Chapter 7

Fundamentals of
Reliability Analysis

7.1 INTRODUCTORY COMMENTS

In general, engineering design consists of proportioning the elements of a system so
that it satisfies various criteria of performance, safety, serviceability, and durability
under various demands. For example, a structure should be designed so that its strength
or resistance is greater than the effects of the applied loads. However, there are numer¬
ous sources of uncertainly in the load- and resistance-related parameters. Methods for
identifying and quantifying uncertainties in random variables were discussed in
Chapters 5 and 6. Incorporating the information on uncertainty into actual design prob¬
lems is the subject of this chapter.

7.2 DETERMINISTIC AND

PROBABILISTIC APPROACHES
In the presence of uncertainty, it is not simple to satisfy the basic design requirements.
Figure 7. 1 shows a simple case considering two variables (one relating to the demand
on the system, e.g., load on the structure, S, and the other to the capacity of the system,
e.g., resistance of the structure, R). Both 5 and R are random in nature; their random¬
ness is characterized by their means (is and [iR, standard deviations os and oR, and cor¬
responding probability density functions fs(s) and fR(r), respectively, as shown in
Figure 7. 1 . Figure 7. 1 also identifies the deterministic (nominal) values of these param¬
eters, SN and Rn, used in a conventional safety factor-based approach. The purpose of
this chapter is to develop a rational procedure to incorporate the information in Figure
7.1 into actual designs.

181
182 Chapter 7 Fundamentals of Reliability Analysis

Figure 7.1
Fundamentals of Risk
Evaluation

The concept of risk-based design was introduced by Freudenthal (1956) and was
summarized by Freudenthal, Garrelts, and Shinozuka (1966). The concept has matured
since then and is presented in the following sections.

7.3 RISK AND SAFETY FACTORS CONCEPT

Before the risk-based design format is developed, the concept of risk and the corre¬
sponding safety factor need some elaboration. The application of the risk-based format
in various engineering disciplines can be described, at best, as nonuniform. Most of the
progress has been made in structural engineering. Design guidelines using the load and
resistance factor design (LRFD) concept are essentially based on the risk-based design
format. In the following discussion, the risk-based design format in structural engi¬
neering is emphasized. However, the same concept can also be applied to other engi¬
neering disciplines by replacing resistance and load by supply and demand as discussed
in Section 1.1 of Chapter 1. We hope the following discussion will help to accelerate
the implementation of the risk-based design concept in other engineering disciplines,
where necessary. Referring to Figure 7.1, design safety is ensured in a deterministic
approach by requiring that RN be greater than SN with a specified margin of safety as
Nominal SF = RN / SN (7.1)
where SF is the safety factor.
The nominal resistance (or capacity) RN is usually a conservative value, perhaps
one, two, or three standard deviations below the mean value. The nominal load (or
demand) SN is also a conservative value; however, it is several standard deviations
above the mean value. Thus, the intended conservatism introduced in designs in the
form of the nominal safety factor depends on many other factors, namely the uncer¬
tainty in the load and resistance and how conservatively the nominal load and resist¬
ance values are selected. The nominal safety factor may fail to convey the actual
margin of safety in a design.
Conceptually, then, in a deterministic design the nominal safety factor can be
applied to the resistance, to the load, or to both. The allowable stress design methods
use a safety factor to compute the allowable stresses in members from the ultimate
stress, and a successful design ensures that the stresses caused by the nominal values
 7.4 Risk-Based Design Concept 183

of the loads do not exceed the allowable stresses. In other words, referring to Figure
7.1 and Equation 7.1, RN is divided by a safety factor to compute the allowable resist¬
ance Ra, and safe design requires that the condition SN < Ra be satisfied. In this case,
the safety factor is used for the resistance only. In the ultimate strength design method,
the loads are multiplied by certain load factors to determine the ultimate loads, and the
members are required to resist various design combinations of the ultimate loads. That
is, in Figure 7.1, SN is multiplied by a load factor to obtain the ultimate load Su, and
safe design requires the satisfaction of the condition S„ < RN. In this case, the safety
factors are used in the loads and in load combinations. In some designs, for example,
in concrete design or in steel design using the load and resistance factor design (LRFD)
concept, the capacity reduction factor (generally less than one) and load factors (gen¬
erally more than one) are used to achieve the same objective. Essentially, the safety
factors are used to estimate both the resistance and the loads.
The intent of these conventional approaches can be explained by considering the
area of overlap between the two curves (the shaded region in Figure 7.1), which pro¬
vides a qualitative measure of the probability of failure. This area of overlap depends
on three factors:
1. The relative positions of the two curves: As the distance between the two
curves increases, reducing the overlapped area, the probability of failure
decreases. The positions of the curves may be represented by the means (|iw
and (is) of the two variables.
2. The dispersion of the two curves: If the two curves are narrow, then the area
of overlap and the probability of failure are small. The dispersion may be
characterized by the standard deviations (oR and os) of the two variables.
3. The shapes of the two curves: The shapes are represented by the probability
density functions fR(r) and fs(s).
The objective of safe design in deterministic design procedures can also be achieved,
perhaps more comprehensively, by selecting the design variables in such a way that the
area of overlap between the two curves is as small as possible, so that the underlying risk
is not compromised within the constraints of economy. Conventional design approaches
achieve this objective by shifting the positions of the curves through the use of safety fac¬
tors. A more rational approach would be to compute the risk by accounting for all three
overlap factors, and selecting the design variables so that an acceptable risk of failure is
achieved. This is the foundation of the risk-based design concept. With this approach,
however, the information on the probability density functions of the resistance and loads
(as in Figure 7.1) is usually difficult to obtain, and engineers must formulate an accept¬
able design methodology using only the information on means and standard deviations.

7.4 RISK-BASED DESIGN CONCEPT AND

THE DEVELOPMENT OF THE RISK-BASED
DESIGN FORMAT
Instead of using the safety factor for the resistance alone, as in the working stress
method, or for the loads alone, as in the ultimate strength method, it is more rational to
apply safety factors to both resistance and loads, as is done in concrete or steel (LRFD)
184 Chapter 7 Fundamentals of Reliability Analysis

structural design. The following idealized discussion is presented to develop a parallel

between the deterministic approaches in terms of the capacity reduction factor and load
factors with risk-based design. Referring to Figure 7.1, where the uncertainties in the
load and resistance variables are expressed in the form of the probability density func¬
tions, we can express the measure of risk in terms of the probability of thqfailure event
or P(R < S) as:
pf = /'(failure) = P{R < S)
(7-2)

= Fft(s)fs(s)ds
where FR(s) is the CDF of R evaluated at 5. Equation 7.2 states that when the load is S
= s, the probability of failure is FR(s), and since the load is a random variable, the inte¬
gration needs to be carried out for all the possible values of S, with their respective like¬
lihoods represented by the PDF of S. Equation 7.2 can be considered to be the basic
equation of the risk-based design concept. The CDF of R or the PDF of S may not
always be available in explicit form, and thus the integration of Equation 7.2 may not
be practical. However, Equation 7.2 can be evaluated easily, without performing the
integration, for some special cases. They are considered first in the following sections.

7.4.1 Load and Resistance Normal Variables: Single Load Case

Consider a structure with resistance R subjected to a single load S. The structure is sub¬
jected to one load at a time (i.e, dead load alone, live load alone, wind load alone, seis¬
mic load alone, etc.). If both R and S are normal variables, that is, N(p.R, oR) and N(\is,
Gs), then another random variable Z can be introduced as
Z = R-S. (7.3)
Since it is quite reasonable to assume that R and S are statistically independent, based
on the discussion in Section [Link], we can infer that Z is also a normal random vari¬
able, that is, N(\ir - [ls,-\jo2R +oÿ). Then, Equation 7.2 can be used to define the
probability of failure as
pf=P(Z <0)
or
0-(.u*-H.s)
Pf = 0
+ c2s
or
Pf = 1-0 I 2 2
(7.4)
where 0 is the CDF of the standard normal variate. To develop the explicit expression
for the risk-based design format, Equation 7.4 can be rewritten as
Li* >g5 +0"'(1 (7-5)
where 0~'(/ - pf) is the value of the standard normal variate at the probability level
(7 - pf). Introducing (3 = 0~'(/ - p}), and considering the equality condition, we can
rewrite Equation 7.5 as
 7.5 Fundamental Concept of Reliability Analysis 193

exp 5.05 x 0.72 x 0.77 x -y ln(l + 0. 132 )

Y D= -k- = 1.42
Vl + 0.132
exp[5.05 x 0.72 x 0.77 x /ln(l + 0.372) = 2.56
YL =
V 1 + 0.37"
([) = 0.62 exp(2 x 0. 1 3) = 0.80
yD = 1.42exp(-2x 0.13) = 1.09
yL = 2.56 exp(-2 x 0.37) = 1.22

Thus, the design equation is

0.8/? = 1.09/7 + 1.22L
or
R = 1.36D + 1.53L.
The two examples given here clearly indicate that it is not difficult to estimate the
underlying resistance and load factors in a particular design for a set of assumptions.
Whether these factors are acceptable is a different question. In actual building design
codes, these factors are usually calibrated to satisfy current practice. Furthermore,
assuming that the acceptable risk is going to remain the same for all the load combi¬
nations to be considered for the design of a structure, the engineer can calculate the cor¬
responding resistance and load factors similarly for all load combinations. Maintaining
uniform risk for different loads and load combinations is not practical in deterministic
designs. The ability to design a structure for uniform risk with several loads and load
combinations is one of the many desirable features of risk-based design.
The discussion and the examples clearly indicate that conventional safety
factor-based deterministic designs, in terms of capacity reduction factor and load fac¬
tors, and probability-based load and resistance factor designs are essentially parallel to
each other. However, probabilistic design addresses the necessary design conservatism
more explicitly, perhaps better and more comprehensively, through treatment of the
uncertainty in the random variables, the conservatism used in selecting the design val¬
ues, and the desired underlying reliability. Engineers are empowered to use judgment in
selecting these factors on a case-by-case basis. This concept is the basis of all the relia¬
bility-based design codes being developed in different areas of engineering worldwide.

7.5 FUNDAMENTAL CONCEPT OF

RELIABILITY ANALYSIS
The basic concept of the classical theory of structural reliability and risk-based design
can now be presented more formally. We have seen that it is not difficult to calculate
the underlying resistance and load factors for a given design, assuming an acceptable
level of risk. However, it is more relevant to calculate the underlying risk of a given
design, as is discussed in the following sections.
The first step in evaluating the reliability or probability of failure of a structure is to
decide on specific performance criteria and the relevant load and resistance parameters,