298 Reliability Estimation for Reinforced Concrete Bridges Connected in a Network (PSAM-0371) Available to Purchase

Transportation networks are critical lifelines and their functionality after an earthquake is of primary importance for life safety and economic recovery of a community, yet they can be severely damaged during earthquakes. Reinforced concrete bridges are a particularly vulnerable segment of these networks. Severe damage in past earthquakes (e.g., Northridge, 1994 and Kobe, 1995) have caused prolonged disruption to transportation systems. Past approaches for estimating traffic disruptions due to earthquakes have relied on computationally expensive simulations and have tended to overestimate post-earthquake traffic volumes [1]. Some potential causes of this overestimation are (i) lack of sound fragility estimates for bridges and (ii) traffic models that may not accurately reflect post-disaster conditions (see [1]). We make use of improved fragility estimates for reinforced concrete bridges from Gardoni et al. [2, 3] and take a simplified approach for estimating network reliability to overcome some of the limitations discussed above.

First, we consider improved predictive fragility estimates based on unbiased probabilistic capacity and demand models. Unlike the point estimates of fragilities that are typically used, predictive fragilities incorporate the epistemic uncertainties inherent in the probabilistic capacity and demand model parameters. The probabilistic models are developed using a Bayesian formulation that (i) incorporates all available sources of information, including basic rules of mechanics, state-of-the-art deterministic models and analysis procedures, laboratory and field data, and expert judgment and (ii) properly accounts for all the prevailing uncertainties, including model errors arising from an inaccurate model form or missing variables, measurement errors, and statistical uncertainty.

Second, the traffic model is removed and the simulation is simplified. We use block sampling-based simulation to estimate network connectivity rather than traffic volume. This eliminates the difficulties associated with post-disaster origin-destination modeling that may account for some of the overestimation in earlier models. It also eliminates the assumption that drivers have perfect information, overcoming another possible source of overestimation. Our approach also decreases the computational cost of the simulation by replacing inefficient Monte Carlo sampling with more efficient block sampling. However, these improvements come at the cost of a weaker link between our measure and the measure of ultimate interest — economic losses. Our approach provides a convenient, quick estimation of network reliability that complements the computationally expensive traffic models.