint Crack Model with Varying Section Thickness .. 171
Figure A.7: BS 8007 Edge and End Restraint Crack Model Data ............................................ 172
Figure A.8: EN 1992 Edge Restraint Data with Varying Restraint Factor .............................. 173
Figure A.9: BS 8007 Edge Restraint Data with Varying Restraint Factor ................................ 174
Figure B.1: Edge Restraint MATLAB Input for FORM Analysis (hc,eff = 2.5(c + ϕ/2)) .......... 176
Figure B.2: End Restraint MATLAB Input for FORM Analysis (hc,eff = 2.5(c + ϕ/2)) ............ 177
Figure B.3: Edge Restraint MATLAB Input for FORM Analysis (hc,eff = h/2)......................... 178
Figure B.4: End Restraint MATLAB Input for FORM Analysis (hc,eff = h/2) .......................... 179
Figure B.5: Selected Data of EN 1992 Edge Restraint Crack Model FORM Analysis (hc,eff = 2.5(c
+ ϕ/2) and h/2- Effective Depth Comparison) .......................................................................... 180
Figure B.6: Selected Data of EN 1992 End Restraint Crack Model FORM Analysis (h c,eff = 2.5(c
+ ϕ/2) and h/2- Effective Depth Comparison) .......................................................................... 181
Figure B.7: Effect of Variation in Elastic Modulus of Concrete............................................... 182
Figure C.1: EN 1992 Edge Restraint Crack Model Sensitivity Factors and Theoretical Partial
Safety Factors (hc,eff = 2.5(c + ϕ/2)).......................................................................................... 184
Figure C.2: EN 1992 Edge Restraint Crack Model Sensitivity Factors and Theoretical Partial
Safety Factors (hc,eff = h/2) ........................................................................................................ 185
Figure C.3: EN 1992 End Restraint Crack Model Sensitivity Factors and Theoretical Partial
Safety Factors (hc,eff = 2.5(c + ϕ/2)).......................................................................................... 186
Figure C.4: EN 1992 End Restraint Crack Model Sensitivity Factors and Theoretical Partial
Safety Factors (hc,eff = h/2) ........................................................................................................ 187
 xii

Figure C.5: EN 1992 Edge and End Restraint Crack Model Sensitivity Factors and Theoretical
Partial Safety Factors with Varying β Values (hc,eff = 2.5(c + ϕ/2) ........................................... 188
Figure C.6: EN 1992 Edge and End Restraint Crack Model Sensitivity Factors and Theoretical
Partial Safety Factors with Varying β Values (hc,eff = h/2) ........................................................ 189
 xiii

List of Tables

Table 2.1: External Restraint Degrees for Various Restraint Conditions (Bamforth, 2007) ...... 13
Table 2.2: Restraint Factors at Centreline of Slab (Table A.3 of BS 8007:1987) ....................... 21
Table 2.3: Restraint Factors (Table 3.3 of BS 8110-2:1985) ...................................................... 22
Table 2.4: Exposure conditions to EN 1992-1-[Link] Table 7.1............................................... 23
Table 2.5: Permissible Crack Widths for Autogenous Healing (Edvardsen, 1999) .................... 34
Table 3.1: Relationship between Failure Probability and Reliability Index JCSS Part 1 (Joint
Committee of Structural Saftey, 2001). ...................................................................................... 41
Table 3.2: Ultimate Limit State Target Reliability Indices and Related Failure Probabilities for a
1 Year Reference Period (Joint Committee of Structural Saftey, 2001). .................................... 41
Table 3.3: Irreversible Serviceability Limit State Target Reliability Indices and Related Failure
Probabilities for a 1 Year Reference Period (Joint Committee of Structural Saftey, 2001). ...... 42
Table 3.4: Suggested Reliability Classes and Recommended Minimum Values for Reliability
Index β from EN 1990 for Ultimate Limit State, Fatigue and Serviceability Limit State (Holický,
2009). .......................................................................................................................................... 43
Table 3.5: ISO 2394 Lifetime Target Reliability Indices (Holický, 2009). ................................ 44
Table 3.6: Design Working Life as Described in SANS 10160-1:2011 ..................................... 45
Table 3.7: Theoretical Models of Basic Variables in EN 1992 Crack Model (Holický, 2009). . 49
Table 4.1: Influence of Cover on EN 1992 Crack Spacing Model ............................................. 69
Table 4.2: The Influence of the φ/pρ, eff Ratio on Crack Spacing as per EN 1992....................... 72
Table 4.3: Influence of section thickness and cover on effective depth of tension zone (φ = 16
mm) ............................................................................................................................................. 74
Table 4.4 Influence of bar diameter on effective depth (h = 250 mm) ....................................... 76
Table 5.1: Feasible limit for reinforcement for select section thicknesses (minimum bar spacing
75 mm) ........................................................................................................................................ 96
Table 5.2: Change of k Coefficient with Increasing Section Thickness (by interpolation) ...... 107
Table 6.1: Sensitivity Factors of Random Variables for Edge Restraint Crack Model (βt = 1.5, hc,
eff = 2.5(c +φ/2)) ........................................................................................................................ 120
Table 6.2: Sensitivity Factors of Random Variables for Edge Restraint Crack Model (βt = 1.5, hc,
eff = h/2) ..................................................................................................................................... 122
Table 6.3: Sensitivity Factors of Random Variables for End Restraint Crack Model (β t = 1.5, hc,
eff = 2.5(c +φ/2)) ....................................................................................................................... 126
Table 6.4: Sensitivity Factors of Random Variables for End Restraint Crack Model (β t = 1.5, hc,
eff = h/2) ..................................................................................................................................... 129
 xiv

Table 6.5: Theoretical Partial Factors of Random Variables for Edge Restraint Crack Model (βt =
1.5, hc, eff =2.5(c +φ/2)) ............................................................................................................. 133
Table 6.6: Theoretical Partial Factors of Random Variables for Edge Restraint Crack Model (βt =
1.5, hc, eff =h/2) ........................................................................................................................... 135
Table 6.7: Theoretical Partial Factors of Random Variables for βt = 1.5 (hc, eff =2.5(c +φ/2)) .. 138
Table 6.8: Theoretical Partial Factors of Random Variables for βt = 1.5 (hc, eff =h/2) ............... 141
Table 6.9: Influence of Reliability Index on the Basic Variables of the EN 1992 Edge Restraint
Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = 2.5(c + ϕ/2)) ................... 146
Table 6.10: Influence of Reliability Index on the Basic Variables of the EN 1992 Edge Restraint
Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = h/2) ................................. 147
Table 6.11: Influence of Reliability Index on the Basic Variables of the EN 1992 End Restraint
Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = 2.5(c + ϕ/2)) ................... 148
Table 6.12: Influence of Reliability Index on the Basic Variables of the EN 1992 End Restraint
Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = h/2) ............................... 148
Table 6.13: Comparison of Deterministic and Probabilistic Analysis for wlim = 0.3, 0.2, and 0.1
mm (hc,eff = 2.5(c + ϕ/2), h = 250 mm, Model Uncertainty CoV = 0.3 and βt = 1.5) ................ 150
 xv

List of Symbols

Symbol Explanation

Ac Concrete area
Ac,eff Effective tension area of concrete
As Steel reinforcement
αe Modular ratio
αT,c Coefficient of thermal expansion of mature concrete
c Concrete cover
ɛca Autogenous shrinkage
εcd Drying shrinkage
εctu Tensile strain capacity of concrete
ε, εr Restrained strain
Es modulus of elasticity of the steel reinforcement
fb average bond strength between concrete and steel
fctm (t), fct,eff Concrete tensile strength at time 't'
h Section thickness
hc,eff Effective depth of tension area
k1 coefficients that accounts for the bonding properties of reinforcement in the concrete
k2 coefficient that accounts for the distribution of strain
kc coefficient for stress distribution
k coefficient that accounts for the effect of self-equilibrating stresses
φ Bar diameter
R Restraint degree
Sr, m mean crack spacing
Sr,max Maximum crack spacing
T1 Fall in temperature from the hydration peak and ambient
T2 Fall in temperature because of season variations
wlim Crack width limit

Reliability

αc Sensitivity factor/direction cosine of concrete cover

αh Sensitivity factor/direction cosine of section thickness
αfct,eff Sensitivity factor/direction cosine of concrete tensile strength
αθ Sensitivity factor/direction cosine of model uncertainty
μx, σx mean and standard deviation of random variables
 xvi

β Reliability index
βt Target reliability index
γc Theoretical partial safety factor of concrete cover
γh Theoretical partial safety factor of section thickness
γfct,eff Theoretical partial safety factor of concrete tensile strength
γθ Theoretical partial safety factor of model uncertainty
 1

Chapter 1: Introduction

1.1 Introduction

Cracking is an expected phenomenon in concrete structures and has been described in Eurocode
1992-1-1: 2004 as being a normal part of reinforced concrete structures subjected to various
loading conditions experienced within its lifetime. Cracking is regarded as a serviceability limit
state problem (where generally its effects on the structural integrity are secondary to those of
ultimate limit state conditions). However, this otherwise secondary problem of cracking becomes
more pressing when dealing with structures where the formation of cracks is detrimental to the
structure’s function. Structures of this nature include liquid retaining structures where
permeability of the structure is an important design criterion. Thus cracking, a serviceability limit
state, becomes the dominant limit state in liquid retaining structure design.

Historically, South African codes of practice for the design of engineering structures have been
based on the British standards for design. In dealing with water retaining structures, where no
equivalent code of practice had been developed in South Africa, the British code BS 8007:1987
and those relevant parts of BS 8110-2:1985 were adopted as they stood. These codes have since
been withdrawn and superseded by Eurocode EN 1992-3: 2006 and EN 1992-1-1:2004. If South
Africa were to go on to adopt those parts of EN 1992-1-1:2004 pertaining to cracking and EN
1992-3:2006 for the design of liquid retaining structures, the question of its performance against
South African reliability requirements comes into effect. It is this concern that warrants the
reliability assessment of the EN 1992 restrained strain crack model (through which the reliability
performance of the crack model may be gauged) for the design of liquid retaining structures under
South African conditions.

The reliability of a structure may be described as the extent to which the structure performs as
designed by the engineer for its intended design life (Green & Bourne, 1972). Not only is it
important to determine the failure probabilities of structures for safety reasons, knowing the
reliability of a structure can prevent dire financial loss. There exists some level of uncertainty in
any engineering undertaking as complete structural reliability cannot be guaranteed. These
uncertainties may be measured and assessed through probability methods of analysis (Holicky,
2009). The First Order Reliability Method (FORM) – said to be the most dependable
computational method for structural reliability analysis (Zhao and Ono, 1999) – was the reliability
assessment methodology employed in this investigation.
 2

1.2 Aim

The aim of this research was to undertake a reliability assessment of the presently adopted
European crack model as outlined in codes of practice EN 1992-1-1: 2004 and EN 1992-3:2006
for cracks due to the restrained deformation of concrete applied to the design of liquid retaining
structures in South Africa. The findings made in this investigation would work towards improving
the reliability of the EN 1992 design code where it is used for the design of liquid retaining
structures in South Africa.

1.3 Objectives

The principal objective is to assess the performance of the reliability of the Eurocode 2 crack
model against the reliability targets and reliability performance requirements stipulated in the
South African design codes for the irreversible serviceability limit state of cracking. This may be
achieved through:
i) Establishing the influence of key identified design parameters on the reliability
performance of the Eurocode 2 crack model as applied to the South African context. A
deterministic (excluding inherent variability of input variables) and a reliability-based
(accounting for inherent variability and uncertainty existing in input variables) parametric
study will be employed for this purpose.
ii) A sensitivity analysis in which greater insight into the relative influence held by the key
identified design parameters on the reliability of the crack model may be determined .

In this way, the reliability of the Eurocode 2 crack model may be improved for use in the South
African environment.

Regarding the restrained shrinkage cracking models, some design parameters of interest include
concrete cover and the reinforcing bar diameter to effective reinforcement ratio. For example, in
BS 8007:1987, the cover was not included in the equation for crack spacing. However, in the now
implemented EN 1992-1-1: 2004, the concrete cover was considered to have a significant
influence on the determination of the crack spacing and ultimately on the estimation of the surface
crack width. Moreover, where the now superseded BS 8007:1987 outlined the same design
approach for both edge and end restraint, EN 1992-3:2006 had completely done away with this
approach- leaving only the methodology for edge restraint as it stood under BS 8007:1987. Also,
EN 1992-3:2006 stipulates more stringent crack width limits than BS 8007:1987. An exploration
 3

into the implication of these and other such significant differences in calculation methodologies
between the above-mentioned design codes on the design of LRS in South Africa was performed
in this research.

1.4 Outline of Thesis

Chapter 1 includes within it the introduction, aims and objectives of this research.
Chapter 2 marks the first half of the literature review in which relevant concepts relating to the
design of LRS in South Africa are explored. The ideology of autogenous healing and issues
around crack width estimation, a background into restrained cracking and a review of the
Eurocode and British design codes for the design of liquid retaining structures were covered. The
historical design of liquid retaining structures in South Africa was also considered.
Chapter 3 deals with the second half of the literature review. It gives a collection of relevant
literature and research pertaining to basic reliability theory. In this chapter FORM, the reliability
method of choice for this thesis was explained and the target reliability index for use in the
reliability analysis selected.
Chapter 4 outlines the methodology for the deterministic parametric study of both the Eurocode
2 and superseded British crack model. Comparisons between the two codes were made. The
design implications of adopting the EN 1992 design code are evaluated. Those input variables
found to bear the most influence on the Eurocode 2 crack model were identified for further
investigation in the reliability analysis. Also, a realistic set of parameters for a representative
liquid retaining structure were thus established for use in the reliability analysis.
Chapter 5 presents the reliability analysis of the EN 1992 crack model as applied to a
representative liquid retaining structure subject to South African conditions. Here the influences
of selected design parameters are investigated through a reliability-based parametric study.
Chapter 6: Those results obtained through the reliability analysis presented in Chapter 5 are
expanded upon through a sensitivity analysis of the random variables of the crack model.
Chapter 7 marks the end of the thesis in which concluding remarks and recommendations for
future research are made.
The Appendices contains relevant data not already included in the main chapters of this
dissertation.
 4

Chapter 2: Review of Restrained Cracking in Liquid Retaining

Structure Design

2.1 Liquid Retaining Structures in South Africa

2.1.1 Introduction
The design of liquid retaining structures (LRS) is one that involves special attention to the
serviceability limit state (which takes precedence over the demands of the ultimate limit states).
Cracking is a serviceability limit state problem, which generally is a secondary concern in the
design of concrete structures. As is commonly understood by all civil engineers, serviceability
limit states deal particularly with the appearance of the structure and its functionality under
working conditions. This is contrary to the ultimate limit states of design which concerns itself
with the collapse of structures and the safety of its occupants and thus has a higher required level
of reliability. However, where cracking forms an integral part of a structure’s design, the problem
of cracking becomes a more pressing issue. Liquid retaining structures are a good example of
structures where cracking is an important part of the structure’s function. At present, South
African engineers tasked with the design of a liquid retaining structure would do so under the
guidance and stipulations of the British codes of practice (Wium, 2008). This is due to the fact
that South Africa had not yet come to develop an equivalent code of its own. The British code of
practice used to design LRS, namely BS 8110-2:1985 and BS 8007: 1987 have since been
superseded by the Eurocodes EN 1992-1-1: 2004 and EN 1992-3:2006 respectively.

At the Structural Eurocode Summit in 2008, held in Pretoria over the issue of the move towards
Eurocode use in South Africa, in a lecture held by Wium (2008) the options of perhaps updating
existing codes for concrete design in South Africa, adopting Eurocode, adapting a foreign code
and using said code as reference; or just developing a new code were put forth. The option to
update was abandoned by a previous working group tasked with this alternative learned of a new
generation of codes being underway (Wium, 2008). Wium (2008) suggested that adopting or
either adapting EN 1992 would require less work and be ideal due to the internationally
widespread use of Eurocode. The development of a new code would be taxing both monetarily
and time wise (Wium, 2008). Thus, regarding liquid retaining structures, it was best to just adopt
Eurocode and incorporate South African nationally determined parameters in a code annex. An
equivalent code for liquid retaining structures, SANS 10100-3, is being developed and has been
said to closely follow the format of EN 1992-3 (Wium, Retief & Barnardo- Viljoen, 2014). SANS
10100-3 will incorporate additional sections, clauses, informative annexes and provisions (some
of which being derived from BS 8007) where necessary (Wium, Retief & Barnardo-Viljoen,
 5

2014). Hence, a reliability assessment of the EN 1992 restrained shrinkage crack model should
contribute to the development of the proposed SANS 10100-3.

2.1.2 Design of Liquid Retaining Structures in South Africa

As previously mentioned, South African engineers generally design LRS in accordance to BS
8007 and BS 8110, with a full transition into the use of Eurocode 2 not yet being realised. This,
along with a few other findings, was made after a survey, which focused particularly on the design
and construction practices of water reservoirs in South Africa (McLeod, 2013). Other noteworthy
findings made from this survey are listed below:
 Whether designed to be rectangular or circular in plan, the specified storage capacity of
the LRS will determine its ultimate size.
 A 40 mm concrete cover to reinforcement is often used for durability. A 50 mm concrete
cover may be used instead where the reservoir is meant to stand against soil.
 Generally, the water reservoirs are constructed with a 35 MPa 28 day concrete strength
with a minimum of 375 kg/m3 good quality binder content. The upper limit for binder
content is set at 500 kg/m3 to control thermal shrinkage. Thermal shrinkage may be
further limited through the avoidance of slagment in cement and through the use of rapid
hardening cement.
 It is also common practice to use concrete with a 0.5 water/cement ratio.
 The permeability restrictions mean that the serviceability requirements of the LRS
dominate the design. This generally translates into more steel reinforcement being
required to meet the crack widths than those required for the fulfilment of ultimate limit
states.
 In South Africa, high-yield steel reinforcement is used with a 450 MPa characteristic
strength.
 LRS may be designed either to be fully or partially buried. Buried reservoirs are normally
built onto excavated land or on a fill embankment. The choice of arrangement depends
ultimately on the design conditions and the client’s own preferences.
 The design of water reservoirs may be broken into three main structural elements: the
wall, floor slabs and foundation as well as roof. Walls are either built as continuous or
joined vertically. In rectangular reservoir walls are usually designed as cantilever walls.
In the case of pre-stressed circular reservoirs in South Africa, the walls are generally
equipped with a sliding joint at the base of the wall stem. For a circular water reservoir
with a capacity of less than 10 Ml, a ratio of diameter to wall height of 4 is usually
employed in South Africa. The floors slabs are normally cast discontinuously with the
walls in the case of larger reservoirs, or may otherwise be cast to be continuous with the
 6

walls. The floors of rectangular reservoirs are constructed with reinforced concrete slabs
that are cast in square panels. A radial pattern is used when casting the floors in circular
reservoirs. The most common roof type used for water reservoirs in South Africa is the
flat slab.

It is obvious that in order for water reservoirs (along with other LRS) to meet their required
function, they must be constructed with a combination of both good design and proper
construction practices.

2.2 Restrained Cracking

Cracking in concrete is a natural phenomenon that occurs may occur due to loading, expansive
chemical reactions in the concrete, plastic settlement just after casting or otherwise due to
restrained deformation (Beeby and Narayanan, 2005). The focus of this thesis is on cracks that
occur from restrained deformation, where the deformation comes after shrinkage and thermal
movement of concrete during hydration. During the hydration reaction the concrete generates heat
at a faster rate than what is lost to the environment, this then increases the temperature in the
concrete. As heat is released from the hydration reaction in concrete, the concrete will expand and
hence be under compression with the steel reinforcement being under tension. The concrete will
then begin to cool to ambient temperature and contract, putting the concrete under tension and the
steel reinforcement in compression (Greensmith, 2005). If the concrete is not restrained in any
way and insulated, the movement induced by the temperature changes will be allowed to take
place (Mosley, Bungey and Hulse, 2012). This allowed movement will result in there being no
changes in concrete stresses. If a restraint is applied onto a concrete member, either from an
external attachment or from temperature differentials within thick concrete members, the free
movement of the concrete will be restricted. This restriction in movement would subsequently
bring about stress changes within the concrete. The introduction of steel reinforcement within the
concrete member would restrain the concrete’s movement during the hydration process Mosley
et al. (2012). Where the concrete reaches a tensile stress that is greater than the tensile capacity
of the concrete, a crack will form.

The Figure 2.1 illustrates the above-mentioned relationship between stress and strain development
in the concrete member as the temperature changes during hydration. It is evident in Figure 2.1
that creep significantly reduces the amount of early age thermal strain and consequently the stress
in the concrete.
 7

Figure 2.1: Relationship between Stress and Strain from Change in Temperature due to
Concrete Hydration (Greensmith, 2005)

2.2.2 Restraint Conditions

The restraint to deformation (or otherwise volume change) of the concrete may occur either
internally, externally or a combination of both (ACI, 2002). Internal restraint takes place where
one part of the concrete section undergoes a change in volume that is different to those parts
adjacent to it. More specifically, one part of the concrete will expand or contract relative to another
part of the concrete section. Thus the relative movement of one part of a concrete section would
be restricted by another. Design guide, CIRIA C660 (Bamforth, 2007) gives an example of such
an occurrence: where in thick sections the concrete core generates heat faster than the concrete’s
surface creating a differential temperature gradient within the concrete section. In this instance,
the core of the concrete section would expand and be in compression whilst the cooler surface of
 8

the concrete section undergoes contraction and experiences tension. Cracks may then form at the
concrete’s surface. As the concrete begins to cool, the core cools faster than the concrete’s
surface– this time contracting whilst the concrete’s surface expands. The crack widths of the
cracks formed on the concrete’s surface subsequently decrease in magnitude.
External restraint may be imposed onto a concrete member through its support conditions. The
support conditions acting on the concrete member prevent the member from fully expanding
during the heating phase. Cases dealt with in both BS 8007 and EN 1992-3 are end restraint and
edge restraint (as illustrated in Figure 2.2).

Figure 2.2: Difference between End and Edge Restraint (Bamforth, 2010).

End restraint comes about where a member is fixed in place along its ends whereas edge
restraint comes after the concrete member being held in place all along its edge. EN 1992-1-1
suggests that the load is transferred entirely to the reinforcement during end restraint. Bamforth
(2010) suggests that end restraint results in the crack width being limited and the number of
cracks that occur being increased.

Edge restraint is assumed to be directly proportional to the strain developed in the concrete.
The crack width then will be increased with an increase in strain since the crack width itself is
directly proportional to the strain and crack spacing. BS 8007 and EN 1992-3 both assume that
the cracking due to edge restraint not only increases the crack width, but also that it has no
influence on the number of cracks formed– the cracks are considered to act independently of
each other (Bamforth, 2010). An illustration of the crack pattern formed from edge restraint is
presented in Figure 2.3. The restraint of the concrete member along the edge provides a
resisting horizontal force that brings about cracking in the mid-span of the concrete member.
The vertical tensile force generated in the concrete to resist the warping in the wall from the
horizontal resisting force produces cracks that spread off diagonally towards the ends of the
concrete member (Highways England, 1987).
 9

Figure 2.3: Crack Pattern of Concrete Member Subjected to Edge Restraint (Highways
England, 1987)

Pure end restraint comes about where the restraining member is short. The distribution of cracks
due to end restraint is illustrated in Figure 2.4:

Figure 2.4: Crack Pattern of Concrete Member Subjected to End Restraint (Highways
England, 1987)

Figure 2.5 illustrates the difference in the cracking between end and edge restraint (as found in
EN 1992-3:2006 annex M). The Y axis represents the crack width, whilst the X axis represents
the imposed deformation. The key to Figure 2.5 are as follows:
 1- The end restraint equation graph.
 2- Cracking due to end restraint,
 3- Cracking due to edge restraint.

Clearly, end restraint may be found to produce crack widths that a larger than those resulting from
concrete members subjected to edge restraint.
 10

Figure 2.5: Difference in Cracking between End and Edge Restraint (Figure M.2 of EN
1992-3:2006)

A combination of both edge and end restraint may occur where a thin section is constructed in an
alternative construction sequence. An advice note of the Design Manual for Roads and Bridges
(Highways England, 1987), puts forth that where the restraining member is up to 5 m, the
restrained member is also under edge restraint and no longer only enduring the effects of end
restraint. A member subjected to both edge and end restraint would result in the above mentioned
typical crack patterns for either restraint conditions (Figures 2.3 and 2.4 respectively) developing
into the pattern presented in Figure 2.6.

Figure 2.6: Crack Pattern of Concrete Member Subjected to Edge and End Restraint
(Highways England, 1987)

External and internal restraint may also act together in the concrete member. An example where
this may occur is the case where alternative panels are constructed for a thick concrete wall. The
 11

design guide for early-age thermal crack control, CIRIA C660 (Bamforth, 2007), explains this
occurrence well. Surface cracks may occur where the concrete’s core generates more heat than at
the surface of the concrete. The cracks on the concrete’s surface are reduced due to the presence
of the external restraint which restricts the expansion of the concrete’s core, subsequently
restricting the potential amount of tension to be developed in the surface zone. Hence the effect
of external restraint cancels that of the internal restraint during concrete heating. However, the
opposite is true when concrete cooling occurs. The concrete’s core will cool faster than the surface
and will thus have its contraction restricted by the surface zone– generating cracks within the
core. This is further aggravated by the effect of the external restraint which also restricts the
contraction of the concrete’s core. Thus, during concrete cooling, the external and internal
restraint work together to restrict movement in the restrained concrete member.

2.2.3 Restraint Degree

The restraint, or otherwise expressed as the degree of freedom of movement, is essentially a ratio
of the actual imposed strain to the imposed strain resulting from full restraint (Antona and
Johansson, 2011):
actual imposed strain
Restrain Factor =
imposed strain in case of full restraint (2.1)

At full restraint the restraint factor will stand at 1 and in instances in which some freedom of
movement is allowed, the restraint factor will be less than 1.
Both EN 1992-3 and BS 8007 prescribe a restraint factor for instances of external restraint in
which the concrete member in question is restrained continuously along its edge. Figure 2.7
illustrates the way in which restraint varies in the case with respect to the geometry of the
restrained member. Those parts of the restrained member farther away from the restraining
member are freer to move and so bear lower degrees of restraint, whilst a greater restriction to
movement may be observed for those parts of the restrained member close to the restrained base.

Figure 2.7: Change in Degree Of Freedom (Kamali, Svedholm and Johansson, 2013)
 12

A methodology for the estimation of the restraint factor was developed by the American Institute
of Concrete (ACI, 2002) for both edge (continuous external restraint) and end restraint conditions
(or otherwise described as being a discontinuous external restraint). For edge restraint, ACI-
207.2R-95 (2002) describes a multiplier to be used in conjunction with the restraint values based
on test data (ACI, 2002). These restraint values obtained (reproduced in Figure 2.8) were related
to the configuration of the concrete member in question, namely the length to height ratio of the
restrained concrete member.

Figure 2.8: Restraint Level at Centre of Section (ACI, 2002)

The ACI-207.2R-95 (2002) given multiplier is described as follows:

1 (2.2)
X= 𝐴𝑔 𝐸𝑐
1+ 𝐴
𝐹 𝐸𝐹

where,
 Ag denotes the gross concrete cross-sectional area of the restrained concrete member (or
otherwise the new concrete pour)
 Ec is the modulus of elasticity of the newly poured concrete
 AF marks the cross-sectional area of the restraining concrete base
 EF represents the modulus of elasticity of the old (restraining) concrete element
 13

The restraint factor calculated from ACI-207.2R-95 (2002) would then be a product of the above-
described multiplier and the applying restraint value read from Figure 2.8. ACI-207.2R-95
suggests that for mass concrete on rock, the maximum effective area of the restraining element
(AF) can be assumed to be 2.5 times greater than the gross cross-sectional area of the restrained
member (Ag). Moreover, it was recommended in design guide CIRIA C660 (Bamforth, 2007) that
an Ec/EF value of 0.7 to 0.8 may be assumed where the ACI approach for restraint estimation is
used. In instances in which cooling in the concrete is found to be most rapid, the lower value of
the Ec/EF ratio should be used (Bamforth, 2007). CIRIA C660 (2007) includes this ACI method
for restraint estimation, adding a factor K1 to account for the effects of creep. Bamforth (2007)
suggests a typical value of K1 = 0.65 for where 35% reduction in the stresses in the concrete is
anticipated to have resulted from the effects of creep.

The true restraints observed from experimental data are listed in BS 8110-2:1985 and HA BD
28/87 (Highways England, 1987) - a supplementary code to BS 5400-4:1984 for bridge design.
The effects of creep are not included in these values. BS 8007 and EN 1992-3 account for the
effect of creep (with a modification factor of 0.5) on the restrained concrete member (Bamforth,
2007). The various restraint levels for a range restraint conditions, after CIRIA C660, are
reproduced in Table 2.1:
Table 2.1: External Restraint Degrees for Various Restraint Conditions (Bamforth, 2007)
Restraint Condition BS 8110-2 HA BD 28/87 BS 8007 EN 1992-3
Base of wall onto a 0.6-0.8 0.6 0.5 0.5
massive base
Top of a wall cast on to a 0.1-0.2 0-0.5 0-0.5
massive base
Edge restraint in box type 0.5
deck cast in stages
Edge element cast onto 0.8
slab
Massive pour cast onto 0.1-0.2 0.2
blinding
Base of massive pour cast 0.3-0.4
onto existing mass
concrete
Suspended slabs 0.2-0.4
Infill bays (e.g. rigid 0.8-1 1 0.5 0.5
restraint)
 14

However, when regarding end restraint, EN 1992-3 does not apply a restraint factor to the
estimation of the restrained strain as BS 8007 did. ACI-207.2R-95 gives a restraint factor to be
used for end restraint (or otherwise, discontinuous external restraint) conditions. The restraint
factor is given by the formula:
1 (2.3)
R=
A h3
1 B
4LI c

in which,
 AB represents the area of the deforming member (the member being restrained)
 h denotes the height of the supporting ends restraining the deforming member
 L is the length of the deforming member (the member being restrained)
 Ic refers to the average moment of inertia of the supporting ends restraining the deformed
member

Much like the definition given for the restraint factor (degree) given by Antona and Johansson
(2011) in Equation 2.1, researcher Gilbert (2016) outlined a rational method to estimate the
restraint degree of the boundary between the wall and concrete base for an edge restrained
concrete element. The restraint degree was said to be estimated using the formula:
R = εr/∆εfree, (2.4)

where εr denotes the restrained strain at the bottom of the wall which is equal to the ratio between
the tensile stress (σcs) at the bottom of the wall (caused by the restraining force acting at some
distance ӯ below the interface of the wall and base) and the age-adjusted effective concrete
modulus (Ēc) of the wall.

Moreover, ∆εfree represents the change in the free contraction at the interface of the wall and base
which is the sum of the strain induced by the changes in the temperature of the concrete, the
autogenous shrinkage as well as the drying shrinkage.

2.3 Review of Design Codes

A review of the parts of design codes used in South Africa for the design of the cracking
serviceability limit state are considered in the subsequent text. The estimation of cracks induced
from the restrained shrinkage was the focus of the review. A look through the South African code
of practice for the structural use of concrete (SANS 10100-1:2000) reveals few references to the
control of crack widths. Formulae to predict the characteristic crack width are given in Annex A
 15

(under section A.3), which covers the methods for the checking of compliance to serviceability
limit state criteria. Annex B gives some general information on the use of movement joints in
controlling cracking. A maximum crack width limit of 0.3 mm is given in clause [Link].1.1 for
the design of concrete elements where the serviceability limit state is not dominant (e.g.
buildings).

Although BS 8007 and BS 8110-2 have both evidently been dependable for use in South Africa,
they have been superseded. The adoption of EN 1992 could, therefore, prove to be beneficial. In
a lecture on the relevance of the Eurocodes in South Africa, Zingoni (2008) listed a few of the
benefits to be had with the Eurocode adoption. These benefits include the added competitiveness
of South African engineers as they will be more equipped to bid for international projects in a
wide variety of countries in which Eurocode adoption had already been established. South African
engineers also stand to gain from international software and design manuals, making for a
smoother transition into complete adoption. Additionally, through the Eurocode’s general
framework and flexibility, the opportunity is available for South African engineers to include
local partial factors and unique geographical and climatic parameters. However, the implications
of this change over into Eurocode use must be quantified. In this way, a full assessment of the
pros and cons of Eurocode adoption in South Africa may be done.

There are several ways in which the British and Eurocode differ in their approach on the
estimation of crack widths formed from restrained shrinkage. The crack spacing equation of
Eurocode 2 include the effects of the concrete cover, whereas the British code of practice does
not. This has come after studies proved that concrete cover plays a significant role in the crack
spacing (Beeby and Narayanan, 2005). Moreover, the ratio of concrete tensile strength for
immature concrete with respect to the bond strength of the steel reinforcement (fct/fb) in BS 8007
has been replaced by the factor k1 in EN 1992-1-1 to account for the bond properties of the
reinforcement used. The estimation of the restrained strain under either edge or end restraint are
dealt with differently between the two codes.

Restrained strain from edge restraint in EN 1992 is dealt with in a similar way to the British codes.
However, historically autogenous shrinkage has been assumed to occur only in concretes with
very low water/cement ratios. For normal strength concretes– where the water/ cement ratio is
greater than 0.4– drying shrinkage is assumed to make up the total measured shrinkage since very
little autogenous shrinkage is said to occur (Addis and Owens, 2001). Eurocode 2, on the other
hand, assumes that autogenous shrinkage comes into effect for all concretes with characteristic
cylinder strength greater than 10 MPa. The greatest difference between both codes lies in their
estimation of restrained strain after end restraint conditions, where completely different
 16

approaches are adopted. It appears that not one parameter is shared between the codes with
regards to this estimation. Even with the differences in approach for most crack models, it is the
general consensus of researchers that the most influential parameters of the crack model – in order
of importance- include the reinforcing steel stress, concrete cover, reinforcement spacing and area
of concrete surrounding each reinforcing bar (Zahalan, 2010). The following section looks closer
into the approaches adopted by both the British and Eurocode cracking models.

2.3.2 BS cracking model

The old British design code of practice for the design and construction of liquid retaining
structures was BS 8007, which covered particularly tanks, reservoirs and other vessels that either
contained or excluded an aqueous liquid (except for the case of aggressive liquids). Liquid
retaining structures designed to BS 8007 were done so together with relevant parts of BS 8110-1
and BS 8110-2.

[Link] Permissible Crack Widths

A limit has been imposed on the maximum design crack width (based on the allowable
permeability for the concrete) for liquid retaining structures depending on the exposure conditions
that are to be endured by the structure. Under the BS 8007 code, it has been recommended that
the maximum design surface crack width be limited to 0.2 mm for severe or very severe exposure
conditions. However, where aesthetic appearance is a matter of concern, a limiting crack width
of 0.1 mm was recommended.

[Link] Minimum Area of Steel

BS 8007 states that after the first crack has formed, the formation of cracks thereafter will be
influenced by the provision of reinforcing in the concrete. The steel reinforcement controls the
distribution of cracks by increasing the number of cracks that form whilst limiting their width
to within the limiting crack width. This occurs where the reinforcement across the initial crack
does not yield.

Where the tensile force experienced by the concrete is beyond the maximum tensile force
capacity of the concrete (Acfct), cracking will occur. The steel reinforcement provided must be
sufficient enough that the resistant tension force of the steel (Asfy) is at least equal to the
maximum tensile force capacity of the concrete (Asfy ≥ Acfct ). For the steel reinforcements to
effectively reduce the crack widths to within the limiting value, the minimum amount of steel
reinforcement in the concrete needs to be as set out by BS 8007:
As = Ac X ρcrit, (2.5)
 17

where,
fct
 ρcrit = (the ratio between the direct tensile strength of the concrete taken at 3 days
fy

and the characteristic strength of the steel reinforcement).

 As is the minimum area of steel
 Ac is the area of concrete effective surface zones which follow the recommendations
listed in figures A.1 and A.2 of Appendix A of BS 8007:1987.

In figure A.1, BS 8007 suggests that the effective tension zone (effective surface zones) for
walls and suspended slabs with thickness ‘h’ less than or equal to 500 mm take up half the
section depth. Where the wall and suspended slab thickness is greater than 500 mm, it is
assumed that each reinforcement face will control 250 mm of the concrete’s depth. figure A.2,
on the other hand, proposes that the effective tension zone of ground floor slabs with thickness
‘h’ under 300 mm will be h/2 on one reinforcement face with no reinforcement required for the
bottom face of the section. Values for the ground slab thicknesses between 300 mm and 500
mm will produce an effective tension zone that is half the section thickness for the top
reinforcement face and 100 mm was recommended for the bottom reinforcement face. Finally,
where the ground slab thickness is found to exceed 500 mm, the surface zone was assumed to
be 250 mm for the top reinforcement face with the bottom reinforcement face set at 100 mm.

[Link] Crack Spacing

A comprehensive discussion of the BS 8007 crack spacing formula is given by Bhatt, Thomas,
McGinley and Choo (2006), a summary of which is presented below:

Slipping between the reinforcement and the concrete begins after the first crack forms. More
cracks will then start occurring where the bond stress (fb) between steel and concrete is greater
than the concrete tensile strength (fct) as such,
fbsƩu ≥ fctAc.
In this inequality‘s’ refers to the development length of bond stress and Ʃu is the total perimeter
of bars at the section. Considering the ratio of the sum of the perimeter of reinforcement bars to
area of reinforcement,
∑u/As=πφ/(πφ2/4),
it is understood that generally the same bar diameter is used at a cross section. The ratio of the
sum of the steel reinforcement perimeter to steel area then becomes:
∑u/As=4/φ
Ultimately the inequality may be rewritten in the form:
fct Φ
s≥ x
fb 4ρ
 18

This describes the minimum crack spacing, with the maximum crack spacing being twice the
minimum (Bhatt et al.,2006). Therefore, the maximum spacing to BS 8007 of the cracks formed
in the concrete is to be determined by the following equation:
fct φ (2.6)
Smax = x ,
fb 2ρ

where:
fct
 The ratio is the relationship between the tensile strength of the concrete and the
fb

average bond strength of the steel reinforcement with respect to the concrete.
 φ is the bar diameter of the steel reinforcement
 And ρ is the ratio of steel based on the effective concrete tension areas defined in
figures A.1 and A.2 of appendix A of BS 8007: 1987 (reproduced here as Figure 2.9)
 19

Figure 2.9 Effective Concrete Area (BS 8007:1987)

 20

[Link] Crack Width Calculation

As per BS 8007 the estimated maximum crack width that can develop in the concrete due to
thermal changes is equated to the product of the maximum crack spacing and the restrained
strain:
wmax = Smax x ɛ (2.7)

where,
 Smax is the maximum spacing of cracks as defined in equation 2.6 of section [Link]
 ɛ is the restrained strain

Here, ɛ, the restrained strain of the concrete is assumed to follow the relationship:
ɛ = R αT,c (T1 + T2), (or otherwise ɛ = 0.5αT,c (T1 + T2) ) (2.8)
where,
 αT,c is the coefficient of thermal expansion of the mature concrete
 R is the restraint factor that ranges from 0 to 0.5 (where creep is accounted for)
 T1 is the drop in temperature from the hydration peak to the ambient temperature
 T2 is the fall in temperature because of season variations

[Link] Restraint Conditions

The restraint that causes cracking may be either internal or external. Internal restraint is
dominant where the concrete member is thick. Various external restraint conditions are given
in figure A.3 of annex A in BS 8007, where the corresponding restraint factor R is given. As is
evident in Figure 2.10, the restraint factor varies with its location within the member, the
member’s proportions, as well as the type of restraint it is subjected to (be it edge or end
restraint). This may be illustrated in Figure 2.10 taken from BS 8007:
 21

Figure 2.10: Restraint Factors (Figure A.3 of BS 8007:1987)

Table 2.2 (initially table A.3 of BS 8007:1987) presents the differences in the restraint factors
from the fixed edge (e.g. the base for a wall slab) of a restrained member to the opposite free edge
(e.g. the top section of a wall slab) of that same restrained member.
 22

Table 2.2: Restraint Factors at Centreline of Slab (Table A.3 of BS 8007:1987)

Ratio L/H Design Centreline Horizontal Restraint Factors
Base of Panel Top of Panel
1 0.5 0
2 0.5 0
3 0.5 0.05
4 0.5 0.3
>8 0.5 0.5
 H is the height or width to the free edge
 L is the distance between full contraction joints
 All values of the restraint factor, except where the restraint is zero at the top panel, may be
less where L < 4.8 m

R = 0.5 is the restraint factor for a ground slab at mid-length cast onto smooth blinding concrete.
This restraint applies for the seasonal change in temperature T2, where the slab length is 30m
or more. In accordance with BS 8007, the restraint factor R = 0.5 is assumed to vary uniformly
from 0.5 to 0 at the ends of the slab.

Some restraint factors based on typical values of restraints that have been recorded for various
pour configurations found in industry have been included in table 3.3 of BS 8110-2:1985,
reproduced here as Table 2.3:

Table 2.3: Restraint Factors (Table 3.3 of BS 8110-2:1985)

Pour Configuration Restraint Factor

Thin wall cast on to massive concrete base 0.6 to 0.8 at base

0.1 to 0.2 at top
Massive pour cast into blinding 0.1 to 0.2
Massive pour cast on to existing mass 0.3 to 0.4 at base
concrete 0.1 to 0.2 at top

Suspended slabs 0.2 to 0.4

Infill bays, i.e. rigid restraint 0.8 to 1
 23

2.3.3 Eurocode cracking model

[Link] Permissible Crack Widths
As with the BS 8007, the permissible crack widths for liquid retaining structures have been
determined for structures depending on their function. The different categories concerning the
degree to which permeability is permitted, as defined by EN1992-3, are as follows:
 Tightness Class 0 are all those structures where some degree of leakage will not be
detrimental to the structures function. Here the permissible crack width will follow the
requirements outlined in clause 7.3.1 of EN 1992-1-1. Under clause 7.3.1 of EN 1992-
1-1 the permissible crack widths are listed in table 7.1 with the limits being:

Table 2.4: Exposure conditions to EN 1992-1-[Link] Table 7.1

Exposure Class Reinforced Members and Pre-stressed
members with Unbonded Tendons
Quasi-permanent Load Combination
X0, XC1 0.4 mm
XC2, XC3, XC4 0.3 mm
XD1, XD2, XD3, XS1, XS2, XS3 0.3 mm

Exposure conditions X0 and XC1 are for where the crack width has no real effect on the
structures durability. Here the crack width limit is set for aesthetic reasons and would not apply
to LRS.
 Tightness Class 1 deals with liquid retaining structures that are allowed to leak to some
extent. There is some surface dampness and surface staining that is allowed to take place.
Where cracks are expected to pass the section thickness, the crack width needs to be limited
to wk1. The crack limit wk1 is based on the ratio of hydrostatic pressure hD to wall thickness
h. If the ratio hD/h ≤ 5 then wk1 = 0.2mm, and where the ratio hD/h ≥ 35 then wk1 =
[Link] of crack width lying somewhere between these ratio may be interpolated.
Where the crack is not expected to pass completely through the section thickness then the
crack width may be limited to those outlined in clause 7.3.1 of EN 1992-1-1.
 Tightness Class 2 covers structures where leakage is to be kept to a minimum and where
surface dampness and surface staining is not permitted. Under this tightness class, cracks
that are expected to pass through the section are to be avoided altogether.
 Tightness Class 3 pertains to structures where no amount of leakage is permissible.

Where these crack limits are met, EN 1992-3 expects that the cracks should be able to heal
themselves under normal changes in temperature and loading in service. This is assumed to
 24

occur where strain under service conditions is expected to range below 150x10-6. It is important
that these crack limits are met so that the self-healing of the cracks is made possible. Where the
self-healing of the reinforced concrete doesn’t take place, it is expected that any crack that
forms will result in leakage.

[Link] Minimum Area of Steel

Under the same school of thought employed for the minimum required reinforcement for crack
control in BS 8007, for the control of crack formation in the concrete the tensile force in the
steel (Fsteel = Asfy) should be at least the tensile force capacity of the concrete (where Fconcrete=
Acfct– is the minimum force necessary to cause cracks to form in the concrete) or otherwise
greater. Where this is satisfied, the steel reinforcement will not yield at the crack-inducing
force. In this way, the steel reinforcement remains within the elastic range, which is an essential
requirement for the validity of the crack width estimation under EN 1992 (Beeby and
Narayanan, 2005). The derivation of the formula for minimum steel reinforcement given in EN
1992 does not account for where the steel reinforcement yields (Beeby and Narayanan, 2005).
The minimum area of reinforcement required to control crack formation is given in EN 1992-
1-1:2004 as follows:
k c f ct,eff A ct (2.9)
As, min =
f ky
where,
 As, min describes the minimum area of reinforcement where the concrete section will be
under tension.
 kc is a coefficient that considers the stress distribution in the concrete section just
before cracking occurs. The coefficient kc is 1 for pure tension
 k is a coefficient that accounts for non-uniform self-equilibrating stresses that reduce
the restraint forces. The coefficient will be 1 where the member’s web, h, is up to 300
mm thick or when the member’s flange is less than 300 mm wide. The coefficient k
will be 0.65 where the member’s web is at least 800 mm or has a flange with a width
greater than 800 mm. Any values lying in between these limits may be interpolated.
 fct,eff is the mean value of the concrete tensile strength for the time where the concrete
is expected to first appear. For early age cracking this time is usually taken to be 3 days
and for long term cracking the mean tensile strength is generally taken at 28 days.
 Act refers to the area of concrete that is under tension just before the formation of a
crack.
 fky is the characteristic yield strength of the reinforcement.
 25

[Link] Crack Spacing

The spacing of cracks in concrete members lies within So and 2S0, where S0 is the minimum crack
spacing and 2S0 is the maximum. Any distance beyond 2S0 will result in the formation of another
crack. The crack spacing depends on the rate of transfer of tensile stress from the crack to the
concrete; this is influenced by bond strength between the concrete and the reinforcement (Beeby
and Narayanan, 2005).

The ensuing description of the crack spacing derivation follows from the works of Beeby and
Narayana (2005) as well as from the design guide for crack control in reinforced concrete beams
produced by the Centre for Construction Technology Research (2000). After the first crack
occurs, slippage between the concrete and reinforcing bars will follow. Bond stress will then
develop between the concrete and the reinforcing steel over a transfer length on either side of the
crack. The minimum crack spacing may be equated to this transfer length,
Sr,min=ltr.
The maximum crack spacing is twice the minimum crack spacing:
Sr,max=2ltr
The transfer length may be described by the formula:
ltr = φfct/4fbρ,
in which φ is the bar diameter, fct is the tensile strength of the concrete, fb is the average bond
stress over the transfer length and ρ denotes the ratio of the gross cross-sectional area of concrete
to the area of steel reinforcing .
This would then make the maximum crack spacing equal to:
Sr,max = φfct/2fbρ.
The average crack spacing between cracks is assumed to be given by multiplying the transfer
length by 1.5, giving the relationship:
Srm= k1φ/4ρ.
Where k1 = 1.5(fct/fb), is a coefficient accounting for the bond characteristics of the concrete. Thus,
including the effects of cover (a parameter found empirically to have a direct effect on the crack
width), the average crack spacing may then be described using the equation:
Srm= 2c + k1k2φ/4ρp,eff.
The inclusion of k2 was to have the crack spacing formula also cater to cracking due to flexure
since the crack spacing formula had been derived for concrete members under pure tension. The
introduction of an effective tension area (ρp,eff) rather than the gross-cross sectional area (ρ) of the
concrete also accounts for instances in which the concrete member is not just under pure tension
in which the full cross section is under tension
 26

EN 1992 puts forth that the maximum crack spacing that will result in the characteristic crack
width having only a 5% probability of exceedance is 1.7 times bigger than the average crack
spacing. This assumption was based on experimental data (Beeby & Narayanan, 2005). Thus the
maximum crack spacing that appears in the EN 1992 is:
Sr, max = 1.7Sr,m= 3.4c + 0.425k1k2φ/ρp,eff (2.10)
where,
 Sr,max is the maximum crack spacing
 c is the value of the cover to the reinforcement
 k1 is a coefficient that considers the bonding properties of reinforcement in the
concrete. This coefficient is 0.8 where high bond steel reinforcement is used and is 1.6
in instances where reinforcement bars with a plain surface are used.
 k2 is a coefficient that accounts for the distribution of strain. This coefficient is 0.5 for
bending and 1 when dealing with pure tension (as for restrained shrinkage). Any values
that lie between these above-mentioned values, the k1 value may be determined using
the formula k1 = (ɛ1 + ɛ2 )/ (2ɛ1). In this formula, ɛ1 represents the greater tensile strain at
the boundary and ɛ2 describes the lesser tensile strain.
 φ is the bar diameter of the steel reinforcement.

As per the requirements specified in EN 1992-3, in a situation where the steel reinforcement
spacing exceeds 5(c+ φ/2) or where there is no bonded reinforcement in the tension zone, the
maximum crack spacing becomes [Link] = 1.3(h-x).

The crack spacing formula includes the use of a reinforcement to effective concrete ratio (ρ p,eff
=As/Ac,eff) rather than the steel reinforcement to gross concrete ratio (ρ =As/Ac) used to determine
the minimum required steel reinforcement for crack control. The effective concrete area is
generally defined as being 2.5 times the distance from the tension face to the centroid of the steel
reinforcement (2.5(h-d)), or limited to a third of the difference between the section thickness (h)
and the neutral axis (x) for slabs ((h – x)/3) (Mosley, Bungey & Hulse, 2012). These limits are
represented graphically in Figure 2.11.
 27

Figure 2.11: Typical Cases of Effective Concrete Area Following (Figure 7.1 of EN 1992-1-
1:2004)

[Link] Crack Width Calculation

The calculation for the maximum crack width may be obtained from the compatibility equation
given as (similar to BS 8007):
wk = Sr, max ɛr,
in which the crack width is equal to the product of the maximum crack spacing and the average
strain. The strain induced by either an end or edge restraint is dealt with separately in EN 1992-
3, unlike in the superseded BS 8007 standard which dealt with both end and edge restraint using
the same equation. Where a member is restrained along its ends the restrained strain follows the
equation:
0.5α e kc kf ct,eff (1+1/α e ρ) (2.11)
εr = (ɛsm - ɛcm) =
Es

where,
 28

 ɛsm is the average strain in the reinforcement

 ɛcm represents the average tensile strain in the concrete between cracks
 αe is the ratio of the steel modulus of elasticity of steel to the modulus of elasticity of
concrete for the appropriate age of concrete
 kc is the coefficient for stress distribution
 k is the coefficient that accounts for the effect of self-equilibrating stresses.
 fct,eff is the mean tensile strength of the concrete at the time of cracking
 ρ is the ratio of steel area to the gross concrete area
 Es is the modulus of elasticity of the steel reinforcement

Members that are restrained along one edge are estimated in EN 1992-3 by the following formula:
(ɛsm - ɛcm) = Rax ɛfree
where,
 Rax is the restraint factor
 ɛfree is the strain that would occur if the member were completely unrestrained

εfree = The free strain may be approximated by:

εfree = εcd+εca+αT,c( T1+T2), (2.12)

where εcd is the drying shrinkage strain, εca is the autogenous shrinkage strain; αT,c is the
coefficient of thermal expansion of concrete, T1 denotes the fall in temperature from hydration
peak to mean ambient temperature in the concrete. T2 is the seasonal fall in temperature.

[Link] Restraint Conditions

The restraint factors that may be obtained for calculations under EN 1992-3 may be obtained in
the same way as BS 8007. As in BS 8007, the factors for common situations and construction
sequences are given in the code (figure A.3 of BS 8007 was reproduced as figure L.1 of EN 1992-
3. Also, table A.3 of BS 8007 is as table L.1 of EN 1992-3).

2.3.4 Issues Surrounding Crack Width Estimation

There has been much deliberation around the way in which crack formation may be estimated
with many variations of the crack width model currently available, according to Caldentey et al.
(2013). The crack spacing has been modelled in as much as 23 different ways (Caldentey et al.,
2013). Many national codes differ in their formulation of the crack spacing (Beeby & Narayanan,
2005). One particular point of contention regarding the Eurocode 2 crack model involves the
inclusion of the φ/ρp,eff, whose influence on the crack spacing has been questioned by Beeby
(2004) article entitled The Influence Of Parameter Φ/Ρ Eff on Crack Widths for the journal
 29

Structural Concrete (as cited in Caldentey et al., 2013). Beeby (2004) stated further that the real
influence the φ/ρp,eff ratio may have on the crack spacing is due to the implicit consideration of
the concrete cover parameter in the ratio of reinforcing steel to effective concrete area, ρ p,eff
(Caldentey et al., 2005).

Caldentey et al. (2013) conducted an experiment on the influence of concrete cover and the φ/ρp,
eff ratio on the crack spacing for cracks induced under flexure, where 12 beam specimens were
loaded at a constant moment span of 3.42 m. All the rectangular cross-sections used were 0.35 m
by 0.45 m and were made of class C25/33 concrete. Different reinforcement configurations were
looked at, one with no stirrups, and another with stirrups (8 mm diameter) spaced at 100 mm and
300 mm centre to centre respectively. Caldentey et al. (2013) showed clearly in this research that
crack spacing, and thus the crack width, increased with an increase in concrete cover to
reinforcement. This result confirmed that cover was, in fact, an important part of the estimation
of crack spacing for the load induced crack case. The inclusion of cover in the EN 1992 crack
spacing formula comes after previous experimental findings revealed the concrete cover to be an
important contributor to crack spacing (Caldentey et al., 2013). The influence of φ/ρp,eff may be
derived, using bond theory, from the equilibrium of the reinforcement bar and the parts of the
concrete cross section found between the crack and the section of zero slip. The concept of transfer
length is applied in this instance (Caldentey et al., 2013).

The bond theory to which the φ/ρp,eff ratio is based still stands for both the load-induced case,
which may mean that its influence in the restrained shrinkage case could be comparable to that of
the load-induced cracking case. The same crack spacing equation used to predict the crack widths
of load induced cracks in EN 1992 is also used for the restrained shrinkage case. The EN 1992
crack spacing model was conceptually derived for concrete members under pure tension.
Accounts are taken for instances of flexure through the introduction of coefficient k 2 and the
effective steel ratio (ρp,eff) where only parts of the concrete section will be experiencing tension.
Thus, the modifications (particularly of the second term) of the EN 1992 crack spacing equation
allows for a crossover in application. The contribution made by concrete cover towards the crack
spacing, as theorized by Caldentey et al. (2013), is through the need for the transfer of stresses
from the reinforcing steel to the centre of the effective concrete area located on either side of the
bar. It must be reiterated that the findings made by Caldentey et al. (2013) on the influence of
either variable are based on cracks due to bending. Data on the influence of concrete cover on
crack spacing for the restrained shrinkage case were not found. However, it could be assumed
that this parameter would be influential in this case. The same may be said about the φ/ρp,eff
variable.
 30

Caldentey (2005) had also compiled an earlier report in which Beeby’s 2004 claim had been
challenged. In this report, Caldentey (2005), put together several tests done by others on this
matter and compared their findings. The data in these tests were obtained with the cover being
kept mainly constant, whilst the φ/ρp,eff ratio was varied. The works of Hartl (1977), Eligehausen
(1976) and Rüsch & Rehm (1963) were reviewed. The experiments reviewed were for concrete
tested in tension.

Hartl’s (1977) test had square concrete elements being subjected to pure tension, reinforced with
just one reinforcing steel rod where it was revealed that, even though the φ/ρp,eff ratio and concrete
cover were both found to influence the crack spacing, the influence of the φ/ρ p,eff ratio could not
be distinguished from that of the cover in this experiment. Here changing the φ/ρp,eff ratio required
changing the bar diameter value, which implied a change in the concrete cover value. Eligehausen
(1976) found that the φ/ρp,eff ratio had a small influence on crack spacing, while Rüsch & Rehm
(1963) determined that crack spacing became smaller with an increase in the φ/ρp, eff ratio. Rüsch
& Rehm’s (1963) results were found to have been effected by the reinforcing configuration, the
difference in the types of ribs used in the reinforcement and reinforcing cross section between
specimens having similar concrete cover values. The paper eventually concluded that Andrew
Beeby’s theory was, in fact, sound and that the φ/ρp, eff ratio had no real effect on crack spacing.
It was further stated that the use of this ratio in the current formula for crack spacing in EN 1992-
1-1 is due to there being a lack of critical examination into a more suitable formula.

Essentially, as may have been deduced from the earlier studies, the crack spacing may be
separated into two terms (Kaethner, 2011). That is, crack spacing is the sum of the cover zone
cracking (k3c, contributing 50- 80% of the crack spacing value) and the cracking near the bar
(k1k2k4φ/ρp,eff, contributing 20-50% of the value). In an investigation carried out by Kaethner
(2011) comparing the two terms found in the crack spacing formula to those values of the crack
width at the concrete surface and at the bar surface found in practice. It was found in this
comparison that the calculated cover term contributed less to the crack spacing than found in
practice. However, the bar slip term predicted a stronger value than what may be observed in
practice. Kaethner’s (2011) findings once again prove the relevance of both the concrete cover
and bar slip term in the EN 1992 crack spacing formulation- both of which having a clear
contribution to the ultimate value of the crack width (particularly the cover term). Although, the
experimental data collected were done so on cracks resulting from flexural loading. The
observation made is represented graphically in Figure 2.12:
 31

Figure 2.12: Comparison of the Cover and Bar Slip Terms of the EN 1992 Crack Spacing
Formula with Experimental Data (Kaethner, 2011)

Most findings relating to the crack spacing formulation, and in particular the debates of the
inclusion of either the cover term or the bond slip term in EN 1992, have included experiments
and observation based on the load induced cracking case. It is recommended for future research
that the same exercise is extended towards cracks resulting from restrained shrinkage.

A study was conducted by Bamforth, Denton and Shave (2010) on the development of a unified
approach to estimating crack width from both end and edge restraint. In this study it was
concluded that the methods used in determining the edge restraint in Eurocode 2 were based on
flawed assumptions. For instance, different parameters are used to calculate edge restraint and
end restraint, resulting in the formulae being considerably different. Bamforth et al. (2010) also
noted in this study that, for the edge restraint case of both BS 8007 and EN 1992, the transfer of
the load from the restrained concrete member onto the reinforcing steel when cracking occurs was
neglected. This stands contradictory to the concept behind the required minimum amount of
reinforcing steel for crack control which is based on the idea that the steel reinforcing carries the
load from the concrete section after cracking (Bamforth et al., 2010).

It was found in this investigation into the EN 1992-3 edge restraint crack model that the geometry
of the restrained member had not been considered in the crack spacing equation. Bamforth et al.
(2010) observed that the tensile strength of the concrete was also not accounted for in the edge
restraint model. In addition to this, the positive effect that the restraining member has on the crack
formation was neglected. Here, Bamforth et al. (2010) suggests that the restraining base member
actually works to prevent the crack widths that are generated from reaching their full potential.
 32

Moreover, the minimum reinforcement area as determined using EN 1992-1-1 was found to be
conservative when applied to the edge restraint crack model (Bamforth et al., 2010). Instead,
Bamforth et al. (2010) suggest that further research is required to develop a crack model that
better models the formation of cracks in structures.

Bamforth’s unified method assumes the formation of the maximum crack width follows the model
for end restraint in EN 1992-3. However, where edge restraint is being considered, Bamforth
(2010) suggests development of the same maximum crack would be reduced due the following
factors:
 Some of the load is transferred to the restraining member and so decreases the amount of
load that would normally get transferred to the reinforcement.
 The edge restraint prevents the crack width from expanding to its full potential width.
 The existence (or non-existence) of cracks in the concrete may determine the degree of
stress relaxation between cracks, potentially affecting any new cracks that forms.

In this new model developed it is suggested that cracks actually develop in 2 stages:
 Stage 1 cracking is based on the EN 1992-3 model for end restraint, the only modification
made in this stage is the inclusion of the effect of the edge restraint. The edge restraint
inhibits the extent to which the cracks may open and attracts some of the load onto the
restraining member. In stage 1, a crack of width wk1 opens instantaneously and part of the
load is transferred onto the reinforcement. The relative lengths of the cracked (assumed
length of debonding) and uncracked lengths are accounted for in this stage.
 The second stage of the crack model considers the continued contraction of the cracked
concrete relative to the steel reinforcing (the crack width opens up further by a value wk2).
As with stage 1, it is assumed that increased restraint reduces the extent to which cracks
may open.

The full crack width would then amount to the sum of the cracks obtained at each stage of cracking
(wk = wk1 + wk2). Bamforth (2010) found, after comparison to observed cracks, that the proposed
unified approach better reflected the formation of crack widths for edge restraint then EN 1992-
3 currently does.
 33

2.4 Autogenous healing

Autogenous healing (self-healing) of concrete, according to EN 1992-1-1:2004 is said to occur to

some degree in all concrete structures. In a technical report, Edvardsen (1999) identified the
participation of calcium carbonate crystals as being the main culprit of self-healing in concrete
(Edvardsen, 1999). Other contributors to concrete’s autogenous healing observed by Edvardsen
(1999) include: the flow of the concrete being blocked by impurities from the water or loose
particles from the cracking and swelling of the cement paste during hydration. Essentially, all of
the mentioned contributors to the concrete’s self-healing serve to reduce the amount of water
flowing through the member with time. Other more secondary causes include the crack width,
water hardness of the retained water, aggregate type as well as the type of cement used for the
structure and the water pressure (Edvardsen, 1999) . Self-healing of the concrete was also said to
generally occur during the first 3 to 5 days of the concrete structure. In EN 1992-1-1: 2004, it is
suggested that a crack width of 0.05 mm with a water pressure gradient (water depth to wall
thickness) limited to 35 would heal itself. This is also true for crack widths of 0.2 mm with a
water pressure gradient of up to 5. Crack widths with water pressure gradients between these two
values may also heal autogenously.

At current, the South African equivalent of the EN 1992-3 (2006) design code, namely SANS
10100-3, is in the draft phase. However, some of what may be expected by designers were
outlined in a research paper by Wium, Retief and Barnardo-Viljoen (2014) and reiterated in a
doctoral thesis by Retief (2015). Wium et al. (2014) noted that the jump in the crack width
limitations set out by BS 8007 to the stricter crack limits of the EN 1992-3 design code would
incur considerable increases in cost to meet those more rigorous limitations in crack width. The
EN 1992 crack width limit required increases in reinforcement of factor 1.4 and 2 where the crack
width limit was reduced from the BS 8007 specified 0.2 mm to crack widths of 0.1 and 0.05 mm
respectively- which are crack limits included within the EN 1992 specified range of permissible
crack widths (McLeod, Retief & Wium 2013). The SANS 10100-3 draft was reported to have
done away with the rigorous crack limits of EN 1992-3 and employ those crack limits of BS 8007
instead (Wium, Retief and Barnardo-Vijoen, 2014). Although, it must be noted that the
implications of the more onerous crack width limits stipulated by the EN 1992 as compared to
those of BS 8007 were considered for the load induced cracking case and not for cracks resulting
from restrained deformation. Extending this investigation to include the restrained strain crack
model would present a complete gauge of the ramifications of the change in crack width limits
from BS 8007 to EN 1992-3. It was suggested by Retief (2015) that a rational basis for the use of
the more onerous EN 1992 crack limits may be established through a probability-based economic
optimisation.
 34

An experimental research aimed at studying the phenomenon of autogenous healing was carried
out and the results reported in a technical paper for the American Concrete Institute Materials
Journal (Edvardsen, 1999). The experiments were conducted on concrete specimens with a single
tension crack set in each specimen. The crack widths tested in this experiment were 0.1, 0.2 and
0.3 mm respectively. The crack lengths varied at 200, 300, and 400 mm with the water head varied
from 2.5 to 20 m of water. The hydraulic gradient (water pressure head /thickness of structure)
for this research varied from 6.25 to 50. It was determined in this investigation that for 50 % of
the specimens with a 0.2 mm crack width and hydraulic gradient of 6.25 (water pressure head of
2.5 m) healed completely in 7 weeks. For 25 % of the specimens with a 0.2 mm crack width and
hydraulic gradient of 25 (water pressure head of 10 m), the concrete specimen also healed within
7 weeks. The experiments showed that the influence of the hydraulic gradient on the water flow
was smaller than that of the crack width. The permissible crack widths, expected to obtain almost
total self- healing, recommended for use after the experimentation were (Table 2.5):

Table 2.5: Permissible Crack Widths for Autogenous Healing (Edvardsen, 1999)
+
Hydraulic gradient (m/m) *wk (mm) wk (mm)
40 0.1 to 0.15 ≤ 0.1
25 0.15 to 0.20 0.10 to 0.15
15 0.2 to 0.25 0.15 to 0.20
Notes:
*∆w≤ 10%
+
10%≤ ∆w ≤ 90%

In a similar study on the influences of both hydraulic pressure and crack width on the water
permeability of crack- induced concrete specimens, Yi, Hyun and Kim (2011) found that as the
crack width and hydraulic pressure increased, so did the transport of water through the concrete.
Three particular crack widths were examined in this study, namely 0.03, 0.05 and 0.1 mm. The
permeability of the water through the crack-induced concrete specimens was measured for
hydraulic pressures of 0.01, 0.025, 0.05 and 0.2 MPa. Yi et al. (2011) determined in this study
that crack widths smaller than 0.05 mm had little effect on the permeability of the concrete due
to autogenous healing. Where the crack widths were found to be between 0.05 mm to 0.1 mm,
with a hydraulic pressure greater than 0.025 MPa, the permeability of concrete increased
considerably. Ultimately, Yi et al. (2011) suggested that in the case where a structure experiences
a hydraulic pressure of less than 0.01 MPa, the allowable crack width may be set at 0.1 mm. For
 35

a hydraulic pressure of 0.025 MPa or greater, the allowable crack widths should be 0.05 mm (or
otherwise be between 0.05 and 0.1 mm).

Yet another test on the phenomenon of autogenous healing carried out at the University of Kwa-
Zulu Natal (Mans, 2012) revealed that concrete samples with a crack width of ±0.2 mm through
cracks showed considerable healing within 72 hours of testing. The test was conducted for 250
hours under a hydraulic gradient of 12.

Conclusively, it may be deduced from the various experiments studied herein that the autogenous
healing does, in fact, occur between the crack widths of 0.05 mm and 0.2 mm as suggested by EN
1992 with a similar range of hydraulic gradients proposed by EN 1992. Moreover, the size of
crack widths was found to have more of an influence on the concrete’s permeability than the
hydraulic gradient. Although, still increasing the size of both the crack width and the hydraulic
gradient would result in an increase in the permeability in of the concrete.

2.5 Conclusion

Liquid retaining structures in South Africa are designed using design code BS 8007 and those
relevant parts of BS 8110-2. The replacement of BS 8007 and BS 8110-2 with EN 1992-3 and
EN 1992-1-1 respectively for the design of liquid retaining structures presents South African
engineers with the opportunity to also changeover into the use of the Eurocodes for LRS design.
Much stands to be gained from Eurocode adoption, namely reaping from the technical expertise
of the Eurocodes with supporting design guides and software easing the transition into adoption.
Additionally, there would be an increase opportunity for local engineers to participate in some
international projects. Moreover, the choice to either adapt or adopt is a less demanding
alternative than the more labour intensive, expensive and time-consuming task of developing a
completely new code. However, the question of the possible implications of this changeover of
codes as applied in the South African context is raised. These implications have already been
quantified for the load induced cracking case by past researchers, making an investigation into
these implications for the restrained shrinkage cracking case relevant.

Points of interest raised by past researchers and to be further investigated in this research include
the implications of the more stringent crack limits of the EN 1992 crack model, the influence of
the concrete cover and ϕ/ρp,eff values on the crack spacing, and issues surrounding the ways in
which the edge and end restrained strain are modelled. Ultimately, a better understanding of the
EN 1992 crack model would aid towards improving its reliability for the South African
 36

environment. The increased demand of the EN 1992 cracking serviceability limit state as
compared to the ultimate limit state gives a reliability analysis into the EN 1992 crack model
significance (evidently, since the failed compliance of the crack limits may lead to the loss of
structural integrity).
 37

Chapter 3: Structural Reliability

3.1 Introduction

The reliability of a structure is described as its ability to successfully perform its function under
working conditions throughout its required working life (Green, 1972). No structure or system
can perform at 100% reliability. One can expect some probability of failure in the structure's
lifetime as the engineer cannot escape from uncertainties that exist in design. Some examples of
where uncertainties in design may arise include, amongst others: the randomness of geometric
data, statistical uncertainties, simplification of actual conditions in determining theoretical models
and errors in design (Holický, 2009). The effect of these uncertainties on the design may be
quantified and evaluated through probabilistic concepts and reliability theory. The idea of using
reliability concepts in engineering design is not a new one, dating as far back as World War One.
During World War One there was increased interest in knowing the failure rate of flights,
consequently a reliability criterion was developed to ensure maintenance of a reasonable failure
rate (Green, 1972). Determining the failure probability of a system is important in that it will not
only help in evaluating whether or not the system performs satisfactorily in its lifespan, but it may
also help to avoid dire financial loss from system failure. The principles involved in reliability
theory are outlined hereunder.

3.2 Limit State

The satisfactory performance of a structure is ensured by the implementation of limit states. The
performance limit states of structures may be thought of as a kind of boundary, beyond which the
structure will be considered inadequate. Limit states may be divided into two major categories,
namely the ultimate limit state and the serviceability limit state. The ultimate limit state deals with
the collapse of the structure as well as the safety of its occupants. Whereas the serviceability limit
state pertains to the normal working conditions of the structure with its most important areas of
concern being: deflection, cracking and durability (Mosley et al., 2012). The serviceability limit
state can be further broken up into irreversible and reversible limit states. The irreversible
serviceability limit state, as can be deducted from the name, is where the damage caused remains
permanent even after the cause of the damage itself has been removed. Contrary to this state is
the reversible serviceability state where the damage incurred does not remain permanent even
after the cause of the damage is removed (Holický, 2009). Generally, the ultimate limit state has
taken on greater relative importance to the serviceability limit state. However, for liquid retaining
structures, the serviceability limit state takes on more importance in the structure’s performance.
 38

A crack width that is wider than what is allowed for under the serviceability limit state will result
in the structure becoming permeable and thus losing its structural integrity.

3.3 Basic Reliability Theory

A model is a representation of an existing object or phenomenon in which some aspects of this

representation vary somewhat from the original object or phenomenon. This is because some
simplifications and assumptions have to be made in the development of the representative model.
Subsequently, uncertainties may arise from the simplifications and assumptions made in forming
the model (Croce, Diamantidis and Vrouwenvelder, 2012). Other sources of uncertainty are the
characteristic randomness of a physical phenomenon, as well as the predictions of states of nature
made with inadequate information (Ang and Tang, 1984). With increased data and information,
models representing physical phenomenon may be improved and made more accurate with
inadvertent biases reduced. However, the inherent randomness of physical phenomena cannot be
avoided. It cannot be guaranteed with absolute certainty that a variable will take on a particular
value; instead, a range of possible outcomes may be attributed to this same variable. The
likelihood of occurrence for a specific value may be determined by its probability distribution
function (Holický, 2009).

As described by Ang et al. (1984), most engineering problems may be described as supply and
demand problems where the safe state of the structure is where the supply exceeds the maximum
amount of demand experienced over a lifetime. The supply and demand may be expressed as
either random variables (Xi) or functions of random variables with their own distribution
functions. In other words, the resistance of a structure (R) needs to be greater than the action effect
(E) of the structure in order for the structure to remain reliable (E<R). The performance function
separating the safe state of the structure or engineering process may be expressed as:
g(Xi) = R-E = 0 (3.1)

A negative value of the performance function is indicative of a failure in performance, whilst a

positive answer shows that the resistance of the structure exceeds the load effect and thus the
structure is safe. The limit state may be defined as the distinct separation between the safe state
of the structure where it performs reliably and the unsafe state where it no longer functions. So
primarily, the performance function, g(Xi), is itself a limit state.

The equivalent normal distributions of the demand and supply variables are used to approximate
the failure probability. Where the reduced variate (equivalent normal variate) of the resistance
(supply) may be described by the equation 3.2:
 39

R - R (3.2)
R’ = ,
R

where μR and σR respectively denote the mean and standard deviation of the resistance variable.
And the reduced variate of the action effect E (demand) may be determined using the formula:
R - E (3.3)
E’ = ,
E
where μE and σE are the respective symbols for the mean and standard deviation of the action
effect, E. Then the performance function may be rewritten as:
g(Xi) = R’-E’ = 0. (3.4)

This then equates to,

g(Xi)= σRR’ - σEE’ + μR – μE = 0
Then the linear failure distance from the origin to the failure line g(Xi) = 0 can be expressed as:

R  Y G (3.5)
β= = ,
 R2   E2 G

This distance β is the safety index, and describes the shortest distance from the reduced variate
origin to the limit state (Wu, Lo and Wang, 2011). In other words, this distance describes the
distance to the most likely point of failure along the limit state (this is illustrated in Figure 3.1).
In general, a structure is said to be in a desirable state where the limit state function is greater than
zero and at values less than zero the structure will be in an undesirable state. At zero, the structure
just meets the limit state as shown in Figure 3.1.

Figure 3.1: Space of Reduced Variates E’ and R’ (as adapted from Ang and Tang. (1984))
 40

For the serviceability limit state, the performance function described above is structured such that
an exceedance of a limiting design criterion (like a set deflection value, or in this case an allowable
crack width) would take the form:
g(Xi) = C – S = 0, (3.6)
where C represents the serviceability criterion in question and S denotes the action effects (as
described in SANS 2394:2004). Clearly, regarding cracking in liquid retaining structures, where
the action effect exceeds the serviceability criterion the limit state would be exceeded and the
undesirable (unsafe) state entered into. The EN 1992 cracking serviceability limit state may be
similarly formulated:
g(Xi) = wlim – θw, (3.7)

where wlim describes the permissible crack limit and w represents the mean crack width based on
the EN 1992 maximum characteristic crack width. In developing the crack width formula for EN
1992-1-1 it was determined through experimental data that a factor of 1.7 be applied to the average
crack spacing in order to calculate the maximum crack width (Beeby and Narayanan, 2005).
However, the reliability analysis undertaken herein (described in chapter 5) requires the use of
the mean crack width and so a reduction in value of 1.7 to the EN 1992 maximum characteristic
crack width formula should return a function for the mean crack width.

The value, θ accounts for the model uncertainty in the EN 1992 crack model and is regarded as
a random variable in the reliability analysis. An elaboration of the formation of the limit state
function to be used in this investigation is given in chapter 5.

The probability of safety may then be determined by:

ps = Ф(β), (3.8)

where Ф is the standardized normal distribution function and β is the reliability index as defined
in equation 3.5. The failure probability may then be determined using the relationship pf = 1 - ps,
so that pf = 1 – Ф (β) = Ф (-β).

A simple illustration of the relationship between the reliability index and the failure probability
may be outlined in Table 3.1:
 41

Table 3.1: Relationship between Failure Probability and Reliability Index JCSS Part 1
(Joint Committee of Structural Saftey, 2001).
Pf 10-1 10-2 10-3 10-4 10-5 10-6 10-7
β 1.3 2.3 3.1 3.7 4.2 4.7 5.2

3.4 Reliability Index and Target Reliability

The reliability index, β, is indicative of a structure or engineering phenomena’s failure probability

over the lifetime of the structure, thus the designer may design a structure to meet a certain
reliability index. This particular reliability index may be referred to as the target reliability index,
βt. In other words, in designing towards a specific target reliability index, the designer is ensuring
that the structure does not fail beyond an accepted failure probability (Holický and Marková,
2012). The reliability index may be determined using reliability methods, the obtained β value
may subsequently be compared to the target reliability (βt) for its class of structure and design
working life (the design working life of the structure may be defined as the period of time in
which the structure is intended to be in use without there being any major repair required.
Although, it is expected that the structure will have to encounter some minor maintenance
attempts throughout its service life (Holický, 2009)). An acceptable structure is one where the
target reliability is either just met or the reliability index calculated is approximately equal to the
target reliability index (Holický and Marková, 2012). The JCSS (Joint Committee on Structural
Safety) have recommended a set of reliability indices for use under both the ultimate limit state
as well as the serviceability limit state. Their recommended values for the ultimate limit states are
as follows:

Table 3.2: Ultimate Limit State Target Reliability Indices and Related Failure Probabilities
for a 1 Year Reference Period (Joint Committee of Structural Saftey, 2001).
1 2 3 4
Relative cost of safety Minor Moderate Large
measure consequences consequences of consequences of
of failure failure failure
Large (A) β = 3.1 (pf ≈10-3) β = 3.3 (pf ≈ 5 10-4) β = 3.7 (pf ≈ 10-4)
Normal (B) β = 3.7 (pf ≈10-4) β = 4.2 (pf ≈ 10-5) β = 4.4 (pf ≈ 5 10-6)
Small (C) β = 4.2 (pf ≈10-5) β = 4.4 (pf ≈ 5 10-6) β = 4.7 (pf ≈ 10-6)

The above-mentioned values are also based on a cost-benefit analysis for a representative set of
engineering structures. As can be deducted from the table, the indices are categorized according
 42

to relative cost of safety measure and level of failure consequence. The failure consequences are
determined from looking at the ratio between the total costs (which includes the sum of the
construction cost and failure costs) and construction cost. Where this ratio is less than 2, the
structure may be categorized as having a minor consequence failure (minor risk to life, with
negligible economic loss in instance of failure). If the total cost to failure cost ratio were to be
somewhere between 2 and 5, then the structure may be seen as having a moderate consequence
of failure (there is a moderate level of risk to life in the event of failure and economic loss, in this
case, would be significant). The last class of failure consequence, the large consequence of failure
class, is for structures where the ratio lays between 5 and 10. In this class, the risk to life and
economic loss in the case of failure is sizeable.

The target reliability indices as recommended in part of 1 of the Probabilistic Model Code (JCSS,
2001) for the irreversible serviceability limit state are listed in the following table:

Table 3.3: Irreversible Serviceability Limit State Target Reliability Indices and Related
Failure Probabilities for a 1 Year Reference Period (Joint Committee of Structural Saftey,
2001).
Relative Cost of Safety Measure Target Index (Irreversible SLS)
High β = 1.3 (pf ≈ 0.1)
Normal β = 1.7 (pf ≈ 0.05)
Low β = 2.3 (pf ≈ 0.01)

Values for the reversible serviceability limit state have been given no real general rule in this
JCSS document. In EN 1990 (Eurocode 0), the target reliability index of an ultimate limit state
for a reference period of 1 year is given by βt,1 = 4.7. The reliability indices then for time periods
other than a year may be calculated from the approximate formula (Holický, 2009):
Φ(βt,n) = [Φ(βt,1)]n,
where n denotes the number of years.

Holický (2009) puts forth that where a structure is to be designed for a particular reliability level
and design working life, the target reliability index for the 1 year reference period may be changed
accordingly. For instance, for a structure designed for a target reliability index of 3.8 and a design
working life of 50 years, the target reliability index for the reference period of 1 year should be
βt,1 = 4.7. It must be noted that the above-mentioned reliability indices represent the same level of
reliability, namely an accepted lethal accident rate of 10-6 per year applied to different reference
 43

periods. Additionally, for the same 3.8 target reliability index and a 25-year design life, the 1
year reference period’s reliability index should be set at 4.5 (Holický, 2009).

The target reliability indices with corresponding reliability classes and limit states as presented in
EN 1990 are reproduced in the following table:

Table 3.4: Suggested Reliability Classes and Recommended Minimum Values for
Reliability Index β from EN 1990 for Ultimate Limit State, Fatigue and Serviceability Limit
State (Holický, 2009).
Minimum Values for β
Ultimate Limit States Fatigue Serviceability
(Irreversible)
Reliability Building 1-Year 50-Year 1-Year 50-Year 1-Year 50-Year
Class Example Reference Reference Reference Reference Reference Reference
Period Period Period Period Period Period
RC-3 High Bridges, 5.2 4.3
public
buildings
RC-2 Residential 4.7 3.8 1.5 to 3.8 2.9 1.5
Normal and office
buildings
RC-1 Agricultural 4.2 3.3
Low buildings,
greenhouses

The class divisions – high, normal (moderate) and low – set out in the above-mentioned EN 1990
table of reliability indices follow much of the same descriptions as those mentioned in the JCSS
part 1 (2001). Another code of practice that recommends values for the target reliability index for
the design of structures is ISO 2394:1998 (SANS 2394:2004), which summaries the indices as
follows:
 44

Table 3.5: ISO 2394 Lifetime Target Reliability Indices (Holický, 2009).
Relative cost of Consequences of Failure
Safety Measures
Small Some Moderate Great
High 0 1.5 2.3 3.1
(reversible serviceability (irreversible (fatigue limit states) (ultimate limit states)
limit state) serviceability limit
states)
Moderate 1.3 2.3 3.1 3.8
(fatigue limit states) (ultimate limit states)
Low 2.3 3.1 3.8 4.3
(ultimate limit states)

The shaded values from this table are those values of the reliability index that are also shared in
EN 1990. SANS 10160-1:2011 also sets target reliability indices for structures categorised
according to their consequence of failure. The target reliability index values presented in Table
3.5 have been obtained assuming lognormal distribution or the Weibull distribution for resistance.
A normal distribution was assumed for permanent loads and a Gumbel distribution was assumed
for variable loads (Holický, 2009).

South Africa also categorises different structure types according to what their expected design
working life should be. The table for the design working life and their respective working life
categories from SANS 10160-1:2011 is reproduced here as Table 3.6:
 45

Table 3.6: Design Working Life as Described in SANS 10160-1:2011

Design Indicative Description of Structure
Working Life Design Working
Category Life In Years
*
1 10 Temporary structures (not pertaining to structures that
are intended for re-use after being dismantled).
2 25 Replaceable structural parts, agricultural structures and
other such structures with a low consequences of
failure
3 50 Building structures and other common structures**
4 100 Essential building structures such as hospitals,
communication centres or rescue centres with high
consequences of failure+
*
Refer to SANS 10160-8 for assessment of temporary structures during execution
**
The design working life category applies to the reference reliability class referred to in clause
[Link].
+
Consequences of structural failure could be determined in accordance with annex A

Knowing the design working life intended for the liquid retaining structure allows for the
appropriate target reliability index to be used in comparison to the reliability index determined
through the reliability assessment of the crack model. In accordance with SANS 10160-1:2011,
an appropriate design working life of 50 years will be used for liquid retaining structures in South
Africa.

The recommended target reliability index for a 50-year design working life would be βt = 1.5, for
the irreversible serviceability state of cracking (ISO 2394:1998/SANS 2394:2004). This target
reliability index is the same as the one found in EN 1990, also for a structure with a 50-year design
life under an irreversible serviceability limit state. However, considering that for liquid retaining
structures the serviceability limit state has an increased level of importance and demand when
compared to the ultimate limits state, it may be plausible that a higher target reliability index be
applied to this specialized structure. This, along with the idea that a reliability class (RC) of 3
classification (as described in SANS 10160-1) be used for liquid retaining structures, was put
forth by researchers Barnardo-Viljoen, Mensah et al. (2014). Typical values of βt = 0.5 for the
reversible limit state and 1.5-2 were said to be appropriate for cracking in buildings (Barnardo-
Viljoen, Mensah et al. (2014)). After an assessment of the influence that a change in target
reliability index would have on the load induced crack model of EN 1992 it was determined that,
 46

where a default target reliability index of 0.5 was selected, an increase in βt from 1.5 to 2 resulted
in a 10 and 15% respective increase in the amount of reinforcement required to meet reliability
targets (Retief, 2015). An assessment of the implications of a change in choice of target reliability
index for the restrained shrinkage cracking case should give a complete understanding of how an
increase in the target reliability index value, βt, of the cracking serviceability limit state would
affect the design of liquid retaining structures.

3.5 The First Order Reliability Method

There are various probability methods available that may be used to determine the reliability index
and thus the failure probability of a structure. As outlined in SANS 2394:2004, these methods for
the determination of a structures failure probability include: exact analytical methods or a
numerical integration approach, as well as other methodologies such as the Monte Carlo
simulation, and lastly approximate methods such as the First Order Reliability Method (FORM).
Exact analytical methods are generally used for exceptional cases, whilst numerical methods are
used more often in reliability assessment with approximate methods being the most frequently
used method (Holický and Marková, 2012). The First Order Reliability Method also acts as a
fundamental procedure for a lot of commercially available software used in reliability assessment
(Holický and Marková, 2012). Simulation methods, such as the Monte Carlo method, are most
appropriate for more complex problems where a closed-form solution may be determined (if many
simplifying assumptions are made) or where closed-form solutions are difficult to get (Nowak
and Collins, 2000). This was not necessary for the reliability assessment undertaken herein. This
thesis adopted the First Order Reliability Method. According to Zhao and Ono (1999), it is one
of the most efficient structural reliability assessment methods and is also one of the methods used
in the development of the Eurocode (Eurocode 2’s restrained shrinkage crack model was
investigated in this research and hence FORM seemed the more appropriate probability method
to use in this study). FORM is an approximate method that was developed to circumvent the
difficult computation of the failure probability integral,
pf =  f x ( x)dx. (3.9)
g(x)  0

The SORM method is a refinement of the FORM method where the failure surface (g(X i) = 0) is
approximated by a quadratic surface at the design point (SANS 2934:2004).

The following basic outline describes the FORM methodology. The steps presented are adapted
after Holický (2009) and Ang & Tang (1984):
 47

1. The performance function, g(Xi) = 0, is defined and the initial values for the limit state
basic variables Xi lying on the failure surface are assumed. The initial assumption is
generally taken to be the mean of the basic variable.
2. The mean and standard deviation of non- normal random variables are transformed into
their normal equivalents. In other words, the non-normal μ becomes μNxi and the non-
normal σ is converted to σNxi. Non-normal random variables are then transformed to the
standardised normal equivalent:
xi’ = xi - µNxi/ σNxi
3. The partial derivatives of the performance function with respect to the standardised
random variables are determined
g
.
X 'i
The derivatives are used to find the direction cosines,α*i, at the failure point.
g
 X i
α*i =
2
 g 
i  X 
 i 

4. The direction cosine found in the previous step is then used to determine the new failure
point in terms of the reliability index, ‘β’,
xi* = μNxi – α*iσNxiβ.
This new failure point is then substituted into the performance function g(X i) = 0 and
solved for β.
5. The β value is then used to find the numerical value for the design point at the limit state.
This failure point may then be used as the new starting failure point in the next iteration.
6. Steps 2 to 5 are repeated until convergence of β and the subsequent design failure point
is reached.
7. The failure probability can then be calculated using the formula pf = Ф (-β).

The sensitivity factors/direction cosines generated from the FORM analysis describes the relative
influence each random variable has with respect to the others utilised in the analysis. This
normalised value is represented in either decimal or percentage form, where the closer the value
is to 1 (or 100%), the stronger the relative influence of the random variable in question is with
respect to the others being analysed in a particular FORM analysis. The square of the sum of the
direction cosines should add up to 1 (or 100%). Needless to say, the variable found to be most
influential contributes the most to the reliability index obtained at the end of the FORM analysis
(Saassouh and Lounis, 2012). The sensitivity factor (direction cosine) can either be found to be
 48

positive or negative. A negative direction cosine represents an unfavourable action (Holický,

2009).

3.6 Statistical Parameters of the EN 1992 Restrained Cracking Serviceability Limit State
3.6.1 Introduction
The probability distribution functions (pdf's), as well as the mean (μ) and standard deviations (σ)
of the random variables, are required for the FORM analysis (as outlined in section 3.5). The
choice of theoretical model assumed for the basic variables significantly affects the reliability
indices obtained in a reliability analysis (Holický, 2009). An investigation on the variability of
the basic variables used in the EN 1992 crack model was therefore conducted. Conventional
models for the time-invariant basic variables used in crack width estimation are summarised in
Table 3.7. These statistical parameters have been derived primarily from the works of Holický
(2009). In addition to this, literature on the stochastic nature of the respective variables as well as
information from the Joint Committee on Structural Safety’s (JCSS, 2001) probabilistic model
code documents were gathered and included in Table 3.7.

In a background on typical probabilistic distributions used to describe random variables, Holický

(2009), gives examples of generally accepted assignments of these distributions for load,
geometric and material variables. Geometric basic variables may be described by a normal
probabilistic distribution, log-normal distribution and beta distribution. Material properties may
usually be described by normal distribution and log-normal (considering material strengths).
Additionally, load effects may be categorised by a normal distribution and Gumbel distribution
(Holický, 2009). Model uncertainty has generally been known to follow a log-normal distribution
function (Holický, 2009).

For this particular investigation the variables:

 Model uncertainty, concrete cover, concrete tensile strength and section thickness were
treated as random variables (where their inherent variabilities accounted for in the
analysis).
 The restraint factor and the remainder of the basic variables of the EN 1992 restrained
shrinkage crack model were regarded as being deterministic.

The investigations into literature regarding the statistical parameters of the variables and the
ultimate choice of said parameters for use in this investigation are addressed in the subsequent
text.
 49

Table 3.7: Theoretical Models of Basic Variables in EN 1992 Crack Model (Holický, 2009).
Name Sym. Units Distribution Mean μx St. Dev.
σx
Cross section h m Normal hk 0.005-
thickness 0.01
Geometry Cross section b m Normal bk 0.005-
depth 0.01
Concrete cover c m Both- Sided ck 0.005-
Limited 0.015
Beta/Gamma
Reinforcement φ mm Deterministic 0
diameter
Concrete fc MPa Lognormal fyk + 2σ 0.1-
Compressive 0.18μk
strength (cube)
Concrete fctm MPa Lognormal fyk + 2σ 0.1-
Material Tensile 0.18μk
Strength
Steel Yield fy MPa Lognormal fyk + 2σ 0.07-
Point 0.1μk
Concrete Ec,eff GPa Deterministic 0
Modulus
Steel Modulus Es GPa Deterministic 0
Coefficient- k1 - Deterministic 0
Reinforcement
Coefficients Coefficient- k2 - Deterministic 0
Tension
Limiting Crack wlim mm Deterministic 0
Width
Cracking θ - Lognormal 1 0.3
Model
Uncertainty

3.6.2 Model uncertainty (θ)

The model uncertainty may be determined by comparing experimental data to those values
obtained through the existing prediction model (JCSS, 2000). There are instances in which not
 50

much data is available on the model uncertainty and experience and professional judgement is
depended upon (Holický, 2009). Considering a sensitivity analysis conducted of the EN 1992
load induced crack model with respect to variations in model uncertainty conducted by McLeod
(2013), model uncertainty had been found to bear the most influence on the tension load case and
was found to be the second most influential random variable of the flexural loading case. The
above-mentioned tension load case may be indicative of how influential model uncertainty might
be on the restrained shrinkage crack model of EN 1992. ISO 2394:1998 (reproduced as SANS
2394:2004) includes model uncertainty as a random variable, θ, to be used in reliability
assessments of performance functions accounting for a) inherent variability within the analysed
model, b) inadequate knowledge and c) statistical uncertainty. Moreover, d) mathematical
simplifications and assumptions made in developing the prediction model generates a certain
degree of uncertainty (McLeod, Viljoen & Retief, 2016).

Looking more carefully into these above-mentioned sources of uncertainty with respect to the
restrained shrinkage crack model it may be gathered that:
a) Cracking is a naturally random phenomenon with inherent variability.
b) The knowledge base regarding the stochastic nature of the restrained thermal and
shrinkage cracking case is limited, meaning that there must be a heavy reliance on
experience and professional judgement in this regard. Increased research in this area
would result in a more accurate depiction of restrained cracking’s statistical parameters
and thus increased accuracy in the reliability assessment of its model. Most knowledge
in the area of reliability-based assessments of the cracking serviceability limit state veered
towards those cracks resulting from load (be it a concrete member under flexure or
tension).

The Eurocode 2 crack model along with other crack models have been tested against
experimental data several times in previous research. One such comparison of the
experimental crack widths to those predicted by EN 1992-3 was found in the
investigations of Kamali et al. (2013) on the crack width control of a concrete slab bridge
under restrained cracking (particularly for tensile forces in the transversal direction). It
was determined in the course of this study that for 90% of all the observed crack widths,
the EN 1992-3 crack model overestimated the crack widths (more crack widths were
found to fall below where the measured crack widths equalled those estimated by EN
1992-3, as denoted by the broken red line of Figure 3.2). This experiment was done for
crack widths greater than 0.2 mm. This fact is made clear in Figure 3.2 where the majority
of the estimated crack widths are either comparable to the measured crack widths (at
lower crack widths) or greater than the measured crack widths where the crack widths are
 51

larger than ±0.4 mm. This finding reinforces the notion that the EN 1992-3 is conservative
in its estimation of the crack width due to restrained strains.

Figure 3.2: Comparison of Measure Crack Widths against the EN 1992-3 Predicted Crack
Widths for a Concrete Member Restrained Along its Base (Kamali, Svedholm and
Johansson, 2013).

In another comparison of the EN 1992-3 and BS 8007 crack prediction models to observe
cracks, both models were found to under-predict the observed crack widths – this is
presented in Figure 3.3 (Bamforth, Shave & Denton, 2011). In some instances, this
underestimation of observed cracks would be by as much as 50%. This is contrary to what
was found in the previous case by Kamali et al. (2013), alluding to the considerable
amount of scatter in model uncertainty of the EN 1992-3 crack model for restrained
shrinkage. Both Kamali et al. (2013) and Bamforth et al.’s (2011) comparisons were
done so against data obtained for research on the control of cracking resulting from
restrained contraction.
 52

Figure 3.3: Comparison of Measured Crack Widths to Predicted Crack Widths of BS

8007:1987 and EN 1992-3:2006 (Bamforth, Shave & Denton, 2011)

Although the determination of model uncertainty depends on the formulation of the

prediction model used (McLeod, 2016), the findings of model uncertainty related to the
load induced cracking model were included in the subsequent text. Given the scarcity of
probability based investigations done on the restrained thermal and shrinkage strain
cracking, the load induced crack case should give indications as to how the EN 1992 may
be described in terms of its statistical parameters. Quan and Gengwei (2002) found the
model uncertainty for the crack widths of reinforced concrete beams to have a coefficient
of variance of 0.298 (or otherwise 0.3) and an estimated mean of 1.05. These results came
after a statistical study of 116 beams with varying configurations, strengths and applied
loads (Quan and Gengwei, 2002). The model uncertainty was found to follow a lognormal
probability distribution model. Thus subsequently, a mean of 1 and a maximum
coefficient of variance of 0.3 will be adopted for the reliability analysis in this thesis.

c) Statistical uncertainty results from there being some uncertainty in the ways in which
statistical parameters are estimated. Increases in the data base and sample size of the
cracking from restrained shrinkage through testing and recording of observations should
increase the accuracy of reliability assessments.

d) Examples of mathematical simplifications or assumptions made in modelling cracking

include, for instance, the crack spacing formula of EN 1992 which contains some
empirical fixed-value coefficients (McLeod, Viljoen, Retief, 2016). Such as the
coefficient k1, accounting for bond properties in EN 1992. A value of 0.8 is stipulated in
EN 1992 for instances of good bond. This coefficient is the equivalent of BS 8007’s fct/fb
 53

(taking on a value of 0.67 for type 2 deformed bars for class C35A concrete. Previous
research has also indicated that 0.67 could be safely applied to all strength classes of
concrete (Bamforth, 2007)). Even though the concrete tensile strength to reinforcing bond
strength ratio (fct/fb) was found in past research to decrease with an increase in concrete
strength class, EN 1992 gives a constant value (0.8 for good bond) that is to be applied
across all strength classes. This would then mean that at higher concrete strength classes,
the k1 coefficient provides an added margin of safety (or otherwise an added degree of
conservatism).

Additionally, creep is accounted for particularly in the restraint factor since it has the
effect of reducing restraint over time. However, where creep test methods are not given
in the South African and British standards, most creep test methods involve loading
concrete cylinders hydraulically and then measuring the deformation that results over
time (Owens, 2013). This would mean that the creep value obtained would be based on
compression rather than tension in the concrete (particularly tension arising from
restrained contraction in the concrete). Thus in applying this same creep factor to tension
cases (such as where there is restrained shrinkage) there could be a margin of error that
arises since the creep prediction model does not necessarily represent the tension case.
Furthermore, it had been found in past research that the tensile creep of concrete is lower
under restrained shrinkage as opposed to where the concrete is under constant stress
(Sajedi et al., 2011).

Bearing all of these sources of uncertainty in mind, the EN 1992-3 restrained shrinkage crack
model’s coefficient of variance value will be varied in the reliability assessment to gauge what
influence it has on the reliability performance of the crack model. This would be a particularly
relevant assessment given model uncertainty’s observed dominance in previous research (Retief,
2015).

3.6.3 Concrete cover (c)

The concrete cover is found to generally have a both-sided beta, or otherwise, gamma distribution
(Holický, 2009). British construction practices are categorised into either high quality (near-
laboratory precision), good, moderate or poor quality. The coefficients of variation associated
with each quality level are 10%, 15%, 20% and 30% respectively (McLeod, 2013). Assuming
that liquid retaining structures in South Africa are constructed under good quality management
practices, the coefficient of variation for concrete cover selected for use in this investigation will
be 15%. Concrete cover has also been found to follow a lognormal distribution (Holický 2007)
 54

and will be the distribution of choice in the reliability analysis of the EN 1992-3 crack model
performed herein (presented in chapter 5).

3.6.4 Limiting crack width (wlim)

An example of an observation of a typical crack pattern for a concrete member restrained at its
edge is shown in Figure 3.4 (Kamali et al., 2013). This was taken from the experimental data of
researcher Kheder (1997) who investigated the control of cracks induced by restrained
deformation. Kheder (1997) found that the largest cracks occurred in the middle of the concrete
member with inclined cracks appearing along the sides. The crack widths established in the
example below from experimental data suggests that there is considerable variability in the crack
widths found in a restrained member.

Figure 3.4: Example of Observed Crack Pattern and their Correlating Restraint Factor
(Kamali et al., 2013).

The limiting crack widths set forth in design standards have been established for both load induced
cracking and thermal cracking. However, past researchers considering the variability of various
crack width models have done so typically on cracking due to loading. The limiting (or allowable)
crack width used in these analyses were either regarded as having a stochastic nature or being
deterministic. Holický (2010) conducted a fuzzy probabilistic analysis (where a broad transition
region exists between the satisfactory and unsatisfactory state of a structure, rather than there
being an abrupt change in state) of the EN 1992 load induced crack model. The crack width limit
was said in this analysis to follow a beta distribution (the lower limit of the transition region was
0.05 mm and the upper limit was set at 0.2 mm). In assessing the reliability of cracking, the
maximum allowable crack width value for load induced cracks in the Chinese design code was
 55

regarded as being deterministic when used to calculate the reliability index of reinforced concrete
beams under service conditions (Quan and Gengwei, 2002). The First Order Reliability Method
of analysis was used in that particular study. Holický et al. (2009) regarded the limiting crack
width in the EN 1992 load induced crack model (looking particularly at cracking in a cylindrical
water retaining structure under pure tension) as being deterministic in a probabilistic analysis.
Thus, considering the above findings on the probabilistic nature of the permissible crack width,
the limiting crack width used for the reliability analysis in this thesis will also be regarded as
being deterministic.

3.6.5 Concrete tensile strength (fct,eff)

The concrete tensile strength may be found to follow a lognormal distribution- much like most
resistance variables (Holický, 2009). The characteristic value for the concrete tensile strength was
found in EN 1992 to be fctk, 0.05 = 2 MPa and mean value of 2.9 MPa for class C30/37 concrete (as
derived from table 3.1 of EN 1992-1-1:2004). As part of the revision of early age cracking design
guide, from CIRIA 91 to CIRIA C660, a probabilistic analysis of the design effective concrete
tensile strength (fct,eff) was conducted to determine its exceedance probability. The findings are
presented in Figure 3.5:

Figure 3.5: Probability Distribution of the Early Age In-Situ Tensile Strength of C30/37
Concrete (Bamforth, 2010).

Input data for concrete tensile strength into the probabilistic analysis assumed a normal
distribution and coefficient of variance of 18% (standard deviation 0.53). A mean of 1.06 was
obtained from the analysed data. The 5% fractile for the concrete tensile strength was found to be
0.65, whilst the 95% fractile was 1.54. Holický (2009) suggests the concrete strength follows a
 56

log-normal distribution and generally has coefficient of variances that range from 0.1 to 0.18.
Researchers investigating the reliability performance of the EN 1992 load induced cracking
serviceability limit state have also regarded the concrete tensile strength has having a log-normal
distribution, with a concrete grade of C30/37 (as with this investigation) having a mean of 2.9MPa
and coefficient of variance of 0.19 (Holický, Reteif & Wium, 2009 and McLeod, 2013). Also
undergoing the reliability analysis of the load induced cracking case, Zahalan (2010) regarded the
concrete strength as following a log-normal distribution. This particular investigation will also be
assuming a log-normal distribution for the concrete tensile strength, with a mean of 2.9 MPa and
coefficient of variance of 0.19.

3.6.6 Restraint degree/factor (R)

The restraint degree also has an inherent variability. The restraint degree’s variations depends on
the elastic modulus of the new concrete pour which varies with time. Figure 3.6 illustrates how
the change in concrete elastic modulus effects the restraint factor at the joint, calculated using the
ACI method (the ratio of the cross-sectional area of the restraining element to the cross-sectional
area of the restrained member, A0/An, was assumed to be 1). Restraint also varies with the length
to height ratio of the restrained member, with the degree of restraint decreasing with increased
distance away from the restraining element. A review of literature on the variability of the restraint
degree returned no real findings on the statistical parameters of this variable. One way to obtain
these statistical parameters would be to compile experimental data on the restraint degree and use
the data to obtain a theoretical probability model (Holický, 2009), a process that should be
undertaken in future. Therefore based on the above mentioned short coming, the restraint degree
will be treated as a deterministic variable. It is suggested that further research is done on the
probability distribution and statistical parameters for the restraint degree.
 57

Figure 3.6: Variation of ACI Calculated Restraint Degree with Change in Concrete Elastic
Modulus at Early Age (A0/An = 1), (Bamforth et al., 2010).

The remainder of the parameters to be used in the reliability analysis of the EN 1992 crack model
will be treated as deterministic variables. These include parameters such as coefficients and a few
other material properties.

3.7 Partial Safety Factors

For a design to be considered safe, it needs to be ensured that the action effect acting on the
structure or structural element is either equal to or does not exceed the capacity (resistance) of the
structure or structural element. To do this, a factor of safety may be applied onto both the demand
and supply values that increases the nominal demand value and decreases the nominal supply
value. The increased demand and decreased supply values are then used in the design of the
engineering facility. This is a notion that is adopted in EN 1990 to EN 1999 (as well as in SANS
10160-1: 2011), in which the design value of a material property becomes:

xk
Xd = .
M
In this case, the characteristic value for the material property is divided by a factor before being
used for design purposes (Holický, 2009). Here the characteristic value, with a recommended
probability of exceedance, for the material property may also be replaced by the nominal (X nom)
 58

value for the material property. The same may be done for the design load F d, where the design
point may be found using the formula:
Fd = γFFk.
Both γM and γF symbolise the partial safety factor for the material property and force respectively
(Holický, 2009).

Partial factors account for there being some level of uncertainty in the design model and for any
limitation in available data for a given level of reliability. For instance, in the Eurocodes the partial
factor for material properties (γM) is made up of the product of the material property factor (γm =
Xk
) and resistance model uncertainty, γRd (Holický, 2009). Similarly, the partial factor for loads,
X𝑑
Xd
γF, comes as the product of both the load intensity uncertainty (γ f = ) and the load model
X𝑘

uncertainty (γEd). The required factors for use in design must satisfy the inequality:

ϕR ≥ i  1 γ i E i ,
n

where ϕ is the supply factor and γi is the demand partial factor (Ang and Tang, 1984). In other
words, for the performance function/limit state function to meet a target reliability index, partial
safety factors are applied to the basic variables of the performance function. Thus, essentially,
the limit state criteria (g(Xi) > 0) must still be met where the characteristic basic variables have
been adjusted by the partial factors to meet the design value for the basic variables in question
(symbolically the limit state criteria for safe design then becomes (g(Xid) > 0). If, say, the basic
variables were represented by Xi, then the limit state equation:
g γ1μ x1 , γ 2 μ x2 ,..., γ n μ xn  = 0.
Each γ1 μ𝑋1 value is representative of the failure points on the performance function failure
surface, x* (the design value), where the target reliability index will most probably be met. Thus
the formula for the failure point,
x*= γ1μ x1
may be rewritten as:

x *i
γ=
μx
It may be found that the design failure points using FORM may be calculated using the subsequent
formula:
x* = μx (1– αi*βwXi).
And so consequently, the theoretical partial factor becomes:

x *i
γ= = 1– αi*βwXi,
μx
 59

where αi* is the direction cosine (sensitivity factor) for a failure point obtained at the end of the
iterative FORM algorithm and wXi is the coefficient of variation of the basic variable in question
(Ang and Tang, 1984).

The theoretical partial safety factor may be obtained using the following algorithm which is
essentially the reverse of the FORM method, where the reliability index is known beforehand (as
outlined by Ang and Tang (1984)):
1. The failure points are assumed (generally taken as the mean of the basic variables in
question)
2. Non-normal means and standard deviations of random variables are converted to their
normal equivalents.

 g 
3. The partial derivatives   of the performance function is determined and the
  X i 

g
 X i
directional cosines found (αi = )
2
 g 
i  
  X 
 i
4. Then the new failure point is determined via the equation:
x* = μNx – ασNxβ,
and substituted into the limit state equation and solved for an unknown basic variable
5. Steps 2 through to 4 are repeated until convergence of the basic variables in question is
reached.

On a very basic level, the calibration of partial factors for a design code would involve selecting
a set of partial factors such that the structural element under design has a reliability level that lies
as close as possible to the stipulated target reliability index whilst meeting the limit state criteria.
The reliability standard code, SANS 2394:2004, puts forth that an array of design conditions
should be considered so as to ultimately determine a combination of partial safety factors that
covers a large scope of expected performance applications. As indicated in SANS 2394:2004, the
set of partial factors that are found to generate a reliability index that has the least amount of
deviation from the target reliability would then be the best set of partial factors to be used in the
design of the structural element under consideration. The calibration process is one that involves
optimization, which is beyond the scope of this investigation.
 60

3.8 Previous Research on the Reliability of the Cracking Serviceability Limit State

Overall, investigations into the reliability assessment of the serviceability limit state returned
research geared more towards the load induced cracking case rather than that of the restrained
shrinkage (as evidenced by this section). This justifies the relevance of assessing the reliability of
the restrained cracking serviceability limit state. Fortunately, some inferences may be made from
the load induced cracking cases- where clues towards the reliability performance of the restrained
shrinkage cracking serviceability limit state may be found. The nature of reliability assessment is
such that the ultimate reliability index or failure probability arrived at after analysis depends
significantly on the probabilistic distribution (or otherwise, theoretical models) to which the basic
variables are assumed to follow (SANS 2394:2004). For example, direct comparisons may be
unfeasible even where the same limit state function was being considered among comparative
research works, but the shared basic variables utilised have been said to follow different
probability distributions - a likely occurrence due to some deficiencies in the knowledge of the
stochastic nature of some basic variables. The formulation of the limit state function also affects
the reliability indices obtained, thus making it difficult to directly compare research findings to
past works where a different limit state function was considered. For instance, the use of a
different physical model describing the same phenomenon (e.g. the various models in existence
modelling cracking) disallows the option of direct comparison. However, that being said,
inferences may be made relevant to restrained cracking. A summary of past investigations on the
reliability assessment of the cracking serviceability limit state has been compiled:

Holický, Retief and Wium (2009) assessed the reliability performance of the EN 1992 load
induced crack model. Here the crack widths of water retaining structures were investigated
probabilistically (using FORM) and compared to a deterministic analysis of the same
representative water retaining structure. The probabilistic method was determined to be more
economical than employing a deterministic design methodology. It was found in this research that
reinforcing required for the serviceability limit state for crack control exceeded that which was
required satisfy the ultimate limit state. The degree of exceedance increased with a decrease in
the permissible crack width limit. Research by Holický et al. (2009) also indicated that 2 to 5
times more reinforcement than the basic reinforcement required for the ultimate limit state was
necessary for crack limit compliance. This was true for the EN 1992 load induced cracking case,
with the enhancement factors applying to crack limits 0.2 mm and 0.05 mm respectively. Clearly,
from Holický et al.’s (2009) findings it may be observed that regarding LRS the serviceability
limit state is the more critical limit state with regards to liquid retaining structures. With restrained
cracking falling under the same serviceability design criterion as that of load induced cracking,
 61

one may presume that it too would require larger amounts of reinforcing as compared to that
which is required for the ultimate limit state design criteria in LRS design.

Holický (2010) conducted a probabilistic optimization of the EN 1992 load induced crack model
to ascertain what optimal value of the ratio of a generic reinforcement to the reinforcement
required for ultimate limit state compliance (ω = A/A0). This ratio may also be described as a
measure of what enhancement in magnitude was required of the reinforcement determined for
ultimate limit state agreement to meet the crack limit. It was determined that the reinforcement
calculated for ultimate limit state agreement needed to be significantly increased for crack limit
compliance. Moreover, in this analysis, Holický (2010) determined that a range of optimal
reliability indices of 0 to 3.5 was calculated for the EN 1992 load induced crack model depending
on the ratio of cost of failure to the cost per unit of ω = A/A0, (Cf/C1). For a high cost of failure,
the reliability indices calculated were as large as those generally required for ultimate limit states
(Holický, 2010). This discovery is one that may also be applied to the restrained shrinkage case
since it is a significant design criterion in LRS design and the cost of failure may most likely be
found to be high. Evidently, increased knowledge of the potential cost of serviceability failure for
liquid retaining structures will give an indication of what reliability index is most appropriate for
liquid retaining structures where the load induced and restrained shrinkage cracking cases are
both accounted for.

Using a target reliability index similar in magnitude to those used for ultimate limit states, Zahalan
(2010) conducted a reliability-based analysis on the reliability index, again, on load induced
cracking. The target reliability used in this exercise was 3.5 for beams. The limit state function in
this particular research was derived from Frosch’s (1999, as cited in Zahalan, 2010, p.55) equation
and principles of reinforced concrete analysis (specifically the force and moment equilibrium in
concrete sections). The Monte Carlo method was adopted for the analysis of the failure probability
of the crack model. In this particular investigation, concrete cover and reinforcement spacing were
found to have the most influence on the overall reliability of the crack model. Beam width,
effective depth, concrete strength and steel strength were found to have a lesser influence on the
reliability indices achieved by the crack model. The reinforcement area was found to have a
limited influence on the reliability of the crack model since only a certain amount of reinforcement
may feasibly be included in the concrete beam.

McLeod, Wium and Retief (2012) also performed a reliability analysis of the EN 1992 crack
model as part of research undertaken for the development of the proposed design code for liquid
retaining structures in South Africa. It was determined in this analysis that the crack width limit
and model uncertainty had a significant effect on the reliability of the EN 1992 crack model. The
 62

limiting equation of the effective depth was also found to bear some influence on the reliability
of the crack model for the tension load case. It is important to extend this analysis to include the
restrained shrinkage case to fully gauge these variables’ overall influence on the EN 1992
cracking serviceability limit state.

3.9 Concluding Remarks

Since the serviceability limit state was found in past research to be the more dominant limit state,
the question of what the appropriate reliability index for this limit state becomes an important
one. Clearly, being the more critical limit state, its target reliability index should be greater than
those set for conventional serviceability limit states. Hence it is important that reliability
performance of this serviceability limit state be assessed. An assessment of the influence of
various design parameters on the load induced cracking model have already been conducted in
past research for a variety of design codes. Thus, it is necessary that a similar investigation is
conducted for the restrained shrinkage case. An investigation on the influence of concrete cover,
the φ/ρp, eff ratio, the effective tension area, section thickness, the reinforcement area as well as
the restraint factor on the reliability of the crack model will be assessed in the subsequent chapter
4. Moreover, model uncertainty has been found in past research to contribute considerably to the
reliability of the EN 1992 load induced crack model - pointing towards its potential on the
restrained shrinkage crack model. Thus an investigation into the influence of model uncertainty
on the EN 1992 restrained shrinkage crack model should be conducted. An understanding of the
reliability performance of the EN 1992 shrinkage cracking model in the South African context
provides an opportunity for improving this code for South African use.
 63

Chapter 4: Parametric Study of the EN 1992-1-1 & 3 and

Corresponding Codes BS 8007 & BS 8110-2 Restrained Shrinkage
Crack Models

4.1 Introduction

The objective of this study was to closely examine relationships of interest within both the
restrained shrinkage crack model of BS 8007 and EN 1992. As far as possible, typical South
African conditions, materials and configurations where used in this study. The values for the
typical South African liquid retaining structures were those taken from a survey done on South
Africa’s practices in the construction of water reservoirs (Holicky, 2009). In addition to
understanding the relationship of parameters within the respective crack models, the parametric
study will serve to indicate which parameters bear the most influence on the EN 1992 crack
model. The parameters to which the crack model is most sensitive are indicative of where
sensitivities might lie within the reliability model. Consequently, the influence of these identified
parameters on the reliability of the EN 1992 crack model may then be more closely examined in
the reliability analysis.

Questions raised from a review of relevant literature included the debate around the equation for
maximum spacing in which EN 1992-1-1 includes concrete cover as an influencing parameter in
its estimation of crack spacing. In the superseded BS 8007 code, concrete cover was not included
in the crack spacing model. The influence held by concrete cover and the φ/ρp,eff (or φ/ρ in the BS
8007 case) parameter was examined. This, as before mentioned, should give an indication of what
bearing these parameters have on the reliability model and also be able to quantify the effects that
the inclusion of the cover value in the EN 1992 crack model has. Moreover, in observing both
codes of practice, it can be noted that the way in which the restrained strain in the end restraint
case is dealt with in EN 1992-1-1 differs markedly from BS 8007, with no two parameters shared
between codes. This is contrary to edge restraint, which both codes have dealt with in a similar
way. The implication of this changeover in methodology in the design of water retaining
structures in South Africa will be examined herein.

4.2 Design Parameters:

The parametric study was deterministic in nature, meaning that the inherent variability and
uncertainty in each input variable were disregarded. Instead, each input variable was regarded as
 64

having a fixed value. A list of both the material and physical parameters used in this study are
given below:

4.2.1 Material parameters

Concrete compressive strength (characteristic value)
A survey of South African water retaining structures revealed that typical concrete grade used
included either C25/30 OPC concrete or otherwise grade C30/37(Holicky, Retief & Wium, 2009).
The latter of the two was chosen as the concrete grade of choice. A concrete grade of C30/37 was
selected.

Concrete tensile strength (mean value)

For C30/37 concrete at 3 days fctm (3) = 1.73 MPa (fctm = fct,eff ). The tensile strength taken at 28
days is fctm (28) = 2.9 MPa for C30/[Link] are for concrete tensile strength are taken from
CIRIA C660, table 3.2 (Bamforth, 2007).

Reinforcement yield strength (characteristic value)

The reinforcement yield strength common in South Africa is 450 MPa.

Modulus of elasticity of steel

The modulus of elasticity of steel is 200 GPa.

Modular ratio
The modular ratio is αe = Es/Ec, where Es denotes the modulus of elasticity of the steel and Ec
relates to the modulus of elasticity of concrete at the appropriate age. This value can be estimated
by from EN 1992-1-1:2004 equation (under clause 3.1.3):

Ecm (t) = (fctm (t)/fcm) 0.3 Ecm,

where,
 Ecm (t) is the modulus of elasticity at ‘t’ days
 fcm (t) is the mean concrete compressive strength at ‘t’ days (EN 1992-1-1 (2004), clause
3.1.2, equation 3.1)
 Ecm is the modulus of elasticity at 28 days
 fcm is the mean concrete compressive strength at 28 days (taken from table 3.1 of EN
1992-1-1:2004)
 65

Ecm (3) = (22.8/38)0.3x33 = 28 GPa, This makes the modular ratio for C30/37 concrete αe = 200/28=
7.14 for concrete at 3 days. At 28 days Ecm = (38/38)0.3x33 = 33 GPa, making αe = 200/33 = 6.06
for C30/37 concrete.

Coefficient of thermal expansion, αc,T

The most used aggregate in South African concrete is quartzite and sandstone (Addis and Owens,
2001) and thus reading from table 4.4 in CIRIA C660 (Bamforth, 2007), the coefficient of thermal
expansion for quartzite containing concrete was (was αT,c = 14µɛ/˚C, which acted as the reference
thermal expansion coefficient for calculations). This was a proposed conservative design value
on the high-end of the observed range of concrete thermal expansion coefficients (after Browne
1972 as cited by Bamforth, 2007 in CIRIA C660).

Autogenous shrinkage, εca

The values for autogenous shrinkage strain obtained from table 4.5 of CIRIA C660 were ɛca =
15µɛ (at 3 days) for C30/37 concrete and ɛca = 33µɛ (28 days) – used in calculations, taking into
consideration long term effects. Or otherwise the autogenous shrinkage may be obtained via the
formulae for autogenous shrinkage in EN 1992-1-1:2004, under clause 3.1.4:
εca(t) = βas(t)εca(∞) where,

εca(∞) = 2.5(fck – 10)x10-6 and

βas(t) = 1-exp(-0.2t0.5).
The time, t, input is given in days. Using these formulae it may be found that:
 For t = 3 days, εca(3)=(1-exp(-0.2(3)0.5)x2.5(30-10)10-6 = 14.64με (≈15με)
 For t = 28 days, εca(28)=(1-exp(-0.2(28)0.5)x2.5(30-10)10-6 = 32.65με (≈33με)

T1
The most common formwork used in South Africa is steel formwork (Addis and Owens, 2001).
Table 4.2 of CIRIA C660 gives 340kg/m3 binder content for C30/37 CEM I (ordinary Portland
cement) concrete. Figure 4.5 from CIRIA C660 (this value is based on a mean ambient
temperature of 15°C and placing temperature of 20°C) gives, T1 value of 15°C.
 66

T2
Considering the concrete placement in summer, the T2 fall in temperature selected for the analysis
was 23°C (estimating from data obtained by [Link], 2015 and the Climate Change
Knowledge Portal, 2009).

Drying shrinkage strain, εcd

For relative humidity was said to be 80% for coastal areas (Addis and Owens, 2001) where the
section thickness h = 250 mm and the width of the section considered is b = 1000 mm. The
effective section thickness h0 may then be obtained by dividing twice the concrete cross-sectional
area by the perimeter of the parts of the cross-section that would be exposed to the drying (2Ac/u).
Applying this formula in this context gave the following result:

2Ac/u=2(250 x 1000)/ (2x1000) = 250 mm (considering a section of wall, top and bottom of
cross section not exposed)

Reading from figure 8.20 from Fulton’s Concrete Technology (Addis and Owens, 2001), the
drying shrinkage strain is interpolate between values for h0 = 150 and 300 in this instance where
h0 = 250 mm, yielding ɛcd = 220µɛ (30 year shrinkage) - this was the value to be adopted in the
subsequent parametric calculations. For inland areas, the relative humidity in South Africa is 60%.
For h0 of 250 mm lying between 150 mm and 300 mm, as before, the drying shrinkage read for
60% relative humidity was ɛcd = 340µɛ (30-year shrinkage).

Tensile strain capacity, εctu

The tensile strain capacity represents the maximum amount of strain that the concrete can sustain
before the formation of a crack (Bamforth, 2007). This value may be obtained by dividing the
mean tensile strength of the concrete by the mean modulus of elasticity of the concrete. The
values of the tensile strain capacity were taken from table 4.11 of CIRIA C660 (Bamforth, 2007)
where the effects of creep and sustained loading were accounted for.
 εctu = 76 με for C30/37 concrete at 3 days.
 εctu = 108 με for C30/37 concrete at 28 days.

4.2.2 Physical parameters

Section thickness, h
The typical section thickness in South Africa for water retaining structures was found to be
250mm (Holicky, Retief & Wium, 2009).
 67

Cover, c
The concrete cover was taken to be 40mm (this value takes into consideration the minimum
concrete covers for the durability of water retaining structures in accordance with BS 8007. It was
also found to be the typical choice for engineers in South Africa (Holicky, Retief & Wium, 2009)).

Diameter of reinforcement, φ
This value may vary depending on the parameter being studied, a reinforcing steel diameter of 16
mm was selected as the reference case.

Area of tension reinforcement, As

Varies as required in the comparisons considered. The maximum amount of area of steel
reinforcement (As) allowed in South Africa as stipulated by SANS10100-1 (clause [Link]) is
4% of the gross cross-sectional area of concrete (Ac). A feasible minimum limit of 75 mm spacing
for single bars of reinforcement was used.

Restraint degree, R
A maximum restraint degree, with creep accounted for, will be used. Otherwise, the reference
value for the restraint factor is taken to be 0.5 (for a concrete member under full restraint with the
effects of creep accounted for).

4.3 Methodology for Crack Width Estimation

Calculations pertaining to both end and edge restraint were done so using both the Eurocode (EN
1992-1-1:2004 and EN 1992-3:2006) and British (BS 8110-2:1985 and BS 8007:1987) codes of
practice.

4.3.1 Crack Estimation Following EN 1992

The Eurocode crack calculation procedure for restrained cracking went as follows:
 The crack width is the product of both the restrained strain and the crack spacing (S r, max
 εr). Thus, initially, both the crack spacing and restrained strain need be determined.
 Determine the retrained strain by substituting the appropriate, above mentioned, input
parameters into the strain equation for either end (αe, k, kc, fct, eff, Es, As and Ac) or edge
(αc, T1, T2, ɛcd, R and εca) restraint. The applicable restrained strain formula for end
restraint is
0.5α e k c kf [Link] 1  1/α e ρ) 
εr = (ɛsm - ɛcm) = ,
Es
 68

whilst that for edge restraint is

εr = (ɛsm - ɛcm) = Rax ɛfree.
 Determine the effective depth of the tension area, hc,eff, which is the lesser of either h/2 or
2.5(c+ φ/2).
 Crack spacing is then calculated the same way for both end and edge restraint conditions
using the equation
Sr, max = 3.4c + 0.425k1k2Φ/ρp, eff.
It must be noted that the ratio of steel reinforcement to gross concrete is represented by
ρ= As/Ac, which is not to be confused with the ratio of steel reinforcement to effective
concrete area ratio symbolised by ρp, eff =As/Ac,eff. The effective tension area can, of
course, be obtained by multiplying the depth of the effective tension area by the section
width (hc,eff  b). The depth of the effective tension area is determined as the lesser of
h/2 and 2.5(c + φ/2).

4.3.2 Crack Estimation Following BS 8007 and BS 8110-2

The crack calculation procedure for restrained cracking as per BS 8110-2:1985 and BS 8007:1987
respectively went as follows:
 As in the case for EN 1992-1-1:2004 and EN 1992-3:2006, the crack width is the product
of the restrained strain and crack spacing.
 Once again, the restrained strain (εr) resulting from end or edge restraint must be
calculated. The strain calculation is the same for both end and edge restraint under the
British codes,
ɛr = Rα (T1 +T2).
 The crack spacing may then be calculating using
f ct,eff 
Smax =  .
fb 2ρ

4.4 Influence of cover versus φ/ρp, eff

The modelling of crack spacing model is an aspect of crack estimation that differs most across
design codes (Beeby & Narayanan, 2005). Such a vast set of possible ways in which crack spacing
may be determined is cause for an investigation. This is particularly of interest here where the
crack spacing equations adopted by BS 8007 and EN 1992-1-1 have some noticeable differences.
An investigation into the influence of cover and the φ/ρp,eff ratio on crack spacing had been
undertaken.
 69

4.4.1 Influence of Cover on Crack Spacing

The first part of this exploration involved assessing the influence of cover on the crack spacing in
the EN 1992-1-1: 2004 equation. To examine the influence of cover on crack spacing, the cover
was varied and its effect compared against a comparable set of φ/ρp, eff ratios. The section
thickness was kept constant at 250 mm and the reinforcing bar diameter remained 16 mm
throughout the analysis. The reinforcing bars were assumed to be spaced at 250 mm centre to
centre. A section width (b) of 1000 mm was selected for this analysis. The φ/ρp, eff ratio is limited
by the effective tension area and only those ratios that were close in value were included in the
study. Given the concrete covers chosen and the choice of bar diameter (parameters onto which
the effective tension area depends) it was determined that the φ/ρp, eff ratio became constant after
the cover value of 50 mm since the limiting effective depth from a 50 mm cover onwards was
limited to 2.5(c + ϕ/2). Hence, the ratio developed from concrete cover values greater than 50
mm was compared against a constant φ/ρp, eff ratio and the subsequent findings (as presented in
Table 4.1) were made:

Table 4.1: Influence of Cover on EN 1992 Crack Spacing Model

he,eff (mm)

cover bar As/face h/2 2.5(c + φ/2) ρp,eff φ/ρp,eff Sr,max Term Term
(mm) dia. (mm2) (mm) (mm) ‘X’ % ‘Y’ %
(mm)
40 16 804 125 120 0.0067 2387 948 14 86
50 16 804 125 145 0.0064 2487 1016 17 83
60 16 804 125 170 0.0064 2487 1050 19 81
70 16 804 125 195 0.0064 2487 1084 22 78
80 16 804 125 220 0.0064 2487 1118 24 76
100 16 804 125 270 0.0064 2487 1186 29 71
Notes:
 Cover varied as presented whilst only the shaded values considered in study
 Reinforcement spacing was set at 250 mm centre to centre
 he, eff is limited to the lesser of h/2 or 2.5(c + φ/2). The value for he, eff stabilised after cover = 50 mm and
so values φ/ρp, eff became constant thereafter.
 Term ‘X’ = 3.4c
 Term ‘Y’ = 0.425k1k2ϕ/ρp, eff
 70

If one were to consider the two terms (the cover and effective reinforcement ratio term) in the
EN 1992-1-1: 2004 crack spacing model separately, an estimation of either one’s influence may
be more clearly assessed. Considering that the crack equation is Sr, max = 3.4c + 0.425k1k2ϕ/ρp, eff,
it may be separated such that the cover term is represented by term ‘X’ = 3.4c and the second half
of the crack spacing which deals with the effective reinforcement ratio is represented by term ‘Y’
= 0.425k1k2ϕ/ρp, eff, then the influence held by each term on the overall crack spacing may be
assessed. It is evident from the results that an increase in the concrete cover term ‘X’ value brings
about an increase in crack spacing calculated, although this increase in crack spacing is marginal.
For cover values 50 and greater used in the assessment presented in Table 4.1, the limiting
effective depth was h/2 meaning that the concrete cover had no influence on term ‘Y’ since it did
not feature. It is evident from results that concrete cover makes a relatively small contribution on
the crack spacing in the EN 1992 crack spacing model. The second term of the crack spacing
formula, term ‘Y’, carries a greater influence on the crack spacing model.

A graphical representation of this data is displayed in Figure 4.1. Here, the estimation of crack
spacing as done under both EN 1992 and BS 8007 was included. Since BS 8007 does not include
the cover variable, it was independent of this variable and thus remained constant as the concrete
cover value varied. It can be seen, in the EN 1992 case that increasing the cover resulted in an
increase in crack spacing. However, this increase in crack spacing was gradual. It may also be
deduced from Figure 4.1 that the EN 1992 crack spacing model predicts larger crack spacing as
compared to those calculated from BS 8007. The difference between the predicted crack spacing
as calculated from EN 1992 and BS 8007, of course, would increase with an increase in cover
value.
 71

1400

1200

1000
Crack Spacing (mm)

800

600

400

200

0
0 20 40 60 80 100 120
Cover (mm)

EN 1992 BS 8007

Figure 4.1: Influence of Cover on Crack Spacing for Both EN 1992 and BS 8007

4.4.2 Influence of the φ/ρp, eff Ratio on Cracking

To determine the influence of the φ/ρp, eff ratio, the bar diameter was varied in order to vary the
φ/ρp, eff ratio. The cover was kept constant at 40 mm and the bar spacing was set at 250 mm centre
to centre. The section thickness in this study was fixed at 250 mm. The results of this analysis are
presented in Table 4.2. Considering the contribution the φ/ρp, eff ratio makes to the estimated crack
spacing (i.e. regarding term ‘Y’ = 0.425k1k2ϕ/ρp, eff) – it can be clearly seen that this ratio is once
again a sizeable contributor to the overall value of the crack spacing. Concrete cover, once again,
has no impact on term ‘Y’ since the limiting effective depth was h/2 for the selection of concrete
covers, bar diameters and section thickness considered in this analysis.
 72

Table 4.2: The Influence of the φ/pρ, eff Ratio on Crack Spacing as per EN 1992
he,eff (mm)
cover bar As/face h/2 2.5(c + φ/2) ρp,eff φ/ρp,eff Sr,max Term
(mm) dia. (mm2) (mm) ‘Y’ %
(mm)
40 16 804 125 120 0.0067 2387 948 86
40 20 1257 125 125 0.0101 1989 812 83
40 25 1963 125 131 0.0157 1592 677 80
40 32 3217 125 140 0.0257 1243 559 77
40 40 5027 125 150 0.0402 995 474 71
Notes:
 250 mm center to center spacing for reinforcement
 Term Y = 0.425k1k2ϕ/ρp, eff

The graphical representation of the effect of the φ/ρp, eff ratio on the crack spacing is displayed in
Figure 4.2. It is clear from Figure 4.2 that increases in the φ/ρp, eff brought about an increase in
the predicted crack spacing, as would be expected.

The influence of the φ/ρp, eff (or φ/ρ) ratio on crack spacing for both the EN 1992 design code and
BS 8007 was compared. The Figure 4.2 shows this comparison between the two codes, here the
section thickness was kept constant at 250 mm and the cover to reinforcement remained 40 mm
throughout the analysis. It is evident from Figure 4.2 that the crack spacing values obtained
through the EN 1992 are greater than those obtained by way of BS 8007. This may be attributed
to the inclusion of the cover term in the EN 1992 crack spacing estimation. For instance,
considering a reinforcing bar diameter to effective reinforcement ratio of 1592 mm, EN 1992
predicted a crack spacing value of 677 mm whilst the BS 8007 crack spacing model estimated a
value of 573 mm (about decrease in value of factor 1.18) – φ/ρp, eff was equal to φ/ρ in this instance
since the limiting effective depth was h/2 under both EN 1992 and BS 8007 for reinforcing bar
diameter 25 mm. Concrete cover plays an even greater role on the EN 1992 calculated crack
spacing at lower reinforcing bar diameters (namely, the 16 and 20 mm wide bars considered in
this analysis), thus the difference between the calculated crack spacing of EN 1992 and BS 8007
increases at lower reinforcing bar diameters.
 73

1000

900

800

700
Sr,max (mm)

600

500

400

300

200

100

0
0 500 1000 1500 2000 2500 3000
ϕ/ρ (mm)

EN 1992-1-1 BS 8007

Figure 4.2: The influence of φ/ρp, eff (or φ/ρ) Ratio on Crack Spacing for BS 8007 and EN
1992-1-1 (40 mm Cover and 250 mm Section Thickness).

It is clear from the analysis conducted that the influence of the bar diameter to effective steel
content ratio is quite great than that of concrete cover under both codes, making up a large portion
of the final crack spacing predicted. This would ultimately affect the crack widths estimated under
both codes and in turn the amount of reinforcement required to maintain the stipulated rack width
limit. Based on the analysis conducted above, it appears as though EN 1992 is the more
conservative of the two codes, requiring more reinforcement to maintain the crack width limit.

4.5 Depth of effective area

Another way in which the British code differs from the Eurocode is the way in which the effective
tension area is defined. This difference in definition affects the overall value of the crack spacing
and thus the crack width. Therefore, a comparison of the effective tension areas between the codes
was undertaken. Table 4.3 gives a summary of the findings on the effect of concrete cover and
section thickness on the effective depth of the tension zone. The discussion corresponding to the
results is referred to in the subsequent section.
 74

Table 4.3: Influence of section thickness and cover on effective depth of tension zone (φ =
16 mm)
EN 1992 BS 8007
Cover (mm) h (mm) hc,eff (mm) hc,eff
(mm)
h/2 2.5(c + φ/2) h/2
250 125 120 125
40 300 150 120 150
350 175 120 175
400 200 120 200
450 225 120 225
500 250 120 250
250 125 145 125
300 150 145 150
50 350 175 145 175
400 200 145 200
450 225 145 225
500 250 145 250
250 125 170 125
300 150 170 150
60 350 175 170 175
400 200 170 200
450 225 170 225
500 250 170 250
Note: Limiting effective depth highlighted.

4.5.1 Influence of Section Thickness and Concrete Cover on Effective Tension Depth
To examine how the choice of section thickness and concrete cover influences the value of the
effective tension depth, the concrete cover (40, 50 and 60 mm) and reinforcing bar (16 mm) were
kept constant whilst the section thicknesses were varied. This was done under the guidelines of
both EN 1992 and BS 8007 and subsequently compared. EN 1992 proposes that the depth of the
effective depth for members in tension is the lesser of h/2 and 2.5(c + φ/2) (denoted by the
highlighted cells in Table 4.4). In the BS 8007, for section thicknesses of walls and suspended
slabs less than 500 mm thick, the effective tension area is taken as being half the section thickness
(h/2). Where the section thickness is greater than 500 mm the effective tension height is said to
be 250 mm thick.
 75

Reading from Table 4.3 above, it was apparent here that for EN 1992 the limiting effective tension
depth was where hc,eff = 2.5(c + φ/2). The effective depth value was 120 mm for all section
thicknesses where the bar diameter was set at 16 mm and concrete cover was 40 mm. The effective
tension depth found using BS 8007 showed values consistently greater than those determined by
EN 1992.

Where the concrete cover was 50 mm, the depth of effective tension depth in accordance to BS
8007 was expectedly unaffected by a change in concrete cover or the choice of bar diameter since
these variables were not included in the formula for effective depth in BS 8007. For EN 1992, the
effective tension depth was limited by hc,eff = 2.5(c + φ/2) only after a section thickness of 300
mm. Section thicknesses determined using BS 8007 were found to either be equal to or greater
than those obtained by EN 1992, as in the previous cases. As the cover was further increased to
60 mm it can be seen that for EN 1992 the effective tension was also mostly limited to h c,eff =
2.5(c+ φ/2).

Conclusively, it may be determined that the effective tension depths determined using EN 1992
were generally smaller than those determined using BS 8007. The limiting effective depth was
2.5(c + φ/2) for most combinations of section thickness and concrete cover values considered.
Thus, the bar diameter to effective steel ratio (φ/ρp, eff) as determined by EN 1992 would generally
be smaller than those obtained using BS 8007 given that the bar diameter and steel content was
the same in both cases. Although, even where EN 1992 may estimate a smaller effective steel
content ratio (φ/ρp, eff), the exclusion of the concrete cover term in the BS 8007 variation of the
crack spacing model would mean that EN 1992 still predicts greater crack spacing values.

4.5.2 Influence on Reinforcement Bar Diameter on Effective Tension Area

The influence of the reinforcement bar diameter may be found by varying the bar diameter whilst
maintaining the same section thickness (250 mm) and cover value (40, 50 and 60 mm). It was
determined that bar diameters 20 mm and greater generally gave equal effective tension depths
of h/2 for both codes EN 1992 and BS 8007 (Table 4.4). This may be attributed more so to the
increasing concrete cover value rather than the choice of reinforcing bar diameter (since where
hc,eff = 2.5(c + φ/2) the cover value has a larger helping of the overall value of the effective tension
depth) .
 76

Table 4.4 Influence of bar diameter on effective depth (h = 250 mm)

EN 1992 BS 8007
Cover Bar hc,eff (mm) hc,eff
(mm) diameter (mm) (mm)
h/2 2.5(c + ϕ/2) h/2
40 16 125 120 125
20 125 125 125
25 125 131 125
50 16 125 145 125
20 125 150 125
25 125 156 125
60 16 125 170 125
20 125 175 125
25 125 181 125

4.6 Comparison of BS 8007 and EN 1992 Edge Restraint Estimation on Crack Width
4.6.1 Influence of Section Thickness on the EN 1992 Edge Restraint Crack Model
To assess the influence of section thickness on the edge restraint crack model, the section
thickness was varied whilst the, amongst other variables, the cover was kept constant at 40 mm
and the reinforcing bar diameter remained 16 mm throughout. Thus, the effective depth was
limited to 2.5(c + φ/2) for all considered section thicknesses.

Increases in the section thickness results in a decrease in the crack width calculated. The restrained
strain as calculated from EN 1992-3 gave a constant restrained strain value for all section
thicknesses considered. It was uncovered, in this analysis, that there was little difference in the
amount of area required to achieve a 0.2 mm crack width for each section thickness considered.
More specifically, a range of reinforcing from 1.4 to 0.7 %As was required to meet the crack width
limit 0.2 mm for 250 to 500 mm thick sections respectively (a ± 15% average relative decrease in
reinforcement with each 50 mm increase in section thickness as observed in Figure 4.3).
 77

1.4

1.2

1
Crack Width (mm)

0.8

0.6

0.4

0.2
wlim = 0.2 mm
0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
As (ratio of reinforcement to concrete area)

h = 250mm h = 300mm h = 350mm

h = 400mm h = 450mm h = 500mm

Figure 4.3: Influence of Section Thickness on Crack Width for Edge Restrained Crack
Model (EN 1992)

The only real difference that is presumed to have come from changing the section thickness was
the amount of tension steel area required to maintain a particular ratio of steel to gross concrete
cross-sectional area for each section thickness considered. This effects the effective steel ratio
(ρp,eff) and thus the crack spacing (and the eventual crack width) calculated.

The limiting crack width of 0.2 mm was also one that satisfied the BS 8007 crack limit – as was
previously established. For BS 8007, the amount of area required to achieve a crack width of 0.2
mm is about 0.8% for all section thicknesses (as was illustrated in Figure 4.4). Thus section
thickness had no impact on the crack width for the BS 8007 edge restraint case. Here, we find
that EN 1992 was quite conservative when compared to BS 8007 requiring ±63, 50, 25, 13, and 0
% more reinforcement for crack width limit satisfaction for 250, 300, 350, 400 and 450 mm thick
section respectively. Whilst for a 500 mm thick section, BS 8007 required 14% more
reinforcement than EN 1992 for crack width limit compliance.
 78

1.4

1.2

1
Crack Width (mm)

0.8

0.6

0.4
wlim = 0.2 mm
0.2

0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
% As (ratio to concrete area)

h=250mm h=300mm h=350mm

Figure 4.4: Influence of Reinforcement Area on Crack Width for Edge Restrained Crack
Model (BS 8007)

4.6.2 Influence of Restraint Factor on EN 1992 Edge Restraint Crack Model

The section thickness was kept constant at 250 mm thick whilst the amount of reinforcement used
was varied (φ =16 mm throughout). This exercise is done for a varied array of restraint factors
ranging from R = 0.1 up to R = 0.5.

As may be expected, an increase in the restraint factor would result in there being an increase in
the estimated crack width. The results of this analysis are presented in Figure 4.5. A larger
restraint factor would bring about an increase in the restrained strain and thus a bigger crack width
value being calculated and an increase in reinforcement required for a particular crack width to
be met. A reinforcement of 0.22% As for a restraint factor of R = 0.1, 0.46% As at R = 0.2, 0.74%
As for R = 0.3, 1% As at R = 0.4 and 1.30% As for R = 0.5 was what was required to meet the 0.2
mm crack width using EN 1992 (a 59% average relative increase in reinforcement required with
every 0.1 increment in restraint factor).
 79

1.4

1.2

1
Crack Width (mm)

0.8

0.6

0.4
wlim = 0.2 mm
0.2

0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
As (ratio of steel reinforcement to concrete area)

R=0.1 R=0.2 R=0.3 R=0.4 R=0.5

Figure 4.5: Influence of Restraint on the Edge Restrained Crack Model (EN 1992)

4.6.3 Influence of Restraint Factor on BS 8007 Edge Restraint Crack Model

A steel reinforcement amount of about 0.16% at a restraint factor of 0.1 to 0.80% at R = 0.5 would
result in the crack width limit being met as shown in Figure 4.6. Comparing this result to those
found in the EN 1992 case, A range of from 38% more reinforcement was required to meet the
0.2 mm crack limit at R = 0.1 under EN 1992 to about 63% more reinforcement was necessary
for the crack limit compliance where R = 0.5 (a significant jump at full restraint).
 80

1.4

1.2

1
Crack Width (mm)

0.8

0.6

0.4
wlim = 0.2 mm
0.2

0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
As (ratio of steel reinforcement to concrete area)

R=0.1 R=0.2 R=0.3 R=0.4 R=0.5

Figure 4.6: Influence of Restraint on the Edge Restrained Crack Model (BS 8007)

It was anticipated that both EN 1992 and BS 8007 would yield very similar results. This
expectation comes particularly considering the fact that the formulae for the edge restraint crack
width model under EN 1992 and BS 8007 are quite similar in composition. However, it was found
that the EN 1992 crack model still gave more conservative results. This might be due to the
inclusion of autogenous shrinkage in the estimation of the free unrestrained strain in the EN 1992
crack model, which is not included in the BS 8007 crack model.

4.7 Comparison of BS 8007 and EN 1992 End Restraint Estimation of Crack Width

Another point of interest, in comparing the major differences between the EN 1992 code of
practice with BS 8007, is the change in the restrained strain estimation for a concrete member
subject to end restraint. It may be observed that no one parameter is shared between either of the
restrained strain code formulas for end restraint. A comparison between the two different
formulae was conducted. For this comparison, the section thicknesses were varied with a constant
40 mm concrete cover and a 75 mm centre to centre reinforcement spacing (As = 2680 mm 2/
section face).
 81

It was determined that the restrained strain from end restraint calculated using EN 1992 estimated
much greater crack widths as compared to those crack widths determined using BS 8007 (as may
be deduced from Figure 4.7). For instance, in the case where reinforcement spacing was 75 mm
centre to centre, cover is 40 mm and 16 mm reinforcing bar diameter, a crack width estimated
under EN 1992 for 250 mm concrete section will amount to 0.15 mm. Whilst under the same
conditions, BS 8007 will determine that the crack width produced by end restraint will equal 0.07
mm (about half of the value of the crack width determined under EN 1992). Thus, the EN 1992
estimation of end restraint was found to be more conservative than that of BS 8007.

0.25

0.2
Wlim = 0.2 mm
Crack Width (mm)

0.15

0.1

0.05

0
0 100 200 300 400 500 600
h (mm)

EN 2 BS 8007

Figure 4.7: Comparing EN 1992 and BS 8007 End Restraint Equation (40 mm cover, 75 mm
reinforcement spacing)

4.7.1 Influence of Section Thickness on the EN 1992 End Restraint Crack Model
The influence of section thickness on the end restraint crack width calculations for EN 1992
against BS 8007 were extended for section thicknesses ranging from 250 mm to 500 mm.

Increasing the section thickness decreased the calculated crack widths of members subjected to
end restraint. Increases in section thickness resulted in an increase in reinforcement required
achieve the crack width limit of 0.2 mm. Clearly, those reinforcement ratios for which section
thicknesses 250 mm to 450 mm meet the 0.2 mm crack limit were beyond the range of
 82

reinforcement ratio considered in this analysis (Figure 4.8). A quick calculation of the reinforcing
ratio at which the crack width limit 0.2 mm may be met returned values in the range of 1.8 to
1.3% As for section thicknesses 250 to 450 mm respectively (a ± 9% average relative decrease in
reinforcement per 50 mm increase in section thickness– smaller than the edge restraint case). For
section thickness 500 mm, it may be read from Figure 4.8 that 1.2% As was required to meet the
0.2 mm crack width limit.

1.4

1.2

1
Crack Width (mm)

0.8

0.6

0.4

wlim = 0.2 mm
0.2

0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013
As (ratio of steel reinforcing to concrete gross cross-sectional area)

h=250mm h=300mm h=350mm h=400mm h=450mm h=500mm

Figure 4.8: Influence of Section Thickness on Crack Width for End Restraint (EN 1992)

The decreasing value of the k coefficient as the section thickness increases decreases the amount
of strain calculated for thicker sections (k varies from to 0.65 for h ≥ 800 mm up to 1 for h ≤ 300
mm). Increases in section thicknesses means that both the ρp,eff (As/Ac,eff) and the ρ (As/Ac) would
increase since the amount of steel included in these ratios was increased for the same steel to gross
concrete ratio– this was especially true for the effective steel content ratio, ρp,eff . The increase in
these ratios results in a decrease in the amount of restrained strain and crack spacing calculated
since the reciprocal of these ratios are included in the estimation of both. However, the increase
in the amount of steel reinforcing required for any given steel to concrete ratio as section thickness
was increased had the seemingly stronger impact of increasing the overall crack width.
 83

4.7.2 Influence of Section Thickness on the BS 8007 End Restraint Crack Model
The end restraint estimation in BS 8007 follows the same approach as for edge restraint. The
relationship of reinforcement to crack width for various section thicknesses will be as mentioned
for edge restraint (in the previous section) where a reinforcement ratio of 0.8% As was necessary
for the 0.2 mm crack width to be met. This would mean that EN 1992 requires from 50% (at 500
mm) to beyond 63% (for section thickness 400 mm down to 250 mm) more reinforcement than
BS 8007 to meet the 0.2 mm crack width limit. It is also evident from this result that the EN 1992
end restraint crack model requires more reinforcement to meet the crack width limit than the EN
1992 edge restraint crack model when compared against the BS 8007 crack model– indicating
that it was the more conservative of the two EN 1992 models.

4.8 Parameter Sensitivities in Crack Model: Summary and Concluding Remarks

Along with comparing the British Code and Eurocode crack model, the parametric study was
conducted to assess the sensitivities of the various parameters of the crack model. The model’s
sensitivity to various parameters may ultimately affect the failure probability calculated in the
reliability analysis of the crack model. Thus, a brief outline of these sensitivities is presented
below:

 Concrete cover:
Concrete cover proved to have a relatively small effect on the predicted crack spacing, especially
where compared to the φ/ρp, eff ratio term in the crack spacing model. Although its influence on
the reliability of the crack model must still be considered since it was determined from
experimental data that concrete cover was a parameter not to be ignored in the assessment of crack
spacing (Caldentey et al., 2013). Thus, variations in the concrete cover value were used to
examine what influence this parameter has on the reliability of the crack model. A reference
concrete cover value of 40 mm was selected for the reliability analysis based on the survey of
water retaining structures (McLeod, 2013).
 84

 φ/ρp, eff ratio:

The investigation into the φ/ρp, eff ratio’s influence on the crack spacing revealed that this ratio
contributes considerably to the estimated crack spacing. This, in turn, would result in φ/ρp, eff
contributing considerably to the ultimate crack model.

 Effective tension area:

The effective tension area had been found to be predominately limited to 2.5(c + φ/2) for similar
covers and bar diamters. In the cases where the bar diameter or the concrete cover values are
increased, h/2 becomes the limiting effective tension area. For a combination of 40 mm cover and
bar diameter of 16 mm or 20 mm, the effective tension area will always give 2.5(c + φ/2) for any
section thickness. For a 250 mm thick concrete section, 40 mm concrete cover and 20 mm
reinforcing bar diameter the effective depth for hc,eff = 2.5(c + ϕ/2) and h/2 both return a value of
125 mm. Thus this combination section thickness, cover and bar diameter will be considered in
the reliability analysis; giving opportunity for the direct comparison of reliability models
containing either effective depth variations. In this way, the effect of the effective tension areas
may be assessed. The effective tension area has an obvious effect on the φ/ρp,eff ratio which had
already been found to contribute considerably to the crack model.

 Crack width limit:

The crack width limit was found to have a considerable influence on the restrained strain crack
model for both EN 1992 and BS 8007 design codes. This influence applied both to the edge and
end restraint conditions. Thus, as found in previous research, the adoption of the more onerous
crack limits of EN 1992 would subsequently have a significantly negative financial implication
in design. A decrease in the crack limit results in an increase in the demand for reinforcing (and
thus an increase in the cost of construction).

 Section thickness:
The section thickness has been found to have some influence on the crack model for both the end
and edge restraint. A variation of the section thickness will be used in the reliability model to
gauge what influence it bears on the crack model’s overall reliability level. Section thickness was
found to bear more of an influence on the end restraint crack model than the edge restraint crack
model. A reference value of 250 mm section thickness will be used in reliability analysis.

 Bar diameter:
Increasing bar diameter, as before mentioned, resulted in h/2 being the limiting effective tension
area. Little variability in the influence on the crack model was experienced where the bar
 85

diameters were varied. This was particularly true for bar diameters beyond 20 mm where the same
effective tension area is produced each time (h/2).

 Reinforcement area:
There was a noticeable correlation between the EN 1992 crack model and the amount of
reinforcement used. Clearly, the decline in the crack widths estimated by the crack model came
with an increase in the amount of tension reinforcement used. It had been determined in past
investigations on the reliability of reinforced concrete members that the failure probability is
sensitive to, among other basic variables, the reinforcement ratio in the member (Holický, Retief
and Wium, 2010). Therefore, it is suggested that the sensitivity of the reliability of the crack
model be tested against a variation of reinforcement ratios for various basic variables. For the
selected reference case of 250 mm section thickness, 40 mm cover and 20 mm rebar diameter, a
2 2
reinforcement ratio of 1748 mm and 2202 mm per section face for edge and end restraint
respectively for the 0.2 mm crack limit to be met.

 Restraint degree:
The restraint degree is not a parameter used in the estimation of end restraint in EN 1992. The
edge restraint crack model has been proven to be noticeably influenced by the restraint under edge
restraint conditions. Historically, according to BS 8110-2, the restraint degrees (restraint factors)
have been found to be greater than a value of 0.7. EN 1992 includes creep into its restraint degree,
giving a value of full restraint of up to 0.5. A fully restrained member will be considered in the
reliability analysis. However, much like in the parametric study, the influence of the restraint
factor on the reliability of the edge restraint crack model will be investigated by varying the
restraint factor from 0.1 to 0.5. Practically, this considers an array of configurations, and thus
restraint conditions under which the restrained member is nearly free to move and then
increasingly restricted up to full restraint.

Following after the parametric study, it was determined that the end restraint model was the most
conservative of the two external restraint conditions assessed using EN 1992 (as BS 8007
modelled edge and end restraint in the same way). In other words, more reinforcement is required
for a member subjected end restraint to meet the crack limit as compared to the edge restraint
condition. Also, larger crack widths were determined using the end restraint model. Comparing
BS 8007 and relevant parts of BS 8110-2 to corresponding codes EN 1992-3 and EN 1992-1-1,
Eurocode was found to be more conservative than the British codes. From the parametric study
conducted above, a reasonable selection of design variables for which the reliability analysis
could be conducted may be established. It was paramount that the effect of influential parameters
be tested in the reliability analysis of the EN 1992 crack model. Hence, the variables found to be
 86

particularly influential to the EN 1992 crack model and thus potentially influential to the
reliability of the crack model include the following: concrete cover, the φ/ρp, eff ratio, the effective
tension area, section thickness, the reinforcement area as well as the restraint degree.
 87

Chapter 5: FORM Analysis of EN 1992 Crack Model: Methodology,

Results and Discussion

5.1 Introduction:

The absence of an equivalent design code for liquid retaining structures in South Africa formed
the primary basis for the reliability analysis conducted herein. In the parametric study of chapter
4, it was found that the crack width limit held a considerable influence on the EN 1992 edge
restraint crack model. This influence was experienced less so on the end restraint crack model,
but still noteworthy. The more stringent crack width limits of EN 1992 brought with it increased
financial demand proportional to the increased requirement for steel reinforcement for the
compliance of the permissible crack width limits. The restraint factor was also determined to be
a particularly influential parameter in the edge restraint crack model in which it appears. Thus,
investigations into the impact had by the above-mentioned variables on the reliability of the crack
model were undertaken in this chapter.

In an attempt to further understand the reliability performance of the EN 1992 crack model in the
South African context, the influence of variables such as the concrete cover, the bar diameter to
effective steel ratio and the section thickness (which were already been assessed deterministically
in the parametric study) will be assessed probabilistically (taking into account their stochastic
nature). This should indicate the ways in which the EN 1992 model may be adjusted to bring
about compliance with South African reliability standards. Moreover, identifications of the
circumstances under which the target reliability index is met under South African conditions may
be found through the reliability analysis.

Investigations into past research on the matter revealed that, considering the load induced case of
the EN 1992 crack model, the serviceability limit state is the more dominant limit state for liquid
retaining structures (McLeod, 2013). Bearing this in mind it had also been argued that a higher
reliability class and target index be considered for use when designing liquid retaining structures
(Barnardo-Viljoen, Mensah et al., 2014). These arguments strengthen the need for a reliability
analysis of this serviceability limit state, particularly as applied to the South African environment.
Additionally, model uncertainty had also been identified in the literature to be a particularly
influential parameter in the reliability of reinforced concrete structures. Hence, its impact on the
reliability of the cracking serviceability limit state was also investigated.
 88

The First Order Reliability Methodology had been employed to assess a representative liquid
retaining structure which had come as a by-product of the parametric study which gave a sense
of realistic sets of variables to be used in the reliability analysis of the EN 1992 crack model. The
representative structure to be used in the analysis is further described in the subsequent text.

5.2 Methodology of Reliability Analysis:

5.2.1 Reliability Analysis Formation

The calculated crack width is described in EN 1992 as being a simple compatibility equation in
which the crack width comes as the product of both the crack spacing (Srm) and the restrained
strain (εr).
w = Srmεr (5.1)
The maximum crack spacing equation found in EN 1992-1-1 is used to calculate the characteristic
crack width. This is the crack width that has a 5% probability of being exceeded and does not
represent the average crack width, being a maximum value.
Sr,max = 3.4c+0.425k1k2φ/ρp,eff (5.2)
In the development of the crack spacing formulae, it was found through experimentation that the
maximum crack spacing may be estimated to be about 1.7 times the mean crack spacing (Beeby
and Narayanan, 2005). The mean crack spacing formula will be adopted in the reliability analysis
of the EN 1992 crack model undertaken in this investigation and is given by the following
formula:
Srm = 2c+0.25k1k2φ/ρp,eff (5.3)
The mean restrained strain (εr) for edge and end restraint are as those given in EN 1992-3 in which
the restrained strain for edge restraint is given by,
εr = (εsm - εcm) = Rax ɛfree (5.4)
Where εsm is the mean strain in the reinforcement and εcm represents the mean strain in the concrete
between cracks. Rax marks the restraint factor (restraint degree) which is indicative of the degree
of freedom of movement which is experienced by the concrete member. And ε free = εcd+εca+αT,c(
T1+T2), where εcd is the drying shrinkage strain, εca is the autogenous shrinkage strain; αT,c is the
coefficient of thermal expansion of concrete, T1 denotes the fall in temperature from hydration
peak to mean ambient temperature in the concrete. T2 is the seasonal fall in temperature.
The mean strain resulting from end restraint is estimated in a different way to edge restraint and
is given by the formula:
0.5α e k c kf ct,eff 1  1/α e ρ  (5.5)
εr = (εsm - εcm) =
Es
 89

The modular ratio is represented by the symbol αe, the factor kc is the coefficient accounting for
the stress distribution within a section and k accounts for the non-uniform self-equilibrating
stresses. The parameter ρ describes the reinforcement to gross concrete cross-sectional area ratio,
whilst Es denotes the steel modulus of elasticity.

For the reliability analysis of the EN 1992 crack model the limit state function was first defined,
where wlim is the limiting crack width treated as a deterministic value in this analysis. Failure of
the performance function will occur where the calculated crack width either just meets or exceeds
the limiting crack width. The calculated crack width, calculated using equations 5.3 and 5.4 or
5.5 (for the appropriate restraint condition), was multiplied by the model uncertainty (θ). The
model uncertainty was treated as a random variable in this analysis with its mean value taken to
be 1.
g = wlim – θw (5.6)
Other random variables of the limit state function include the section thickness (h), the concrete
cover (c) and the tensile strength of the concrete (fct, eff).

5.2.2 Representative Liquid Retaining Structure

Two restraint conditions were considered: edge and end restraint. Under both restraint conditions,
the section thickness was 250 mm with a C30/37 specified strength for OPC concrete (T1 = 15 ͦC,
for 340kg/m3 binder content), cast in the summer (giving a seasonal change in temperature of T 2
= 23 ͦC). Cover to reinforcement was 40 mm and steel formwork was selected for the analysis.
The shrinkage strain was determined to be 220με for the South African coastal regions with 80%
relative humidity. The steel reinforcement selected was 20 mm diameter high yield bars. Using
EN 1992, the autogenous shrinkage strain was determined to be 33με (at 28 days) and the
coefficient of thermal expansion of quartzite concrete had been found to amount to 14με. The
crack width limit was set at 0.2 mm for the reliability analysis. The choice of section modulus
and thus the modular ratio obtained through calculations was found to bear little significance to
the overall reliabilities calculated (about a 2% average increase in reliability indices was
calculated between the 3, 7 and 28-day elastic moduli). For these analyses, the section modulus
for concrete at 3 days was selected since that is the critical time in which cracking most likely
occurs.

5.2.3 Probabilistic Theoretical Models of Basic Variables

The basic variables entered into the limit state function are stochastic in nature. The various
statistical distributions to which each variable belonged and moreover what statistical parameter
best represented they have been collected. Basic variables found to have very little to no real
 90

variability were regarded as being deterministic. Those variables that were found to have a
random nature include the concrete cover (c), the concrete section thickness (h), the concrete
tensile strength (fct,eff) and the model uncertainty (θ). The basic variables were entered into the
limit state function in meters (m) and kilonewtons (kN). A summary of the basic variables
featured in the limit state function for the FORM analysis and their respective statistical
properties, as entered in Excel, are summarised in Figure 5.1:
 91

Basic Variables Symbols Dimimensions Distrbtribution Char. Value Mean CoV Std. Dev.
μx in kN and m σx in kN and m
Cross section thickness h mm Normal 250 0.25 0.01 0.0025
Geometry Cross section width b mm Deterministic 1000 1 0 0
Concrete cover c mm Gamma 40 0.04 0.15 0.006
Reinforcement diameter φ mm Deterministic 20 0.02 0 0
Concrete compressive strength (cube) fcu MPa Deterministic 37 37000 0 0
Concrete compressive strength (cylinder) fck MPa Deterministic 30 30000 0 0
Concrete tensile strength fct,eff MPa Lognormal 2 2900 0.19 551
Concrete elastic modulus Ec,eff GPa Deterministic 28 28000000 0 0
Modular ratio αe none Deterministic 7.14 7.14 0 0
Steel modulus Es GPa Deterministic 200 200000000 0 0
Material Autogenous Shrinkage Strain εca με Deterministic 33 3.30E-05 0 0
Drying shrinkage strain εcd με Deterministic 220 2.20E-04 0 0
2
Area of Steel As mm Deterministic 2513 0.002513 0 0
Coefficient (reinforcement) k1 none Deterministic 0.8 0.8 0 0
Coefficient (tension) k2 none Deterministic 1 1 0 0
Coefficient (for self-equilibrating stresses) k none Deterministic 1 1 0 0
Coefficient for stress distribution (pure tension) kc none Deterministic 1 1 0 0
Coefficients Coefficient of Thermal Expansion αc µɛ/˚C Deterministic 14 1.40E-05 0 0
Temperature T1 ˚C Deterministic 23 23 0 0
Temperature T2 ˚C Deterministic 15 15 0 0
Restraint Degree R none Deterministic 0.5 0.5 0 0
Limiting Crack Width wlim mm Deterministic 0.2 0.0002 0 0
Model Uncertainty θ none Lognormal 1 1 0.3 0.3

Figure 5.1: Summary of Statistical Parameters

 92

The FORM calculation was conducted following the logarithm (after A. H-S. Ang and W. H. Tang
(1984)):
1) Define the performance function (g(x) = wlim – θw).
2) Convert non-normal means and standard deviations of variates to normal equivalent.
3) The initial failure points are usually taken to be the mean values of the variates.
4) Determine the derivative of the performance function, g(x), with respect to each random
g
variables using MATLAB. The partial derivatives ( ) are then evaluated at the failure
X ' i
points (which were initially assumed to be the means of the random variables).
5) The direction cosines/sensitivity factors (αi*) may be subsequently obtained by dividing each
derivative by the root of the sum of the square of each derivative (i.e. each derivative is
normalised)
 g 
 
  X 
αi*=  i *
 g  2
i  
  X 
 i *
6) The failure point at the end of the iteration may then be determined through the equation
(represented initially as functions of the unknown 'β' value):
xi*=μXiN- αxi*σXiN*β
The normal equivalent of the mean and standard deviation of the random variables are used to
determine the failure points of the performance function.
7) This failure point is then substituted into the failure function, g(x) =0), and solved for β. Excel
solver is used to solve for β. Microsoft Excel solver is a function in the Excel programme that
finds the optimal value of a target cell by adjusting the values in variable cells which are used
to calculate the value in the target cell. The limit state function was entered into the target cell
whilst the reliability index was entered into the variable cell. The GRG linear is the solving
method selected. This method is meant for problems that have a smooth nonlinear nature.
The performance function is set into Excel solver as being the target (or objective) cell and
excel solver is set to solve for zero (0). The reliability index, β, is then entered in as the variable
cell in which that value of β that will make the performance function zero is determined by
excel solver. Thus a new set of failure points may be found using the now obtained β value.
The numerical values of these failure points- found after substituting β into the failure point
equation- are then used as the starting points in the next iteration.
8) Steps 2 through to 7 are repeated until convergence of β is reached. The failure points obtained
with the final reliability index, β, will then represent the final failure points of the performance
function.
 93

Figure 5.2 illustrates the steps taken in the FORM analysis of the EN 1992 crack model. The first
two iterations are shown in the figure. For this particular analysis, four iterations (Figure 5.3) were
enough to bring β into convergence. The influence of select parameters on the EN 1992 restrained
strain crack model was assessed in the FORM analysis. These were the parameters identified in the
parametric study as being particularly influential to the crack model.
 94

1st Iteration

convert non-normals to normal 2

lognormal: pdf ξ = c.o.v
f ξ λ μN σN λ= lnμ-0.5*(c.o.v)^2
Concrete cover c 0.15 -3.230125825 0.03955 0.006 μN = y*(1-lny*+λ)
N
Model Uncertainty θ 0.3 -0.045 0.955 0.3 σ = y*ξ

1st iteration logrithm *σ x*=μN-α*σN*β

partial direction failure failure
3 4 5 6 7
uncorrelated variables assumed derivatives cosine point equation point
h,c, fctm, θ Xi xi μNxi σNxi (δg/δXi) αxi σ*α
Cross section thickness h 0.25 0.25 0.0025 0 0 0.25+0*β 0.25 0
Concrete cover c 0.04 0.03955 0.006 -1.40813E-05 -0.394001578 0.03955+0.002364009*β 0.045398449 -0.002364009
Model Uncertainty θ 1 0.955 0.3 -3.28482E-05 -0.919109763 0.955+0.275732929*β 1.637150565 -0.275732929
Ʃα^2 1
substitute β 2.473953936

failure function: g(x)= wlim - θwk = 0.0002-th*wk 0.0002-(0.955+0.275732929*β)*(2*(0.03955+0.002364009*β)+0.25*0.8*1*phi/(As/(2.5*(0.03955+0.002364009*β+0.01)*b)))*(R*(alpha*(T1+T2)+ eca +ecd))

solve for β using β g(x)

excel solver 2.168419926 -6.87046E-11

2nd Iteration

convert non-normals to normal

lognormal: pdf ξ = c.o.v
N N
f ξ λ μ σ λ= lnμ-0.5*(c.o.v)^2
Concrete cover c 0.15 -3.230125825 0.039140341 0.006809767 μN = y*(1-lny*+λ)
Model Uncertainty θ 0.3 -0.045 0.756433516 0.49114517 σN = y*ξ

N N
2nd iteration logrithm *σ x*=μ -α*σ *β
partial direction failure failure
uncorrelated variables assumed derivatives cosine point equation point
N N
h,c, fctm, θ Xi xi μ xi σ xi (δg/δXi) αxi σ*α
Cross section thickness h 0.25 0.25 0.0025 0 0 0.25+0*β 0.25 0
Concrete cover c 0.045398449 0.039140341 0.006809767 -2.61644E-05 -0.399721592 0.039140341+0.002722011*β 0.044616338 -0.002722011
Model Uncertainty θ 1.637150565 0.756433516 0.49114517 -6E-05 -0.916636596 0.756433516+0.450201636*β 1.662125239 -0.450201636
Ʃα^2 1
β 2.011746849

failure function: g(x)= wlim - θwk = 0.0002-th*wk 0.0002-(0.756433516+0.450201636*β)*(2*(0.039140341+0.002722011*β)+0.25*0.8*1*0.02/(0.002513/(2.5*(0.039140341+0.002722011*β+0.01)*1)))*(0.5*(0.000014*(15+23)+ 0.000033+0.00022))

Figure 5.2: First and Second Iteration of FORM Analysis of EN 1992 Restrained Shrinkage Crack Model (Edge Restraint, hc,eff = 2.5(c + ϕ/2))
 95

β g(x)
1st iteration 2.47395 9.61306E-11
2nd iteration 2.01175 -3.11178E-11
3rd iteration 2.01099 -2.75106E-11
4th iteration 2.01099 -2.74877E-11

variable cell objective cell

iterations stopped after convergence was reached for β

Figure 5.3: Example of Convergence Achieved After Four Iterations (Edge Restraint, hc,eff
= 2.5(c + ϕ/2))

A separate spreadsheet was generated for the β values and the corresponding variation of the
above-mentioned parameters under which β was attained. The influence of the selected
parameters were assessed against a wide array of reinforcement. The steel reinforcement was
represented as a percentage amount against the gross cross-sectional area of the concrete member
(%As). The range selected for each analysis reached up to 3% of steel to concrete. This limit falls
just below the feasible limit for reinforcement in a 250 mm thick concrete section (the limit being
3.35% As). The feasible minimum bar spacing of 75 mm results in reinforcement to concrete
ratios that fall below the SAN 10100-1 maximum limit (As/Ac = 4%), thus making the bar spacing
the limiting criteria for the amount of reinforcement considered. A summary of the maximum
amount of reinforcing that may be implemented for other considered section thicknesses is given
in Table 5.1:
 96

Table 5.1: Feasible limit for reinforcement for select section thicknesses (minimum bar
spacing 75 mm)
h (mm) Max.
Feasible As
(%)
250 3.35
300 2.79
350 2.39
400 2.09
450 1.86
500 1.68

It must be reiterated that comparisons of reliability assessments with those of similar past research
findings are quite difficult in that unless the same limit state and statistical parameters (mean
value and standard deviation of the random variables) are used, direct comparisons would prove
to be somewhat inaccurate. Thus alternative measures of results verification may have to be
employed. As a measure of assurance that the FORM analysis was correctly executed, hand
calculations were conducted alongside those calculations done via Microsoft Excel (acting as a
double check of the results obtained).

Reliability models used in the analysis are:

a) Edge restraint with depth of effective tension zone taken to be 2.5(c +φ/2)
b) Edge restraint with depth of effective tension zone taken to be h/2
c) End restraint with depth of effective of tension zone taken to be 2.5(c +φ/2)
d) End restraint with depth effective of tension zone taken to be h/2

The majority of the reliability analysis was conducted with 2.5(c +φ/2) being the depth of effective
tension area as it is the limiting depth of effective tension area for most combinations of section
thicknesses, concrete covers and bar diameters. However there were certain instance where
models containing h/2 was more appropriate, these instances are mentioned where they apply.

5.3 Results and Discussion

5.3.1 Influence of cover and φ/ρp, eff
The effects of the concrete cover and bar diameter to effective steel content ratio were assessed
to gauge what impact either variable had on the reliability performance of the restrained shrinkage
 97

crack model of EN 1992. The concrete cover affects the formulation of the limit state function by
dictating the limiting effective tension depth. Thus the cover value was selected such that the
same limit state function was used in the reliability analysis- making direct comparisons of results
between concrete covers possible.

[Link] Edge Restraint

For a 250 mm thick section, the influence of concrete cover and the φ/ρ p, eff was examined by
varying the cover value against a selection of steel reinforcement ratios. Cover values of 50, 60
and 70 mm were used in this analysis, making h/2 the appropriate effective depth of tension area
for the analysis. The reliability of the crack model increased with a decrease in concrete cover (as
illustrated in Figure 5.4). For example for a steel ratio of say 2%, the reliability indices for 70, 60
and 50 mm are 1.48, 1.68, and 1.90 respectively. A decrease in cover would result in a decrease
in the crack spacing attained, thus decreasing the overall crack width obtained. Crack widths
beyond the crack width limit are then less likely to occur, hence the increase in reliability with
the decrease in concrete cover.

Moreover, irrespective of the cover value selected, the reliability index would increase with an
increase in the amount of reinforcement used. Undoubtedly, where more reinforcement is applied
to a concrete section the more resistant the member will become against tensile stresses, and so
less cracking occurs.

More reinforcement is required to meet the target reliability index as concrete cover increases.
The target reliability index (βt = 1.5) is met at steel to concrete ratios of about 1.69%, 1.85% and
2.02% for a cover value of 50 mm, 60 mm and 70 mm respectively. This amounts to about an
average increase in reinforcement of 10% with every 20% relative increase in covers selected (or
otherwise per 10 mm absolute increase in concrete cover value).
 98

Reliability Index (β) 2.5

βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
% As

c=50mm c=60mm c=70mm

Figure 5.4: Influence of Cover and φ/ρp, eff (Edge Restraint)

On comparing the reliability index directly against the φ/ρp,eff ratio (as shown in Figure 5.5) the
φ/ρp,eff ratios that ensure that the target reliability index is met for concrete covers 50, 60 and 70
mm are: 1.18, 1.08 and 0.99 m respectively (Figure 5.5). Since an increase in the amount of steel
reinforcement used will result in a decrease in the φ/ρp,eff ratio, it may be concluded that the
reliability increases with a decrease in the φ/ρp,eff ratio. Where the bar diameter had been found in
the deterministic analysis to have little influence on the crack width model, the influence of the
φ/ρp,eff ratio may be deduced to have come mostly from the steel reinforcing and partially from
the section thickness’s stochastic nature within the limiting effective tension depth equation. The
gradient of the graph of Figure 5.5 was close to -2 for the cover values considered indicating a
strong relationship.
 99

Reliability Index (β) 2.5

βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
ϕ/ρp,eff (m)

c=50mm c=60mm c=70mm

Figure 5.5: Influence of φ/ρp, eff Ratio on Reliability Index (Edge Restraint)

[Link] End Restraint

For end restraint conditions, with the same selection of concrete cover values, as with edge
restraint, it can clearly be deduced that a decrease in cover results in an increase in the reliability
index calculated (Figure 5.6). For instance, the reliability indices for end restraint at 2% As would
go from 1.17, 1.34, 1.52 for concrete values 70, 60 and 50 mm respectively (achieving lower
reliability indices than edge restraint for the same %As value).

A steel reinforcement to concrete percentage of 1.99%, 2.09% and 2.19% is required for the
selected concrete covers 50, 60 and 70 mm respectively to meet the target reliability index (βt =
1.5). This translates to about a 5% increase in reinforcement required per 20% relative increase
in the concrete cover- half the value found for edge restraint. This is indicative of the greater
influence had by concrete cover on the edge restraint crack model as compared to that of the end
restraint. Comparing this result to that of the edge restraint crack model, it is evident that slightly
more reinforcement is required to meet the target reliability index for end restraint. This makes
end restraint the more conservative of the two variations of the restrained shrinkage crack model.
 100

Reliability Index (β) 2.5

βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

c=50mm c=60mm c=70mm

Figure 5.6: Influence of Cover and φ/ρp, eff (End Restraint)

Much like for the edge restraint condition, the reliability index decreases as the φ/ρ p, eff ratio
increases. The target reliability index (βt = 1.5) is met where the φ/ρp, eff ratio is at 1, 0.96 and 0.92
m for covers 50, 60 and 70 mm respectively (Figure 5.7). The gradient of the φ/ρp, eff ratio to
reliability indices graphs across the concrete cover values selected was about -4, having a strong
impact on the reliability of the end restraint crack model (reading from Figure 5.7). This was a
stronger relationship than in the edge restraint case.
 101

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
ϕ/ρp,eff (m)

c=50mm c=60mm c=70mm

Figure 5.7: Influence of φ/ρp, eff Ratio on Reliability Index (End Restraint)

Clearly, the concrete cover selected has some bearing on the reliability of the crack model for
both end and edge restraint. For both restraint conditions, increasing the cover decreases the
reliability of the crack model. The amount of reinforcing also influences the reliability levels that
may be achieved by the crack model and thus the contributions of the φ/ρp, eff ratio cannot be
ignored. This is evident in the figures representing the change in reliability indices with respect
to the φ/ρp, eff ratio for both end and edge restraint (Figures 5.5 and 5.7, respectively). So therefore,
as the EN 1992 crack model stands, both terms (c and the φ/ρp, eff ratio) of the crack spacing have
a noteworthy influence on the eventual reliability of the crack model.

5.3.2 Influence of Effective Tension Area

An assessment was done on what influence the depth of the effective tension area had on the
reliability of the crack model. For a combination of a 40 mm concrete cover, 250 mm wide
concrete section and 20 mm reinforcing bar diameter, the depth of the effective tension area hc,eff
may either be taken as h/2 or 2.5(c +φ/2) since both returned the same value (hc,eff = 125 mm in
either case). Hence, the crack model for edge restraint containing both variations of the effective
depth could be compared directly.
 102

[Link] Edge Restraint

The edge restraint model with hc,eff = 2.5(c +φ/2) resulted in a lower reliability than where hc,eff =
h/2. The variation is slight, with the greatest absolute difference in reliability being about 0.19 for
a reinforcement percentage of 3% (the limit for feasible reinforcement to be used for a section
thickness of 250 mm is 3.35% for reinforcement applied on both faces of the section). This
difference in reliability indices translates to a difference in failure probability of p f = 0.11%
(Figure 5.8). However slight the difference between the effective depths might be, the result is
contrary to what one would expect. Considering that for most combinations of cover, section
thickness and bar diameter, the more limiting effective depth was found to be 2.5(c + φ/2); this
limiting effective depth generally predicts lower crack widths than crack width models containing
an effective depth of h/2. It may naturally be expected that the generally more conservative model
(where hc,eff = h/2) would produce lower reliability indices since the most conservative variation
of the crack model predicts wider crack widths that are more likely to exceed the stipulated crack
width limit.

Reinforcement to gross concrete ratios of about 1.55% and 1.62% for h/2 and hc,eff = 2.5(c +φ/2)
respectively mark the ratios at which the target reliability index of 1.5 was met (Figure 5.8). That
2
would mean a difference of about 175 mm (or 85 mm 2/section face per m of wall) in
reinforcement is required to meet the target reliability index which is a relatively small margin.
 103

3.5

2.5
Reliability Index (β)

2
βt = 1.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

h/2 2.5(c+ 0.5*φ)

Figure 5.8 Influence of Effective Tension Area (Edge Restraint)

[Link] End Restraint

For end restraint, the variation in reliability for the different effective depth formulations follows
after the edge restraint condition in which hc,eff =2.5(c +φ/2) generates smaller reliability indices
than the reliability indices achieved for where hc,eff = h/2. The biggest difference in reliability
index observed was 0.16, for 3% reinforcement to concrete. This result is close to the difference
found in the case of edge restraint (namely 0.19). This amounts to a difference in failure
probability of pf = 0.024% (as may be seen in Figure 5.9). Again, however slight, this is contrary
to what may be expected, considering that for most cases 2.5(c + φ/2) is the limiting effective
depth.

The target reliability index is met at 1.92% and 1.94% for %As for the crack models containing
an effective depth of h/2 and 2.5(c +φ/2) respectively (Figure 5.9). This corresponds to a
reinforcement amount of 4800 mm 2 and 4850 mm 2 (a 50 mm 2 difference in reinforcement). This
is a smaller difference than that found in the edge restraint crack model, and is negligible.
 104

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
% As

h/2 2.5(c+0.5*φ)

Figure 5.9: Influence of Effective Tension Area (End Restraint)

Clearly, the choice of effective depth of tension area affects the overall reliability of the crack
model for both restraint conditions– although this effect was slight for both restraint conditions.
During the course of the analysis, where the stochastic nature of the variables had been accounted
for, the random variables (cover in particular) increased in magnitude. Meaning that essentially,
the concrete cover value at the limit state/ failure point is greater than the initial cover value (40
mm). This increase in value would subsequently lead to an effective depth value that was greater
than what would normally be prescribed for the hc,eff = 2.5(c + φ/2) variation of the crack model
in a deterministic analysis. Thus, a comparably larger than usual effective tension area would be
in effect. This would then result in an increase in the likelihood of the limit state being exceeded,
and hence a decrease in the reliability indices obtained. A result that is somewhat counterintuitive
and could be better explained through a closer examination of the limit state function. This was
done by performing a sensitivity analysis of the EN 1992 restrained shrinkage crack model and
the results of which are reported in chapter 6.

5.3.3 Influence of Section Thickness

For a cover of 40 mm and reinforcement bar diameter of 20 mm, 2.5(c+ φ/2) is the limiting
effective area for section thicknesses considered in the analysis (250 mm to 500 mm). However,
section thickness (h) does not feature in the edge restraint crack model where hc,eff = 2.5(c+ φ/2).
 105

Covers beyond 50 mm make h/2 the limiting effective tension depth for a 250 mm thick section,
whilst bar diameter contributes little to the effective depth (cover is more influential, as was
uncovered in the parametric study). Therefore a small selection of section thicknesses were
analysed; namely, 250, 300 and 350 mm for the edge restraint case whilst for end restraint section
thicknesses analysed varied from 250 mm to 500mm since section thickness featured in the end
restraint case.

[Link] Edge Restraint

As the section thickness was increased, so did the reliability index (Figure 5.10). The only real
difference that section thickness made in the edge restraint case (where hc,eff = 2.5(c+ φ/2)) was
that thicker sections require more reinforcement to meet a particular steel to gross concrete cross-
sectional area ratio. For instance for a steel to gross cross-sectional concrete area of say 2% a 250
mm section would require 2500 mm 2/sectional face, while 3000 mm 2 and 3500 mm2/sectional
face is required for 300 and 350 mm respectively. The increase in reinforcement area results in
an increase in the effective steel content (ρp,eff) and subsequently a decrease in crack spacing and
crack widths obtained (since the model contains the reciprocal of the effective steel content ratio).
Thus it appears as though the reliability index of the edge restraint crack model increases with an
increase in the section thickness.

For the range of section thicknesses analysed, the amount of area required to maintain the target
reliability index (βt = 1.5) varied slightly, going from 1.62 %As (giving a steel reinforcing area of
4050 mm 2 for a 250 mm thick section), 1.36 %As (4080 mm 2 for a 300 mm section thickness
and 1.17%As (4095 mm 2 for a 350 mm thick concrete section) – an average decrease of 18%
with each 50 mm increase in section thickness. All in all, the steel content is believed to have
had the most impact of the reliability performance of the edge restraint crack model rather than
the inherent variabilities of the random variables within the model. Thus the physical model rather
than the reliability model influenced the outcome of this particular reliability assessment (where
section thickness was varied).
 106

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

h=250mm h=300mm h=350mm

Figure 5.10: Influence of Section Thickness (Edge Restraint)

[Link] End Restraint

It can be clearly seen that an increase in section thickness results in an increase in the reliability
index (Figure 5.11), as was found in the edge restraint case. For reinforcement ratio of say 2%,
the reliability index is 1.64, 1.97, 2.3, 2.65, 2.93 and 3.19 for section thicknesses 250, 300, 350,
400, 450 and 500 mm sequentially. Comparing the reliability indices obtained for h = 250, 300
and 350 mm at 2% As in the end restraint crack model to those of edge restraint it is observed that
the edge restraint crack model produced higher reliability indices (2, 2.38 to 2.69 corresponding
to h = 250, 300 and 350mm respectively for the edge restraint case as compared to the 1.64, 1.97
and 2.33 of the end restraint case for h = 250, 300 and 350 mm).

A few factors come into effect in this result. As the section thickness varies, so did the k
coefficient- this coefficient accounts for the presence of a non-linear stress distribution (varying
between 1 for section thicknesses less than 300 mm and 0.65 for section thicknesses greater than
800 mm, values between these limits being interpolated). Table 5.2 gives those k values used and
their corresponding section thickness. Being directly proportional to the restrained strain, the
decrease in this coefficient with the increase in section thickness resulted in a decrease in the
restrained strain. A decrease in the restrained strain means that a smaller crack width is attained,
 107

resulting in a decrease in the likelihood of the crack width limit being exceeded (increasing the
model’s reliability).

Table 5.2: Change of k Coefficient with Increasing Section Thickness (by interpolation)
h (mm) k
250 1
300 1
350 0.965
400 0.93
450 0.895
500 0.86

Also, much like in the case for edge restraint, the increase in amount of reinforcement area
required to meet particular ratio of reinforcing steel to gross concrete cross-sectional area may
have also influenced the results. Once again, larger section thicknesses require larger amounts of
reinforcement to meet a certain steel to concrete ratio. This then decreases the likelihood of crack
limit exceedance and increases the reliability performance of the crack model. The amount of
reinforcement required for the reliability index to be met ranged from 1.94%As (giving a
reinforcing steel area of 4859 mm 2) to 1.33%As (6655 mm 2) for the range of section thickness
from 250 mm to 500 mm considered in this analysis.

To directly compare the results of the end restraint crack model with that of the edge restraint
crack model, the amount of reinforcement required to meet the target reliability index for section
thicknesses from 250 to 350 mm were examined. At βt = 1.5, a reinforcement to concrete
percentage of 1.94, 1.80 and 1.65% are required for section thicknesses 250, 300 and 350 mm
respectively (larger percentages of reinforcement are required here than the 1.62, 1.36 and 1.17%
As respectively found for edge restraint). An average decrease of about 8% was observed with
every 50 mm increase in section thickness, almost half that experienced in the edge restraint case
(Figure 5.11). For the remainder section thicknesses of 400, 450 and 500 mm the corresponding
percentage of reinforcement required to meet the reliability index are 1.5, 1.41 and 1.33%A s
respectively (Figure 5.11). None of the reinforcement requirements for the section thicknesses
analysed exceeded the maximum feasible limit for reinforcement at the target reliability index.
 108

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

h=250mm h=300mm h=350mm h=400mm h=450mm h=500mm

Figure 5.11: Influence of Section Thickness (End Restraint)

Evidently, section thickness does influence the reliability of the crack model, more so in the end
restraint crack model than for the edge restraint crack model in that more reinforcing is required
to meet the target reliability index. Although, the effect held by section thickness in the edge
restraint case has more so to do with the increase in reinforcing required maintaining a particular
reinforcement ratio (thus being more a testament to the influence held by the steel reinforcing in
the edge restraint crack model) - the variable itself has no role in the edge restraint crack model.
Hence, the section thickness actually appearing in the end restraint crack model by default has
more of an effect on the end restraint model. A closer examination of the sensitivity factors
obtained for a predetermined reliability index should expose to what extent section thickness
influences the end restraint crack model.

5.3.4 Influence of Restraint Factor

The influence of the restraint factor was assessed by varying the restraint factor in the edge
restraint crack model when performing the FORM analysis. The restraint factors selected (ranging
from 0.1 to 0.5) are analysed against an increasing percentage of reinforcement to gross concrete-
cross sectional area.
 109

[Link] Edge Restraint

The restraint factor was found to have a considerable influence on the reliability of the EN
1992 crack model. The smaller the restraint factor applied to the model the greater the
reliability (as may be deduced from Figure 5.12). The increase in reliability index for the case
of 1% As, for instance, goes from 0.35 for restraint factor R = 0.5 to 1.037 (at R = 0.4), 1.92
(at R = 0.3), 3.16 (at R = 0.2) up to 5.28 for restraint factor of R = 0.1. The effects of restraint
factor 0.1 are not featured in Figure 5.12 due to the high reliability indices achieved (beyond
the considered range). At R = 0.1, the concrete member is virtually free to contract and thus
almost unlikely to crack. It’s the restriction in movement during the hydration process and
shrinkage of the concrete that results in the formation of cracks.

More reinforcement is required to maintain the target reliability as the restraint factor
increases. The target reliability index is met at 0.55, 0.88, 1.23, and 1.62 %A s for restraint
factors 0.2, 0.3, 0.4 and 0.5 sequentially (about a 40% increase in reinforcement required with
each 36 % relative increase in the restraint factor- equivalent to a 0.1 incremental/absolute
increase in the restraint factor). This has obvious financial implications. Thus essentially, the
choice of pour configuration and construction sequence significantly affects the reliability of
the crack model– resulting in sizeable increases in the reinforcement required to meet the
required reliability of the liquid retaining structure.

The clear influence of the restraint factor supports the need for increased knowledge of this
parameter. In this way, its influence on the crack model may be better quantified-particularly
where the restraint factor is treated as a random variable in reliability assessments rather than
a deterministic one. Further research on the restraint factor’s stochastic nature is then
necessary.
 110

4.5

3.5
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

R=0.1 R=0.2 R=0.3 R=0.4 R=0.5

Figure 5.12: Influence of Restraint Factor (Edge Restraint)

[Link] End Restraint

There is no restraint factor in the end restraint formula and thus no need to assess the
influence of the restraint factor in this case.

5.3.5 Influence of Model Uncertainty

The influence of model uncertainty was assessed by changing its coefficient of variance
(CoV) for each FORM analysis of the restrained strain crack model. The CoV’s assessed
ranged from 0.1 to 0.3. The section thickness and other variables of the crack model stood as
they did before (h = 250 mm, c = 40 mm, φ = 20 mm with coefficients and strains as
mentioned in the table of statistical parameters given in Figure 5.1).

[Link] Edge Restraint

A decrease in CoV value for the model uncertainty in edge restraint resulted in an increase in
the reliability index of the crack model. At 2%As the reliability index attained varied from β
= 2.0, 2.27, 2.63, 3.11 and 3.70 for CoV = 0.3, 0.25, 0.2, 0.15, 0.1 in turn. This marks a 17%
average increase in reliability with every 0.05 incremental increase in model uncertainty
variation.
 111

Hence, where the variation between predicted results and empirical results are minimised,
good construction practice is followed and other such quality control measures are employed
to reduce uncertainties, the results obtained from the existing model would be more reliable.
A smaller CoV for model uncertainty would result in a smaller crack width being estimated
by the reliability model for cracking and thus produce a crack width that is less likely to
surpass the crack width limit, hence the increase in the reliability index calculated. The target
reliability index of 1.5 is reached at the following percentages of steel to gross concrete cross-
sectional area for the CoV’s presented in the Figure 5.13: 1.27%, 1.33%, 1.41%, 1.49% and
1.62% for the corresponding coefficient of variances 0.1, 0.15, 0.2, 0.25 and 0.3 respectively.
The increases in the reinforcement required as model uncertainty increases are slight (an
average rise of factor 1.06 between increasing model uncertainties).

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
% As

CoV=0.1 CoV=0.15 CoV=0.2 CoV=0.25 CoV=0.3

Figure 5.13: Influence of Model Uncertainty (Edge Restraint)

[Link] End Restraint

Much like in the case of edge restraint, a decrease in model uncertainty resulted in an increase in
the reliability of the crack model (Figure 5.14). More precisely, the increase in reliability at say
2%As was 1.64, 1.78, 1.95, 2.13 and 2.31 for corresponding model uncertainties 0.3, 0.25, 0.2,
0.15 and 0.1 respectively. This gives an average increase in reliability of 9% for every 0.05
increment of model uncertainty variation, a smaller margin of change as compared to edge
 112

restraint for the same percentage of reinforcement to concrete of 2% As (an average increase of
17%). This indicates that variability in the model uncertainty has a larger effect on the edge
restraint crack model than it does on the end restraint crack model.

Again, as for the edge restraint case, the increase in the amount of reinforcement required results
from an increase in the CoV of the model uncertainty. Percentages at which the target reliability
index for liquid retaining structures was met for model uncertainty CoV’s of 0.1, 0.15, 0.2, 0.25,
and 0.3 are: 1.78%, 1.81%, 1.85%, 1.90% and 1.94% respectively (an average increase in
percentage area of factor 1.02, a smaller increase than for edge restraint condition which had an
increase factor of 1.06). It may thus be deduced that model uncertainty variations have a greater
influence on edge restraint crack model as compared to the end restraint case. It can also be
observed here that more reinforcement is required to reach the target reliability index for
corresponding CoV’s of model uncertainty of the end restraint as compared to the edge restraint
crack model- indicating, once again that the end restraint model is the most conservative model
of the two.

3.5

3
Reliability Index (β)

2.5

2
βt = 1.5
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
%As

CoV=0.1 CoV=0.15 CoV=0.2 CoV=0.25 CoV=0.3

Figure 5.14: Influence of Model Uncertainty (End Restraint)

Past research into the reliability of reinforced concrete structures have proven that model
uncertainty has a significant influence on the reliabilities of reinforced concrete structures
 113

(Holický, Retief, Wium, 2010).This is contrary to what was found in this analysis of the EN 1992
restrained strain crack model. However, a better estimate of what influence model uncertainty has
on the crack model may be obtained from a sensitivity analysis.

5.4 Summary

An investigation into the influence key identified parameters of the EN 1992 crack model had on
the overall reliability of the model was presented in this chapter. The following main findings of
the FORM analysis of the EN 1992 crack model for edge and end restraint are summarised below:
 Increasing the concrete cover and the φ/ρp,eff ratio decreased the model’s reliability.
 The effective depth of tension area bears some influence on the reliability of both the
edge and end restraint crack model, with hc,eff = h/2 proving to achieve higher reliability
indices than hc,eff = 2.5(c + φ/2). However, the difference in reliability between the
variations of effective depth is slim. This applies to both the edge and end restraint crack
model.
 Section thickness is not a parameter in the edge restraint crack model. An increase in
reliability resulting from an increase in section thickness was attributed to the increase in
the amount of reinforcement required to meet particular steel to gross concrete ratio for
thicker sections. Increased reinforcement used meant an increase in the reliability of the
model.
 Similarly, for end restraint, an increase in section thickness resulted in an increase in
reliability of the crack model– less so in the edge restraint case.
 The restraint factor had a significant effect on crack model reliability. Increases in
restraint factor decreased the reliability of edge restraint crack model. Thus further
research into this variable’s stochastic nature is necessary so that more can be known
about its impact on the reliability of the crack model.
 Increases in model uncertainty CoV resulted in a decrease in reliability for both the edge
and end restraint crack model.
 The end restraint model proved to be the more conservative of the two models.

A sensitivity analysis of the reliability crack model for edge and end restraint was then performed
to give a clearer perspective of the relative influence held by each random variable. In this way,
greater insight would be gained into the influence held by each random variable on the overall
reliability of the crack model. The relative influence, or otherwise sensitivity, of each random
variable may be measured through an extended reliability analysis. Those sensitivity factors
 114

associated with the target reliability index (βt = 1.5) are found by way of a reverse FORM analysis
(as reported in chapter 6).
 115

Chapter 6: Sensitivity Analysis of EN 1992 Crack Model:

Methodology, Results and Discussion

By determining the relative influence held by each random variable for the target reliability index
for the irreversible limit state, the opportunity of determining which variables most affect the
reliability of the crack model may be realised. The sensitivity factors give the relative influence
of random variables for a given reliability index (βt) using the reverse of the FORM analysis.
Additionally, theoretical partial safety factors, which indicated the adjustments necessary for
reliability compliance were also obtained from the reverse FORM analysis. This analysis serves
the main objective of, not only further understanding the reliability performance of the EN 1992
crack model as applied to liquid retaining structures in South Africa, but also presents the
opportunity of finding the ways in which to improve this crack model for use in the local
environment. This sensitivity analysis was conducted for the random variables of the cracking
serviceability limit state (namely, the effective concrete cover, section thickness, concrete tensile
strength and model uncertainty) against variations in the model uncertainty. Model uncertainty
had been determined in previous research to have a significant influence on the reliability of
reinforced concrete structures (Holicky, Retief and Wium, 2010). And since the model
uncertainty for cracking is not really known, its influence on the EN 1992, restrained cracking
serviceability limit state needs to be assessed. This was previously done for the load-induced
cracking case (Mcleod, 2013) and so the same will be explored for the restrained shrinkage case.

The influence held by the choice of target reliability index was investigated in this chapter.
Considering that the exceedance of the crack limit may result in the loss of structural integrity in
the case of liquid retaining structures with a potentially large consequence of loss, an investigation
into the influence of the reliability index on the basic variables of the crack model was conducted.
The need for this investigation also comes after understanding the increased importance of the
serviceability limit state as compared to the ultimate limit state in the design of liquid retaining
structures. Previous investigations on the matter that the cracking serviceability limit state was
the more demanding limit state, requiring more reinforcing to satisfy its design criteria.

Insights gained through this analysis may be used towards the calibration of the cracking
serviceability limit state for a larger scope of liquid retaining structure configurations and
expected uses for local conditions.
 116

6.1 Methodology of the Reverse FORM Analysis:

6.1.1 Reliability Analysis Formation
The reverse FORM analysis follows much of the same methodology as for the conventional
FORM analysis of chapter 5. The major difference in this instance is that the reliability index is
selected from the onset and the steel reinforcement (the unknown parameter) required to meet this
target reliability index is then calculated. The statistical parameters of the basic variables are as
for the conventional FORM calculation presented in chapter 5. The limit state function remains
as for the FORM analysis in the previous section and the reverse FORM algorithm used is as
follows:
1) The limit state function is first defined (as defined in equation 5.6)
2) Convert the mean and standard deviation of non-normal variates to their normal
equivalent.
3) Assume initial failure points- normally taken as the mean values of the random variables
in question.
4) Determine the partial derivatives of the performance function with respect to each random
variable using MATLAB. Evaluate these derivatives at the failure points. Then substitute
the previously determined value for the area of steel required to satisfy the performance
function (g(x) = wlim - θw) when evaluating the derivatives.
5) Determine the direction cosines (sensitivity factors) by dividing each partial derivative
by the root of the sum of the squared partial derivatives (in other words, normalise the
partial derivatives).
6) The failure points may then be determined by substituting the target reliability index in
the failure point equation (the normal equivalent of the mean and the standard deviation
are used in this equation).
7) The new found failure points are then substituted into the performance function and the
amount of area required for the performance function to equal zero is calculated using
excel solver. Here the area of steel required is set as the variable cell, whilst the
performance function is the objective cell set to meet an objective value of zero in excel
solver. The area of steel determined is then substituted into the partial derivatives of the
next iteration where the partial derivatives are also evaluated at the failure points
determined after the current iteration.
8) Steps 2 through to 7 are repeated until convergence of the required area of steel is reached.

Figure 6.1 illustrates the use of the reverse FORM algorithm for the EN 1992 restrained
shrinkage crack model,
 117

1st Iteration

convert non-normals to normal 2

lognormal: pdf ξ = c.o.v
f ξ λ μN σN λ= lnμ-0.5*(c.o.v)^2
Concrete cover c 0.15 -3.230125825 0.03955 0.006 μN = y*(1-lny*+λ) These new failure
points go into failure
Model Uncertainty θ 0.3 -0.045 0.955 0.3 σN = y*ξ
equation to find 'new'
value for 'A'

1st iteration logrithm *σ x*=μN-α*σN*β

3 partial 4 direction failure failure
5 6 7
uncorrelated variables assumed derivatives cosine point equation point
N N
h,c, fctm, θ Xi xi μ xi σ xi (δg/δXi) αxi σ*α
Cross section thickness h 0.25 0.25 0.0025 0 0 0.25+0*β 0.25 0
Concrete cover c 0.04 0.03955 0.006 -2.4942E-05 -0.38385468 0.03955+0.002303128*β 0.043004692 -0.002303128
Model Uncertainty θ 1 0.955 0.3 -6E-05 -0.923393516 0.955+0.277018055*β 1.370527083 -0.277018055
Ʃα^2 1
Target reliability index β
β 1.5

failure function: g(x)= wlim - θwk = 0.0002-th*wk 0.0002-(0.955+0.277018055*β)*(2*(0.03955+0.002303128*β)+0.25*0.8*1*0.02/(As/(2.5*(0.03955+0.002303128*β+0.01)*b)))*(R*(alpha*(T1+T2)+ eca +ecd))

Substitute into next

iteration (becomes new
'A' value in derivatives)

after 1st iteration Solve using excel solver A g(x)

0.001854706 0.0002-(0.955+0.277018055*1.5)*(2*(0.03955+0.002303128*1.5)+0.25*0.8*1*0.02/(As/(2.5*(0.03955+0.002303128*1.5+0.01)*1)))*(0.5*(0.000014*(15+23)+ 0.000033 +0.000220))

2nd Iteration

convert non-normals to normal

lognormal: pdf ξ = c.o.v
f ξ λ μN σN λ= lnμ-0.5*(c.o.v)^2
Concrete cover c 0.15 -3.230125825 0.039406069 0.006450704 μN = y*(1-lny*+λ)
Model Uncertainty θ 0.3 -0.045 0.876869535 0.411158125 σN = y*ξ

N N
2nd iteration logrithm *σ x*=μ -α*σ *β
direction failure failure
uncorrelated variables assumed cosine point equation point
h,c, fctm, θ Xi xi μNxi σNxi (δg/δXi) αxi σ*α
Cross section thickness h 0.25 0.25 0.0025 0 0 0.25+0*β 0.25 0
Concrete cover c 0.043004692 0.039406069 0.006450704 -2.56495E-05 -0.393079788 0.039406069+0.002535641*β 0.043209531 -0.002535641
Model Uncertainty θ 1.370527083 0.876869535 0.411158125 -6E-05 -0.919504366 0.876869535+0.378061691*β 1.443962072 -0.378061691
Ʃα^2 1
β 1.5

failure function: g(x)= wlim - θwk = 0.0002-th*wk 0.0002-(0.876869535+0.378061691*β)*(2*(0.039406069+0.002535641*β)+0.25*0.8*1*0.02/(As/(2.5*(0.039406069+0.002535641*β+0.01)*b)))*(R*(alpha*(T1+T2)+ eca +ecd))

A g(x)
after 2nd iteration 0.001996852 0.0002-(0.876869535+0.378061691*1.5)*(2*(0.039406069+0.002535641*1.5)+0.25*0.8*1*0.02/(As/(2.5*(0.039406069+0.002535641*1.5+0.01)*1)))*(0.5*(0.000014*(15+23)+ 0.000033 +0.00022))

Figure 6.1: First and Second Iteration of the Reverse FORM Analysis of EN 1992 Crack Model
 118

A total of four iterations were required for the convergence of β (as shown in Figure 6.2). Again,
a separate spreadsheet is generated for select data obtained after each analysis. Data of particular
interest include the direction cosines (sensitivity factors) of each random variable achieved at the
end of each calculation. The sensitivity factors are indicative of the influence that each random
variable has on the crack model relative to the other random variables. The closer the sensitivity
factors are to the number one (+1 or -1) the more influential the variable. Being a normalised
factor, the sum of the square of the sensitivity factor of each random variable should add up to
one (Σ (αi*) 2 = 1). The sensitivity factors of each random variable were plotted against the
coefficient of variance (CoV) of the model uncertainty.

A (mm2) g(x)
from failure function 0.001164 -6.9133E-11
1st iteration 0.001855 2.0535E-12
2nd iteration 0.001997 1.2922E-10
3rd iteration 0.002001 3.1129E-11
4th iteration 0.002001 3.1208E-11

Variable Objective

iterations stopped after convergence is reached

Figure 6.2: Example of Convergence Achieved After Four Iterations (Edge Restraint, hc,eff
= 2.5(c + ϕ/2))

The theoretical partial safety factors (psf’s) for each random variable were calculated with the
eventual failure point and mean of the random variable.
𝑋𝑖 ∗ (6.1)
γ𝑖 = = 1– αi*βwXi
μX

The theoretical partial safety factors are indicative of the amounts of adjustment that are required
to be made to the input random variables in order for the limit state function to be satisfied and
for the given target reliability index to be met. Both the sensitivity factor and the theoretical partial
factors were assessed for the crack width limits corresponding to the tightness classes and
functions to which liquid retaining structures are designed. The intent of this being that the partial
factors and sensitivity factors obtained are meant to represent and work across all expected
performance applications.

The restraint crack models assessed were as for the standard FORM analysis, consisting of:
 119

a) Edge restraint with depth of effective tension zone taken to be 2.5(c +φ/2)
b) Edge restraint with depth of effective tension zone taken to be h/2
c) End restraint with depth of effective tension zone taken to be 2.5(c +φ/2)
d) End restraint with depth of effective tension zone taken to be h/2

6.2 Results and Discussion

6.2.1 Sensitivity Factors at Varying Model Uncertainty
The sensitivity factors (direction cosines) reveal the degree to which random variables are
influential to the model in question relative to each other. The sensitivity factors were determined
for the random variables concrete cover (αc), section thickness (αh), concrete tensile strength
(αfct,eff) and model uncertainty (αθ). The size, or otherwise strength, of the sensitivity factors are
regarded in terms of their absolute magnitude. The sensitivity factors ranged from -1 to 1 and the
closer the value was to  1 , the larger its influence. The sensitivity factors were compared against

changes in the coefficient of variance of the model uncertainty varying from 0.1 to 0.3. The
reliability index was set at 1.5. The crack models containing both variations of the effective depth
of the tension zone (hc, eff = 2.5(c +φ/2) and h/2) were compared directly, since the combination
of concrete cover at 40 mm, a 250 mm section thickness and 20 mm bar diameter meant both
effective depth formulations were equally limiting.

[Link] Edge Restraint (hc, eff = 2.5(c +φ/2))

Table 6.1 presents a comparison of the sensitivity factors obtained for the respective random
variables where the effective depth was 2.5(c + ϕ/2). Across all crack width limits assessed, model
uncertainty proved to be the most influential variable of the edge restraint crack model with its
sensitivity factors going up to about -0.92 for all crack limits considered at model uncertainty
CoV = 0.3. Concrete cover follows after model uncertainty with its influence on the crack model
being about half that of model uncertainty at a model uncertainty CoV of 0.3 for all crack width
limits considered in the analysis. Section thickness clearly had no influence on the crack model
since this random variable did not feature in the model in question, since the effective depth
formulation was 2.5(c +φ/2).
 120

Table 6.1: Sensitivity Factors of Random Variables for Edge Restraint Crack Model (βt =
1.5, hc, eff = 2.5(c +φ/2))
wlim Model %As required αc αθ
(mm) uncertainty CoV (concrete (model
cover) uncertainty)

0.3 0.1 0.756 -0.786 -0.618

0.15 0.797 -0.646 -0.764
0.2 0.848 -0.535 -0.845
0.25 0.902 -0.451 -0.892
0.3 0.963 -0.388 -0.922
0.2 0.1 1.229 -0.791 -0.612
0.15 1.302 -0.651 -0.759
0.2 1.390 -0.541 -0.841
0.25 1.491 -0.458 -0.889
0.3 1.600 -0.394 -0.919
0.1 0.1 3.294 -0.802 -0.597
0.15 3.553 -0.668 -0.745
0.2 3.884 -0.559 -0.829
0.25 4.280 -0.476 -0.880
0.3 4.743 -0.412 -0.911

Figure 6.3 illustrates how the concrete cover decreased in influence as the model uncertainty
variability and the crack width limit are increased. The influence of the model uncertainty in the
edge restraint crack model increased as the crack width limit and the variability of model
uncertainty increased (as indicated in Figure 6.4).
 121

0.8

Sensitivity Factor (|αc|) 0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.3: Edge Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2))

0.8
Sensitivity Factor (|αθ|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.4: Edge Restraint Sensitivity of Model Uncertainty (θ) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))

[Link] Edge Restraint (hc, eff = h/2)

Table 6.2 demonstrated the relative influence of the random variables in the edge restraint crack
model where hc, eff = h/2. Model uncertainty, once again, proved to be the most influential
 122

parameter with concrete cover and section thickness following after respectively. Model
uncertainty were generally found to be above about -0.86, whilst the sensitivity factors of section
thickness were found to be negligible, mostly being greater than about -0.02. Section thickness’s
negligible influence relates to its indirect influence on the restrained strain resulting from edge
restraint- thicker sections would mean that a greater differential in temperature within the concrete
may occur.

Table 6.2: Sensitivity Factors of Random Variables for Edge Restraint Crack Model (βt =
1.5, hc, eff = h/2)
wlim Model %As Required αh αc αθ
(mm) uncertainty CoV (section (concrete (model
thickness) cover) uncertainty)

0.3 0.1 0.690 -0.086 -0.183 -0.980

0.15 0.745 -0.057 -0.130 -0.990
0.2 0.804 -0.043 -0.104 -0.994
0.25 0.866 -0.034 -0.089 -0.996
0.3 0.930 -0.028 -0.079 -0.997
0.2 0.1 1.114 -0.078 -0.273 -0.959
0.15 1.210 -0.052 -0.196 -0.979
0.2 1.314 -0.039 -0.157 -0.987
0.25 1.424 -0.031 -0.134 -0.991
0.3 1.541 -0.025 -0.118 -0.993
0.1 0.1 2.935 -0.052 -0.508 -0.860
0.15 3.251 -0.036 -0.381 -0.924
0.2 3.617 -0.027 -0.309 -0.951
0.25 4.035 -0.021 -0.265 -0.964
0.3 4.511 -0.017 -0.236 -0.972

Figures 6.5 and 6.6 both represent graphically how the sensitivity factors of section thickness and
concrete cover decreased with an increase in model uncertainty CoV. A trend that was contrary
to that of the model uncertainty sensitivity factor which increased with an increase in the model
uncertainty CoV (as shown in Figure 6.7). For concrete cover, as the crack width limit was
increased, its influence decreased. In the case for section thickness, an increase in the crack width
limit meant an increase in its relative influence, although the overall influence of section thickness
was found to be negligible. Increasing the crack width limit increased the influence of model
uncertainty on the edge restraint crack model (as found in Figure 6.7).
 123

0.8

Sensitivity Factor (|αh) 0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.5: Edge Restraint Sensitivity of Section Thickness (h) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = h/2)

0.8
Sensitivity Factor (|αc|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.6: Edge Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc, eff = h/2)
 124

0.8

Sensitivity Factor (|αθ|) 0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.7: Edge Restraint Sensitivity of Model Uncertainty (θ) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = h/2)

Comparing the effective depth of tension (hc,eff =2.5(c +φ/2) and h/2) it may be observed that the
edge restraint model containing hc,eff =h/2 has random variables that held a greater influence on
the crack model in comparison to the same random variables acting in the crack model containing
hc,eff =2.5(c +φ/2). Considering a model uncertainty CoV of 0.3, where the effective depth was h/2
the relative influence had by model uncertainty at 0.3 mm and 0.2 mm was greater by factor 1.08
than where the effective depth was hc,eff = 2.5(c +φ/2);for crack width limit 0.1 mm this factor
decreased slightly to 1.07. At a model uncertainty CoV of 0.3, where the relative influence of
concrete cover was at its lowest for either edge restraint models, the concrete cover was about 5
times more influential at a crack width limit of 0.3 mm for the crack model where the effective
depth of tension zone was 2.5(c +φ/2). This factor decreases to about 3 at 0.2 mm and 1.75 for
wlim = 0.1 mm. Section thickness was only a variable in the edge restraint model where the
effective depth was h/2, even the small relative influence held by section thickness in this model
was obviously greater than the no influence had by section thickness where hc,eff = 2.5(c + φ/2).

It is suspected, then, that the relative influence held by the concrete cover where hc,eff = 2.5(c +φ/2)
resulted in the edge restraint crack model generating smaller reliability indices in the FORM
analysis, especially when considering that the relative influence of model uncertainty was
comparable for either variations of the edge restraint crack model. Considering the impact of the
concrete cover’s relative influence in either variation of edge restraint crack model, this variable
serves to increase the crack spacing and the eventual crack widths calculated (since the model is
 125

directly proportional to concrete cover). The reliability model of the edge restraint crack model
containing hc,eff = 2.5(c +φ/2) would then generate crack widths considerably greater than those
of the reliability crack model where hc,eff = h/2. This would mean that the edge restraint crack
model would produce reliability indices that were lower than where hc,eff =h/2 since the crack
width limit would more likely be exceeded (as observed where the influence of the effective depth
for edge restraint was assessed in chapter 5). Although, the dominance in reliability performance
where h/2 is limiting was notably slight.

An analysis of the sensitivity factors of the edge restraint revealed that model uncertainty was, in
fact, the most influential random variable and adjustments made with respect to this variable by
way of partial factors should make the most impact on the crack model for edge restraint.

[Link] End Restraint (hc, eff = 2.5(c +φ/2))

Unlike in the case for edge restraint, the end restraint crack model containing the effective depth
of 2.5(c + φ/2) has section thickness included in the model. However, the influence of section
thickness was found to be negligible with the highest obtained value being -0.036 (wlim = 0.3 mm
and CoV = 1). Model uncertainty still remains the most influential random variable. Model
uncertainty sensitivity factors values were up to about -0.79 across all crack width limits
considered in the analysis at a model uncertainty CoV of 0.3. The effective concrete tensile
strength was the second most influential random variable followed by concrete cover and then
finally section thickness. The above mentioned observed trends were evident for model
uncertainty CoV’s of 0.2 and greater. At lower CoV’s for model uncertainty the effective concrete
tensile strength was the most influential random variable with section thickness being the least
influential. These results are tabulated in Table 6.3.
 126

Table 6.3: Sensitivity Factors of Random Variables for End Restraint Crack Model (βt =
1.5, hc, eff = 2.5(c +φ/2))
wlim Model %As αh αc αfct,eff αθ
(mm) uncertainty CoV required (section (concrete (effective (model
thickness) cover) concrete uncertainty)
tensile
strength)
0.3 0.1 1.380 -0.036 -0.513 -0.759 -0.399
0.15 1.404 -0.033 -0.468 -0.693 -0.547
0.2 1.435 -0.030 -0.421 -0.624 -0.657
0.25 1.471 -0.027 -0.378 -0.560 -0.737
0.3 1.510 -0.024 -0.339 -0.503 -0.794
0.2 0.1 1.762 -0.035 -0.517 -0.757 -0.398
0.15 1.794 -0.032 -0.472 -0.692 -0.546
0.2 1.835 -0.029 -0.425 -0.623 -0.656
0.25 1.883 -0.026 -0.381 -0.559 -0.736
0.3 1.935 -0.023 -0.343 -0.503 -0.793
0.1 0.1 2.751 -0.033 -0.524 -0.753 -0.396
0.15 2.807 -0.030 -0.479 -0.688 -0.543
0.2 2.877 -0.027 -0.432 -0.621 -0.653
0.25 2.959 -0.024 -0.388 -0.557 -0.733
0.3 3.049 -0.022 -0.349 -0.501 -0.791

As crack width limit decreased so did the (negligible) influence of section thickness (as shown
in Figure 6.8), effective concrete tensile strength (referring to Figure 6.10) and model uncertainty
(Figure 6.11). The influence of concrete cover tended to increase with a decrease in the crack
width limit (Figure 6.9). As the coefficient of variance of model uncertainty increased, the relative
influence of section thickness, concrete cover and the effective concrete tensile strength would
decrease (as illustrated in Figures 6.8, 6.9 and 6.10 respectively), whilst model uncertainty
increased in influence (as observed in Figure 6.11).
 127

0.8

Sensitivity Factor (|αh|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.8: End Restraint Sensitivity of Section Thickness (h) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))

0.8
Sensitivity Factor (|αc|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.9: End Restraint Sensitivity of Concrete Cover (c) with Varying Model Uncertainty
Coefficient of Variance (hc, eff = 2.5(c +φ/2))
 128

Sensitivity Factor (|αfct,eff |)

0.8

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.10: End Restraint Sensitivity of the Effective Concrete Tensile Strength (fct,eff) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))

0.8

0.6
Sensitivity Factor (|αθ|)

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.11: End Restraint Sensitivity of Model Uncertainty (θ) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))
 129

[Link] End Restraint (hc, eff = h/2)

The most influential random variable was the model uncertainty where the effective depth was
h/2. Model uncertainty's sensitivity factors go up to about -0.83 across all crack width limits at
model uncertainty CoV of 0.3 (slightly larger in magnitude to those values observed where the
effective depth of the crack model was 2.5(c+ φ/2)). The effective concrete tensile strength
follows after model uncertainty with sensitivity factors of about -0.53 at the same model
uncertainty CoV of 0.3 for the crack width limits considered. Concrete cover follows after the
effective concrete tensile strength. Then lastly, section thickness was found to have the least
influence on the crack model. These results may be observed in Table 6.4.

Table 6.4: Sensitivity Factors of Random Variables for End Restraint Crack Model (βt =
1.5, hc, eff = h/2)
wlim Model %As αh αc αfct,eff αθ
uncertainty CoV Required (section (concrete (effective (model
thickness) cover) concrete uncertainty)
tensile
strength)
0.3 0.1 1.347 -0.078 -0.149 -0.872 -0.459
0.15 1.375 -0.069 -0.134 -0.776 -0.613
0.2 1.409 -0.061 -0.120 -0.683 -0.718
0.25 1.447 -0.053 -0.105 -0.601 -0.791
0.3 1.489 -0.047 -0.097 -0.532 -0.840
0.2 0.1 1.718 -0.074 -0.179 -0.868 -0.457
0.15 1.755 -0.066 -0.161 -0.773 -0.610
0.2 1.800 -0.058 -0.144 -0.680 -0.716
0.25 1.851 -0.051 -0.129 -0.599 -0.788
0.3 1.906 -0.045 -0.117 -0.531 -0.838
0.1 0.1 2.679 -0.067 -0.242 -0.856 -0.451
0.15 2.741 -0.059 -0.219 -0.764 -0.604
0.2 2.818 -0.052 -0.196 -0.674 -0.710
0.25 2.906 -0.046 -0.176 -0.595 -0.783
0.3 3.001 -0.040 -0.159 -0.528 -0.833

The graphical representation of the findings are presented in the Figures 6.12 to 6.15. As the crack
width limit was decreased, so did the sensitivity factors of section thickness (as shown in Figure
6.12), concrete tensile strength (referring to Figure 6.14) and model uncertainty (Figure 6.15).
Concrete cover increases in relative influence as the crack width limit decreases (Figure 6.13).
The sensitivity of section thickness, concrete cover and concrete tensile strength decreased as the
model uncertainty CoV was increased. Contrary to this, the sensitivity factor of model uncertainty
increased with an increase in model uncertainty variability (as illustrated in Figure 6.15).
 130

0.8
Sensitivity Factor (|αh|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.12: End Restraint Sensitivity of Section Thickness (h) for Varying Model
Uncertainty Coefficient of Variance (hc,eff = h/2)

0.8
Sensitivity Factor (|αc|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.13: End Restraint Sensitivity of Concrete Cover (c) for Varying Model Uncertainty
Coefficient of Variance (hc,eff = h/2)
 131

Sensitivity Factor (|αfct,eff|)

0.8

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.03mm wlim=0.2mm wlim=0.1mm

Figure 6.14: End Restraint Sensitivity of the Effective Concrete Tensile Strength (fct,eff) for
Varying Model Uncertainty Coefficient of Variance (hc,eff = h/2)

0.8
Sensitivity Factor (|αθ|)

0.6

0.4

0.2

0
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.15: End Restraint Sensitivity of Model Uncertainty (θ) for Varying Model
Uncertainty Coefficient of Variance (hc,eff = h/2)

At a model uncertainty variation of 0.3 and crack width limits 0.3, 0.2 and 0.1 the concrete cover
sensitivity factor for the hc,eff = 2.5(c +φ/2) were larger by factors 3.5, 2.9 and 2.3 respectively
as compared to the sensitivity factors of the end restraint crack model where the effective depth
 132

was h/2. The sensitivity factors obtained for section thickness where hc,eff = h/2 are greater in
magnitude than those obtained where hc,eff = 2.5(c +φ/2), but still negligible. The sensitivity
factors of concrete tensile strength are quite comparable between the two variations of the end
restraint crack model, with the hc,eff = h/2 containing end restraint crack model slightly larger in
magnitude (larger by factor 1.06) at wlim = 0.2 mm and model uncertainty CoV 0.3.

The biggest difference between the two variations of the end restraint crack model was in the
influence of the concrete cover variable. The larger magnitude of the concrete cover sensitivity
factor in the model containing hc,eff = 2.5(c +φ/2) will mean that larger crack widths are calculated
with this reliability model. Thus the reliability indices produced under this model will be lower
as compared to the end restraint crack model where the effective depth of tension was h/2.

6.2.2 Theoretical Partial Safety Factors

The theoretical partial safety factors (psf’s) of section thickness (γh), concrete cover to
reinforcement (γc), effective concrete tensile strength (γfct,eff) and model uncertainty (γθ) were
calculated. These partial factors were calculated with respect to changes in the variability of the
model uncertainty. Variations in degree of restraint applied to edge restraint members were also
considered by calculating the theoretical partial factors under a selection of restraint factors.

[Link] Edge Restraint (hc, eff = 2.5(c +φ/2))

Table 6.5 gives the compilation of the theoretical psf’s of the random variables calculated for the
edge restraint crack model with the target reliability index set at 1.5 and effective depth of tension
2.5(c + ϕ/2). Section thickness did not feature in the edge restraint crack model where hc,eff =2.5(c
+φ/2). Model uncertainty required the largest theoretical psf’s with values from  1.1. This
dominance occurs only from a model uncertainty CoV of 0.2. Theoretical psf’s for concrete cover
were less than those of model uncertainty with values from  1.08.
 133

Table 6.5: Theoretical Partial Factors of Random Variables for Edge Restraint Crack
Model (βt = 1.5, hc, eff = 2.5(c +φ/2))
wlim Model %As required γc γθ
(mm) uncertainty CoV (concrete (model
cover) uncertainty)
0.3 0.1 0.756 1.18 1.09
0.15 0.797 1.14 1.17
0.2 0.847 1.12 1.26
0.25 0.902 1.09 1.35
0.3 0.963 1.08 1.45
0.2 0.1 1.229 1.18 1.09
0.15 1.302 1.15 1.17
0.2 1.390 1.12 1.26
0.25 1.491 1.10 1.35
0.3 1.600 1.08 1.45
0.1 0.1 3.294 1.18 1.09
0.15 3.553 1.15 1.17
0.2 3.884 1.12 1.26
0.25 4.280 1.10 1.35
0.3 4.743 1.09 1.44

For concrete cover the theoretical psf’s were relatively constant irrespective of the model
uncertainty CoV and the crack width limit, Figure 6.16 illustrates this (ranging from
approximately 1.1, up to a maximum value of 1.2). A larger range of variations in the theoretical
psf’s of model uncertainty were found as the model uncertainty CoV increased (approximately
from 1.1 to 1.4, as shown in Figure 6.17). An increase in the crack width limit resulted in a
decrease in the influence of the concrete cover (as indicated in Figure 6.16) with an increase in
influence being found for model uncertainty (referring to Figure 6.17).
 134

1.5

1.4

Partial Factor (γc) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.16: Edge Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =2.5(c +φ/2))

1.5

1.4
Partial Factors (γθ)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.17: Edge Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =2.5(c +φ/2))

[Link] Edge Restraint (hc, eff = h/2)

Even with where the effective depth of tension area was h/2, no real adjustment was required for
section thickness to meet the desired reliability index. This was evident across all assessed crack
 135

width limits and model uncertainty CoV’s considered in the analysis. Model uncertainty required
the most adjustments to meet the target reliability index, with theoretical psf’s from about 1.1
(model uncertainty CoV = 0.1) to 1.5 (model uncertainty CoV = 0.3) for the crack width limits
considered. The concrete cover follows after model uncertainty with psf’s from 1.01 (wlim = 0.3
mm, model uncertainty CoV = 0.3). This was indicative of the slight influence the concrete cover
had on the edge restraint crack model where h/2 was the effective depth of tension area. Presented
in Table 6.6 are the theoretical psf’s obtained for the edge restraint crack model where hc,eff = h/2.

Table 6.6: Theoretical Partial Factors of Random Variables for Edge Restraint Crack
Model (βt = 1.5, hc, eff =h/2)
wlim Model %As Required γh γc γθ
(mm) uncertainty CoV (section (concrete (model
thickness) cover) uncertainty)

0.3 0.1 0.690 1.00 1.03 1.15

0.15 0.745 1.00 1.02 1.24
0.2 0.804 1.00 1.01 1.32
0.25 0.866 1.00 1.01 1.41
0.3 0.930 1.00 1.01 1.50
0.2 0.1 1.114 1.00 1.05 1.15
0.15 1.210 1.00 1.03 1.23
0.2 1.316 1.00 1.02 1.32
0.25 1.424 1.00 1.02 1.41
0.3 1.541 1.00 1.02 1.49
0.1 0.1 2.935 1.00 1.11 1.13
0.15 3.251 1.00 1.08 1.22
0.2 3.617 1.00 1.06 1.30
0.25 4.0354 1.00 1.05 1.39
0.3 4.511 1.00 1.04 1.48

Figure 6.18 illustrates how the theoretical psf’s required for section thickness were generally
unaffected by the increase in crack width limit. For concrete cover, increases in the crack width
resulted in a decrease in the theoretical psf required to attain the target reliability index (observing
from Figure 6.19). Model uncertainty psf’s increased with an increase in crack width limit
(referring to Figure 6.20). Increases in the model uncertainty CoV resulted in decreases in the
theoretical psf attained for concrete cover and an increase in those theoretical psf’s values
obtained for model uncertainty (reading from Figure 6.19 and 6.20 respectively). In the case of
section thickness, little variation was experienced across the range model uncertainty CoV’s
considered in this analysis (as illustrated in Figure 6.18).
 136

1.5

1.4

Partial Factor (γh) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.18: Edge Restraint Theoretical Partial Safety Factors of Section Thickness (h) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2)

1.5

1.4
Partial Factor (γc)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.19: Edge Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2)
 137

1.5

1.4

Partial Factor (γθ) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.20: Edge Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2)

The adjustments required for concrete cover to meet the target reliability were greater by a factor
of 1.07 where the effective depth of tension zone was 2.5(c +φ/2) –for wlim = 0.2 mm and model
uncertainty CoV of 0.3 (a small increase in the psf required between the effective depth hc,eff = h/2
to where hc,eff is 2.5(c +φ/2). The theoretical psf’s calculated for model uncertainty where the
effective depth is h/2 were 1.03 times greater than where the effective depth of the tension was
2.5(c +φ/2). This slight increase was found where the crack width limit was 0.2 mm and the
variability of model uncertainty was set at 0.3.

[Link] End Restraint (hc, eff = 2.5(c +φ/2))

It may be observed from Table 6.7 that the section thickness had obtained negligible theoretical
psf’s. Being the most influential random variable, model uncertainty had partial factors from
about 1.1 to 1.4for the range of crack width limits considered (for corresponding model
uncertainty CoV’s 0.1 and 0.3). This was followed by concrete cover with theoretical psf’s from
1.07 to 1.1 (at model uncertainty CoV of 0.3). The effective concrete tensile strength is a material
property and thus a resistance variable, the theoretical partial safety factor for concrete tensile
strength would be implemented in design codes as 1/γfct,eff to obtain the design value for this
variable . For the effective concrete tensile strength theoretical psf’s were generally from 1.1 at
model uncertainty CoV of 0.3 to about 1.2 for model uncertainty CoV of 0.1 – implemented as
0.91 to 0.83 respectively (referring to Table 6.7).
 138

Table 6.7: Theoretical Partial Factors of Random Variables for βt = 1.5 (hc, eff =2.5(c +φ/2))
wlim Model %As required γh γc γfct,eff 1/γfct,eff γθ
(mm) uncertainty CoV (section (concrete (effective (model
thickness) cover) concrete uncertainty)
tensile
strength)
0.3 0.1 1.380 1.00 1.11 1.22 0.82 1.06
0.15 1.404 1.00 1.10 1.20 0.84 1.12
0.2 1.435 1.00 1.09 1.17 0.85 1.19
0.25 1.471 1.00 1.08 1.15 0.87 1.28
0.3 1.510 1.00 1.07 1.13 0.88 1.37
0.2 0.1 1.762 1.00 1.11 1.22 0.82 1.06
0.15 1.794 1.00 1.01 1.20 0.84 1.12
0.2 1.835 1.00 1.09 1.17 0.85 1.19
0.25 1.883 1.00 1.08 1.15 0.87 1.28
0.3 1.935 1.00 1.07 1.13 0.88 1.37
0.1 0.1 2.751 1.00 1.11 1.22 0.82 1.06
0.15 2.807 1.00 1.10 1.20 0.84 1.12
0.2 2.877 1.00 1.09 1.17 0.85 1.19
0.25 2.959 1.00 1.08 1.15 0.87 1.28
0.3 3.049 1.00 1.07 1.13 0.88 1.37

The theoretical partial safety factors obtained for the random variables remained mostly steady
across all crack width limits considered as may be deduced from the Figures 6.21 to 6.24,
particularly for section thickness (as shown in Figure 6.21). Nonetheless, as the crack width limit
decreased the theoretical psf’s of the effective concrete tensile strength and model uncertainty
decreased (referring to Figures 6.23 and 6.24 respectively). The theoretical partial safety factors
of concrete cover increased with a decrease in crack width limit (reading from Figure 6.22).
Increases in the variability of the model uncertainty decreased the theoretical partial safety factors
required for section thickness (negligible decrease), concrete cover (decrease was also found to
be marginal, but not as small as for section thickness) and for the effective concrete tensile
strength for reliability compliance (as may be observed in Figures 6.21, 6.22 and 6.23
respectively). Model uncertainty’s theoretical partial safety factors increased considerably with
an increase in model uncertainty variability (Figure 6.24).
 139

1.5

1.4

Partial Factor (γh) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.21: End Restraint Theoretical Partial Safety Factors of Section Thickness (h) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))

1.5

1.4
Partial Factors (γc)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.22: End Restraint Partial Safety Factors of Concrete Cover (c) for Varying Model
Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))
 140

1.5

1.4

Partial Factor (γfct,eff) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.23: End Restraint Theoretical Partial Safety Factors of the Effective Concrete
Tensile Strength (fct,eff) for Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c
+φ/2))

1.5

1.4
Partial Factor (γθ)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.24: End Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff = 2.5(c +φ/2))

[Link] End Restraint (hc, eff = h/2)

Model uncertainty, once again, required the largest theoretical partial safety factor with values
from approximately 1.1 to 1.4 (at model uncertainty CoV’s of 0.1 and 0.3 respectively) for the
 141

crack width limits considered in this analysis. The theoretical psf’s for the effective concrete
tensile strength follows after model uncertainty with theoretical psf’s from about 1.1. As before
mentioned, when applying the calculated theoretical partial safety factor for the effective concrete
tensile in a design code the factor 1/γfct,eff will be used rather than γfct,eff. since it is a material
property (and thus a resistance variable). The theoretical psf’s obtained for concrete cover was
generally around 1.02. Section thickness had a small influence on the end restraint crack model
(where hc,eff = h/2) and hence obtained theoretical partial safety factors of about 1 for all crack
widths limits and model uncertainty CoV’s considered. These results may be observed in Table
6.8.

Table 6.8: Theoretical Partial Factors of Random Variables for βt = 1.5 (hc, eff = h/2)
wlim Model Uncertainty %As Required γh γc γfct,eff 1/γfct,eff γθ
(mm) CoV (section (concrete (effective (model
thickness) cover) concrete uncertainty
tensile
strength)
0.3 0.1 1.347 1.00 1.02 1.26 0.79 1.07
0.15 1.375 1.00 1.02 1.23 0.82 1.14
0.2 1.409 1.00 1.02 1.19 0.84 1.22
0.25 1.447 1.00 1.01 1.17 0.86 1.30
0.3 1.489 1.00 1.01 1.14 0.88 1.40
0.2 0.1 1.718 1.00 1.03 1.26 0.80 1.07
0.15 1.755 1.00 1.03 1.22 0.82 1.13
0.2 1.800 1.00 1.02 1.19 0.84 1.22
0.25 1.851 1.00 1.02 1.17 0.86 1.30
0.3 1.906 1.00 1.02 1.14 0.88 1.39
0.1 0.1 2.679 1.00 1.04 1.25 0.80 1.07
0.15 2.741 1.00 1.04 1.22 0.82 1.13
0.2 2.818 1.00 1.03 1.19 0.84 1.21
0.25 2.906 1.00 1.03 1.16 0.86 1.30
0.3 3.001 1.00 1.03 1.14 0.88 1.39

There were slight variations in the theoretical partial safety factors obtained across the crack width
limits considered for all random variables (particularly for section thickness). Increases in the
crack width limit meant an increases in the theoretical partial safety factors required for the
effective concrete tensile strength and model uncertainty (as shown in Figures 6.27 and 6.28
respectively). The concrete cover had theoretical partial safety factors that decreased in value as
 142

the crack width limit was increased (referring to Figure 6.26). Increases in the variability of the
model uncertainty resulted in there being a decrease in values of the theoretical partial safety
factors obtained for concrete cover and the effective concrete tensile strength (as shown in Figures
6.26 and 6.27). The theoretical partial factors obtained for model uncertainty increased as the
variability in the model uncertainty was increased (as observed in Figure 6.28).

1.5

1.4
Partial Factor (γh)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.25: End Restraint Theoretical Partial Safety Factors of Section Thickness (h) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =h/2)
 143

1.5

1.4

Partial Factors (γc) 1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.26: End Restraint Theoretical Partial Safety Factors of Concrete Cover (c) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =h/2)

1.5

1.4
Partial Factor (γfct,eff)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.27: End Restraint Theoretical Partial Safety Factors of the Effective Concrete
Tensile Strength (fct,eff) for Varying Model Uncertainty Coefficient of Variance (hc, eff = h/2)
 144

1.5

1.4

Partial Factors (ϒθ)

1.3

1.2

1.1

1
0.1 0.15 0.2 0.25 0.3
Model Uncertainty CoV

wlim=0.3mm wlim=0.2mm wlim=0.1mm

Figure 6.28: End Restraint Theoretical Partial Safety Factors of Model Uncertainty (θ) for
Varying Model Uncertainty Coefficient of Variance (hc, eff =h/2)

Comparing the theoretical psf’s obtained for the end restraint crack model where the effective
depth was h/2 to the end restraint crack model where hc,eff = 2.5(c +φ/2), it may be found that the
theoretical psf’s obtained for most variables were greater in value, but only slightly. For a 0.2
mm crack width limit and at a model uncertainty CoV of 0.3, factors of 1.02 and 1.01 were where
the model uncertainty and the effective concrete tensile strength’s respective theoretical partial
safety factors were greater in the case where the effective depth was hc,eff = h/2 as compared to
the end restraint crack model where hc,eff = 2.5(c +ϕ/2). Considering concrete cover, the
theoretical partial safety factors where hc,eff = 2.5(c +φ/2) was greater in magnitude by factor 1.06
than where the effective depth were h/2 (a larger difference in magnitude than those experienced
by model uncertainty and the effective concrete tensile strength at the same crack limit of 0.2 mm
and model uncertainty CoV of 0.3). Additionally, section thickness had theoretical psf’s
amounting to 1 in either variations of the end restraint crack model. Overall, the theoretical partial
safety factors obtained for the respective variables were quite comparable.

6.2.3 Potential Partial Factors for Code Calibration (Edge vs. End Restraint):
The theoretical implication (or otherwise practical application) of the above-mentioned
comparison may be that comparable psf’s obtained for the respective random variables indicate
that the same partial factor may be applied to those variables to obtain the desired reliability
irrespective of the restraint model (whether edge or end restraint). This is ideal for developing a
design code, where simplification is preferred. However, the theoretical partial factors obtained
 145

for the edge restraint crack model were found to be slightly greater than those obtained for the
end restraint model. This implies that if the same partial factors were to be applied to the
respective random variables for both the edge and end restraint crack model, these partial factors
would be slightly conservative for the end restraint crack model.

The findings of the comparison in theoretical psf’s obtained for the edge and end restraint crack
model are listed below. In this exercise the edge and end restraint crack model, the theoretical
partial safety factors that were obtained for where crack width limit was 0.2 mm and model
uncertainty CoV was 0.3 were considered (as for the representative liquid retaining structure
case).
 Model uncertainty:
 Where hc,eff = 2.5(c + φ/2), model uncertainty was comparable with (ratio of edge
to end partial factor was 1.06)
 Where hc,eff = h/2, model uncertainty was– once again– comparable. The ratio of
edge to end was 1.07.
 Effective concrete tensile strength:
 Only found in the end restraint model, so no comparison could be made between
the edge and end restraint crack model in this respect. However, based on the
findings of the end restraint for both where hc,eff was 2.5(c + ϕ/2) and h/2 ranged
from about 1.1 to 1.2, thus a value of 1.2 may then be recommended for the
effective concrete tensile strength for use in a design standard.
 Concrete cover:
 Where hc,eff = 2.5(c +φ/2), concrete cover also returned comparable theoretical
partial safety factors for edge and end restraint. A ratio of 1.01 was calculated for
edge to end restraint theoretical partial safety factors.
 Where hc,eff = h/2, the ratio of edge to end restraint theoretical partial factors was
1.
 Section thickness:
 Theoretical partial factor obtained were either 1 or very close to 1 across all the
restraint models assessed.
 146

6.2.4 Influence of the Choice of Reliability Index (β)

[Link] Edge Restraint
To determine the impact a change in the target reliability index on the EN 1992 edge restraint
crack model with hc,eff = 2.5(c +ϕ/2) and where hc,eff = h/2 (referring to Table 6.9 and Table 6.10
respectively), the reliability index was changed from 0.5 to 2 whilst maintaining the crack width
limit (wlim = 0.2 mm) and model uncertainty CoV (0.3). In the case where the effective depth of
the tension zone was given by hc,eff = 2.5(c + ϕ/2), the relative influence held by both concrete
cover and model uncertainty remained effectively constant through the changes in reliability
index. Both the theoretical psf’s for concrete cover and model uncertainty increased with an
increase in the reliability index. Regarding β = 0.5 as a base, theoretical partial safety factors for
concrete cover increased by 6% and 10% with an increase in reliability index to 1.5 and 2
respectively. And for model uncertainty, an increase in theoretical partial factors of 32% and
51% where corresponding reliability indices β = 1.5 and 2 were compared against selected base
reliability index 0.5. Clearly, a change in the reliability index has the most effect on the model
uncertainty of the EN 1992 edge restraint model. Section thickness, having no part in the edge
restraint model where hc,eff = 2.5(c + ϕ/2), has no relative influence with a sensitivity factor of 0
and theoretical partial safety factor of 1.

A 50% and 89% respective increase in reinforcement was obtained for reliability index 1.5 and 2
to be met with respect to reliability index 0.5– a substantial increase. A 25% increase in
reinforcement was required where the reliability index was increased from 1.5 to 2. Thus a change
in reliability index could have a considerable financial effect on the design of liquid retaining
structures with elements restrained along their edge.

Table 6.9: Influence of Reliability Index on the Basic Variables of the EN 1992 Edge
Restraint Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = 2.5(c + ϕ/2))
Sensitivity Factors Partial Factors
β %As required αh αc αθ γh γc γθ
0.5 1.060 0 -0.386 -0.922 1.000 1.018 1.098
1.5 1.600 0 -0.394 -0.919 1.000 1.081 1.446
2 2.000 0 -0.399 -0.917 1.000 1.115 1.657

The same exercise was extended to where the effective depth of tension zone, h c,eff,, was h/2
(results of which were presented in Table 6.10). The sensitivity factors for section thickness and
model uncertainty were slightly influenced by the change in reliability index. Concrete cover, on
the other hand, increased by 37% and 62% for reliability indices 1.5 and 1.2 when compared
against the sensitivity factor when the reliability index was set at 0.5. Section thickness and
 147

concrete cover obtained theoretical psf’s that varied only slightly as the reliability index was
increased. Model uncertainty’s theoretical partial safety factor increased by 35% and 56% for
reliability indices 1.5 and 2 respectively when compared against the theoretical partial factor
obtained for where the reliability index was set at 0.5. A comparable result to those obtained for
when the effective depth of the tension zone was 2.5(c + ϕ/2).

Reinforcements required to meet a reliability index of 1.5 and 2 as compared to those required for
β = 0.5 are 46% and 79% respectively. A considerable increase in reinforcement, which would
have a proportional impact on the cost of design where the reliability index is changed. An amount
of 23% more reinforcement was required where the reliability index was changed from 1.5 to 2.
These were overall smaller increases as compared to the results for the case where the effective
depth of tension was 2.5(c + ϕ/2).

Table 6.10: Influence of Reliability Index on the Basic Variables of the EN 1992 Edge
Restraint Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = h/2)
Sensitivity Factors Partial Factors
β %As required αh αc αθ γh γc γθ
0.5 1.054 -0.027 -0.086 -0.996 1.000 0.995 1.110
1.5 1.541 -0.025 -0.118 -0.993 1.000 1.015 1.495
2 1.889 -0.024 -0.139 -0.990 1.000 1.031 1.732

[Link] End Restraint

For end restraint, the sensitivity factors for all basic variables remained relatively constant
(referring to Table 6.11 where the effect depth of tension was 2.5(c + φ/2)). As the reliability
index was increased from 1.5 to 2 the theoretical partial safety factors of concrete cover increases
by 5% and 8% correspondingly. For the effective concrete tensile strength this increase was about
10% and 15% for reliability index 1.5 and 2 respectively as compared to the corresponding
theoretical partial safety factor for reliability index 0.5. Considering the theoretical partial safety
factors of model uncertainty, an increase of 27% and 43% was experienced for reliability indices
1.5 and 2 respectively as compared against the theoretical partial safety factor obtained where β
was 0.5. Again, a change in the reliability index had the largest effect on the model uncertainty.

Increases of 26% and 42% in steel reinforcement would be required to meet a reliability index of
1.5 and 2 as compared against β = 0.5. An increase of the reliability index from 1.5 to 2 resulted
in an increase in reinforcement of 13%. These increases were smaller than those required for
edge restraint, although there are still significant. Evidently, the target reliability index set for the
 148

EN 1992 cracking serviceability limit state has a considerable impact on the cost of the design of
liquid retaining structures.

Table 6.11: Influence of Reliability Index on the Basic Variables of the EN 1992 End
Restraint Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = 2.5(c + ϕ/2))
Sensitivity Factors Partial Factors
β %As required αh αc αfct,eff αθ γh γc γfct,eff 1/γfct,eff γθ
0.5 1.540 -0.024 -0.337 -0.504 -0.795 1.000 1.020 1.030 0.971 1.077
1.5 1.935 -0.023 -0.343 -0.503 -0.793 1.000 1.068 1.133 0.882 1.366
2 2.181 -0.023 -0.345 -0.502 -0.793 1.000 1.097 1.189 0.841 1.538

In the case where the effective depth of tension zone was h/2, the sensitivity factors for section
thickness, concrete tensile strength and model uncertainty were only slightly affected by the
change in reliability index (as evident in Table 6.12). A similar trend may be found for the
theoretical partial safety factors where a small variation was experienced as the reliability index
was increased. Concrete cover obtains sensitivity factors that increased in value by 19% and 31%
for reliability indices 1.5 and 2 correspondingly as compared to the sensitivity factor obtained for
a reliability index of 0.5. The same comparison being applied to the theoretical partial safety
factors (with β = 0.5 as the base) of concrete cover showed an increase in value of 2% and 3% for
β = 1.5 and 2 respectively. However, model uncertainty had increases in value of 29% and 47%
for reliability indices 1.5 and 2 respectively, where β = 0.5 was the base of comparison – a
comparable finding to where the hc,eff = 2.5(c + φ/2).

The demand in reinforcement increased by 24% and 39% for reliability indices 1.5 and 2 as
compared to that which was required for a 0.5 reliability index. An increase in reliability index
from 1.5 to 2 results in a 12% steel reinforcement. Once again, a comparable result to where hc,eff
= 2.5(c + φ/2).

Table 6.12: Influence of Reliability Index on the Basic Variables of the EN 1992 End
Restraint Crack Model (wlim = 0.2 mm, model uncertainty CoV = 0.3, hc,eff = h/2)
Sensitivity Factors Partial Factors
β %As αh αc αfct,eff αθ γh γc γfct,eff 1/γfct,eff γθ
required
0.5 1.532 -0.047 -0.098 -0.532 -0.840 1.000 0.996 1.033 0.968 1.077
1.5 1.906 -0.045 -0.117 -0.531 -0.838 1.001 1.015 1.143 0.875 1.394
2 2.135 -0.044 -0.128 -0.530 -0.837 1.001 1.027 1.201 0.832 1.580
 149

Overall, a large increase in reinforcement was observed where the reliability index was increased
from 0.5 to 2. This was particularly evident for the edge restraint case. Results for where the
effective depth of the tension zone was either 2.5(c + ϕ/2) or h/2 were comparable for both the
edge and end restraint case. The above-mentioned observations were much greater than the EN
1992 load-induced cracking case (Retief, 2015) in which the amount of tension steel increased by
10% and 15% for β = 1.5 and 2 respectively (where β = 0.5 was set as a default value). Clearly, a
change in the choice of reliability index of the EN 1992 restrained shrinkage crack model may be
deduced to have a considerable effect on the cost of design. However, these increases in cost may
be minor when compared to those required for structural failure where this serviceability limit
state is not met. Further research into the cost of failure for the serviceability limit state is required;
this should give clearer insight into what the target reliability index should be for liquid retaining
structures.

6.3 Comparison of Results for Deterministic and Probabilistic Analysis

Comparisons of the reinforcement required to meet the considered crack width limits for the
deterministic analysis and those obtained from the probabilistic analysis (for βt = 1.5) were made.
The crack width limits was varied (considering only crack width limits 0.3, 0.2 and 0.1 mm) and
the section thickness was kept constant at 250 mm. The effective depth of tension zone, hc,eff, was
2.5(c + ϕ/2) for a cover of 40 mm and reinforcing bar diameter of 20 mm (this was also the
variation of effective depth that was found to be limiting for most combinations of section
thickness, concrete cover and reinforcing bar diameters). It may be observed that analysing the
crack model deterministically (ignoring the stochastic nature of the input variables) would result
in greater amounts of reinforcement being required for the crack width limit to be met. This was
evident for both the 0.2 and 0.3 mm crack width limits considered in this analysis (referring to
Table 6.13). The dominance held by the reinforcement requirements of the deterministic analysis
in the case of the edge restraint condition was about 15% more than that of the probabilistic case
and for end restraint there was a 2% increase in demand of steel reinforcing required than the
probabilistic analysis for all the crack width limits considered. Those results obtained for the
edge restraint condition are comparable to those obtained by Holický, Retief and Wium (2009) in
which 15% more reinforcement was required to meet a 0.2 mm crack width limit using
deterministic methods as opposed to a reliability based assessment for the EN 1992 tension load
case.

Evidently, applying a probabilistic method of analysis provides a more economically viable

design. This was apparent more so in the edge restraint case rather than the end restraint, where
 150

the dominance held by the deterministic analysis was slight. This result suggests that, however
more conservative the end restraint crack model may be to that of the edge restraint crack model,
those reinforcement amounts calculated for crack limit satisfaction through the end restraint crack
model produced results that were close to those required to meet reliability requirements.

Table 6.13: Comparison of Deterministic and Probabilistic Analysis for wlim = 0.3, 0.2 and
0.1 mm (hc,eff = 2.5(c + ϕ/2), h = 250 mm, Model Uncertainty CoV = 0.3 and βt = 1.5)
Edge Restraint
crack width limit (mm) Area Required/Face (mm2)
Deterministic Probabilistic D/P
0.3 1353 1215 1.11
0.2 2275 2020 1.13
0.1 7156 5928 1.21
End Restraint
crack width limit (mm) Area Required/Face (mm2)
Deterministic Probabilistic D/P
0.3 1930 1890 1.02
0.2 2474 2430 1.02
0.1 3906 3811 1.02

This finding provides an interesting argument for the need of a more unified approach for
restrained strain estimation as proposed by Bamforth (2010). Bamforth (2010) developed in his
research a unified alternative means of estimating crack widths in which the restrained strain
separated crack formation into two stages. Those parts of the overall restrained strain coming
from the first stage of crack formation were based on the formula for the end restrained strain
under EN 1992-3: 2006. Further, Bamforth (2010) found in his investigation of the EN 1992
crack model that many of the assumptions made in the development of the edge restraint crack
model were not sufficiently robust. It seems that perhaps the development of a crack model that
better reflects the occurrence of cracks in practice, as Bamforth (2010) attempted to do, may in
fact lead to a model that is more compliant to South African reliability requirements for liquid
retaining structures. Further research into a model of crack formation that is more reflective of
observations made in practice and its reliability in the South African context is needed. This may
be further corroborated by an investigation into the model uncertainty of the EN 1992 restrained
shrinkage crack model (particularly for the edge restraint case).

6.4 Conclusion

The relative influence held by each random variable on the reliability of the restrained strain crack
models was considered. The relative influence was measured through a reverse FORM calculation
 151

of the crack models. Also, the theoretical partial safety factors of each random variable considered
in the analysis were calculated. This extended reliability analysis of the EN 1992 restrained strain
crack model provided with it greater insight into the ways in which EN 1992 may be adjusted for
compliance to South African reliability requirements. The following observations were made
from this sensitivity analysis:
Sensitivity factors indicate which basic variable is most influential on the reliability of the model
thus indicating to which random variable applying a partial factor to would have the most effect
on reliability of the model.
 Model uncertainty was found to be the most influential random variable for both the edge
and end restraint crack model. Model uncertainty was also found to be comparable for
the different effective depths considered (hc,eff = 2.5(c +φ/2) and h/2). Hence, applying a
partial factor to this variable could make the most impact on achieving the desired
reliability.
 The major difference was found with the relative influence of the concrete cover where
the effective depth of tension area (hc,eff) was different. Where the effective depth was
2.5(c +φ/2), the concrete cover’s relative influence was notably greater. This was true for
both edge and end restraint. It was then concluded that concrete cover was the variable
that influenced the difference in reliability generated by the effective depth of tension
area for both end and edge restraint.

The partial factor is essentially a factor which scales the nominal value of an input variable
to the value of the variable at the failure point of the performance function. A larger partial
factor indicates that there is a larger variation from nominal to failure point.
 Model uncertainty was found to require the largest theoretical partial safety factor across
all models considered.
 The edge and end restraint crack models were found to require comparable theoretical
partial factors for the random variables of the restrained strain crack model. This indicates
that the same partial factor may potentially be applied to the same random variable used
in either the edge or end restraint crack model.
 All models obtained theoretical partial factors for section thickness that were γh = 1 or
close to 1 (both variations of edge and end restraint crack model)
 The choice of reliability index has a significant effect on the design of liquid retaining
structure. This was particularly true for the edge restraint case.

The knowledge gained from the sensitivity analysis may then be used towards the full calibration
of the EN 1992 restrained shrinkage crack model. Unquestionably, a full calibration involves
 152

more than what was carried out for this research. Particularly since this research looks at a specific
configuration of a liquid retaining structure (LRS), with the parameters varied around this
particular set up – not considering at a large scope of liquid retaining structures and performance
applications. However, taking the above findings into consideration, the observations made from
this analysis may be used as a pilot towards a complete calibration for the EN 1992 restrained
strain crack model.
 153

Chapter 7: Final Conclusions

7.1 Introduction

The EN 1992-3:2006 and EN 1992-1-1:2004 design code have come to replace the corresponding
codes withdrawn of BS 8007:1987 and BS 8110-2:1985 (which South African engineers had
conventionally adopted). Currently, with South Africa having not yet developed and
implemented its own equivalent code, South African designers are investigating the adoption of
the EN 1992 design code. An investigation into the reliability performance of the EN 1992 crack
model as applied in the South African context was undertaken in this thesis. Understanding the
reliability performance of the EN 1992 crack model as compared to those reliability requirements
stipulated in the South African codes provides an opportunity for improvements of the design
code for use in the South African environment. Research into current South African practice, with
a review of the relevant British and Eurocode was undertaken. This was followed by calculations
conducted to quantify the implications of a change in code on the design of LRS under South
African conditions. Background knowledge of the reliability theory was also obtained with a
compilation of relevant parts presented herein– these tasks consequently fed into the reliability
analysis conducted in this dissertation. Important findings made through the above mentioned
undertakings are summarised in the subsequent text.

7.2 Literature Review

A review of current practices for liquid retaining structure design returned information on the
typical configurations and design selections that could be used in both the deterministic and
reliability based analyses of this research. Past research on liquid retaining structure design have
highlighted the dominance held by the cracking serviceability limit state as compared to the
ultimate limit state (Mcleod, 2013; Holický, Reteif and Wium, 2009). This substantiated the need
to conduct an investigation on the serviceability limit state, especially where a foreign code was
being applied in the South African environment.

7.3 Parametric study

A parametric study was conducted with the intent to both compare the BS 8007 and relevant parts
of BS 8110-2 crack models with that of EN 1992 and establish a reasonable representative liquid
retaining structure upon which the reliability assessment would be assessed. Additionally,
variables to which the EN 1992 crack model were found to be most sensitive were revealed as a
 154

by-product of the parametric study- the influence of these variables on the crack model were then
assessed in a reliability-based assessment of the crack model. Those key identified variables were
found to include the following: concrete cover (c), the reinforcing bar diameter to effective steel
content ratio (φ/ρp,eff), the effective tension area (Ac,eff), section thickness (h), the reinforcement
area (As) as well as the restraint degree (R).

7.4 FORM analysis of EN 1992

The influence of concrete cover, the reinforcing bar diameter to effective steel content ratio,
section thickness, restraint factor and model uncertainty were measured against increases in the
steel reinforcement to gross concrete cross-sectional area ratio (which had been found in previous
research to be a particularly influential variable for reinforced concrete structures). The reliability
of the crack model would decrease where concrete cover, the φ/ρp,eff ratio, restraint factor and
model uncertainty were increased. This was found to be true for both edge and end restraint
conditions. Section thickness was found to have the opposite effect on the reliability of the crack
model. The difference in reliability amounting from the effective depth (hc,eff) of tension zone was
found to be slight for both the edge and end restraint crack models. The end restraint crack model
was uncovered to be the more conservative of the two restrained shrinkage crack models–
requiring more reinforcement to achieve the target reliability index. The restraint factor was found
to have a significant influence on the reliability performance of the edge restraint crack model.
Increases in restraint factor was also found to decrease the reliability of edge restraint crack
model. Data on the restraint factor was found to be limited, thus further research is recommended
for this parameter.

7.5 Sensitivity Analysis

The relative influence, or otherwise sensitivity factor, of each random variable may be measured
through a sensitivity analysis of the reliability models. Model uncertainty was found to bear the
most influence on both the edge and end restraint crack models, thus applying a partial factor to
this variable would bring about the most effective adjustments (of the all random variables
accounted for in this analysis of the EN 1992 crack model) for compliance of South African
reliability performance requirements. Understandably, model uncertainty’s theoretical partial
factors were found to be the largest amongst all the random variables considered. Moreover, both
the edge and end restraint crack models (containing both hc,eff = h/2 and 2.5(c + φ/2)) returned
theoretical partial factors for section thickness (h) that were either 1 or close to 1.
 155

The difference in reliability between crack models with effective depths of tension zone h c,eff =
h/2 and 2.5(c + φ/2) was primarily attributed to the effects of concrete cover.

A comparison of the edge and end restraint crack models found that theoretical partial factors
required for both restraint conditions were quite comparable. This indicates that the same partial
factor may be used irrespective of the restraint condition being considered- ideal for the simple
application of a fully calibrated design standard.

An assessment of the implications of a change in the target reliability index was also carried out.
Target reliability indices 0.5, 1.5 and 2 were considered. It was found that significant increases in
reinforcement were required for increases in the stipulated target reliability index. This was
particularly evident for the edge restraint case.

7.6 Deterministic Versus Probabilistic Approach

A comparison between the reinforcing areas required for crack width limit satisfaction obtained
by means of deterministic and reliability based analysis was conducted. It was determined in this
exercise that reliability-based calculations returned more economically viable designs. The
demand of the deterministic calculation was experienced more so in the edge restraint case. The
amount of reinforcement required, found deterministically, for specified crack limit compliance
for the end restraint crack model were close to those required to meet the target reliability index.
This uncovering provides an interesting argument for a more unified approach to crack width
estimation for cracks due to restrained deformation, particularly for application in the South
African context. Perhaps an adoption of a crack model that lends itself more so towards the EN
1992 end restraint crack model would result in a crack model that is more conducive to local
reliability requirements. This also highlights the need to further investigate the model uncertainty.

7.7 Recommendations

 As mentioned earlier, further research is required on the statistical parameters and

characteristics of the restraint factor. Namely, more information is required on the mean,
standard deviation and probability distribution that best describe the restraint factor.
 In view of the considerable influence held by model uncertainty on the crack models of
both edge and end restraint, more information is required on the model uncertainty of the
cracking model for reinforced concrete structures.
 156

 It may be that those crack width limits stipulated by EN 1992 may be stricter than
necessary for South African design requirements. Further research is required on crack
width limits which are better suited to South African conditions and for South African
design practices, especially given their influence on the reliability of the crack model.
 Investigations into a crack model that is more reflective of the formation of cracks in
practice may be necessary.

By way of this investigation into the reliability of the EN 1992 crack model, South African
engineers are presented with the opportunity of selecting those combinations of variables for
which reliability was found to be satisfied. Alternatively, the identification of those variables
found to bear the most influence on the reliability of the EN 1992 crack model provide an
indication of where adjustments may be most effectively made for compliance to South
African safety requirements. Taking the above findings into consideration, a full calibration
may thus be attempted for a larger range of liquid retaining structure configurations and
design conditions (i.e. considering a larger scope of LRS)– subsequently improving the use
of the EN 1992 crack model for restrained deformation in South Africa.
 157

References

1) Addis, B. & Owens, G., eds., 2001. Fulton's concrete technology. 8th ed. Midrand:
Cement and Concrete Institute.

2) American Concrete Institute, 2002. ACI 207.2R-[Link] effect of restraint, volume

change, and reinforcement of cracking of mass concrete (reported by ACI committee
207). American Concrete Institute (ACI).

3) Ang, A. H-S. & Tang, W.H., 1984. Probability concepts in engineering planning and
design: decision, risk and [Link] 2- decision risk and reliability. John Wiley
& Sons.

4) Antona, B. & Johansson, R., 2011. Crack control of concrete structures. MSc thesis,
Chalmers University Of Technology.

5) Bamforth, P. B., 2007. Early-age thermal crack control in concrete. London: CIRIA.

6) Bamforth, P. B., Shave, J. & Denton S., 2011. Bridge Design to Eurocodes. [e-book]
Thomas Telford Limited. Available through: Iinstitute of Civil Engineers (ICE) Virtual
Library. Available at:
[Link]

7) Bamforth, Dr. P., Denton, Dr. S. & Shave, Dr. J., 2010. The development of a revised
unified approach for the design of reinforcement to control cracking in concrete resulting
from restrained contraction. Institute of Civil Engineers.

8) Barnardo-Viljoen C., Mensah K.K., Retief, J.V., Wium, J.A., Van Zijl GPAG., 2014.
Background to the Draft SA National Standard for the Design of Water Retaining
Structures. Concrete/Beton
 158

9) Beeby, A.W. & Narayanan, R.S., 2005. Designer's guide to eurocode 2: design of
concrete structures. London: Author and Thomas Telford Limited.

10) Bertagnoli, G., Gino, D., Mancini, G., 2016. Effect of endogenous deformations in
composite bridges. In: Marcinowski, J., ed. Proceedings of the XIII International
Conference on Metal Structures (CMS2016). Zielona Góra, Poland, 15-17 June 2016.
London: Taylor & Francis Group, pp. 287-298.

11) Bhatt, P., MacGinley, T. & Choo S., 2006. Reinforced concrete design theory and
examples. 3th ed. Abingdon: Taylor and Francis.

12) British Standards Institute, 1984. BS 5400-4:1984. Steel, concrete and composite
bridges-part 4: code of practice for design of concrete bridges.. London: British
Standards Institute (BSI).

13) British Standards Institute, 1985. BS 8110-2:1985. Structural use of concrete- part 2:
code of practice for special circumstances. London: British Standards Institute (BSI).

14) British Standards Institute, 1987. BS 8007:1987. Design of concrete structures for
retaining aqueous [Link]: British Standards Institute (BSI).

15) Caldentey, A.P., Peiretti, H. C., Iribarren, J.P. & Soto, A.J., 2013. Cracking of RC
members revisited: influence of cover, φ / ρs,ef and stirrup spacing– an experimental and
theoretical study. Structural Concrete. 14(1), pp. 69-78.

16) Caldentey, A. P., 2005. Cracking of reinforced concrete. is φ/ρeff a relevant parameter?
investigating a. beeby’s statement. [pdf] Polytechnic University of Madrid (UPM).
Availabe at:

[Link]

17) Centre of Construction Technology Research, 2000. Crack Control of Beams. Part 1: AS
3600 Design. [pdf] One Steel Reinforcing and the University of Western Sydney.
Available at:
 159

[Link]
hnical%20Resources%20PDFs/Crack_Control_of_Beams_Design_Booklet.pdf

18) Climate Change Knowledge Portal, 2009. Average monthly Temperature and Rainfall for
South Africa from 1900-2012. [online] Available at:
[Link]
gion=Africa&ThisCCode=ZAF

19) Croce, P., Diamantidis, D., & Vrouwenvelder, T., 2012. Structural reanalysis. In:
Diamantidis D. & Milan, H. eds. Innovative methods for the assessment of existing
structures. Prague: Czech Technical University, pp. 45-50.

20) Edvardsen, C., 1999. Water permeability and autogenous healing of cracks in concrete.
ACI Materials Journal, 96(4). pp. 448-454.

21) Eligehausen, R., Malle, R. & Rehm, G., 1976. Rissverhalten von Stahlbetonkörpern bei
Zugbeanspruchung. s.n.

22) European Committee For Standardization, 2004. EN 1992-1-1:2004. Eurocode 2: design

of concrete structures - part 1-1 : general rules and rules for buildings. CEN.

23) European Committee For Standardization, 2006. EN 1992-3:2006. Eurocode 2 - design

of concrete structures - part 3: liquid retaining and containment structures. CEN.

24) European Concrete Platform ASBL, 2008. Eurocode 2 commentary. [online] Brussels:
European Concrete Platform ASBL. Available at:
[Link]

25) Gilbert, R. I., 2016. Control of Cracking Caused by Early-Age Deformation in Edge
Restrained Concrete Walls. In: Zingoni, A., ed. Proceedings of the 6th International
Conference on Structural Engineering Mechanics and Computation (SEMC 2016). Cape
Town, South Africa, 5-7 September 2016. London: Taylor 7 Francis Group, pp. 539-540.

26) Green, A. E. & Bourne A.J., 1972. Reliability [Link].

 160

27) Greensmith, C.G., 2005. The effects of cement extenders and water binder ratio on the
heat evolution characteristics of concrete. MSc thesis, University of the Witwatersrand.

28) Harrison, T. A., 1981. Early-age thermal crack control in concrete. London: CIRIA.

29) Hartl, G., 1977. Die Arbeitslinie eingebetteter Stähle bei Erst- und Kurzzeitbelastung.
Universität Innsbruck.

30) Highways England, 1987. Designers Manual for Roads and Bridges (DMRB)) Volume
1, Section 3 Part 14- BA 24/87. Early thermal cracking of concrete. Highways England.
Available at:
[Link]

31) Highways England, 1987. Designers Manual for Roads and Bridges (DMRB) Volume 1
section 3 Part 14- HA BD 28/87. Early thermal cracking of concrete. Highways England.
Available at:
[Link]

32) Holicky, M., 2007. Probabilistic optimisation of concrete cover designed for durability.
In: International Congress, Concrete: construction’s sustainable option. Dundee,
September 2007. International Congress.

33) Holicky, M., 2009. Reliability analysis for structural design. Stellenbosch: Sun Media
Stellenbosch.

34) Holicky, M., Retief, J., Wium, J., 2009. Probabilistic Design for Cracking of Concrete
Structures. In: Vrijling, H., van Gelder, P. & Proske, D., eds. Proceedings of the 7th
international probabilistic workshop. Delft, the Netherlands, 2009. Dresden, Germany:
Proske, D. pp. 87-98.

35) Holický, M., Retief J. & Wium J., 2010. Partial factors for selected reinforced concrete
members: background to a revision of SANS 10100-1. Journal of The South African
Iinstitution Of Civil Engineering, 56(1), pp. 36-44.
 161

36) Holický, M. & Marková, J., 2012. Probablistic assessment. In: Diamantidis D. & Milan,
H. eds. Innovative methods for the assessment of existing structures. Prague: Czech
Technical University, pp. 51-75.

37) Joint Committee on Structural Safety (JCSS), 2000. Probabilistic model code part 3:
material properties. [pdf] JCSS. Available at:
[Link]

38) Joint Committee on Structural Safety (JCSS), 2001. Probabilistic model code part 1:
basis of design. [pdf] JCSS. Available at:
[Link]

39) Jones, Dr. T., 2008. Eurocode 2- Design of concrete structures - part 3: liquid retaining
and containment structures. Dissemination of information workshop. Brussels, 18-20
February.

40) Kamali, A.Z., Svedholm C. & Johansson, M., 2013. Effects of restrained thermal strains
in transversal direction of concrete slab frame bridges. Stockholm: Kamali A.Z.,
Svedholm C., Pacoste C.

41) Kaethner, S., 2011. Have EC2 Cracking Rules Advanced the Mystical Art of Crack Width
Prediction?. The Structural Engineer, 89(19), pp. 14-22.

42) Kheder, G.F., 1997. A New Look at the Control of Volume Change Cracking of Base
Restrained Concrete Walls. ACI Structural Journal. 94(3), pp. 262-271.

43) Mans, R. 2012. Will autogenous healing Seal a Concrete Crack with a Surface Width of
0.2mm? BSc, University of Kwa-Zulu Natal.

44) McLeod, C., Wium, J. & Retief, J.V., 2012. Reliability model for cracking in South
African reinforced concrete water retaining structures. In: Moormann, C., Huber, M.,
Proske, D., eds. Proceedings of the 10th International Probabilistic Workshop. Stuttgart,
15-16 November 2012. Institute of Geotechnical Engineering, University of Stuttgart,
Germany, pp. 141-156.
 162

45) McLeod, C., 2013. Investigation into cracking in reinforced concrete water-retaining
structures. MSc thesis. Stellenbosch University.

46) McLeod, C.H., Retief, J.V., Wium, J.A., 2013. Reliability of EN 1992 crack model
applied to South African water retaining structures. In: Zingoni, A., ed. Proceedings of
the 5th International Conference on Structural Engineering Mechanics and Computation
(SEMC 2013). Cape Town, South Africa, 2-4 September 2013. London: Taylor & Francis
Group, pp. 553-555.

47) McLeod, C., Viljoen, C., Retief, J., 2016. Quantification of Model Uncertainty of EN
1992 Crack Width Prediction Model. In: Zingoni, A., ed. Proceedings of the 6th
International Conference on Structural Engineering Mechanics and Computation (SEMC
2016). Cape Town, South Africa, 5-7 September 2016. London: Taylor & Francis Group,
pp. 475-476.

48) Mosley, B., Bungey, J. & Hulse, R., 2012. Reinforced concrete design to eurocode 2.
Seventh Edition. Hampshire: Macmillan Publishers Limited.

49) Nesset, J. & Skoglund, S., 2007. Reinforced concrete subjected to restraint forces. MSc
thesis, Chalmers University of Technology.

50) Nowak, A. S. & Collins, K.R., 2000. Reliability of structures. Mcgraw Hill Companies,
Inc.

51) Owens, G., ed., 2013. Fundamentals of Concrete. Third Edition. Midrand, South Africa:
The Concrete Institute.

52) Quan, Q. & Gengwei, Z., 2002. Calibration of reliability index of rc beams for
serviceability limit state of maximum crack width. Reliability engineering and system
safety. 75(3), pp. 359-366.

53) Rehm, G. & Rüsch, H., 1963. Versuche mit Betonformstählen. Deutscher Ausschuss für
Stahlbeton Heft.
 163

54) Retief, J. V., 2015. Contributions to the Implementation of the Principles of Reliability to
the Standardized Basis of Structural Design. PhD thesis, Stellenbosch University.

55) Saassouh, B. & Lounis, Z., 2012. Probabilistic modeling of chloride-induced corrosionin
concrete structures using First and Second Order reliability methods. Cement and
Concrete Composites, Volume 34, pp. 1082-1093.

56) Sajedi, S., Razavizadeh, A., Minaii, Z., Ghassemzadeh, F., Shekarchi, M., 2011. A
rational method for calculation of restrained shrinkage stresses in repaired concrete
members. In: Grantham, M., Mechtcherine, V., Schneck, U., eds. 4th international
conference on concrete repair. Dresden, Germany, 26-28 September 2011. London:
Taylor & Francis Group.

57) South African Bureau of Standards, 2000. SANS 10100-1:2000. The structural use of
concrete: part 1: design. Pretoria: SABS.

58) South African Bureau of Standards, 2011. SANS 10160-1:2011. Basis of structural
design and actions for buildings and industrial structures, part 1: basis of structural
design. Pretoria: SABS.

59) South African Bureau of Standards, SANS 2394:2004. General procedures for loading
in buildings and industrial structures. SABS.

60) [Link], 2015. South Africa's weather and climate. [online] Brand South Africa.
Available at:
[Link]

61) Wium, J., 2008. Concrete code. Structural eurocode summit. Pretoria, South Africa, 8
Holicky, Retief and Wium (2009) February.

62) Wium, J., Retief J. V. & Viljoen C. B., 2014. Lessons from development of design
standards in South Africa. In: International Association for Bridge and Structural
Engineering (IABSE), 37th Madrid IABSE Symposium 2014. Madrid, Spain, 3-5
September, 2014. Madrid, Spain: International Association for Bridge and Structural
Engineering, pp. 3198-3205.
 164

63) Wu, W., Lo, S-C R. & Wang, X.H., 2011. An objective approach to reliability-based risk
assessment and mitigation for coastal infrastructure development. Labour and
Management in Development Journal, [e-journal] Volume 2, pp. 0-16. Available through:
National Library of Australia website
[Link]

64) Yi, S., Hyun, T., & Kim, J., 2011. The effects of hydraulic pressure and crack width on
water permeability of penetration crack- induced concrete. Construction and building
materials. Volume 25, pp. 2576-2583.

65) Zahalan, R., 2010. Reliability-Based Analysis of Crack Control in Reinforced Concrete
Beams. MSc thesis, The State University of New Jersey.

66) Zhao, Y-G. & Ono, T., 1999. A general procedure for first/second-order reliability
method (FORM/SORM). Structural Safety. 21(2), pp. 95-112.

67) Zingoni, A., 2008. The case for the eurocodes for South Africa. Structural eurocode
summit. Pretoria, South Africa, 8 February.
 165

Appendix
 166

Appendix A: Deterministic Parametric Study

 167

Edge Restraint
EN 1992

Input (Constants)
bar diameter, Φ 16
ρeff=ρcrit =fctm(t)/fyk = 1.73/450= 0.384% 0.00384
Φ/ρeff = 4166.67
k1 0.8
k2 1
cover (mm) 40
R 0.5
T1 (°C) 15
T2 (°C) 23
αT (με/°C) 14
εca (με) 33
εcd (με) 220
ɛ (με) 392.5
ɛ (με) 0.00039

A = he,ef x 1000
he,ef is lesser of h/2 or 2.5(c + Φ/2)

reinforcement spacing (mm) 75

b (mm) 250
h (mm) 1000

BS 8007

fctm (t) Mpa 1.73

fb (deformed bars, type 2) Mpa 2.4
ϕ, mm 16
Figure A.1: Edge Restraint Crack Model Inputs for Deterministic Analysis
 168

End Restraint

Eurocode 2
restraint strain
(ɛsm - ɛcm ) =

concrete grade (Mpa) C30/37

section height (mm) 1000
kc 1
k 1
k 0.65
fct m(28) (Mpa) 2.9
Es (Gpa) 200
Ecm(t ) 28
αe 7.14286
bar diameter, Ф 16
bar spacing (mm) 75
As (mm2 ) / face 804.248
Act (mm2 )/ face 200000
ρ = As /Act 0.00402
1+1/αe ρ 35.8151

Crack Spacing

Sr, max = 3.4c + 0.425k1 k2 Φ/ρeff

c, cover 40
k1 (for high strength bonds) 0.8
k2 (for pure tension) 1
Φ 16
As 804.248
Ac,eff
ρeff
Φ/ρeff

BS 8007
restraint strain
ɛ = R α (T1 + T2 )

R 0.5
thermal expansion, α 14
T1 , (°C) 15
T2 , (°C) 23

Crack Spacing

Smax = (fct )/fb x Φ/2ρ

fct m (Mpa) 1.73

fb (Mpa), deformed bars- type 2 2.4
Φ (mm) 16

Figure A.2: End Restraint Crack Model Inputs for Deterministic Analysis
 169

BS 8007
cover (mm) Ф, bar dia. As / face h/2
2
A (area/ face, mm ) ρ ϕ/ρ (mm) ϕ/2ρ (mm) Sr,max (mm) ɛ wmax (mm)
40 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686
50 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686
60 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686
70 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686
80 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686
100 16 804.24772 125 125000 0.006433982 2486.796 1243.39799 896.2827199 0.000105 0.094109686

EN 1992
he,ef
cover (mm) Ф, bar dia. As / face h/2 2.5(c + Φ/2) Ac,eff (area/ face, mm2 ) ρeff Φ/ρeff (mm) Sr,max (mm) wmax (mm)
40 16 804.24772 125 120 120000 0.0067021 2387.32415 947.6902098 0.371968407
50 16 804.24772 125 145 125000 0.006434 2486.79599 1015.510635 0.398587924
60 16 804.24772 125 170 125000 0.006434 2486.79599 1049.510635 0.411932924
70 16 804.24772 125 195 125000 0.006434 2486.79599 1083.510635 0.425277924
80 16 804.24772 125 220 125000 0.006434 2486.79599 1117.510635 0.438622924
100 16 804.24772 125 270 125000 0.006434 2486.79599 1185.510635 0.465312924

Figure A.3: BS 8007 and EN 1992 Data for Varying Concrete Cover Value

BS 8007
cover (mm) Ф, bar dia. As / face h/2 Ac,eff (area/ face, mm2) ρ ϕ/ρ (mm) ϕ/2ρ (mm) Sr,max (mm) ɛ wmax
40 16 804.2477 125 125000 0.006433982 2486.795986 1243.397993 896.2827199 0.000266 0.238411203
40 20 1256.637 125 125000 0.010053096 1989.436789 994.7183943 717.0261759 0.000266 0.190728963
40 25 1963.495 125 125000 0.015707963 1591.549431 795.7747155 573.6209407 0.000266 0.15258317
40 32 3216.991 125 125000 0.025735927 1243.397993 621.6989965 448.1413599 0.000266 0.119205602
40 40 5026.548 125 125000 0.040212386 994.7183943 497.3591972 358.513088 0.000266 0.095364481

EN 1992
he,ef
cover (mm) Ф, bar dia. As h/2 2.5(c + Φ/2) Ac,eff (area/ face, mm2) ρp, eff Φ/ρp,eff Sr,max (mm) wmax ρp, eff (%)
40 16 804.2477 125 120 120000 0.006702064 2387.324146 947.6902098 0.371968407 85.64931888
40 20 1256.637 125 125 125000 0.010053096 1989.436789 812.4085081 0.318870339 83.25965341
40 25 1963.495 125 131.25 125000 0.015707963 1591.549431 677.1268065 0.265772272 79.91513573
40 32 3216.991 125 140 125000 0.025735927 1243.397993 558.7553176 0.219311462 75.6601869
40 40 5026.548 125 150 125000 0.040212386 994.7183943 474.2042541 0.18612517 71.32037538

Figure A.4: BS 8007 and EN 1992 Data for Varying ϕ/ρp,eff ratio
 170

Edge Restraint
EN 1992
he,eff
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
250 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.0003925 2.103172
250 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.0003925 1.078276
250 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.0003925 0.736644
250 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.0003925 0.565828
250 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.0003925 0.4633384
250 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.0003925 0.395012
250 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.0003925 0.346207429
250 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.0003925 0.309604
250 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.0003925 0.281134667
250 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.0003925 0.2583592
250 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.0003925 0.239724727
250 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.0003925 0.224196
250 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.0003925 0.211056308

he,eff
2
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
300 150000 0.001 150 120 120000 150 0.00125 12800 4488 0.0003925 1.76154
300 150000 0.002 150 120 120000 300 0.0025 6400 2312 0.0003925 0.90746
300 150000 0.003 150 120 120000 450 0.00375 4266.666667 1586.666667 0.0003925 0.622766667
300 150000 0.004 150 120 120000 600 0.005 3200 1224 0.0003925 0.48042
300 150000 0.005 150 120 120000 750 0.00625 2560 1006.4 0.0003925 0.395012
300 150000 0.006 150 120 120000 900 0.0075 2133.333333 861.3333333 0.0003925 0.338073333
300 150000 0.007 150 120 120000 1050 0.00875 1828.571429 757.7142857 0.0003925 0.297402857
300 150000 0.008 150 120 120000 1200 0.01 1600 680 0.0003925 0.2669
300 150000 0.009 150 120 120000 1350 0.01125 1422.222222 619.5555556 0.0003925 0.243175556
300 150000 0.01 150 120 120000 1500 0.0125 1280 571.2 0.0003925 0.224196
300 150000 0.011 150 120 120000 1650 0.01375 1163.636364 531.6363636 0.0003925 0.208667273
300 150000 0.012 150 120 120000 1800 0.015 1066.666667 498.6666667 0.0003925 0.195726667
300 150000 0.013 150 120 120000 1950 0.01625 984.6153846 470.7692308 0.0003925 0.184776923

he,eff
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
350 175000 0.001 175 120 120000 175 0.00145833 10971.42857 3866.285714 0.0003925 1.517517143
350 175000 0.002 175 120 120000 350 0.00291667 5485.714286 2001.142857 0.0003925 0.785448571
350 175000 0.003 175 120 120000 525 0.004375 3657.142857 1379.428571 0.0003925 0.541425714
350 175000 0.004 175 120 120000 700 0.00583333 2742.857143 1068.571429 0.0003925 0.419414286
350 175000 0.005 175 120 120000 875 0.00729167 2194.285714 882.0571429 0.0003925 0.346207429
350 175000 0.006 175 120 120000 1050 0.00875 1828.571429 757.7142857 0.0003925 0.297402857
350 175000 0.007 175 120 120000 1225 0.01020833 1567.346939 668.8979592 0.0003925 0.262542449
350 175000 0.008 175 120 120000 1400 0.01166667 1371.428571 602.2857143 0.0003925 0.236397143
350 175000 0.009 175 120 120000 1575 0.013125 1219.047619 550.4761905 0.0003925 0.216061905
350 175000 0.01 175 120 120000 1750 0.01458333 1097.142857 509.0285714 0.0003925 0.199793714
350 175000 0.011 175 120 120000 1925 0.01604167 997.4025974 475.1168831 0.0003925 0.186483377
350 175000 0.012 175 120 120000 2100 0.0175 914.2857143 446.8571429 0.0003925 0.175391429
350 175000 0.013 175 120 120000 2275 0.01895833 843.956044 422.9450549 0.0003925 0.166005934

he,eff
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
400 200000 0.001 200 120 120000 200 0.00166667 9600 3400 0.0003925 1.3345
400 200000 0.002 200 120 120000 400 0.00333333 4800 1768 0.0003925 0.69394
400 200000 0.003 200 120 120000 600 0.005 3200 1224 0.0003925 0.48042
400 200000 0.004 200 120 120000 800 0.00666667 2400 952 0.0003925 0.37366
400 200000 0.005 200 120 120000 1000 0.00833333 1920 788.8 0.0003925 0.309604
400 200000 0.006 200 120 120000 1200 0.01 1600 680 0.0003925 0.2669
400 200000 0.007 200 120 120000 1400 0.01166667 1371.428571 602.2857143 0.0003925 0.236397143
400 200000 0.008 200 120 120000 1600 0.01333333 1200 544 0.0003925 0.21352
400 200000 0.009 200 120 120000 1800 0.015 1066.666667 498.6666667 0.0003925 0.195726667
400 200000 0.01 200 120 120000 2000 0.01666667 960 462.4 0.0003925 0.181492
400 200000 0.011 200 120 120000 2200 0.01833333 872.7272727 432.7272727 0.0003925 0.169845455
400 200000 0.012 200 120 120000 2400 0.02 800 408 0.0003925 0.16014
400 200000 0.013 200 120 120000 2600 0.02166667 738.4615385 387.0769231 0.0003925 0.151927692

he,eff
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
450 225000 0.001 225 120 120000 225 0.001875 8533.333333 3037.333333 0.0003925 1.192153333
450 225000 0.002 225 120 120000 450 0.00375 4266.666667 1586.666667 0.0003925 0.622766667
450 225000 0.003 225 120 120000 675 0.005625 2844.444444 1103.111111 0.0003925 0.432971111
450 225000 0.004 225 120 120000 900 0.0075 2133.333333 861.3333333 0.0003925 0.338073333
450 225000 0.005 225 120 120000 1125 0.009375 1706.666667 716.2666667 0.0003925 0.281134667
450 225000 0.006 225 120 120000 1350 0.01125 1422.222222 619.5555556 0.0003925 0.243175556
450 225000 0.007 225 120 120000 1575 0.013125 1219.047619 550.4761905 0.0003925 0.216061905
450 225000 0.008 225 120 120000 1800 0.015 1066.666667 498.6666667 0.0003925 0.195726667
450 225000 0.009 225 120 120000 2025 0.016875 948.1481481 458.3703704 0.0003925 0.17991037
450 225000 0.01 225 120 120000 2250 0.01875 853.3333333 426.1333333 0.0003925 0.167257333
450 225000 0.011 225 120 120000 2475 0.020625 775.7575758 399.7575758 0.0003925 0.156904848
450 225000 0.012 225 120 120000 2700 0.0225 711.1111111 377.7777778 0.0003925 0.148277778
450 225000 0.013 225 120 120000 2925 0.024375 656.4102564 359.1794872 0.0003925 0.140977949

he,eff
2
h Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
500 250000 0.001 250 120 120000 250 0.00208333 7680 2747.2 0.0003925 1.078276
500 250000 0.002 250 120 120000 500 0.00416667 3840 1441.6 0.0003925 0.565828
500 250000 0.003 250 120 120000 750 0.00625 2560 1006.4 0.0003925 0.395012
500 250000 0.004 250 120 120000 1000 0.00833333 1920 788.8 0.0003925 0.309604
500 250000 0.005 250 120 120000 1250 0.01041667 1536 658.24 0.0003925 0.2583592
500 250000 0.006 250 120 120000 1500 0.0125 1280 571.2 0.0003925 0.224196
500 250000 0.007 250 120 120000 1750 0.01458333 1097.142857 509.0285714 0.0003925 0.199793714
500 250000 0.008 250 120 120000 2000 0.01666667 960 462.4 0.0003925 0.181492
500 250000 0.009 250 120 120000 2250 0.01875 853.3333333 426.1333333 0.0003925 0.167257333
500 250000 0.01 250 120 120000 2500 0.02083333 768 397.12 0.0003925 0.1558696
500 250000 0.011 250 120 120000 2750 0.02291667 698.1818182 373.3818182 0.0003925 0.146552364
500 250000 0.012 250 120 120000 3000 0.025 640 353.6 0.0003925 0.138788
500 250000 0.013 250 120 120000 3250 0.02708333 590.7692308 336.8615385 0.0003925 0.132218154

Figure A.5: EN 1992 Data for Edge Restraint with Varying Section Thickness
 171

End Restraint
EN 1992
he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c + ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff φ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
250 125000 0.001 141 125 120 120000 125 0.0010417 15360 5358.4 0.0073018 39.125889
250 125000 0.002 71 125 120 120000 250 0.0020833 7680 2747.2 0.0036768 10.100866
250 125000 0.003 47.6666667 125 120 120000 375 0.003125 5120 1876.8 0.0024685 4.6327914
250 125000 0.004 36 125 120 120000 500 0.0041667 3840 1441.6 0.0018643 2.6875543
250 125000 0.005 29 125 120 120000 625 0.0052083 3072 1180.48 0.0015018 1.772828
250 125000 0.006 24.3333333 125 120 120000 750 0.00625 2560 1006.4 0.0012601 1.2681838
250 125000 0.007 21 125 120 120000 875 0.0072917 2194.2857 882.05714 0.0010875 0.9592371
250 125000 0.008 18.5 125 120 120000 1000 0.0083333 1920 788.8 0.000958 0.7556986
250 125000 0.009 16.5555556 125 120 120000 1125 0.009375 1706.6667 716.26667 0.0008573 0.614085
250 125000 0.01 15 125 120 120000 1250 0.0104167 1536 658.24 0.0007768 0.5113114
250 125000 0.011 13.7272727 125 120 120000 1375 0.0114583 1396.3636 610.76364 0.0007109 0.4341776
250 125000 0.012 12.6666667 125 120 120000 1500 0.0125 1280 571.2 0.000656 0.37468
250 125000 0.013 11.7692308 125 120 120000 1625 0.0135417 1181.5385 537.72308 0.0006095 0.3277304

he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c +ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff ϕ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
300 150000 0.001 141 150 120 120000 150 0.00125 12800 4488 0.0073018 32.770414
300 150000 0.002 71 150 120 120000 300 0.0025 6400 2312 0.0036768 8.5007286
300 150000 0.003 47.6666667 150 120 120000 450 0.00375 4266.6667 1586.6667 0.0024685 3.9166111
300 150000 0.004 36 150 120 120000 600 0.005 3200 1224 0.0018643 2.2818857
300 150000 0.005 29 150 120 120000 750 0.00625 2560 1006.4 0.0015018 1.5113971
300 150000 0.006 24.3333333 150 120 120000 900 0.0075 2133.3333 861.33333 0.0012601 1.0853825
300 150000 0.007 21 150 120 120000 1050 0.00875 1828.5714 757.71429 0.0010875 0.8240143
300 150000 0.008 18.5 150 120 120000 1200 0.01 1600 680 0.000958 0.6514643
300 150000 0.009 16.5555556 150 120 120000 1350 0.01125 1422.2222 619.55556 0.0008573 0.5311705
300 150000 0.01 15 150 120 120000 1500 0.0125 1280 571.2 0.0007768 0.4437
300 150000 0.011 13.7272727 150 120 120000 1650 0.01375 1163.6364 531.63636 0.0007109 0.3779279
300 150000 0.012 12.6666667 150 120 120000 1800 0.015 1066.6667 498.66667 0.000656 0.3271016
300 150000 0.013 11.7692308 150 120 120000 1950 0.01625 984.61538 470.76923 0.0006095 0.2869235

he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c +ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff ϕ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
350 175000 0.001 141 175 120 120000 175 0.0014583 10971.429 3866.2857 0.0070462 27.242712
350 175000 0.002 71 175 120 120000 350 0.0029167 5485.7143 2001.1429 0.0035481 7.1002514
350 175000 0.003 47.6666667 175 120 120000 525 0.004375 3657.1429 1379.4286 0.0023821 3.2858769
350 175000 0.004 36 175 120 120000 700 0.0058333 2742.8571 1068.5714 0.001799 1.9223982
350 175000 0.005 29 175 120 120000 875 0.0072917 2194.2857 882.05714 0.0014492 1.2782977
350 175000 0.006 24.3333333 175 120 120000 1050 0.00875 1828.5714 757.71429 0.001216 0.9213918
350 175000 0.007 21 175 120 120000 1225 0.0102083 1567.3469 668.89796 0.0010494 0.7019666
350 175000 0.008 18.5 175 120 120000 1400 0.0116667 1371.4286 602.28571 0.0009245 0.5568158
350 175000 0.009 16.5555556 175 120 120000 1575 0.013125 1219.0476 550.47619 0.0008273 0.4554278
350 175000 0.01 15 175 120 120000 1750 0.0145833 1097.1429 509.02857 0.0007496 0.3815669
350 175000 0.011 13.7272727 175 120 120000 1925 0.0160417 997.4026 475.11688 0.000686 0.3259283
350 175000 0.012 12.6666667 175 120 120000 2100 0.0175 914.28571 446.85714 0.000633 0.2828579
350 175000 0.013 11.7692308 175 120 120000 2275 0.0189583 843.95604 422.94505 0.0005881 0.2487536

he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c +ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff ϕ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
400 200000 0.001 141 200 120 120000 200 0.0016667 9600 3400 0.0067907 23.088246
400 200000 0.002 71 200 120 120000 400 0.0033333 4800 1768 0.0034194 6.0455181
400 200000 0.003 47.6666667 200 120 120000 600 0.005 3200 1224 0.0022957 2.8098887
400 200000 0.004 36 200 120 120000 800 0.0066667 2400 952 0.0017338 1.650564
400 200000 0.005 29 200 120 120000 1000 0.0083333 1920 788.8 0.0013967 1.101686
400 200000 0.006 24.3333333 200 120 120000 1200 0.01 1600 680 0.0011719 0.7968993
400 200000 0.007 21 200 120 120000 1400 0.0116667 1371.4286 602.28571 0.0010114 0.6091367
400 200000 0.008 18.5 200 120 120000 1600 0.0133333 1200 544 0.000891 0.4846894
400 200000 0.009 16.5555556 200 120 120000 1800 0.015 1066.6667 498.66667 0.0007973 0.3976006
400 200000 0.01 15 200 120 120000 2000 0.0166667 960 462.4 0.0007224 0.3340427
400 200000 0.011 13.7272727 200 120 120000 2200 0.0183333 872.72727 432.72727 0.0006611 0.2860826
400 200000 0.012 12.6666667 200 120 120000 2400 0.02 800 408 0.00061 0.2488946
400 200000 0.013 11.7692308 200 120 120000 2600 0.0216667 738.46154 387.07692 0.0005668 0.2194008

he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c +ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff ϕ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
450 225000 0.001 141 225 120 120000 225 0.001875 8533.3333 3037.3333 0.0065351 19.849272
450 225000 0.002 71 225 120 120000 450 0.00375 4266.6667 1586.6667 0.0032907 5.2212808
450 225000 0.003 47.6666667 225 120 120000 675 0.005625 2844.4444 1103.1111 0.0022093 2.4370646
450 225000 0.004 36 225 120 120000 900 0.0075 2133.3333 861.33333 0.0016685 1.4371654
450 225000 0.005 29 225 120 120000 1125 0.009375 1706.6667 716.26667 0.0013441 0.9627327
450 225000 0.006 24.3333333 225 120 120000 1350 0.01125 1422.2222 619.55556 0.0011278 0.6987388
450 225000 0.007 21 225 120 120000 1575 0.013125 1219.0476 550.47619 0.0009733 0.5357854
450 225000 0.008 18.5 225 120 120000 1800 0.015 1066.6667 498.66667 0.0008574 0.4275777
450 225000 0.009 16.5555556 225 120 120000 2025 0.016875 948.14815 458.37037 0.0007673 0.351717
450 225000 0.01 15 225 120 120000 2250 0.01875 853.33333 426.13333 0.0006952 0.2962578
450 225000 0.011 13.7272727 225 120 120000 2475 0.020625 775.75758 399.75758 0.0006362 0.2543396
450 225000 0.012 12.6666667 225 120 120000 2700 0.0225 711.11111 377.77778 0.0005871 0.2217848
450 225000 0.013 11.7692308 225 120 120000 2925 0.024375 656.41026 359.17949 0.0005455 0.1959262

he,eff
2
h Act / face ρ 1+1/αe ρ h/2 2.5(c +ϕ/2)Ac,eff (area/ face, mmA )s ρp,eff ϕ/ρp,eff (mm)Sr,max EN 2 ɛ (EN 2) Wmax EN 2
500 250000 0.001 141 250 120 120000 250 0.0020833 7680 2747.2 0.0062795 17.251141
500 250000 0.002 71 250 120 120000 500 0.0041667 3840 1441.6 0.003162 4.5583907
500 250000 0.003 47.6666667 250 120 120000 750 0.00625 2560 1006.4 0.0021229 2.1364554
500 250000 0.004 36 250 120 120000 1000 0.0083333 1920 788.8 0.0016033 1.2646718
500 250000 0.005 29 250 120 120000 1250 0.0104167 1536 658.24 0.0012915 0.8501405
500 250000 0.006 24.3333333 250 120 120000 1500 0.0125 1280 571.2 0.0010837 0.6190108
500 250000 0.007 21 250 120 120000 1750 0.0145833 1097.1429 509.02857 0.0009353 0.476069
500 250000 0.008 18.5 250 120 120000 2000 0.0166667 960 462.4 0.0008239 0.3809763
500 250000 0.009 16.5555556 250 120 120000 2250 0.01875 853.33333 426.13333 0.0007373 0.3141939
500 250000 0.01 15 250 120 120000 2500 0.0208333 768 397.12 0.000668 0.2652903
500 250000 0.011 13.7272727 250 120 120000 2750 0.0229167 698.18182 373.38182 0.0006114 0.2282684
500 250000 0.012 12.6666667 250 120 120000 3000 0.025 640 353.6 0.0005641 0.1994725
500 250000 0.013 11.7692308 250 120 120000 3250 0.0270833 590.76923 336.86154 0.0005242 0.1765663

Figure A.6: EN 1992 Data for End Restraint Crack Model with Varying Section Thickness
 172

Edge and End Restraint

BS 8007 (same equation for both)
walls, suspended slabs
h h (BS 8007, mm) Ac (BS 8007) As ρ ϕ/2ρ Sr,max (BS 8007) ɛ (BS 8007) Wmax BS 8007
250 125 125000 125 0.001 8000 5766.666667 0.000266 1.533933333
250 125 125000 250 0.002 4000 2883.333333 0.000266 0.766966667
250 125 125000 375 0.003 2666.666667 1922.222222 0.000266 0.511311111
250 125 125000 500 0.004 2000 1441.666667 0.000266 0.383483333
250 125 125000 625 0.005 1600 1153.333333 0.000266 0.306786667
250 125 125000 750 0.006 1333.333333 961.1111111 0.000266 0.255655556
250 125 125000 875 0.007 1142.857143 823.8095238 0.000266 0.219133333
250 125 125000 1000 0.008 1000 720.8333333 0.000266 0.191741667
250 125 125000 1125 0.009 888.8888889 640.7407407 0.000266 0.170437037
250 125 125000 1250 0.01 800 576.6666667 0.000266 0.153393333
250 125 125000 1375 0.011 727.2727273 524.2424242 0.000266 0.139448485
250 125 125000 1500 0.012 666.6666667 480.5555556 0.000266 0.127827778
250 125 125000 1625 0.013 615.3846154 443.5897436 0.000266 0.117994872

walls, suspended slabs

h h (BS 8007, mm) Ac (BS 8007) As ρ ϕ/2ρ Sr,max (BS 8007) ɛ (BS 8007) Wmax BS 8007
300 150 150000 150 0.001 8000 5766.666667 0.000266 1.533933333
300 150 150000 300 0.002 4000 2883.333333 0.000266 0.766966667
300 150 150000 450 0.003 2666.666667 1922.222222 0.000266 0.511311111
300 150 150000 600 0.004 2000 1441.666667 0.000266 0.383483333
300 150 150000 750 0.005 1600 1153.333333 0.000266 0.306786667
300 150 150000 900 0.006 1333.333333 961.1111111 0.000266 0.255655556
300 150 150000 1050 0.007 1142.857143 823.8095238 0.000266 0.219133333
300 150 150000 1200 0.008 1000 720.8333333 0.000266 0.191741667
300 150 150000 1350 0.009 888.8888889 640.7407407 0.000266 0.170437037
300 150 150000 1500 0.01 800 576.6666667 0.000266 0.153393333
300 150 150000 1650 0.011 727.2727273 524.2424242 0.000266 0.139448485
300 150 150000 1800 0.012 666.6666667 480.5555556 0.000266 0.127827778
300 150 150000 1950 0.013 615.3846154 443.5897436 0.000266 0.117994872

walls, suspended slabs

h h (BS 8007, mm) Ac (BS 8007) As ρ ϕ/2ρ Sr,max (BS 8007) ɛ (BS 8007) Wmax BS 8007
350 175 175000 175 0.001 8000 5766.666667 0.000266 1.533933333
350 175 175000 350 0.002 4000 2883.333333 0.000266 0.766966667
350 175 175000 525 0.003 2666.666667 1922.222222 0.000266 0.511311111
350 175 175000 700 0.004 2000 1441.666667 0.000266 0.383483333
350 175 175000 875 0.005 1600 1153.333333 0.000266 0.306786667
350 175 175000 1050 0.006 1333.333333 961.1111111 0.000266 0.255655556
350 175 175000 1225 0.007 1142.857143 823.8095238 0.000266 0.219133333
350 175 175000 1400 0.008 1000 720.8333333 0.000266 0.191741667
350 175 175000 1575 0.009 888.8888889 640.7407407 0.000266 0.170437037
350 175 175000 1750 0.01 800 576.6666667 0.000266 0.153393333
350 175 175000 1925 0.011 727.2727273 524.2424242 0.000266 0.139448485
350 175 175000 2100 0.012 666.6666667 480.5555556 0.000266 0.127827778
350 175 175000 2275 0.013 615.3846154 443.5897436 0.000266 0.117994872

Figure A.7: BS 8007 Edge and End Restraint Crack Model Data
 173

Edge Restraint
EN 1992
h 250 mm constant
he,eff
R Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
0.1 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.0000785 0.4206344
0.1 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.0000785 0.2156552
0.1 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.0000785 0.1473288
0.1 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.0000785 0.1131656
0.1 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.0000785 0.09266768
0.1 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.0000785 0.0790024
0.1 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.0000785 0.069241486
0.1 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.0000785 0.0619208
0.1 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.0000785 0.056226933
0.1 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.0000785 0.05167184
0.1 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.0000785 0.047944945
0.1 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.0000785 0.0448392
0.1 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.0000785 0.042211262

h 250 mm
he,eff
R Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
0.2 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.000157 0.8412688
0.2 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.000157 0.4313104
0.2 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.000157 0.2946576
0.2 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.000157 0.2263312
0.2 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.000157 0.18533536
0.2 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.000157 0.1580048
0.2 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.000157 0.138482971
0.2 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.000157 0.1238416
0.2 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.000157 0.112453867
0.2 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.000157 0.10334368
0.2 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.000157 0.095889891
0.2 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.000157 0.0896784
0.2 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.000157 0.084422523

h 250 mm
he,eff
R Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
0.3 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.0002355 1.2619032
0.3 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.0002355 0.6469656
0.3 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.0002355 0.4419864
0.3 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.0002355 0.3394968
0.3 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.0002355 0.27800304
0.3 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.0002355 0.2370072
0.3 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.0002355 0.207724457
0.3 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.0002355 0.1857624
0.3 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.0002355 0.1686808
0.3 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.0002355 0.15501552
0.3 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.0002355 0.143834836
0.3 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.0002355 0.1345176
0.3 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.0002355 0.126633785

h 250 mm
he,eff
R Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
0.4 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.000314 1.6825376
0.4 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.000314 0.8626208
0.4 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.000314 0.5893152
0.4 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.000314 0.4526624
0.4 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.000314 0.37067072
0.4 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.000314 0.3160096
0.4 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.000314 0.276965943
0.4 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.000314 0.2476832
0.4 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.000314 0.224907733
0.4 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.000314 0.20668736
0.4 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.000314 0.191779782
0.4 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.000314 0.1793568
0.4 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.000314 0.168845046

he,eff
R Act / face ρ h/2 2.5(c + ϕ/2) Ac,eff (area/ face, mm2 ) As ρp ,eff φ/ρp ,eff (mm) Sr,max EN 2 ɛ (EN 2) Wmax EN 2
0.5 125000 0.001 125 120 120000 125 0.00104167 15360 5358.4 0.0003925 2.103172
0.5 125000 0.002 125 120 120000 250 0.00208333 7680 2747.2 0.0003925 1.078276
0.5 125000 0.003 125 120 120000 375 0.003125 5120 1876.8 0.0003925 0.736644
0.5 125000 0.004 125 120 120000 500 0.00416667 3840 1441.6 0.0003925 0.565828
0.5 125000 0.005 125 120 120000 625 0.00520833 3072 1180.48 0.0003925 0.4633384
0.5 125000 0.006 125 120 120000 750 0.00625 2560 1006.4 0.0003925 0.395012
0.5 125000 0.007 125 120 120000 875 0.00729167 2194.285714 882.0571429 0.0003925 0.346207429
0.5 125000 0.008 125 120 120000 1000 0.00833333 1920 788.8 0.0003925 0.309604
0.5 125000 0.009 125 120 120000 1125 0.009375 1706.666667 716.2666667 0.0003925 0.281134667
0.5 125000 0.01 125 120 120000 1250 0.01041667 1536 658.24 0.0003925 0.2583592
0.5 125000 0.011 125 120 120000 1375 0.01145833 1396.363636 610.7636364 0.0003925 0.239724727
0.5 125000 0.012 125 120 120000 1500 0.0125 1280 571.2 0.0003925 0.224196
0.5 125000 0.013 125 120 120000 1625 0.01354167 1181.538462 537.7230769 0.0003925 0.211056308

Figure A.8: EN 1992 Edge Restraint Data with Varying Restraint Factor
 174

Edge Restraint
cover = 40 mm
BS 8007 h=250mm

Restraint Degree h/2 Act/ face (mm2)As / face (mm2)ρ ϕ/2ρ S r, max (mm) ε Wmax (mm)
0.1 125 125000 125 0.001 8000 5766.666667 0.0000532 0.30678667
0.1 125 125000 250 0.002 4000 2883.333333 0.0000532 0.15339333
0.1 125 125000 375 0.003 2666.7 1922.222222 0.0000532 0.10226222
0.1 125 125000 500 0.004 2000 1441.666667 0.0000532 0.07669667
0.1 125 125000 625 0.005 1600 1153.333333 0.0000532 0.06135733
0.1 125 125000 750 0.006 1333.3 961.1111111 0.0000532 0.05113111
0.1 125 125000 875 0.007 1142.9 823.8095238 0.0000532 0.04382667
0.1 125 125000 1000 0.008 1000 720.8333333 0.0000532 0.03834833
0.1 125 125000 1125 0.009 888.89 640.7407407 0.0000532 0.03408741
0.1 125 125000 1250 0.01 800 576.6666667 0.0000532 0.03067867
0.1 125 125000 1375 0.011 727.27 524.2424242 0.0000532 0.0278897
0.1 125 125000 1500 0.012 666.67 480.5555556 0.0000532 0.02556556
0.1 125 125000 1625 0.013 615.38 443.5897436 0.0000532 0.02359897

h=250mm

Restraint Degree h/2 Act/ face (mm2)As / face (mm2)ρ ϕ/2ρ S r, max (mm) ε Wmax (mm)
0.2 125 125000 125 0.001 8000 5766.666667 0.0001064 0.61357333
0.2 125 125000 250 0.002 4000 2883.333333 0.0001064 0.30678667
0.2 125 125000 375 0.003 2666.7 1922.222222 0.0001064 0.20452444
0.2 125 125000 500 0.004 2000 1441.666667 0.0001064 0.15339333
0.2 125 125000 625 0.005 1600 1153.333333 0.0001064 0.12271467
0.2 125 125000 750 0.006 1333.3 961.1111111 0.0001064 0.10226222
0.2 125 125000 875 0.007 1142.9 823.8095238 0.0001064 0.08765333
0.2 125 125000 1000 0.008 1000 720.8333333 0.0001064 0.07669667
0.2 125 125000 1125 0.009 888.89 640.7407407 0.0001064 0.06817481
0.2 125 125000 1250 0.01 800 576.6666667 0.0001064 0.06135733
0.2 125 125000 1375 0.011 727.27 524.2424242 0.0001064 0.05577939
0.2 125 125000 1500 0.012 666.67 480.5555556 0.0001064 0.05113111
0.2 125 125000 1625 0.013 615.38 443.5897436 0.0001064 0.04719795

h=250mm

Restraint Degree h/2 Act/ face (mm2)As / face (mm2)ρ ϕ/2ρ S r, max (mm) ε Wmax (mm)
0.3 125 125000 125 0.001 8000 5766.666667 0.0001596 0.92036
0.3 125 125000 250 0.002 4000 2883.333333 0.0001596 0.46018
0.3 125 125000 375 0.003 2666.7 1922.222222 0.0001596 0.30678667
0.3 125 125000 500 0.004 2000 1441.666667 0.0001596 0.23009
0.3 125 125000 625 0.005 1600 1153.333333 0.0001596 0.184072
0.3 125 125000 750 0.006 1333.3 961.1111111 0.0001596 0.15339333
0.3 125 125000 875 0.007 1142.9 823.8095238 0.0001596 0.13148
0.3 125 125000 1000 0.008 1000 720.8333333 0.0001596 0.115045
0.3 125 125000 1125 0.009 888.89 640.7407407 0.0001596 0.10226222
0.3 125 125000 1250 0.01 800 576.6666667 0.0001596 0.092036
0.3 125 125000 1375 0.011 727.27 524.2424242 0.0001596 0.08366909
0.3 125 125000 1500 0.012 666.67 480.5555556 0.0001596 0.07669667
0.3 125 125000 1625 0.013 615.38 443.5897436 0.0001596 0.07079692

h=250mm

Restraint Degree h/2 Act/ face (mm2)As / face (mm2)ρ ϕ/2ρ S r, max (mm) ε Wmax (mm)
0.4 125 125000 125 0.001 8000 5766.666667 0.0002128 1.22714667
0.4 125 125000 250 0.002 4000 2883.333333 0.0002128 0.61357333
0.4 125 125000 375 0.003 2666.7 1922.222222 0.0002128 0.40904889
0.4 125 125000 500 0.004 2000 1441.666667 0.0002128 0.30678667
0.4 125 125000 625 0.005 1600 1153.333333 0.0002128 0.24542933
0.4 125 125000 750 0.006 1333.3 961.1111111 0.0002128 0.20452444
0.4 125 125000 875 0.007 1142.9 823.8095238 0.0002128 0.17530667
0.4 125 125000 1000 0.008 1000 720.8333333 0.0002128 0.15339333
0.4 125 125000 1125 0.009 888.89 640.7407407 0.0002128 0.13634963
0.4 125 125000 1250 0.01 800 576.6666667 0.0002128 0.12271467
0.4 125 125000 1375 0.011 727.27 524.2424242 0.0002128 0.11155879
0.4 125 125000 1500 0.012 666.67 480.5555556 0.0002128 0.10226222
0.4 125 125000 1625 0.013 615.38 443.5897436 0.0002128 0.0943959

h=250mm

Restraint Degree h/2 Act/ face (mm2)As / face (mm2)ρ ϕ/2ρ S r, max (mm) ε Wmax (mm)
0.5 125 125000 125 0.001 8000 5766.666667 0.000266 1.53393333
0.5 125 125000 250 0.002 4000 2883.333333 0.000266 0.76696667
0.5 125 125000 375 0.003 2666.7 1922.222222 0.000266 0.51131111
0.5 125 125000 500 0.004 2000 1441.666667 0.000266 0.38348333
0.5 125 125000 625 0.005 1600 1153.333333 0.000266 0.30678667
0.5 125 125000 750 0.006 1333.3 961.1111111 0.000266 0.25565556
0.5 125 125000 875 0.007 1142.9 823.8095238 0.000266 0.21913333
0.5 125 125000 1000 0.008 1000 720.8333333 0.000266 0.19174167
0.5 125 125000 1125 0.009 888.89 640.7407407 0.000266 0.17043704
0.5 125 125000 1250 0.01 800 576.6666667 0.000266 0.15339333
0.5 125 125000 1375 0.011 727.27 524.2424242 0.000266 0.13944848
0.5 125 125000 1500 0.012 666.67 480.5555556 0.000266 0.12782778
0.5 125 125000 1625 0.013 615.38 443.5897436 0.000266 0.11799487

Figure A.9: BS 8007 Edge Restraint Data with Varying Restraint Factor
 175

Appendix B: Data for FORM Analysis

 176

1) Edge Restraint Crack Model (hc,eff = 2.5(c + φ/2))

Symbols

Concrete cover c
Bar diameter phi
Steel reinforcing A
Section thickness H

Effective tension depth hc

Section width b
Coefficient of thermal
alpha
expansion
Early age change in
T1
temperature

Seasonal fall in temperature T2

Autogenous shrinkage strain eca

Drying shrinkage strain ecd
Crack width limit wl
Model uncertainty th

syms c k1 k2 k3 k4 phi A h hc b R alpha T1 T2 eca ecd wl th

d=h-c-0.5*phi;
hc=2.5*(h-d);
pe=A/(hc*b);
s=k3*c+(k1*k2*k4*phi)/pe;
e= R*(alpha*(T1+T2)+eca+ecd);
g= wl - th*s*e;
diff(g,h)
diff(g,c)
diff(g,th)

Partial Derivatives

diff(g,h) =

diff(g,c) =

-th*(k3+5/2*k1*k2*k4*phi/A*b)*R*(alpha*(T1+T2)+eca+ecd);

diff(g,th) =

-(k3*c+k1*k2*k4*phi/A*(5/2*c+5/4*phi)*b)*R*(alpha*(T1+T2)+eca+ecd)

Figure B.1: Edge Restraint MATLAB Input for FORM Analysis (hc,eff = 2.5(c + ϕ/2))
 177

2) End Restraint Crack Model (h c,eff = 2.5(c + φ/2))

syms c k1 k2 k3 k4 phi A h hc b alphae kc k f E wl th

d=h-c-0.5*phi;
hc=2.5*(h-d);
pe=A/(hc*b);
p=A/(0.5*h*b);
s=k3*c+k1*k2*k4*phi/pe;
e=(0.5*alphae*kc*k*f*(1+(alphae*p)^-1))/E;
g=wl-th*s*e;
diff(g,h)
diff(g,c)
diff(g,f)
diff(g,th)

Partial Derivatives

diff(g,h)

ans =

-1/4*th*(k3*c+k1*k2*k4*phi/A*(5/2*c+5/4*phi)*b)*kc*k*f/A*b/E;

diff(g,c)

ans =

-1/2*th*(k3+5/2*k1*k2*k4*phi/A*b)*alphae*kc*k*f*(1+1/2/alphae/A*h*b)/E;

diff(g,f)

ans =

-1/2*th*(k3*c+k1*k2*k4*phi/A*(5/2*c+5/4*phi)*b)*alphae*kc*k*(1+1/2/alphae/A*h*b)/E;

diff(g,th)

ans =

-1/2*(k3*c+k1*k2*k4*phi/A*(5/2*c+5/4*phi)*b)*alphae*kc*k*f*(1+1/2/alphae/A*h*b)/E;
Figure B.2: End Restraint MATLAB Input for FORM Analysis (hc,eff = 2.5(c + ϕ/2))
 178

1) Edge Restraint Crack Model (hc,eff = h/2)

syms c k1 k2 k3 k4 phi A h hc b R alpha T1 T2 eca ecd wl th

hc=0.5*h;
pe=A/(hc*b);
s=k3*c+(k1*k2*k4*phi)/pe;
e= R*(alpha*(T1+T2)+eca+ecd);
g= wl - th*s*e;
diff(g,h)
diff(g,c)
diff(g,th)

derivatives:

>> diff(g,h)

ans =

-1/2*th*k1*k2*k4*phi/A*b*R*(alpha*(T1+T2)+eca+ecd)

>> diff(g,c)

ans =

-th*k3*R*(alpha*(T1+T2)+eca+ecd)

>> diff(g,th)

ans =

-(k3*c+1/2*k1*k2*k4*phi/A*h*b)*R*(alpha*(T1+T2)+eca+ecd)
Figure B.3: Edge Restraint MATLAB Input for FORM Analysis (hc,eff = h/2)
 179

2) End Restraint Crack Model (hc,eff = h/2)

syms c k1 k2 k3 k4 phi A h hc b alphae kc k f E wl th

hc=0.5*h;
pe=A/(hc*b);
p=A/(0.5*h*b);
s=k3*c+k1*k2*k4*phi/pe;
e=(0.5*alphae*kc*k*f*(1+(alphae*p)^-1))/E;
g=wl-th*s*e;
diff(g,h)
diff(g,c)
diff(g,f)
diff(g,th)

Partial Derivatives

diff(g,h)

ans =

-1/4*th*k1*k2*k4*phi/A*b*alphae*kc*k*f*(1+1/2/alphae/A*h*b)/E-1/4*th*(k3*c+1/2*k1*k2*k4*phi/A*h*b)*kc*k*f/A*b/E

diff(g,c)

ans =

-1/2*th*k3*alphae*kc*k*f*(1+1/2/alphae/A*h*b)/E

diff(g,f)

ans =

-1/2*th*(k3*c+1/2*k1*k2*k4*phi/A*h*b)*alphae*kc*k*(1+1/2/alphae/A*h*b)/E

diff(g,th)

ans =

-1/2*(k3*c+1/2*k1*k2*k4*phi/A*h*b)*alphae*kc*k*f*(1+1/2/alphae/A*h*b)/E

Figure B.4: End Restraint MATLAB Input for FORM Analysis (hc,eff = h/2)
 180

Edge Restraint cov: 0.3

Constant Variables
As unkown variable εca = 33με R=0.5
c=40mm εcd = 220με
h=250mm αT ,c = 14με
250mm thick cover = 40mm
2.5*(c+ϕ/2)
model uncertainty (cov) Area (%) Area (mm2) Area/face β(mm2) h* c* θ* Srm ε wcalc g(x) probability(s) probability(f)
0.3 0.5 1250 625 solver could not find feasible solution therefore gave 0
0.3 0.93 2328 1164 0.16729791 0.25 0.039935 1.001347 0.508869 0.0003925 0.0002 1.11678E-12 0.566432177 0.433567823 min reinf. required
0.3 1 2500 1250 0.35113627 0.25 0.040363 1.053603 0.48363 0.0003925 0.00019 1.74267E-11 0.637256941 0.362743059
0.3 1.5 3750 1875 1.34889791 0.25 0.042826 1.386947 0.367393 0.0003925 0.000144 -1.43176E-11 0.911315119 0.088684881
0.3 2 5000 2500 1.99974707 0.25 0.044579 1.657214 0.307476 0.0003925 0.000121 -2.55916E-11 0.977236209 0.022763791
0.3 2.5 6250 3125 2.46699742 0.25 0.045921 1.881738 0.270789 0.0003925 0.000106 7.62381E-11 0.993187433 0.006812567
0.3 3 7500 3750 2.82232372 0.25 0.046993 2.071627 0.245968 0.0003925 9.65E-05 -9.84676E-11 0.997616149 0.002383851

250mm thick cover = 40mm

h/2
model uncertainty (cov) Area (%) Area (mm2) Area/face β(mm2) h* c* θ* Srm ε wcalc g(x) probability(s) probability(f)
0.3 0.5 1250 625 solver could not find feasible solution therefore gave 0
0.3 0.93119824 2328 1164 0.17189348 0.250012 0.039632 1.006413 0.508838 0.0003925 0.0002 -9.99751E-07 0.568239363 0.431760637 min reinf. required
0.3 1 2500 1250 0.35404227 0.250025 0.039726 1.062696 0.479492 0.0003925 0.000188 -1.51783E-12 0.638346397 0.361653603
0.3 1.5 3750 1875 1.43253849 0.250091 0.040549 1.464822 0.347861 0.0003925 0.000137 3.93235E-12 0.924005114 0.075994886
0.3 2 5000 2500 2.13387707 0.250124 0.041429 1.800812 0.282958 0.0003925 0.000111 -4.82536E-12 0.983573578 0.016426422
0.3 2.5 6250 3125 2.63408494 0.250141 0.042306 2.082344 0.244702 0.0003925 9.6E-05 -3.46189E-11 0.995781781 0.004218219
0.3 3 7500 3750 3.01110895 0.250149 0.043149 2.319209 0.21971 0.0003925 8.62E-05 -6.58466E-11 0.998698523 0.001301477

Figure B.5: Selected Data of EN 1992 Edge Restraint Crack Model FORM Analysis (hc,eff = 2.5(c + ϕ/2) and h/2- Effective Depth Comparison)
 181

End Restraint cov: 0.3

Constant Variables
As unkown variable αe=7
c=40mm kc=1 Es = 200GPa
h=250mm k=1
250mm thick cover = 40mm
2.5*(c+ϕ/2)
model uncertainty (cov) Area (%) Area (mm2) Area/face β(mm2) h* c* fct,eff θ* Srm ε wcalc g(x) probability(s) probability(f)
0.3 0.5 1250 625 solver could not find feasible solution therefore gave 0
0.3 1 2500 1250 solver could not find feasible solution therefore gave 0
0.3 1.43267655 3582 1791 0.19237675 0.250012 0.039938 2901.059 1.000911 0.3587251 0.000557 0.000199818 8.01721E-12 0.576276448 0.423723552 min reinf. req.
0.3 1.5 3750 1875 0.39691205 0.250024 0.040354 2958.374 1.05094 0.3492613 0.000545 0.000190306 -2.65663E-12 0.654283841 0.345716159
0.3 2 5000 2500 1.64017775 0.250095 0.043039 3330.802 1.412409 0.2982366 0.000475 0.000141602 -9.60073E-11 0.949515893 0.050484107
0.3 2.5 6250 3125 2.55455526 0.250143 0.045201 3632.844 1.753656 0.2670477 0.000427 0.000114047 1.01278E-10 0.994683822 0.005316178
0.3 3 7500 3750 3.26556127 0.250178 0.047004 3885.534 2.073729 0.2460188 0.000392 9.64446E-05 8.37744E-12 0.999453763 0.000546237

250mm thick cover = 40mm 1.00038 1.075986 1.148553 1.412409

h/2
model uncertainty (cov) Area (%) Area (mm2) Area/face β(mm2) h* c* fct,eff θ* Srm ε wcalc g(x) probability(s) probability(f)
0.3 0.5 1250 625 solver could not find feasible solution therefore gave 0
0.3 1 2500 1250 solver could not find feasible solution therefore gave 0
0.3 1.43267655 3581.691376 1790.8 0.18357641 0.250022 0.039654 2901.481 1.001274 0.3585297 0.000557 0.000199746 5.95144E-12 0.572827116 0.427172884 min reinf. req.
0.3 1.5 3750 1875 0.39979824 0.250047 0.039782 2965.571 1.057325 0.3462808 0.000546 0.000189157 -1.22998E-11 0.655347435 0.344652565
0.3 2 5000 2500 1.71339766 0.25019 0.040805 3385.107 1.47052 0.2817623 0.000483 0.000136006 -6.49981E-11 0.956680297 0.043319703
0.3 2.5 6250 3125 2.67703137 0.250281 0.041898 3727.549 1.869857 0.2439752 0.000438 0.00010696 3.13578E-11 0.996286116 0.003713884
0.3 3 7500 3750 3.42315354 0.25034 0.043009 4013.709 2.248497 0.2195327 0.000405 8.89483E-05 4.651E-11 0.999690504 0.000309496

Figure B.6: Selected Data of EN 1992 End Restraint Crack Model FORM Analysis (hc,eff = 2.5(c + ϕ/2) and h/2- Effective Depth Comparison)
 182

End Restraint

250mm thick
2.5*(c+ϕ/2)
Elastic Modulus 3 days
2 2
αe = 7 Gross Area (%)Gross Area (mm ) Area/face (mm ) β h* c* fct ,eff θ* Srm ε wcalc g(x) probability (s)probability (f)
7 0.5 1250 625 solver could not find feasible solution therefore gave 0
7 1 2500 1250 solver could not find feasible solution therefore gave 0
7 1.43267655 3582 1791 0.192376749 0.25001159 0.039937687 2901.059074 1.000910714 0.35872512 0.000557023 0.000199818 8.01721E-12 0.57627645 0.4237236 min reinf. req.
7 1.5 3750 1875 0.39691205 0.2500238 0.040353818 2958.374419 1.050939977 0.349261334 0.000544881 0.000190306 -2.65663E-12 0.65428384 0.3457162
7 2 5000 2500 1.640177755 0.25009507 0.04303944 3330.802409 1.412409479 0.298236641 0.000474798 0.000141602 -9.60073E-11 0.94951589 0.0504841
7 2.5 6250 3125 2.554555258 0.250143359 0.045201488 3632.844251 1.753655694 0.267047738 0.000427068 0.000114047 1.01278E-10 0.99468382 0.0053162
7 3 7500 3750 3.26556127 0.250177662 0.047004019 3885.534025 2.073728658 0.246018755 0.000392021 9.64446E-05 8.37744E-12 0.99945376 0.0005462

250mm thick
2.5*(c+ϕ/2)
Elastic Modulus 7 days
2 2
αe = 6 Gross Area (%)Gross Area (mm ) Area/face (mm ) β h* c* fct ,eff θ* Srm ε wcalc g(x) probability (s)probability (f)
6.5 0.5 1250 625
6.5 1 2500 1250
6.5 1.42715822 3567.895547 1783.947773 0.192374776 0.25001167 0.039937627 2901.059267 1.00091088 0.359802884 0.000555354 0.000199818 8.01763E-12 0.57627568 0.4237243 min reinf. req.
6.5 1.5 3750 1875 0.414962421 0.250025052 0.040390744 2963.486151 1.055473004 0.349532123 0.000542121 0.000189489 -2.80245E-11 0.66091529 0.3390847
6.5 2 5000 2500 1.663464977 0.25009727 0.04309161 3338.201438 1.420244525 0.298549661 0.000471683 0.000140821 -9.98417E-11 0.9518903 0.0481097
6.5 2.5 6250 3125 2.582764092 0.250146494 0.0452691 3642.595444 1.765414423 0.267399318 0.000423665 0.000113288 1.24461E-10 0.99509938 0.0049006
6.5 3 7500 3750 3.298407179 0.250181692 0.047087016 3897.655473 2.089894632 0.246406075 0.000388378 9.56986E-05 1.87998E-11 0.99951382 0.0004862

250mm thick 0.990041888 1.010058274

2.5*(c+ϕ/2) 0.989078045 1.011042562
Elastic Modulus 28 days
2 2
αe = 6 Gross Area (%)Gross Area (mm ) Area/face (mm ) β h* c* fct ,eff θ* Srm ε wcalc g(x) probability (s)probability (f)
6 0.5 1250 625 solver could not find feasible solution therefore gave 0
6 1 2500 1250 solver could not find feasible solution therefore gave 0
6 1.42166718 3554.167958 1777.083979 0.192372828 0.25001175 0.039937568 2901.05946 1.000911046 0.360883622 0.000553691 0.000199818 8.01803E-12 0.57627491 0.4237251 min reinf. req.
6 1.5 3750 1875 0.433135362 0.250026328 0.040427963 2968.641306 1.060056374 0.349805061 0.000539355 0.000188669 6.23999E-12 0.66754178 0.3324582
6 2 5000 2500 1.686957883 0.250099521 0.043144319 3345.682071 1.428192391 0.298865912 0.000468562 0.000140037 -6.50273E-11 0.95419427 0.0458057
6 2.5 6250 3125 2.61127902 0.250149714 0.045337568 3652.478414 1.77738011 0.267755353 0.000420254 0.000112525 -3.62227E-11 0.99548979 0.0045102
6 3 7500 3750 3.331666358 0.250185847 0.047171238 3909.966942 2.106391046 0.246799109 0.000384722 9.49491E-05 2.10335E-11 0.99956836 0.0004316

E37/E24 0.99998987
E38/E25 1.043794183
28 to 7 days E39/E26 1.014122874
E40/E27 1.011040469
E41/E28 1.010083406
ave (7 and 28 day) AVERAGE(E44:E48) 1.01580616

E37/E10 0.999979614
28 to 3 days E38/E11 1.091262817
E39/E12 1.028521377
E40/E13 1.022204946
E41/E14 1.020243101
ave (3 and 28 day) AVERAGE(E51:E55) 1.032442371 1.024124266 average

Figure B.7: Effect of Variation in Elastic Modulus of Concrete

 183

Appendix C: Sensitivity Analysis of EN 1992

 184

Edge Restraint
hc,e ff = 2.5(c+ ϕ/2)

Constant Variables
As unkown variable εca=33με R = 0.5 Ω = c.o.v.=[Link]./mean
c = 40mm εcd=220με Final failure point
h = 250mm αT,c=14με ϒxi* = 1-αi*βΩ = x*/mean
wl= 0.3mm
Final direction cosine
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
model uncertainty c.o.v. Area/face (m ) Gross Area (m )As (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.000944393 0.001888787 0.7555146 0.25 0.0472085 1.0916103 0.70018675 0.000393 0.00027482 4.20768E-11 0 0 -0.7864191 0.78641912 -0.6176933 0.61769327 1 1.18021226 1.09161035
0.15 0.000996228 0.001992456 0.7969824 0.25 0.045738 1.1741475 0.65096691 0.000393 0.0002555 5.25105E-12 0 0 -0.6457821 0.64578211 -0.7635218 0.76352175 1 1.14345109 1.17414755
0.2 0.00105844 0.00211688 0.8467518 0.25 0.04461 1.2630059 0.60516827 0.000393 0.00023753 4.23328E-11 0 0 -0.5347951 0.53479514 -0.8449818 0.84498175 1 1.11525036 1.26300591
0.25 0.001128038 0.002256077 0.9024306 0.25 0.0437798 1.3544421 0.56431441 0.000393 0.00022149 8.35648E-12 0 0 -0.4512998 0.45129983 -0.8923724 0.89237238 1 1.09449431 1.3544421
0.3 0.001203377 0.002406755 0.962702 0.25 0.0431633 1.4472941 0.52811044 0.000393 0.00020728 3.13308E-11 0 0 -0.3882692 0.38826925 -0.921546 0.92154598 1 1.0790819 1.44729412

wlim= 0.2mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
model uncertainty c.o.v. Area/face (m ) Gross Area (m )As (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.001536328 0.003072656 1.2290622 0.25 0.0472526 1.090739 0.46716411 0.000393 0.00018336 1.36415E-11 0 0 -0.7905716 0.79057162 -0.6123696 0.61236959 1 1.18131546 1.09073899
0.15 0.001627396 0.003254792 1.3019167 0.25 0.045796 1.172881 0.43444658 0.000393 0.00017052 5.45515E-12 0 0 -0.6514113 0.65141125 -0.7587248 0.75872484 1 1.14490026 1.17288097
0.2 0.001737945 0.00347589 1.390356 0.25 0.0446717 1.2615217 0.40392024 0.000393 0.00015854 5.44181E-12 0 0 -0.5409384 0.54093844 -0.8410622 0.84106219 1 1.11679294 1.26152166
0.25 0.001863172 0.003726345 1.4905379 0.25 0.0438409 1.3528365 0.3766561 0.000393 0.00014784 3.11378E-12 0 0 -0.4575005 0.45750048 -0.8892094 0.88920937 1 1.09602239 1.35283652
0.3 0.002000583 0.004001165 1.600466 0.25 0.0432228 1.4456177 0.35248189 0.000393 0.00013835 3.11973E-11 0 0 -0.3943258 0.39432577 -0.9189707 0.91897072 1 1.08056923 1.4456177

wlim= 0.1mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
model uncertainty c.o.v. Area/face (m ) Gross Area (m )As (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.004116912 0.008233823 3.2935293 0.25 0.0473779 1.0881995 0.23412715 0.000393 9.1895E-05 9.93702E-12 0 0 -0.802367 0.80236704 -0.5968309 0.5968309 1 1.18444823 1.08819949
0.15 0.004441849 0.008883697 3.5534789 0.25 0.0459635 1.1691383 0.21791827 0.000393 8.5533E-05 1.89297E-10 0 0 -0.667602 0.66760202 -0.7445183 0.74451833 1 1.14908635 1.16913829
0.2 0.004855315 0.00971063 3.8842518 0.25 0.044851 1.2570841 0.20267302 0.000393 7.9549E-05 1.7998E-11 0 0 -0.5587797 0.55877969 -0.8293161 0.82931614 1 1.12127483 1.25708407
0.25 0.0053503 0.0107006 4.2802399 0.25 0.0440193 1.3480012 0.18900357 0.000393 7.4184E-05 1.78965E-11 0 0 -0.4756025 0.47560248 -0.8796603 0.87966032 1 1.1004823 1.34800116
0.3 0.005928176 0.011856352 4.7425408 0.25 0.043396 1.4405273 0.17686372 0.000393 6.9419E-05 2.13316E-11 0 0 -0.4121156 0.41211563 -0.9111316 0.91113155 1 1.08490113 1.44052732

wlim= 0.05mm
Figure C.1: EN 1992 Edge Restraint Crack Model Sensitivity Factors and Theoretical Partial
β=1.5 Safety
Direction Cosines/Factors (hc,eff = 2.5(c + ϕ/2)).
Sensitivity Factors Partial Factors
2 2
model uncertainty c.o.v. Area/face (m ) Gross Area (m )As (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.025743027 0.051486054 20.594422 0.25 0.0476011 1.0834415 0.11757765 0.000393 4.6149E-05 9.90309E-12 0 0 -0.8232938 0.8232938 -0.5676155 0.56761547 1 1.19002785 1.08344152
0.15 0.032908533 0.065817067 26.326827 0.25 0.0462691 1.1619141 0.10963676 0.000393 4.3032E-05 1.73404E-11 0 0 -0.6971008 0.69710077 -0.7169732 0.71697316 1 1.15672665 1.16191408
0.2 0.047270777 0.094541554 37.816622 0.25 0.045187 1.2483115 0.10204864 0.000393 4.0054E-05 1.70247E-11 0 0 -0.5919501 0.59195009 -0.8059746 0.80597462 1 1.12967488 1.24831146
0.25 0.08401427 0.16802854 67.211416 0.25 0.0443596 1.3382632 0.09518942 0.000393 3.7362E-05 1.54004E-11 0 0 -0.5097418 0.50974181 -0.8603274 0.86032743 1 1.10898928 1.33826317
0.3 0.331524058 0.663048115 265.21925 0.25 0.0437271 1.4301295 0.08907478 0.000393 3.4962E-05 2.59423E-11 0 0 -0.4460008 0.44600083 -0.8950325 0.89503255 1 1.09317715 1.43012947
 185

Edge Restraint
hc,eff = h/2
As varies εca=33με R=0.5
c=40mm εcd=220με
h=250mm αT,c=14με ϒxi*= 1-αi*βΩ = x*/mean
wlim= 0.3mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
Area/face (m G
model uncertainty c.o.v. ) ross Area (mAs) (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.000862 0.001724 0.68961 0.2503212 0.041211 1.152482 0.6632 0.00039 0.00026 5.61674E-10 -0.08566 0.085662 -0.18259 0.182588 -0.979451 0.979451 1.001285 1.030282 1.152482
0.15 0.0009316 0.0018631 0.74525 0.2502147 0.04073 1.235479 0.61865 0.00039 0.00024 1.29216E-09 -0.05725 0.05725 -0.13033 0.130329 -0.989817 0.989817 1.000859 1.018238 1.235479
0.2 0.0010051 0.0020102 0.80406 0.2501601 0.040491 1.320604 0.57877 0.00039 0.00023 1.75781E-11 -0.0427 0.042703 -0.10427 0.104273 -0.993632 0.993632 1.000641 1.012286 1.320604
0.25 0.0010823 0.0021646 0.86584 0.2501271 0.040351 1.407835 0.54291 0.00039 0.00021 2.38073E-11 -0.03388 0.033883 -0.08878 0.088783 -0.995475 0.995475 1.000508 1.008764 1.407835
0.3 0.0011631 0.0023262 0.93046 0.2501049 0.040258 1.496954 0.51059 0.00039 0.0002 3.97022E-11 -0.02797 0.027967 -0.07857 0.078571 -0.996516 0.996516 1.00042 1.006449 1.496954

wlim= 0.2mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
Area/face (m G
model uncertainty c.o.v. ) ross Area (mAs) (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.0013929 0.0027858 1.1143 0.2502911 0.042056 1.14894 0.4435 0.00039 0.00017 2.7208E-11 -0.07762 0.077616 -0.2728 0.272802 -0.958934 0.958934 1.001164 1.051408 1.14894
0.15 0.0015129 0.0030258 1.21033 0.2501957 0.041334 1.232545 0.41342 0.00039 0.00016 6.72761E-11 -0.05219 0.052188 -0.19581 0.195814 -0.979251 0.979251 1.000783 1.033352 1.232545
0.2 0.001642 0.003284 1.31359 0.2501457 0.040974 1.317922 0.38663 0.00039 0.00015 5.06629E-11 -0.03886 0.038862 -0.15687 0.156872 -0.986854 0.986854 1.000583 1.024338 1.317922
0.25 0.0017796 0.0035592 1.42369 0.2501151 0.04076 1.405242 0.36261 0.00039 0.00014 6.76818E-11 -0.0307 0.030701 -0.13361 0.133614 -0.990558 0.990558 1.000461 1.018991 1.405242
0.3 0.0019257 0.0038514 1.54054 0.2500945 0.040619 1.494361 0.34098 0.00039 0.00013 2.03629E-11 -0.0252 0.025196 -0.11825 0.118249 -0.992664 0.992664 1.000378 1.015475 1.494361

wlim= 0.1mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
Area/face (m G
model uncertainty c.o.v. ) ross Area (mAs) (%) h* c* θ* Sr,m ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.1 0.0036689 0.0073377 2.93509 0.2501952 0.044343 1.131966 0.22507 0.00039 8.8E-05 4.84367E-11 -0.05206 0.052055 -0.50813 0.508127 -0.859708 0.859708 1.000781 1.108578 1.131966
0.15 0.0040635 0.0081269 3.25077 0.2501358 0.043088 1.217339 0.20929 0.00039 8.2E-05 6.79237E-12 -0.03622 0.036219 -0.38049 0.380489 -0.924076 0.924076 1.000543 1.077195 1.217339
0.2 0.0045216 0.0090431 3.61725 0.2501008 0.042404 1.303651 0.19543 0.00039 7.7E-05 1.64703E-11 -0.02689 0.026893 -0.30937 0.309369 -0.950562 0.950562 1.000403 1.060095 1.303651
0.25 0.0050443 0.0100886 4.03544 0.2500783 0.041985 1.391297 0.18312 0.00039 7.2E-05 9.60355E-12 -0.02087 0.020871 -0.26521 0.265211 -0.963964 0.963964 1.000313 1.049614 1.391297
0.3 0.0056386 0.0112773 4.5109 0.2500626 0.041705 1.480355 0.17211 0.00039 6.8E-05 1.8395E-12 -0.01669 0.016689 -0.23547 0.235471 -0.971738 0.971738 1.00025 1.042614 1.480355

Figure C.2: EN 1992 Edge Restraint Crack Model Sensitivity Factors and Theoretical Partial Safety Factors (hc,eff = h/2)
 186

End Restraint
hc,eff = 2.5(c+ ϕ/2)

Constant Variables
As unkown variableαe=7
c=40mm kc=1 Es = 200GPa
h=250mm k=1
wlim= 0.3mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2
Area/face (m
model uncertainty c.o.v. Gross
) Area (m As) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff fct,eff∗ αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.001725 0.00345016 1.380064 0.25014 0.044393 3535.77 1.056444 0.404095 0.000703 0.000284 2.94E-10 -0.03640258 0.0364026 -0.5131462 0.513146 -0.758843 0.75884295 -0.399391 0.399391 1.001 1.11 1.219 0.82 1.056
0.15 0.001755 0.0035109 1.40436 0.25012 0.043947 3470.063 1.118349 0.395207 0.000679 0.000268 3.89E-10 -0.0331952 0.0331952 -0.4682618 0.468262 -0.693025 0.6930248 -0.5471248 0.5471248 1 1.099 1.197 0.836 1.118
0.2 0.001794 0.00358802 1.435208 0.25011 0.043487 3402.701 1.193785 0.385117 0.000653 0.000251 3.55E-10 -0.02984328 0.0298433 -0.4214923 0.421492 -0.624242 0.62424172 -0.6570965 0.6570965 1 1.087 1.173 0.852 1.194
0.25 0.001839 0.00367707 1.470828 0.2501 0.043063 3340.946 1.27769 0.374741 0.000627 0.000235 3.92E-10 -0.02671155 0.0267115 -0.3779204 0.37792 -0.559976 0.55997593 -0.7368104 0.7368104 1 1.077 1.152 0.868 1.278
0.3 0.001887 0.00377454 1.509817 0.25009 0.042692 3287.242 1.366807 0.36458 0.000602 0.000219 2.7E-10 -0.02394094 0.0239409 -0.3394699 0.33947 -0.503116 0.50311599 -0.7943937 0.7943937 1 1.067 1.134 0.882 1.367

wlim= 0.2mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2 fct,eff∗
Area/face (m
model uncertainty c.o.v. Gross
) Area (m
As) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.002202 0.00440483 1.761931 0.25013 0.04443 3533.836 1.056284 0.335998 0.000564 0.000189 3.39E-10 -0.03544735 0.0354474 -0.5168192 0.516819 -0.756924 0.75692402 -0.3983811 0.3983811 1.001 1.111 1.219 0.821 1.056
0.15 0.002243 0.00448537 1.794146 0.25012 0.043983 3468.591 1.118053 0.328673 0.000544 0.000179 3.48E-10 -0.03232165 0.0323217 -0.4718829 0.471883 -0.691536 0.69153553 -0.5459491 0.5459491 1 1.1 1.196 0.836 1.118
0.2 0.002294 0.00458782 1.835128 0.25011 0.043521 3401.623 1.193366 0.320363 0.000523 0.000168 3.86E-10 -0.02905168 0.0290517 -0.4250025 0.425003 -0.623129 0.62312944 -0.6559257 0.6559257 1 1.088 1.173 0.853 1.193
0.25 0.002353 0.00470637 1.88255 0.2501 0.043095 3340.163 1.277172 0.311823 0.000502 0.000157 8.44E-11 -0.02599364 0.0259936 -0.381278 0.381278 -0.559154 0.55915437 -0.7357294 0.7357294 1 1.077 1.152 0.868 1.277
0.3 0.002418 0.00483645 1.934582 0.25009 0.042723 3286.67 1.366215 0.303467 0.000482 0.000146 2.95E-10 -0.02328613 0.0232861 -0.342656 0.342656 -0.502506 0.5025061 -0.7934307 0.7934307 1 1.068 1.133 0.882 1.366

wlim= 0.1mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
2 2 fct,eff∗
Area/face (m
model uncertainty c.o.v. Gross
) Area (m
As) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.003439 0.00687872 2.751488 0.25012 0.044507 3529.727 1.055944 0.247494 0.000383 9.47E-05 6.6E-10 -0.03321025 0.0332103 -0.5245255 0.524525 -0.752841 0.75284134 -0.3962323 0.3962323 1 1.113 1.217 0.822 1.056
0.15 0.003508 0.00701636 2.806543 0.25011 0.044058 3465.456 1.117423 0.242209 0.000369 8.95E-05 3.17E-10 -0.03026924 0.0302692 -0.479488 0.479488 -0.688363 0.68836278 -0.5434443 0.5434443 1 1.101 1.195 0.837 1.117
0.2 0.003596 0.00719229 2.876917 0.2501 0.043594 3399.324 1.192472 0.236219 0.000355 8.39E-05 6.57E-10 -0.02718559 0.0271856 -0.4323804 0.43238 -0.620757 0.62075705 -0.6534285 0.6534285 1 1.09 1.172 0.853 1.192
0.25 0.003698 0.00739694 2.958776 0.25009 0.043164 3338.494 1.276068 0.230074 0.000341 7.84E-05 3.01E-10 -0.02429598 0.024296 -0.3883382 0.388338 -0.557401 0.55740064 -0.7334219 0.7334219 1 1.079 1.151 0.869 1.276
0.3 0.003811 0.00762281 3.049122 0.25008 0.042787 3285.451 1.364951 0.224072 0.000327 7.33E-05 3.87E-10 -0.0217333 0.0217333 -0.3493554 0.349355 -0.501204 0.50120373 -0.7913743 0.7913743 1 1.07 1.133 0.883 1.365

wlim= 0.05mm
Figure
β=1.5 C.3: EN 1992 End Restraint Crack Model Sensitivity Factors and Theoretical
Direction Cosines/ Partial Safety Factors (hc,eff = 2.5(c + ϕ/2)).
Sensitivity Factors Partial Factors
2 2
Area/face (m
model uncertainty c.o.v. Gross
) Area (mAs) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff fct,eff∗ αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.005682 0.01136314 4.545255 0.25011 0.044604 3524.482 1.055509 0.185314 0.000256 4.74E-05 4.66E-10 -0.02984085 0.0298409 -0.5341701 0.53417 -0.747623 0.74762334 -0.393486 0.393486 1 1.115 1.215 0.823 1.056
0.15 0.005813 0.01162504 4.650018 0.2501 0.044153 3461.445 1.116617 0.181472 0.000247 4.48E-05 4.51E-11 -0.02716309 0.0271631 -0.48902 0.48902 -0.6843 0.68429966 -0.5402366 0.5402366 1 1.104 1.194 0.838 1.117
0.2 0.005981 0.01196249 4.784994 0.25009 0.043685 3396.377 1.191327 0.177124 0.000237 4.2E-05 2.72E-10 -0.02434727 0.0243473 -0.4416374 0.441637 -0.617714 0.61771405 -0.6502253 0.6502253 1 1.092 1.171 0.854 1.191
0.25 0.006179 0.01235858 4.943431 0.25008 0.04325 3336.353 1.274651 0.172675 0.000227 3.92E-05 2.32E-10 -0.02170166 0.0217017 -0.3972003 0.3972 -0.555149 0.55514896 -0.7304592 0.7304592 1 1.081 1.15 0.869 1.275
0.3 0.0064 0.01280012 5.120047 0.25007 0.042868 3283.885 1.36333 0.168342 0.000218 3.67E-05 2.72E-10 -0.01935028 0.0193503 -0.3577629 0.357763 -0.499531 0.49953102 -0.7887332 0.7887332 1 1.072 1.132 0.883 1.363
 187

End Restraint

αe=7
As unkown variable
c=40mm kc=1 Es = 200GPa
h=250mm k=1
hc,eff=0.5h
wlim= 0.3mm
β=1.5 Direction Cosines/ Sensitivity Factors fct,eff
Partial Factors
2 2
model uncertainty Gross
c.o.v. (m
Area/face As) (%) h*
) Area (m c* fct,eff* θ* Srm ε wcalc g(x) αh αc c
αfct,eff αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.001684 0.0033678 1.34713 0.2503 0.0409 3652.05 1.06596 0.37907 0.00074 0.00028 3E-11 -0.07788 0.077878 -0.14861 0.14861 -0.8724 0.872374 -0.459144 0.459144 1.0011682 1.022435 1.259326 0.794076 1.065956
0.15 0.001718 0.0034366 1.37464 0.2503 0.04076 3553.02 1.13494 0.37281 0.00071 0.00026 3E-10 -0.06909 0.069092 -0.133954 0.13395 -0.7759 0.775916 -0.612566 0.612566 1.0010364 1.019069 1.225178 0.816208 1.134937
0.2 0.001761 0.0035218 1.40873 0.2502 0.04063 3459.68 1.21596 0.36547 0.00068 0.00025 3E-10 -0.06058 0.060579 -0.119815 0.11981 -0.6825 0.682514 -0.718436 0.718436 1.0009087 1.015832 1.192994 0.838227 1.215956
0.25 0.001809 0.0036182 1.44729 0.2502 0.0405 3380.1 1.30373 0.3576 0.00064 0.00023 2E-11 -0.05335 0.053354 -0.10522 0.10522 -0.6009 0.600858 -0.790602 0.790602 1.0008003 1.012502 1.165552 0.857963 1.303725
0.3 0.001861 0.0037221 1.48885 0.2502 0.04043 3314.35 1.39508 0.3497 0.00061 0.00022 5E-11 -0.04685 0.046847 -0.097092 0.09709 -0.5319 0.531934 -0.839896 0.839896 1.0007027 1.010652 1.14288 0.874983 1.395082

wlim= 0.2mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
fct,eff∗
2 2
model uncertainty c.o.v. (m
Area/face ) Area (m
Gross As) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c
αfct,eff αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.002148 0.004296 1.7184 0.2503 0.04118 3647.63 1.0656 0.31539 0.0006 0.00019 3E-10 -0.07447 0.074467 -0.178965 0.17897 -0.8681 0.868134 -0.456913 0.456913 1.001117 1.029442 1.257805 0.795036 1.065599
0.15 0.002193 0.0043869 1.75475 0.2502 0.04101 3549.92 1.13432 0.31021 0.00057 0.00018 3E-10 -0.06609 0.066086 -0.161345 0.16135 -0.7729 0.772861 -0.610153 0.610153 1.0009913 1.025369 1.224111 0.816919 1.134321
0.2 0.00225 0.0044998 1.79992 0.2502 0.04086 3457.57 1.21513 0.30414 0.00054 0.00016 2E-10 -0.05794 0.057943 -0.144316 0.14432 -0.6804 0.68037 -0.716179 0.716179 1.0008692 1.021448 1.192266 0.838739 1.215133
0.25 0.002314 0.0046279 1.85115 0.2502 0.04072 3378.51 1.30266 0.29769 0.00052 0.00015 1E-10 -0.05079 0.050792 -0.12942 0.12942 -0.5992 0.599207 -0.78843 0.78843 1.0007619 1.01803 1.165003 0.858367 1.302664
0.3 0.002383 0.0047662 1.90649 0.2502 0.04061 3313.32 1.394 0.29116 0.00049 0.00014 2E-10 -0.04477 0.044768 -0.116895 0.11689 -0.5308 0.530844 -0.838174 0.838174 1.0006715 1.015165 1.142525 0.875255 1.394002

wlim= 0.1mm
β=1.5 Direction Cosines/ Sensitivity Factors Partial Factors
fct,eff∗
2 2
model uncertainty Gross
c.o.v. (m
Area/face As) (%) h*
) Area (m c* fct,eff* θ* Srm ε wcalc g(x) αh αc c
αfct,eff αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.1 0.003349 0.0066985 2.6794 0.2503 0.04177 3635.54 1.06462 0.23298 0.0004 9.4E-05 2E-10 -0.06681 0.066811 -0.242457 0.24246 -0.8565 0.856476 -0.450777 0.450777 1.0010022 1.044254 1.253633 0.797682 1.064618
0.15 0.003426 0.0068523 2.74094 0.2502 0.04155 3541.41 1.13263 0.22916 0.00039 8.8E-05 8E-11 -0.05935 0.059353 -0.218839 0.21884 -0.7644 0.76444 -0.603505 0.603505 1.0008903 1.038719 1.221177 0.818882 1.132626
0.2 0.003522 0.0070447 2.81789 0.2502 0.04133 3451.75 1.21287 0.22473 0.00037 8.2E-05 4E-10 -0.05205 0.052052 -0.195873 0.19587 -0.6745 0.674456 -0.709953 0.709953 1.0007808 1.033366 1.190258 0.840154 1.212866
0.25 0.003632 0.0072644 2.90577 0.2502 0.04115 3374.49 1.29998 0.22005 0.00035 7.7E-05 5E-10 -0.0456 0.045601 -0.175686 0.17569 -0.595 0.595033 -0.782938 0.782938 1.000684 1.028683 1.163618 0.859388 1.299984
0.3 0.003752 0.0075034 3.00135 0.2502 0.04099 3310.49 1.39103 0.21533 0.00033 7.2E-05 7E-10 -0.04014 0.04014 -0.158648 0.15865 -0.5278 0.527839 -0.83343 0.83343 1.0006021 1.024747 1.141547 0.876004 1.391029

Figure C.4: EN 1992 End Restraint Crack Model Sensitivity Factors and Theoretical Partial Safety Factors (hc,eff = h/2)
 188

Edge Restraint
hc,eff = 2.5(c+ ϕ/2)

Constant Variables
As unkown variableεca=33με R = 0.5 Ω = c.o.v.=[Link]./mean
c=40mm εcd=220με Final failure point
h=250mm αT,c=14με ϒxi*= 1-αi*βΩ = x*/mean
Wlim = 0.2 mm Final direction cosine

β Area/face (m2) Gross Area (m2) As (%) h* c* θ* Srm ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ

0.5 0.001325177 0.002650353 1.060141 0.25 0.04072 1.09786 0.46413 0.0004 0.0001822 3.42E-11 0 0 -0.386219 0.386219 -0.922407 0.922407 1 1.0178775 1.09786
1.5 0.002000583 0.004001165 1.600466 0.25 0.04322 1.44562 0.35248 0.0004 0.0001383 3.12E-11 0 0 -0.394326 0.394326 -0.918971 0.918971 1 1.0805692 1.44562
2 0.002500293 0.005000586 2.000234 0.25 0.04458 1.65733 0.30746 0.0004 0.0001207 6.955E-11 0 0 -0.398861 0.398861 -0.917011 0.917011 1 1.1145027 1.65733

End Restraint
hc,eff = 2.5(c+ ϕ/2)

Constant Variables
As unkown variableαe=7
c=40mm kc=1 Es = 200GPa
h=250mm k=1
Wlim = 0.2 mm Direction Cosines/ Sensitivity Factors Partial Factors
β Area/face (m2) Gross Area (m2) As (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff fct,eff∗ αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.5 0.001919165 0.003838329 1.535332 0.25 0.04057 2987.67 1.07708 0.3446 0.0005388 0.0001857 9.9E-11 -0.023926 0.0239264 -0.337406 0.3374063 -0.503513 0.503513 -0.795021 0.79502 1 1.014 1.03 0.971 1.077
1.5 0.002418227 0.004836454 1.934582 0.25 0.04272 3286.67 1.36621 0.3035 0.0004824 0.0001464 2.95E-10 -0.023286 0.0232861 -0.342656 0.342656 -0.502506 0.502506 -0.793431 0.79343 1 1.068 1.133 0.882 1.366
2 0.00272579 0.00545158 2.180632 0.25 0.04387 3446.69 1.53812 0.2854 0.0004556 0.00013 1.81E-10 -0.022912 0.0229122 -0.345354 0.3453541 -0.501982 0.501982 -0.792603 0.7926 1 1.097 1.189 0.841 1.538

Figure C.5: EN 1992 Edge and End Restraint Crack Model Sensitivity Factors and Theoretical Partial Safety Factors with Varying β Values (hc,eff =
2.5(c + ϕ/2)
 189

Edge Restraint
hc,eff = 0.5h

Constant Variables
εca=33με R = 0.5
As unkown variable Ω = c.o.v.=[Link]./mean
c=40mm εcd=220με Final failure point
h=250mm αT,c=14με ϒxi*= 1-αi*βΩ = x*/mean
Wlim = 0.2 mm Final direction cosine

β Area/face (m2Gross
) 2
Area (mAs
) (%) h* c* θ* Srm ε wcalc g(x) αh αc c αθ ϒh ϒc ϒθ
0.5 0.00131795 0.00263591 1.054363 0.250034 0.03981 1.110025 0.459047 0.000393 0.00018 2.4322E-11 -0.027434739 0.02743474 -0.086365 0.0863653 -0.995886 0.9958857 1.00013717 0.9952388 1.1100254
1.5 0.00192568 0.00385136 1.540545 0.250094 0.040619 1.494361 0.340985 0.000393 0.000134 2.0363E-11 -0.025196033 0.02519603 -0.118249 0.1182488 -0.992664 0.9926643 1.00037794 1.0154745 1.4943614
2 0.00236177 0.00472354 1.889415 0.250119 0.041233 1.731569 0.294273 0.000393 0.000116 1.5176E-10 -0.023741961 0.02374196 -0.138725 0.1387247 -0.990046 0.9900464 1.00047484 1.0308332 1.7315688

End Restraint

Constant Variables
αe=7
As unkown variable
c=40mm kc=1 Es = 200GPa
h=250mm k=1
Wlim = 0.2 mm Direction Cosines/ Sensitivity Factors Partial Factors
β Area/face (m2Gross
) 2
Area (mAs
) (%) h* c* fct,eff* θ* Srm ε wcalc g(x) αh αc c αfct,eff fct,eff∗ αθ ϒh ϒc ϒfct,eff 1/ϒfct,eff ϒθ
0.5 0.00191557 0.00383113 1.532452 0.250058 0.039845 2995.734 1.077079 0.34077 0.000541 0.00018444 3.81648E-10 -0.0467186 0.0467186 -0.098195 0.0981945 -0.531879 0.53187923 -0.839809 0.8398093 1.0002336 0.99612 1.033012 0.968043 1.077079
1.5 0.00238312 0.00476623 1.906494 0.250168 0.040607 3313.322 1.394002 0.291163 0.000493 0.00014347 1.82688E-10 -0.0447676 0.0447676 -0.116895 0.1168945 -0.530844 0.53084381 -0.838174 0.8381744 1.0006715 1.01517 1.142525 0.875255 1.394002
2 0.00266823 0.00533646 2.134582 0.250218 0.041095 3483.805 1.579748 0.269744 0.000469 0.0001266 2.63001E-10 -0.0436424 0.0436424 -0.127532 0.1275321 -0.530169 0.5301691 -0.837109 0.8371091 1.0008729 1.02738 1.201312 0.832423 1.579748

Figure C.6: EN 1992 Edge and End Restraint Crack Model Sensitivity Factors and Theoretical Partial Safety Factors with Varying β Values
(hc,eff = h/2)