Vulnerability Reliability and Exposure Current Design Paradigms

4 Vulnerability, Reliability,
and Exposure: Current

Design Paradigms
4.1 INTRODUCTION
4.1.1 overvIeW
The design of structures for multihazard (MH) conditions required the development of appropriate
specifications to account for the combined effects of the hazards under consideration. Alipour et al.
(2013) note that according to AASHTO load and resistance factor design (LRFD) bridge design
specifications (AASHTO 2007), the effects of scour on changing the structural characteristics of
a bridge should be evaluated to make sure that the bridge is not vulnerable to extreme loads such
as wind and earthquake. Hence, authors studied the seismic performance of a series of RC bridges
subjected to expected local scour scenarios and made an effort to calibrate the design load factors
following an MH approach.
They defined the probability of failure of a bridge as the probability of exceeding an acceptable
state of damage when a scoured bridge is subjected to earthquake loads. Since the chance of a simul-
taneous occurrence of extreme scour and earthquake events during the design life of a bridge is very
low, they developed a matrix of joint probabilities of two events for a range of possible intensities.
Uncertainties in the estimation of the frequency and intensity of natural hazards were addressed by
generating the required seismic hazard and scour risk curves. For the RC bridges considered in this
study, they developed bridge fragility curves that provided the probability of exceeding a specific
state of damage under any given earthquake and scour demands. Failure probabilities obtained for
various bridge types and scour-earthquake scenarios were compared with the maximum acceptable
probability of failure (or its equivalent target reliability index), to determine scour-load modification
factors to satisfy the code-specified design requirements.
One of the most common and studied MH affinities in civil engineering is the affinity between
earthquake and wind hazards. Duthinh and Simiu (2010) note that according to ASCE standard
7-05, in regions prone to wind and earthquake, structures are designed for each of the hazards
separately, and then final design is based on the more demanding of these two loading conditions.
They show that these standard provisions are not risk-consistent, that is, structures in regions with
significant wind and seismic hazards can have risks of exceedance of limit states that can be up to
twice as high as those in regions where only one of these hazards dominates. In this approach, the
increase in the probability of exceeding a limit state in the presence of two hazards is not taken into
account, that is, this would wrongfully imply that the insurance rates for structures subjected to
two significant hazards should be the same as for their counterparts subjected to one hazard. They
illustrate their point using an example of a water tank situated in South Carolina where both wind
and earthquake-resistant designs are important.
We also note that procedures for design of infrastructure against various hazards using the
LRFD method are from different research and practicing perspectives and are based on data from
single-hazard demands even though structures are subjected to multiple hazards in their service life.
Most structures are designed for the most demanding load combinations by using the appropriate
77
78 Multihazard Considerations in Civil Infrastructure
nominal strength. Thus, risk associated with such designs against various hazards is not generally
uniform. Building upon the work in the literature (Ettouney et al. 2005, Freeman et al. 2005, Hayes
et al. 2005, Sasani and Sagiroglu 2008), Crosti et al. (2011) investigated the risk consistency in MH
design and proposed an integrated approach that achieves consistent risk. They show that, even
though difficult, it is possible to quantify the risks of arriving at a particular lateral drift state for
structures exposed to multiple nonsimultaneous hazards and to compare them to the risks for the
same structures subjected to a single hazard. Using a 10-story, 5-bay steel frame building, they note
that it is possible to take advantage of synergies between designs to resist various hazards. Their
case study shows that the reduced beam section connections, which are intended to enhance ductil-
ity in seismic design, do not reduce the risk of structural damage caused by exposure to wind alone
or exposure to wind or earthquake.
As noted in the earlier studies in the literature, the terms vulnerability and exposure are used in
different contexts with a different meaning. Reliability, on the other hand, due to its objective roots,
is well defined and more uniformly used. This chapter develops processes for MH applications
when vulnerability, reliability (in particular current design practices), and exposure are of interest
to the stakeholders.
4.1.2 thIs chapter

This chapter is devoted to the discussion of vulnerability (Section 4.2), reliability (Sections 4.3
and 4.4), and exposure (Section 4.5). Each section will define the management paradigm from this
book’s viewpoint and then offer two MH methods: a method for applying MPT to the paradigm
and a method for applying MDT for the paradigm. We offer a practical case study for each of these
methods. Figure 4.1 illustrates the composition of this chapter.
4.2 VULNERABILITY OF INFRASTRUCTURES

4.2.1 overvIeW
Vulnerability is a widely used term, but unfortunately, there is no universally accepted definition for
it among infrastructure stakeholders. We utilize the concept of vulnerability in numerous occasions
in this book to indicate the susceptibility of the physical or operational entity under consideration
to degrade when subjected to a given hazard or set of hazards. The physical entity can be a small
component, such as a beam or a structural connection, a system that is a collection of components,
or even an entire asset network. An operational entity can describe an operational component or sys-
tem such as the security of a building, office operations, information flow in an IT system, or traffic
on a bridge. We can, conceptually, consider vulnerability as the inverse of capacity. Even though
the use of capacity has more of an objective tone to it, in many situations, capacity is objectively a
well-defined limit state. Vulnerability can be defined either objectively or subjectively, as pertinent.
We finally note that vulnerability has been used extensively as one of the risk components, as we
will discuss in the next chapter.
This section will discuss two aspects of MH vulnerability. We first develop the MH interaction
matrix of vulnerability, which is a direct application of MPT. We then show how using vulner-
ability as a decision paradigm in the MH environment using MDT to generate the MH decision
interaction matrix.
4.2.2 physIcal Mh vulnerabIlIty InteractIons

We start by investigating simple exposure definitions. Let us assume that vulnerability to the ith
hazard, Hi, is VHi . We now define the vulnerability vector as {V} with VHi as a vulnerability rating
due to the ith hazard. The order of {V} is NH, that is, number of hazards.
Vulnerability, Reliability, and Exposure: Current Design Paradigms 79
Vulnerability, reliability, and exposure
Metrics or measure involved with how and/or

Threats/hazards/demands
why systems degrade
Subjective Objective Objective

For example, wear For example, Subjective For example,
and tear, corrosion, ultimate stresses For example, own weight, live loads,
security, and physical and wear and tear and earthquake loads,
protection ultimate strains security and impact forces
Vulnerability Capacity
Reliability Exposure
Section 4.2 Section 4.3 Section 4.5
Section 4.4
FIGURE 4.1 Overview of this chapter.
In its simplest form, the vulnerability rating can be expressed as a weighted average of several
parameters (US-DHS 2011a,b,c, Fenton and Neil 2013), such that
{V } = éëWTV ùû {VR} (4.1)
The matrix [WT V] includes a set of weights that are related to the system and a set of hazards under
consideration. The order of {VR} is the number of issues/considerations that affect different hazard
vulnerability ratings, nv.
We define MH vulnerability interaction here as the change in the vulnerability to the ith hazard,
VHi , due to a change in the vulnerability to the jth hazard, VH j . Formally, MH vulnerability interac-
tion is defined as ¶VHi /¶VH j .
It is reasonable to state that ¶VH j /¶VH j > 0, that is, this interaction term is always finite, no mat-
ter how small it is.
Recall that
æ k = nv ö
VHi = ç
ç å
è k =1
WTV ik (VR)k ÷
÷
ø
(4.2)
By assuming a change in the kth vulnerability component, the resulting change in vulnerability is
¶VHi
= WTV (4.3)
¶(VR)k ik
The resulting MH vulnerability interaction is
¶VHi ¶VHi ¶ (VR )k

= (4.4)
¶VH j ¶ (VR )k ¶VH j
This equation can be expressed as
¶VHi ¶V /¶(VR)k
= Hi (4.5)
¶VH j ¶VH j /¶(VR)k
We reach the desired MH interaction between vulnerabilities as
¶VHi ( WTV )ik

= (4.6)
¶VH j ( WTV )jk
From Equation 4.6, ¶VHi /¶VHi = 1.0 and ¶VHi /¶VH j = 0.0 if ( WTV )ik = 0.0.
Equation 4.6 is useful when a single parameter affecting vulnerability changes. Similar develop-
ment can be made if more than one parameter changes and if those changes are related. In such situ-
ations, the development of the interaction expressions becomes more involved. The final interaction
terms will be dependent on the actual changes in the parameters such that

¶VHi
=
(
f D ( WTV )ik , D ( WTV ) jk

) (4.7)
¶VH j (
f D ( WTV ) jk , D ( WTV ) jk )

with D ( WTV )ik as a vector containing all actual kth changes, k = 1,2,…,nv. We leave these develop-
ments to the reader as an exercise.
4.2.3 case study 4.1: Mh vulnerabIlIty InteractIons oF Mass transIt statIons

Mass transit stations are important civil infrastructure in any society and may be subjected to a
multitude of hazards, such as flooding, blast, or fire. We are interested in how vulnerability to these
hazards interacts. We first develop the vulnerability weights [WT V] as shown in Table 4.1. Note that
these weights are relative to each other, and thus experts in the field who have a great deal of expe-
rience in infrastructure performance due to different hazards should develop them. There are five
hazards of interest and nine issues that can affect the vulnerability of the particular mass transit sta-
tion under consideration. Details of these issues can be found in US-DHS (2011a,b,c). We can now
apply Equation 4.6 to obtain the MH vulnerability interaction matrices for each of the nine issues
as shown in Tables 4.2 through 4.10. The average MH vulnerability interaction matrix (Table 4.11)
can be obtained by averaging the nine matrices (Tables 4.2 through 4.10).
These matrices reveal how vulnerability of the mass transit station under consideration is due
to interaction between hazards. For example, on average, in Table 4.11, the subjective interaction
between fire and manmade hazards is 1.763, while the interaction between wind and flood is 0.981.
This means that, on average, changes in the vulnerability of the system to manmade hazards due to
changes of the vulnerability to fire hazard are almost 180% (i.e., 1.763/0.981) of the changes of the
vulnerability to flood hazard due to changes of vulnerability due to wind hazard. Due to the subjec-
tive nature of the process, only the relative interaction values can be obtained from these matrices,
thus limiting the usefulness of the process. We utilized nine issues for this example, which do not
TABLE 4.1
Vulnerability Weights Matrix [WTV]
Vulnerability Fire Operational Nonstructural Physical
Issues Site Architecture Structure Ventilation Protection Systems Systems Security IT Issues
Manmade 13 4 8 12 8 12 2 38 4
Earthquakes 6 11 39 4 12 7 9 3 9
Fire 5 10 20 9 33 10 7 2 6
Floods 44 3 4 9 4 18 5 5 8
Wind 50 3 10 4 5 20 5 2 2
Vulnerability, Reliability, and Exposure: Current Design Paradigms
81
TABLE 4.2
Multihazard Interaction Matrix for Site Vulnerability
Changes
Hazard Manmade Earthquakes Fire Floods Wind
Manmade 1.000 2.067 2.864 0.296 0.264
Earthquakes 0.484 1.000 1.386 0.143 0.128
Fire 0.349 0.722 1.000 0.104 0.092
Floods 3.374 6.973 9.661 1.000 0.891
Wind 3.786 7.825 10.842 1.122 1.000
TABLE 4.3
Multihazard Interaction Matrix for Architectural
Vulnerability Changes
Manmade 1.000 0.365 0.385 1.453 1.295
Earthquakes 2.742 1.000 1.056 3.984 3.550
Fire 2.597 0.947 1.000 3.774 3.363
Floods 0.688 0.251 0.265 1.000 0.891
Wind 0.772 0.282 0.297 1.122 1.000
TABLE 4.4
Multihazard Interaction Matrix for Structural
Manmade 1.000 0.196 0.394 1.744 0.777
Earthquakes 5.095 1.000 2.008 8.884 3.958
Fire 2.538 0.498 1.000 4.425 1.972
Floods 0.574 0.113 0.226 1.000 0.446
Wind 1.287 0.253 0.507 2.244 1.000
TABLE 4.5
Multihazard Interaction Matrix for Ventilation
Manmade 1.000 2.736 1.263 1.293 2.913
Earthquakes 0.366 1.000 0.462 0.473 1.065
Fire 0.791 2.165 1.000 1.024 2.306
Floods 0.773 2.115 0.977 1.000 2.253
Wind 0.343 0.939 0.434 0.444 1.000
TABLE 4.6
Multihazard Interaction Matrix for Fire Protection
Manmade 1.000 0.644 0.236 1.744 1.554
Earthquakes 1.554 1.000 0.367 2.709 2.414
Fire 4.229 2.722 1.000 7.375 6.572
Floods 0.574 0.369 0.136 1.000 0.891
Wind 0.644 0.414 0.152 1.122 1.000
TABLE 4.7
Multihazard Interaction Matrix for Operational
Systems Vulnerability Changes
Manmade 1.000 1.641 1.155 0.654 0.583
Earthquakes 0.609 1.000 0.704 0.398 0.355
Fire 0.866 1.421 1.000 0.566 0.504
Floods 1.529 2.510 1.767 1.000 0.891
Wind 1.716 2.817 1.982 1.122 1.000
TABLE 4.8
Multihazard Interaction Matrix for Nonstructural
Manmade 1.000 0.253 0.337 0.466 0.466
Earthquakes 3.960 1.000 1.334 1.844 1.846
Fire 2.968 0.749 1.000 1.382 1.384
Floods 2.148 0.542 0.724 1.000 1.001
Wind 2.145 0.542 0.723 0.999 1.000
TABLE 4.9
Multihazard Interaction Matrix for Physical Security
Manmade 1.000 13.678 22.462 7.759 19.423
Earthquakes 0.073 1.000 1.642 0.567 1.420
Fire 0.045 0.609 1.000 0.345 0.865
Floods 0.129 1.763 2.895 1.000 2.503
Wind 0.051 0.704 1.156 0.399 1.000
TABLE 4.10
Multihazard Interaction Matrix for IT
Manmade 1.000 0.438 0.674 0.485 1.942
Earthquakes 2.285 1.000 1.540 1.108 4.438
Fire 1.484 0.650 1.000 0.720 2.882
Floods 2.062 0.903 1.390 1.000 4.005
Wind 0.515 0.225 0.347 0.250 1.000
TABLE 4.11
Average Multihazard Interaction Matrix
Manmade 1.000 2.446 3.308 1.766 3.246
Earthquakes 1.907 1.000 1.166 2.234 2.130
Fire 1.763 1.165 1.000 2.190 2.216
Floods 1.317 1.727 2.004 1.000 1.530
Wind 1.251 1.556 1.827 0.981 1.000
provide enough accuracy. For more issues associated with vulnerability of tunnels, see US-DHS
(2011c). The reader is invited to build a higher-resolution vulnerability model from this resource.
The very nature of the weighted averages process, which is the basis of this approach, does not allow
for any negative terms in the interaction matrices. This limits the accuracy of such an approach.
GN models would allow for negative interactions, as shown in the following chapters. But the added
accuracy of GN processes will come at the cost of more computational demand.
4.2.4 decIsIon-based vulnerabIlIty InteractIons

Decision-based MH vulnerability interaction is different from physical-based MH interaction as it
explores the changes in vulnerability due to ith hazard, VHi , to a decision regarding the jth hazard,
based on the decision MH theory. Let us assume that a particular decision, D = d, is made that would
affect {VR} so that
{VR} = {VR ( d )} (4.8)
From Equations 4.2 and 4.8, we express changes of vulnerability due to ith hazard, VHi to a decision
D = d as
= nv
¶VHi
¶d
= å WT
=1
V
i
æ ¶(VR(d )) ö
ç
è ¶d ÷
ø
(4.9)
If the decision, D = d, is concerned with a particular hazard, j, such that
D = dj (4.10)
we obtain the required MH vulnerability decision interaction as
¶VHi ê
é = nv (
æ ¶ VR ( d j ) ) ö÷ ùú
¶d j
= å
ê =1
WTV ç
ç
i è
¶d ÷ ú
ø û
(4.11)
ë
In matrix notations, the MH decision vulnerability matrix [V]DECISION with ¶VHi /¶d j in its ith row
and jth column is expressed as
é ¶V ù
[V ]DECISION = ê ú (4.12)
ë ¶d û
with the ith row and jth column cell [∂V/∂d] is
é = nv (
æ ¶ VR ( d j ) ) ö÷ ùú
ê
å
ê =1
WTV ç
ç
i è
¶d ÷ ú
ø û
ë
The key in this process is to estimate ∂(VR(dj))/∂d that indicates how issues of different vul-
nerability to different hazards change according to the decision dj. An experienced practitioner
should do this step. Note that the values of ∂(VR(dj))/∂d are all relative and can be positive (which
indicate improvements in vulnerability) or negative (which indicates a reduction in vulnerability).
4.2.5 case study 4.2: Mh vulnerabIlIty InteractIons oF tunnels

We consider now the application of the MH decision vulnerability interaction matrix shown in
Equation 4.12, using tunnels as infrastructure of interest. The vulnerability weight matrix [WT V]
of the tunnel under interest is shown in Table 4.12. We use four hazards (manmade, earthquakes,
fire, and flood) with 10 issues that would affect the tunnel vulnerability. If the tunnel under con-
sideration is assessed so as to produce the vector {VR} as in Table 4.13 and the estimated changes
to hazards decisions, di, will produce the matrix of derivatives ∂(VR(dj))/∂d, also displayed in
Table 4.13. The desired MH decision matrix is computed from Equation 4.12 and is shown in
Table 4.14. The ith row and jth column of this matrix indicate the changes of vulnerability of the
tunnel of the jth hazard due to a decision made in regard to the ith hazard. Thus, if the decision-
maker makes a decision regarding earthquake, it is expected that the behavior of the tunnel due to
floods will change by 32%.
Similar to the developments of Section 4.2.2 and due to the subjective nature of the process, only
the relative interaction values can be obtained from these matrices, thus limiting the usefulness of
the process. For more issues regarding vulnerability, US-DHS (2011c) offers more resolutions. The
reader is invited again to build a higher-resolution vulnerability model from that resource. Unlike
the physical vulnerability interaction process of Section 4.2.2, these developments do allow for
negative terms in the interaction matrices, since the signs of the interaction matrices are controlled
via the sign of the parameters ∂(VR(dj))/∂d, which can take a negative value. A final observation on
the interaction matrices is that they are not symmetric (which indicates that different decisions will
have unequal effects on the different hazards vulnerabilities), and they offer only relative terms (due
to the underlying subjectivity of the weighted averages method). We also note that the diagonals are
not unity. Again, this is due to the underlying subjectivity of the weights chosen in the matrix [WT V].
A normalized version of the matrices might be more useful that can be obtained by dividing each
row by its main diagonal term. The normalized matrix is shown in Table 4.15.
86
TABLE 4.12
Structural Method of Liner Cross Liner Relative Construction Known Longest Controlling Type of Seismic Overall Structural
Vulnerability Issues Construction Section Thickness Material Retrofits Span Height Framing Design Conditions
Manmade 13 4 7 11 7 11 2 37 4 4
Earthquakes 6 10 38 4 12 7 9 3 9 2
Fire 4 10 19 9 32 10 7 2 6 3
Floods 41 2 4 8 4 16 5 5 7 7
Multihazard Considerations in Civil Infrastructure
TABLE 4.13
Vulnerability Assessment Vector {VR} and Decision Matrix ∂(VR(dj))/∂d
¶(VR(d j ))
¶d
Issues Affecting
Vulnerability VR Manmade, j = 1 Earthquakes, j = 2 Fire, j = 3 Floods, j = 4
Method of construction 3 1 1 1 3
Liner cross section 9 0 0 0 0
Liner relative thickness 8 6 7 5 0
Construction material 8 2 2 8 0
Known retrofits 7 4 4 2 2
Longest span 3 6 5 0 0
Controlling height 2 0 0 5 0
Type of framing 3 8 8 1 0
Seismic design 2 8 10 0 0
Overall structural conditions 5 3 4 3 1
TABLE 4.14
Multihazard Decision Vulnerability Matrix [V]DECISION
Hazard Manmade (%) Earthquakes (%) Fire (%) Floods (%)
Manmade 51 52 21 6
Earthquakes 43 48 31 4
Fire 39 41 28 8
Floods 31 32 19 14
TABLE 4.15
Row-Wise Normalized Multihazard Decision Vulnerability Matrix [V]DECISION
Manmade 100 101 41 11
Fire 140 148 100 29
Floods 227 234 134 100
4.2.6 reMarks
The methods and examples developed so far utilized a weighted averages approach. Such an
approach has two limitations:
1. Interactions between parameters in matrices are not accommodated.

2. The uncertainties are not accounted for.
Examples of vulnerability interactions will be revisited in later chapters (Chapters 6 through 10)
with case studies where the two assumptions are released using GN processes.
4.3 RELIABILITY-BASED DESIGN: MH VIEWPOINTS

OF GOVERNING EQUATIONS
4.3.1 overvIeW
Current civil infrastructure design process is generally based on accommodating several hazards
in a multilimit states (MLS) design loading conditions (DLCs). Those loading conditions and limit
states are based on setting adequate reliability (as measured by probability of failure, i.e., demands
does not exceed capacity) index (e.g., AASHTO 2007). The design process is based on satisfying
those design MLS for all pertinent hazards in an iterative manner. Figure 4.2 illustrates the concep-
tual design process. The MLS is based on potential occurrences of pertinent hazards in a simultane-
ous fashion (loading conditions), which is based on load factors and strength reduction factors. In
general, the design process tries to optimize capacity to demand ratios and MLS in a “serial” fash-
ion, that is, one limit state at a time. The process will result in a safe but perhaps inefficient design.
Is it possible to improve the design efficiency through consideration of all limit states in a parallel
fashion? We explore this question next.
4.3.2 theoretIcal background

The concept of reliability, Rel, in civil engineering has a long history (see Elishakoff 2004, Rausand
and Hoyland 2004). In the most general sense, reliability is a relationship between the capacity of
Current design practice
Hazard 1 Hazard 2 Hazard N Initialize design
Multilimit states (MLS)

Current design
(loading conditions)
Apply to current design Modify current design
All loading No
conditions
acceptable?
Yes
Need to Yes
improve design
further?
No
Modify current design Stop
FIGURE 4.2 Current design practice.

the system of interest, Capacity, and the demands experienced by that system, Demand (recall that
we use in this book demands, hazards, or threats interchangeably). We express this general reli-
ability relationship as
Rel = f RELIABILITY ( Capacity, Demand ) (4.13)
In order to ensure safety of the system of interest, Equation 4.13 is always subjected to the condition
Capacity ³ Demand (4.14)
The failure limit state* of the system is reached when
Capacity = Demand (4.15)
Generally speaking, for civil infrastructures, there are two types of reliability: structural reliability,
Rel|STR, and system reliability, Rel|SYS. Structural reliability accounts only for the structural system,
while system reliability accounts for reliability of the entire system including Rel|STR. As such,
Rel STR Î Rel SYS (4.16)
Obviously, dealing with Rel|SYS is more involved than handling Rel|STR. In this section, we focus only
on MH aspects of Rel|STR. We look at MH issues of Rel|SYS in later chapters.
Structural reliability, Rel|STR, coupled with the requirement (4.14), has been the cornerstone of
structural design from its inception. In modern days, the two structural design paradigms (allowable
stress design and LRFD) are simply the objective representation of the general reliability expression
(4.13); see AASHTO (2007) and ASCE 7-10 (2013). Let us consider the MH aspects of the more
popular LRFD method that uses Rel|STR at different hazard (load) combinations while accommo-
dating the probabilities of the occurrences of the combinations through application of load factors.
Each load combination is considered a DLC where structural reliability needs to be ensured. For the
ith DLC, we express demands as
j = NHi
Demands i = ågL
j =1
ij j (4.17)
where NHi is the number of hazards (loads) that are included in this DLC. The expression γij is an
appropriate load factor and the expression Lj an appropriate representation of the jth hazard demand
measure. The demand measure can be internal force, displacement, or stress.
The popular LRFD DLC can now be expressed as
j = NH
g ij L j
DCi = å Capacity
j =i i
(4.18)
* Reliability definition in the field of structural design follows the limit states of Equations 4.14 and 4.15. However, in our
handling of the reliability paradigm in this book, we use the more general definition of Equation 4.13. This will allow
us to include both objective and subjective treatments of reliability to make it easier to contrast the concept of reliability
with the concept of risk in the later chapters.
TABLE 4.16
Selected Load Factors for Reinforced Concrete Construction
Fluid Soil
Loading Weights Weights
Combination Dead and/or Live Temperature and or Snow
Condition Load Pressures Loads Loads Pressures Loads Earthquakes Wind
1 1.4 1.4 0 0 0 0 0 0
2 1.2 2.2 1.6 1.2 1.2 0.5 0 0
3 1.2 0 0 0 0 1.6 0 0.8
4 1.2 0 1.0 0 0 0.5 0 1.6
5 1.2 0 1.0 0 0 0.2 1.0 0
6 0.9 0 0 0 1.6 0 1.0 0
It is obvious that DCi is an expression of structural reliability Rel|STR,i since it relates the ith loading
condition (AASHTO 2007, ASCE 7-10 2013). If we drop the subscript STR, we can state
Reli = f ( DCi ) (4.19)
with the requirement
DCi £ 1.0 (4.20)
Note that Equations 4.19 and 4.20 ensure a reliable (safe) structure, with i = 1,2,…,NI, where NI is
the number of DLCs. For the ith loading condition, the jth hazard demand, Dij, is expressed as
Dij = g ij L j (4.21)
while the capacity is expressed as Capacity|j. The design metric for the jth hazard of the DLC
(which can be stresses, displacements, strains, etc.) is Lj. The load factor for the jth hazard and ith
loading condition is γij. Load factors are well developed and can be found in most design-related
references. As can be deduced from Equation 4.18, they are functions of type of hazard and the
particular loading condition. They are also dependent on the material of construction as well as the
type of system of interest, for example, buildings, bridges, and tunnels. An example of a partial list
of γij for reinforced concrete structures is shown in Table 4.16 (ACI 318-05 2005).
The Capacity|i in Equation 4.18 is often expressed as
Capacity i = fi ( Nominal _ Capacity)i (4.22)
where
ϕi is an appropriate reduction factor
(Nominal_Capacity)i is the specified capacity of the ith loading condition
4.3.3 Mh structural relIabIlIty MetrIcs

We now consider applying either the MH physical theory or the MH decision theory to structural
reliability as exemplified by Equation 4.18. A rudimentary examination of the equation shows that
the MH effect is not possible since the relationship between Rel|i and Lj is linear. This would not
produce a relationship between the jth hazard and the th hazard in Equation 4.18. This observation
is not too surprising since the very concept of reliability does not represent a direct physical state
(unlike the direct physical Newtonian states that produced MH interactions in Chapter 3). In order
to explore the MH considerations for Rel|i, we need to have a closer look at the composition of
Equations 4.18 and 4.19.
Let us assume that the jth design measure Lj is a function of x , the independent variables that
control the design measure, with = 1, 2,…, nn. These variables can be geometric, material, or envi-
ronmental properties. We can now rewrite Equation 4.18 as
j = NH
g ij ( L ( x1, x2 ,¼, xnn )) j
DCi = åj =i
Capacity i
(4.23)
It is now clear that Rel|i, DCi, as well as Lj all interact through the system as exemplified by x , as
the physical and decision MH theories imply. The form of Equation 4.23 offers us several ways to
understand how reliability and MH intersect, and how we can use such an understanding to produce
efficient design processes. We offer two fairly useful but different MH structural reliability plat-
forms in Sections 4.3.4 and 4.3.5.
4.3.4 Mh InteractIon through dlc (Mh-dlc)

We are interested in exploring the affinities between different hazards as they respond to different
DLCs as expressed by Equations 4.18 and 4.23. We need first to rewrite Equation 4.18 as
ì DC1, j ü
j = NHi ï ï
ï DC2, j ï
{DC} = å j =1
í
ï ï
ý (4.24)
ïî DC NI , j ïþ
The ith component in {DC} is DCi. Also,
g ij L j
DCi, j = (4.25)
Capacity i
We now define the matrix [Γ] as a matrix with NH columns and NI rows. The jth column is repre-
sented by the vector
ì DC1, j ü
ï ï
ï DC2, j ï
{DC} j = í ý (4.26)
ï ï
ïî DC NI , j ïþ
We realize the vector {DC}j as containing the different demand to capacity ratios for the jth for
all DLCs.
Following the procedure of Section 3.4.2, Chapter 3, we define the average interaction coefficient
between the ith hazard and the jth hazard as
aij = {DC}Ti {DC} j (4.27)

TABLE 4.17
Hypothetical [Γ]
Hazards
DLC Hazard #1 Hazard #2 Hazard #3
DLC1 0.1 0.55 0.3
DLC2 0.2 0 0.6
DLC3 0.2 0 0.6
DLC4 0.02 0.3 0.06
As in Chapter 3, Section 3.4.2, we desire to produce a normalized expression that has a more useful
form. Again, we now define a general-purpose MH interaction coefficient between the ith and the
jth hazards as
( aij )
2
MHIC = MHI ij = (4.28)

( aii ) ( a jj )
Each MHIij represents a component in the desired MH interaction matrix [MH‐DLC]. Again, [MH‐
DLC] is a square symmetric matrix with components that are ≥ 0 and ≤ 1.
We offer a simple example of the use of [MH‐DLC]. Consider Table 4.17, which shows the com-
ponents of a hypothetical [Γ].
In Table 4.17, NH = 3 and NI = 4. When we apply Equations 4.27 and 4.28, we get
é1.000 0.104 1.000 ù

ê ú
[ MH -DLC ] = ê0.104 1.000 0.104 ú (4.29)
êë1.000 0.104 1.000 úû
Now, we can make four useful observations:
1. The unity values in the main diagonals of Equation 4.29 are expected.
2. A perfect affinity (value of 1.0) is evident between hazards 1 and 3. This can be explained
by observing that {DC}1 is linearly related to {DC}3. This means that the distributions of
reliability as expressed by demand to capacity ratios of the two hazards are similar across
all DLCs.
3. A small affinity (value of 0.104) exists between hazards 1 and 2. This can be explained by
observing that the distributions of {DC}1 and {DC}2 are different across all DLCs.
4. Affinities (interactions) between hazards 2 and 3 are identical to those between hazards 1
and 2 in view of the earlier observation #2.
Let us make a slight change in [Γ] as shown in Table 4.18:

Table 4.18 indicates NH = 3 and NI = 4. By applying Equations 4.27 and 4.28, we obtain
é1.000 0.000 0.485 ù

ê ú
[ MH -DLC ] = ê0.000 1.000 0.399 ú (4.30)
êë 0.485 0.399 1.000 úû
We notice that the affinities (interactions) between reliability distributions of hazards 1 and 2 are
nonexistent. Average affinities (interactions) exist between hazards 1 and 3 as well as 2 and 3.
TABLE 4.18
Modified Hypothetical [Γ]
Hazards
DLC Hazard #1 Hazard #2 Hazard #3
DLC1 0.0 0.55 0.4
DLC2 0.2 0 0.6
DLC3 0.2 0 0.32
DLC4 0.0 0.3 0.5
The MH-DLC matrices can help the designer and the analyst to understand the behavior of the
structural system as it relates to DLCs and multiple hazards. Such an understanding can lead to
resolving some design inconsistencies of problems should they arise and streamline the design/
detailing process, especially for complex structural systems. These can ultimately result in an opti-
mal system design and costs.
4.3.5 Mh desIgn usIng desIgn loadIng condItIons (Mh-dlc)

4.3.5.1 Overview
The iterations of the current DLC are derived from reliability-based equations for improving design
efficiency. It is usually done in a serial manner, either manually or automatically, without studying
objectively how reliability of different hazards for different design conditions behaves during the
design iterations.
4.3.5.2 Case Study 4.3: Evaluating MH-DLC Using Simulated Current Practices
As an example of a current procedure for iteration efficiency improvements, consider the design
process of the 2D frame of Figure 3.2. The frame is subjected to five hazards: dead load DL, live
load LL, wind load W, horizontal seismic loads E, and blast load B. There are four design load condi-
tions, NI = 4, with their load factors, γ, as shown in Table 4.19. The applied design loads, as analyzed
by the FEM of Figure 3.2, are shown in Table 4.55 (see Appendix 4A). The resulting displacements
due to all five hazards are shown in Table 4.56 (see Appendix 4A).
Let us assume that the design demand for this example is the horizontal displacement. As such,
the governing (absolute maximum displacements) design values for each hazard, as obtained from
Table 4.56 (see Appendix 4A), can be ascertained as shown in Table 4.20.
Applying the displacement demands of Table 4.20 to the different DLCs identified in Table 4.19,
we obtain the total design demands, Demand|j, shown in Table 4.21, j = 1, 2, 3, 4. Assuming a
TABLE 4.19
Load Factors
Loading
Condition # DL LL W E B
1 1.20 1.60 0.00 0.00 0.00
2 1.20 1.00 1.60 0.00 0.00
3 1.20 1.00 0.00 1.00 0.00
4 1.20 0.50 0.00 0.00 1.00
TABLE 4.20
Maximum Hazard Displacements (Displacement Demands)
Hazard DL LL W E B
Abs. maximum 6.85E−08 4.50E−07 5.99E−04 7.22E−04 2.45E−04
horizontal displacement
demands (in.)
TABLE 4.21
Governing Design Horizontal Displacements for Each Design
Loading Condition
Design Loading
Condition, j Demand|j (in.) Capacity|j (in.) DCj
1 8.02E−07 1.00E−05 0.08
2 9.58E−04 1.00E−03 0.96
3 7.23E−04 1.00E−03 0.72
4 2.45E−04 1.00E−03 0.25
design capacity for each of the DLCs, Capacity|j, as shown in Table 4.21, we finally obtain the DCj
ratio for each condition, also shown in Table 4.21. As expected, the DC ratios satisfy Equation 4.20.
Some of the DCj ratios of Table 4.21 are far from the desired unity. Because of this, the designer
might want to improve the design by opting for design iterations. Normally, DLC number 1 would
be ignored since it pertains to horizontal displacement to predominantly vertical gravity loads.
Because of this, the designer will prioritize the interest of improvements to loading conditions 4,
then 2, and then finally 3, respectively. This is following the DLC from the least value to the most
value, the least value DLC requiring the most attention and prioritization.
We ask ourselves whether this is the most efficient prioritization technique for improving design.
We will have a closer look using an MH technique in the following.
4.3.6 Mh-desIgn eFFIcIency MatrIx

4.3.6.1 Overview
In the previous section, we looked at the efficiencies of DLCs after the contributions of different
hazards summed within each loading condition. By doing so, we cannot really investigate the effi-
ciencies of the designs according to each hazard in a true MH sense. In order to address this limita-
tion, we recast the DLC expression of Equation 4.18 as
j = NH
DCi = å DC
j =1
ij (4.31)
subjected to the conditions of Equation 4.20. In Equation 4.31, the jth hazard demand capacity
ratio contribution within the ith DLC is DCij and the total number of hazards of interest is NH
as usual.
Assuming the linear behavior of the system under consideration, it is not obvious at first how the
different expressions DCij would interact or affect each other. In order to uncover the MH interac-
tions in this situation, we first explore how Equations 4.18 and 4.20 are used in current design prac-
tices with the aid of Figure 4.2. When a designer initiates the design process, Equations 4.18 and
4.20 are evaluated for the initial system. An iterative process follows with the goal of optimizing
Equation 4.20 by bringing the inequalities as close to unity as possible. Thus, for each iteration, ni,
the designer evaluates the components of Equation 4.31 as DCij( n ) . The iterative design process ends
when the designer is satisfied that the inequalities of Equation 4.20 is optimal.
Let us take a closer look at the expression DCij( n ) . Let us assume, for ease of development, that
the design metrics of the system under consideration are the kth displacements within the general
system displacements vector {U}j that can be expressed as ujk. The order of this vector is N that
denotes the total number of degrees of freedom of the system. The displacements are the response
of the system to the jth hazard. Recall that
{U} j = [ Á]ni {P} j (4.32)
The matrix [ Á]ni denotes the flexibility matrix of the system during the nith design iteration, while
the vector {P}j denotes the numerical representation of the jth hazard. An assumption of system
nonlinearity is implied in Equation 4.32 via the independence of {P}j from the design iteration
process. Thus,
u jk = {P} j ëé Áùû ni { A}k

T
(4.33)
The superscript T denotes a matrix (or vector) transpose. The cells of the vector {A}k are all null,
with the exception of the kth cell that we assign a value of unity. From Equations 4.18, 4.31, and
4.33, we obtain
({P} éëÁùû { A}k )

g ij T
DCij = (4.34)
Capacity i j ni
The matrix éë Áùû ni is an extremely complex matrix. It is usually a full matrix with highly interactive
components. In order to illustrate the MH concepts of this section in a simple manner, we assume
the matrix éë Áùû ni in the form
( )
éë Áùû ni = f x ( ni )1 , x ( ni )2 ,…, x ( ni )m ëé Áùû (4.35)
The function f() is an analytic function, while x ( ni ) are scalar variables that control the nith itera-
tive design process. Also, = 1, 2,…, nn, where nn, represent an adequate number of variables, which
are changed during each nith design iteration process. The simplifying assumption of Equation 4.35
can be released of course. However, the resulting MH process will become more computationally
demanding.
We have all the needed tools now to evaluate the MH-DDM matrix, which includes the relative
effects of changes of design iterations on the demand capacity components of each hazard. These
are expressed as
¶ Dij
( MHDDM )ij = (4.36)
¶x(ni )1 ¶x(ni )2 … ¶x(ni )
The term (MHDDM)ij forms the ith row and the jth column in the MH-DDM, [MH‐DDM]. The
number of columns of the matrix is NH, while the number of the rows is NDLS, which represents
the number of DLCs.
4.3.6.2 Case Study 4.4: Simple Analytic Expressions of MH-DEM

In some situations, analytic expressions of Equation 4.36 can be obtained. For example, if the flex-
ibility matrix éë Áùû is expressed as
éë Áùû = Com éë Áùû (4.37)
with the compliance of the system material is
1
Com = (4.38)
E
where E is the modulus of elasticity of the system material. The implicit assumptions in these
expressions are as follows:
1. The structural system is built using only a single material with modulus of elasticity E.
2. The design iterations involve only changes to the value of E. This situation might happen if
the material of interest is concrete and the designer tries to optimize the system design via
changes in the concrete stiffness.
The components of the MH design efficiency matrix (MH-DEM) are now expressed as
¶Dij
( MH -DEM )ij = (4.39)
¶Com
({P} éëÁùû { A}k )

¶Dij g ij T
= (4.40)
¶ Com Capacity i j ni
To show how MH-DEM is computed and utilized, we continue the example of Section 4.3.5.2. By
applying Equation 4.40, we obtain [MH‐DEM] as shown in Table 4.22. A normalized version of the
matrix (i.e., maximum value in the matrix = 1.0) is shown in Table 4.23.
MH-DEM of Table 4.22 and its normalized counterpart of Table 4.23 show that the second DLC
has the highest sensitivity to design iterations and is from the wind hazard contribution. The second
highest contribution is the third DLC with the high contribution coming from the seismic hazard.
The remaining order is third DLC followed by the fourth DLC with contributions from blast and
TABLE 4.22
Multihazard–Design Efficiency Matrix (Using Simple Analytical Methods)
Hazard Contribution
Design Loading
Condition DL LL W E B
1 8.22E−03 7.20E−02 0.00E + 00 0.00E + 00 0.00E + 00
2 8.22E−05 4.50E−04 9.58E−01 0.00E + 00 0.00E + 00
3 8.22E−05 4.50E−04 0.00E + 00 7.22E−01 0.00E + 00
4 8.22E−05 2.25E−04 0.00E + 00 0.00E + 00 2.45E−01
TABLE 4.23
Normalized Multihazard–Design Efficiency Matrix
(Using Simple Analytical Methods)
Hazard Contribution
Design Loading
1 8.58E−03 7.51E−02 0.00E + 00 0.00E + 00 0.00E + 00
2 8.58E−05 4.70E−04 1.00E + 00 0.00E + 00 0.00E + 00
3 8.58E−05 4.70E−04 0.00E + 00 7.54E−01 0.00E + 00
4 8.58E−05 2.35E−04 0.00E + 00 0.00E + 00 2.56E−01
live loads, respectively. This ordering shows that the sensitivity to design iterations will be most
if the designer concentrates on the second DLC rather than the fourth DLC as obtained from the
example in Section 4.3.5.2. The implication here is that we need to consider hazards’ contributions
to DLC in order to achieve an efficient design iteration process.
4.3.6.3 Case Study 4.5: Numerical Solution of MH-DEM for Portal Frame Design
In general, it is not easy to find the analytical expressions for flexibility matrix [ Á]. Thus, there is
a need to develop a more generic solution for Equation 4.36. We rely in this task on developing the
different expressions for the differentials in Equation 4.36. By inspecting closely Equation 4.34, we
can rewrite it in the form
=N
g ij
Dijni =
Capacity i å( p f )
=1
ni
j k (4.41)
with p j as the th component of {P}j and fk as the component of éë Áùû in the jth row and the th
column (recall that éë Áùû is symmetric). Note that ni is the nith design iteration and k is the design
metric (design metric is assumed to be a displacement measure, without any loss of generality of the
equations, since similar developments can be produced for any other design metric). We can now
approximate the DEM during the nith design iteration as
Dijni - Dijni -1
( MH -DEM )ijni = (4.42)
Dijni -1
Note that we used a backward difference expression in Equation 4.42. Obviously, a forward differ-
ence expression also can be used in the expression.
We can use the numerical expressions of Equations 4.41 and 4.42 for optimal MH design itera-
tions of systems of any size.
To show how MH-DEM is used for situations where only the difference process, Equation 4.42,
is possible, we continue with the example of Section 4.3.5.2. We increase the depth of the beam
from 24 to 26.4 in. Also, we change the depth of the columns from 12 to 14.4 in. These changes led
to a reduction to DC ratios, except in DLC #1 because the design condition is mainly a vertical load-
ing condition that did not affect horizontal displacements. The new total design demands, Demand|j,
are shown in Table 4.24 for loading conditions, j = 1, 2, 3, 4. Using the same design capacity for
each of the DLCs, Capacity|j, as before, we obtain the new DCj ratio for each condition as shown in
Table 4.24 using corresponding displacements from Table 4.57 (Appendix 4B).
We can now compute MH-DEM by applying Equation 4.42; the result is shown in Table 4.25.
Neglecting DLC #1, since it is only a vertical loading condition, as we are interested in horizontal
TABLE 4.24
Governing Design Horizontal Displacements for
Each Design Loading Condition (Improved System)
Design Loading
Condition, j Demand|j (in.) Capacity|j (in.) DCj
1 7.52E−07 1.00E−05 0.08
2 6.11E−04 1.00E−03 0.61
3 4.60E−04 1.00E−03 0.46
4 1.55E−04 1.00E−03 0.16
TABLE 4.25
Normalized Multihazard–Design Efficiency Matrix
(Using Simple Analytical Methods)
Hazard Contribution
Design Loading
1 0.054 1.000 0.000 0.000 0.000
2 0.001 0.006 0.052 0.000 0.000
3 0.001 0.006 0.000 0.033 0.000
4 0.001 0.003 0.000 0.000 0.033
loading conditions, the DLC that is most sensitive to hazards (wind) is #2, followed equally by
#3 and #4. The results are coincidentally similar, but not identical, to the analytical result of
Section 4.3.6.2.
4.4 RELIABILITY-BASED DESIGN: MH SERIAL

VERSUS PARALLEL DESIGN PROCESS
4.4.1 overvIeW
The two main obstacles in applying an MH approach to tall buildings are as follows: (1) abnormal
conditions vary immensely in almost all aspects of design and analysis and (2) analyzing a tall
building is, by default, a daunting task. In addition, all known design guidelines, standards, and
codes have not followed MH strategies. These factors, among others, have made it extremely dif-
ficult to accommodate a true MH design, both qualitatively and quantitatively.
In this book, we explore several aspects of MH designs of buildings. We follow some of the
principles of the theory of multihazard that was introduced in this book (also see Ettouney et al.
2005). The theory argued for the inherent resiliency of any structural system to resist all types of
hazards. The first example covers the inherent resiliency of the structural system to resist wind and
progressive collapse of a tall building. Next, the varied efficiency of structural systems according to
the types of hazards (single or multi) is studied. Finally, the life-cycle cost of hazards and the effects
of MH considerations are shown in an example. It is shown that the cost effects of ignoring MH
demands can be as large as 18% in this example.
4.4.2 applIcatIon oF Mpt: WInd and progressIve collapse oF tall buIldIngs

Consider the situation of a tall building that is located in an area without any seismic provision
requirement. Only wind loading is required for the design of the building. However, the building
is also required to follow measures for progressive collapse mitigation. There are two procedures
for designing such a situation: a one-hazard-at-a-time procedure and an MH procedure. The one-
hazard-at-a-time procedure will design the building for wind and progressive collapse indepen-
dently. At first, a wind design will be evaluated and then the measures for progressive collapse
demands will be applied. The second approach, an MH approach, will design for wind first, then
apply progressive collapse measures, and then readjust the wind designs. The differences between
the two approaches are illustrated in the following.
In order to compare the two design procedures mentioned earlier for wind and progressive col-
lapse hazards, a test case building was created with the intent of closely simulating possible real-
world design. At the same time, the design was kept as simple as possible to improve the efficiency
of effort in computations without sacrificing the accuracy of results. The building will be referred
to as Building A. The building is 49 stories with a typical floor-to-floor height of 12 ft 6 in. and a
double-height first floor giving a total building height of 625 ft.
The typical building framing plan is shown in Figure 4.3. It is a simple 5-bay by 5-bay plan with
columns spaced at 30 ft, giving plan dimensions of 150 ft × 150 ft. The corner bays are removed to
create a cruciform shape in the plan. The elevator core mimics the shape of the building with a pair
of central corridors that cross at the center. Two braced frames in each orthogonal lie on the edges
of the edges of the core, keeping with the cross theme of the design.
Figure 4.4 shows the elevation of the typical braced frame. The central bay uses an eccentric
bracing to allow for the passage of the corridor. The adjacent bays, which lie behind the elevators in
the plan, are concentrically braced. An outrigger truss is located between the 33rd and 35th floors
to engage the exterior columns.
FIGURE 4.3 Floor layout of Building A.

FIGURE 4.4 Elevation of typical braced frame.
The building is intended to be an office structure with a live load of 80 psf. Mechanical spaces
are located at the roof and outrigger floors. The curtain wall weight is 20 psf. The floor is composite
construction using a 3 in. metal deck with 3¼ in. normal-weight concrete topping. A yield stress of
50 ksi, conforming to A992 steel, was used for all structural steel members.
Wind loads were calculated based upon ASCE 7-98 (1998). The intensity of the wind loading
is similar to that in New York City. The wind loads were determined for a basic wind speed of 110
mph, exposure category A, and importance factor 1.15. The gust factor was determined based upon
the dynamic properties of the structure.
The first step in the study was to analyze and design this structure for the combined effects
of wind and gravity loads in much the same manner as one would design a typical project. The
program ETABS (2004) was used for both analysis and design. The steel member sizes were deter-
mined through an iterative optimization process. For the purposes of this study, serviceability cri-
teria, such as human comfort under wind-induced vibrations, were ignored. The building design in
this case is referred to as Case A1.
Next, a progressive collapse upgrade to the building was performed without any considerations
of the wind resiliency already present. Building A, as described earlier, was used. The upgrade was
applied by following the procedures outlined in Smith et al. (1998). This was achieved by adding
moment connections at the perimeter of the building and increasing the sizes of the perimeter col-
umns and beams to withstand the progressive collapse-induced forces. The interior column, beam,
and brace sizes remained as determined by the previous wind load design.
The progressive collapse design procedure assumed the loss of any one column immediately
above the ground floor. For simplicity, the design process employed a linear static design procedure;
see Smith et al. (1998). This was done by applying a dynamic load factor of 2.0 to the gravity loads
in the influence area of the lost column. The design also assumes a strength increase factor of 5%
(Fy = 52.5 ksi). The analysis is performed for service level loads, 1.0× dead load and 0.25× live load.
Members are designed elastically, with a small increase in the allowable stress ratio to allow for
some ductile behavior. The building design in this case is referred to as Case A2.
TABLE 4.26
Wind versus Progressive Collapse Study Results
One Hazard at a Time (Independent) MH
Case A1 Case A2 (Wind and Case A3 (Wind and
(Wind Only) Progressive Collapse) Progressive Collapse)
Steel weight (t) 8351 9564 9412
Cost ($3150/t) $26.31M $30.13M $29.65M
The third design of Building A was performed as follows: given the progressive collapse require-
ments for the exterior beams and columns determined earlier in Case A2, redesign of the interior
beams, columns, and braces for wind and gravity forces. For this case, the exterior member sizes
remained fixed and the interior sizes were reoptimized using methods similar to those employed in
Case 1. The building design of this case is referred to as Case A3.
Utilizing simple cost estimation procedures that are based on steel tonnage, the structural costs
of Cases A1, A2, and A3 were evaluated. Table 4.26 summarizes the results of the relative costs of
the structural steel framing for the three cases.
Table 4.26 shows the advantage of an MH design procedure. When the building was designed
using a one-hazard-at-a-time approach, the total cost of the design was $30.1M. When the build-
ing was designed using an MH approach, the cost of design was $29.65M. This represents a sav-
ings of $480,000.00. The savings amount to almost 1.5% of the total structural steel framing cost.
Note that the iterations that were needed to evaluate the MH design required almost no additional
costs since the designs were all done using a building model that was already available. The
results in Table 4.26 show that the progressive collapse mitigation measures that were adopted in
Case A2 included some inherent wind resiliency, which was then taken into consideration during
the design of Case A3. This is the MH inherent resiliency that the theory of multihazard predicted
(Ettouney et al. 2005).
4.4.3 InteractIon oF Mh buIldIng systeMs: WInd and seIsMIc hazards

The MH considerations of buildings, as presented by the MPT, imply an interaction between how
the buildings respond to hazards, that is, the capacities to resist hazards. This means that the effi-
ciencies of different systems that resist a given hazard depend on the magnitude and type of another
hazard. To amplify this concept, let us consider the MH effects of wind and seismic loads as applied
to a tall building. We use a simplified version of Building A for this situation. For the simplified
building, referred to as Building B, the floor-to-floor height (12 ft 6 in.), number of stories (49), and
total height (625 ft.) are unchanged from Building A. The column spacing (30 ft.) and number of
bays (5 × 5) are also unchanged, but the corner bay is no longer removed so that the building has a
simple square plan. For simplification, the mechanical floors and elevator openings were removed.
The plan drawing of Building B is shown in Figure 4.5.
The lateral load system for this study is located in frames at the building exterior rather than
the interior frames used in the previous study. Two different lateral load resisting systems were
examined, a concentrically braced frame (CBF) and an eccentrically braced frame (EBF). Typical
elevations of these frames are shown in Figures 4.6 and 4.7.
The wind loads used were the same as those used for Building A (basic wind speed 110 mph,
Exposure Category A, and Importance Factor 1.15). Seismic loads were determined as per IBC
(2000), with short and long period spectral response values of Ss = 1.0 s and S1 = 0.4 s. These magni-
tudes of wind and seismic loading are comparable to those that would be used for a building located
near the coast of North Carolina. The building is assumed to be located on soft soil, categorized as
FIGURE 4.5 Floor plan of Building B.
FIGURE 4.6 Eccentrically braced frame of Building B.

FIGURE 4.7 Concentrically braced frame of Building B.
Site Class E. The building is categorized as Seismic Use Group II. Linear static analysis was used
for seismic loads.
A total of four analysis/design cases, B1, B2, B3, and B4, were studied, as defined in Table 4.27.
For the wind-only case, B1, a CBF was chosen since it is expected to be more efficient than an EBF.
For the seismic-only case, an EBF was chosen since it is expected to be more efficient than a CBF.
However, for our MH purpose, we are more interested in Cases B2 and B4.
Note that while the site conditions were exactly the same for the four cases, the applied lateral
base shear varied for each model. The wind gust factor was determined based upon the dynamic
properties of the structure. Wind forces were reduced when the building stiffness was increased.
Several factors used in the determination of the seismic loads are dependent on the structural sys-
tem. The coefficient for the approximate fundamental period, CT, has a value of 0.02 for the CBF
and 0.03 for the EBF. Also, the response modification coefficient, R, has a value of 5 for the CBF
and 8 for the EBF.
For each case, structural steel member sizes were optimized to produce the minimum weight of
structural steel required to satisfy the design loads.
It is interesting that the seismic only EBF case, B3, produced the least expensive design. Yet
when the wind is added to the design conditions, the system produced the most expensive design.
TABLE 4.27
Wind versus Seismic Test Cases and Results
Case Frame Applied Lateral Loads Max Base Shear (Kips) Steel Wt. (t) Cost ($3150/t)
Case B1 CBF Wind only 4137 7430 $23.40M
Case B2 CBF Wind and seismic 5398 7607 $23.96M
Case B3 EBF Seismic only 3317 7219 $22.74M
Case B4 EBF Wind and seismic 4818 7867 $24.78M
For the two MH designs, B2 and B4, the CBF system is the least expensive. The CBF, in an MH
condition, is the preferred solution, even though it is not the preferred solution in single hazard
conditions. This leads us to this important conclusion. In an MH design environment, the structural
systems interact in a different manner than they do in a single hazard condition (i.e., if hazards are
considered separately). The savings can be substantial if this system interaction is accounted for. In
the Building B case, the savings of using Case B2 instead of Case B4 is $820,000 or about 3.5% of
the total cost.
4.4.4 lIFe-cycle cost analysIs: seIsMIc versus blast

One of the least considered aspects of structural design of buildings is the life-cycle cost aspects of
design decisions. The life-cycle cost decision implications also depend on whether the building is
subjected to MH. In this example, the effects of MH considerations on life-cycle cost analysis are
investigated. In order to simplify the example, we consider only a single time period as the basis of
the example. For multiperiod extensions, the discount rate must be considered; see Hawk (2003).
However, since our immediate purpose is to show implication of MH considerations, a single period
study can be used without loss of any accuracy. Life-cycle cost analysis for the ith hazard for a single
time period can be defined as
j=N
Ci = åp c
j =1
ij ij (4.43)
where pij is the probability of occurrence of the ith hazard with jth intensity in the time period. The
number of subdivisions in the hazard intensity space is N. The cost of ith hazard with jth intensity
is cij. Note that, by definition, for a given hazard, that is, constant i,
j=N
åp
j =1
ij = 1.0 (4.44)
In order to perform life-cycle cost analysis, we have to develop a method for computing the costs cij
that result from different hazards.
In this example, we will study life-cycle costs of seismic and blast MH effects. Building B with
an EBF will be used in this example. The level of seismicity was reduced slightly (Ss = 0.75 s and
S1 = 0.3 s). The other seismic load coefficients remained the same (CT = 0.03, R = 8, Seismic Use
Group II, Site Class E). The steel member sizes were optimized for gravity and seismic loads. After
the initial optimization, the member sizes were unchanged for all subsequent analyses.
Since life-cycle cost analysis is done for a given geometry, the approach to this example is dif-
ferent from the previous two. Rather than optimizing member sizes for different loading and design
criteria, the building design was kept constant, while stress levels were investigated for various
levels of loading. In this study, the building was subjected to three separate earthquake loadings
(EQ1, EQ2, EQ3). These earthquakes represent three hazard levels (low, medium, high), with EQ2
being equivalent to the design earthquake. The spectral accelerations of the three earthquakes are
listed in Table 4.28.
Each earthquake was applied as an equivalent static load based on the response spectrum values
listed in Table 4.28. Since this study is looking at actual member stress levels, the applied loads are
not reduced by the full response modification coefficient, R. In order to get an estimate of the actual
response level, a value of R = 1.5 is used for the investigation. Additionally, the members are checked
for the load combination 1.0DL + 0.25LL + 1.0EQ rather than the usual design load combinations,
which have an inherent factor of safety.
TABLE 4.28
Seismic Hazard Spectral Accelerations
Earthquake Ss (s) S1 (s)
EQ1 (low) 0.5 0.2
EQ2 (medium: design level) 0.75 0.3
EQ3 (high) 1.0 0.4
For each analysis, an accounting is made of the structural members and their stress level. Each
member is categorized as undamaged, lightly damaged, moderately damaged, or highly damaged
depending on the stress ratio, r. For the definition of the stress ratio, r, see ETABS (2004). The rela-
tionship between stress ratio, r, and the damage level is shown in the left column (“Components and
Damage Levels”) of Table 4.29. Note that this is a qualitative estimate, as a more accurate estimate
would require a nonlinear dynamic analysis. Because beams and braces exhibit greater ductility
than columns, the damage in these members is assumed to occur at higher stress levels.
A cost is assigned to each level of damage of each member. Because the failure of a column is
more catastrophic than the failure of a beam or brace, the cost of column damage is greater than that
of other members. The results of the number of damaged components and the costs of their retrofits
(individual and total) for each of the three postulated seismic events are shown in Table 4.29. We
estimate the probabilities of occurrence pij of each seismic event (EQ1, EQ2, and EQ3) during the
period under consideration. By applying Equation 4.43, we can estimate the seismic life-cycle cost
as summarized in Table 4.30.
Next, we perform similar life-cycle cost analysis for blast events. The study assumes that each
blast event will affect a different envelope of the building. The low-level blast event affects only one
bay, while the high-level blast event affects one entire façade. The study assumes that 50% of the
members within the blast envelope sustain light damage, 30% sustain moderate damage, and 20%
are highly damaged. Finally, probabilities of occurrence are assigned to each blast event. The results
of the blast life-cycle analysis are also summarized in Tables 4.31 and 4.32. The total life-cycle costs
for seismic and blast hazards are then combined to reach $7,308,625, according to Equation 4.43.
Next, the study investigates the effects on life-cycle cost when a blast retrofit is applied to the
structure. The retrofit consists of a 24 in. thick concrete blast wall constructed at the perimeter of
the building between the ground and fourth floor. The analysis model of Building B is modified to
include this wall, and the complete seismic life-cycle cost analysis is performed on the retrofitted
structure. The blast life-cycle cost analysis is also performed for the retrofitted structure with the
assumption that blast effects are greatly reduced. The results are summarized in Tables 4.33 through
4.36. The total life-cycle costs for seismic and blast hazards are then combined to reach $8,636,875
after applying Equation 4.43.
Comparing the total life-cycle cost of the original building and the blast-hardened building shows
that the blast-hardened building is $1,328,250 more expensive than the original building, nearly an 18%
increase in cost. This means that the hardening of the buildings actually costs more than if the build-
ing was left in its original state. This result is due to the fact that by hardening the lower few floors of
the building to resist blast effects, the building seismic properties are changed. The changes include
lowering the natural period of the building, thus increasing the seismic demand on the building. These
changes increased seismic vulnerability of the building, thus the damage levels of different building
components are changed. Note that the blast costs in the hardened building decrease considerably;
however, the increase in seismic costs more than offsets the decreased blast costs. The blast hardening
of the lower floors is in conflict with the seismic need for a lower period, which increases the life-cycle
cost of the building; for more discussion, see Ettouney et al. (1998) and Ettouney (2001). This shows
that MH life-cycle cost considerations can result in increased costs if the two hazards have conflicting
106
TABLE 4.29
Original Building B, Damage and Costs for Different Forecasts of Future Three Seismic-Level Events (i = 1)
Seismic Event (EQ1), j = 1 Seismic Event (EQ2), j = 2 Seismic Event (EQ3), j = 3
Cost
Components and Stress Component No. of Damaged No. of Damaged No. of Damaged
Damage Levels Ratio Retrofit Components Cost Components Cost Components Cost
Columns
Undamaged r ≤ 1.5 $0 1764 $0 1748 $0 1700 $0
Low 1.5 < r ≤ 2 $10,000 0 $0 16 $160,000 64 $640,000
Moderate 2 < r ≤ 2.5 $50,000 0 $0 0 $0 0 $0
High 2.5 < r $150,000 0 $0 0 $0 0 $0
Beams
Undamaged r≤2 $0 974 $0 402 $0 176 $0
Low 2 < r ≤ 2.5 $5,000 6 $30,000 514 $2,570,000 496 $2,480,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 64 $1,280,000 308 $6,160,000
High 3<r $50,000 0 $0 0 $0 4 $200,000
Braces
Undamaged r≤2 $0 1664 $0 644 $0 356 $0
Low 2 < r ≤ 2.5 $5,000 296 $1,480,000 948 $4,740,000 1056 $5,280,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 368 $7,360,000 548 $10,960,000
High 3<r $50,000 0 $0 0 $0 276 $13,800,000
Total event costs Cij $1,510,000 $16,110,000 $39,520,000
TABLE 4.30
Life-Cycle Costs for Future Three Seismic-Level Events
(Original Building B)
Seismic Event Seismic Event Seismic Event
(EQ1), j = 1 (EQ2), j = 2 (EQ3), j = 3
pij pijcij pij pijcij pij pijcij
0.8 $1,208,000 0.15 $2,416,500 0.05 $1,976,000
Seismic cost $5,600,500
demands, as in this case. Ettouney et al. (2005) have predicted this phenomenon. They also predicted
that as the two hazards have consistent demands, the life-cycle costs will decrease for the MH situation.
4.4.5 reMarks
MH considerations for tall buildings were considered in this section. The inherent capacities of tall
buildings to resist MH were evaluated for a wind and progressive collapse example. Next, a wind
and seismic example showed that the efficiency of structural systems can differ depending on the
number and level of hazards under consideration. Finally, a life-cycle cost example showed that
those life-cycle costs can increase immensely if the different hazards and the underlying building
systems interact in a conflicting manner.
It is concluded that a considerable cost implication can result from considering, or ignoring, MH
effects. In the examples shown earlier, those cost implications ranged from 1.5% to 18%. In tall
buildings, this would be a major cost that could be easily saved if proper MH considerations were
followed during the design stage of the building.
4.5 EXPOSURE: ARE INFRASTRUCTURES EXPOSED

TO ONE HAZARD AT A TIME?
4.5.1 overvIeW
Reliability, as a relationship between capacity and demands, is suited for numerous applications;
perhaps the most popular is in design applications. Reliability is expressed as a ratio between capac-
ity and demand in one form or another. However, there are several other instances where this type of
relationship is not convenient. One of those instances is when we are interested in the likelihood of
an event. In such a situation, the likelihood is expressed as a multiplier of demand* (expressed in the
probability of the event) and capacity† (expressed as the outcome of the event). Another useful form
of expressing relationship between capacity and demands is exposure. We introduce the exposure
definition next. We then provide methods of defining MH exposure interaction and some examples
of MPT and MDT as applied to exposure of civil infrastructures.
4.5.2 exposure deFInItIons

Exposure has several definitions depending on the industry or community that uses it. For example,
the financial industry (Jorion 2007), in some situations uses the term “exposure” in a similar fashion
that we are using absolute risk; see section 6.4 for a case study on using exposure and section 8.3.3
for a case study on using absolute risk. In this book, we define exposure as a function of the hazard,
* We use expressions of demands, hazards, and threats interchangeable in this chapter.

† Capacity or vulnerability. The relationship between capacity and vulnerability is discussed in Section 4.2.1.
108
TABLE 4.31
Original Building B, Damage and Costs for Different Forecasts of Future Three Blast-Level Events (i = 2)
Blast Event #1, j = 1 Blast Event #2, j = 2 Blast Event #3, j = 3
Cost
Columns
Undamaged r ≤ 1.5 $0 1762 $0 1746 $0 1176 $0
Low 1.5 < r ≤ 2 $10,000 1 $10,000 9 $90,000 294 $2,940,000
Moderate 2 < r ≤ 2.5 $50,000 1 $50,000 5 $250,000 176 $8,800,000
High 2.5 < r $150,000 0 $0 4 $600,000 118 $17,700,000
Beams
Undamaged r≤2 $0 978 $0 964 $0 539 $0
Low 2 < r ≤ 2.5 $5,000 1 $5,000 8 $40,000 220 $1,100,000
Moderate 2.5 < r ≤ 3 $20,000 1 $20,000 5 $100,000 132 $2,640,000
High 3<r $50,000 0 $0 3 $150,000 89 $4,450,000
Braces
Undamaged r≤2 $0 1954 $0 1922 $0 1176 $0
Low 2 < r ≤ 2.5 $5,000 3 $15,000 19 $95,000 392 $1,960,000
Moderate 2.5 < r ≤ 3 $20,000 2 $40,000 11 $220,000 235 $4,700,000
High 3<r $50,000 1 $50,000 8 $400,000 157 $7,850,000
Total event costs Cij $190,000 $1,945,000 $52,140,000
TABLE 4.32
Life-Cycle Costs for Future Three Blast-Level Events
(Original Building B)
(EQ1), j = 1 (EQ2), j = 2 (EQ3), j = 3
0.85 $161,500 0.125 $243,125 0.025 $1,303,500
Blast cost $1,708,125
H, (or demand) and the vulnerability, V, (or capacity) of an asset network, a single asset, or a com-
ponent of an asset, to such a hazard. Formally, exposure can be expressed as
E = fEXPOSURE ( H , V ) (4.45)
Note that by following such a definition and noting the general reliability relationship (see
Equation 4.13), we are relating exposure to be the inverse of reliability in the general sense that
higher reliability implies lower exposure and lower reliability implies higher exposure. In general,
exposure is usually expressed in a subjective manner, whereas reliability (and likelihood) is usu-
ally expressed and computed in an objective manner. Perhaps, the most popular exposure use is in
the field of infrastructure inspection. When the inspector assigns a subjective condition rating to
a structure or its components, for example, they are actually assigning an exposure rating to the
entity they are observing. The rating, say a fair condition rating, which is observed by an inspec-
tor, accounts for the combined effects of the hazard (say fatigue) and the vulnerability of the entity
to fatigue: the result is a subjective estimate of the exposure of the entity being observed. In many
situations, the observed exposure is a resultant of several hazards. For our purposes, we are inter-
ested in an MH exposure and its uses.
4.5.3 physIcal Mh exposure InteractIons

We start by investigating simple exposure definitions. Let us assume that exposure to the ith hazard,
Hi, is E Hi . It can be expressed as
E Hi = VHi PHi (4.46)
where
VHi is a vulnerability rating due to the ith hazard
PHi is a threat (hazard) rating of the ith hazard
Equation 4.46 can be expressed as
{E} = {V} Ä {H} (4.47)
with E Hi , VHi , and PHi as the ith components of the vectors {E}, {V}, and {H}, respectively. The order
of Equation 4.47 is NH. The notation ⊗ indicates a vector-by-vector multiplication.
In its simplest form, the vulnerability rating can be expressed as a weighted average of several
parameter (see US-DHS 2011a,b,c, Fenton and Neil 2013), such that
{V } = éëWTV ùû {VR} (4.48)

110
TABLE 4.33
Blast-Hardened Building B, Damage and Costs for Different Forecasts of Future Three Seismic-Level Events (i = 1)
Seismic Event (EQ1), j = 1 Seismic Event (EQ2), j = 2 Seismic Event (EQ3), j = 3
Cost
Columns
Undamaged r ≤ 1.5 $0 1760 $0 1716 $0 1680 $0
Low 1.5 < r ≤ 2 $10,000 4 $40,000 44 $440,000 84 $840,000
Moderate 2 < r ≤ 2.5 $50,000 0 $0 4 $200,000 0 $0
High 2.5 < r $150,000 0 $0 0 $0 4 $600,000
Beams
Undamaged r≤2 $0 850 $0 346 $0 168 $0
Low 2 < r ≤ 2.5 $5,000 130 $650,000 430 $2,150,000 492 $2,460,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 204 $4,080,000 320 $6,400,000
High 3<r $50,000 0 $0 0 $0 96 $4,800,000
Braces
Undamaged r≤2 $0 1560 $0 620 $0 420 $0
Low 2 < r ≤ 2.5 $5,000 400 $2,000,000 1000 $5,000,000 844 $4,220,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 340 $6,800,000 696 $13,920,000
High 3<r $50,000 0 $0 132 $6,600,000 384 $19,200,000
Total event costs Cij $2,690,000 $25,270,000 $52,440,000
TABLE 4.34
Life-Cycle Costs for Future Three Seismic-Level Events
(Blast-Hardened Building B)
(EQ1), j = 1 (EQ2), j = 2 (EQ3), j = 3
0.8 $2,152,000 0.15 $3,790,500 0.05 $2,622,000
Seismic cost $8,564,500
Similarly, the hazard (threat) rating can be expressed as
{H} = éëWTH ùû {HR} (4.49)
Both matrices [WT V] and [WTH] include a set of weights that are related to the system and a set of
hazards under consideration. The order of {VR} is the number of issues/considerations that affect
different hazard vulnerability ratings, nv. Similarly, the order of {HR} is the number of issues/con-
siderations that affect different hazards/threat ratings nh.
We define MH exposure interaction here as the change in the exposure to the ith hazard, E Hi, due
to a change in the exposure to the jth hazard, E H j . Formally, MH exposure interaction is defined as
¶E Hi æ ¶E Hi ö æ ¶VH j ö
=ç ÷÷ çç ÷÷ (4.50)
¶E H j çè ¶VH j ø è ¶E H j ø
This can be rearranged as

-1
¶E Hi æ ¶E Hi ö æ ¶E H j ö
=ç ÷÷ çç ÷÷ (4.51)
¶E H j çè ¶VH j ø è ¶VH j ø
From Equation 4.51, we know that ¶E Hi /¶E H j satisfies
¶E Hi
<¥ (4.52)
¶E H j
Since ¶E H j /¶VH j > 0, which simply states that exposure to a hazard that causes such an exposure is
always finite, no matter how small such an exposure is, we consider Equations 4.50 through 4.52 as
another proof for the physical theory of MH.
We need now to quantify expressions of Equation 4.50. From Equations 4.48 and 4.49, we surmise
æ k = nv ö
VHi = ç
ç
è k =1
å WTV ik (VR)k ÷
÷
ø
(4.53)
æ k = nh ö
H Hi = ç
ç å
è k =1
WTH ik ( HR)k ÷
÷
ø
(4.54)
112
TABLE 4.35
Blast-Hardened Building B, Damage and Costs for Different Forecasts of Future Three-Blast Level Events (i = 2)
Blast Event #1, j = 1 Blast Event #2, j = 2 Blast Event #3, j = 3
Components and Stress Cost Component No. of Damaged No. of Damaged No. of Damaged
Columns
Undamaged r ≤ 1.5 $0 1764 $0 1762 $0 1746 $0
Low 1.5 < r ≤ 2 $10,000 0 $0 1 $10,000 9 $90,000
Moderate 2 < r ≤ 2.5 $50,000 0 $0 1 $50,000 5 $250,000
High 2.5 < r $150,000 0 $0 0 $0 4 $600,000
Beams
Undamaged r≤2 $0 980 $0 978 $0 964 $0
Low 2 < r ≤ 2.5 $5,000 0 $0 1 $5,000 8 $40,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 1 $20,000 5 $100,000
High 3<r $50,000 0 $0 0 $0 3 $150,000
Braces
Undamaged r≤2 $0 1960 $0 1954 $0 1922 $0
Low 2 < r ≤ 2.5 $5,000 0 $0 3 $15,000 19 $95,000
Moderate 2.5 < r ≤ 3 $20,000 0 $0 2 $40,000 11 $220,000
High 3<r $50,000 0 $0 1 $50,000 8 $400,000
Total event costs Cij $0 $190,000 $1,945,000
TABLE 4.36
Life-Cycle Costs for Future Three Blast-Level Events
(Blast-Hardened Building B)
(EQ1), j = 1 (EQ2), j = 2 (EQ3), j = 3
0.85 $0 0.125 $23,750 0.025 $48,625
Blast cost $72,375
And from Equations 4.47 through 4.49, we express exposure as
æ k = nv ö æ k = nh ö
E Hi = ç
ç å
è k =1
WTV ik (VR)k ÷ ç
÷ç
ø è k =1
å WTH ik ( HR)k ÷
÷
ø
(4.55)
Assuming a change in the kth vulnerability component, the resulting change in vulnerability is
¶VHi
= WTV (4.56)
¶(VR)k ik
The resulting exposure change is
¶E Hi ¶E Hi ¶(VR)k
= (4.57)
¶VH j ¶(VR)k ¶VH j
that is expressed as
¶E Hi æ æ = nh öö
= ç (WTV )ik ç
¶VH j ç
è
ç
è =1
å
WTH i ( HR) ÷ ÷ (WTV ) jk
÷÷
øø
(4.58)
We reach the desired MH interaction between exposures as
æ æ
å ö
WTH i ( HR ) ö÷ ÷ ( WTV ) jk
= nh
¶E Hi ç ( WTV )ik ç
è øø
= è
=1
(4.59)
¶E H j æ æ
å ö
WTH j ( HR ) ö÷ ÷ ( WTV ) jk
= nh
ç ( WTV ) jk ç
è è =1 øø
From Equation 4.59, ¶E Hi /¶E Hi = 1.0 and ¶E Hi /¶E H j = 0.0 if (WTV )ik = 0.0.
Equation 4.59 is useful when a single parameter affecting vulnerability changes. Similar devel-
opment can be made if more than one parameter changes and if those changes are related. In such a
situation, the development of the interaction expressions becomes more involved. The final interac-
tion terms will be dependent on the actual changes in the parameters such that
æ
f ç D ( WTV )ik , D ( WTV ) jk , æç å ö
WTH i ( HR ) ö÷ ÷
= nh
¶E Hi è øø
= è
=1
(4.60)

¶E H j æ
f ç D ( WTV ) jk , D ( WTV ) jk , æç å ö
WTH j ( HR ) ö÷ ÷
= nh
è è =1 øø

with D(WTV )ik a vector containing all actual kth changes, k = 1,2,…,nv. Similar expressions can be
made if the hazard-related parameters, (WTH )ik, are changed producing D(WTH )ik with k = 1,2,…,nh.
We leave these developments to the reader as an exercise.
4.5.4 case study 4.6: Mh exposure InteractIon For Mass transIt statIons
We continue Case Study 4.1 by exploring MH exposure interactions for mass transit stations. We
are now interested in how exposure to the five hazards of that case study interacts as we developed
in the last section. This will entail adding the threat matrices to the vulnerability matrices we have
used in Case Study 4.1. We now develop the vulnerability as well as the threat weights, [WT V] and
[WTH], as shown in Tables 4.37 and 4.38, respectively. The five hazards of interest are interacting
with the nine vulnerability issues and the seven threat issues for that particular mass transit station
under consideration. Details of these issues can be found in US-DHS (2011b). Let us assume that
for the mass transit station under consideration, the assessed vulnerability vector {VR} and threat
vector {TR} are as shown in Tables 4.39 and 4.40, respectively. We can now apply Equation 4.59 to
obtain the MH exposure interaction matrices for each of the nine vulnerability issues as shown in
Tables 4.41 through 4.49. The average MH exposure interaction matrix can be obtained by averag-
ing the nine matrices as shown in Table 4.50.
These matrices can reveal how exposures of the mass transit station under consideration due
to different hazards do interact. For example, on average, Table 4.50, the subjective interaction
between fire and manmade hazards is 1.624, while the interaction between wind and flood is 1.518.
These exposure interaction matrices are different from the vulnerability matrices, as expected,
since they accommodate the additional information of the threats to the mass transit station under
consideration. For a more accurate and higher-resolution analysis of this type, the reader should
consult US-DHS (2011b).
4.5.5 decIsIon-based exposure InteractIons

Decision-based MH exposure is different from physical-based MH exposure in which it explores
the changes in exposure due to ith hazard, E Hi , to a decision regarding the jth hazard, as per the
decision MH theory. Let us assume that a particular decision, D = d, is made that would affect {VR}
and/or {HR}, so that
{VR} = {VR ( d )} (4.61)
and
{HR} = {HR ( d )} (4.62)

TABLE 4.37
Vulnerability Fire Operational Nonstructural Physical
Issues Site Architecture Structure Ventilation Protection Systems Systems Security IT Issues
Manmade 13 4 8 12 8 12 2 38 4
Earthquakes 6 11 39 4 12 7 9 3 9
Fire 5 10 20 9 33 10 7 2 6
Floods 44 3 4 9 4 18 5 5 8
Wind 50 3 10 4 5 20 5 2 2
Vulnerability, Reliability, and Exposure: Current Design Paradigms
115
TABLE 4.38
Threat Weights Matrix [WTH]
Threat/Hazards Traffic Issues Redundancy Attractiveness Earthquake Fire Floods Wind
Manmade 14 11 43 1 21 4 4
Earthquakes 0 0 0 100 0 0 0
Fire 9 13 13 15 19 15 15
Floods 4 0 0 0 0 74 22
Wind 0 0 0 0 0 23 77
TABLE 4.39
Vulnerability Assessment Vector {VR}
Site 3
Architecture 9
Structure 8
Ventilation 8
Fire protection 7
Operational systems 3
Nonstructural systems 2
Physical security 2
IT issues 5
TABLE 4.40
Threat Assessment Vector {TR}
Traffic issues 4
Redundancy 8
Attractiveness 8
Earthquake 7
Fire 5
Floods 3
Wind 7
TABLE 4.41
Multihazard Exposure Matrix for Site
Manmade 1.000 1.923 3.109 0.492 0.283
Earthquakes 0.520 1.000 1.617 0.256 0.147
Fire 0.322 0.619 1.000 0.158 0.091
Floods 2.033 3.911 6.322 1.000 0.576
Wind 3.532 6.793 10.981 1.737 1.000
TABLE 4.42
Multihazard Exposure Matrix for Architectural
Manmade 1.000 0.339 0.418 2.411 1.388
Earthquakes 2.946 1.000 1.232 7.103 4.089
Fire 2.392 0.812 1.000 5.767 3.320
Floods 0.415 0.141 0.173 1.000 0.576
Wind 0.720 0.245 0.301 1.737 1.000
TABLE 4.43
Multihazard Exposure Matrix for Structural
Manmade 1.000 0.183 0.428 2.893 0.833
Earthquakes 5.475 1.000 2.342 15.840 4.560
Fire 2.337 0.427 1.000 6.762 1.947
Floods 0.346 0.063 0.148 1.000 0.288
Wind 1.201 0.219 0.514 3.474 1.000
TABLE 4.44
Multihazard Exposure Matrix for Ventilation
Manmade 1.000 2.546 1.372 2.146 3.123
Earthquakes 0.393 1.000 0.539 0.843 1.227
Fire 0.729 1.856 1.000 1.564 2.277
Floods 0.466 1.186 0.639 1.000 1.455
Wind 0.320 0.815 0.439 0.687 1.000
TABLE 4.45
Multihazard Exposure Matrix for Fire Protection
Manmade 1.000 0.599 0.257 2.893 1.666
Earthquakes 1.669 1.000 0.429 4.830 2.781
Fire 3.896 2.333 1.000 11.271 6.489
Floods 0.346 0.207 0.089 1.000 0.576
Wind 0.600 0.360 0.154 1.737 1.000
TABLE 4.46
Multihazard Exposure Matrix for Operational Systems
Manmade 1.000 1.527 1.254 1.085 0.625
Earthquakes 0.655 1.000 0.821 0.710 0.409
Fire 0.797 1.218 1.000 0.865 0.498
Floods 0.922 1.408 1.156 1.000 0.576
Wind 1.601 2.445 2.008 1.737 1.000
TABLE 4.47
Multihazard Exposure Matrix for Nonstructural
Manmade 1.000 0.235 0.366 0.772 0.500
Earthquakes 4.255 1.000 1.557 3.287 2.126
Fire 2.734 0.642 1.000 2.112 1.366
Floods 1.295 0.304 0.474 1.000 0.647
Wind 2.001 0.470 0.732 1.546 1.000
TABLE 4.48
Multihazard Exposure Matrix for Physical Security
Manmade 1.000 12.729 24.387 12.875 20.821
Earthquakes 0.079 1.000 1.916 1.011 1.636
Fire 0.041 0.522 1.000 0.528 0.854
Floods 0.078 0.989 1.894 1.000 1.617
Wind 0.048 0.611 1.171 0.618 1.000
TABLE 4.49
Multihazard Exposure Matrix for IT
Manmade 1.000 0.407 0.732 0.805 2.082
Earthquakes 2.455 1.000 1.796 1.976 5.112
Fire 1.367 0.557 1.000 1.100 2.846
Floods 1.243 0.506 0.909 1.000 2.587
Wind 0.480 0.196 0.351 0.386 1.000
TABLE 4.50
Average Multihazard Exposure Matrix
Manmade 1.000 2.277 3.591 2.930 3.480
Earthquakes 2.050 1.000 1.361 3.984 2.454
Fire 1.624 0.998 1.000 3.347 2.187
Floods 0.794 0.968 1.312 1.000 0.989
Wind 1.167 1.350 1.850 1.518 1.000
From Equation 4.47,

¶E Hi ¶H Hi ¶V
= VHi + H Hi Hi (4.63)
¶d ¶d ¶d
From Equations 4.48, 4.49, 4.61, and 4.62,

= nv
¶VHi
¶d
= å WT
=1
V
i
æ ¶(VR(d )) ö
ç
è ¶d ÷
ø
(4.64)
and
¶H Hi
= nh
(
æ ¶ HR ( d ) ) ö÷
¶d
= å WT
=1
H ç
ç
i è
¶d ÷
ø
(4.65)
From Equations 4.63 through 4.65, we express changes of exposure due to ith hazard, E Hi to a deci-
sion D = d as
¶E Hi é
= nv ù é = nh (
æ ¶ HR ( d ) ) ö÷ ùú
¶d
=ê å
êë =1
WTV (VR ( d )) ú êêå WT
úû =1
H ç
ç
i è
¶d ÷ ú
ø û
i ë
é = nh ù é = nv (
æ ¶ VR ( d ) ) ö÷ ùú
+ê å
êë =1
WTH HR ( d ) ( ) ú êêå WT

úû =1
V ç
ç
i è
¶d ÷ ú
ø û
(4.66)
i ë
If the decision, D = d, is concerned with a particular hazard, j, such that

D = dj (4.67)
we obtain the required MH exposure decision interaction as
¶E Hi é
= nv ù é = nh (
æ ¶ HR ( d j ) ) ö÷ ùú
¶d j
=ê å
êë =1
WTV
i
(VR ( d )) úú êêå WT
j

û ë =1
H
ç
ç
i è
¶d ÷ ú
ø û
é = nh ù é = nv (
æ ¶ VR ( d j ) ) ÷ö ùú
+ê å
êë =1 i
(
WTH HR ( d j ) ) úú êêå WT

û ë =1
V
ç
ç
i è
¶d ÷ ú
ø û
(4.68)
In matrix notations, the MH decision exposure matrix [E]DECISION with ¶E Hi /¶d j in its ith row and
jth column is expressed as
é ¶H ù é ¶V ù
éë E ùû DECISION = {VR} Ä ê ú + {VH } Ä ê ¶d ú (4.69)
ë ¶d û ë û
with the ith row and jth column cell of [∂H/∂d] is
é = nh (
æ ¶ HR ( d j ) ) ÷ö úù
ê
å
ê =1
WTH ç
ç
i è
¶d ÷ ú
ø û
ë
and the ith row and jth column cell of [∂V/∂d] is
é = nv (
æ ¶ VR ( d j ) ) ö÷ ùú
ê
å
ê =1
WTV ç
ç
i è
¶d ÷ ú
ø û
ë
The notation ⊗ indicates a vector-by-vector multiplication.

Similar to Sections 4.2.4 and 4.2.5, the key in this process is to estimate both ∂(VR(dj))/∂d and
∂(HR(dj))/∂d, which indicate how issues of different vulnerability and threats change according to
the decision dj. An experienced practitioner should do this step. Note that the values of ∂(VR(dj))/∂d
and ∂(HR(dj))/∂d are all relative and can be positive (which indicate improvements in vulnerability)
or negative (which indicates a reduction in vulnerability).
4.5.6 case study 4.7: decIsIon-based Mh exposure InteractIon For tunnels

We apply the earlier method to the same tunnel example of Section 4.2.5 to obtain its MH decision
exposure interaction matrix of Equation 4.69. The vulnerability weight matrix [WT V] and the threat
weight matrix [WTH] of the tunnel under interest are in Tables 4.12 and 4.51, respectively. We use
the same four hazards (manmade, earthquakes, fire, and flood) as in Section 4.2.5 with 10 issues that
would affect the tunnel vulnerability and 8 issues that would affect the tunnel threats. The assess-
ment vectors of the vulnerability {VR} and threats {TR} of the tunnel are in Tables 4.13 and 4.52,
respectively. The changes of vulnerability and threats, ∂(VR(dj))/∂d and ∂(HR(dj))/∂d, according to
decisions di are also estimated and are displayed in Tables 4.13 and 4.52. The desired MH decision
matrix is computed from Equation 4.69 and is shown in Table 4.53. The ith row and jth column of
this matrix indicate the changes of exposure of the tunnel to the jth hazard due to a decision made in
TABLE 4.51
Threat Weights Matrix [WTH]
Peak Function
Threat/ No. of within the
Hazards Locality Users Attractiveness Significance Region Earthquakes Fire Floods
Manmade 13 11 40 5 6 1 20 4
Earthquakes 0 0 0 12 15 73 0 0
Fire 9 13 13 7 8 15 19 15
Floods 3 0 0 21 18 0 0 59
TABLE 4.52
Threat Assessment Vector {TR} and Decision
Matrix ∂(HR(dj))/∂d
∂(HR(dj ))/∂d: Decisions in Regard to a
Hazard, j, Will Affect HR as Follows, Used
Only for i Type = 1
HR Manmade Earthquakes Fire Floods
Locality 9 6 3 0 8
Peak no. of users 8 2 1 9 3
Attractiveness 8 7 1 3 0
Significance 7 4 2 0 0
Function within 7 2 1 8 3
the region
Earthquake 2 0 10 0 0
Fire 3 2 2 10 3
Floods 2 0 0 0 10
TABLE 4.53
Multihazard Decision Exposure Matrix [E]DECISION
Manmade 55 42 35 15
Fire 36 38 40 25
Floods 18 16 13 32
regard to the ith hazard. Thus, if the decision-maker makes a decision regarding exposure to earth-
quake, it is expected that the behavior of the tunnel due to exposure to floods will change by 4%.
Similar to the developments of Section 4.5.3 and due to the subjective nature of the process, only
the relative interaction values can be obtained from these matrices, thus limiting the usefulness of
the process. For more issues regarding exposure (vulnerability and threats), US-DHS (2011c) offers
more resolutions. The reader is invited again to build a higher-resolution vulnerability model from
that resource. Unlike the physical exposure interaction process of Section 4.5.3, these developments
do allow for negative terms in the interaction matrices. Since the signs of the interaction matrices
are controlled via the sign of parameters ∂(HR(dj))/∂d and/or ∂(VR(dj))/∂d, which can take a negative
value. A final observation on the interaction matrices is that they are not symmetric, which indicates
that different decisions will have unequal effects on the different hazards’ vulnerabilities, and they
offer only relative terms due to the underlying subjectivity of the weighted averages method. We
also note that the diagonals are not unity. Again, this is due to the underlying subjectivity of the
weights chosen in the matrix [WTH] and [WT V]. A normalized version of the matrices might be more
useful that can be obtained by dividing each row by its main diagonal term. The normalized matrix
is shown in Table 4.54. We end this section by comparing the MH decision vulnerability matrices,
Table 4.14 (not normalized) and Table 4.15 (normalized), and their exposure equivalents, Table
4.53 (not normalized) and Table 4.54 (normalized). We note that there are large differences. Since
exposure is more realistic as a basis than vulnerability, as it includes the combined effects of both
vulnerability and threats, it is more accurate to use exposure-based matrices for decision-making.
The added computational efforts are a small price to pay for the additional decision accuracy.
TABLE 4.54
Normalized Multihazard Decision Exposure Matrix [E]DECISION
Manmade 100 77 64 28
Fire 91 95 100 64
Floods 57 50 41 100
4.5.7 reMarks
The methods and examples we developed so far utilized a weighted averages approach. Such an
approach has two limitations:
1. Interactions between parameters in matrices are not accommodated.

2. The uncertainties are not accounted for.
Examples of exposure interactions will be revisited in the later chapters (Chapters 6 through 10)
with case studies where these two assumptions are released using GN processes.
APPENDIX 4A MATRICES OF CASE STUDY 4.3
TABLE 4.55
Load Amplitudes
Amplitudes of Five Hazards (lb)
Node
Number Direction DL LL W E B
1 x1 0.00 0.00 0.00 0.00 0.00
x2 0.00 0.00 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
2 x1 0.00 0.00 1.00 2.00 5.00
x2 −0.50 −0.10 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
3 x1 0.00 0.00 2.00 2.00 2.00
x2 −0.50 −0.25 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
4 x1 0.00 0.00 3.00 2.00 1.00
x2 −0.50 −0.35 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
5 x1 0.00 0.00 0.00 0.00 0.00
x2 −1.00 −2.00 0.00 2.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
6 x1 0.00 0.00 0.00 2.00 0.00
x2 −1.00 −3.00 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
7 x1 0.00 0.00 0.00 2.00 0.00
x2 −1.00 −4.00 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
8 x1 0.00 0.00 0.00 0.00 0.00
x2 0.00 0.00 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
9 x1 0.00 0.00 1.00 2.00 0.00
x2 −0.50 −0.75 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
10 x1 0.00 0.00 2.00 2.00 0.00
x2 −0.50 −0.50 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
11 x1 0.00 0.00 3.00 2.00 0.00
x2 −0.50 −0.25 0.00 0.00 0.00
θ3 0.00 0.00 0.00 0.00 0.00
TABLE 4.56
Hazard Displacements and Rotations
Displacements (in.)/ Rotations (rad) due to Five Hazards
Node
Number Direction DL LL W E B
1 x1 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
x2 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
θ3 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
2 x1 −2.35E−08 −3.36E−08 2.20E−04 2.90E−04 1.38E−04
x2 −8.33E−07 −1.44E−06 1.98E−06 2.61E−06 7.11E−07
θ3 −7.01E−08 −1.33E−07 −8.97E−07 −9.75E−07 −4.39E−07
3 x1 −1.32E−08 −1.02E−08 4.55E−04 5.73E−04 2.17E−04
x2 −1.39E−06 −2.58E−06 3.01E−06 3.74E−06 9.60E−07
θ3 −5.28E−08 −1.54E−07 −6.99E−07 −8.17E−07 −1.85E−07
4 x1 6.85E−08 4.50E−07 5.99E−04 7.22E−04 2.45E−04
x2 −1.67E−06 −3.24E−06 3.31E−06 4.05E−06 1.01E−06
θ3 −1.01E−07 −4.16E−07 −3.04E−07 −3.06E−07 −5.69E−08
5 x1 −4.91E−12 3.53E−08 2.20E−04 2.90E−04 1.37E−04
x2 −1.30E−05 −2.52E−05 9.84E−12 2.64E−05 −9.35E−08
θ3 −8.86E−15 −6.31E−10 4.32E−07 5.48E−07 2.13E−07
6 x1 −9.06E−12 1.14E−07 4.55E−04 5.73E−04 2.17E−04
x2 −1.20E−05 −3.41E−05 4.18E−11 −1.96E−06 1.50E−07
θ3 −3.34E−15 −7.39E−10 3.24E−07 3.67E−07 8.53E−08
7 x1 −1.09E−11 1.97E−07 5.99E−04 7.22E−04 2.45E−04
x2 −1.66E−05 −6.42E−05 9.35E−12 7.20E−07 1.91E−08
θ3 −1.81E−15 −6.54E−10 1.24E−07 1.24E−07 2.01E−08
8 x1 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
x2 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
θ3 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
9 x1 2.35E−08 1.04E−07 2.20E−04 2.90E−04 1.37E−04
x2 −8.33E−07 −1.67E−06 −1.98E−06 −2.05E−06 −7.11E−07
θ3 7.01E−08 1.32E−07 −8.97E−07 −1.30E−06 −4.37E−07
10 x1 1.32E−08 2.38E−07 4.55E−04 5.73E−04 2.16E−04
x2 −1.39E−06 −2.85E−06 −3.01E−06 −3.19E−06 −9.60E−07
θ3 5.28E−08 1.53E−07 −6.99E−07 −7.67E−07 −1.88E−07
11 x1 −6.85E−08 −5.65E−08 5.99E−04 7.22E−04 2.45E−04
x2 −1.67E−06 −3.47E−06 −3.31E−06 −3.50E−06 −1.01E−06
θ3 1.01E−07 4.15E−07 −3.04E−07 −3.16E−07 −5.73E−08
APPENDIX 4B MATRICES OF CASE STUDY 4.5
TABLE 4.57
Hazards Displacements and Rotations for the Improved Design
Displacements (in.)/ Rotations (rad) due to Five Hazards
Node Number Direction DL LL W E B
1 x1 0 0 0 0 0
x2 0 0 0 0 0
θ3 0 0 0 0 0
2 x1 −2.2E−08 −3.5E−08 0.000136 0.000179 8.4E−05
x2 −6.9E−07 −1.2E−06 1.63E−06 2.16E−06 5.86E−07
θ3 −4.4E−08 −8.4E−08 −6.6E−07 −7.3E−07 −3.2E−07
3 x1 −1.5E−08 −3.7E−08 0.000288 0.000361 0.000136
x2 −1.2E−06 −2.2E−06 2.5E−06 3.11E−06 7.97E−07
θ3 −3.2E−08 −9.2E−08 −5.2E−07 −6.1E−07 −1.4E−07
4 x1 6.88E−08 4.19E−07 0.000381 0.000459 0.000155
x2 −1.4E−06 −2.7E−06 2.76E−06 3.37E−06 8.44E−07
θ3 −6.7E−08 −2.8E−07 −2.3E−07 −2.4E−07 −4.5E−08
5 x1 −2.9E−12 2.9E−08 0.000136 0.000179 8.36E−05
x2 −9.1E−06 −1.8E−05 1.89E−11 1.83E−05 −9.6E−08
θ3 −1.2E−14 −5.3E−10 3.17E−07 4.02E−07 1.56E−07
6 x1 −5.4E−12 9.39E−08 0.000288 0.000361 0.000135
x2 −8.4E−06 −2.4E−05 −7.5E−12 −1.4E−06 1.51E−07
θ3 −2.7E−15 −6.2E−10 2.41E−07 2.73E−07 6.42E−08
7 x1 −6.3E−12 1.63E−07 0.000381 0.00046 0.000155
x2 −1.2E−05 −4.5E−05 1.9E−11 5.82E−07 1.7E−08
θ3 2.51E−15 −5.4E−10 9.43E−08 9.47E−08 1.55E−08
8 x1 0 0 0 0 0
x2 0 0 0 0 0
θ3 0 0 0 0 0
9 x1 2.21E−08 9.34E−08 0.000136 0.000179 8.31E−05
x2 −6.9E−07 −1.4E−06 −1.6E−06 −1.7E−06 −5.9E−07
θ3 4.45E−08 8.35E−08 −6.6E−07 −9.4E−07 −3.2E−07
10 x1 1.5E−08 2.25E−07 0.000288 0.000361 0.000135
x2 −1.2E−06 −2.4E−06 −2.5E−06 −2.6E−06 −8E−07
θ3 3.17E−08 9.04E−08 −5.2E−07 −5.8E−07 −1.4E−07
11 x1 −6.9E−08 −9.3E−08 0.000381 0.000459 0.000155
x2 −1.4E−06 −2.9E−06 −2.8E−06 −2.9E−06 −8.4E−07
θ3 6.7E−08 2.77E−07 −2.3E−07 −2.5E−07 −4.5E−08
REFERENCES
AASHTO. 2007. AASHTO LRFD Bridge Design Specifications, 4th Ed., American Association of State
Highway and Transportation Officials (AASHTO), Washington, DC.
ACI 318-05. 2005. Building Code Requirements for Structural Concrete and Commentary, American
Concrete Institute (ACI), Farmington Hills, MI.
Alipour, A., Shafei, B., and Shinozuka, M. Reliability-based calibration of load and resistance factors for design
of RC bridges under multiple extreme events: Scour and earthquake, Journal of Bridge Engineering,
18(5), May 1, 2013, 362–371.
American Society of Civil Engineers (ASCE). 1998. Minimum design loads for buildings and other structures.
ASCE 7-98.
ASCE 7-10. 2013. Minimum Design Loads for Buildings and Other Structures, Standard ASCE/SEI 7-10,
American Society of Civil Engineers (ASCE), Reston, VA.
Crosti, C., Duthinh, D., and Simiu, E. Risk consistency and synergy in multihazard design, Journal of
Structural Engineering, 137(8), August 1, 2011, 844–849.
Duthinh, D. and Simiu, E. Safety of structures in strong winds and earthquakes: Multihazard considerations,
Journal of Structural Engineering, 136(3), March 1, 2010, 330–333.
Elishakoff, I. 2004. Safety Factors and Reliability: Friends or Foes?, Kluwer Academic Publishers, Dordrecht,
the Netherlands.
ETABS. 2004. Extended 3D analysis of building systems, Version 8.5.2, Computers and Structures, Inc.,
Berkeley, CA.
Ettouney, M. 2001. Is seismic design adequate for blast?, Society of American Military Engineers National
Symposium on Comprehensive Force Protection, October 2001, Charleston, SC.
Ettouney, M., Alampalli, S., and Agrawal, A. Theory of multihazards for bridge structures, Journal of Bridge
Engineering, 2005, 1(3), 281–291.
Ettouney, M., Smilowitz, R., and Daddazio, R. 1998. Earthquake vs. blast—A short comparison, Proceedings
of Structural Engineers World Congress, San Francisco, CA.
Fenton, N. and Neil, M. 2013. Risk Assessment and Decision Analysis with Bayesian Networks, CRC Press,
Boca Raton, FL.
Freeman, G., Lee, J.C., and Ettouney, M.M. 2005. Multihazard design of tall buildings, Proceedings of the
Council on Tall Buildings and Urban Habitat, 7th World Congress, New York.
Hawk, H. 2003. Life cycle cost analysis, NCHRP Report 483, Transportation Research Board, Washington, DC.
Hayes, J.R., Woodson, S.C., Pekelnicky, R.G., Poland, C.D., Corley, W.G., and Sozen, M. 2005. Can strength-
ening for earthquake improve blast and progressive collapse resistance?, Journal of Structural
Engineering, 131(8), 1157–1177.
IBC-2000. 2000. International Building Code, International Code Council (ICC), Country Club Hills, IL.
Jorion, P. 2007. Value at Risk: The New Benchmark for Managing Financial Risk, McGraw-Hill, New York.
Rausand, M. and Hoyland, A. 2004. System Reliability Theory: Models, Statistical Methods, and Applications,
Wiley, Hoboken, NJ.
Sasani, M. and Sagiroglu, S. Progressive collapse of reinforced concrete structures: A multi-hazard perspec-
tive, ACI Structural Journal, 105(1), 2008, 96–103.
Smith, J.L., Brokaw, J.T., and Swatzell, S.R. 1998. Progressive collapse analysis and design guidance sum-
mary report documenting the literature review findings, ARA-TR-98-4854-1, Report submitted to
General Services Administration (GSA), Washington, DC.
US-DHS. 2011a. Integrated Rapid Visual of Screening of Buildings, Building and Infrastructure Protection
Series, Washington, DC.
US-DHS. 2011b. Integrated Rapid Visual Screening of Mass Transit Stations, Building and Infrastructure
Protection Series, Washington, DC.
US-DHS. 2011c. Integrated Rapid Visual Screening of Tunnels, Building and Infrastructure Protection Series,
Washington, DC.

Vulnerability Reliability and Exposure Current Design Paradigms

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Vulnerability Reliability and Exposure Current Design Paradigms

Transféré par

Droits d'auteur :

Formats disponibles

4 Vulnerability, Reliability,

and Exposure: Current

4.1.2 thIs chapter

4.2 VULNERABILITY OF INFRASTRUCTURES

4.2.2 physIcal Mh vulnerabIlIty InteractIons

Vulnerability, reliability, and exposure

Metrics or measure involved with how and/or

Subjective Objective Objective

Section 4.2 Section 4.3 Section 4.5

FIGURE 4.1 Overview of this chapter.

The resulting MH vulnerability interaction is

¶VHi ¶VHi ¶ (VR )k

This equation can be expressed as

We reach the desired MH interaction between vulnerabilities as

¶VHi ( WTV )ik

4.2.3 case study 4.1: Mh vulnerabIlIty InteractIons oF Mass transIt statIons

4.2.4 decIsIon-based vulnerabIlIty InteractIons

{VR} = {VR ( d )} (4.8)

If the decision, D = d, is concerned with a particular hazard, j, such that

we obtain the required MH vulnerability decision interaction as

with the ith row and jth column cell [∂V/∂d] is

4.2.5 case study 4.2: Mh vulnerabIlIty InteractIons oF tunnels

1. Interactions between parameters in matrices are not accommodated.

4.3 RELIABILITY-BASED DESIGN: MH VIEWPOINTS

4.3.2 theoretIcal background

Current design practice

Hazard 1 Hazard 2 Hazard N Initialize design

Multilimit states (MLS)

Apply to current design Modify current design

FIGURE 4.2 Current design practice.

Rel = f RELIABILITY ( Capacity, Demand ) (4.13)

Capacity ³ Demand (4.14)

The failure limit state* of the system is reached when

Capacity = Demand (4.15)

Rel STR Î Rel SYS (4.16)

Reli = f ( DCi ) (4.19)

with the requirement

DCi £ 1.0 (4.20)

Capacity i = fi ( Nominal _ Capacity)i (4.22)

4.3.3 Mh structural relIabIlIty MetrIcs

4.3.4 Mh InteractIon through dlc (Mh-dlc)

The ith component in {DC} is DCi. Also,

aij = {DC}Ti {DC} j (4.27)

MHIC = MHI ij = (4.28)

é1.000 0.104 1.000 ù

Now, we can make four useful observations:

Let us make a slight change in [Γ] as shown in Table 4.18:

é1.000 0.000 0.485 ù

4.3.5 Mh desIgn usIng desIgn loadIng condItIons (Mh-dlc)

4.3.6 Mh-desIgn eFFIcIency MatrIx

{U} j = [ Á]ni {P} j (4.32)

u jk = {P} j ëé Áùû ni { A}k

({P} éëÁùû { A}k )

4.3.6.2 Case Study 4.4: Simple Analytic Expressions of MH-DEM

éë Áùû = Com éë Áùû (4.37)

with the compliance of the system material is

({P} éëÁùû { A}k )

4.4 RELIABILITY-BASED DESIGN: MH SERIAL

4.4.2 applIcatIon oF Mpt: WInd and progressIve collapse oF tall buIldIngs

FIGURE 4.3 Floor layout of Building A.

FIGURE 4.4 Elevation of typical braced frame.

4.4.3 InteractIon oF Mh buIldIng systeMs: WInd and seIsMIc hazards

FIGURE 4.5 Floor plan of Building B.

FIGURE 4.6 Eccentrically braced frame of Building B.