Bridge Engineering

Bridge Engineering
Design Principles

Nirjhar Dhang

√
Version : 1.414 : 2
Created on : January 01, 2017
Last revision : July 24, 2024
Module A : Introduction, Planning, Design Process and
Design Principles
Module A-5.1 : Design Principles
Overview

1 Structural Safety

2 Partial safety factors

3 Functions of Random Variables

4 Reliability analysis

5 Example : calibration of partial safety factors

6 Summary
Structural Safety
Structural Safety

Structural safety appears in different forms when criteria for

design are stated in codes and specifications.
The most commonly encountered criteria for design are :
Allowable stress design(ASD)
Ultimate strength design(USD)
Load and resistance factor design(LRFD)
Structural Safety(contd...)

Allowable stress design(ASD)

Structural members are designed such that the computed
elastic stress is less than or equal to a fraction of a limiting
value such as the yield stress of the material
The computed elastic stress is obtained from an analysis of
the structure using specified design loads.
Structural Safety(contd...)

Ultimate strength design(USD)

Design loads are increased by a factor
The computed strength of a structural member is reduced
by a factor
The computed strength of a structural member is based on
ultimate behaviour
An elastic analysis is used to obtain the member forces

Vu ≤ ϕVn (1)

where ϕ is a strength reduction factor, Vn is the nominal

shear strength of the beam and Vu is the factored shear
force at the section under consideration
Structural Safety(contd...)

Load and resistance factor design(LRFD)

This method has been introduced by the American Institute
of Steel Construction(AISC).
It is characterized by an expression of the form
X
γi Qi ≤ ϕRn (2)

where γi is the load factor for the ith load effect, Qi is the
ith load effect, ϕ is a resistance factor, and Rn is the
nominal resistance(strength).
A similar method is used in the AASHTO specification
Standard Specifications for Highway Bridges
Partial safety factors
Partial safety factors

µQ

Figure 1: Relationships among nominal load, mean load and factored load
Partial safety factors

µR

Figure 2: Relationships among nominal resistance, mean resistance and

factored resistance
Partial safety factors

µQ µR

Figure 3: Relationships among nominal, mean and factored load and

resistance
Partial safety factors(contd...)

In load and resistance factor design (LRFD) or limit state design

(LSD), load components are multiplied by load factors and
resistance is multiplied by a resistance factor.
The basic form of the LRFD equation is
X
ϕR ≥ γi Qi (3)
where γi is a load factor applied to load components (or load
effect) Qi and ϕ is a resistance factor applied to the
resistance(measure of load carrying capacity) R.
Partial safety factors(contd...)

The basic form of the LRFD equation is

X
ϕR ≥ γi Qi (4)
In words, the equation says that the capacity of the structural
member (modified by the factor ϕ) must be larger than the total
effect of all the loads acting on the member.
Partial safety factors(contd...)

Forward problem: To determine β given a limit state function

(or performance function) and the mean and variance (or
standard deviation or coefficient of variation) for each of the
random variables used in the limit state function.
Inverse problem: A target β is specified and it is then necessary
to determine the required mean values of the resistance and
loads to achieve the target.
This means that we need to find the design point {z ∗ }
corresponding to the target β.
Partial safety factors(contd...)

Assume that the design points {z ∗ } (in reduced coordinates)

corresponding to a target β is known. To get the corresponding
design point {x ∗ } in original coordinates, the following relation
is used for each variable.
xi∗ = µxi + zi∗ σxi (5)
Since the design point must be on the failure boundary, the limit
state function must satisfy
g(x1∗ , x2∗ , ..., xi∗ ) = 0 (6)
For design purposes, it is necessary to relate each design point
value xi∗ to a value of the variable used in design (e.g. a nominal
design value specified by the code).
Partial safety factors(contd...)

If the nominal design value of Xi is denoted by xei , then the

partial safety factor γi is defined as
x∗
γi = i (7)
xei
The partial safety factor is nothing more than a scaling factor
that allows the designer to convert a nominal design value of a
variable to the value needed to satisfy the limit state function for
a target β.
g(x1∗ , x2∗ , ..., xi∗ ) = 0 (8)
Functions of Random Variables
Functions of Random Variables
Let Y be a linear function of random variables X1 ,X2 ,. . .,Xn :
n
X
Y = a0 + a1 X1 + a2 X2 + ... + an Xn = a0 + ai Xi (9)
i=1
The mean value of Y is :
n
X
µY = a0 + a1 µX1 + a2 µX2 + ... + an µXn = a0 + ai µXi (10)
i=1
The variance of Y is :
h i
σY2 = E (Y − µY )2 (11)
= E [Y 2 ] − µ2Y (12)
n X
X n
= ai aj CoV (Xi , Xj ) (13)
i=1 i=1
Xn X n
= ai aj ρxi xj σxi σxj (14)
i=1 i=1
Functions of Random Variables
n
X
Y = a0 + a1 X1 + a2 X2 + ... + an Xn = a0 + ai Xi (15)
i=1

The variance of Y is :
n X
n
σY2 =
X
ai aj CoV (Xi , Xj ) (16)
i=1 i=1
Xn X n
= ai aj ρxi xj σxi σxj (17)
i=1 i=1

Note that the constant a0 does not affect the variance

If the n random variables are uncorrelated with each other, then
CoV (Xi , Xj ) = 0 for i ̸= j
n
σY2 = ai2 σx2i
X
(18)
i=1
Functions of Random Variables

n
X
Y = a0 + a1 X1 + a2 X2 + ... + an Xn = a0 + ai Xi (19)
i=1

The mean value of Y is :

n
X
µY = a0 + a1 µX1 + a2 µX2 + ... + an µXn = a0 + ai µXi (20)
i=1

The variance of Y for uncorrleated random variables is :

n
σY2 = ai2 σx2i
X
(21)
i=1
Probability of failure

Performance function
g(R, Q) = R − Q (22)
The structure fails when the load exceeds the resistance.
Reliability Index
Performance function
g(R, Q) = R − Q (23)
For the basic variables R and Q, the standard forms can be
expressed as
R − µR
ZR = (24)
σR
Q − µQ
ZQ = (25)
σQ
The variables ZR and ZQ are sometimes called random variables
The resistance R and the load Q can be expressed in terms of
the reduced variables as follows:
R = µ R + ZR σ R (26)
Q = µQ + ZQ σQ (27)
Reliability Index

R = µ R + ZR σ R (28)
Q = µQ + ZQ σQ (29)

The limit state function can be expressed in terms of the

reduced variables

g(ZR , ZQ ) = µR + ZR σR − µQ − ZQ σR (30)
= (µR − µQ ) + ZR σR − ZQ σQ (31)
Reliability Index

µR − µQ
β=q (32)
σR2 + σQ2

β = −Φ−1 (Pf ) or Pf = Φ(−β) (33)

First-Order Second-Moment Reliability Index

The linear limit state function

g(X1 , X2 , . . . , Xn ) = a0 + a1 µX1 + a2 µX2 + ... + an µXn (34)

n
X
= a0 + ai µXi (35)
i=1
Pn
a0 + ai µXi
β= qP i=1 (36)
n
a2 σ 2
i=1 i xi

If variance, V is given, σ = V µ
Reliability analysis
Reliability analysis
Calculate the reliability index for the three span continuous
beam shown in Figure 1.

L L L

Figure 4: Three span beam

Reliability analysis(contd...)

The Random variables in the problem are distributed load (w),

span length (L), modulus of elasticity (E), moment of inertia (I).
The limit state to be considered is deflection, and the allowable
deflection is specified as L/360.
The maximum deflection is 0.0069wL4 /EI, and it occurs at
0.446L from either end.
The limit state function is
L wL4
g(w , L, E , I) = − 0.0069 (37)
360 EI
Reliability analysis(contd...)

The means and standard deviations of the random variables are

listed in Table:1
Table 1 : Parameters of random variables
Variable Mean Standard deviation
w 10kN/m 0.4kN/m
L 5m ~0
E 2 × 107 kN/m2 0.5 × 107 kN/m2
I 8 × 10−4 m4 1.5 × 10−4 m4
Reliability analysis(contd...)

The limit state function is

L wL4
g(w , L, E , I) = − 0.0069 (38)
360 EI
Put L=5
5 (54 )w
g =0 ⇒ − 0.0069 = 0 ⇒ EI − 310.5w = 0 (39)
360 EI
Define the reduced variables:
I − µI
Z1 =
σI
E − µE
Z2 = (40)
σE
w − µw
Z3 =
σw
Reliability analysis(contd...)
Define the reduced variables:
I − µI
Z1 =
σI
E − µE
Z2 = (41)
σE
w − µw
Z3 =
σw
I = µ I + Z1 σ 1
E = µE + Z2 σE (42)
w = µw + Z3 σw

5 (54 )w
g =0 ⇒ − 0.0069 = 0 ⇒ EI − 310.5w = 0 (43)
360 EI
Substituting into g:
(µE + Z2 σE )(µI + Z1 σI ) − 310.5(µw + Z3 σw ) = 0 (44)
Reliability analysis(contd...)

5 (54 )w
g =0 ⇒ − 0.0069 = 0 ⇒ EI − 310.5w = 0 (45)
360 EI
Substituting into g:
(µE + Z2 σE )(µI + Z1 σI ) − 310.5(µw + Z3 σw ) = 0 (46)

[2×107 +Z2 (0.5×107 )][8×10−4 +Z1 (1.5×10−4 )]−310.5[10+Z3 (0.4)] = 0

(47)

(3000)Z1 + (4000)Z2 + (750)Z1 Z2 − (124.2)Z3 + 12.895 = 0 (48)

Reliability analysis(contd...)
Formulate g in terms of β and αi :
zi∗ = 0 (49)

(3000)βα1 +(4000)βα2 +(750)β 2 α1 α2 −124.2βα3 +12.895 = 0 (50)

−12.895
β= (51)
3000α1 + 4000α2 + 750βα1 α2 − 124.2βα3
The iterations start with a guess for β, α1 , α2 , α3 .
α1 = α2 = −0.58; α3 = 0.58 and let β = 3 (52)

Finally , α1 = −0.182 α2 = − − 0.983; α3 = 0.034 and β = 3.173

(53)
Reliability analysis(contd...)

−12.895
β= (54)
3000α1 + 4000α2 + 750βα1 α2 − 124.2βα3
The iterations start with a guess for β, α1 , α2 , α3 .
α1 = α2 = −0.58; α3 = 0.58 and let β = 3 (55)

Finally , α1 = −0.182 α2 = −0.983; α3 = 0.034 and β = 3.173

(56)
Reliability index, β=3.173
Example : calibration of partial safety factors
Example : calibration of partial safety factors

Consider the fundamental case

g =R −Q (57)
where R is the resistance and Q is the load.
A possible design equation (in LRFD format) is
γR µR ≥ γQ µQ (58)
The nominal values are assumed to be equal to the mean values
Assume, VR as 10 percent and VQ as 12 percent
To determine the partial safety factors that must be used in
design to achieve the reliability index, βtarget =3.0
Assume that both R and Q are normally distributed and
uncorrelated
Example : calibration of partial safety
factors(contd...)

The limit state function :

g =R −Q (59)
where R is the resistance and Q is the load.
Obtain and initial design point, say r ∗ = q ∗ .
There are two variables, so we have two unknown mean values
Use limit state function at the mean values to relate the two
unknown mean values. This results, µR = µQ
Example : calibration of partial safety
factors(contd...)

Determine the {G} vector:

∂g
G1 = − σR = −σR = −VR µR = −0.1µR = −0.1µQ
∂R (60)
∂g
G2 = − σQ = σQ = VQ µQ = 0.12µQ
∂Q
Note that µR = µQ is used to evaluate G1
Example : calibration of partial safety
factors(contd...)

Calculate α. The correlation matrix [ρ] is the identity matrix. It

means that variables are uncorrelated
[ρ]{G}
{α} = q
{G}T [ρ]{G}
( )
1 −0.1µQ
=q
(−0.1µQ )2 + (0.12µQ )2 0.12µQ
( ) (61)
1 −0.1µQ
=
0.156µQ 0.12µQ
( )
−0.641
=
0.769
Example : calibration of partial safety
factors(contd...)

Determine a new design point for n-1 variables. Calculaute zQ∗ in

this way
zQ∗ = αQ βtarget = 0.769(3.0) = 2.31 (62)
Determine q ∗ using
q ∗ = µQ + zQ∗ σQ = µQ (1 + zQ∗ VQ ) = µQ (1 + 2.31(0.12)) = 1.28µQ
(63)
∗ ∗
Determine r by solving g=0. Thus r = 1.28µQ .
Before iterating again, get an improved estimate of µR in terms
of µQ for use in calculating {G} and {α}.
r∗ 1.28µQ
µR = = = 1.58µQ (64)
(1 + αR βtarget VR ) 1 − 0.641(3.0)(0.10)
Example : calibration of partial safety
factors(contd...)

The results of subsequent iterations are shown in Table-2

Table 2 : Iteration results
Iteration number
1 2 3
∗
r (start) µQ 1.28µQ 1.22µQ
q ∗ (start) µQ 1.28µQ 1.22µQ
µR (start) µQ 1.58µQ 1.60µQ
αR -0.641 -0.796 -0.800
αQ 0.769 0.605 0.600
∗
r (end) 1.28µQ 1.22µQ 1.22µQ
q ∗ (end) 1.28µQ 1.22µQ 1.22µQ
µR (end) 1.58µQ 1.60µQ 1.60µQ
Example : calibration of partial safety
factors(contd...)

Assuming the mean values are the nominal design values, the
design factors are
r∗ 1.22µQ
γR = = = 0.763
µR 1.60µQ
(65)
q∗ 1.22µQ
γQ = = = 1.22
µQ µQ
Assume µQ = 10. From the design equation, γR µR ≥ γQ µQ , the
minimum required µR to achieve β = 3 would be
γQ µQ (1.22)(10)
µR = = = 16.0 (66)
γR 0.763
Example : calibration of partial safety
factors(contd...)
If the partial safety factors are calibrated correctly, a nominal
resistance of 16 will ensure a target β = 3 is acheieved if the
nominal loading is equal to 10.
For this simple case of a linear limit state function, it can be
checked with the following equation
µR − µQ
β=q (67)
σR2 + σQ2
Substituting all the variables determined into this expression and
noting that σR = VR µR
µR − µQ 16.0 − 10 6.0
β=q =q = = 3.0
σR2 + σQ2 [(0.1)(16.0)]2 + [(0.12)(10)]2 2.00
(68)
Summary
Summary

In modern codes, a target reliability index is fixed to find partial

safety factors of materials and loads.
References

Reliability of structures, A. S. Nowak and K. R. Collins,

McGraw-Hill