Report - PEER - University of California, Berkeley

Report - PEER - University of California, Berkeley Report - PEER - University of California, Berkeley

peer.berkeley.edu
from peer.berkeley.edu More from this publisher
12.07.2015 Views

P(C LVCC i |IM )1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]P(C LVCCi |IM )1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]P(C|IM)1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]Figure 5. Different steps of estimation of the probability of collapse of thesystem conditioned on IM .3.4 Repair or Replacement Costs EstimationFor each component loss functions are developed to estimate the cost of repair or costto replace each component. Loss functions are functions that provide information onthe probability of exceeding a certain level of repair or replacement cost given that thecomponent is in the damage measure, DM. Examples of these functions are given inAslani and Miranda (2004a).3.5 Modeling Correlation between Losses in Individual ComponentsEstimation of the correlation between losses in individual components requiresinformation on the correlation at three different levels; EDP | IM level, DM | EDPlevel and DV | DM level. The correlation at the response level, EDP | IM is estimatedbased on the results from nonlinear response history analyses. The correlation at thedamage level, DM | EDP, is mathematically modeled by categorizing componentsinto certain groups in terms of their damageability and estimating the joint probabilityof two components being at different damage states conditioned on the level ofdeformation each of them is subjected to. The correlation at the repair cost level, DV |DM, is estimated from the information on the correlation between different tasksrequired to repair the component.PFA 3, PFA roofρ EDPi,EDPj | IM, NC1.000.801.00.60 IDR 1, PFA 30.400.50.20IDR 1, IDR 30.00.000 10 20 30 40 50IM [S d (cm)]P(DM ki,DMkj|EDPi,EDP j)0.35Component iDM3| EDP iDM2| EDPi0.00.050.200.3DM1| EDP iComponentj0.0DM 1| EDPjDM 2| EDPjDM 3| EDPj0.00.100.0ρ (Li,Lj|DMki,DMkj,NC)1.00.50.0ColumnL | | DM 30.29L | | DM21.000.291.00L | | DM1Beam-column1.000.29L | DM 1 1Figure 6. Variations of the required parameters to estimate the correlation oflosses in individual components.L | DM2157

E [ L T| IM ]$10 M$8 M$6 M$4 Mσ [ L T| IM ]$10 M$8 M$6 M$4 MCorrelat edNon-correlat edν ( L T > $ )$2 M$2 M0.001(a) (b) (c)$0 M$0 M0.00010 20 40 60 80 100 0 20 40 60 80 100 $ 0 $ 4 $ 8 $ 12 $ 16IM [ S d(cm) ]IM [ S d(cm) ]L T [ million $ ]Figure 7. (a) Expected loss at different levels of intensity, (b) dispersion ofloss at different levels of intensity, (c) building loss curve.Figure 6 presents examples of each of the correlation at each of the above threelevel. Figure 6a shows how the correlation between different types of EDP varies asthe ground motion intensity increase. Shown in Figure 6b is an example of the jointprobability distribution of two components being at different damage states. Figure 6cshows the correlation between repair costs for a column and a beam-columnconnection.3.6 Building Loss EstimationFigure 7a presents the variations of the expected loss at different levels of intensity,E[L T |IM], estimated for the case study building. It can be seen that for this buildinglosses rapidly increase at small levels of ground motion intensity. Figure 7b presentsthe variations of the dispersion of the loss of the building with increasing level ofground motion intensity for two cases: when losses in individual components areassumed to be correlated and when they are assumed non-correlated. It can be seenthat correlation has significant effects on the uncertainty of the loss. For example, atS d =20 cm assuming that the losses are uncorrelated leads to an underestimation of25% of the dispersion of the loss.The loss curve for the case study building is shown in Figure 7c where it can beseen that losses smaller than $1,000,000 have relatively high mean annual frequenciesof exceedance.4. LOSS DEAGGREGATIONSimilarly to seismic hazard deaggregation (McGuire, 1995) building losses can alsobe disaggregated. In particular, it is interesting to investigate the ground motionintensities that most contribute to expected annual losses in a building. Figure 8provides three examples of loss deaggregation. Figure 8a presents the contribution ofcollapse and non-collapse expected loss to the total loss at different levels of intensity.It can be seen in the figure that at small levels of intensity, (S d

P(C LVCC i |IM )1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]P(C LVCCi |IM )1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]P(C|IM)1.00.80.60.40.20.00 20 40 60 80 100IM [ S d (cm) ]Figure 5. Different steps <strong>of</strong> estimation <strong>of</strong> the probability <strong>of</strong> collapse <strong>of</strong> thesystem conditioned on IM .3.4 Repair or Replacement Costs EstimationFor each component loss functions are developed to estimate the cost <strong>of</strong> repair or costto replace each component. Loss functions are functions that provide information onthe probability <strong>of</strong> exceeding a certain level <strong>of</strong> repair or replacement cost given that thecomponent is in the damage measure, DM. Examples <strong>of</strong> these functions are given inAslani and Miranda (2004a).3.5 Modeling Correlation between Losses in Individual ComponentsEstimation <strong>of</strong> the correlation between losses in individual components requiresinformation on the correlation at three different levels; EDP | IM level, DM | EDPlevel and DV | DM level. The correlation at the response level, EDP | IM is estimatedbased on the results from nonlinear response history analyses. The correlation at thedamage level, DM | EDP, is mathematically modeled by categorizing componentsinto certain groups in terms <strong>of</strong> their damageability and estimating the joint probability<strong>of</strong> two components being at different damage states conditioned on the level <strong>of</strong>deformation each <strong>of</strong> them is subjected to. The correlation at the repair cost level, DV |DM, is estimated from the information on the correlation between different tasksrequired to repair the component.PFA 3, PFA ro<strong>of</strong>ρ EDPi,EDPj | IM, NC1.000.801.00.60 IDR 1, PFA 30.400.50.20IDR 1, IDR 30.00.000 10 20 30 40 50IM [S d (cm)]P(DM ki,DMkj|EDPi,EDP j)0.35Component iDM3| EDP iDM2| EDPi0.00.050.200.3DM1| EDP iComponentj0.0DM 1| EDPjDM 2| EDPjDM 3| EDPj0.00.100.0ρ (Li,Lj|DMki,DMkj,NC)1.00.50.0ColumnL | | DM 30.29L | | DM21.000.291.00L | | DM1Beam-column1.000.29L | DM 1 1Figure 6. Variations <strong>of</strong> the required parameters to estimate the correlation <strong>of</strong>losses in individual components.L | DM2157

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!