Seismic risk analysis using Bayesian belief networks - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 172

importance/consequence of failure. The seismic risk assessment requires
methods that combine human knowledge and experience as well as expert
judgment. The challenge of seismic risk assessment and decision making is
further compounded due to ubiquitous uncertainty (Wen et al. 2003). The
level of uncertainty associated with any system is proportional to its com-
plexity, which arises as a result of vaguely known relationships among
various entities, and randomness in the mechanisms governing the system
(Ross 2004).

7.1.2 Soft computing

The confi dence level on seismic risk assessment can be enhanced by con-
sidering soft computing techniques that account for different sources of
uncertainty (see Chapter 6). Soft computing is a conglomerate of computing
techniques that include fuzzy-based methods, neuro-computing, probabilis-
tic reasoning, genetic algorithms, chaotic systems, belief networks, and
learning theory (Zadeh 1994). Different classical (probabilistic) and non-
classical methods (including possibility theory, fuzzy sets, fuzzy measures,
random sets) are used to represent different types of uncertainties. When
extensive historical data exist, model-free methods, such as artifi cial neural
networks (ANN), can provide insights into cause-effect relationships and
uncertainties through learning from data. In cases where historical data are
scarce and/or available information is ambiguous and imprecise, other soft
computing techniques, such as fuzzy sets and Bayesian belief network
(BBN), can provide an appropriate framework to handle such relationships
and uncertainties. BBN has a utility where the physics-based models are
not readily available and intuitive knowledge of the expert is used to
develop the causal network (e.g. Li et al. 2010).
In this chapter, applications of BBN to site-specifi c seismic hazard (liq-
uefaction), regional damage estimation, and individual building damage
estimation are presented. There are increasing applications of the BBN in
regional risk assessment (Cockburn and Tesfamariam 2012), seismic hazard
assessment (Bensi et al. 2009; Bayraktarli et al. 2011), risk assessment for
bridges (Bensi et al. 2009) and buildings (Faizian et al. 2004; Tesfamariam
et al. 2010; Bayraktarli and Faber 2011), loss assessment (Li et al. 2010;
Schubert and Faber 2011); and optimization and reliability (Nishijima et al.
2009; Straub and Der Kiureghian 2010a,b). Weber et al. (2010) provided a
thorough review of the application of BBN in risk analysis.

7.2 Bayesian belief networks (BBN)

BBN, also known as Bayesian net, causal probabilistic network, Bayesian
network or simply belief network, is a graphical model that permits a proba-