Dynamic Pile Testing Technology:

Validation and Implementation

Robert Y. Liang, Ph.D., P.E.

Luo Yang, Ph.D., P.E.

for the
Ohio Department of Transportation
Office of Research and Development

and the
U.S. Department of Transportation
Federal Highway Administration

State Job Number 14794

Final Report
FHWA/OH-2007-08
May 2007
1. Report No. 2. Government Accession No. 3. Recipient’s Catalog No.

FHWA/OH-2007-08
4. Title and subtitle 5. Report Date

May 2007
Dynamic Pile Testing Technology: Validation and Implementation 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report

No.

Dr. Robert Liang

Luo Yang 10. Work Unit No. (TRAIS)

9. Performing Organization Name and Address 11. Contract or Grant No.

University of Akron 147940

13. Type of Report and Period
Department of Civil Engineering Covered

Akron, OH 44325
12. Sponsoring Agency Name and Address 14. Sponsoring Agency Code
Ohio Department of Transportation
1980 West Broad Street
Columbus, OH 43223
15. Supplementary Notes

16. Abstract
Driven piles are widely used as foundations to support buildings, bridges, and other structures. In 2007, AASHTO
has adopted LRFD method for foundation design. The probability based LRFD approach affords the mathematical
framework from which significant improvements on the design and quality control of driven piles can be achieved.
In this research, reliability-based quality control criteria for driven piles are developed based on the framework of
acceptance-sampling analysis for both static and dynamic test methods with the lognormal distribution
characteristics. As a result, an optimum approach is suggested for the number of load tests and the required
measured capacities for quality control of driven piles. Furthermore, this research has compiled a large database of
pile set-up, from which the reliability-based approach of FORM is employed to develop separate resistance factors
for the measured reference (initial) capacity and predicted set-up capacity. This report also provides a Bayesian
theory based approach to allow for combining the information from the static pile capacity calculation and dynamic
pile testing data to improve pile design process. Specifically, the results from dynamic pile tests can be utilized to
reduce the uncertainties associated with static analysis methods of pile capacity by updating the corresponding
resistance factors. This research has also developed one-dimensional wave equation based algorithm to interpret the
High Strain Testing (HST) data for the estimation of the shaft and toe resistance of driven piles. The closed form
solution is obtained for determining the Smith damping factor and the static soil resistance. Finally, a set of new
wireless dynamic testing equipment (both hardware and software) is developed for more efficient dynamic pile
testing.
17. Key Words 18. Distribution Statement
LRFD, Driven piles set-up, Wireless dynamic piles No restrictions. This document is available to
testing the public through the National Technical
Information Service, Springfield, Virginia 22161
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified
Form DOT F 1700.7 (8-72) Reproduction of completed pages authorized
DYNAMIC PILE TESTING TECHNOLOGY: VALIDATION

AND IMPLEMENTATION

By

Robert Y. Liang
Department of Civil Engineering
The University of Akron
Akron, Ohio 44325-3905

Submitted to
The Ohio Department of Transportation
and the U.S. Department of Transportation
Federal Highway Administration
 Dynamic Pile Testing Technology:
Validation and Implementation

Robert Y. Liang, Ph.D., P.E.

Luo Yang, Ph.D., P.E.

431 ASEC, Civil Engineering Department

University of Akron
Akron, OH, 44325

Credit Reference: Prepared in cooperation with the Ohio Department of Transportation and the
U.S. Department of Transportation, Federal Highway Administration.

Disclaimer Statement: The contents of this report reflect the views of the authors who are
responsible for the facts and accuracy of the data presented herein. The contents do not
necessarily reflect the official views or policies of the Ohio Department of Transportation or the
Federal Highway Administration. This report does not constitute a standard, specification or
regulation.

May 2007
 ABSTRACT

Driven piles are widely used as foundations for buildings, bridges, and other

structures. Since 1994, AASHTO has been in process to change from ASD method to

LRFD method for foundation design. The adoption of LRFD approach makes possible

the application of reliability analysis to quantify uncertainties associated with various

load and resistance components, respectively. A successful application of the probability

approach will definitely result in significant improvements on the design and quality

control of driven piles. Therefore, there is a need to develop the quality control criterion

and to improve the LRFD design of driven piles in the framework of reliability-based

analysis.

Reliability-based quality control criteria on driven piles are developed based on the

framework of acceptance-sampling analysis for both static and dynamic test methods

with lognormal statistical characteristics. The lognormal distribution of statistical data of

various load test methods is converted to standard normal distribution for evaluating the

confidence interval to assure the structure reliability. An optimum approach is suggested

for the selection of the number of load tests and the required measured capacities for

quality control of various load test methods of driven piles.

The databases containing a large number of pile testing data are compiled for piles

driven into clay and into sand, respectively. Based on the compiled databases, the set-up

effect is statistically analyzed. The reliability-based approach of FORM is employed to

i
develop separate resistance factors to account for different degrees of uncertainties

associated with the measured reference capacity and predicted set-up capacity in the

LRFD of driven piles.

Bayesian theory is employed to combine the information from the static calculation

and dynamic pile testing in the reliability-based LRFD of driven piles. Specifically, the

results from dynamic pile tests are incorporated to reduce the uncertainties associated

with static analysis methods by updating the resistance factors.

A one-dimensional wave equation based algorithm to interpret the High Strain Testing

(HST) data for estimation of the shaft and toe resistance of driven piles is proposed. Both

the force and velocity time histories measured at the pile head are used as input boundary

conditions in the new algorithm. The Smith damping factor and the static soil resistance

can be directly determined based on the closed form solution of the one-dimensional

wave equation.

A set of new wireless dynamic testing equipment is developed for dynamic pile testing.

This new technology utilizes remote data acquisition and management unit to build the

wireless connection between the sensors and the monitoring PC. The new wireless

dynamic testing equipment is anticipated to be successfully applied in the field testing

that will advance pile driving control and monitoring to a more secure environment.

ii
 TABLE OF CONTENTS

Page

LIST OF TABLES …………………………………………………………………… ..vii

LIST OF FIGURES …………………………………………………………………...…ix

CHAPTER

I. INTRODUCTION......................................................................................................1

1.1 Statement of the Problem..................................................................... 1

1.2 Objectives of the Study........................................................................ 8

1.3 Organization of Report ...................................................................... 10

II. LITERATURE REVIEW ....................................................................................... 12

2.1 Introduction........................................................................................ 12

2.2 Quality Control on Driven Piles ........................................................ 12

2.3 Reliability Analysis in LRFD ............................................................ 18

2.4 Set-up Effect ...................................................................................... 22

2.5 Current Wireless Applications in Dynamic Pile Testing ................... 27

2.6 Summary ............................................................................................ 31

III. QUALITY CONTROL METHOD FOR PILE DRIVING.………………………33

3.1 Introduction………………………………………………………….33

3.2 Distribution Function of Static and Dynamic Pile Test Results ........ 35

iii
 3.3 Determination of Number and Criterion of Load Tests……………..36

3.4 Target Reliability Index vs. t ............................................................ 40

3.5 Number of Load Tests vs. t ………………………………………...41

3.6 Correction Factor for Finite Population ............................................. 42

3.7 Quality Control Criteria ..................................................................... 43

3.8 Conclusions ........................................................................................ 45

IV. INCORPORATING SET-UP INTO RELIABILITY-BASED

DESIGN OF DRIVEN PILES IN CLAY ............................................................ 55

4.1 Introduction........................................................................................ 55

4.2 Statistical Study of Pile Set-up........................................................... 57

4.3 Correlation between Reference Capacity Q0 and Set-up

Capacity Qset-up ................................................................................... 61

4.4 First-Order Reliability Method (FORM) ........................................... 63

4.5 Calibrate Partial Load and Resistance Factors Based on FORM....... 66

4.6 Practical Load and Resistance Factors for Piles Driven into Clay..... 69

4.7 Proposed Procedure to Incorporate Set-up into LRFD of

Driven Piles in Clay ........................................................................... 72

4.8 Conclusions......................................................................................... 73

V. INCORPORATING SET-UP INTO RELIABILITY-BASED

DESIGN OF DRIVEN PILES IN SAND ........................................................... 88

5.1 Introduction........................................................................................ 88

5.2 A Compiled Database of Case Histories of Driven Piles in Sand...... 89

5.3 Empirical Relationships ..................................................................... 91

iv
 5.4 Statistical Study of Pile Set-up........................................................... 95

5.5 Calibrate Parital Load and Resistance Factors Based on FORM....... 97

5.6 Practical Load and Resistance Factors for Piles Driven into Sand .... 99

5.7 Proposed Procedure to Incorporate Set-up into LRFD of

Driven Piles in Sand......................................................................... 101

5.8 Conclusions ...................................................................................... 103

VI. BAYESIAN BASED FORMALISM FOR UPDATING LRFD

RESISTANCE FACTORS FOR DRIVEN PILES .......................................... 117

6.1 Introduction..................................................................................... 117

6.2 Characteristics of LRFD Method..................................................... 119

6.3 First Order Reliability Method (FORM).......................................... 119

6.4 Uncertainty of Static Analysis Methods .......................................... 120

6.5 Uncertainty of Dynamic Pile Testing Methods................................ 121

6.6 Improving Static Method with Dynamic Test Results..................... 122

6.7 Combination of Static Analysis and Dynamic Pile

Testing Method ................................................................................ 124

6.8 Updated Resistance Factors versus RDS and Number

of Dynamic Pile Tests...................................................................... 125

6.9 Case Study I ..................................................................................... 127

6.10 Case Study II.................................................................................... 128

6.11 Conclusions ...................................................................................... 130

VII. WAVE EQUATION TECHNIQUE FOR ESTIMATING

DRIVEN PILE CAPACITY............................................................................. 146

v
 7.1 Introduction...................................................................................... 146

7.2 Soil-Pile Interaction Model .............................................................. 148

7.3 Identification of Shaft Parameters.................................................... 149

7.4 Identification of Pile Capacity at the Pile Toe ................................. 154

7.5 Case Study........................................................................................ 156

7.6 Comparisons .................................................................................... 156

7.7 Conclusions...................................................................................... 157

VIII. NEW MEASUREMENT AND REMOTE SENSING TECHNOLOGY..... 162

8.1 Introduction...................................................................................... 162

8.2 Hardware .......................................................................................... 163

8.3 Software ........................................................................................... 166

8.4 System Installation and Configuration............................................. 169

8.5 NetDaq Client/Server Architecture .................................................. 174

8.6 Trouble Shooting ............................................................................. 178

8.7 Other Errors ..................................................................................... 183

8.8 Laboratory Testing of the System.................................................... 184

8.9 Conclusions...................................................................................... 186

IX. SUMMARIES AND CONCLUSIONS ................................................................ 188

9.1 Summaries and Conclusions ........................................................... 188

9.2 Recommendations for Implementations ......................................... 192

9.3 Recommendations for Future Research .......................................... 193

REFERENCES ............................................................................................................. 197

vi
 LIST OF TABLES

Table Page

3-1 Distribution Parameters of Various Dynamic Test Methods

for Driven Piles………………………………………………………… 47

3-2 Recommended Number of Load Tests to Be Conducted for Quality

Control of Driven Piles (Reliability β=2.33)………….………….…….. 48

3-3 Recommended Number of Load Tests to Be Conducted for Quality

Control of Driven Piles (Reliability β=3.00)………....………….……... 49

4-1 Summary of Load Test Database for Driven Piles in Clay…………….. 76

4-2 Statistical Details and Resistance Factors of CAPWAP

EOD and BOR…………………………………………………………. 77

4-3 Probabilistic Characteristics of Random Variables of

Loads and Resistances for Piles Driven into Clay……………………. 78

4-4 Recommended LRFD Partial Safety Factors for

Driven Piles in Clay ………………………………………………….. 79

5-1 Summary of Load Test Database for Driven Piles in Sand……………. 105

5-2 Empirical Formulas for Predicting Pile Capacity with 106

Time…………...

5-3 Probabilistic Characteristics of Random Variables of Loads and

Resistances for Piles Driven into Sand………………………………… 107

vii
5-4 Recommended Partial Safety Factors (i.e. load and resistance factors)

for Driven Piles in Sand……………………………………………….. 108

6-1 Statistical Parameters and Resistance Factors for Driven Piles from

NCHRP Report 507 (β = 2.33) (Paikowsky et. Al. 2004)...…………… 132

6-2 Summary of Calculated Capacities from Static Analysis Method for

Test Piles #2……………………………………………...…………… 133

6-3 Pile Capacities of Prior Static Analysis, Dynamic Testing, Posterior

Static Analysis, and Static Load Test (Test Pile #2 in 134

Newbury)………

6-4 Summary of Calculated Capacities from Static Analysis Method for

Test Piles #3……………………………………………...…………… 135

6-5 Pile Capacities of Prior Static Analysis, Dynamic Testing, Posterior

Static Analysis, and Static Load Test (test pile #3 in Newbury)………. 136

7-1 Comparison of Results from Different Methods ……………………… 158

8-1 List of Commands from Client………………………………………… 176

8-2 List of Commands from 178

Server…………………………………………

viii
 LIST OF FIGURES

Figure Page

3-1 Tail Area of the Normal and Lognormal Distributions……………....... 50

3-2 Reliability Index and t with Different n……………………………..... 51

3-3 Relationship between t and Number of Load Tests at Target

Reliability Index β = 2.33 ……............................................................ 52

3-4 Relationship between t and Number of Load Tests at Target

Reliability Index β = 3.00……………………………………………… 53

3-5 Effect of Finite Population on Different Total Numbers

of Load 54
Tests …………………………………………………………...

4-1 Variation of Normalized Capacity with Time for

Driven Piles in Clay……………………………………………………. 80

4-2 Frequency Distribution of the Ratio of Measured Set-up

Capacities to the 81
Predicted ……………………………………………..

4-3 Performance Space in Reduced Coordinates…….. …………………… 82

4-4 Relationship between Partial Safety Factor and Target

Reliability Index When Incorporating Set-up into

the Design of Driven Piles………………………………………........... 83

ix
4-5 Relationship between Partial Safety Factor and Target

Reliability Index without Considering Set-up in the

Design of Driven Piles ……………………………………………….... 84

4-6 Comparison between Considering and Not Considering Set-up 85

Effect...

4-7 Variation of Resistance Factor for Qset-up in Clay versus the Ratio

of Dead Load to Live Load…………………………………………….. 86

4-8 Variation of the F.S versus the Ratio of Dead Load to Live

Load for Piles Driven into Clay………………………………………... 87

5-1 Case Histories of Pile Set-up in the Compiled Database………………. 109

5-2 Variation of Normalized Capacity with Time Based on

Skov and Denver’s Equation…………………………………………... 110

5-3 Variation of Normalized Capacity with Time Based on

Svinkin et al.’s Equation……………………………………………….. 111

5-4 Frequency Distribution of the Ratio of Measured Set-up Capacities

to the Predicted, Based on Skov and Denver’s Equation……………… 112

5-5 Kolmogorov-Smirnov Test for Assumed Lognormal Distribution

of the Ratio of the Measured to the Predicted Set-up Capacity……….. 113

5-6 Relationship between Partial Safety Factor and

Target Reliability Index………………………………………………... 114

5-7 Variation of Resistance Factor for Qset-up in Sand versus the Ratio

of Dead Load to Live Load ……………………………………………. 115

5-8 Variation of the F.S versus the Ratio of Dead Load to Live Load For

x
 Piles Driven into Sand…………………………………………………. 116

6-1 Prior and Posterior Distribution of Different Static Design

Methods When CAPWAP BOR Is Used for Updating………………... 137

6-2 Prior and Posterior Distribution of Static λ-Method when

Different Dynamic Pile Tests Are Used for Updating………………… 138

6-3 Prior and Updated Distribution of Static λ-Method with

Different RDS………………………………………………………….. 139

6-4 Prior and Updated Distribution of Static λ-Method with

Different Numbers of Dynamic Tests…………………………………. 140

6-5 Variation of Updated Resistance Factors with RDS…………………... 141

6-6 Variation of Updated Resistance Factors with Number

of Dynamic Tests………………………………………………………. 142

6-7 Soil Profile in Newbury Site…………………………………………… 143

6-8 Comparison of Ratio of Static Load Test Capacity to Factored

Prediction Capacity for Test Pile #2..………………………………….. 144

6-9 Comparison of Ratio of Static Load Test Capacity to Factored

Prediction Capacity for Test Pile #3…………………………………… 145

7-1 Soil-Pile Interaction Model………….………………………………… 159

7-2 Measured Force and Velocity at the Pile Head……………………… 160

7-3 CAPWAP Results vs. the Results Calculated from

the Proposed Method…………………………………………………... 161

8-1 New Wireless Dynamic Pile Testing System………………………….. 164

8-2 Block Diagram of Wireless Data Acquisition 165

xi
 System………………….

8-3 User Interface of the Server……………………………………………. 179

8-4 Control of User Interface of NetDaq Client 182

8-5 User Interface of NetDaq Client……………………………………….. 183

8-6 Installation of Strain Gages and Accelerometers………………………. 184

xii
 CHAPTER I

INTRODUCTION

1.1 Statement of the Problem

Deep foundations have been frequently used to support structures, such as buildings,

bridges, towers, and dams, in areas where the soil conditions are unfavorable for shallow

foundations. Two basic types of deep foundation are well known as drilled shafts and

driven piles. Drilled shafts are usually installed by boring of the ground and filling the

void with concrete. Driven piles are usually installed by using the hammer to drive the

teel or concrete piles into the ground. This report is concentrated on the improvement of

design and quality control of driven piles.

Traditionally, the design of pile foundations is based on static analysis methods, such

as α-method, β-method, λ-method, Nordlund’s method, Meyerhof’s method, SPT

(Standard Penetration Test) method, and CPT (Cone Penetration Test) method. However,

the pile capacity estimated from static analysis based on the soil parameters obtained

from the laboratory test and in-situ test may not be accurate. It can vary widely due to

inherent heterogeneous nature of soil deposit, and different soil testing and evaluation

methods. Consequently, some type of proof of load-carrying capacity of driven piles is

needed in the field.

1
 Static load test can serve as the ultimate verification of driven pile capacity, despite

the unsettled question concerning definition of failure load. Nevertheless, the static load

test is expensive and time-consuming to carry out and therefore can only be performed on

a few selected driven piles, often one or two piles at most. The information obtained from

static load test results cannot be easily extrapolated to other piles at the site, leaving those

other piles to be unverified. Besides, the information gathered from the static load tests

cannot help field engineer to decide when to stop pile driving. Estimation of pile capacity

based on observation during pile driving is therefore an essential ingredient in a sound

pile driving practice.

With the advent of digital computers, dynamic analysis of pile driving problem

became possible. In 1960, Smith proposed the first applicable wave equation method for

pile-driving analysis. Since then, the wave equation method was accepted by the

engineers in the piling industry as a more accurate pile-driving analysis. With the

development and acceptance of PDA (Pile Driving Analyzer), the Case method (Rausche

et al. 1985), utilizing the measured force and velocity histories in approximate simple

algebraic equations, has become one of the widely used methods to evaluate the pile

capacities during pile driving. However, the static soil resistances determined by the Case

method are very sensitive to assumed Case damping factor, and shaft and toe resistance

cannot be separated by the Case method.

2
 In an effort to overcome some of the shortcomings of the Case method, an alternative

procedure, known as the CAse Pile Wave Analysis Program (CAPWAP), was developed

by Rausche et al. (1972, 1985). The CAPWAP approach is based on the one-dimensional

wave propagation model suggested by Smith (1960). In the CAPWAP analysis, one of

the two HST data (either force or velocity) is used as input to match closely with the

other by adjusting Smith model parameters. Once the acceptable match is achieved, then

the pile capacities and the Smith model parameters can be determined. In addition, some

other researchers such as Paikowsky et al. (1994), Hirsch et al. (1976), Liang and Zhou

(1996), and Liang (2003) have proposed various approaches to interpret the measured

force and velocity histories for evaluating the pile capacity. However, the capacity

estimation and drivability predictions from dynamic pile tests are, sometimes, far from

satisfactory. Thus, the dynamic testing technique still needs to be further improved.

Currently, dynamic pile tests are widely carried out to verify when the design pile

capacity is reached during pile driving and to monitor the installation process to avoid

pile damage due to hammer impact. However, there is a lack of acceptance criteria for the

number of pile tests and measured capacity. Hannigan et al. (1998) suggested that a

minimum of two dynamic pile tests be conducted for a small project. For large projects or

small projects with anticipated installation difficulties, or significant time dependent

capacity issues, a greater number of dynamic pile tests are recommended. If the test piles

do not reach a prescribed design capacity, the design load for the piles must be reduced or

additional number of piles or longer pile length must be installed. The recommendations

of the number of dynamic pile tests are entirely based on engineering experience.

3
 Paikowsky et al. (2004) provided recommendations for the number of piles to be

dynamically tested as well as the acceptance criterion for a set of driven piles. The

approach of quality assurance testing in manufacturing was employed to determine the

number of dynamic tests to be performed on production piles. However, the adoption of

normal distribution for dynamic test methods, which deviated from the actual cases that

the probabilistic characteristics of dynamic test methods can be properly represented by

the lognormal distribution, makes the recommendations very conservative. The

acceptance criterion for a set of dynamic test results is subjectively chosen such that the

average capacity of the tested piles is no less than 85% of the nominal ultimate capacity.

In addition, the application of arithmetic mean of measured capacity is far away from the

actual cases of geometric mean of measured capacity when the lognormal distribution is

used to represent the probabilistic characteristics of dynamic test methods. Therefore,

there is a need to develop more reasonable, reliability-based quality control criteria for

driven piles.

Since 1994, AASHTO (American Association of State Highway and Transportation

Officials) has been in process to change from ASD (Allowable Stress Design) method to

LRFD (Load and Resistance Factor Design) method for foundation design. Whereas ASD

considers all uncertainties in the applied load and ultimate geotechnical or structural

capacity in a single value of factor of safety (FS), LRFD separates the variability of these

design components by applying load and resistance factors to the load and material

capacity, respectively. Comparing to ASD, LRFD has the following advantages:

4
 • Accounts for variability in both resistance and load;

• Achieves relatively uniform levels of safety based on the strength of soil and rock

for different limit states and foundation types;

• Provides more consistent levels of safety in the superstructure and substructure as

both are designed using the same loads for predicted or target probabilities of

failure.

The adoption of LRFD approach makes possible the application of reliability analysis

to quantify uncertainties associated with various methods for estimating loads and

resistances, respectively. In AASHTO LRFD specifications (2003), the resistance factors

for various design methods are recommended with calibrations mostly based on the

reliability analysis using available statistical data. The design of foundation piles is

usually performed with static analytical calculations using the soil parameters from local

geotechnical site investigations and accompanied laboratory test results. The uncertainties

with respect to the prediction method, the errors of calculation model, and the spatial

variability of soil parameters, are considered in a single resistance factor recommended

for a specific design method regardless of individual site-specific situation. For each

construction site, soil profiles, soil types, pile driving equipment, and hammer

performance, will be unique. Thus, it is advantageous if a site-specific calibration for the

resistance factors can be performed to improve design.

With the use of both static load test and dynamic pile test, the pile length estimated

using static analysis methods during the design stage would be proven either adequate or

5
inadequate based on field pile tests during pile driving. Therefore, the uncertainties of

static analysis design method could be reduced by incorporating dynamic test results.

Vrouwenvelder (1992) presented an approach for incorporating either the static or

dynamic test results to update the factor of safety in the ASD method of driven piles.

Zhang et al. (2002) demonstrated that the results from static pile load tests could be

incorporated into pile design using Bayesian theory by updating the resistance factor in

LRFD. The practice in Ohio Department of Transportation nevertheless shows that more

and more dynamic pile test methods have been utilized compared to static load test, due

to savings in cost and time. Recognizing that dynamic pile testing is by far the much

preferred pile capacity verification method, a methodology needs to be developed to

update the resistance factors for static analysis method by utilizing dynamic pile test

results.

The fact that axial capacity of a driven pile may change over time after initial pile

installation has been reported by a number of geotechnical engineers for many years. The

increase of pile capacity with time is usually referred to as set-up; conversely, the

decrease of pile capacity with time is often termed as relaxation. Due to pile driving, soils

around the pile are significantly disturbed and remolded, and excess pore pressures are

generated. With passage of time, the excess pore pressure will dissipate and consequently

pile capacity is regained. Decrease in excess pore pressure is inversely proportional to the

square of the distance from the pile (Pestana et al. 2002). The time to dissipate excess

pore pressure is proportional to the square of horizontal pile dimension (Holloway and

Beddard 1995; Soderberg 1961), and inversely proportional to the soil’s horizontal

6
coefficient of consolidation (Soderberg 1961). Accordingly, larger-diameter piles take

longer time to set-up than small-diameter piles (Long et al. 1999; Wang and Reese 1989).

As excess pore pressures dissipate, the effective stress of the affected soil increases, and

set-up predominately occurs as a result of increased shear strength and increased lateral

stress against the pile. Piles driven into clay tend to experience greater set-up than piles

driven into sand and silt. Piles may experience relaxation when driven into dense and

saturated sand and silt. Based on observations in the field, numerous geotechnical

engineers developed various empirical formulas to predict the set-up behavior (e.g., Skov

and Denver 1988; Svinkin et al. 1994; Huang 1988; Zhu 1988). In particular, the semi-

logarithmic empirical relationship, proposed by Skov and Denver, has been widely used

to predict post-installation pile capacity increase with time.

With an accumulation of more experience and knowledge on set-up phenomenon,

some researchers have suggested that the set-up be formally incorporated into the

prediction method to determine total pile capacity. For example, Bullock et al. (2005)

proposed a conservative method for incorporating side shear set-up into the total pile

capacity. The predicted set-up capacity was assumed to have the same degree of

uncertainties as the measured reference capacity and a single safety factor was used to

account for all uncertainties of loads and resistances. Due to different uncertainties

associated with measured capacity and predicted set-up capacity, Komurka et al. (2005)

proposed a method to apply separate safety factors to EOD (End of Driving) and set-up

components of driven pile capacity. Furthermore, the set-up capacity was characterized as

a function of pile penetration based on dynamic monitoring during both initial driving

7
and restrike testing. The separate safety factors recommended for EOD and set-up

capacity, however, are based purely on judgment with no attendant database and

statistical analysis. Therefore, the development of a reliability analysis methodology on

set-up capacity will be desirable to separate the resistance factors in LRFD of driven piles.

1.2 Objectives of the Study

The objectives of this study can be enumerated as follows:

1. To develop reliability-based quality control criteria on driven piles from the

framework of acceptance-sampling analysis for both static and dynamic test

methods with lognormal statistical characteristics. The lognormal distribution of

statistical data of various load test methods is converted to standard normal

distribution for evaluating the confidence interval of the driven piles in achieving

the desired capacity. An optimum approach is developed for the determination of

the number of dynamic pile tests and the required measured capacities for quality

control of various dynamic test methods of driven piles.

2. To present the compilation of a database containing a large number of pile testing

data where piles were driven into clay and into sand, respectively. Based on the

compiled databases, the set-up effect will be statistically analyzed. The

compilation of the statistical database of set-up capacity makes it possible to

apply a reliability-based analysis method to develop separate resistance factors to

account for different degrees of uncertainties associated with the measured

reference capacity and predicted set-up capacity in the LRFD framework. The

reliability-based approach of FORM (First Order Reliability Method) will be used

8
 to determine the desired separate resistance factors for various target reliability

levels. The test results obtained during convenient short-time restrike can be used

to predict the long-term capacity based on the proposed formula by Skov and

Denver. The application of separate resistance factors in LRFD for the measured

reference capacity and the predicted set-up capacity can improve the prediction of

capacity of driven piles. Furthermore, by applying separate resistance factors, pile

length or number of piles can be reduced and economical design of driven piles

can be achieved.

3. To present a methodology for pile design by combining the information from the

static calculation and dynamic pile testing. Bayesian theory is employed to

combine two design methods in the framework of LRFD approach using

reliability analysis. Specifically, the results from dynamic pile tests will be

incorporated to reduce the uncertainties associated with static analysis methods by

updating the corresponding resistance factors. The combination of two design

methods could improve accuracy and confidence level on the predicted pile

capacity as opposed to the case of using only one design method.

4. To develop a one-dimensional wave equation based algorithm to interpret the

High Strain Testing (HST) data for estimation of the shaft and toe resistance of

driven piles. Both the force and velocity time histories measured at the pile head

are used as input in the new algorithm. The Smith damping factor and the static

soil resistance are directly determined based on the derived closed form solution

of the one-dimensional wave equation. The developed solution algorithm is

computationally efficient compared to the existing numerical procedure.

9
 Additionally, the shaft resistance and the toe resistance can be separated using the

derived wave equation method.

5. To develop a set of new wireless dynamic testing device for dynamic pile testing.

This new technology utilizes remote data acquisition and management unit to

build the wireless connection between the sensors and the monitoring PC. The

new wireless dynamic testing equipment is anticipated to be successfully applied

in the field testing that will advance pile driving control and monitoring to a more

secure environment.

1.3 Organization of Report

The organization of this report is as follows:

• Chapter II presents the literature review of analysis methods of driven piles.

• Chapter III presents the developed reliability-based quality control method on

driven piles, particularly regarding the criteria for the number of dynamic pile

tests and the acceptable measured capacity.

• Chapter IV presents a compiled database containing a large number of pile testing

data in clay, from which the set-up effect is statistically analyzed. The statistical

database is used in the reliability-based analysis framework to develop separate

resistance factors in LRFD for the measured reference capacity in short term and

the predicted set-up capacity in clay.

• Chapter V presents a database containing a large number of pile testing data in

sand, from which the set-up effect is statistically analyzed. The statistical database

is used in the reliability-based analysis framework to develop separate resistance

10
 factors in LRFD for the measured short-term reference capacity and the predicted

long-term set-up capacity in sand.

• Chapter VI presents the application of Bayesian theory to develop a methodology

for updating the resistance factors of driven piles by combining the static analysis

method and the dynamic test methods.

• Chapter VII presents a closed form solution to the one-dimensional wave

equation based algorithm for estimation of shaft and toe resistance of

driven piles using High Strain Testing (HST) data.

• Chapter VIII presents the developed wireless data communication device for field

application in the dynamic pile testing.

• Chapter IX presents summaries and conclusions of the research work. Also,

recommendations for implementations as well as future research topics are

provided.

11
 CHAPTER II

LITERATURE REVIEW

2.1 Introduction

Numerous analysis methods for the estimation of capacity of driven piles have been

proposed during the past hundred years. These methods can be generally categorized into

two broad groups: 1) static analysis methods based on the soil parameters obtained from

the laboratory test and in-situ test; 2) direct field pile testing methods which include static

load tests and dynamic pile tests. Static analysis methods are usually used prior to the

installation of driven piles to determine the preliminary design of driven piles, such as

pile length, pile size, pile location, and the estimation of capacity of driven piles. The

dynamic and static load tests are performed during and after the installation of driven

piles to verify the design capacity. Dynamic pile testing also serves as quality control of

pile driving.

2.2 Quality Control on Driven Piles

The most common static analysis methods used for evaluating the static axial capacity

of driven piles are as follows: α-method (Tomlinson, 1986), β-method (Esrig & Kirby,

1979), λ-method (Vijayvergiya and Focht, 1972), Nordlund’s method (Nordlund, 1963),

Nottingham and Schmertmann’s CPT method (Nottingham and Schmertmann, 1975), and

12
Meyerhof’s SPT method (Meyerhof, 1976). Nordlund’s method, β-method, Nottingham

and Schmertmann’s CPT method, and Meyerhof’s SPT method are generally used when

calculating the design capacity of driven piles in cohesionless soils, while α-method, β-

method, λ-method, and Nottingham and Schmertmann’s CPT method are used to predict

the pile capacity when piles are driven into cohesive soils. However, the pile capacity

estimated from static analysis based on the soil parameters obtained from the laboratory

test and in-situ test may not be accurate. Thus, static load test, dynamic pile test, or both,

which are believed to have higher accuracy in estimation of pile capacity, have been

performed to verify the design capacity calculated from static analysis method.

The static load tests have been performed to verify that the behavior of the driven

piles agreed with the assumption of the design for decades. There are various definitions

of pile capacity evaluated from load-movement records of a static load test. Four of them

are of particular interests; namely, the Davisson Offset Limit (Davisson 1972), the

DeBeer Yield Limit (DeBeer 1968), the Hansen Ultimate Load (Hansen 1963), and

Decourt extrapolation (Decourt 1999). NCHRP Report 507 (Paikowsky et al. 2004)

presented that Davisson’s Pile failure criterion could be used to properly determine the

reference pile capacity for driven piles, irrespective of the pile diameter and the static

load test procedure. The static load tests have been accepted by most geotechnical

engineers as the most accurate evaluation method of pile capacity. However, the cost and

time needed for a static load test hindered its extensive application in field testing.

Dynamic pile testing, an alternative approach of verification of pile design capacity in the

13
field, has become more and more attractive in geotechnical engineering due to its savings

in cost and time.

Over the past one hundred years or so, many attempts have been made to predict the

driving characteristics and the bearing capacity of piles through the use of dynamic

energy formulas. Dynamic energy formulas are based on simple energy balance

relationship in which input energy is equal to the sum of consumed energy and lost

energy. In USA, most of the state highway departments still widely use the Engineering

News Record (ENR) formula (Wellington, 1892) and its modified version for estimating

pile capacities. However, there are some reasons that make these formulas less than

satisfactory: 1) rigid pile assumption; 2) no consideration of soil-pile interaction, 3) no

incorporation of damping factor. Although many efforts have been made to improve the

dynamic energy formulas (Gates 1957, Liang and Husein 1993, Paikowsky et al. 1994,

Liang and Zhou 1996), the accuracy of estimation from dynamic energy methods is still

far from satisfactory.

As a better alternative to energy formulas for pile driving, the wave equation method

was proposed by Smith (1960) for the first practical use in estimating the pile capacity.

The governing one-dimensional wave equation was derived to describe the motion of pile

particles by applying Newton’s Second Law to a differential element of an elastic rod. In

Smith model, the pile is discretized into lumped masses interconnected by pile “spring”.

The soil resistance to driving is provided by a series of springs that are assumed to

behave in a perfectly elastic-plastic manner, and the spring stiffness is defined by the

14
ratio of the maximum static resistance of the soil Rs and the maximum elastic

deformation or quake. Damping coefficients are introduced to account for the viscous

behavior of the soil. The total soil resistance Rt is given by

Rt = Rs (1 + JV ) (2-1)

where Rs = the static soil resistance; J = the Smith damping coefficient; V = pile velocity.

In 1975 and 1976, the other two of the still widely used soil models were proposed.

They are Case model and TTI model, by Goble at Case Western Reserve University and

Hirsch at Texas Transportation Institute, respectively. These two models can be viewed

as modified version of Smith model. In Case model, the soil damping force is uncoupled

from the spring force and is dependent on the pile particle velocity. Case damping factor

Jc was introduced to account for the soil damping effect when multiplied with pile

impendence Zp and toe velocity V. The total resistance Rt is given by

Rt = Rs + J c Z pV (2-2)

It is worth noting that Jc is not related to soil properties, but a pure empirical value

calibrated with the results of static load tests.

In TTI model, the nonlinearity of soil damping force with velocity is taken into

account. The total resistance Rt during driving is given by

Rt = Rs (1 + JV N ) (2-3)

where N = an exponent less than unity to reflect the nonlinearity of the damping force

with velocity; 0.2 can be taken for N when lack of information.

15
 With the development and acceptance of pile driving analyzer (PDA), Case method

has become one of the widely used methods to evaluate the pile load capacity. However,

its inaccuracy has also been reported (Lai and Kuo 1994; Paikowsky et al. 1994). The

static soil resistances determined by the Case method are very sensitive to the assumed

Case damping factor, and the shaft and the toe resistance cannot be separated by the Case

method. In an effort to overcome these shortcomings, Rausche (1972, 1985) developed an

alternative procedure, known as the CAse Pile Wave Analysis Program (CAPWAP). The

CAPWAP approach is also based on the one-dimensional wave propagation model

suggested by Smith (1960). In CAPWAP analysis, one of the two HST data on the pile

head (either force or velocity) is used as input to generate output that would match

closely with the other HST data by adjusting Smith model parameters. Once the

acceptable match is achieved, then the pile capacities and the Smith model parameters

can be determined. However, the numerical procedure in CAPWAP (lumped mass and

springs) is computationally time-consuming, and the capacity estimation and drivability

predictions from dynamic pile tests are, sometimes, far from satisfactory. Thus the

dynamic testing technique still needs to be further improved.

Currently, dynamic pile tests are widely carried out to verify the accomplishing of

design pile capacity during pile driving and to monitor the installation process for

avoiding pile damage due to hammer impact. Hannigan et al. (1998) presented that the

number of piles that should be dynamically tested on a project depends on the project size,

variability of the subsurface conditions, the availability of static load test information,

16
and the reasons for performing the dynamic tests. A minimum of two dynamic pile tests

is recommended to be conducted for a small project. For large projects or small projects

with anticipated installation difficulties, or significant time dependent capacity issues, a

larger number of dynamic pile tests are recommended. On larger projects, CAPWAP

analyses are typically performed on 20 to 40% of the dynamic test data obtained from

both initial driving and restrike dynamic tests. If the test piles do not reach a prescribed

design capacity, the design load for the piles must be reduced or additional piles or pile

lengths must be installed. The recommendations of the number of dynamic pile tests are

entirely based on engineering experience.

Paikowsky et al. (2004) provided recommendations for the number of piles to be

dynamically tested as well as the acceptance criterion for a driven pile project. The

method of quality assurance testing in manufacturing industry is adopted to determine the

number of dynamic tests to be performed on production piles. The acceptance criterion

for a set of driven piles is subjectively recommended as the average capacity of the tested

piles to be no less than 85% of the nominal ultimate capacity. However, the adoption of

normal distribution the probabilistic characteristics of dynamic test methods deviates

from the more representative lognormal distribution, thus making the recommendation

very conservative. In addition, the application of arithmetic mean of measured capacity is

far away from the actual cases of geometric mean of measured capacity when the

lognormal distribution is used to represent the probabilistic characteristics of dynamic test

methods. Therefore, there is a need to develop improved reliability-based quality control

criteria on driven piles.

17
2.3 Reliability Analysis in LRFD

Allowable Stress Design (ASD) method has been used in civil engineering since the

early 1800s. Under ASD, the design loads, which consist of the actual forces estimated to

be applied to the structure (or a particular element of the structure), are compared to

resistance, or strength through a factor of safety (FS):

Rn Qult
Q ≤ Qall = = (2-4)
FS FS

where Q = design load; Qall = allowable design load; Rn = resistance of the element or

the structure; Qult = ultimate geotechnical pile resistance.

In the 1997 version of Standard Specifications for Highway Bridges (AASHTO 1997),

the traditional factors of safety are used in conjunction with different levels of control in

analysis and construction. However, the recommended factors of safety do not consider

the different degrees of uncertainties associated with load and resistance performance.

Thus, the reasonableness and economy of design in ASD is questionable.

In reality, both load and capacity have various sources and levels of uncertainties.

Since 1994, AASHTO (American Association of State Highway and Transportation

Officials) has been in process to supersede ASD method with LRFD (Load and

Resistance Factor Design) method for foundation design. Beginning in 2007, all federally

funded projects are required to follow the new LRFD procedure. Whereas ASD considers

all uncertainties in the applied load and the estimated ultimate geotechnical or structural

18
capacity in a single value of the factor of safety (FS), LRFD separates the variability of

these design components by applying load and resistance factors to the load and material

capacity, respectively. LRFD Bridge Design Specifications (AASHTO 2006) states that

the ultimate resistance (Rn) multiplied by a resistance factor (Φ), which thus becomes the

factored resistance, must be greater than or equal to the summation of loads (Qi)

multiplied by corresponding load factors (γi), and modifiers (ηi). LRFD criterion for

strength limit states can be expressed as:

φR n ≥ ∑ η i γ i Q i (2-5)

where

ηi = η D η R η I > 0.95 (2-6)

where ηD = effects of ductility; ηR = redundancy; ηI = operational importance.

McVay et al. (2000) presented the statistical parameters and the corresponding

resistance factors for various dynamic methods in LRFD of driven piles based on a

database of 218 pile cases in Florida. Eight dynamic methods were studied: ENR,

modified ENR, FDOT, Gates driving formulas, CAPWAP, Case method for PDA,

Paikowsky’s energy method, and Sakai’s energy method. It was demonstrated that the

modern methods based on wave mechanics, such as CAPWAP, PDA, and Paikowsky’s

energy methods, are roughly twice as cost effective to reach the target reliability indices

of 2.0 to 2.5 (probability of failure =0.62 to 2.5%) as the ENR and modified ENR driving

formulas.

19
 Paikowsky et al. (2004) compiled a large database in NCHRP Report 507 that was

used for the quantitative assessment of pile capacity evaluation methods during both

design and construction. The resistance factors for various design methods were

calibrated according to soil types, pile types, and time of testing. Static analysis methods

used in most design practices were found overall over-predicting the observed pile

capacities. Most dynamic capacity evaluation methods used for quality control were

found overall under-predicting the observed pile capacities. Both findings demonstrated

the shortcoming of safety evaluation based on absolute values (i.e., resistance factors or

factors of safety) and the need for an effective measurement index to objectively assess

the performance of methods of analysis. An efficiency factor is recommended by McVay

et al. (2000) and Paikowsky et al. (2004) to account for the bias of the analysis methods

and to provide an objective evaluation regarding the effectiveness of the capacity

prediction method.

Based on the NCHRP Report 507 (Paikowsky et al. 2004), AASHTO (2006) divided

the design methods of driven piles into two major categories: static calculation and

dynamic testing, and proposed an independent resistance factor for each design method.

Where nominal pile axial resistance is determined during pile driving by dynamic

analysis, dynamic formula, or static load test, the uncertainty in the pile axial resistance is

strictly due to the reliability of the resistance determination method used in the field

during pile installation. Actually, each type of design method, however, has its own

merits and one method may eliminate uncertainties that otherwise may not be eliminated

by the others. With the widespread uses of driven piles, the fact that the static calculation

and dynamic testing play a relatively independent role in the design of driven piles has

20
lead to a large variability in the estimation of pile capacities and consequently high

project cost.

Vrouwenvelder (1992) presented an approach for incorporating dynamic test results

to update factor of safety in the ASD design of driven piles. It was stated that the rules for

dynamic testing and pile driving analysis are formulated in a more strict way than those

for static analysis methods. The formulation for static analysis method seems to be less

restrictive. Nevertheless, the dynamic pile tests can only determine the pile capacity at the

testing time and cannot predict the long-term pile capacity as a function of time after pile

installation. It could be found that the various methods are not treated in a fully

symmetric manner. Furthermore, it should open the possibility to combine the results of

various methods. The application of reliability theory makes it possible to combine

various methods.

Zhang et al. (2002) demonstrated that the results from static pile load tests could be

incorporated into pile design using Bayesian theory by updating the resistance factor in

LRFD. The practice in Ohio Department of Transportation nevertheless shows that more

and more dynamic pile test methods have been utilized compared to static load test, due

to savings in cost and time. Recognizing that dynamic pile testing is by far the much

preferred pile capacity verification method, there will be a need to update the resistance

factors for static analysis method by utilizing dynamic pile test results.

21
2.4 Set-up Effect

Set-up (time-dependent increase in pile capacity) has long been recognized as a well-

known phenomenon, and can contribute significantly to long-term pile capacity.

Economical design of driven piles would not be achieved without the recognition of set-

up effect. When a pile is driven into the ground, the soils surrounding the pile are

significantly disturbed and excess pore pressure is generated. After pile driving, the pile

regains its capacity with time primarily due to the dissipation of excess pore pressure.

Set-up behaviors have been reported for various types of soil and pile during the past

decades. Numerous researchers have also developed various formulas to predict the

development of set-up capacity after the installation of driven piles.

Skov and Denver (1988) investigated the set-up effects by arranging restrike on piles

of different age. It was found that, after a certain period of time, depending on soil

conditions, the time-dependent increase in capacity can be considered approximately

linear with the logarithm of time. Skov and Denver also presented an empirical equation

for set-up based on a logarithmic increase of pile capacity with time as follows

t
Qt = Q0 (1 + A log ) (2-7)
t0

where

Qt = axial capacity at time t after driving,

Q0 = axial capacity at time t0,

A = a factor that is a function of soil type and can be determined as the slope of

the linear portion of the normalized capacity (Qt/Q0) versus log(t) plot.

22
 t = time since installation, and

t0 = time after installation at which point the capacity gain becomes linear on a

log(t) plot.

They also recommended numerical values for A and t0 as 0.6 and 1 day in clay and 0.2

and 1 day in sand, respectively.

Zhu (1988) presented an empirical expression for the set-up coefficient in terms of

sensitivity of cohesive soil for the determination of regain of capacity by performing pile

tests in the coastal area of East China. According to collected information, the maximum

adhesion at soil-pile interface, when reinforced concrete piles are driven into clay, is

approximately equal to cohesion of undisturbed soil. The bearing capacity of the pile, as a

function of time, is related to the reconsolidation of soil and can be expressed in terms of

sensitivity as Eq. (2-8). The more sensitive the soil, the more obvious the regain of

bearing capacity after pile installation.

Q14 = QEOD (1 + 0.375S t ) (2-8)

where

Q14 = pile capacity 14 days after pile installation;

QEOD = pile capacity at end of driving;

St = sensitivity of soil.

Svinkin (1994) studied a number of piles driven into mostly sandy soil and presented

the upper and lower boundaries to predict the development of set-up of driven piles in

sandy soil. Pile capacity at end of initial driving is the important value in sandy soil and
23
can be taken as the lower limit of pile capacity. It was found that the tendency of soil set-

up for all five investigated piles was generally the same, but the set-up coefficient ranged

between upper boundary which can be expressed as

Qt = 1.4QEOD t 0.1 (2-9)

and lower boundary which can be expressed as

Qt = 1.025QEOD t 0.1 (2-10)

where

Qt = pile capacity at the time t after pile driving;

QEOD = pile capacity at end of driving;

t = time after pile driving (days).

Svinkin also concluded that the set-up effect in sandy soils to a certain degree depends on

the level of ground water table.

The effects of time on pile capacity are believed affected by the type of soil into

which the pile is driven. Set-up is recognized as occurring in organic and inorganic

saturated clay, and loose to medium dense silt, silty sand, and fine sand (Astedt and Holm,

1992; Attwooll et al., 1999; Hannigan et al., 1998). Holloway and Beddard (1995)

observed little or no set-up in very silty low-plasticity cohesive materials. Walton and

Borg (1998) indicated that set-up in sand and gravel may not be a significant factor in

long-term pile capacity. Percentage-wise, piles driven into soft clay tend to experience

greater set-up than piles driven into stiff clay. Piles driven into loose sands and silts

generally experience a smaller magnitude of set-up than those in soft clays. Piles driven

24
into saturated, dense sands and silts or into shales may experience relaxation (Long et al.,

1999; Svinkin, 1994, 2002; Titi and Wathugala, 1999).

Set-up is not only affected by soil type, but also related to the pile type. Set-up rate

decreases as pile size increases (Camp and Parmar, 1999). Long et al. (1999) offered that

there is no clear evidence of difference in set-up between small- and large-displacement

piles. A study by Finno et al. (1989) reported that a pipe pile generated higher excess

porewater pressures during installation than did an H-pile, but that unit shaft resistance

for the two piles were approximately equal after 43 weeks. Wood piles tend to set-up

faster than steel or concrete piles, and more permeable wood piles set-up faster than less-

permeable wood piles (Bjerrum, et al., 1958; Yang, 1956). For piles installed in organic

silt, Yang (1956) observed greater set-up for wood piles than for steel piles. Pre-stressed

concrete piles generally exhibit more set-up than steel piles. This phenomenon has been

attributed to a higher soil/pile interface coefficient of friction (Priem et al. 1989).

Long et al. (1999) used a database of pile load tests (both static and dynamic)

collected from the literature to quantify time effects on pile capacity. The database is

divided into three groups according to the primary subsurface soil profile: clay, sand, and

mixed soil. Set-up in clay and mixed profiles varied from one to more than six times the

axial capacity estimated immediately after driving. The gain in axial pile capacity with

time is explained only partially by the dissipation of excess pore pressures. Aging can

also contribute to the increase in capacity with time.

25
 Camp and Parmar (1999) observed that the piles driven into the Cooper Marl, a stiff

and overconsolidated cohesive calcareous soil, experienced significant set-up. The data

available from 14 sites involving 114 tests were analyzed to evaluate whether the Skov

and Denver formulas would apply. They proposed the use of 2 days as the time for the

onset of the linear set-up on the semi-logarithm plot. With this assumption, the back-

calculated A values generally follow the expected trends. The average A value decreases

as the pile size increases. This means that the smaller piles gain capacity at a faster rate

than the larger piles, which is in agreement with Soderberg’s work (1961).

With an accumulation of more experience and knowledge on set-up phenomenon,

some researchers have suggested that the set-up be formally incorporated into the

prediction method to determine total pile capacity. For example, Bullock et al. (2005)

proposed a conservative method for incorporating side shear set-up into the total pile

capacity by analyzing the results of a driven pile side shear set-up (SSS) test program

performed at the University of Florida. It was observed that all pile segments showed SSS,

with similar magnitude in both cohesionless and cohesive soils, and irrespective of depth.

Pile load tests and standard penetration tests with torque measurement (SPT-T) confirmed

the approximate semilog-linear time versus SSS behavior and usefulness of the set-up

factor A. They believed that engineers can include SSS in routine design using a

minimum set-up factor A=0.1 for soils. The predicted set-up capacity was assumed to

have the same degree of uncertainties as the measured reference capacity and a single

safety factor was used to account for all uncertainties of loads and resistances.

26
 Due to different uncertainties associated with measured capacity and predicted set-up

capacity, Komurka et al. (2005) proposed a method to apply separate safety factors to

EOD (End of Driving) and set-up components of driven pile capacity. CAPWAP results

were used to estimate unit shaft resistance distribution (unit shaft resistance as a function

of depth) at EOD and at BOR (Beginning of Restrike). Furthermore, the set-up capacity

was characterized as a function of pile penetration based on dynamic monitoring during

both initial driving and restrike testing. The separate safety factors recommended for

EOD and set-up capacity, however, are based purely on judgment with no attendant

database and statistical analysis.

2.5 Current Wireless Applications in Dynamic Pile Testing

Dynamic pile testing has evolved from the traditional energy method to the more

accurate wave equation method since the first applicable one-dimensional wave

propagation model was suggested by Smith (1960). The work done by Smith opened new

horizons for subsequent researchers to follow and to improve on. In 1976, two widely

used computer programs, TTI and WEAP (Wave Equation Analysis of Pile), were

developed by Hirsch at Texas Transportation Institute and Goble et al. at Case Western

Reserve University, respectively.

Aiming at solving the actual input hammer energy to piles and providing a better

understanding of soil-pile interaction during pile driving, PDA (Pile Driving Analyzer)

was developed in the 1970’s as a method to directly measure dynamic pile response

during driving. As the name implies, it was developed to analyze pile driving and

27
evaluate pile driveability, including the range of stresses imparted to the pile, hammer

efficiency, etc. During dynamic pile testing, the strain gages and accelerometers are

usually attached near the pile head to record the waveforms. The sensors are typically

connected to the monitoring computer with 100-ft cables hanging down from pile top.

The measured strains are converted into axial forces and the measured accelerations are

integrated over time to velocity, both as a function of real time. With the development

and acceptance of PDA (Pile Driving Analyzer), the Case method (Rausche et al. 1985),

utilizing the measured force and velocity time histories in approximate simple algebraic

equations, has become one of the widely used methods to evaluate the pile capacities

during pile driving. Rausche et al. (1972, 1985) also developed a numerical program

known as CAPWAP in which one of the two HST data (either force or velocity) from

PDA is used as input in a numerical simulation to match closely with the other HST data

by adjusting Smith model parameters. Once the acceptable match is achieved, then the

distribution of side and toe resistances and the Smith model parameters can be

determined.

The current dynamic testing hardware and analysis software are rather expensive. The

operation of the PDA and attendant CAPWAP analysis requires highly trained personnel

to carry out. Furthermore, the typically 100-ft long sensor cables are difficult to handle

and often time can lead to safety hazards to engineers and operators attempting to install

the gages to the piles. These cables can be damaged by the construction and pile driving

equipments, resulting in considerable delays. Therefore, there is a need to develop an

28
alternative, robust, easy to handle wireless communication capability between the sensors

and a data acquisition unit from a distance.

McVay et al. (2002) developed a technology consisting of non-recoverable (cast into

the pile) wireless transmitter and instrumentation package at the University of Florida for

pile monitoring. The new testing system can be divided into two modules – Non-

Recoverable Unit and Receiver and Data Processing Unit. The Non-Recoverable Unit

embedded into concrete piles consists of two strain gages, two accelerometers, a signal

conditioning and sending unit, and antenna. The Non-Recoverable Unit is capable of

broadcasting the stress wave information to the Receiver and Data Processing Unit during

pile driving. The Receiver and Data Processing Unit receives the information from the

Non-Recoverable Unit to calculate force traces, velocity traces and pile capacities. The

Receiver and Data Processing Unit is made of receiver antenna, receiver and conditioning

unit, data acquisition card, laptop computer, and data acquisition and processing software.

Since the Non-Recoverable Unit is pre-installed directly into the concrete piles, no

installation set-up is required at the time of pile driving. The wireless operation of

proposed technology minimizes the accident potential and construction delays (e.g.,

engineers attempting to install the gages to the piles and connection cables being

damaged). The Non-Recoverable Unit can be installed into both pile top and bottom so

that the force-time and velocity-time histories at the pile top and bottom can be obtained,

thus resulting in a more accurate analysis on the distributions of pile stresses and pile

capacities. Additional benefits include that the wireless system sleeps and is capable of

29
waking itself up during an extreme event (e.g., earthquake, collision); therefore, it can

report the magnitude of the event and the resulting structural damage.

Laboratory and full-scale field tests have been performed by the University of Florida

for calibrating and validating equipment responses. The output signals from wireless

analysis package (WAP) system were very strong with less noise and can be achieved

within the distance of 60 m. The WAP output signals show a very satisfactory

comparison with the PDA outputs. An acquisition and data processing software was

developed by LABVIEW to process the testing data for pile capacity determination, as

well as pile stress monitoring during driving. The program is called Pile-Monitoring. vi

including three theoretical methods of capacity assessment: Case Method, Paikowsky

energy approach, and UF method.

Although the favorable field testing results could be obtained, many problems arose

in the development of the equipment, the installation of the Non-Recoverable Unit, and

the wireless transmission. The sensors, cables, and conditioners in Non-Recoverable Unit

are easily broken and misaligned due to the concrete pouring. Due to the embedment of

the Non-Recoverable Unit into the piles, it is not easily checked if errors happened during

the pile driving. The wireless analysis package (WAP) system was not able to wake up

from the low hammer strike accelerations.

The cost is a major limitation for the application of the developed WAP system. The

approximate cost of $300 is needed for Non-Recoverable Unit (transmitters, conditioners,

30
and instruments: accelerometers, strain gages, etc.). The cost is about $6,400 for

acquiring the Receiver and Data Processing Unit (receiver, PCMCA card, laptop

computer, and LABVIEW software). The requirement of the experienced engineers on

the installation and operation of WAP system may potentially hinder its extensive

application.

2.6 Summary

From the review presented in this chapter, it can be seen that, over the last ten or

fifteen years, several researchers have made major advances in applying reliability

analysis to driven pile problems. Although the geotechnical engineers long ago learned

practical ways to deal with uncertainties, they have been reluctant to embrace the more

formal and rational approaches of reliability theory while other fields of civil engineering

have made major commitments to probabilistic approaches. However, it is undeniable

that reliability analyses provide insight into the relative contributions of different factors

to the uncertainty of the prediction results and thus giving guidance for where further

investigations are more fruitful. The reliability-based methodology could be used to

further improve the quality control on pile driving and refine the LRFD of driven piles.

The purpose of this study is to develop a reliability-based quality control method on

pile driving, to develop an updating methodology for improving resistance factors in

LRFD by combining static analysis methods with dynamic test methods, and finally to

incorporate set-up into reliability-based design of driven piles in clay and in sand. In

addition to enhancing reliability based approach outlined above, this study also provides

31
new wave equation simulation technique for pile driving, which can minimize the need

for CAPWAP trial-and-error procedure in achieving best matches. Finally, the research

has resulted in a development of a new generation wireless communication technology

with the necessary hardware and software to support its operation in dynamic pile testing

applications.

32
 CHAPTER III

QUALITY CONTROL METHOD FOR PILE DRIVING

3.1 Introduction

Static load tests have been used to verify the load-carrying capacity of driven piles.

Additionally, dynamic pile tests have also been used as a quality control measure during

pile driving to ensure adequacy of pile capacity. It has been suggested that the number of

static load tests required to support the use of the resistance factor in the LRFD (Load and

Resistance Factor Design) should be established based on site subsurface conditions

characterized by field exploration and laboratory testing program. One or more static load

tests should be performed per site to justify using the specified resistance factors for piles

installed within a site (AASHTO 2006). Although dynamic pile tests have been widely

used for quality control of driven piles, there is a lack of acceptance criteria for linking

the number of dynamic pile tests and the required measured pile capacity.

Hannigan et al. (1998) suggested that a minimum of two dynamic pile tests be

conducted for a small project. For large projects or small projects with anticipated

installation difficulties, or significant time dependent capacity issues, a greater number of

dynamic pile tests is recommended. If the test piles do not reach a prescribed design

capacity, the design load for the piles is reduced or additional piles or pile lengths are

33
installed. Paikowsky et al. (2004) provided recommendations for the number of piles to

be dynamically tested as well as the corresponding acceptance criterion for driven pile

project. The quality assurance testing used approach in the manufacturing industry was

employed to determine the number of dynamic tests to be performed on production piles.

However, the use of normal distribution to characterize the statistics of dynamic test

methods is very conservative, since the lognormal distribution has been known to better

represent the probabilistic distribution characteristics of dynamic pile test results. The

acceptance criterion for a set of piles was subjectively chosen so that the average capacity

of the tested piles was no less than 85% of the nominal ultimate capacity. The use of

arithmetic mean of the measured capacities is far different from the geometric mean of

the measured capacities when the lognormal distribution is used to represent the

probabilistic characteristics of dynamic pile tests.

In this chapter, quality control criteria for driven piles are developed from the

framework of acceptance-sampling analysis for both static and dynamic pile test methods,

adopting a more realistic lognormal distribution function to characterize the probabilistic

properties of results of various prediction methods. The lognormal distribution of

statistical data of various load test methods is converted into a standard normal

distribution for evaluating the confidence interval to assure that the population mean of

pile capacities be no less than the factored design loads. The number of load tests and the

required measured capacities are interrelated, since reducing one of them invariably

increases the other. An optimum approach is recommended for the selection of the

34
number of load tests and the required measured capacities for different quality control

methods in driven piles.

3.2 Distribution Function of Static and Dynamic Pile Test Results

A database of 16 static pile load tests at a site in southern Italy, reported by

Evangelista et al. (1977), showed that the lognormal distribution could represent the

range of the pile capacities. Static pile load tests directly measure driven pile capacity;

therefore, the mean of bias factor, λ1, is 1.0. The coefficient of variation, COV1, reflects

the effects of the spatial variation from one pile location to another, the soil disturbance

caused by pile driving, and set-up or relaxation after pile installation. Empirical data of

sufficient quality to estimate within-site variability is not available. The site spatial

variation, in general, could be categorized as low, medium, or high variability with the

corresponding COV of 0.15, 0.25, and 0.35, respectively (Phoon and Kulhawy 1996,

1999a, and 1999b). A study by Zhang and Tang (2002) has shown that the COV of the

capacity of driven piles from static pile load tests at eight sites ranged from 0.12 to 0.28,

thus providing support for the suggested classification by Phoon and Kulhawy.

The pile capacities measured by dynamic load tests, Rd, can be related to the pile

capacities from static pile load tests, Rs, by the bias factor, λ2, as follows

R s = λ 2 Rd (3-1)

Other than the uncertainty of static pile load tests, the dynamic pile test itself also

introduces variability due to model and systematic errors. For example, the dynamic

formulas, such as ENR and Gates formulas, suffer a great deal of uncertainties in
35
determining input energy. Although the wave equation methods, such as CAPWAP

(CAse Pile Wave Analysis Program) and Case methods using the PDA (Pile Driving

Analyzer), may partially eliminate the uncertainties associated with the energy methods,

these methods nevertheless introduce other uncertainties, such as the error of signal

recording of the transducers as well as simplification of damping factors in the soil-pile

interaction model. The set-up or relaxation after pile installation, especially in clayey soil,

could not be properly accounted for in these dynamic pile testing methods. The

variability of hammer performance and pile material properties could also introduce the

errors to the dynamic pile testing method. The statistical characteristics for the bias factor

for various dynamic test methods are needed for formulating the reliability-based pile

acceptance criteria. Recently, McVay et al. (2000) and Paikowsky et al. (2004) have

presented these statistical characteristics and they are summarized in Table 3-1. It has

been found that the lognormal function could properly represent the distribution of the

bias factor, λ2.

3.3 Determination of Number of Load Tests and Corresponding Acceptance Criterion

The reliability-based acceptance-sampling analysis could provide a theoretical

framework from which to estimate the mean of pile population. In the acceptance-

sampling analysis, confidence intervals offer a means of establishing a range of values in

which a true value is expected to lie, thus providing a probability statement about the

likelihood of correctness (Ayyub and McCuen 2003). A test statistics Z can be used to

compute the confidence interval for the case of normal distribution where the standard

36
deviation, σ, is known. The lower one-sided confidence intervals on the population mean,

μ, are given by:

σ
X − Zα ( )≤μ ≤∞ (3-2)
n

where X is the sample mean; n is the sample size; Zα is reliability index corresponding

to the value of random variables having the standard distribution and cutting off α percent

in the tail of the distribution; and α is the level of significance and equal to 1-γ where γ is

the level of confidence.

For quality control purpose of pile driving, field engineers usually select a certain

number of test piles to determine the population mean of the pile capacity and to make

sure that the population mean is no less than the factored design load at a certain level of

reliability. It can be stated as:

σR
L ≤ R − Zα ( )≤∞ (3-3)
n

where L = factored design load; R = the mean of equivalent static resistance that is

computed as the product of the mean of bias factor and the dynamically measured

capacity; σR = the standard deviation of equivalent static resistance; n = the number of

dynamic pile tests. The reliability index Zα is derived from the hypotheses test based on a

test statistics Z. The reliability index in this chapter relates to the probability that a set of

pile is incorrectly accepted as good, where the population mean is less than the factored

design load. Therefore, the reliability index in this chapter carries different significance

37
from that commonly accepted in the LRFD design outlined by Withiam et al. (2001).

Assuming

Y = λt (3-4)

where Y = ratio of equivalent static capacity to the factored design load; λ = bias factor

for a type of dynamic pile test method; t = ratio of static or dynamic measured capacity to

the factored design load. The variable λ is a random variable that accounts for the

variability of the predicted capacity from various load test methods. The variable t is

treated as a deterministic variable obtained as the ratio of actual static or dynamic

measured capacity to the factored design load. Although there is a limitation in ignoring

the uncertainties in design load herein, for simplification, the variability associated with

design load is taken into account by the load factors because the uncertainties associated

with load components are much smaller than those associated with resistance components.

Therefore, the following expressions can be obtained for the ratio Y:

μY = μ λ t (3-5)

σ Y = σ λt (3-6)

where μλ and σλ are mean value and standard deviation of bias factor, λ, respectively. The
lognormal distribution can be employed to express the variable Y as Y ~ LN(ψ Y , ζ Y )
2
1 2 2 2 σ
where ψ Y = lnμ Y − ζ Y , and ζ Y = ζ λ = ln(1 + λ 2 ) . Therefore, by utilizing the
2 μλ

transformation relationship between the lognormal distribution Y ~ LN(ψ Y , ζ Y ) and the

standard normal distribution Z ~ N(0,1) , and the integration results for the standard
normal, the cumulative distribution function for the lognormal distribution can be
evaluated using the following transformation:

38
 ln Y − ψ Y
Z= (3-7)
ζY

Fig. 3-1 shows the tail areas of the normal and lognormal distributions in which μ = 1

and σ = 0.3. It can be seen that the difference between the corresponding deviations from

mean is very large for the normal and lognormal distribution when the required levels of

significance (i.e., value of tail area) is very small. For example, if the mean of a normal

distribution lies two standard deviations above a single lower specification limit, the

levels of confidence accepting the set of piles is 97.7% for the population with the normal

distribution and 99.8% for the population with the lognormal distribution, respectively. It

is obvious that when requiring high level of confidence, the use of sampling

characteristics must be extremely careful (Phoon and Kulhawy 2005).

Now, considering ln(y) as a sample value, Eq. (3-2) can be expressed as:

ζY
ln y − Z α ( ) ≤ψ Y ≤ ∞ (3-8)
n

1
Noting that ψ Y = ln μ Y − ζ 2Y , y = λt , and ζ Y = ζ λ , Eq. (3-8) becomes:
2

1 ζ
ln λt + ζ λ2 − Z α ( λ ) ≤ ln μ Y ≤ ∞ (3-9)
2 n

In order to assure that μY is no less than 1 (i.e. ln μY ≥ 0), the following expression

can be obtained:

39
 ____
1 ζ
0 ≤ ln t + ψ λ + ζ λ2 − Z α λ ≤ ∞ (3-10)
2 n

1
where ψ λ = ln λ = ln μ λ − ζ λ2 . By requiring that the mean capacity of population piles
2

be no less than the factored design load, the relationship of the required ratio, t, and the

number of load tests can be derived for a particular dynamic pile test method and the

chosen level of reliability.

3.4 Target Reliability Index vs. t

The test statistics Z is applied to determine the lower one-sided confidence interval,

from which the relationship between target reliability index, β, and level of significance,

α, can be expressed as:

β = Z α = Φ −1 (1 − α ) (3-11)

where Φ-1 is the inverse of the cumulative distribution function of the standard normal

distribution. The target reliability index increases approximately proportionally with an

exponential decrease in level of significance.

The lognormal distribution has been demonstrated to represent the statistics of static

and dynamic pile test results, thus the mean of t can be expressed as:

t = n t1 t 2 ⋅ ⋅ ⋅ t n (3-12)

where tn is the ratio of nth measured capacity of pile test to the factored design load. When

the test method and the number of test piles are pre-set, the relationship between

reliability indexes β and the geometric mean of t could be obtained from Eq. (3-10)

40
based on the statistics of the test method. Fig. 3-2 shows the relationship between

reliability indexes β and the geometric mean of t for the static load test method with the

high site variability. It can be seen that the required t increases with an increase in the

desired reliability index. The observation is consistent with experience-based quality

control approach, where the higher target reliability index demands the higher measured

mean capacity of load test for a given number of pile load tests. Furthermore, the required

t is very sensitive to the selected target reliability index when the number of load tests is

small. With an increase in the number of load tests, the sensitivity of t to the reliability

index is lessened. At a certain target reliability level, an increase in the number of load

tests provides higher confidence on the estimation of the mean capacity of pile population,

thus allowing for the adoption of lower required t .

3.5 Number of Load Tests vs. t

As suggested in NCHRP Report 507 (Paikowsky et al. 2004), the target reliability

index of β = 2.33 corresponding to the level of significance α = 1% is recommended for

redundant piles defined as 5 or more piles per pile cap; the target reliability index of β =

3.00 corresponding to the level of significance α = 0.1% is recommended for

nonredundant piles defined as 4 or fewer piles per pile cap. Figs. 3-3 and 3-4 present the

relationship between the number of pile load tests and t for different dynamic pile test

methods for reliability index β = 2.33 and 3.00, respectively. It can be seen that the

required t decreases with an increase in the number of load test for all dynamic pile test

methods. The first several load tests exert more pronounced effects on the reduction of

the required t , compared to the rest of load tests. The required t is not only dependent
41
on the number of load tests, but also related to the statistical distribution of load test

capacity. For static load test with lowest COV, the required t exhibits the least sensitivity

to the number of load tests. On the other hand, for the CAPWAP EOD method with the

largest COV, the required t exhibits the highest sensitivity to the number of load tests.

Figs. 3-3 and 3-4 can be adopted as an efficient and economical basis to determine the

number of load tests and the corresponding required acceptance criterion for various pile

test methods at the selected target reliability indices.

3.6 Correction Factor for Finite Population

When the number of tested sample n is comparable to the total number of population

N, that is, the sampling fraction, f = n/N, is greater than about 10%, the effect of finite

population comes into play. Ayyub and McCuen (2003) have proposed a finite population

correction factor, (N-n)/(N-1), to account for the effect of finite population. Thus, the

variance of the sample mean, X , can be calculated with the known standard deviation, σ,

as follows:

( N − n) σ 2
Var ( X ) ≅ (3-13)
( N − 1) n

The application of correction factor for finite population reduces the sampling variance,

because the assumption of sampling without replacement in entire population is no longer

reasonable. For most piling projects, the effect of finite population must be considered in

determining the number of pile load tests and the corresponding acceptance criterion.

42
 Fig. 3-5 illustrates the effect of finite population on the required number of pile load

tests and the corresponding acceptance criterion for static load tests at high site variability

for β = 3.00. It can be seen that an increase of sampling fraction corresponds to a

decrease of the variance of the sample mean, thus resulting in a decrease of the required

t for a given number of pile load tests. For the cases where the number of population

piles is low, the relationship between the number of load tests and the required t exhibits

a drastic change from the case where the number of population piles is infinitely large. It

should be noted that the required t is 1.0 for the case of 10 test piles when the total

number of piles is 10. It means that there is no variance in the sample average since all

the production piles have been accounted for. It is consistent with engineering experience

that the mean of the measured capacity of test piles is required to be no less than the

factored design load when all the production piles are tested.

3.7 Quality Control Criteria

When the correction factor on finite population is considered and a target reliability

index is preset, the relationship between the number of pile load tests and the required

t can be obtained for different load test methods from Eq. (3-10). Tables 3-2 and 3-3

provide the number of pile load tests and the corresponding acceptance criteria for the

quality control of driven piles using various load test methods for the target reliability

indices β = 2.33 and 3.00, respectively. The number of piles that should be tested at a

given construction site depends on the project size, the variability of the subsurface

conditions, the adopted load test method, the target reliability index, and the relationship

between the measured capacity of load test and the factored design load. In general, an

43
increase in the number of load tests gives engineers more confidence on the prediction of

the overall behavior of population piles.

For all dynamic testing methods with signal measurements, the selection of the

number of load tests and the corresponding acceptance criterion is suggested as the point

at which the required t decreases about 10 percent when an additional load test is

performed. It means, at that point, increasing the number of pile load tests can no longer

effectively reduce the required t . The number of pile load tests and the required t at this

point are recommended as an optimum point for selection of quality control criterion. In

Tables 3-2 and 3-3, the difference between statistical characteristics for various dynamic

test methods results in the different required t . However, it can be found that the ratios

of equivalent static capacity to factored design load, Y = λt , are about 1.87, 1.85, 1.81,

1.76, 1.79, and 1.73 for dynamic test methods of CAPWAP EOD, CAPWAP BOR, PDA

EOD, PDA BOR, EA EOD, and EA BOR, respectively. It means that various dynamic

test methods demand almost the same required design capacities to verify the structure

reliability, based on the above approach of determining the required t and the number of

load tests. On the other hand, it is noted that the dynamic test methods with relatively

small COV, such as CAPWAP BOR, demand a less number of load tests. In other words,

in order to obtain the same level of reliability, the more accurate dynamic test methods

require a less number of test piles for a given construction site.

For static load tests, the total number of production piles has small effect on

determining the quality control criteria due to the method’s high accuracy. Due to high
44
cost and time-consuming nature of static load tests, it is not economical to increase the

number of load tests to reduce the required t . For a project, if the site has similar

subsurface soil conditions, the overall project piles can be considered as the entire

population. Otherwise, the criteria should be independently applied to each part of the

project site where similar subsurface soil conditions exist. If the measured pile capacity

cannot satisfy the required criteria, the alternative solution is to install additional piles or

to drive piles deeper. It should be noted that, when pile relaxation occurs at a project site,

the quality control criteria should be carefully used for dynamic pile tests at EOD, since

higher degree of confidence could be obtained from static load test or dynamic pile tests

at BOR.

3.8 Conclusions

In this chapter, acceptance-sampling analysis has been employed to develop quality

control criteria for driven piles. The more representative lognormal distribution of

statistical data of various load test methods is used to evaluate the confidence interval for

assuring that the population mean of pile capacities be no less than the factored design

loads. The required measured capacity decreases with an increase in the number of load

test in achieving the same level of reliability. Theoretically, an increase of both the

required measured capacity and the number of load tests enhances the reliability level of

population pile capacity. However, there is an optimum point where the number of load

test and the required measured capacity are economically associated to yield reasonable

quality control criteria based on acceptance-sampling analysis techniques. The correction

factor of finite population has been incorporated in interpreting the cases of limited pile

45
population that represent most of geotechnical projects. The number of load tests

associated with the criteria of the measured capacities for driven piles has been

recommended in Tables 3-2 and 3-3 for different load test methods with different levels

of reliability.

46
Table 3-1 Distribution Parameters of Various Dynamic Test Methods for Driven Piles

Number Mean of bias Standard Coefficient of References

Load test
of cases factor, λ2 deviation variation,
method
COV
CAPWAP EOD 125 1.626 0.797 0.490 Paikowsky et al. (2004)

CAPWAP BOR 162 1.158 0.393 0.339 Paikowsky et al. (2004)

PDA EOD 48 1.344 0.443 0.329 McVay et al. (2000)

PDA BOR 42 1.036 0.322 0.311 McVay et al. (2000)

EA EOD 128 1.084 0.431 0.398 Paikowsky et al. (2004)

EA BOR 153 0.785 0.290 0.369 Paikowsky et al. (2004)

CAPWAP = Case Pile Wave Analysis Program; PDA = Case Method Using the Pile Driving Analyzer; EA
= Paikowsky’s Energy Approach; EOD = End of Driving; BOR = Beginning of Restrike.

47
 Table 3-2 Recommended Number of Load Tests to Be Conducted for Quality Control of Driven Piles (Reliability β=2.33)

Site Variability

Low Medium High

Test methods CAPWAP PDA EA CAPWAP PDA EA CAPWAP PDA EA

SLT SLT SLT
Time of test EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR

Required ratio t 1.45 1.15 1.60 1.35 1.70 1.65 2.20 1.75 1.15 1.60 1.35 1.70 1.65 2.20 1.75 1.15 1.60 1.35 1.70 1.65 2.20

<=10 1 3 2 2 2 3 3 1 3 3 3 3 3 3 2 4 3 3 3 4 4

11~20 1 3 2 2 2 3 3 1 4 3 3 3 3 3 2 4 3 3 4 4 4
No. of pile
21-50 1 4 2 2 2 3 3 1 4 3 3 3 4 4 2 4 3 4 4 4 4
population
51-100 1 4 2 2 2 3 3 1 4 3 3 3 4 4 2 5 3 4 4 4 5

>=100 1 4 2 2 2 3 3 1 4 3 3 3 4 4 2 5 3 4 4 4 5

SLT = Static Load Test; CAPWAP = Case Pile Wave Analysis Program; PDA = Case Method Using the Pile Driving Analyzer; EA = Paikowsky’s Energy

Approach; EOD = End of Driving; BOR = Beginning of Restrike.

t = n t 1 t 2 ⋅ ⋅ ⋅ t n , where tn is the ratio of nth measured capacity of pile test to the factored design load.

48
 Table 3-3 Recommended Number of Load Tests to Be Conducted for Quality Control of Driven Piles (Reliability β=3.00)

Site Variability

Low Medium High

Test methods CAPWAP PDA EA CAPWAP PDA EA CAPWAP PDA EA

SLT SLT SLT
Time of test EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR EOD BOR

Required ratio t 1.60 1.15 1.60 1.35 1.70 1.65 2.20 2.10 1.15 1.60 1.35 1.70 1.65 2.20 2.10 1.15 1.60 1.35 1.70 1.65 2.20

<=10 1 4 3 3 3 4 4 1 5 4 4 4 4 4 2 5 4 4 4 5 5

11~20 1 5 3 3 3 4 4 1 5 4 4 4 4 5 2 6 5 5 5 6 6
No. of pile
21-50 1 5 3 3 4 5 5 1 6 4 4 4 5 6 2 6 5 5 5 6 7
population
51-100 1 6 3 4 4 5 5 1 6 4 4 4 5 6 2 7 5 6 6 7 7

>=100 1 6 4 4 4 5 5 1 6 4 4 5 6 6 2 7 5 6 6 7 7

SLT = Static Load Test; CAPWAP = Case Pile Wave Analysis Program; PDA = Case Method Using the Pile Driving Analyzer; EA = Paikowsky’s Energy

Approach; EOD = End of Driving; BOR = Beginning of Restrike.

t = n t 1 t 2 ⋅ ⋅ ⋅ t n , where tn is the ratio of nth measured capacity of pile test to the factored design load.

49
 0.1
Lognormal distribution
Normal distribution

0.01
Tail area

0.001

0.0001
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Deviation from mean (σ)
Figure 3-1 Tail Area of the Normal and Lognormal Distributions

50
 6.00
n = No. of test piles n=1
5.00

4.00
n=2
3.00
t

n=5
2.00
n=10
n=50
1.00

0.00
0 1 2 3 4 5 6
Reliability Index

Figure 3-2 Reliability Index and t with Different n

51
 3.50

3.00 Static Load Test

CAPWAP EOD
CAPWAP BOR
EA EOD
2.50 EA BOR
PDA EOD
PDA BOR

2.00
t

1.50

1.00

0.50

0.00
0 2 4 6 8 10
Number of load tests
Figure 3-3 Relationship between t and Number of Load Tests at Target Reliability
Index β =2.33

52
 4.50

4.00
Static Load Test
CAPWAP EOD
3.50 CAPWAP BOR
EA EOD
EA BOR
3.00 PDA EOD
PDA BOR

2.50
t

2.00

1.50

1.00

0.50

0.00
0 2 4 6 8 10
Number of load tests
Figure 3-4 Relationship between t and Number of Load Tests at Target Reliability
Index β =3.00

53
 3.00

total number of piles N=10

total number of piles N=20
2.50 total number of piles N=50
total number of piles N=100
total number of piles N=infinite

2.00
t

1.50

1.00

0.50
0 2 4 6 8 10 12
Number of Load Test
Figure 3-5 Effect of Finite Population on Different Total Numbers of Load Tests

54
 CHAPTER IV

INCORPORATING SET-UP INTO RELIABILITY-BASED DESIGN OF DRIVEN

PILES IN CLAY

4.1 Introduction

The fact that axial capacity of a driven pile may change over time after initial pile

installation has been reported by a number of geotechnical engineers for many years.

The significance of time effects on pile capacity, however, is dependent upon the type

of soil which piles are driven into. The increase of pile capacity with time is usually

referred to as set-up, while the decrease of pile capacity with time is often termed as

relaxation. Due to pile driving, soils around the pile are significantly disturbed and

remolded, and excess pore pressures are generated in saturated cohesive soils. With

passage of time, the excess pore pressure will dissipate and consequently pile capacity

is regained. Decrease in excess pore pressure is inversely proportional to the square of

the distance from the pile (Pestana et al. 2002). The time to dissipate excess pore

pressure is proportional to the square of the horizontal pile dimension (Holloway and

Beddard 1995; Soderberg 1961), and inversely proportional to the soil’s horizontal

coefficient of consolidation (Soderberg 1961). Accordingly, Larger-diameter piles take

longer to set-up than smaller-diameter piles (Long et al. 1999; Wang and Reese 1989).

On a percentage-basis, piles driven into clay generally tend to experience greater set-up

55
than piles driven into sand and silt. Piles may experience relaxation when driven into

dense and saturated sand and silt (Long et al. 1999). Numerous geotechnical engineers

developed various empirical formulas to predict the set-up behavior (e.g., Skov and

Denver 1988; Svinkin et al. 1994; Huang 1988). In particular, the semi-logarithmic

empirical relationship, proposed by Skov and Denver, has been widely used to predict

post-installation pile capacity increase with time.

With an accumulation of more experience and knowledge on set-up phenomenon,

some researchers have suggested that the set-up be formally incorporated into the

prediction method to determine total pile capacity. For example, Bullock et al. (2005)

proposed a conservative method for incorporating side shear set-up into the total pile

capacity. The predicted set-up capacity was assumed to have the same degree of

uncertainties as the measured reference capacity and a single safety factor was used to

account for all uncertainties of loads and resistances. Due to different uncertainties

associated with measured capacity and predicted set-up capacity, Komurka et al. (2005)

proposed a method to apply separate safety factors to EOD (End of Driving) and set-up

components of driven pile capacity. Furthermore, the set-up capacity was characterized

as a function of pile penetration based on dynamic monitoring during both initial

driving and restrike testing.

This chapter presents a database containing a large number of pile testing data in

clay, from which the set-up effect is statistically analyzed. The compilation of the

statistical database of set-up capacity makes it possible to apply reliability-based

56
analysis techniques to develop separate resistance factors to account for different

degrees of uncertainties associated with the measured reference capacity and the

predicted set-up capacity in the LRFD (Load and Resistance Factor Design) of driven

piles. The First-Order Reliability Method (FORM) is used to determine the separate

resistance factors for various target reliability levels. The test results obtained during a

convenient short-time restrike (usually 1 day after initial pile driving) can be used to

predict the long-term capacity based on the proposed formula by Skov and Denver

(1988). The application of separate resistance factors in LRFD for the measured

reference capacity and the predicted set-up capacity can remarkably improve the

prediction of driven pile capacity. Furthermore, by incorporating pile set-up capacity,

pile length or number of piles can be reduced and economical design of driven piles can

be achieved.

4.2 Statistical Study of Pile Set-up

Pile set-up has been widely observed in piles driven into saturated clays. Due to pile

driving, soils around the pile are significantly disturbed and remolded, and excess pore

pressures are generated. With passage of time, the excess pore pressure dissipates and

consequently pile capacity is increased. Based on observations in the field, some

researchers have developed predictive equations for the time-dependent increase in pile

capacity. Skov and Denver (1988) presented an empirical equation for set-up based on a

logarithmic increase of pile capacity with time as follows

t
Qt = Q0 (1 + A log ) (4-1)
t0

57
where

Qt = axial capacity at time t after driving,

Q0 = axial capacity at time t0,

A = a factor that is a function of soil type and can be determined as the slope of the

linear portion of the normalized capacity Qt/Q0 versus log(t) plot,

t = time since pile installation, and

t0 = time after installation at which point the capacity gain becomes linear on a log(t)

plot.

Based on Eq. (4-1), the set-up capacity can be expressed as

t
Qset −up = Q0 A log (4-2)
t0

Skov and Denver also recommended numerical values for A and t0 as 0.6 and 1 day in

clay and 0.2 and 0.5 days in sand, respectively. Reasearchers, such as Svinkin et al.

(1994), Camp and Parmar (1999), and Bullock et al. (2005), have also provided a range

of numerical values for these two parameters.

Because of the highly disturbed state of the soil, there is a brief period following

installation during which the initial rate of capacity gain with time is different from the

later rate. During this period, the affected soil experiences an increase in effective and

horizontal stress, consolidates, and gains strength in a manner which is not well-

understood and is difficult to model and predict. After the time t0, the affected soil

experiences an increase in effective vertical and horizontal stress, consolidates, and

58
gains shear strength according to conventional consolidation theory. The time t0 is a

function of soil type, permeability, and sensitivity and pile type, permeability, and size.

The less permeable the soil and pile and the greater volume of soil displaced by pile

driving, the longer the time t0. The A parameter is a function of soil type, pile material,

type, size, and capacity to account for the degree of the capacity gain with the elapse of

time (Camp and Parmar 1999; Svinkin et al. 1994; Svinkin and Skov 2000). Although

there should be physical meanings behind these parameters, both the parameters t0 and

A are back-calculated from field data, or gleaned from empirical relationships in the

literatures (Skov and Denver 1988; Svinkin et al. 1994; Chow 1998; Axelsson 1998;

Bullock 1999; Camp and Parmar 1999; Bullock et al. 2005). It should be noted that

determination of A is a function of the value used for t0, and visa-versa; these two

variables are not independent (Bullock 1999). On a percentage-basis, piles driven into

clay generally tend to experience greater set-up than piles driven into sand and silt

(Long et al. 1999); therefore, the focus of this chapter is on piles driven into clay.

A database containing both dynamic pile and static load tests is compiled in Table

4-1. The table includes information about the test piles and source of references. For

each pile in the table, at least two load tests at different time after pile installation

(whether dynamic or static methods) are performed so that the development of set-up

capacity can be quantified. The database is statistically analyzed to compare with the

time-capacity relationship proposed by Skov and Denver (1988).

59
 Based on the compiled database, most of pile testing data show that the pile capacity

gain becomes approximately linear with the logarithm of time 1 day after pile

installation. Researchers, such as Svinkin et al. (1994), Alexsson (1998), McVay (1999),

and Bullock et al. (2005), have proposed the use of t0 = 1 day. Using t0 = 1 day, the

normalized axial pile capacity (i.e., Qt/Q0) versus the logarithm of time is plotted in Fig.

4-1 for all cases in the database. The time corresponding to the pile capacity measured

at EOD is taken as 0.01 day in the plot. The dashed line defines an approximate range

of A from 0.1 to 1.0. The mean value of A for all 37 testing piles is close to the value of

0.5. Chow (1998) reported the values of A ranging from 0.25 to 0.75. Studies by

Axelsson (1998) yielded the values of A ranging from 0.2 to 0.8. The observations from

this compiled database seem to conform with the previous experiences. Therefore, the

reference time t0 = 1 day and A = 0.5 are used for the subsequent statistical analyses in

this chapter. When further data becomes available, the values of t0 and A could be

modified. It is interesting to note that, in some cases, the axial capacity after installation

reaches 8 times the axial capacity at EOD. A couple of cases show that pile capacity

increases linearly with the logarithm of time until 100 days after driving. After time

elapse of 100 days, the increase of pile capacity with time becomes less pronounced

compared to those during the first 100 days after driving. The compiled database

supports the general notion that 100 days after EOD can be taken as the time after

which the set-up effect would be minimal.

The accuracy of the Skov and Denver Equation is analyzed by examining the ratio

of the measured set-up capacities to the predicted using t0 = 1 day and A = 0.5. The

60
frequency distribution for all 127 cases of 37 piles is shown in Fig. 4-2. It can be seen

that a normal distribution seems to approximately represent the distribution. The mean

value μ, standard deviation σ, and Coefficient of Variation, COV, of the fitted normal

distribution are 1.141, 0.543 and 0.475, respectively. To verify the assumed theoretical

distributions, the Kolmogorov-Smirnov test (Ang and Tang 1975) was carried out. The

maximum differences in cumulative frequency between the observed data and the

theoretical distributions are Dn = 0.058 that is smaller than the critical value Dnα = 0.121

at the 5% significance level. Therefore, the assumed normal distribution for the

predicted set-up capacity is valid. However, the current use of a normal distribution to

describe the statistical characteristics of the predicted set-up capacity does involve

limitations. For a normal random variable with mean = 1.141 and standard deviation =

0.543, the probability of realizing negative values is about 2%. The database shown in

this chapter, on the other hand, fails to detect any negative value. Therefore, it is

expected that, with the accumulation of additional pile test data, the fitted probability

model and the relevant statistical parameters may require further modifications if a

larger dataset still fails to detect negative value.

4.3 Correlation between Reference Capacity Q0 and Set-up Capacity Qset-up

A statistical analysis result on CAPWAP EOD and BOR (Beginning of Restrike)

methods by Paikowsky et al. (2004) is summarized in Table 4-2. It was found that the

time of testing (EOD vs. BOR) significantly affected the accuracy of CAPWAP

prediction, regardless of soil type (Paikowsky and Stenerson 2000; Paikowsky et al.

2004). The BOR statistics included only the last blow of BOR for all the case histories,

61
and the time of the dynamic test is generally close to the time of static load tests.

Therefore, CAPWAP BOR results were used to evaluate the accuracy of CAPWAP

approach. Comparing the statistical parameters, COV of CAPWAP EOD and CAPWAP

BOR, one can see that CAPWAP BOR provides better accuracy than CAPWAP EOD.

The statistical database in Fig. 4-1 shows that the pile capacity at BOR is significantly

larger than that at EOD for piles in clay. Therefore, the factored capacity from EOD test

result is much less than that from BOR test result for the same pile driven into clay. The

unreasonableness of different factored capacity for the same pile can be attributed to

higher COV value for EOD, caused primarily by more varied set-up effects during the

time period from EOD to the static load test. The database of pile set-up in clay

supports the notion that predicting pile set-up effect involves higher uncertainties than

the measured CAPWAP results. Therefore, it is necessary to separate the resistance

factors to account for the different degrees of uncertainties associated with the measured

capacity and the predicted set-up capacity.

The relationship between the measured reference capacity Q0 and the predicted pile

set-up capacity Qset-up is not at all clear. However, since the uncertainties of pile set-up

prediction are contributed by empirical relationship of Skov and Denver, it may be

reasonable to assume that the measured CAPWAP results and the prediction of set-up

capacity are independent of each other. To verify this assumption, a total of 16 piles in

the database (each has 2 or more CAPWAP dynamic testing results and 1 static load test

result) is used to investigate the correlation between the measured CAPWAP results and

62
the predicted set-up capacity. The correlation coefficient ρ, based on Eq. (4-3), is

calculated as -0.243.

Cov( X , Y ) E[( X − μ X )(Y − μ Y )]

ρ= = (4-3)
σ XσY σ XσY

where μX and μY = means of the random variables X and Y, respectively; σX and σY =

standard deviation of X and Y, respectively. In this case, X = the measured reference

capacity from CAPWAP method; Y = the predicted set-up capacity from Eq. (4-2). It

can be seen that the correlation coefficient between Q0 and Qset-up is very small, thus

supporting the notion that Q0 and Qset-up can be assumed to be independent of each other.

In this chapter, the uncertainties associated with dead and live loads as well as the

measured reference capacity Q0 and the predicted set-up capacity Qset-up are systemically

accounted for in a framework of First-Order Reliability Method (FORM). Separate

resistance factors are derived for Q0 and Qset-up for adoption in the Load and Resistance

Factor Design (LRFD) of driven piles.

4.4 First-Order Reliability Method (FORM)

Load and Resistance Factor Design (LRFD) method for foundations has evolved

beyond First-Order Second Moment (FOSM), which is being used in the existing

AASHTO specifications to calibrate the resistance factors, to the more invariant FORM

approach. Phoon et al. (2003a and 2003b), Paikowsky et al. (2004), and Allen et al.

(2005) have proposed the application of FORM approach to calibrate resistance factors

for deep foundation. Ang and Tang (1984) provide detailed descriptions about the

procedure of using FORM to calibrate LRFD partial safety factors. FORM can be used

63
to assess the reliability of a pile with respect to specified limit states as well as to

calculate partial safety factors φi and γi for resistances and loads, respectively, for a

chosen target reliability index β. In design practice, the limit state can be represented by

a performance function of the form,

g ( X ) = g ( X 1 , X 2 ,..., X n ) (4-4)

in which X=(X1, X2,…,Xn) is a vector of basic random variables of strengths and loads.

The limit is defined as g(X)=0, implying failure when g(X)<0. When performing

FORM analysis, it can be stated that a measure of safety in the form of the reliability

index β is defined as the shortest distance from the origin to the failure surface (limit

state) in the reduced coordinates (Ayyub and McCuen 2003). The reliability index

according to this definition is shown in Fig. 4-3 for a nonlinear performance function g.

The point on the failure surface that corresponds to the shortest distance is called the

design (or failure) point at which the load and resistance factors can be respectively

determined for the components of load and resistance in the reliability-based design.

Iterative steps for determining β using FORM (Ang and Tang 1984) are outlined

herein:

1. Assume an initial value for the design point. It is common to start with the mean

values of the basic random variables. The design point in the reduced coordinates

should then be computed as

xi* − μ X i
x =
'*
(4-5)
σX
i
i

64
where μ X i = mean value of the basic random variable Xi, σ X i = standard deviation of

the basic random variable Xi, the notations x* and x’* are used to denote the design point

in the regular coordinates and in the reduced coordinate system, respectively.

2. If the distribution of basic random variables is non-normal, approximate this

distribution with an equivalent normal distribution at the design point, having the same

tail area and ordinate of the density function with equivalent mean,

μ XN = x * − Φ −1 ( FX ( x*))σ XN (4-6)

and equivalent standard deviation

φ (Φ −1 ( FX ( x*)))
σ XN = (4-7)
f X ( x*)

where μ XN = mean value of the equivalent normal distribution, σ XN = standard deviation

of the equivalent normal distribution, FX ( x*) = original cumulative distribution function

(CDF) of Xi evaluated at the design point, f X ( x*) = original PDF of Xi evaluated at the

design point, Φ (•) =CDF of the standard normal distribution, and φ(•) = PDF of the

standard normal distribution.

3. Set x i'* = α*iβ , in which the α *i are direction cosines. Compute the directional

cosines ( α *i , i = 1, 2,…, n) using,

∂g
( )*
∂xi'
α i* = for I =1,2,…,n (4-8)
n
∂g
∑ ( ' ) *2
i =1 ∂xi

where

65
 ∂g ∂g
( ) = ( ) * σ XNi
' *
(4-9)
∂x i ∂x i

4. With α *i , μ XNi , σ XNi now known, the following equation is solved for β:

g[( μ XN1 − α *X1 σ XN1 β ),..., ( μ Xn

N
− α *Xnσ Xn
N
β )] = 0 (4-10)

5. Use the β obtained from step 4, a new design point is obtained from

xi* = μ XNi − α *X i σ XNi β (4-11)

6 Repeat steps1 to 5 until convergence of β is achieved.

4.5 Calibrate Partial Load and Resistance Factors Based on FORM

The database shows that driven piles in clay experience set-up. The magnitude of

set-up is so significant that economical design of driven piles is not possible without

recognition of the anticipated capacity gain with time. Within the last decade, dynamic

pile testing during restrikes for capacity determination has been adopted in practice to

account for the set-up. However, the set-up may continue 100 days after driving. Thus,

the restrike capacity measured at 7 days or less after initial pile installation is likely to

underpredict the actual capacity. Therefore, it would be advantageous to develop a

method to enable predication of long-term capacity based on the test data obtained

during short-term restrike.

When the set-up capacity is incorporated into the driven pile design, the strength

limit state at the failure point can be expressed as

g = R * − L* = Q0* + Qset
*
− up − L D − L L = 0
* *
(4-12)

66
where R* = total resistance, L* = total load, Q0* = measured pile axial capacity at

* *
reference time t0, Qset −up = predicted pile set-up capacity after time t0, LD = dead loads,

and L*L = live loads. The probabilistic characteristics of the random variables LD and LL

are well documented in AASHTO (Nowak 1999). The probabilistic characteristics of

the random variables Q0 is taken as that of CAPWAP BOR presented by Paikowsky et

al. (2004). The probabilistic characteristics of the random variables Qset-up is assumed as

a normal distribution:

1 1 x−μ 2
f ( z) = exp[− ( ) ] (4-13)
2π σ 2 σ

where μ = 1.141 and σ = 0.543 in Fig. 4-2, based on Eq. (4-2) (A = 0.5 and t0 = 1 day).

Table 4-3 summarizes the statistical parameters and distribution function for each

random variable. With the known load and resistance statistical characteristics, the

iterative solution of FORM approach can be applied to determine the load and

resistance factors for the chosen target reliability index (Ang and Tang 1984; Phoon et

al. 2003a and 2003b; Ayyub and McCuen 2003; Paikowsky et al. 2004).

Based on AASHTO LRFD Highway Bridges Design Specifications (2003), the ratio

of dead to live load is a function of a bridge’s span length. The following values of

LD/LL are recommended as: 0.52, 1.06, 1.58, 2.12, 2.64, 3.00, and 3.53 for span lengths

of 9, 18, 27, 36, 45, 50, and 60 m, respectively. The relationship between the partial

safety factor (i.e., load and resistance factors) and the target reliability index is shown in

Fig. 4-4, where the ratio of dead load to live load is selected as 1.58:1 to represent a

bridge span length of 27m, and the ratio of Q0 to Qset-up is 1:1 based on the Skov and
67
Denver equation with A = 0.5 and t =100 days. The partial safety factors for all loads

and resistances vary approximately linearly with the target reliability index. The

resistance factors for Q0 are higher than those for Qset-up at a given reliability level

because uncertainties for Q0 are less than those for Qset-up. The fact that COV of two

loads are much smaller than those of two resistances has resulted in smaller variation of

load factors with the target reliability index for the live and dead load in Fig. 4-4. The

variance of target reliability index imposes the largest effect on the resistance factor for

the set-up capacity with the highest uncertainty (the highest COV value). Fig. 4-4 shows

that the percentage of the capacity component with the higher uncertainty (i.e., Qset-up),

which can be utilized in design to resist the load, decreases with an increase of the

desired target reliability index. The negative value for the resistance factor of Qset-up in

Fig. 4-4 does not mean that the consideration of set-up effect will reduce the total

factored design capacity. It can be interpreted as the most possible failure point at which

the combination of Q0 and Qset-up is optimized to resist the given load effects. The same

statistics of load and resistance components can be used to determine the relationship

between partial safety factors and reliability index in Fig. 4-5, where the resistance

components only include Q0 without considering the set-up effects. Compared to the

approximately invariant load factors with the change of target reliability index for the

case of considering set-up effect, the load factors for the case of not considering set-up

effects vary more pronouncedly. This may be attributed to the fact that the resistance

component (i.e., Q0) possesses the relatively similar COV values with the load

components. When the ratio of dead load to live load is considered as 1.58:1 and the

nominal live load is assumed as 1.0, Fig. 4-6 shows that the required resistances Q0 to

68
satisfy the predefined structural reliability for the cases with and without considering

the set-up effects, respectively. It can be seen that although the resistance factor of Qset-

up is negative for the target reliability index β=2.35 in Fig. 4-4, the required Q0 with

considering the set-up effect is still smaller than that without considering the set-up

effect until the target reliability index β=3.00 is chosen as shown in Fig. 4-6. It means

that, when the target reliability index is low (β<3.00), the incorporation of set-up effect

into the design of driven piles can advantageously enhance the prediction of design

capacity so that shorter pile length or less number of piles can be chosen to achieve

economical design. When the reliability index is greater than 3.00, the incorporation of

set-up effect can no longer improve the prediction of design capacity of driven piles due

to higher uncertainties associated with the prediction of set-up capacity.

4.6 Practical Load and Resistance Factors for Piles Driven into Clay

For a given reliability index β and probability distribution for resistance and load

effects, the partial safety factor determined by the FORM approach may differ for

different failure modes. For this reason, calibration of load and resistance factors is

important in order to maintain the same values of load factors for all loads at different

failure modes. In order to be consistent with the existing experience in LRFD, the

calibrations of resistance factor for set-up is performed for a set of load factors already

specified in the structural code. The resistance factors of Q0 for CAPWAP approach are

taken as the same as those recommended in NCHRP Report 507. When incorporating

the set-up effect into the design of driven pile, the LRFD criterion can be expressed as

φ 0 Q0 + φ set −up Q set −up ≥ γ L L L + γ D L D (4-14)

69
where φ 0 and φset-up = resistance factors for reference resistance at t0 and set-up

resistance, respectively; γL and γD = load factors for live and dead loads, respectively.

The relationship between the reliability index and the probability of failure can be

expressed as

β = Φ −1 (1 − Pf ) (4-15)

where Φ-1 is the inverse of the cumulative distribution function of the standard normal

distribution and Pf is probability of failure. The target reliability index increases

approximately exponentially with a decrease of probability of failure. As suggested in

NCHRP Report 507 (Paikowsky et al. 2004), the target reliability index of β=2.33

corresponding to the probability of failure Pf = 1% is recommended for redundant piles

defined as 5 or more piles per pile cap; the target reliability index of β=3.00

corresponding to the probability of failure Pf = 0.1% is recommended for nonredundant

piles defined as 4 or fewer piles per pile cap. The resistance factors are φ0 = 0.65 and φ0

= 0.50 for CAPWAP approach at β=2.33 and β=3.00, respectively. As specified by

AASHTO (Novak 1999), the load factors γL=1.75 and γD=1.25 are used for live and

dead loads, respectively. The above load and resistance factors can be used to determine

the resistance factor φ set-up for predicted set-up capacity in the following algorithm

based on the FORM approach.

γ LμL γ DμL φ0 μ Q
L
+ D
− 0

λL λL λQ
φ set −up = l D 0
(4-16)
μQ set − up

λQ set − up

70
where μ LL , μ LD , μ Q0 , and μ Qset −up are the mean values of random variables LL, LD, Q0,

and Qset-up, respectively, calculated from FORM approach; λLl , λLD , λQ0 , and λQset−up are

the bias factors of random variables LL, LD, Q0, and Qset-up, respectively.

Fig. 4-7 shows the effect of the variation of the ratio of dead to live load on the

calculated resistance factors for set-up capacity at the reliability level β=2.33 and

β=3.00, respectively. The load factors for dead and live load and the resistance factor

for Q0 are fixed as mentioned earlier when calculating the resistance factor for Qset-up. It

can be seen that the calculated resistance factor for Qset-up decreases slightly with an

increase in the ratio of dead to live load. In general, however, the calculated resistance

factor for Qset-up is not very sensitive to the ratio of dead to live load. The calculated

resistance factor for Qset-up is approximately equal to 0 at the target reliability index

β=3.00. It is because that, at higher reliability level, the higher uncertainty associated

with the set-up effect results in its lower contribution to the total pile capacity. It means

that the set-up effect could be ignored in driven pile design at target reliability index

β=3.00 when the predefined resistance factor for Q0, dead load factor, and live load

factor are 0.50, 1.25, and 1.75, respectively. The resistance factor for Qset-up can be

conservatively taken as 0.3 for the span length less than 60 m at target reliability index

β=2.33, when the predefined resistance factor for Q0, dead load factor, and live load

factor are 0.65, 1.25, and 1.75, respectively. The recommended partial safety factors are

tabulated in Table 4-4 for different reliability levels. It should be noted that the

recommended resistance factors for Qset-up in Table 4-4 are calibrated based on the

compiled database presented in this chapter. If additional pile test data could be
71
obtained to supplement the current database, the recommended resistance factors for

Qset-up should be modified according to the potential change of the probability model

and the relevant statistical parameters for the updated database.

The corresponding F.S (Factor of Safety) in ASD (Allowable Stress Design) can be

calculated as follows:

γ D LD / LL + γ L
LD / LL + 1
F .S . = (4-17)
φ 0 Q0 / Qset −up + φ set −up
Q0 / Qset −up + 1

Fig. 4-8 shows the relationship between the corresponding F.S. in ASD and the ratio of

dead to live load. The corresponding F.S. decreases slightly with an increase of the ratio

of dead to live load. The corresponding F.S. is about 3.00 and 5.80 for the

recommended load and resistance factors at target reliability index β=2.33 and β=3.00,

respectively.

4.7 Proposed Procedure to Incorporate Set-up into LRFD of Driven Piles in Clay

The incorporation of set-up effect at the lower target reliability level can efficiently

improve the estimation of capacity of driven piles so that pile length or numbers of piles

can be economically reduced. The proposed design procedure for incorporating set-up

into LRFD of driven piles in clay is illustrated as follows:

1. Assign a pile size, length, and type for preliminary design using available site

information and a static analytical design method.

72
 2. Perform CAPWAP dynamic testing 1 day after pile installation and record the

measured capacity as Q0.

3. Estimate the set-up capacity based on the proposed semi-logarithmic empirical

t
relationship: Qset −up = Q0 A log (where A = 0.5, t = 100 days, and t0 = 1 day).
t0

4. Use the recommended resistance and load factors in Table 4-4 and the following

formula, φ 0 Q0 + φ set −up Q set −up ≥ γ L L L + γ D L D , to evaluate the reasonableness of

the preliminary design.

5. Based on the result from step 4, change the pile length or number as desired to

optimize the design.

4.8 Conclusions

A database of pile set-up capacity has been compiled and presented in this chapter.

The compiled database showed that the logarithmic empirical relationship proposed by

Skov and Denver (1988) could be used to predict pile set-up. The reference time t0 = 1

day and mean value A = 0.5 were found to be optimum for the prediction equation. The

pile set-up database indicated that the time duration of 100 days after EOD may be

considered as the point after which the set-up effect would be diminished. The normal

distribution was shown to adequately represent the probabilistic characteristics of the

predicted pile set-up capacity. The mean value μ, standard deviation σ, and COV of the

fitted normal distribution are 1.141, 0.543 and 0.475, respectively. The calculated

correlation coefficient between the measured reference pile capacity, Q0, and the

73
predicted set-up capacity, Qset-up, suggested that these two pile capacity components

could be treated as if they were statistically independent.

The statistical parameters of the pile set-up capacity derived from the compiled

database, together with previous statistical analysis of CAPWAP approach (Paikowsky

et al. 2004), were systematically incorporated within the framework of FORM to

develop separate resistance factors using the load conditions specified in AASHTO

(Novak 1999). The incorporation of set-up effect into the prediction of total pile

capacity gives advantageous contribution to the estimated total pile capacity at a low

target reliability index (β<3.00). However, due to higher uncertainty associated with the

predicted pile set-up capacity, the incorporation of the set-up capacity into the pile

design yields more conservative prediction on the total pile capacity at higher reliability

index (β≥3.00). To be consistent with the existing experience about the load and the

resistance effects, the load factors and resistance factors for Q0 are pre-set in the present

chapter when performing FORM analysis to determine the resistance factors for the pile

set-up capacity. The set-up effect could be ignored in driven pile design if a target

reliability index β=3.00 is chosen and the pre-set resistance factor for Q0, dead load

factor, and live load factor are 0.50, 1.25, and 1.75, respectively. The resistance factor

for Qset-up can be conservatively taken as 0.3 for a bridge span length less than 60 m if a

target reliability index β=2.33 is chosen and the pre-set resistance factor for Q0, dead

load factor, and live load factor are 0.65, 1.25, and 1.75, respectively. The

corresponding F.S. in ASD is about 3.00 and 5.80 for the recommended load and

74
resistance factors in LRFD at the target reliability index β=2.33 and β=3.00,

respectively.

75
Table 4-1 Summary of Load Test Database for Driven Piles in Clay

Pile series # Source of references Paper pile # Pile type

1 Paikowsky et al. (2004) FN1 HP 10x42
2 Paikowsky et al. (2004) FL3 PSC24”sq
3 Paikowsky et al. (2004) CHA1 CEP 12.75”
4 Paikowsky et al. (2004) CHB2 HP 12x63
5 Paikowsky et al. (2004) CHB3 HP 12x63
6 Paikowsky et al. (2004) CH39 CEP9.63”
7 Paikowsky et al. (2004) LB3 PSC24”sq
8 Paikowsky et al. (2004) LB4 PSC30”sq
9 Paikowsky et al. (2004) LB5 PSC30”sq
10 Paikowsky et al. (2004) LB6 PSC36”cyl
11 Paikowsky et al. (2004) LB7 PSC36”cyl
12 Paikowsky et al. (2004) NBTP2 HP 12x74
13 Paikowsky et al. (2004) NBTP3 HP 12x74
14 Paikowsky et al. (2004) UMLNB2 CEP 324mm
15 Paikowsky et al. (2004) UMLNB3 PSC 356mm sq
16 Skov and Denver (1988) Case1 P9 Concrete 25x25 cm2
17 Skov and Denver (1988) Case1 P3 Concrete 25x25 cm2
18 Skov and Denver (1988) Case1 P5 Concrete 25x25 cm2
19 Skov and Denver (1988) Case1 P6 Concrete 25x25 cm2
20 Long et al. (1999) Pile 70 Close-end pipe
21 Long et al. (1999) Pile 71 Close-end pipe
22 Long et al. (1999) Pile 72 Wood
23 Long et al. (1999) Pile 101 Timber
24 Long et al. (1999) Pile 102 Timber
25 Long et al. (1999) Pile 103 Reinf Concrete
26 Long et al. (1999) Pile 104 Reinf Concrete
27 Long et al. (1999) Pile 105 Reinf Concrete
28 Long et al. (1999) Pile 106 NP30 Steel Grdr
29 Long et al. (1999) Pile 107 Timber, Box
30 Long et al. (1999) Pile 108 Timber, Box
31 Long et al. (1999) Pile 109 Reinf Concrete
32 Long et al. (1999) Pile 110 Reinf Concrete
33 Long et al. (1999) Pile 111 Capped Pie-pile
34 Long et al. (1999) Pile 112 Capped Pie-pile
35 Long et al. (1999) Pile 113 Capped Pie-pile
36 Long et al. (1999) Pile 114 Capped Pie-pile
37 Long et al. (1999) Pile 115 Monotube

76
Table 4-2 Statistical Details and Resistance Factors of CAPWAP EOD and BOR

Method No. of cases Bias factor λ COV Φ (β=2.33) Φ (β=3.00)

CAPWAP EOD 125 1.626 0.490 0.64 0.46

CAPWAP EOD

AR<350 & Bl. Ct. 37 2.589 0.921 0.41 0.23

<16 BP10cm

CAPWAP BOR 162 1.158 0.339 0.65 0.51

EOD = end of driving; BOR = beginning of restrike; AR= ratio of surface area in contact with soil to
area of pile tip; Bl. Ct. = blow count; BP10cm =blows per 10cm

77
Table 4-3 Probabilistic Characteristics of Random Variables of Loads and Resistances
for Piles Driven into Clay
Random Bias factor λ Standard Coefficient of Distribution
variables deviation σ variation COV type
Q0 1.158 0.393 0.339 Lognormal
Qset-up 1.141 0.543 0.475 Normal
LL 1.150 0.230 0.200 Lognormal
LD 1.050 0.105 0.100 Lognormal

78
Table 4-4 Recommended LRFD Partial Safety Factors for Driven Piles in Clay
Random Bias Coefficient of Partial safety Partial safety
variables factor λ variation COV factor (β=2.33) factor (β=3.00)
Q0 1.158 0.339 0.65 0.50
Qset-up 1.242 0.474 0.30 0.00
LL 1.150 0.200 1.75 1.75
LD 1.050 0.100 1.25 1.25

79
 4.0

A=1.0
Series1 Series2
3.5 Series3 Series4
Series5 Series6
Series7 Series8
Normalized Pile Capacity Qt/Q0
Series9 Series10
Series11 Series12
3.0 Series13 Series14
Series15 Series16
Series17 Series18
Series19 Series20
Series21 Series22
2.5 Series23 Series24
Series25 Series26
Series27 Series28 A=0.5
Series29 Series30
Series31 Series32
2.0 Series33 Series34
Series35 Series36
Series37 A=0.1
A=0.5 A=1.0
1.5

A=0.1
1.0

0.5

0.0
0.01 0.1 1 10 100 1000
Time After EOD (days)
Figure 4-1 Variation of Normalized Capacity with Time for Driven Piles in Clay

80
 0.16
Relative Frequency of Number of Cases

0.14

0.12

0.10
Normal distribution
μ=1.141 σ=0.543
0.08

0.06

0.04

0.02

0.00
0 1 2

Ratio of Measured Set-up Capcacity over

Predicted Set-up Capacity from Eq. [1] (A=0.5)

Figure 4-2 Frequency Distribution of the Ratio of Measured Set-up Capacities to the
Predicted

81
Reduced Load Effect, L’ Failure Region Failure Surface
g=0
Design
Point

Survival Region

Reliability
Index

Reduced Strength, R’

Figure 4-3 Performance Space in Reduced Coordinates

82
 1.4
1.2
1
Partial Safety Factors

0.8
0.6
0.4
0.2
Resistance factor for Qo
0 Resistance factor for Qset-up
Load factor for live load
-0.2 Load factor for dead load

-0.4
0 0.5 1 1.5 2 2.5 3 3.5
Target Reliability Index β

Figure 4-4 Relationship between Partial Safety Factor and Target Reliability Index
When Incorporating Set-up into the Design of Driven Piles

83
 1.4

1.2
Partial Safety Factors

0.8

0.6
Resistance Factor for Qo
0.4
Load Factor for Live Load
Load Factor for Dead Load
0.2

0
0 0.5 1 1.5 2 2.5 3 3.5
Target Reliability Index β

Figure 4-5 Relationship between Partial Safety Factor and Target Reliability Index
without Considering Set-up in the Design of Driven Piles

84
 9

8
Required Resistance Q0 to Satisfy Considering set-up effect
Predifined Structural Reliability Not considering set-up effect
7

0
0 1 2 3 4
Target Reliability Index β
Figure 4-6 Comparison between Considering and Not Considering Set-up Effect

85
 0.6

0.5
Resistance factor for Qset-up

0.4
β=2.33

0.3

0.2

0.1
β=3.00
0.0

-0.1

-0.2
0 1 2 3 4
Ratio of Dead Load to Live load
Figure 4-7 Variation of Resistance Factor for Qset-up in Clay versus the Ratio of Dead
Load to Live Load

86
 7.0

Corresponding F.S (Factor of Safety)

6.0

5.0 β=3.00

4.0

3.0
β=2.33
2.0

1.0

0.0
0 0.5 1 1.5 2 2.5 3 3.5 4

Ratio of Dead to Live Load

Figure 4-8 Variation of the F.S versus the Ratio of Dead Load to Live Load for Piles
Driven into Clay

87
 CHAPTER V

INCORPORATING SET-UP INTO RELIABILITY-BASED DESIGN OF DRIVEN

PILES IN SAND

5.1 Introduction

Set-up is defined as an increase in bearing capacity with time that takes place after

pile installation. It is believed that excess pore pressures induced during driving

dissipates very quickly in sand, usually within a few hours. Tavenas and Audy (1972)

first reported the well-documented cases of pile long-term set-up in sand, which was

not attributable to pore pressure changes. During the last decade, the phenomenon of

long-term set-up of piles driven in sand has been highlighted in many studies. Long-

term set-up in sand can roughly be divided into two vital but substantially different

causes, based on Chow et al. (1997, 1998) and Axelsson (1998, 2002). The first is

stress relaxation in the soil arching generated during pile driving, which leads to an

increase in horizontal effective stress on the pile shaft. The second is the soil aging

caused mainly by rearrangement and interlocking of soil particles, which results in an

increase in stiffness and dilatancy of the soil. Numerous geotechnical engineers have

developed various empirical formulas to predict the set-up behavior in sand (e.g., Skov

and Denver 1988, Svinkin et al. 1994, and Long et al. 1999). However, the beneficial

effect of long-term set-up of driven piles in sand has very seldom been utilized to any

88
significant extent in piling projects for a long time, due to high uncertainties of

predicting the development of set-up.

With the AASHTO (American Association of State Highway and Transportation

Officials) migration from ASD (Allowable Stress Design) method to LRFD (Load and

Resistance Factor Design) method for foundation design, there is a need to develop

separate resistance factors for initial capacity and set-up capacity, respectively. This

chapter presents a database containing a large number of pile testing data in sand, from

which the set-up effect is statistically analyzed. Follow the procedure proposed in

Chapter IV, the separate resistance factors are obtained to account for different degrees

of uncertainties associated with measured short-term capacity and predicted set-up

capacity in sand at various reliability levels.

5.2 A Compiled Database of Case Histories of Driven Piles in Sand

Pile set-up in sand was first reported by Tavenas and Audy (1972) and Samson and

Authier (1986). Subsequently, a number of other case histories have been reported in

the literature. Chow et al. (1998) summarized the results from 11 case histories with

long-term set-up effects for driven pile in sand, indicating that most of pile capacities

increase by around 50% (±25%) per log cycle of time from 1 day after pile driving.

Long et al. (1999) also compiled a database of set-up cases, and categorized the soil

into three main groups: clayey soil, mixed soil, and sand. The sand group contained 6

case histories; many of them are the same as in Chow et al. (1998). Axial pile capacity

was found to increase from 30% to more than 100% of the pile capacity at the end of

89
driving (EOD). Axelsson (2002) presented a database showing the set-up rate ranging

from 15% to 65% per log cycle of time. Komurka et al. (2003) provided an extensive

bibliography related to the topic of pile set-up. In general, the increase in the capacity

of driven piles in sand is believed to be due to the aging effect and the increase of

horizontal effective stress (Chow 1997 and 1998, Axelsson 1998 and 2002).

A database containing both static and dynamic load tests was collected and listed in

Table 5-1 based on the previous studies of the driven piles in sand. The database

consists of 190 loading tests in total, performed on 73 piles where at least two separate

tests at different times are performed for each pile to determine the set-up. Table 5-1

provides information about the test pile, soil conditions of the site, and the sources of

references. However, it should be noted that the database suffers from the following

shortcomings:

• There exists a difference in the definition of the reference capacity Qre. Tavenas

and Audy (1972) and Bullock et al. (2005a, 2005b) assessed Qre from tests

conducted about 0.5 days after driving. In the remaining cases, Qre was defined

as the measured capacity at EOD.

• The capacity was determined either by static or dynamic testing, but both

dynamic and static load test results could have been used in the same test series

without distinguishing them. Furthermore, the different static loading failure

criteria have been used in the references.

• It is generally believed that the set-up mainly takes place along the pile shaft.

However, the total capacity was used in this database, since it was not possible

90
 to separate the toe and shaft capacities with sufficient accuracy in many cases in

the database. The testing results from Bullock et al. (2005a, 2005b) are for

changes in shaft capacity only.

Pile axial capacities are shown to increase with time after driving in Fig. 5-1. It is

noted that, in some cases, the long-term axial capacity after the installation reaches 3-4

times the axial capacity at EOD. Most of piles experience 20%~100% set-up after the

installation. No cases of relaxation in sand were found in the compiled database.

Parsons (1966) reported the relaxation of driven piles in New York area based on the

reduction of pile penetration resistance in granular soils from EOD to BOR (beginning

of restrike), but provided no alternative explanations for the observations. Yang (1970)

also reported the relaxation of piles in sand, but did not provide the actual load test

results. York et al. (1994) argued that the relaxation could be initiated in pile group

when driving pile into saturated sand. Pile-driving vibrations densify the surrounding

soils. Subsequently, driving additional piles may cause the sand to dilate rather than

compress, resulting in the conditions for potential relaxation. However, York et al.

(1994) also proposed that set-up can occur following cessation of relaxation and result

in a net increase of pile capacity.

5.3 Empirical Relationships

Empirical relationships have been proposed for quantifying the set-up by many

researchers. Skov and Denver (1988) presented an empirical equation for the set-up

based on a logarithmic increase of pile capacity with time. Svinkin et al. (1994)

91
developed a formula predicting the set-up based on load test data on five concrete piles

in dense silty sand. Long et al. (1999) proposed an equation predicting the rate at which

the pile capacity increases with time. The equation proposed by Long et al. is very

similar to Svinkin et al.’s where the pile capacity after installation is proportional to the

time with an exponential efficiency. An alternative method of using a hyperbolic

function to predict the set-up was proposed by Bogard and Matlock (1990) and Tan et

al. (2004). Zhu (1988) presented an equation for the set-up in soft clay based on the soil

sensitivity. Huang (1988) also presented a formula predicting the development of pile

capacity in the soft soil of Shanghai, China. These empirical formulas are listed in

Table 5-2.

Among those proposed formulas, Skov and Denver’s and Svinkin et al.’s equations

are most commonly used for the prediction of the set-up in sand, since they require only

a short-term capacity to be measured as the reference capacity. Based on Skov and

Denver’s equation, the predicted set-up capacity can be expressed as

t
Qset −up = Q0 A log (5-1)
t0

where

Qset-up = predicted set-up capacity at time t after driving,

Q0 = measured axial capacity at time t0,

A = a factor that is a function of soil type and can be determined as the slope of the

linear portion of the normalized capacity Qset-up/Q0 versus log(t) plot,

t = time since pile installation, and

92
 t0 = time after installation at which the capacity gain becomes linear on a log(t) plot.

Skov and Denver also recommended numerical values for A and t0 as 0.6 and 1 day

in clay and 0.2 and 0.5 day in sand, respectively. Researchers such as Svinkin et al.

(1994), Camp and Parmar (1999), and Bullock et al. (2005b), also provided a range of

numerical values for these two parameters.

Based on Skov and Denver’s equation, the normalized axial pile capacity (i.e.,

Qt/Q0) versus the logarithm of time is plotted in Fig. 5-2 for the 47 pile cases where at

least two separate load tests are performed after EOD. The reference capacity Q0 was

measured between 0.5 days to 2 days after EOD. The capacity Qt was measured days,

weeks, and months after EOD. The dashed lines define an approximate range of A from

0.1 to 0.9. The mean value of A for all 47 testing piles is close to the value of 0.4.

Based on Svinkin et al.’s equation, the predicted set-up capacity can be expressed

as:

Qset −up = BQEOD (t 0.1 − 1) (5-2)

where

QEOD = measured pile capacity at EOD,

t = time since pile installation (>1 day),

B = 1.025 and 1.4 as lower and upper bound, respectively.

93
 The normalized axial pile capacity (i.e., Qt/QEOD) versus the time is plotted in Fig.

5-3. The dash lines define the lower and upper boundaries as proposed by Svinkin et al.

(1994). The mean value of B is close to the value of 1.25 for all 42 test piles where the

measured capacities at EOD are available.

Comparing Fig. 5-2 to Fig. 5-3, it can be seen that the set-up capacity predicted by

Svinkin et al.’s equation shows the larger scatter than that predicted by Skov and

Denver’s equation. The calculated mean value μ, standard deviation σ, and Coefficient

of Variation (COV) of the ratio of measured to predicted capacity are 1.023, 0.593, and

0.580, respectively, when Skov and Denver’s equation is used (A = 0.4, t0 = 1 days).

However, The calculated mean value μ, standard deviation σ, and Coefficient of

Variation (COV) of the ratio of measured to predicted capacity are 1.211, 0.911, and

0.752, respectively, when Svinkin et al’s equation is used (B = 1.25). The comparison

of the COV values demonstrates that Skov and Denver’s equation provides the higher

accuracy on predicting the set-up than Svinkin et al.’s equation. The higher

uncertainties associated with Svinkin et al.’s equation may be primarily due to the

selection of the reference capacity. Because of the highly disturbed state of the soil,

there is a brief period following installation during which the initial rate of capacity

gain with time is different from the later rate. During this period, the affected soil

experiences an increase in effective and horizontal stress, consolidates, and gains

strength in a manner which is not well-understood and is difficult to model and predict

(Komurka et al. 2003). The use of QEOD as the reference capacity inevitably needs to

account for the high variability of the changes of pile capacity in this period. Therefore,

94
Skov and Denver’s equation is used to predict the set-up in the following statistical

study.

5.4 Statistical Study of Pile Set-up

Based on the compiled database, most of pile testing data show that the pile

capacity gain becomes approximately linear with the logarithm of time 1 day after pile

installation. Researchers, such as Svinkin et al. (1994), Axelsson (1998, 2002), McVay

(1999), and Bullock et al. (2005a, 2005b), have proposed the use of t0 = 1 day in Skov

and Denver’s equation.

Chow (1998) reported the values of A around 0.5 (±0.25). Studies by Axelsson

(2002) yielded the average value of 0.4 for A, with a range from 0.15 to 0.65. The

observations from the compiled database in Fig. 5-2 seem to conform with the previous

experiences. Therefore, the reference time t0 = 1 day and A = 0.4 are used for the

subsequent statistical analyses in this chapter. When further data becomes available, the

values of t0 and A could be modified. Most of pile tests are performed in 30 days after

EOD in the compiled database. It can be seen that pile capacity increases approximately

linearly with the logarithm of time within 30 days after EOD. Tevenas and Audy (1972)

concluded that the ultimate pile capacity increase during the first 15 to 20 days after

EOD could be 70% higher than that observed at 0.5 day. Studies by York et al. (1994)

showed that set-up could have approached a maximum value within 15 to 25 days.

Although some piles in the compiled database exhibit set-up after months or years, the

increase of pile capacity with time after 30 days becomes less pronounced compared to

95
those during the first 30 days after driving. Therefore, 30 days after EOD can be

conservatively taken as the time after which the set-up effect in sand would be minimal.

The accuracy of Skov and Denver’s equation is analyzed by examining the ratio of

the measured set-up capacities to the predicted using t0 = 1 day and A = 0.4. The

frequency distribution for 47 piles is shown in Fig. 5-4. It can be seen that a lognormal

distribution seems to approximately represent the distribution. The mean value μlnx and

standard deviation σlnx of the fitted lognormal distribution are -0.122 and 0.538,

respectively. To verify the assumed theoretical lognormal distributions, the

Kolmogorov-Smirnov test (Ang and Tang 1975) was carried out. Fig. 5-5 shows the

empirical cumulative frequency and the theoretical cumulative distribution of the ratio

of the measured to the predicted set-up capacity. The maximum differences in

cumulative frequency between the observed data and the theoretical distributions are Dn

= 0.096 that is smaller than the critical value Dnα = 0.200 at the 5% significance level.

Therefore, the assumed lognormal distribution for the predicted set-up capacity is valid.

CAPWAP approach (Rausche et al. 1972, 1985) has been widely used for quality

control and capacity determination of driven piles. Also, most of pile testing data in the

compiled database are from CAPWAP results. Therefore, CAPWAP approach is

recommended for determining the measured short-term capacity Q0. The statistical

analysis results of the CAPWAP BOR measured capacity by Paikowsky et al. (2004)

and the predicted set-up capacity by the Skov and Denver’s equation are summarized in

Table 5-3. Comparing the COV values, it is found that predicting pile set-up effect

96
involves significantly higher uncertainties than measuring capacity Q0 by CAPWAP.

Therefore, it is necessary to separate the resistance factors to account for the different

degrees of uncertainties associated with the measured capacity and the predicted set-up

capacity. The relationship between the measured reference capacity Q0 and the

predicted pile set-up capacity Qset-up is not at all clear. As discussed in Chapter 4, it may

be reasonable to assume that the measured Q0 by CAPWAP and the predicted Qset-up are

independent of each other. In this chapter, the uncertainties associated with dead and

live loads as well as the measured reference capacity Q0 and the predicted set-up

capacity Qset-up are systemically accounted for in a framework of FORM. Separate

resistance factors are derived for Q0 and Qset-up for adoption in the LRFD of driven piles.

5.5 Calibrate Partial Load and Resistance Factors Based on FORM

The database shows that driven piles in sand experience the set-up. The magnitude

of the set-up is so significant that economical design of driven piles is not possible

without recognition of the anticipated capacity gain with time. When the set-up

capacity in sand is incorporated into the driven pile design, the strength limit state at the

failure point can be expressed as

g = R * − L* = Q0* + Qset
*
− up − L D − L L = 0
* *
(5-3)

where R* = total resistance, L* = total load, Q0* = measured pile axial capacity at

* *
reference time t0, Qset −up = predicted pile set-up capacity after time t0, LD = dead loads,

and L*L = live loads.

97
 The probabilistic characteristics of the random variables LD and LL are well

documented in AASHTO (Nowak 1999). The probabilistic characteristics of the

random variables Q0 is taken as that of CAPWAP BOR presented by Paikowsky et al.

(2004). The probabilistic characteristics of the random variables Qset-up is taken as a

lognormal distribution with mean value μlnx of -0.122 and standard deviation σlnx of

0.538 in Fig. 5-4, based on Eq. (5-1) (A = 0.4 and t0 = 1 day). Table 5-3 summarizes the

statistical parameters and distribution function for each random variable. With the

known load and resistance statistical characteristics, the iterative solution of FORM

approach (See Section 4.4 in Chapter IV) can be applied to determine the load and

resistance factors for the chosen target reliability index (Ang and Tang 1984; Phoon et

al. 2003a and 2003b; Ayyub and McCuen 2003; Paikowsky et al. 2004).

Based on AASHTO LRFD Highway Bridges Design Specifications (2006), the

ratio of dead to live load is a function of a bridge’s span length. The following values of

LD/LL are recommended as: 0.52, 1.06, 1.58, 2.12, 2.64, 3.00, and 3.53 for span lengths

of 9, 18, 27, 36, 45, 50, and 60 m, respectively. The relationship between the partial

safety factors (i.e., load and resistance factors) and the target reliability index is shown

in Fig. 5-6, where the ratio of dead load to live load is selected as 1.58:1 to represent a

bridge span length of 27 m, and the ratio of Q0 to Qset-up is 1:0.59 based on the Skov and

Denver’s equation with A = 0.4, t0 = 1 day, and t = 30 days. The partial safety factors

for all loads and resistances vary approximately linearly with the target reliability index.

The resistance factors for Q0 are higher than those for Qset-up at a given reliability level

because uncertainties for Q0 are less than those for Qset-up. The fact that COV values of

98
two load components are much smaller than those of two resistances has resulted in

smaller variation of load factors with the target reliability index in Fig. 5-6.

5.6 Practical Load and Resistance Factors for Piles Driven into Sand

For a given reliability index β and probability distributions for resistance and load

effects, the partial safety factors determined by the FORM approach may differ for

different failure modes. For this reason, calibration of load and resistance factors is

important in order to maintain the same values of load factors for all loads at different

failure modes. In order to be consistent with the current experience in LRFD, the

calibrations of resistance factor for the set-up is performed for a set of load factors

already specified in the AASHTO code (Novak, 1999). The resistance factors of Q0 for

CAPWAP approach are taken as the same as those recommended in NCHRP Report

507 (Paikowsky et al. 2004). When incorporating the set-up effect into the design of

driven pile, the LRFD criterion can be expressed as

φ 0 Q 0 + φ set − up Q set − up ≥ γ L L L + γ D L D (5-4)

where φ0 and φset-up = resistance factors for reference resistance at t0 and set-up

resistance, respectively; γL and γD = load factors for live and dead loads, respectively.

The target reliability index increases approximately exponentially with a decrease

of probability of failure. As suggested in NCHRP Report 507, the target reliability

index of β=2.33 corresponding to the probability of failure Pf = 1% is recommended for

redundant piles defined as 5 or more piles per pile cap; the target reliability index of

β=3.00 corresponding to the probability of failure Pf = 0.1% is recommended for

99
nonredundant piles defined as 4 or fewer piles per pile cap. The resistance factors are

φ0 = 0.65 and φ0 = 0.50 for CAPWAP approach at β=2.33 and β=3.00, respectively. As

specified by AASHTO (Novak 1999), the load factors γL=1.75 and γD=1.25 are used

for live and dead loads, respectively. The above load and resistance factors can be used

to determine the resistance factor φset-up for the predicted set-up capacity in the

following algorithm based on the FORM approach.

γ LμL γ DμL φ0 μ Q
L
+ D
− 0

λL λL λQ
φ set −up = l D 0
(5-5)
μQ set − up

λQ set − up

where μ LL , μ LD , μ Q0 , and μ Qset −up are the mean values of random variables LL, LD, Q0,

and Qset-up, respectively, calculated from FORM approach; λLl , λLD , λQ0 , and λQset−up are

the bias factors of random variables LL, LD, Q0, and Qset-up, respectively.

Fig. 5-7 shows the effect of the variation of the ratio of dead to live load on the

calculated resistance factors for set-up capacity at the reliability level β=2.33 and

β=3.00, respectively. The load factors for dead and live loads and the resistance factor

for Q0 are fixed as mentioned earlier when calculating the resistance factor for Qset-up. It

can be seen that the calculated resistance factor for Qset-up decreases slightly with an

increase in the ratio of dead to live load. In general, however, the calculated resistance

factor for Qset-up is not very sensitive to the ratio of dead to live load. The calculated

resistance factor for Qset-up can be conservatively taken as 0.4 for the span length less

than 60 m at the target reliability index β=3.00, when the predefined resistance factor

100
for Q0, dead load factor, and live load factor are 0.50, 1.25, and 1.75, respectively. The

resistance factor for Qset-up can be conservatively taken as 0.5 for the span length less

than 60 m at target reliability index β=2.33, when the predefined resistance factor for

Q0, dead load factor, and live load factor are 0.65, 1.25, and 1.75, respectively.

The recommended load and resistance factors are tabulated in Table 5-4 for

different reliability levels. The corresponding F.S. (Factor of Safety) in ASD

(Allowable Stress Design) can be calculated as follows:

γ D LD / LL + γ L
LD / LL + 1
F .S . = (5-6)
φ 0 Q0 / Qset −up + φ set −up
Q0 / Qset −up + 1

Fig. 5-8 shows the relationship between the corresponding F.S. in ASD and the

ratio of dead to live load. The corresponding F.S. decreases with an increase in the ratio

of dead to live load. The corresponding F.S. are about 2.40 and 3.00 for the

recommended load and resistance factors at target reliability index β=2.33 and β=3.00,

respectively.

5.7 Proposed Procedure to Incorporate Set-up into LRFD of Driven Piles in Sand

The incorporation of set-up effect can efficiently improve the estimation of total

capacity of driven piles so that pile length or numbers of piles can be economically

reduced. The proposed design procedure for incorporating set-up into LRFD of driven

piles in sand is illustrated as follows:

101
 1. Assign a pile size, length, and type for preliminary design using available site

information and a static analytical design method.

2. Perform CAPWAP dynamic testing 1 day after pile installation and record the

measured capacity as Q0.

3. Estimate the set-up capacity based on the proposed semi-logarithmic empirical

t
relationship: Qset −up = Q0 A log (where A = 0.4, t = 30 days, and t0 = 1 day).
t0

4. Use the recommended resistance and load factors in Table 5-4 and the following

formula, φ 0 Q0 + φ set −up Q set −up ≥ γ L L L + γ D L D , to evaluate the reasonableness of

the preliminary design.

5. Based on the result from step 4, change the pile length or number as desired to

optimize the design.

The engineering judgment is very important in a given project site when using the

proposed procedure to consider the set-up capacity, because the database in this chapter

is collected based on a wide range of site variability. The site-specific experience on

set-up capacity is more desirable when the pile test results are available to estimate the

development of long-term set-up at the given site. It is recommended that additional

pile testing data be obtained to supplement the current database, thus allowing for more

comprehensive statistical analysis of the pile set-up.

102
5.8 Conclusions

A database of pile set-up capacity in sand has been compiled and presented in this

chapter. The compiled database showed that the logarithmic empirical relationship

proposed by Skov and Denver (1988) could be used to predict pile set-up. The

reference time t0 = 1 day and mean value A = 0.4 were suggested for Skov and

Denver’s equation. The pile set-up database indicated that the time duration of 30 days

after EOD may be considered as the point after which the set-up effect would be

minimal. Based on the Kolmogorov-Smirnov test, the lognormal distribution was

shown to adequately represent the probabilistic characteristics of the predicted pile set-

up capacity. The mean value μlnx and standard deviation σlnx of the fitted lognormal

distribution are -0.122 and 0.538, respectively.

The statistical parameters of the pile set-up capacity derived from the compiled

database, together with previous statistical analysis of CAPWAP approach (Paikowsky

et al. 2004), were systematically incorporated within the framework of FORM to

develop separate resistance factors using the load conditions specified in AASHTO

(Novak 1999). The incorporation of set-up effect into the prediction of total pile

capacity gives advantageous contribution to the estimated total pile capacity. To be

consistent with the existing experience about the load and the resistance effects, the

load factors and resistance factors for Q0 are pre-set in the present chapter when

performing FORM analysis to determine the resistance factors for the pile set-up

capacity. For piles driven into sand, the resistance factor for Qset-up can be taken as 0.4

for a bridge span length less than 60 m if a target reliability index β=3.00 is chosen and

103
the pre-set resistance factor for Q0, dead load factor, and live load factor are 0.50, 1.25,

and 1.75, respectively. The resistance factor for Qset-up can be taken as 0.5 for a bridge

span length less than 60 m if a target reliability index β=2.33 is chosen and the pre-set

resistance factor for Q0, dead load factor, and live load factor are 0.65, 1.25, and 1.75,

respectively. The corresponding F.S. in ASD are about 2.40 and 3.00 for the

recommended load and resistance factors in LRFD at the target reliability index β=2.33

and β=3.00, respectively.

104
Table 5-1 Summary of Load Test Database for Driven Piles in Sand

References Site Number Main soil type Pile type Depth Testing
locations of piles (m) type
Tevenas and Audy Canada 27 Uniform medium D=305mm Concrete 8.5~13 Static
(1972) sand, SPT N≈23 hex.
Samson and Authier Canada 1 Medium sand, HP 12*63 22 Stat. &
(1986) gravel Dyn.
Seidel et al. (1988) Australia 1 Loose to dense 450*450mm 10.5 Stat. &
sand prestressed concrete Dyn.
Skov and Denver Germany 2 Medium to coarse 350*350mm 21, 33.7 Stat. &
(1988) sand concrete, and Dyn.
D=762mm pipe
Zai (1988) China 5 Fine sand D=610mm pipe 40~45 Stat. &
Dyn.
Preim et al. (1989) USA 2 Loose to medium 355*355mm 27, 25 Stat. &
fine sand concrete, and Dyn.
D=323mm closed-
end pipe
Svinkin et al. (1994) USA 6 Silty sand 457~915mm square 19.5~22.9 Stat. &
concrete Dyn.
York et al. (1994) USA 15 Medium dense D=355mm 10.7~21.6 Stat. &
sand monotube Dyn.
Chow et al. (1998) France 2 medium to very D=324mm open-end 11, 22 Stat. &
dense sand, pipe Dyn.
Dr ≈ 75%
Axelsson (1998) Sweden 3 Loose to medium 235*235mm 19 Dynamic
dense glacial sand concrete
Attwooll et al. USA 1 Unsaturated, D=324mm pipe 10.1 Stat. &
(1999) dense sand Dyn.
Tan et al. (2004) USA 5 Loose to medium 356mm H-pile, 34~37 Dynamic
dense sand D=610mm closed-
end pipe
Bullock et al. (2005) USA 3 Dense fine sand 457*457mm 9~25 O-Cell
prestressed concrete

105
Table 5-2 Empirical Formulas for Predicting Pile Capacity with Time

References Equation Comments

Skov and t where
Qt = Q0 (1 + A log ) (5-7)
Denver t0 t0=0.5 and A=0.2 in sand
(1988) t0=1.0 and A=0.6 in clay
Svinkin et Qt = BQEOD t 0.1 (5-8) B = 1.4 upper bound
al. (1994) B = 1.025 lower bound
Long et al. Qt = 1.1QEOD t α (5-9) values of α: average = 0.13,
(1999) lower bound = 0.05, upper
bound = 0.18,
Bogard t / T50 Qu = ultimate capacity with
Qt = Qu [0.2 + 0.8( )] (5-10)
and 1 + t / T50 100% of setup realized
Matlock T50 = time required to realize
(1990) 50% of pile set-up
Zhu Q14 = (0.375S t + 1)QEOD (5-11) Q14 = pile capacity at 14 day
(1998) after EOD
St = sensitivity of soil
Huang Qt = Q EOD + 0.236[1 + (Qmax − QEOD ) log t ] Qmax = maximum pile capacity
(1998) (5-12) with 100% of setup realized

106
Table 5-3 Probabilistic Characteristics of Random Variables of Loads and Resistances

for Piles Driven into Sand

Random Bias factor Standard Coefficient of Distribution References

variables λ deviation σ variation, COV type

Q0 1.158 0.393 0.339 Lognormal Paikowsky et al.(2004)

Qset-up 1.023 0.593 0.580 Lognormal Current Study

LL 1.150 0.230 0.200 Lognormal AASHTO (Nowak 1999)

LD 1.050 0.105 0.100 Lognormal AASHTO (Nowak 1999)

107
Table 5-4 Recommended Partial Safety Factors (i.e. load and resistance factors) for

Driven Piles in Sand

Random Bias Coefficient of Partial safety Partial safety References

variables factor λ variation COV factor (β=2.33) factor (β=3.00)

Q0 1.158 0.339 0.65 0.50 Paikowsky et al.(2004)

Qset-up 1.023 0.580 0.50 0.40 Current Study

LL 1.150 0.200 1.75 1.75 AASHTO (Nowak 1999)

LD 1.050 0.100 1.25 1.25 AASHTO (Nowak 1999)

108
 4
Tavenas and Audy (1972)
Samson and Authier (1986)
Skov and Denver (1988)
3.5 Seidel et al. (1988)
Normalized Pile Capacity Qt/Qre
Zai (1988)
Priem et al. (1989)
York et al. (1994)
3 Svinkin et al. (1994)
Chow et al. (1998)
Attwooll et al. (1999)
Axelsson (1998)
Tan et al. (2004)
2.5 Bullock et al. (2005)

1.5

0.5
0.01 0.1 1 10 100 1000
Time after EOD (days)

Figure 5-1 Case Histories of Pile Set-up in the Compiled Database

109
 4.0
Tavenas and Audy (1972)
Samson and Authier (1986)
3.5
Skov and Denver (1988)
A=0.9
Normalized Pile Capacity Qt/Q0
Seidel et al. (1988)
Svinkin et al. (1994)
3.0
Axelsson (1998)
Tan et al. (2004)

2.5 Bullock et al. (2005)

A=0.4
2.0

1.5

A=0.1
1.0

0.5

0.0
1 10 100 1000
Time after EOD (days)

Figure 5-2 Variation of Normalized Capacity with Time Based on Skov and Denver’s
Equation

110
 4.000

3.500

3.000

2.500 B=1.4
Qt/QEOD

B=1.25
2.000
B=1.025
1.500

1.000
Samson and Authier (1986) Skov and Denver (1988)
Seidel et al. (1988) Zai (1988)
Priem et al. (1989) York et al. (1994)
0.500 Svinkin et al. (1994) Chow et al. (1998)
Attwooll et al. (1999) Axelsson et al. (1998)
Tan et al. (2004)

0.000
1 10 100 1000
Time after EOD (days)

Figure 5-3 Variation of Normalized Capacity with Time Based on Svinkin et al.’s
Equation

111
 0.25

0.20
Relative Frequency of Number of Pile Cases

Lognormal distribution
0.15 μlnx=-0.122 σlnx=0.538

0.10

0.05

0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Ratio of Measurd Set-up Capacity over Predicted
Set-up Capacity from Eq. (1) (A=0.4)
Figure 5-4 Frequency Distribution of the Ratio of Measured Set-up Capacities to the
Predicted, Based on Skov and Denver’s Equation

112
 100

Theoretical Lognormal
80 distribution
Cumulative Distribution

60
Empirical ratio of the measured to
the predicted set-up capacity
40

20

0
0.0 0.5 1.0 1.5 2.0 2.5
Ratio of the measured to the predicted set-up capacity
Figure 5-5 Kolmogorov-Smirnov Test for Assumed Lognormal Distribution of the
Ratio of the Measured to the Predicted Set-up Capacity

113
 1.6

Partial Load and Resistance Factors

1.4
1.2
1
0.8
0.6
Resistance factor for Qo
0.4
Resistance factor for Qset-up

0.2 Load factor for live load

Load factor for dead load
0
0.5 1 1.5 2 2.5 3 3.5
Target Reliability Index β

Figure 5-6 Relationship between Partial Safety Factor and Target Reliability Index

114
 1.0
Resistance factor for Qset-up 0.9
0.8
0.7
0.6
β=2.33
0.5
β=3.00
0.4
0.3
0.2
0.1
0.0
0 1 2 3 4
Ratio of Dead Load to Live load

Figure 5-7 Variation of Resistance Factor for Qset-up in Sand versus the Ratio of Dead
Load to Live Load

115
 4.0

Corresponding Factor of Safety 3.5

3.0
β=3.00
2.5
β=2.33
2.0

1.5

1.0

0.5

0.0
0 1 2 3 4
Ratio of Dead to Live Load

Figure 5-8 Variation of the F.S versus the Ratio of Dead Load to Live Load for Piles
Driven into Sand

116
 CHAPTER VI

BAYESIAN BASED FORMALISM FOR UPDATING LRFD RESISTANCE

FACTORS FOR DRIVEN PILES

6.1 Introduction

Beginning in 2007, AASHTO (American Association of State Highway and

Transportation Officials) requires that the Load and Resistance Factor Design (LRFD)

method be used for foundation design. Whereas Allowable Stress Design (ASD) in

prior AASHTO Guide lumps all uncertainties in the applied load and ultimate

geotechnical or structural capacity into a single value of factor of safety (FS), LRFD

accounts for the variability of loads and resistances by respective load and resistance

factors. The resistance factors for various design methods are recommended in 2006

AASHTO LRFD specifications (AASHTO 2006). The preliminary design of foundation

piles is usually performed with static analytical calculations using the soil parameters

from geotechnical site investigations and the accompanied laboratory test results. In

current LRFD method (AASHTO 2006), the uncertainties with respect to the prediction

method, the errors of calculation model, and the spatial variability of soil parameters,

are incorporated in a single resistance factor for a particular design method, regardless

of individual site-specific situation. For each construction site, soil profiles, soil types,

pile driving equipment, and hammer performance, may be different. Therefore, it may

117
be desirable to develop a strategy for site-specific calibration of the resistance factor to

improve foundation design.

Currently, dynamic pile tests are carried out to verify the pile capacity during pile

driving and to monitor the installation process for avoiding pile damage due to hammer

impact. The pile length estimated using a static analysis method requires verification by

dynamic pile tests during pile installation. Recognizing that either static pile load test or

dynamic pile test may be performed for most pile driving projects, Vrouwenvelder

(1992) presented an approach for incorporating either the static or dynamic test results

to update FS in the ASD design of driven piles. Recently, Zhang et al. (2002)

demonstrated that the results from static pile load tests could be incorporated into pile

design using Bayesian theory by updating the resistance factor in LRFD. Dynamic pile

testing, however, has become a commonly used method. Accordingly, a methodology

needs to be developed for improving static pile capacity calculation by updating the

corresponding resistance factor using Bayesian theory and the dynamic pile test results.

This chapter presents a methodology for static method based pile design by combining

the static calculation with dynamic pile testing data. Bayesian theory is employed to

combine the two methods in the framework of LRFD approach. The First Order

Reliability Method (FORM) is used to derive the updated resistance factor. Specifically,

the results from dynamic pile tests are incorporated to reduce the uncertainties

associated with static analysis methods via updating of the corresponding resistance

factor. It is shown that by a combination of two methods, one can improve accuracy and

118
confidence level on predicting pile capacity. Two case studies are presented to illustrate

the proposed methodology and its applicability.

6.2 Characteristics of LRFD Method

The LRFD criterion for substructure is stated as follows (AASHTO, 2006):

∑η γ Q
i i i ≤ φRn (6-1)

where Rn= nominal resistance, φ= resistance factor, Qi=load effect, γi=load factor, and

ηi= load modifier to account for the effects of ductility, redundancy and operational

importance.

In AASHTO (2003), the recommended resistance factors of various static analysis

methods for axially loaded driven piles include a parameter, λv, to account for the effect

of quality control procedure during pile driving. For example, if α method is used to

predict the pile’s friction resistance in clay, a resistance factor of 0.7 is recommended. If,

an ENR (Engineering News-Record) driving formula is used to verify the pile capacity,

a λv factor of 0.8 is recommended. Therefore, the resistance factor to be used in the

above scenario is 0.56 (i.e. 0.7×0.8). The concept of incorporating the parameter, λv, to

static analysis is attractive; nevertheless, a single value of λv for each quality control

method appears to be less desirable.

6.3 First Order Reliability Method (FORM)

Barker et al. (1991) and Withiam et al. (1998) described a first-order second-moment

(FOSM) method for developing the resistance factors based on load tests in a large
119
number of sites. However, LRFD for structural design has evolved beyond FOSM to

the more invariant FORM approach. Paikowsky, et al. (2004) suggested that the use of

FOSM method might result in the resistance factors that could be approximately 10%

lower than those obtained by FORM. In order to be consistent with the current

structural code, Phoon et al. (2003a and 2003b), Ayyub and McCuen (2003), Paikowsky

et al. (2004), and Allen et al. (2005) have proposed the application of FORM approach

to calibrate resistance factors for deep foundations. Ang and Tang (1984) provide

detailed descriptions about the procedure of using FORM to calibrate LRFD partial

safety factors. FORM can be used to assess the reliability of a pile with respect to

specified limit states as well as to calculate partial safety factors φi and γi for resistances

and loads, respectively, for a chosen target reliability index β.

6.4 Uncertainty of Static Analysis Methods

The sources of error in predicting axial driven pile capacity via static analysis methods

can be enumerated as follows (Withiam et al. 1998):

1. Model error that relates to an overall bias of the prediction method;

2. Systematic error that can be divided into systematic error in the trend and bias in

the procedures;

3. Inherent spatial variability – only a limited number of soil borings and

samplings at a site to characterize the subsurface condition for a large volume of

soil and rock;

4. Random testing error and statistical uncertainty due to insufficient number of

tests;
120
 5. Error in measuring actual pile capacities in static load tests that are used as

baseline to judge accuracy of the prediction method.

It is important to bear in mind that different error sources may exert different degrees of

effect on the reliability analysis (Christian 2004). For example, the contributions of soil

shear strength are generally summed along the shaft length of driven piles in which the

value of soil shear strength is estimated from soil boring samples. The contribution of

the scatter in the shear strength to the uncertainty in the predicted axial driven pile

capacity may be reduced greatly by increasing the numbers of soil borings. On the other

hand, the model error and systematic error propagate throughout the analysis.

The lognormal function was generally used for describing the variable λR, defined as

the ratio of the measured pile capacity from the static load test, Rs, to the capacity

predicted with a static analysis method, Rp (Withiam et al. 1998).

RS
λR = (6-2)
RP

6.5 Uncertainty of Dynamic Pile Testing Methods

Compared to the static analysis methods, the dynamic pile testing methods also have

model and systematic errors. For example, the errors associated with the energy

methods (Paikowsky et al. 1994, Liang and Zhou 1997) are the errors of energy input in

the prediction equation. The wave equation based methods, such as CAse Pile Wave

Analysis Program (CAPWAP) and Case Method (Rausche et al. 1972, 1985), improved

121
on energy methods by partially eliminating the errors of energy input; nevertheless, they

simultaneously introduced other uncertainties, such as the errors of signal recording of

the transducers as well as the assumed damping models.

The correction factor, K, of the dynamic pile testing method is introduced as the ratio of

the pile capacity determined from the static load test, RS, divided by the pile capacity

determined from the dynamic pile testing method, RD:

RS = KR D (6-3)

McVay et al. (2000) and Paikowsky et al. (2004) showed that the lognormal distribution

could be used to describe the distribution function of K.

6.6 Improving Static Method with Dynamic Test Results

The uncertainty of a static analysis method can be reduced if the available dynamic pile

test results are incorporated in an updating process. Supposing that a number of

dynamic pile tests are conducted at a given site, the dynamic pile test results consist of a

set of n dynamic pile capacities RD1, RD2, …, RDn. For simplicity, assuming that the

static load test results involve negligible uncertainty, then the following expressions can

be obtained,

RS R
λR = = K D = K (RDS) (6-4)
RP RP

RD
λR = μk (6-5)
Rp

122
 RD
σ λR = σ k (6-6)
Rp

where RP = pile capacity predicted by a static method, RDS = ratio of the dynamic

prediction capacity over the static prediction capacity, λ R = mean value of RDS; μk =

mean value of K; σ λ R = standard deviation of RDS; σk = standard deviation of K.

Bayesian sampling theory can be expressed as:

P (ε / θ ) f ' (θ )
f " (θ ) = +∞
(6-7)
∫ P(ε / θ ) f ' (θ )dθ
−∞

where f ’(θ) and f ”(θ) are the prior and the posterior density functions of the parameter

θ, respectively. The term P(ε/θ) is the conditional probability or likelihood of observing

the experimental outcome ε, assuming that the value of the parameter is θ. In the

present analysis, the dynamic pile test results are treated as the experimental outcome ε

assuming that the pile capacity computed by a static analysis method is θ. Therefore, the

posterior distribution parameters (updated) of the variable λR is expressed as

ψ ' (ζ K 2 / n) + ζ ' 2 ln x
ψ '' = (6-8)
ζ K 2 / n + ζ '2

2
ζ ' (ζ K 2 / n)
ζ "= 2
(6-9)
ζ ' +ζ K2 /n

Then, based on FORM analysis, the updated resistance factors can be recalculated using

the obtained posterior statistical parameters of the variable λR.

123
6.7 Combination of Static Analysis and Dynamic Pile Testing Methods

For illustration purpose, the statistics of loads is taken from the LRFD Bridge Design

Specifications (AASHTO, 2006). They are as follows: bias factor for dead load λQD =

1.05, bias factor for live load λQL = 1.15, load factor for dead load γD =1.25, load factor

for live load γL = 1.75, COV value of dead load COVQD = 0.1, and COV value of live

load COVQL = 0.2, ratio of dead load to live load QD/QL = 1.58 corresponding to the

bridge span length of 27 m. A target reliability index of β = 2.33, corresponding to the

probability of failure Pf = 1%, is used as an illustration to determine the resistance

factors for the driven piles.

A study of the statistics of various design methods for driven piles was summarized by

Paikowsky, et al. (2004) and reproduced in Table 6-1. As can be seen, α –API method

exhibits the best accuracy among the static analysis methods. Among different dynamic

methods, CAPWAP BOR (Beginning of Restrike) provides more accurate prediction

than CAPWAP EOD (End of Driving) and EA (Energy Approach, Paikowsky et al.

1994).

To investigate the potential benefits of combining two methods, Fig. 6-1 is presented to

show the effect of combining CAPWAP BOR with several different static analysis

methods. After updating, the most accurate static analysis method in prior prediction, α–

API, still provides the smallest COV for posterior distribution. The dynamic pile testing

methods provides less benefit in updating the high-accuracy static analysis methods,

particularly when the prior static analysis method possesses higher accuracy than the

124
dynamic pile testing method chosen for the updating purpose. Fig. 6-2 shows the effect

of using different dynamic pile testing methods on the updated λ-method distributions.

The incorporation of CAPWAP BOR, which exhibits higher accuracy than the other

dynamic methods, yields the smallest COV of the updated distribution of λ-method.

This demonstrates that, when the static analysis and dynamic pile testing methods are

combined to obtain the posterior distribution, combination of two accurate design

methods generate more reliable updating of posterior distributions.

The effect of RDS on updating is illustrated in Fig. 6-3, where CAPWAP BOR is used

to update the λ-method. When the value of RDS increases, the mean of the posterior

distribution increases and the COV remains the same. It shows that the latter is not

related to the value of RDS, but a function of the standard deviations of prior static

analysis method and dynamic pile testing method as well as the number of dynamic

tests. Fig. 6-4 shows the effect of the number of dynamic tests on the updated

distributions. One can see that COV decreases with an increase in the number of

dynamic pile tests. Thus, the posterior prediction of pile capacity becomes more reliable,

as the number of dynamic pile tests incorporated in updating increases.

6.8 Updated Resistance Factors versus RDS and Number of Dynamic Pile Tests

After obtaining the statistical parameters of the posterior distribution, the resistance

factors can be recalculated using FORM. Fig. 6-5 shows the variation of the updated

resistance factors with RDS. The updated resistance factors are approximately

proportional to RDS for all the combinations examined. The slope in Fig. 6-5 represents

125
the effect of RDS on the updated resistance factors for various combinations. For the

combination of β-method and CAPWAP BOR, RDS exerts the largest influences on the

updated resistance factors for the posterior static analysis method. The least effect of

RDS on the updated resistance factor is observed in the combination of α-API and

CAPWAP EOD. This demonstrates that the use of more reliable dynamic pile testing

method (e.g., CAPWAP BOR) can provide better beneficial updating on the resistance

factors for the static analysis methods. On the other hand, the resistance factors for more

accurate static analysis method (e.g., α-API method) is generally not affected by using

dynamic pile test for updating.

The resistance factors can be significantly increased not only by using dynamic method

that yields favorable RDS, but also by incorporating a larger number of dynamic pile

testing results. Fig. 6-6 shows the relationship between the updated resistance factors

and the number of dynamic pile tests used in different combinations of dynamic pile

testing methods and static analysis methods. For all the combinations considered, the

resistance factor increases as the number of the dynamic pile tests incorporated in

updating increases. The static analysis methods with higher accuracy, such as α-API

and λ-method, are less affected by the number of dynamic pile tests on the updated

resistance factors. The first two dynamic pile tests give more contribution to the

increase of the resistance factors for the α-API and λ-method. The use of dynamic pile

tests of more than two can only result in a small improvement of the reliability of the

static methods. It suggests that two dynamic tests could be considered sufficient to

update the high accuracy static analysis methods. However, for the static methods with

126
low accuracy (e.g., α-Tomlinson and β-method), the effect of the number of dynamic

pile tests used to update resistance factor is very pronounced. The updated resistance

factors are remarkably increased by the first four dynamic tests. Therefore, the number

of dynamic tests to be conducted to update the static calculation depends largely on the

static methods used.

6.9 Case Study I

Test piles #2 at Newbury site is selected from the Appendix D of NCHRP Report 507

(Paikowsky et al. 2004) as a case study. Test pile #2 is a closed-end steel pipe pile, with

O.D. = 324 mm (12.76 in), thickness = 12.7 mm (0.5 in), and penetration depth = 24.4m

(80 ft). The soil profile at the Newbury site is shown in Fig. 6-7. The ground soil

primarily consists of the normally consolidated clay with interbedded silty sand. The

water level is at the depth of 1.5 m (4.9 ft).

The α-API/Nordlund/Thurman method and the β/Thurman method are used to predict

the pile capacities as shown in Table 6-2. Applying the resistance factor (NCHRP

Report 507), the factored capacities are 376 kN (84.5 kip) and 310 kN (69.7 kip) for the

α-API/Nordlund/Thurman method and β/Thurman method, respectively.

The predicted ultimate capacity from the dynamic pile testing methods are 418 kN (94.0

kip), 1112 kN (250.0 kip), and 792 kN (178.0 kip) using CAPWAP EOD, CAPWAP

BOR, and EA EOD, respectively. The factored capacities are 268 kN (60.2 kip), 723 kN

127
(162.5 kip), and 420 kN (94.4 kip) based on CAPWAP EOD, CAPWAP BOR, and EA

EOD, respectively.

Table 6-3 provides a summary of the capacities from the prior static analysis, the

dynamic testing method, and the posterior static analysis methods, in comparison to the

actual static load test results. It can be seen that the updated resistance factors using

CAPWAP BOR for updating are highest for both static analysis methods.

Fig. 6-8 shows the effect of combining two methods. It can be seen that, for β /Thurman

method, the ratio of static load test capacity to factored prediction capacity varies from

prior 2.16 to posterior 1.64, 1.06, and 1.40 after updating using CAPWAP EOD,

CAPWAP BOR, and EA EOD, respectively. The pile capacity prediction using α-

API/Nordlund/Thurman method is also improved when updating the resistance factors

by incorporating the results of dynamic pile tests.

6.10 Case Study II

Test piles #3 at Newbury site from the Appendix D of NCHRP Report 507 (Paikowsky

et al. 2004) is a square (356 mm by 356 mm) precast, prestressed concrete (PPC) pile

with penetration depth of 24.4 m in the same soil condition as that in case study I.

The details of calculation using α-API/Nordlund/Thurman and β/Thurman methods are

summarized in Table 6-4. The pile capacities from the dynamic pile testing methods

based on CAPWAP EOD, CAPWAP BOR, and EA EOD are given in Table 6-5 as well.

128
The prior resistance factors in the table are taken from NCHRP Report 507 (Paikowsky

et al. 2004).

Table 6-5 shows the comparison between the predicted capacities from the prior static

analysis, the dynamic testing, and the posterior static analysis methods, and the actual

static load test results. In this case, the updated resistance factors by the three dynamic

testing methods are very close to each other. It could be due to an accurate prediction of

pile capacities by the original static analysis methods. Thus, if the static analysis

methods provide relatively accurate prediction of pile capacity, the improvement on the

updated resistance factors in predicting pile capacity using dynamic testing results may

be negligible.

Fig. 6-9 shows the comparisons of the factored pile capacities based on different

combinations of two methods. It can be seen that, combining β/Thurman method with

CAPWAP EOD, CAPWAP BOR, and EA EOD, the ratio of static load test capacity to

posterior factored prediction capacity ranges from 1.08, 1.00, and 1.03, respectively.

The predictions of pile capacities by combining α-API/Nordlund/Thurman method with

different dynamic pile testing method are also shown in Fig. 6-9. Although the

combining of two methods did not significantly alter the predicted pile capacity, but did

result in more consistent comparisons with the static load test results, the important

point is that combining two methods adds confidence to the predicted pile capacities.

129
The two case studies demonstrate the benefit of combining two methods to improve pile

capacity prediction as opposed to using only one method. The improvement is mainly

due to reduction of uncertainties associated with one method. By combining more

accurate static and dynamic methods, one can predict the updated, factored pile capacity

to be more consistent with the static load test result.

6.11 Conclusions

One of the benefits of LRFD approach is its suitability for applying reliability analysis

to quantify uncertainties associated with various methods for estimating loads and

resistances, respectively. With the reliability theory, Bayesian approach is employed to

combine different analysis methods to improve accuracy and confidence level of

predicted factored pile capacity as opposed to the case where a single analysis method

is employed. The main benefit of combining two different methods is the reduction of

uncertainties associated with each method itself.

The example analyses presented in this chapter for steel pipe piles and PPC piles in

mixed soil profiles have demonstrated the applicability and reasonableness of the

proposed approach. The updated, factored pile capacities are closer to the static load test

results than the originally predicted factored capacities by the two methods used in a

stand-alone analysis.

Specific conclusions drawn from this study can be enumerated as follows:

130
• When the static analysis and dynamic pile testing methods are combined to obtain

the posterior distribution, combination of two accurate design methods generate

more reliable updating of posterior distributions.

• The dynamic pile testing methods provides less benefit in updating the high-

accuracy static analysis methods, particularly when the prior static analysis method

possesses higher accuracy than the dynamic pile testing method chosen for the

updating purpose.

• The posterior prediction of pile capacity becomes more reliable, as the number of

dynamic pile tests incorporated in updating increases.

• Two dynamic tests could be considered sufficient to update the high accuracy static

analysis methods. For the static methods with low accuracy, the updated resistance

factors are remarkably increased by the first four dynamic tests.

131
Table 6-1 Statistical Parameters and Resistance Factors for Driven Piles from NCHRP
Report 507 (β = 2.33) (Paikowsky et al. 2004)
Number Bias factor Coefficient Resistance
Prediction of cases λR or of factor, Φ
Method category
method Correction variation
factor K
Static analysis λ- Method 18 0.76 0.29 0.48
methods (Concrete α –API 17 0.81 0.26 0.54
piles in Clay) β – Method 8 0.81 0.51 0.32
α –Tomlinson 18 0.87 0.48 0.36
Dynamic methods CAPWAP 125 1.63 0.49 0.64
with measurements EOD
CAPWAP 162 1.16 0.34 0.65
BOR
EA EOD 128 1.08 0.40 0.53

132
Table 6-2 Summary of Calculated Capacities from Static Analysis Method for Test Pile
#2
Layer α-API/Nordlund β- Method
Layer # Thickness Unit Skin Friction Unit Skin Friction
(m) Friction (kPa) Capacity (kN) Friction (kPa) Capacity (kN)
1 2.5 41.8 104 71.2 180

2 6.1 26.6 162 26.6 162

3 4.9 26.6 130 39.4 193

4 7.9 43.7 345 68.4 540

Tip Resistance (kN) –Thurman Method 309 309

Total resistance (kN) 1050 1384

Layer α-API/Nordlund β- Method

Layer # Thickness Unit Skin Friction Unit Skin Friction
(ft) Friction (psf) Capacity (kip) Friction (psf) Capacity (kip)
1 8.2 873 23.4 1487 40.5

2 20.0 556 36.4 556 36.4

3 16.1 556 29.2 823 43.4

4 25.9 913 77.6 1429 121.4

Tip Resistance (kip) –Thurman Method 69.5 69.5

Total resistance (kip) 236 311.1

133
 Table 6-3 Pile Capacities of Prior Static Analysis, Dynamic Testing, Posterior Static Analysis, and Static Load Test
(Test Pile #2 in Newbury)
Static
Posterior static analysis
Prior static analysis method Dynamic pile testing method load test
method
(kN)
Pile Prior Factored Dynamic Pile Factored Pile Posterior Factored
Static design Resistance
capacity resistance capacity Design capacity capacity capacity resistance capacity
method factor, Φ
(kN) factor, Φ (kN) Method (kN) (kN) (kN) factor, Φ (kN)
CAPWAP
418 0.64 268 0.40 420
α-API/ EOD
Nordlund/ 1050 0.36 378 CAPWAP 1050
1112 0.65 723 0.62 651
Thurman BOR
658
EA EOD 792 0.53 420 0.46 483
CAPWAP
418 0.64 268 0.29 401
EOD
1384 0.22 304 CAPWAP 1384
1112 0.65 723 0.45 623
β / Thurman BOR
EA EOD 792 0.53 420 0.34 471

134
Table 6-4 Summary of Calculated Capacities from Static Analysis Method for Test Pile
#3
Layer α-API/Nordlund β- Method
Layer # Thickness Unit Skin Friction Unit Skin Friction
(m) Friction (kPa) Capacity (kN) Friction (kPa) Capacity (kN)
1 2.5 41.8 146 71.2 249

2 6.1 26.6 227 26.6 227

3 4.9 26.6 182 39.4 270

4 7.9 93.1 1030 68.4 756

Tip Resistance (kN) –Thurman Method 473 473

Total resistance (kN) 2058 1975

135
Table 6-5 Pile Capacities of Prior Static Analysis, Dynamic Testing, Posterior Static Analysis, and Static Load Test
(Test Pile #3 in Newbury)
Static
Posterior static analysis
Prior static analysis method Dynamic pile testing method load test
method
(kN)
Pile Prior Factored Dynamic Pile Factored Pile Posterior Factored
Static design Resistance
capacity resistance capacity Design capacity capacity capacity resistance capacity
method factor, Φ
(kN) factor, Φ (kN) Method (kN) (kN) (kN) factor, Φ (kN)
CAPWAP
738 0.64 472 0.37 761
α-API/ EOD
Nordlund/ 2058 0.36 741 CAPWAP 2058
1228 0.65 798 0.42 864
Thurman BOR
872
EA EOD 1228 0.53 651 0.40 823
CAPWAP
738 0.64 472 0.41 810
EOD
1975 0.42 830 CAPWAP 1975
1228 0.65 798 0.44 869
β / Thurman BOR
EA EOD 1228 0.53 651 0.43 849

136
 3
Prior λ-method, mean=0.76, COV=0.29
Prior α-API, mean=0.81, COV=0.26
Prior β-method, mean=0.81, COV=0.51
2.5 Prior α-Tomlinson, mean=0.87,COV=0.48
CAPWAP BOR, mean=1.16, COV=0.34
Posterior λ-method, mean=0.76, COV=0.22
Posterior α-API, mean=0.81, COV=0.20
Posterior β-method, mean=0.81, COV=0.28
2
Probability density

Posterior α-Tomlinson, mean=0.87,COV=0.27

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4

Bias factor λR
Figure 6-1 Prior and Posterior Distribution of Different Static Design Methods When
CAPWAP BOR Is Used for Updating

137
 3

CAPWAP EOD, mean=1.63, COV=0.49

CAPWAP BOR, mean=1.16, COV=0.34
Energy Approach EOD, mean=1.08, COV=0.4
2.5
Prior λ-method, mean=0.76, COV=0.29
Posterior λ-method w ith CAPWAP EOD, mean=0.76, COV=0.25
Posterior λ-method w ith CAPWAP BOR, mean=0.76, COV=0.22
Posterior λ-method w ith EA EOD, mean=0.76, COV=0.23
2
Probability density

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Bias factor λR
Figure 6-2 Prior and Posterior Distribution of Static λ-Method when Different Dynamic
Pile Tests Are Used for Updating

138
 3.5

Prior λ-method, mean=0.76, COV=0.29

3
Posterior λ-method, RDS=0.5, mean=0.64, COV=0.22

Posterior λ-method, RDS=λR=0.76, mean=0.76, COV=0.22

2.5 Posterior λ-method, RDS=1.0, mean=0.85, COV=0.22

Probability density

Combined w ith CAPWAP BOR

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Bias factor λR
Figure 6-3 Prior and Updated Distribution of Static λ-Method with Different RDS

139
 5

Prior λ-method, mean=0.76, COV=0.29

Posterior λ-method, One CAPWAP BOR, mean=0.76, COV=0.22

4 Posterior λ-method, Tw o CAPWAP BOR, mean=0.76, COV=0.18
Posterior λ-method, Five CAPWAP BOR, mean=0.76, COV=0.13
Posterior λ-method, Ten CAPWAP BOR, mean=0.76, COV=0.10
Probability density

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Bias factor λR
Figure 6-4 Prior and Updated Distribution of Static λ-Method with Different Numbers of
Dynamic Tests

140
 1.2
Combination of λ-method and CAPWAP EOD
Combination of α-API and CAPWAP EOD
1.1 Combination of β-method and CAPWAP EOD
Combination of α-Tomlinson and CAPWAP EOD
Combination of λ-method and CAPWAP BOR
1 Combination of α-API and CAPWAP BOR
Combination of β-method and CAPWAP BOR
Updated resistance factor Φ

Combination of α-Tomlinson and CAPWAP BOR

0.9

0.8

0.7

0.6

0.5

0.4

0.3
0 0.5 1 1.5 2 2.5 3

RDS
Figure 6-5 Variation of Updated Resistance Factors with RDS

141
 0.8

0.7
Updated resistance factor Φ

0.6

0.5

Combination of λ-method and CAPWAP EOD

0.4 Combination of α-API and CAPWAP EOD
Combination of β-method and CAPWAP EOD
Combination of α-Tomlinson and CAPWAP EOD
Combination of λ-method and CAPWAP BOR
Combination of α-API and CAPWAP BOR
0.3
Combination of β-method and CAPWAP BOR
Combination of α-Tomlinson and CAPWAP BOR

0.2
0 2 4 6 8 10 12
Number of dynamic tests
Figure 6-6 Variation of Updated Resistance Factors with Number of Dynamic Tests

142
Depth (m)
0 Pile
1.5 Water level Pile is in casing.
3
Layer 1: O.C. Clay Su = 95.76kPa γ=1922kg/m3
5.5

Layer 2: N.C. Clay Su = 31.6kPa γ=1762kg/m3

11.6

Layer 3: N.C. Clay Su = 26.81kPa γ=1842kg/m3

16.5
Layer 4: Interbedded Clay and silty sand γ=1972kg/m3
19.2
Layer 5: Silty sand φ=35º γ=2002kg/m3
21.6
Layer 6: Interbedded Clay and silty sand γ=1972kg/m3
24.1
24.4
Pile tip
Layer 7: Fine to medium sand φ=35º γ=2002kg/m3

Figure 6-7 Soil Profile in Newbury Site

143
 3.5
Measured static load test capacity/ factored prediction capacity Prior static calculation prediction
Dynamic testing Prediction
3 Posterior static calculation prediction

2.5

1.5

0.5

0
α CE Co α CB Co α EE Co β CE Co β CB Co β EE Co

Figure 6-8 Comparison of Ratio of Static Load Test Capacity to Factored Prediction
Capacity for Test Pile #2
α =α-Tomlinson/ Nordlund/ Thurman; β=β-Method/ Thurman; CE=CAPWAP EOD;
CB=CAPWAP BOR; EE=EA EOD; Co=Combination of two design methods.

144
 3

Measured static load test capacity/ factored prediction capacity Prior static calculation prediction
Dynamic testing Prediction
2.5 Posterior static calculation prediction

1.5

0.5

0
α CE Co α CB Co α EE Co β CE Co β CB Co β EE Co

Figure 6-9 Comparison of Ratio of Static Load Test Capacity to Factored Prediction
Capacity for Test Pile #3
α =α-Tomlinson/ Nordlund/ Thurman; β=β-Method/ Thurman; CE=CAPWAP EOD;
CB=CAPWAP BOR; EE=EA EOD; Co=Combination of two design methods.

145
 CHAPTER VII

WAVE EQUATION TECHNIQUE FOR ESTIMATING DRIVEN PILE CAPACITY

7.1 Introduction

Dynamic pile testing has been widely used in estimating the pile capacity and

deciding the driving criteria during pile driving. During dynamic pile testing, the strain

gages and accelerometers are usually attached near the pile head to record the waveforms.

The measured strains are converted into axial forces and the measured accelerations are

integrated over time to obtain velocity, both as a function of real time. Both the measured

force and the product of measured velocity and pile impendence are usually proportional

during the initial rising period at each hammer blow, then separated from each other once

the shaft and toe resistance are reflected upward to the pile head. Some researchers such

as Rausche et al. (1972, 1985), Hirsch et al. (1976), Liang and Zhou (1996), and Liang

(2003) have proposed various approaches to interpret the measured force and velocity

histories on determining the pile capacity.

With the development and acceptance of PDA (Pile Driving Analyzer), the Case

method (Rausche et al. 1985), utilizing the measured force and velocity histories in

approximate simple algebraic equations, has become one of the widely used methods to

evaluate the pile capacities. However, the static soil resistances determined by the Case

146
method are very sensitive to the assumed Case damping factor, and the shaft and the toe

resistance cannot be separated by the Case method. In an effort to overcome some of the

shortcomings of the Case method, an alternative procedure, known as the CAPWAP

(Goble Rausche Likins and Associates, Inc., 1996), was developed by Rausche et al.

(1972, 1985). The CAPWAP approach is based on the one-dimensional wave

propagation model suggested by Smith (1960). In CAPWAP analysis, one of the two

HST data (either force or velocity) is used as input to generate simulation results to match

closely with the other HST data by adjusting Smith model parameters. Once the

acceptable match is achieved, then the pile capacities and the Smith model parameters

can be determined. However, the requirement of experienced program operator and

substantial time-consuming computation hinders the further application of CAPWAP

procedure.

In this chapter, a one-dimensional wave equation based algorithm to interpret the

HST data for estimation of the shaft and toe resistance of driven piles is introduced. Both

the force and velocity time histories measured at the pile head are used as input boundary

conditions in the new algorithm. The Smith damping factor and the static soil resistance

can be directly determined based on the closed form solution of the one-dimensional

wave equation, which is computationally efficient compared to the existing numerical

procedure. Furthermore, the shaft resistance and the toe resistance could be successfully

separated by the proposed wave equation method as well. Numerical examples are

presented to illustrate the methodology and applicability of the proposed analysis

approach.

147
7.2 Soil-Pile Interaction Model

The soil-pile interaction model is shown in Fig. 7-1 where the following assumptions

are made: 1) the pile is represented by a linear elastic bar with Young’s modulus E, total

length L, cross-section area A, and mass density ρ; 2) the soil around the pile is a

homogenized isotropic medium; 3) the resistance along pile shaft is a function of the

displacement and velocity related to the depth and time.

The general equation of motion for the soil-pile interaction model can be formulated

as follows:

∂σ ∂ 2 u ( x, t )
A dx − dm − dF = 0 (7-1)
∂x ∂t 2

∂u ( x, t )
σ = Eε = E (7-2)
∂x

dm = ρAdx (7-3)

dF = R( x, t )dx (7-4)

where

σ = stress of unit pile cross-sectional area;

ε = strain of unit pile length;

m = mass of unit volume of pile body;

R(x,t) = soil resistance along pile shaft, as a function of time and depth.

A general expression of the soil resistance along pile shaft can be written as follows:

148
 R ( x, t ) = c1 f1 (u ( x, t )) + c 2 f 2 (v( x, t )) (7-5)

where c1 and c2 are the coefficients of stiffness and viscous damping,

⎧u ( x, t ), u ( x, t ) < Q s ,
c1 = K S , f1 (u ( x, t )) = ⎨ (7-6)
⎩Qs , u ( x, t ) ≥ Q s

c 2 = c1 J s f 1 (u ( x, t )) , f 2 (v( x, t )) = v( x, t ) (7-7)

where Ks is soil stiffness, Qs is shaft quake, and Js is shaft damping factor.

In most cases, the displacement of pile particles along the shaft is greater than the

recoverable soil deformation (i.e. the soil quake). Therefore, it is reasonable to assume

that the static shaft resistance be totally mobilized during pile driving. The pile shaft

resistance can be expressed as follows

∂u ( x, t )
R ( x, t ) = K s Qs (1 + J s v( x, t )) = RsS (1 + J s ) (7-8)
∂t

where RsS (= K s Qs ) is the static pile shaft resistance per unit embedded pile length.

7.3 Identification of Shaft Parameters

In Eq. (7-1), the analytical model only considers the soil resistance along the pile

shaft that is expressed in Eq. (7-8) as a function of static pile shaft resistance, R Ss , shaft

damping factor, Js, and the displacement of pile particles along the pile shaft. In the

following section, the unknown parameters, R Ss and Js, along the pile shaft will be

identified based on the solution of the proposed analytical model.

149
 Substituting Eqs. (7-2) ~ (7-8) into Eq. (7-1), the original equation for soil-pile

interaction can be written as follows:

∂ 2 u ( x, t ) ∂u ( x, t ) ∂ 2 u ( x, t )
ρA + Rs (1 + J s
S
) − EA =0 (7-9)
∂t 2 ∂t ∂x 2

or

∂ 2 u ( x, t ) RsS ∂u ( x, t ) 2 ∂ u ( x, t )
2
+ (1 + J ) − c =0 (7-10)
ρA
s
∂t 2 ∂t ∂x 2

where, c = E / ρ is the longitudinal wave velocity in the pile.

The initial conditions are:

⎧⎪u ( x,0) = 0
⎨• (7-11)
⎪⎩u ( x,0) = 0

where the dot • represents the time derivative.

The boundary conditions at the pile head are given by the measured force, Fm(t), and

the measured velocity, Vm(t), as follows:

⎧ ∂u (0, t ) Fm (t )
⎪⎪ ∂x = − EA
⎨ (7-12)
⎪ ∂u (0, t ) = V (t )
⎪⎩ ∂t m

Let

u ( x, t ) = e − μt w( x, t ) (7-13)

where

150
 RsS J s
μ= (7-14)
2 ρA

Eq. (7-10) becomes:

RsS μt
wtt ( x, t ) − μ 2 w( x, t ) + e − c 2 wxx ( x, t ) = 0 (7-15)
ρA

The Laplace’s transform (Thomson 1950) of Eq. (7-15) gives:

RsS 1
c 2 wxx ( x, s ) − ( s 2 − μ 2 ) w( x, s ) = (7-16)
ρA s − μ

The general solutions of the above equation are:

x s2 −μ 2 x s2 −μ 2
− RsS
w( x, s) = A1e c
+ A2 e c
− (7-17)
ρA( s + μ )( s − μ ) 2

The Laplace’s transform of the boundary conditions can be expressed as:

⎧ ∂w(0, s ) Fm ( s − μ )
⎪⎪ ∂x = − EA
⎨ (7-18)
⎪w(0, s ) = Vm ( s − μ )
⎪⎩ s−μ

Hence, the coefficients A1 and A2 can be solved as follows:

⎧ c Fm ( s − μ ) RsS 1 V (s − μ )
A
⎪ 1 = − + + m
⎪ 2 EA s 2 − μ 2 ρA 2( s + μ )(s − μ ) 2
2( s − μ )
⎨ (7-19)
⎪ A = c Fm ( s − μ ) + Rs V (s − μ )
S
1
+ m
⎪ 2 2 EA s 2 − μ 2 ρA 2( s + μ )( s − μ ) 2 2( s − μ )
⎩

151
 Then, w(x,s) can be determined as:

x s2 −μ 2
c Fm ( s − μ ) RsS 1 V (s − μ )
w( x, s) = ( − + + m )e c
+
2 EA s − μ
2 2 ρA 2( s + μ )( s − μ ) 2
2( s − μ )
x s2 −μ 2
c Fm ( s − μ ) RsS 1 Vm ( s − μ ) − RsS
( + + )e c
+
2 EA s 2 − μ 2 ρA 2( s + μ )( s − μ ) 2 2( s − μ ) ρA( s + μ )( s − μ ) 2

(7-20)

Taylor’s series expansion theory is applied to simplify the above equation:

⎧1 x s 2 −μ 2
−
x s2 −μ 2
1 x 2 (s 2 − μ 2 )
⎪ (e c
+e c
) ≈ 1+
⎪2 2! c2
⎨ (7-21)
s 2 −μ 2 x s2 −μ 2
⎪1 x − x s −μ
2
1 x s −μ 3
2 2 2

⎪ (e c
−e c
)≈ + ( )
⎩2 c 3! c

Thus, w(x,s) becomes

x 1 c x 3 2 RsS
w( x, s ) = − Fm ( s − μ ) − ( ) Fm ( s − μ )((s − μ ) 2 + 2μ ( s − μ )) +
EA 6 EA c ρA( s + μ )( s − μ ) 2
x 2 RsS 1 Vm ( s − μ ) Vm ( s − μ ) x 2
+ 2 + + (( s − μ ) + 2 μ )
2c ρA s − μ s−μ 2 c2
(7-22)

Neglecting the high order of (s-μ) and performing the inverse Laplace’s transform of

w(x,s), w(x,t) can be obtained as follows:

152
w( x, t ) = L−1 ( w( x, s ))
x c x 3 μt ' RsS x 2 μt t x 2 μt
∫
μt μt
= −e Fm (t ) − ( ) μe Fm (t ) + e + e V (τ ) d τ + e [2 μVm (t ) + Vm' (t )]
ρA 2c 2 m 2
EA 3EA c 0 2c
(7-23)

Then,

u ( x, t ) = e − μt w( x, t )
x c x 3 RS x2 t x2
=−
EA
Fm (t ) − ( ) 2μFm' (t ) + s
6 EA c ρA 2c 2
+ ∫0
V m (τ ) d τ +
2c 2
[Vm' (t ) + 2 μVm (t )]

(7-24)

The boundary condition at the pile toe is considered to be motionless before the

downward wave reaches the toe. It means, for t < L/c , the displacement at the pile toe is

considered to be zero.

L c L 3 R S L2 t L2
u ( L, t ) = −
EA
Fm (t ) − ( ) 2 μFm' (t ) + s
6 EA c ρA 2c 2
+ ∫0
V m (τ ) d τ +
2c 2
[Vm' (t ) + 2 μVm (t )] = 0

(7-25)

Eq. (7-25) can be used to determine the representative parameters, RsS and Js, for the

time period before the stress wave introduced at the pile head reaches the pile toe. Since

Fm(t) and Vm(t) measured at the pile head are used in Eq. (7-25) as input, the motion at

the pile toe is not recorded by the pile head transducers during the time period 0<t<2L/c.

By adjusting these two parameters, RsS and Js, Eq. (7-25) can be satisfied within the

specific tolerance for this time period.

153
7.4 Identification of Pile Capacity at the Pile Toe

After the expression of displacement u(x,t) has been obtained without unknown

parameters, the force at the pile toe, for t ≥ L / c , can be expressed as follows:

∂u ( x, t )
Rt ( L, t ) = EA
∂x x=L
(7-26)
L2 L
= − Fm (t ) − 2 2 μFm' (t ) − RsS L + EA 2 (Vm' (t ) + 2 μVm (t )]
2c c

The velocity at the pile toe can be expressed as follows:

∂u ( x, t ) L ' L3 L2
v ( L, t ) = =− Fm (t ) − 2 μ Fm
"
(t ) + V m (t ) + {Vm" (t ) + 2 μVm' (t )]
∂t x=L
EA 6 EAc 2
2c 2

(7-27)

The soil resistance at the pile toe, similar to that along the pile shaft, can also be

expressed as follows:

R ( L, t ) = c3 f 3 (u ( L, t )) + c4 f 4 (v( L, t )) (7-28)

where c3 and c4 are coefficients of stiffness and viscous damping at the pile toe,

respectively.

⎧u ( L, t ), u ( L, t ) < Qt ,
c3 = K t , f 3 (u ( L, t )) = ⎨ (7-29)
⎩Qt , u ( L, t ) ≥ Qt

c 4 = c3 J t f 3 (u ( L, t )) , f 4 (v( L, t )) = v( L, t ) (7-30)

where Kt is soil stiffness at the pile toe, Qt is toe quake, and Jt is toe damping factor.

154
 When the pile toe resistance is totally mobilized, the force at the pile toe can be

expressed as follows:

R ( L, t ) = K t Qt (1 + J t v( L, t )) = RtS (1 + J t v( L, t )) (7-31)

where RtS is the static resistance at the pile toe.

For the case in which the pile toe resistance is totally mobilized, Eq. (7-31) can be

used to determine the static resistance, RtS, and damping factor, Jt, at the pile toe, when

making the calculated force match the calculated velocity within the specific tolerance.

Sometimes, the pile toe resistance, however, is not totally mobilized, that is, the

displacement at the pile toe does not reach the elastic deformation limit (i.e., the toe

quake, Qt). The following equation, in lieu of Eq. (7-31) can be employed to describe the

relationship for this case:

R ( L, t ) = K t u ( L, t )(1 + J t v( L, t )) (7-32)

The coefficient of stiffness, Kt, and the damping factor, Jt, at the pile toe can be

determined by making the calculated pile toe forces match with the calculated pile toe

displacement and velocity within the specific tolerance. The static resistance at the pile

toe can be obtained by multiplying the calculated Kt with the maximum calculated

displacement at the pile toe.

155
7.5 Case Study

A steel pipe pile was driven about 15.6 m below the ground surface. The soils along

the pile shaft and under the pile toe are clay. The Young’s modulus of the pile E =

210,000,000 Pa; the cross-section area A = 48.5 cm2; the mass density ρ = 7850 kg/m3.

The measured HST data are shown in Fig. 7-2.

By extracting the force and velocity time histories before the downward wave reaches

the pile toe, two unknown parameters, Rs and Js, can be determined when satisfying Eq.

(7-25) within the specified tolerance. The static resistance and damping factor along the

pile shaft are 59.2 KN and 3.882 s/m, respectively. After Rs and Js are determined, Rt and

Jt can be further calculated as 568.1 KN and 0.558 s/m, respectively. The calculated

results from the proposed method are summarized in Table 7-1 with the CAPWAP

analysis results and static load test results. It can be seen that the total resistance

calculated from the proposed method shows good match with the results from static load

test and CAWAP analysis. The distribution of pile resistance and Smith damping factor

from the proposed approach do not match CAPWAP results very well. The sources of

inconsistency may be identified as follows: 1) CAPWAP results may not be correct; 2)

the assumption of homogeneous soil condition along the pile deviates from the actual

conditions; 3) the model of soil-pile interaction needs to be further improved.

7.6 Comparisons

The performance of the proposed method is evaluated by the comparisons with the

CAPWAP approach which is widely accepted as one of the most accurate procedure of

156
dynamic pile testing. An index, denoted as K, is used to define the ratio of the CAPWAP

capacity to the predicted capacity from the proposed method. A total of 26 pile testing

cases were used in this comparison study. The scattergram of the CAPWAP results

versus the predicted capacities from the proposed method is shown in Fig. 7-3. It can be

seen that the K values for most of pile cases fall in the range 0.7~1.5, which means that

the capacities calculated from the proposed method match well with the CAPWAP results.

The mean value, standard deviation, and COV value of index K are 1.167, 0.498, and

0.427, respectively. It should be pointed out that the high COV value is not unusual for

the pile-foundation engineering. On the other hand, the computational efficiency and the

close-form solution of one-dimensional wave equation make the proposed method

attractive.

7.7 Conclusions

A wave equation-based algorithm has been developed in this chapter to estimate the

driven pile capacity based on the High Strain Testing (HST) data. The Smith model is

employed to illustrate the soil-pile interaction effect in the one-dimensional wave

equation. The static shaft resistance is reasonably assumed to be totally mobilized during

pile driving at the most cases. The pile static resistances and Smith model parameters can

be directly determined in an analytical solution of one-dimensional wave equation. The

proposed wave equation method may also separate the pile shaft and toe resistances. The

numerical example is presented to illustrate the application of the proposed wave

equation method. The predicted capacities from proposed method exhibit a favorable

comparison when compared with the CAPWAP analysis results.

157
Table 7-1 Comparison of Results from Different Methods
Shaft Toe Total
Shaft damping Toe damping
Methods resistance resistance resistance
factor Js (s/m) factor Jt (s/m)
(KN) (KN) (KN)
Proposed
59.2 3.882 568.1 0.558 627.3
method
CAPWAP
361.2 0.561 211.7 0.108 572.9
method
Static load
516.2
test

158
F(t)

∂ 2 u ( x, t )
dm
∂t 2
u(x,t) Aσ
x dF = R( x, t )dx
∂σ
A(σ+ dx )
∂x
R(x,t)

Figure 7-1 Soil-Pile Interaction Model

159
 1000
Measured force
800
Measured velocity*EA/c
600
Force (KN)

400

200

-200
0 20 40 60 80 100
Time (ms)
Figure 7-2 Measured Force and Velocity at the Pile Head

160
 3000
K=1.5

2500 K=1.0
CAPWAP Predicted Capacity (KN)

2000

K=0.7
1500

1000

500

0
0 500 1000 1500 2000 2500 3000

Capacity Calculated by Proposed Method (KN)

Figure 7-3 CAPWAP Results vs. the Results Calculated from the Proposed Method

161
 CHAPTER VIII

NEW MEASUREMENT AND REMOTE SENSING TECHNOLOGY

8.1 Introduction

The NetDaq system is designed to acquire data remotely through wireless network. It

is intend for the application where real time data monitor is necessary, but the location is

hard to reach or human being cannot stay at the site because of safety or other reasons.

The system consists of two major components: Stationary PC (Laptop or Desktop PC)

and Remote Data Acquisition Unit.

The stationary PC has the application program allowing for user’s interaction with

system for data acquisition, data display and data process on real time basis. The

developed system provides four channels of data flow plus the channel monitoring the

remote system power supply.

The Remote Data Acquisition Unit will take the commands from stationary PC. Then

it will command the data acquisition card to get data. Finally, it will send the data back to

the stationary PC.

162
8.2 Hardware

Fig. 8-1 shows the developed wireless dynamic pile testing equipment. The new

system consists of sensors and wireless communication for remote data acquisition and

management. Accordingly, hanging cables from the sensor location to the data

acquisition unit are no longer needed, so that dynamic pile testing can be conducted in a

safer environment. The system is physically divided into two modules. The first module

is the remote data acquisition unit. The second module is the stationary PC for the data

processing.

Remote Data Acquisition Unit gathers seven components as follows:

• MicroSys SBC 1586 single board computer With PCMCIA adapter and network card

adapter. OS: Microsoft Windows 2000.

• National Instrument data acquisition card: DaqCard AI-16XE-50.

• National Instrument Signal conditioner: SCC 2341 (for strain gauge), SCC-XXX (for

Accelerator gauge).

• Strain gauge sensors and Accelerator gauges (same as PDI’s sensors)

• Wireless Network: D-Link wireless bridge, Model DWL-810+.

• Customized signal conditioner module (Compact version for the NI SCC 2345

module)

• Power: 10 NiMH Batteries (12 V 4000mAh)

163
 Stationary PC includes two components as follows:

• Laptop PC: Dell model Inspiron 700M operated with Microsoft Windows XP

• Remote wireless Presentation pointer: Kensington Model 33062

a) Remote Data Acquisition Unit b) Stationary PC

Figure 8-1 New wireless dynamic pile testing system

The block diagram of wireless data acquisition system is shown in Fig. 8-2.

164
 Remote Module Stationary Computer
Data Acquisition Data Monitor

MicroSys SBC 1586

Single Board Computer
With Cardbus interface
5V
Operating under
Windows 2000

Wireless

Control Computer with

National Instrument D-Link wireless TCP/IP interface
DaqCard Wireless Bridge
AI-16XE-50
• Data Input
5V • Configuration
National Instrument 12 V Power • Connection/Discon
Signal Conditioner supply nection
SCC-2345 • SelfTesting
• Testing
• Analysis
Strain Gage Accelerometer Accelerometer

Strain gage 2 Accelerometer 4

Figure 8-2 Block Diagram of Wireless Data Acquisition System

i) Hardware Function of Remote Data Acquisition Unit

SBC 1586 is a core part for the remote unit. It commands the DaqCard to obtain data

and establish the network connection. The signal from sensors is first processed by the

165
signal conditioner to reduce noise and spike. Then it feeds into the DaqCard. DaqCard

AI-16XE-50, connected to the PCMCIA interface of SBC1586, converts analog signal

into digital data. Finally, the digital data is transmitted to the stationary PC through

wireless bridge. The wireless network is also used to communicate with the PC to take

commands from stationary PC and send the status of remote unit to stationary PC.

ii) Hardware Function of Stationary PC

The stationary PC is major platform for the data acquisition. It interacts with the user

to get user’s command. It also displays graphs of raw data and processed data. The

wireless presentation remote is used to increase the depth value remotely during the data

acquisition process.

8.3 Software

i) Software programs on Remote Data Acquisition Unit

• Microsoft Windows 2000

• Symantec pcAnywhere server version 11.0

• National Instruments Measurement & Automation application package

• NetDaq server application

• Other software tools and components: Daq controls, Winsock, compression tools etc.

166
ii) Software programs on Stationary PC

• Microsoft Windows XP

• Symantec pcAnywhere client (version 11.0).

• PileTest application program

• NetDaq client application

• Other software tools and components: Graph server, DAQ controls, etc.

iii) Function of software programs on Remote Data Acquisition Unit

The Windows 2000 is the foundation of the remote unit. It communicates with

DaqCard through national instrument driver software. Windows also sets up the

communication with network. NetDaq server communicates with NetDaq client through

wireless network to receive commands and parameters from the DaqCard and send data

from DaqCard to NetDaq client.

Symantec pcAnywhere is the tool for remote control of the Remote unit, including

configuration and Parameters settings.

iv) Function of software programs on Stationary PC

The software for data processing on the stationary PC was called PDC and coded with

Visual Basic, C language, and a real time graphic display facility. The software of PDC

includes three capacity-decision methods: UA energy method, Case Method, and UA

wave equation method.

167
a) UA Energy Method

1 1
(S + Q s ) (S + Q s ) 2 + 4E in β
RS = − 2 + 2 (8-1)
2β(1 + js V) 2β(1 + js V)

where:

Ein = energy transferred to pile during hammer impact;

V = the maximum velocity at the pile head;

RS = static pile capacity;

js = shaft damping;

Qs = shaft quake;

S = the permanent set per hammer blow;

L
β= for triangular shaft resistance distribution, typical for sand;
10EA

L
β= for rectangular shaft resistance distribution, typical for clay;
6EA

in which L = embedded pile length; E = Young’s modulus of the pile; A = cross section

area of the pile.

b) Case Methods (Rausche et al. 1985)

1) Total resistance, RTL

RTL=[FT1+FT2]/2+[(VT1-VT2)*EA/c]/2 (8-2)

where T1=TP+DELAY and TP is the time of the first relative maximum velocity peak

and DELAY is a time required in uneven impacts or in soils capable of large deformation

before achieving full resistance.

168
 2) Damping Factor Method, RSP

RSP=RTL-Jc*(VT1*EA/c+FT1-RTL) (8-3)

where Jc is case damping constant.

c) UA Wave Equation Method

UA wave equation method can be used to determine the pile capacity in the post

analysis. The Kelvin model is employed to represent the soil-pile interaction. The two

unknown model parameters are spring constant K, and the damping coefficient C. By

matching the calculated displacement with the measured displacement at the pile head,

the two unknown parameters can be determined. The identified soil-pile interaction

parameters can be used in the forward calculation scheme to predict the pile shaft and toe

capacities using the measured force and velocity histories.

The PDC software also includes probability analysis and Bayesian updating approach.

The detailed description of the above methods and approaches can be found in the report

FHWA/OH-99/008 (Liang, 1999).

8.4 System Installation and Configuration

The SBC 1586 of the remote data acquisition unit is a compact, embedded computer.

The installation of Windows 2000 is much different with regular PC. Special procedures

have to be followed properly to install Windows 2000. The Stationary PC communicates

with the Remote wireless data acquisition unit through peer-to-peer network, and there is

169
no router or access point between computers. Proper configuration is necessary to

establish the connection.

i) Remote Unit

a) Windows installation on SBC 1586

The SBC 1586 core is Pentium 586 processor. The Windows 2000 system can be

installed on it. But it can not be directly loaded to the Compact Flash (CF) card because it

considers the CF card as removable storage. So the Windows and other software

components have to be installed on a hard drive first. Then, this hard drive image could

be transferred to the CF card. After this transfer, the SBC 1586 becomes fully functional

Windows computer system. The processes are described below:

1. Set up the SBC 1586 like regular PC by connect keyboard, mouse, monitor, hard

driver, CD ROM and others.

2. Install Windows 2000 professional, NI measurement & automation, Symantec

pcAnywhere on the hard drive.

3. Configure the network for the TCP/IP connection, pcAnywhere, and DaqCard

driver. And test them separately.

4. Shut down the machine, insert the CF card, and restart the SBC 1586.

5. Run the image transfer utility (provide name) to transfer image of hard drive to

CF card.

6. Shut down the machine, disconnect the hard drive, and restart machine. The

system should run on the Windows on the CF card

170
 Now the SBC 1586 becomes a compact fully functional system. All of the controls

can be done through pcAnywhere.

b) Other software installation on SBC 1586

The NetDaq can be installed on the SBC 1586 using the following procedure:

1. Copy the NetDaq folder of the developer machine to the SBC 1586.

2. Register the Winsock, Compress, and other components (list all details).

3. Setup the NetDaq server to be run at windows startup by dragging the

[Link] to Startup program item.

c) Network connection configuration on SBC 1586

Computer Name: SBC1586

Work Group: Workgroup

IP address: [Link]

d) Wireless configuration of D-Link 810+

Operating Mode: Ad-hoc (peer to peer)

SSID: NetDaq

Channel: 6

WEP: Enabled

WEP Encryption: 64Bit

WEP Mode: HEX

Key1: 0000000000

171
Key2: 0000000000

The above parameters can be set through Configuration utility at: [Link] .

Login: Admin with blank password

e) Symantec pcAnywhere server configuration

Start Symantec pcAnywhere by click Start -> Symantec pcAnywhere to launch the

Symantec pcAnywhere. At the host window, right click on the “Network,Cable,DSL”

item, then property. At the property window, the parameter can be changed to receive the

connection from Symantec pcAnywhere client.

f) DaqCard configuration

Click Start -> Programs -> National Instruments -> Measurement & Automation to

launch the Measurement & Automation Explore. On the left panel of the Measurement &

Automation Explore, expand the Device and Interface; the DaqCard AI-16XE-50 should

be listed. Right click on the DaqCard Item, then go to and left click properties. The

Configuring Device 1:… window should appear. From this window, the basic parameters

for the DaqCard can be changed.

ii) Stationary PC

Follow the standard installation instructions to install Windows on the stationary PC

(Laptop PC) if necessary.

172
a) Network connection configuration on Laptop PC

Computer Name: SD-Notebook1

Work Group: Workgroup

IP address: [Link]

b) Wireless configuration

Operating Mode: Ad-hoc (Computer to computer)

SSID: NetDaq

Channel: 6

WEP: Enabled

WEP Encryption: 64Bit

WEP Mode: HEX

Key1: 0000000000

Key2: 0000000000

Configuration Parameter: [Link] Login: Admin with blank password

c) Symantec pcAnywhere client configuration

Start Symantec pcAnywhere by click Start -> Symantec pcAnywhere to launch the

Symantec pcAnywhere. At the Remotes window, right click on blank space, then new to

create new TCP/IP connection. Rename it properly, and right click on this new item, then

click on properties to start the connection configuration screen. To connect with the

remote unit, the IP address of remote unit has to be used. The login name and password

have to be matched with the one assigned on the server or host of pcAnywhere.

173
8.5 NetDaq Client/Server Architecture

The NetDaq client and server are the same as typical client/server application. The

server runs on server machines to serve the request from client. The client makes request

to server to get data.

When the server started, it starts listening to the network for connection. If it gets

connected to client, it changes itself to stand by mode waiting for command. During this

process, a timer is used to periodically check the status of connection. If it finds the client

disconnected, it automatically changes to listening mode, waiting another client to

connect.

i) Server

When the server started, it tries to listen to the connections from client. If it detects

the client is connecting, it will respond and establish connection. Now the client can send

requests to ask server to do the tasks. After client finishes the job, it will disconnect from

the network. Then the server will detect this action and change to the “listen” status

waiting for another client to make call.

Functions of server application components:

FrmNetDAQServer: It holds the CWAI, WinSock, and Timer controls. Through

Winsock control, the server application communicates with client

174
 application to get and send commands and data. CWAI control is

used to configure DaqCard and acquire data from DaqCard.

ModMain: Having the basic functions for server application.

ClsDaqCard: Processing commands for DaqCard from client

ModZip: It will compress the data before sending to network for large

amount of data.

ii) Client

When client needs to request some information from server, it will start by connecting

with server. If a connection is made successfully, it can send command to server to

request data or other action.

Function of client application components:

UcNetDAQ: User control which holds Winsock and Timer control

communicates with server application.

ModMain: Having the basic functions for client application.

ClsDB: It processes server’s data and saves it to MS Access database for

further process.

ModZip: It will compress the data before sending to network for large

amount of data.

iii) Command Format

The client and server communicate with each other in the predefined command

protocol. The format of these commands is pipe delimited string commands with

175
parameters, such as Data|…, Command|Start, ScanClock|Period|5. The first string before

the first pipe is the top level command, the second string between first pipe and second

pipe is second level command, and so on. After commands for each level, it should

follow by the parameters, each separated by pipe, if the parameters are required. Taking

the ScanClock|Period|5 as example, the “ScanClock” is the top level command; the

“Period” is the second level command, and the “5” is the parameter. In this case, it only

requires one parameter.

Table 8-1 List of commands from Client

Top level Second level Parameter 1 Parameter 2

command command (example) (example) Note
Command Start Command to
DaqCard
Configure
Reset
Stop
NScans
Device
Channel RemoveAll DaqCard channel
setup
Add
ChannelString “0,1,2,8”
Coupling 1
Gain 0
LowerLimit 2
Polarity 1
Range 0
InputMode 1
ScaleMultiplier 2
ScaleOffset 1
UpperLimit 10
ScanClock ActualFrequency DaqCard scan
parameters
ActualPeriod
ActualTimebaseD
ivisor
176
 ClockSourceSign
al
ClockSourceType
Frequency
InternalClockMo
de
Period
TimebaseDivisor
TimebaseSignal
TimebaseSource
StartCondition ActualHysteresis DaqCard start
condition
ActualLevel
Hysteresis
Level
Mode
PreTriggerScans
SkipCount
Source
Type
StopCondition ActualHysteresis DaqCard stop
condition
ActualLevel
Level
Mode
PreTriggerScans
SkipCount
Source
Type
DataFormat Format of returned
data
DataFilter Data space
Close Close network
Connection connection
Battery 1 or 2 Get the battery
voltages for 5 or 12
volts supplies
ResetTrigger Reset trigger level

177
Table 8-2 List of commands from Server:

Top level Parameter 1 Parameter 2

command (example) (example) Note
DATA 1.0|2.3|3|… Acquired data from
DaqCard
MSG Connected Server message
ERROR Divided by 0 Server error message
Battery 1.2 Battery volts

8.6 Trouble Shooting

i) Remote Unit

1. Make sure that the batteries are fully charged. It takes about 15 hours to have the

batteries fully changed, therefore, it has to be done early.

2. Install the remote unit and attach the remote unit to its holder.

3. Turn on the switch on the remote unit. The power indicator should light up.

4. After finished, turn off the system by switching it to “off”.

The NetDaq server should be running on the remote unit silently. The user interface

of the server is basically for tracking the error and checking the status of the application.

Fig. 8-3 shows the user interface of the server.

Inside the window, just below the blue title area, the top line displays the network

information of local unit: name, IP address, and port.

178
 Figure 8-3 User interface of the server

The second line displays the information about remote machine if it is connected. The

third line displays trigger information of sensors. The main text box below holds all of

the text sent out to the client PC (Stationary PC). It is refreshed every time when the

server sends something to client. The little text below, captioned with “In” is used to

display the information receive from Stationary PC. Below the “In” box, on the left side,

is the area to show the network connection status, e.g., closed, connected, listening. On

the right side, below the “in” box, there are several buttons.

“Send”-- will send whatever in the main text box to client.

“Acquire” -- will start data acquisition on the DaqCard continually

“Stop” -- to stop the data acquisition process started by ‘Acquire”

“One Acq” -- will run the data acquisition one time only.

179
 “Close” -- exit the server application.

ii) Stationary PC

Start up:

1. Turn on the PC and login

2. Click the PileTest icon on the desktop to start data acquisition.

3. After data acquisition, shut down the remote unit first, then shut down the

stationary PC.

Check network:

1. On the laptop PC, click start -> Run…

2. At the open box of the run window, type ping <IP address>, then ok.

Example: ping [Link]

3. A DOS window will show up with the results. If the network connection is

established, one can see the message on this DOS window.

4. If the network connection can not be established, please refer to the

troubleshooting part to correct the problem.

Remote control of remote unit by pcAnywhere:

1. Start up the Symantec pcAnywhere by click its icon on the desktop, or click Start

-> programs -> Symantec pcAnywhere. The pcAnywhere window will be

displayed.

180
 2. On the left side, there is pcAnywhere manager panel. If the item “remotes” is not

highlight, click on it to highlight. Now the right panel should have caption

“remotes” with different remote location items.

3. Double click the item “SBC 1586” to start connecting to remote unit.

4. After a while, the pcAnywhere connecting window will appear. Then the

Symantec pcAnywhere Remote windows will display the desktop windows of

remote unit.

5. Check and configure remote unit as necessary.

6. Close the pcAnywhere client application by click the x on the top right corner or

exit on the file submenu.

Description of the user interfaces of NetDaq Client:

The NetDaq client is component referred by PileTest application on stationary PC.

The user interface of the NetDaq client sits behind the regular PileTest window. On both

selftest and Test window, there is check box on the left side of big command button like

Start (see Fig. 4). When the check box is checked, the user interface of NetDaq client will

show up in the place where the parameters are shown before.

181
 Figure 8-4 Control of user interface of NetDaq client

The items on the part of UI are shown in Fig. 8-5. The top two rows are the network

information of local and remote computers. The two “out” and “in” text boxes are used to

display the message send to server (out) and come from server (in). The two buttons

“Connect” and “Disconnect” are used to connect and disconnect to server. If the “Save

Data” check box is checked, the data from server will be saved to the MS Access

database in the application’s database folder of NetDaq client. Periodically, the NetDaq

client will send command to server to retrieve battery voltage. If the battery voltage drop

to certain voltage, the speaker will sound a beep to warn user that the battery is running

low. This period is defined by the text box right after the in text message box(in minute).

The “Reset Trigger” button will force the NetDaq server to acquire the current value of

the trigger channel so the new trigger value can be obtained.

182
 Figure 8-5 User interface of NetDaq client

8.7 Other errors

Sometime, the stationary PC can not connect to the remote unit. There are several

items that could cause this problem.

First use ping to check the connection and to see whether the user can not ping to the

remote unit ([Link]). If it works the problem should be in the NetDAQ server. Use

the pcAnywhere to connect to the remote unit to check the server is running or not.

If you cannot ping to the ([Link]), try to ping the wireless adapter at

([Link]). If it works, the remote SBC1586 may not start yet. Wait for 3-5 minutes

to retry. If it cannot ping to the ([Link]), the wireless network cannot communicate

with each other. Try to start Internet Explore and go to the wireless configuration site at

[Link]. If it still does not work, the setup parameter on the wireless box must be

wrong. Then, open the remote unit box; unplug the network cable and plug the one from

your local computer; use the Internet Explorer and go to the [Link]; go to the

wireless tab to reconfigure the wireless connection; restart the remote unit.

183
8.8 Laboratory Testing of the System

Laboratory testing of the whole system has been performed. The results are positive.

The two strain gauges and two accelerometers are mounted on an iron tube with a

diameter of 1 inch and 5 foot long. A cap to avoid the damage of the tube and to

distribute the force of the hammer protects both ends of the tube.

The two strain gauges are mounted on the opposite sides of the tube, about 10 inches

below the top of the tube. The two accelerometers are installed about the same height, but

perpendicular to the two strain gauges. See Fig. 8-6 for the details.

Figure 8-6 Installation of strain gages and accelerometers

One problem we find during lab test is trigger level. The system was designed to use

one of the strain gauge signals as trigger signal. Every time there is force applied on the

strain gauge, the signal of the strain gauge will increase. At certain predefined level, the

184
data acquisition card will be triggered to start collecting data. These data plus the data

collected by pre-trig will be constructed as a whole trace of the data points.

When the system was in test, initially certain predetermined trigger level had been

used. This trigger level is just above the level when the strain gages are in a total released

state. After the strain gauge is mounted, every hammer hit will create stress on the strain

gauge. This stress will take much longer time to release because the strain gauge is

bounded onto the pole. This time interval is usually much longer than that of a hammer

hit. Therefore, the signal level of strain gauge, when the second hammer hit, will be

different because of this stress. It reflects that signal level some time is high and some

time is low. This will lead to the system sometime not responding to the hammer hit after

first hammer hit and data is lost.

To solve the problem, a dynamic trig level mechanism is introduced. For every

hammer hit, the last certain data points are collected and averaged to form the base signal

of the trigger level for next hammer hit. This signal level of last hammer hit adds a small

increment will be the trigger level for the next hammer hit. This will guarantee that the

signal of next hammer hit will be acquired at appropriate time. All of this mechanism

now has been built into the application program of a remote unit.

Another problem we found during lab test is data transmission speed from remote

unit to the base computer. Initially the system is using the sequential method mechanism

and data compression for the data transmission from remote unit to base computer

185
wirelessly. This will give about 4 seconds for the system to acquire data, to transfer data

to base computer, and to display and to store data in the base computer. This speed is too

slow in practice because the hammer should hit the pole about 1~2 second interval. After

checking the software carefully, it turns out that the bottleneck is at the remote unit.

Because the remote unit is the embedded computer, the CPU has much lower speed than

the regular desktop or laptop computer. The data compression process will take about 1

second to be finished. Data compression truly gives much faster data transmission speed

wirelessly. Since the data transmission speed on the wireless network is very fast (11 Mb

per second), it really does not have to compress the data. After the data compression on

the remote unit is taken out, the system response time was reduced less than one second.

Now most of time used in the system is to display the data points on the base computer. It

takes about 0.5 second to show the trace. This also can be solved by multiple threading

and/or show the traces selectively (show one trace for two or three data acquisition).

The detailed operation instruction on the program of Pile Driving Controller in the

stationary PC can be found in the section VI.3 of the report FHWA/OH-99/008 (Liang,

1999).

8.9 Conclusions

Evolving from the operating risk encountered in the current dynamic testing

technology, a set of new wireless dynamic testing equipment has been successfully

developed. This new technology utilizes remote data acquisition and management unit to

build the wireless connection between the sensors and the monitoring PC. The new

186
measurement technology is also provided with a user-friendly operational software

package, supplemented with a computer graphical interface that allows continuous

monitoring and detection during pile driving or testing. The laboratory testing has been

performed to validate the applicability of hardware and software of the wireless dynamic

pile testing equipment and to verify the accuracy of signal recordings by comparing the

laboratory testing results from the developed wireless equipment with those from the

PDA testing equipment. The new wireless dynamic testing equipment is anticipated to be

successfully applied in the field testing that will advance pile driving control and

monitoring to a more secure environment.

187
 CHAPTER IX

SUMMARIES AND CONCLUSIONS

9.1 Summaries and Conclusions

Driven piles, as an important type of deep foundations, have been commonly used for

the buildings, bridges, towers, and dams. In AASHTO, the design of driven piles has

been moving from the traditional ASD method to the more formal and rational LRFD

method. The reliability analyses in LRFD allow the evaluation of the relative

contributions of uncertainties of load and resistance components to the design of driven

piles in the framework of probability theory. The application of reliability analyses has

dramatically motivated the evolution and the improvement on the design and quality

control of driven piles. However, the reliability analyses on driven piles needs to be

further developed in order to make the LRFD method more understandable and

applicable by designers. Towards the objective of improving the design and quality

control of driven piles, a series of research works on the reliability-based analyses have

been carried out in this research.

In Chapter III, based on the acceptance-sampling analysis, the reliability-based

quality control criteria on driven piles were developed for both static and dynamic test

methods. The lognormal distribution function was used to characterize the statistical

188
characteristics of the methods. The relationship between the number of load tests and the

criterion on acceptance of the measured capacity was elucidated. It was shown that the

measured capacity required to assure adequate quality control decreases with an increase

in the number of load tests performed. Either the number of load tests or the target

measured capacity or both needs to be increased in order to obtain a higher target

reliability index. An optimum approach was developed for the selection of the number of

load tests and the required measured capacities for quality control using various types of

load test methods of driven piles.

In Chapter IV, a statistical database was compiled to describe the increase of axial

pile capacity with time, known as set-up, when piles were driven into clay. Based on the

collected pile testing data, pile set-up is significant and continues to develop for a long

time after pile installation. The statistical parameters for set-up effect combined with the

previously documented statistics of the loads and the resistances were systematically

accounted for in the framework of reliability-based analysis using First Order Reliability

Method (FORM). Separate resistance factors in Load and Resistance Factor Design

(LRFD) were obtained to account for different degrees of uncertainties associated with

the measured short-term capacity and the predicted set-up capacity at various reliability

levels. The design procedure to incorporate the set-up into the LRFD of driven piles in

clay was proposed. The incorporation of set-up effect in LRFD could improve the

prediction of design capacity of driven piles. Thus, pile length or numbers of pile could

be reduced and economical design of driven piles could be achieved.

189
 In Chapter V, a statistical database was developed to describe the set-up when piles

driven into sand. Based on the compiled database of driven piles in sand, the various

prediction methods for the development of the set-up of driven piles in sand were

evaluated. The same reliability-based analysis approach as proposed for driven piles in

clay was employed to develop the separate resistance factors for the measured short-term

capacity and the predicted set-up capacity for driven piles in sand. The design procedure

was also proposed for incorporating the set-up into LRFD of driven piles in sand.

In Chapter VI, a mathematic algorithm based on Bayesian theory was developed to

allow for updating the resistance factors for the driven piles when static analysis methods

are combined with dynamic pile testing in a given project site. The results of dynamic

pile tests could be incorporated to minimize the uncertainties of static design methods by

updating the resistance factors in LRFD. The main benefit of combining the two methods

was the added confidence and accuracy on the predicted pile capacity, thus allowing for

the adoption of a more reasonable resistance factor as opposed to the case in which only

one analysis method is used for design. The example analyses have demonstrated that the

combination of two methods would make the predictions of driven pile capacity more

consistent with the static load test results.

In Chapter VII, a new one-dimensional wave equation based algorithm for estimation

of shaft and toe resistance of driven piles using High Strain Testing (HST) data was

presented. The pile was represented by a uniform elastic bar, while the soil around the

190
pile was assumed as a homogenized isotropic medium, and Smith model was adopted to

represent soil-pile interaction in the wave equation analysis. The static shaft resistance

was reasonably assumed to be totally mobilized during pile driving. Smith damping

factor was used to account for the combined dynamic and rate effect. Force and velocity

time histories measured at the pile head were used as input boundary conditions in an

analytical solution of the one-dimensional wave equation. Two unknown parameters,

static shaft resistance and Smith damping factor, were incorporated into the one-

dimensional wave equation and determined analytically for the time duration prior to the

first downward wave reaching the pile toe. Once these two shaft parameters are

determined, the soil resistance and Smith damping factor at pile toe can be determined.

Numerical examples were presented to demonstrate the validity of the proposed method

on the determination of the static resistances.

In Chapter VIII, a set of new wireless dynamic testing device for dynamic pile testing

was successfully developed. This new technology utilizes remote data acquisition and

management unit to build the wireless connection between the sensors and the monitoring

PC. The laboratory testing has been performed to validate the applicability of hardware

and software of the wireless dynamic pile testing equipment. The application of this new

technology to dynamic pile testing should significantly improve the operating safety and

advance pile driving control and monitoring to a more secure environment.

191
9.2 Recommendations for Implementations

Three new methodologies and one new product have been developed through the

efforts of this research and they are recommended for adoption by Ohio DOT.

1) New Wireless Dynamic Pile Testing Equipment

The newly developed wireless dynamic pile testing equipment incorporates the

innovative application of wireless communication technology that is outlined in detail in

Chapter VIII. The application of this new technology to dynamic pile testing should

significantly improve the operating safety. The elimination of the need for long cables

should also help speed up the testing process. It is recommended that ODOT adopt and

promote the new wireless dynamic pile testing equipment in the field.

2) New Quality Control Method on Driven Piles

The developed quality control method systematically account for the statistical

characteristics of various pile testing method. In the framework of reliability-based

analysis with consideration of finite population, the number of testing piles and the

required measured capacity are derived for a target reliability index. The recommended

quality control criteria can be adopted by ODOT in Bridge Design Manual.

3) Incorporating Set-up into Design of Driven Piles

The separate resistance factors for the measured CAPWAP capacity and the predicted

set-up capacity have been developed based on the reliability-based FORM in Chapters IV

and V for driven piles in sand and in clay, respectively. The step-by-step procedure has

192
been recommended for the incorporation of the set-up into LRFD of driven pile. The

incorporation of the set-up into design of driven piles will generally reduce the required

pile length or number of piles. The recommended resistance factor, along with the set-up

prediction method, can be incorporated in ODOT Bridge Design Manual.

4) Bayesian Based Formalism for Updating LRFD Resistance Factors for Driven Piles

The Bayesian based formalism to combine dynamic testing and static calculation

results for LRFD of driven piles has been developed. The current design practices of

driven piles treat the static calculation and dynamic testing independently, resulting in a

large variability in the estimated pile capacities. The combination of various design

methods could improve accuracy and confidence level on predicting pile capacity as

opposed to the case of using only one design method. Furthermore, the uncertainties on

the estimation of pile capacity can be effectively reduced so that shorter pile length or

less pile number can be used.

9.3 Recommendations for Future Research

Areas for future studies are identified in the following:

1) LRFD Resistance Factor Calibration for Deep Foundations in Ohio

In 2006, ODOT will supersede the traditional ASD with the new LRFD. The

resistance factors for deep foundations recommended by AASHTO are calibrated based

on the compiled database with a wide range of soil conditions, soil testing techniques,

analysis and design procedures, foundation types, and foundation construction procedures.

193
Therefore, there is a need for ODOT to establish the resistance factors which are

appropriate for Ohio soil conditions and ODOT foundation practice.

In order to calibrate the resistance factors for ODOT practice, comprehensive

databases for driven piles and drilled shafts should be respectively compiled based on the

testing data in Ohio. The pile or shaft testing data should be categorized based on soil

conditions, pile driving time, pile or shaft sizes, and pile or shaft types. The restrike

testing should be separately performed for the driven piles at 1 day after EOD and longer

time after EOD so that the development of set-up can be evaluated for the Ohio soil

conditions.

Based on the compiled databases, the recommended quality control method can be

validated for the ODOT foundation practice. The separate resistance factors for the

measured CAPWAP capacity and predicted set-up capacity can be further calibrated for

the appropriate application in Ohio. With the availability of more driven pile data on

static analysis results, dynamic testing results, and static load test results, the Bayesian-

based formalism can be further improved and validated.

2) Validation and Implementation of Wireless Dynamic Pile Testing Equipment

The newly developed wireless dynamic pile testing equipment utilizes wireless data

communication technology to eliminate the need for hanging cables from the sensor

locations to the data acquisition unit. As a result, dynamic pile testing can be conducted

more efficiently in safer environment. The new measurement device also provides a user-

194
friendly operational software package, supplemented with a computer graphical interface,

that allows continuous monitoring and detection during pile driving. Due to its safety and

cost-saving features, the new equipment is particularly suited for ODOT implementation.

This new wireless dynamic pile testing equipment was developed based on the laboratory

testing and calibration; therefore there is a need to validate the applicability of the new

technology for field applications.

3) Reliability-Based Analysis of the Effect of Pile Group on Calibration of Resistance

Factors in LRFD of Drive Piles

It is generally accepted that a pile group system can be more robust than a single pile

due to a combination of three key factors: a) redundancy (indeterminacy, extra load

carrying paths); b) ductility (ability to tolerate and absorb large inelastic deformations

and shift excess loadings to other piles); and c) excess capacity (piles all driven to the

same penetration as the maximum loaded pile, thus having the ability to carry overloads

and displacements). Without the recognition of robustness of a pile group system, the

economical design of driven piles may not be realized.

In AASHTO code, a target reliability β of 2.3 was used for the calibrations of

resistance factors of deep foundation elements, taking into account that piles are almost

always driven in groups, resulting in additional redundancy in the foundation. However,

the pile group effect on the calibrated resistance factors is dependent on the design

methods and types of structures. A single target reliability β of 2.3 set for all structures

195
and prediction methods would be too generic. Therefore, there is a need to exam more

thoroughly the resistance factors or various prediction methods of driven piles in a pile

group.

196
 REFERENCES

1. AASHTO, (1997). Standard Specifications for Highway Bridges: 16th Edition

(1996 with 1997 interims). AASHTO, Washington, DC.

2. AASHTO, (2003). LRFD Bridge Design Specifications. Second Edition, 1998

with 2003 interim. AASHTO, Washington, DC.

3. AASHTO (2006). LRFD Bridge Design Specifications Section 10: Foundations.

Interim, AASHTO, Washington, DC.

4. Allen, T. M., Nowak, A. S., and Bathurst, R.J. (2005). “Calibration to Determine

Load and Resistance Factors for Geotechnical and Structural Design.” Transportation

Research Circular. No. E-C079.

5. Ang, A. H-S. and Tang, W. H. (1975). Probability Concepts in Engineering

Planning and Design. Vol. 1, Basic Principles. John Wiley & Sons, New York.

6. Ang, A. H-S., and Tang, W. H. (1984). Probability Concepts in Engineering

Planning and Design. Volume II. John Wiley & Sons, New York.

197
7. Astedt, B., Weiner, L., and Holm, G. (1992). “Increase in Bearing Capacity with

Time of Friction Piles in Sand.” Proceedings of Nordic Geotechnical Meeting, pp. 411-

416.

8. Attwooll, W. J., Holloway, D. M., Rollins, K. M., Esrig, M. I., Sakhai, S., and

Hemenway, D. (1999). “Measured Pile Set-up during Load Testing and Production Piling

– I-15 Corridor Reconstruction Project in Salt Lake City, Utah.” Transportation Research

Record 1663, Paper No. 99-1140, pp. 1-7.

9. Axelsson, G. (1998). “Long-Term Set-Up of Driven Piles in Non-Cohesive Soils

Evaluated from Dynamic Tests on Penetration Rods.” Proceedings of the First

International Conference on Site Characterization, Vol. 2, pp. 895-900.

10. Axelsson G. (2002). “A Conceptual of Pile Set-up for Driven Piles in Non-

cohesive Soils.” Conference Proceedings, Deep Foundations 2002, ASCE GSP No. 116,

pp. 64-79.

11. Ayyub, B. M., and McCuen, R. H. (2003). Probability, Statistics, and Reliability

for Engineers and Scientists. Second Edition, Published by Chapman & Hall, ISBN 1-

58488-286-7.

198
12. Barker, R. M., Duncan, J. M., Jojiani, K. B., Ooi, P. S. K., Tan, C. K., and Kim, S.

G. (1991). “Manuals for the Design of Bridge Foundations.” NCHRP Report 343,

Transportation Research Board, National Research Council, Washington, DC.

13. Bjerrum, L., Hansen, and Sevaldson (1958). “Geotechnical Investigations for A

Quay Structure in Ilorton.” Norwegian Geotech. Publ. No. 28, Oslo.

14. Bogard, J. D., and Matlock, H. (1990). “Application of Model Pile Tests to Axial

Pile Design.” Proceedings, 22nd Annual Offshore Technology Conference, Houston,

Texas, Vol. 3, pp. 271-278.

15. Bullock, P. J. (1999). Pile Friction Freeze: A Field and Laboratory Study, Volume

1, Ph.D. Dissertation, University of Florida, USA.

16. Bullock, P. J., Schmertmann, J. H., McVay, M. C., and Townsend, F. C. (2005a).

“Side Shear Setup. I: Test Piles Driven in Florida.” Journal of Geotechnical and

Geoenvironmental Engineering, ASCE, Vol. 131, No. 3, pp. 292-300.

17. Bullock, P. J., Schmertmann, J. H., McVay, M. C., and Townsend, F. C. (2005b).

“Side Shear Setup. II: Results from Florida Test Piles.” Journal of Geotechnical and

Geoenvironmental Engineering, ASCE, Vol. 131, No. 3, pp. 301-310.

199
18. Camp, W. M., and Parmar, H.S. (1999). “Characterization of Pile Capacity with

Time in the Cooper Marl – Study of Application of A Past Approach to Predict Long-

Tern Pile Capacity.” Journal of the Transportation Research Board, No. 1663,

Washington, DC, pp. 16-24.

19. Chow, F. C. (1997). Investigations into Displacement Pile Behaviour for Offshore

Foundations. Ph.D Thesis, University of London (Imperial College).

20. Chow, F. C., Jardine, R. J., Brucy, F., and Nauroy, J. F. (1998). “Effects of Time

on Capacity of Pipe Piles in Dense Marine Sand.” Journal of Geotechnical and

Geoenvironmental Engineering, Vol. 124, No. 3, ASCE, pp. 254-264.

21. Christian, J. T. (2004), “Geotechnical Engineering Reliability: How Well Do We

Know What We Are Doing?” Journal of Geotechnical and Geoenvironmental

Engineering, ASCE, Vol. 130, No. 10, pp. 985-1003.

22. Davisson, M. T. (1972). “High Capacity Piles.” Proceedings of Lecture Series on

Innovations in Foundation Construction, ASCE, Illinois Section, Chicago, pp. 81-112.

23. DeBeer, E. E. (1968). “Proefondervindlijke bijdrage tot de studie van het

grensdraag vermogen van zand onder funderingen op staal.” Tijdshift der Openbar

Verken van Belgie, No. 6, 1967 and No. 4, 5, and 6, 1968.

200
24. Decourt, L. (1999) “Behavior of Foundations under Working Load Conditions.”

Proceedings of the 11th Pan-American Conference on Soil Mechanics and Geotechnical

Engineering, Foz DoIguassu, Brazil, Vol. 4, pp. 453-488.

25. Esrig, M. I., and Kirby, R. C. (1979). “Soil Capacity for Supporting Deep

Foundation Members in Clay.” ASTM STP. No. 670, pp. 27-63.

26. Evangelista, A., Pellegrino, A., and Viggiani, C. (1977). “Variability among Piles

of the Same Foundation.” Proc. 9th ICSMFE, Tokyo, pp. 493-500.

27. Finno, R. J., Achille, J., Chen, H. Cosmao, T., Park, J. B., Picard, J., Smith, D. L.,

and Williams, G. P. (1989). “Summary of Pile Capacity Predictions and Comparison with

Observed Behavior.” Predicted and Observed Axial Behavior of Piles, Geotechnical

Special Publication, No. 23, ASCE, pp. 356-385.

28. Gates, M. (1957). “Empirical Formula for Predicting Pile Bearing Capacity.”

Civil Engineering, Vol. 27, No. 3, pp. 65-66.

29. Goble Rausche Likins and Associates, Inc. (1996). CAPWAP Introduction to

Dynamic Pile Testing Methods, Cleveland, OH, USA.

30. Hannigan, P.J., Goble, G. G., Thendean, G., Likins, G. E., and Rausche, F. (1998).

Design and Construction of Driven Pile Foundations, Workshop Manual. Vol. 2,

Publication No. FHWA-HI-97-014, Federal Highway Administration, Washington, DC.

201
31. Hansen, J. B. (1963). “Discussion on Hyperbolic Stress-Strain Response.

Cohesive Soils.” Journal for Soil Mechanics and Foundation Engineering, ASCE, Vol. 89,

SM4, pp. 241-242.

32. Hirsch, T. J., Carr, L., and Lowery, L. L. (1976). Pile Driving Analysis—Wave

Equation Users Manual, Vol. 1: Background, Texas Transportation Institute, College

Station, Texas A&M, USA.

33. Holloway, D. M., and Beddard, D. L. (1995). ‘Dynamic Testing Results, Indicator

Pile Test Program – I-880, Oakland, California.” Deep Foundations Institute 20th Annual

Members Conference and meeting, pp. 105-126.

34. Huang, S. (1988). “Application of Dynamic Measurement on Long H-Pile Driven

into Soft Ground in Shanghai.” In Proceedings of 3rd International Conference on the

Application of Stress-Wave Theory to Piles, Bi-Tech Publishers, Ottawa, Ontario,

Canada, pp. 635-643.

35. Komurka, V. E., Wanger, A. B., and Tuncer, B. E. (2003). “Estimating Soil/Pile

Set-up.” Final Report, Wisconsin Highway Research Program #0092-0014, Prepared for

The Wisconsin Department of Transportation.

36. Komurka, V. E., Winter, C. J., and Maxwell, S. (2005). “Applying Separate

Safety Factors to End-of-Drive and Set-up Components of Driven Pile Capacity.”

202
Proceedings of the 13th Great Lakes Geotechnical and Geoenvironmental Conference,

Milwaukee, Wisconsin, USA, pp. 65-80.

37. Kulhawy, F. H., and Phoon, K. K. (1996). “Engineering Judgment in the

Evolution from Deterministic to Reliability-Based Foundation Design.” Uncertainty in

the Geologic Environment; from Theory to Practice; Proceedings of Uncertainty ’96, C.

D. Shackelford, P. P. Nelson, and M. J. S. Roth eds., Madison, WI, July 31-Augest 3,

ASCE, New York, pp. 29-48.

38. Lacasse, S., Guttormsen, T. R., and Goulois, A. (1989). “Bayesian Updating of

Axial Capacity of Single Pile.” Proc., 5th International Conference on Structural Safety

and Reliability, ASCE, New York, pp. 287-290.

39. Lai, P., and Kuo, C. L. (1994). “Validity of Predicting Pile Capacity by Pile

Driving Analyzer.” Proceedings of International Conference on Design and Construction

of Deep Foundations, Florida, USA, Dec., 1095-1102.

40. Liang, R. Y. (1999). “Development and Implementation of New Driven Pile

Technology.” Report FHWA/OH-99/008, Ohio Department of Transportation.

41. Liang, R. (2003). “New Wave Equation Technique for High Strain Impact Testing

of Driven Piles,” Geotechnical Testing Journal, ASTM, Vol. 26, No. 1, pp. 1-7.

203
42. Liang, R. and Husein, A. I. (1993). “Simplified Dynamic Method for Pile-Driving

Control.” Journal of Geotechnical Engineering, ASCE, Vol. 119, No. 4, pp. 694-713.

43. Liang, R. and Zhou, J. (1996). “Pile Capacity Estimation Using New HST

Interpretation Method,” In Proceedings of 5rd International Conference on the

Application of Stress-Wave Theory to Piles, Orlando, Florida, USA, pp. 367-381.

44. Liang, R. and Zhou, J. (1997). “Probability Method Applied to Dynamic Pile-

Driving Control.” Journal of Geotechnical and Geoenvironmental Engineering, ASCE,

Vol. 123, No. 2, pp. 137-144.

45. Long, J. H., Kerrigan, J. A., and Wysockey, M. H. (1999). “Measured Time

Effects for Axial Capacity of Driven Piles.” Journal of the Transportation Research

Board, No. 1663, Washington, DC, pp. 8-15.

46. McVay, M. C., Alvarez, V., Zhang, L., Perez, A., and Gibsen, A. (2002).

“Estimating Driven Pile Capacities during Construction.” Research Report WPI 0510852,

Florida Department of Transportation.

47. McVay, M. C., Biergisson, B., Zhang, L. M., Perez, A., and Putcha, S. (2000).

“Load and Resistance Factor Design (LRFD) for Driven Piles Using Dynamic Methods-

A Florida Perspective.” Geotechnical Testing Journal, ASTM, 23 (1), pp. 55-66.

204
48. McVay, M. C., Schmertmann, J., Townsend, F., and Bullock, P. J. (1999). “Pile

Friction Freeze: A Field and Laboratory Study.” Research Report WPI 0510632, Florida

Department of Transportation.

49. Meyerhof, G. (1976). “Bearing Capacity and Settlement of Pile Foundations.” Journal

of the Geotechnical Engineering Division, ASCE, Vol. 102, No. 3, pp.195-228.

50. Nordlund, R. L. (1963). “Bearing Capacity of Piles in Cohesionless Soils.” Journal of

Soil Mechanics and Foundation Engineering, JSMFE, Vol. 89, SM 3, pp. 1-36.

51. Nottingham, L., and Schmertmann, J. (1975). “An Investigation of Pile Capacity

Design Procedures.” Final Report D629 to Florida Department of Transportation,

Department of Civil Engineering, University of Florida.

52. Nowak, A. (1999). “Calibration of LRFD Bridge Design Code.” National

Cooperative Highway Research Program Report 368, Transportation Research Board,

National Research Council, Washington, D.C.

53. Paikowsky, S. G., Birgisson, B., McVay, M., Nguyen, T., Kuo, C., Beacher, G.,

Ayyub, B., Stenersen, K., O’Malley, K., Chernauskas, L., and O’Neill, M. (2004). “Load

and Resistance Factor Design (LRFD) for Deep Foundations.” NCHRP Report 507.

Transportation Research Board, Washington, D. C.

205
54. Paikowsky, S. G., Regan, J. E., and MaDonnell, J. J. (1994). “A Simplified Field

Method for Capacity Evaluation of Driven Piles.” Publication FHWA-RD-94-042, U.S.

Department of Transportation and Federal Highway Administration.

55. Paikowsky, S. G., and Stenerson K. (2000). “The Performance of the Dynamic

Methods, Their Controlling Parameters and Deep Foundation Specifications.” Keynote

Lecture in the Proceeding of the Sixth International Conference on the Application of

Stress-Wave Theory to Piles, Niyama S. and Beim J. edt., September 11-13, Sao Paulo,

Brazil, Balkema, pp.281-304.

56. Parsons, J. D. (1966). “Piling Difficulties in the New York Area.” Journal of the

Soil Mechanics and Foundations Division, ASCE, Vol. 92, No. 1, pp. 43-64.

57. Pestana, J. M., Hunt, C. E., and Bray, J. D. (2002). “Soil Deformation and Excess

Pore Pressure Field around A Closed-Ended Pile.” Journal of Geotechnical and

Geoenvironmental Engineering, ASCE, Vol. 128, No. 1, pp. 1-12.

58. Phoon, K. K., and Kulhawy, F. H. (1996). “On Qualifying Inherent Soil

Variability.” Uncertainty in the Geologic Environment; from Theory to Practice;

Proceedings of Uncertainty ’96, C. D. Shackelford, P. P. Nelson, and M. J. S. Roth eds.,

Madison, WI, July 31-Augest 3, ASCE, New York, pp. 326-340.

206
59. Phoon, K. K., and Kulhawy, F. H. (1999a). “Characterization of Geotechnical

Variability.” Canadian Geotechnical Journal, 36(4), 612-624.

60. Phoon, K. K., and Kulhawy, F. H. (1999b). “Evaluation of Geotechnical Property

Variability.” Canadian Geotechnical Journal, 36(4), 625-639.

61. Phoon K. K., and Kulhawy, F. H. (2005). “Characterization of Model

Uncertainties for Laterally Loaded Rigid Drilled Shafts”, Geotechnique, 55(1), 45-54.

62. Phoon, K. K., Kulhawy, F. H., and Grigoriu, M. D. (2003a). “Development of A

Reliability-Based Design Framework for Transmission Line Structure Foundations.”

Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 129, No. 9, pp.

798 - 806.

63. Phoon, K. K., Kulhawy, F. H., and Grigoriu, M. D. (2003b). “Multiple Resistance

Factor Design (MRFD) for Shallow Transmission Line Structure Foundations.” Journal

of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 129, No. 9, pp. 807 -

818.

64. Preim, M. J., March, R., and Hussein, M. (1989). “Bearing Capacity of Piles in

Soils with Time Dependent Characteristics.” Piling and Deep Foundations, Vol. 1, pp.

363-370.

207
65. Rausche F., Goble, G., and Likins, G. (1985). “Dynamic Determination of Pile

Capacity.” Journal of Geotechnical Engineering, ASCE, Vol. 113, No. 3, pp. 367-383.

66. Rausche, F., Moses, F., and Goble, G. G., (1972) “Soil Resistance Predictions

from Pile Dynamics,” Journal of the Soil Mechanics and Foundations Division, ASCE,

Vol. 98, No. SM9, pp. 917-937.

67. Samson, L. and Authier, J. (1986). “Change in Pile Capacity with Time: Case

Histories.” Canadian Geotechnical Journal, 23(1), pp. 174-180.

68. Seidel, J. P., Haustorfer, I. J., and Plesiotis, S. (1988). “Comparison of Static and

Dynamic Testing for Piles Founded into Limestone.” Proceedings of 3rd International

Conference on the Application of Stress-Wave Theory to Piles, Bi-Tech Publishers,

Ottawa, Ontario, Canada, pp. 717-723.

69. Skov, R., and Denver, H. (1988). “Time Dependence of Bearing Capacity of

Piles.” In Proceedings of 3rd International Conference on the Application of Stress-Wave

Theory to Piles, Bi-Tech Publishers, Ottawa, Ontario, Canada, pp. 879-888.

70. Smith, E. A. L. (1960). “Pile Driving Analysis by the Wave Equation,” Journal of

Soil Mechanics and Foundation Division, ASCE, Vol. 86, No. SM4, pp. 35-61.

208
71. Soderberg, L. O. (1961), “Consolidation Theory Applied to Foundation Pile Time

Effects.” Geotechnique, London, Vol. 11, No. 3, pp. 217-225.

72. Svinkin, M. R. (2002). “Engineering Judgment in Determination of Pile Capacity

by Dynamic Methods.” Deep Foundations Congress, Geotechnical Special Publication,

No. 116, Vol. 2, ASCE, pp. 898-914.

73. Svinkin, M.R., Morgano, C. M., and Morvant, M. (1994). “Pile Capacity as A

Function of Time in Clayey and Sandy Soils.” In Proceedings of 5th International

Conference and Exhibition on Piling and Deep Foundations, Belgium, pp. 1.11.1-1.11.8.

74. Svinkin, M. R., and Skov, R. (2000). “Set-up Effect of Cohesive Soils in Pile

Capacity.” Proceedings of 6th International Conference On Application of Stress Waves

to Piles, Sao Paulo, Brazil, pp. 107-111.

75. Tan, S. L., Cuthbertson, J., and Kimmerling, R. E. (2004). “Prediction of Pile Set-

up in Non-cohesive Soils.” Current Practices and Future Trends in Deep Foundations,

Geotechnical Special Publication No. 125, pp. 50-65.

76. Tavenas, F. and Audy, R. 1972. “Limitation of the Driving Formulas for

Predicting the Bearing Capacities of Piles in Sand.” Canadian Geotechnical Journal, 9(1),

pp. 47-62.

209
77. Titi, H. H. and Wathugala, G. W. (1999). “Numerical Procedure for Predicting

Pile Capacity – Setup/Freeze.” Transportation Research Record 1663, Paper No. 99-0942,

pp. 25-32.

78. Thomson, W. T., (1950). Laplace Transformation, Prentice Hall, Inc., New York,

NY, USA.

79. Tomlinson, M. J. (1986). Foundation Design and Construction. Longman

Scientific and Technical, Essex, England.

80. Vijayvergiya, V. N., and Focht Jr., J. A. (1972). “A New Way to Predict Capacity

of Piles in Clay.” Proceedings of the 4th Annual Offshore Technology Conference, Texas,

Vol. 2, pp. 856-874.

81. Vrouwenvelder, A. (1992). “Effects of inspection on the reliability of foundation

piles.” 4th International Conference on Application of Stress-Wave Theory to Piles,

Balkema, Rotterdam, pp.657-663.

82. Walton, P. A., and Borg, S. L. (1998). “Dynamic Pile Testing to Evaluate Quality

and Verify Capacity of Driven Piles.” Transportation Research Board, pp. 1-7.

210
83. Wang, S., and Reese, L. C. (1989). “Predictions of Response of Piles to Axial

Loading.” Predicted and Observed Axial Behavior of Piles, Geotechnical Special

Publication No. 23, ASCE, pp. 173-187.

84. Wellington, A. (1892). “Discussion of ‘The Iron Wharf at Fort Monroe, Va. By J.

B. Cuncklee.” Transactions, ASCE Vol. 27, Paper No. 543, Aug. 1982, pp. 129-137.

85. Withiam, J. L., Voytko, E. P., Barker, R. M., Duncan, J. M., Kelly, B. C., Musser,

S. C., and Elias V. (2001). “Load and Resistance Factor Design (LRFD) for Highway

Bridge Substructures.” FHWA HI-98-032, National Highway Institute.

86. Yang, N. (1956). “Redriving Characteristics of Piles.” Journal of the Soil

Mechanics and Foundations Division, Vol. 82, Paper 1026, SM 3, July, ASCE.

87. Yang, N. (1970). “Relaxation of Piles in Sand and Inorganic Silt.” Journal of the

Soil Mechanics and Foundations Division, ASCE, Vol. 96, SM2, pp. 395-409.

88. York, D. L., Brusey, W. G., Clemente, F., M., and Law, S. K. (1994). “Set-up and

Relaxation in Glacial Sand.” Journal of Geotechnical Engineering, ASCE, Vol. 120, No.

9, pp. 1498-1513.

89. Zai, J. (1988). “Pile Dynamic Testing Experience in Shanghai.” Proceedings of

3rd International Conference on the Application of Stress-Wave Theory to Piles, Bi-Tech

Publishers, Ottawa, Ontario, Canada, pp. 781-792.

211
90. Zhang, L. M. (2004). “Reliability Verification Using Proof Pile Load Tests.”

Journal of Geotechnical and Geoenvironmental Engineering, Vol. 130, No. 11, pp. 1203-

1213.

91. Zhang, L. M. and Tang, W. H. (2002). “Use of Load Tests for Reducing Pile

Length.” Geotechnical Special Publication No. 116, Deep Foundation 2002 – An

International Perspective on Theory, Design, Construction, and Performance, Vol. 2, pp.

993-1005.

92. Zhu, G. (1988). “Wave Equation Applications for Piles in Soft Ground.”

Proceedings of 3rd International Conference on the Application of Stress-Wave Theory to

Piles, Ottawa, Ontario, Canada, pp. 831-836.

212