The 3rd International Conference on Earthquake Engineering and Disaster Mitigation 2016 (ICEEDM-III 2016)

Loss Optimization Seismic Design

Rajesh P. Dhakal1, Sandip K. Saha1
1
Department of Civil and Natural Resources Engineering,
University of Canterbury, Christchurch, New Zealand.

1. INTRODUCTION
Present performance based seismic design philosophy primarily aims to restrict the structural
response within some prescribed limits such that the structure satisfies the desired performance.
Considering the performance of modern earthquake resistant buildings observed in recent
earthquakes, it could be broadly concluded that the collapse prevention objective is satisfactorily
achieved in most cases. Nevertheless, the financial consequences due to the inevitable damage to
structural and non-structural components of such earthquake resistant modern buildings have been
found to be substantially high, and not acceptable to the stakeholders. Merely achieving the code
based performance objectives (such as life safety and collapse prevention) does not satisfy the
stakeholders concern when it comes to financial losses. To address this issue, FEMA P-58 (FEMA
2012) has put together a methodology based on the framework developed by the Pacific Earthquake
Engineering Research (PEER) centre (Moehle and Deierlein 2004) for seismic loss assessment of
structures.

Considering the enormous attention paid towards the financial loss of earthquake induced
damages, the next generation seismic design is expected to be based on the criteria that minimize the
financial losses in addition to ensuring life safety. Earlier, Krawinkler et al. (2004) presented how the
acceptable monetary loss can be viewed as the „design target‟ at the design level within the
framework of performance based design. They further pointed out that the relationships between the
monitory loss and the engineering demand parameters (EDPs) are to be developed extensively for
efficient application of the loss-targeted design. Later, Dhakal (2010) proposed the concept of the loss
optimization seismic design (LOSD). He conceptually presented how the widely used performance
criteria, such as immediate occupancy or the collapse prevention, can be satisfied in the LOSD
framework even without explicitly defining and checking for those performance criteria during the
design. The need for easily understandable (for the nontechnical stakeholders such as owner, insurer
or common user) performance objectives is growing rapidly after each earthquake occurrence; and
minimisation of seismic losses (financial or otherwise) has emerged as an easily understandable and
widely acceptable design criteria. Therefore, the concept of LOSD needs to be developed further to
cater a seismic design approach that is acceptable to and meets the expectation of all stakeholders.

2. FRAMEWORK FOR LOSD

LOSD, like any other design approach, also works on the motto capacity greater than
demand. Nevertheless, the demand is specified and capacity is calculated in terms of likely losses
(dollars, downtime and injury). Once LOSD is fully developed, the demands pertinent to LOSD can
be specified in the form of tolerable impact levels in national standards/guidelines. Nevertheless, if
the specified tolerances are not acceptable to risk-averse stakeholders, they would always have the
freedom to specify stricter limits of expected losses they are prepared to accept in different levels of
earthquake during the design life of their buildings. All building configurations should satisfy the life
safety requirement for which the estimated losses are less than their tolerable limits. A generic loss
assessment methodology is key to the successful implementation of LOSD. The design process is
schematically presented in Figure 1.
As introduced by Dhakal (2010), the performance is to be measured in RDI format where, R
denotes the expected repair cost, D denotes the expected downtime, and I denotes the injury vector
that include minor/moderate/major injuries and causalities. Evidently, these terms, i.e. repair cost,
downtime and injuries/causalities, are more familiar to the nontechnical stakeholders instead of the
currently used performance criteria in engineering terms such as damage and deformation measures.
Table 1 presents an example of performance requirements in RDI format for buildings in different
levels of seismicity. The design objectives are to be defined in terms of the acceptable losses during
the design life of a structure for different levels of seismicity. Different performance objectives can be
selected based on the use of the buildings, such as residential, commercial, emergency etc.
Computation of the expected losses in RDI format is a crucial and challenging step in LOSD.
Therefore, the methodology to estimate the expected losses needs to be simplified.

Fig. 1. Schematic flow diagram of loss optimization seismic design (LOSD) process
 Table 1. Example of performance requirements in RDI format for buildings in different levels of
seismicity (Dhakal 2010)

3. METHODOLOGY
During the last decade, significant amount of research has been carried out on seismic loss
assessment framework (Moehle and Deierlein 2004; Aslani and Miranda 2005; Bradley et al. 2009).
The framework developed by the Pacific Earthquake Engineering Research (PEER) centre to
interrelate seismic hazard, structural response, damage and losses is presented in Figure 2. As per
FEMA P-58 methodology, the engineering demand parameters (EDPs) are computed from seismic
response analysis for a given seismic hazard. For loss estimation, this has to be followed by
probabilistic computation considering the interrelationships between the EDPs, damage measure
(DM) and the decision variable (DV). The probabilistic calculations can be very cumbersome and
should be performed using computer based performance assessment calculation tools (such as PACT).
However, the selection of appropriate interrelationships and details of the component inventory
require a reasonable understanding and expertise on the probabilistic framework, which may not be
suitable for routine design practice.

Fig. 2. Schematic illustrations of the key steps in the PEER loss estimation methodology

An important simplification can be achieved in the loss estimation process by condensing the
damage measure; in other words by expressing the expected seismic loss directly with respect to the
EDPs, such loss-EDP relationships are commonly termed as loss functions. Ramirez and Miranda
(2009) presented building specific story level loss functions assuming typical values of the component
density (or quantity) at a particular story and cost of the component. However, these loss functions do
not account for the variation of the component density in different buildings. As a result, the predicted
seismic loss may be significantly over/under-estimated when the component distributions differ from
the assumed values. A refined approach would be to consider the component distributions from
extensive construction data collection for a typical building category, then probabilistically introduce
the repair/replacement cost for different levels of damages to arrive at the floor level expected loss for
that component. These loss functions need to be normalized as a percentage of the total building cost
for more generic use. To develop such generalized relationships, the information required are: (i)
database of structural and non-structural components that primarily contributes to the cost of a
particular class of building, (ii) database of component fragility functions, and (iii) repair/replacement
costs for different damage states for each component. Expected loss due to a component at a floor
level for a given value of EDP (Lc /EDP) can be computed deterministically as,

where Ac is the total area/quantity of the component in the floor (i.e. component density); lc Di is the
repair cost per unit area of the component for the damage state Di; nDS is the number of discrete
damage states considered in the component fragility; and P(Di | EDP) is the probability of damage
being in the ith damage state for the given EDP value. Here, lc Di is generated using the assigned
probability distributions based on the collected cost data, whereas P(Di|EDP) can be computed from
the fragility functions. Now, Ac needs to be defined in terms of a probability distribution from
extensive data collection. For a given EDP, this calculation is repeated with different randomly
generated values of the inputs using Monte Carlo (MC) simulations to obtain the expected loss at the
given EDP. This process is then repeated for different values of EDP; and finally the expected
component loss functions are generated in the form of loss vs. EDP plots.

4. EXAMPLE GENERIC LOSS FUNCTIONS FOR TYPICAL COMPONENTS

Extensive construction data are necessary to develop the probability distributions of the components
for a typical building category. Dhakal et al. (2016) presented generic loss functions of two important
non-structural components (suspended ceiling and drywall partition), for typical reinforced concrete
(RC) office buildings based on large construction data collected from Christchurch CBD, New
Zealand. The component level loss functions, normalized with respect to the total building cost, are
presented in Figure 3. The generated normalized loss functions suggest that the loss from suspended
ceilings and partition damage in RC office buildings could be up to 4.5% and 3.4% of the total cost of
the building, respectively. The usefulness of the developed loss functions were compared with the
expected losses calculated using actual suspended ceiling distributions from five floors of a case study
building, and drywall partition from three case study buildings.

Figure 3: Generalized loss functions for suspended ceilings and partition (Dhakal et al. 2016)
5. FLOOR LEVEL LOSS FUNCTION
It can be noted that the damage of almost all the structural and non-structural components can be
associated either with peak inter story drift ratio (IDR) or with peak floor acceleration (PFA) at a
particular floor level. The components that are attached to the roof slab are sensitive to the PFA, and
the components that are connected to the frames or walls, or placed between two floors are sensitive
to the IDR, in general. Therefore, the total losses are divided corresponding to two different categories
of EDPs, i.e. expected losses associated with components sensitive to IDR, and expected losses
associated with components sensitive to PFA. Once the loss-EDP relationships are developed for all
major contributing components of a particular building category, the storey level loss functions can be
generated. Farokhnia (2013) proposed to assign some judgmental weightage factors to different
building components based on their quantity and probable contribution towards the loss at a particular
floor level. Mathematically, the expected loss per square meter of floor area due to drift sensitive
components E[Lidr | IDR  idr] and due to acceleration sensitive components E[Lpfa | PFA  pfa]
are expressed as,

where wc(i) is the weightage factor for ith component; Nidr and Npfa are the number of drift sensitive
and acceleration sensitive components, respectively; and E[li | IDR = idr] and E[li | PFA = pfa] are to
be computed from the MC simulations using Equation 1. The story level losses can be combined to
obtain the total expected loss in terms of repair/replacement cost. These losses should meet the
prescribed limit given in the RDI format. It can be noted here that the structural analysis at the specific
level of seismic intensity is not included in the process. Nevertheless, considerable computational
effort is required for the traditional structural analysis (such as Incremental Dynamic Analysis IDA),
which needs to be simplified substantially to make the LOSD framework more efficient and widely
acceptable to the industry.

6. SUMMARY AND RESEARCH NEEDS

There is a growing demand from non-engineering stakeholders of building stocks to gauge the
building performance during earthquakes not only against the life safety and collapse prevention
criteria, but also in terms of financial losses due to repair, downtime and injury (or causalities). A
framework for a loss targeted seismic design of structures, called Loss Optimisation Seismic Design
(LOSD), has been envisaged, wherein; the structural engineer can estimate the expected seismic loss
for a building in its lifetime during the design phase without involving any rigorous computational
effort or time, and revise the structural design to meet the prescribed loss limits. There exists
significant research needs to fully develop and successfully implement the LOSD framework.
Combining the fragility functions and repair/replacement costs of each component, generalized loss
functions can be developed for all major contributing components in various building categories. It is
necessary to derive the appropriate weightage factors for the components from the collected
construction inventories and loss deaggregation analyses. Once these two tasks are completed, floor
level generalised loss functions can be developed which will tremendously simplify the estimation of
direct seismic loss in buildings. The next step then would be to extend the procedures to incorporate
downtime and injury in line with the loss estimation due to the repair/replacement cost.

REFERENCES

[1] Aslani H. and Miranda E. (2005), “Probabilistic earthquake loss estimation and loss
disaggregation in buildings”, Blume Center Technical Report 157, Stanford University,
California, USA.
[2] Bradley B.A., Dhakal R.P., Cubrinovski M., MacRae G.A. and Lee D.S. (2009), “Seismic loss
 estimation for efficient decision making”, Bulletin of New Zealand Society for Earthquake
Engineering, 42(2), 96-110.
[3] Dhakal R.P. (2010), “First step towards loss optimisation seismic design (LOSD)”, 3 rd Asia
Conference on Earthquake Engineering (ACEE 2010), Bangkok, Thailand.
[4] Dhakal R.P., Pourali A. and Saha S.K. (2016), “Simplified seismic loss functions for suspended
ceilings and drywall partitions”, Bulletin of the New Zealand Society for Earthquake Engineering
(NZSEE), 49(1), 65-80. (In Press)
[5] Farokhnia K. (2013), “Nonstructural vulnerability functions for building categories”, University of
Colorado, Boulder, Colorado, USA.
[6] FEMA (2012), “Next-generation methodology for seismic performance assessment of buildings”,
Applied Technology Council for the Federal Emergency Management Agency, Report No.
FEMA P-58, Washington, D.C., USA.
[7] Krawinkler H., Zareian F., Medina R.A. and Ibarra L. (2004), “Contrasting performance-based
design with performance assessment”, International workshop on performance-based seismic
design-concepts and implementation, Bled, Slovenia.
[8] Moehle J.P. and Deirelein G.G. (2004), “A framework methodology for performance-based
earthquake engineering”, 13th World Conference on Earthquake Engineering, Vancouver, BC,
Canada.
[9] Ramirez C.M. and Miranda E. (2009), “Building-specific loss estimation methods & tools for
simplified performance-based earthquake engineering”, Blume Center Technical Report 171,
Stanford University, USA.