Seismic risk assessment of road networks
Rodrigo Silva-Lopez, Mahalia Miller, Gitanjali Bhattacharjee and Jack W. Baker
Mahalia Miller, a former student working with Professor Jack W. Baker at Stanford University, developed a seismic risk assessment of transportation for the San Francisco Bay Area. The proposed process starts by obtaining spatially correlated values of IM at the location of bridges in the system. Then, using information from fragility functions, damage realizations of each bridge can be obtained, which ultimately lead to a damage map of the whole system. After computing a damage map, to compute the performance of the network after the occurrence of a disruptive event, it is necessary to use a traffic model to simulate how people commute in the damaged condition. As a result of this process, an estimation of traffic metrics such as an increase in aggregate commute time, or maximum travel time for users, or number of trips disconnected, can be performed.
The attached Figure was developed in collaboration with Alan Poulos.
Use of bridge Corridors to support optimal seismic risk management of road networks
Rodrigo Silva-Lopez and Jack W. Baker
This study proposes the use of bridge clusters, defined as Corridors, to support optimal bridge retrofitting strategies for seismic risk management of road networks. A Corridor is defined as a set of bridges that works jointly to ensure connectivity and traffic flow between different areas of a region. To detect Corridors, a Markov Clustering Algorithm is proposed. Using the San Francisco Bay Area road network as a testbed, this clustering technique selects sets of bridges that correspond to main traffic arteries such as highways and high-capacity road segments. After Corridors have been detected, a two-stage stochastic optimization is implemented to detect which bridges should be retrofitted to ensure an acceptable performance under uncertain conditions. The proposed optimization couples retrofitting actions over bridges in a Corridor with the repair actions to individual damaged bridges after an earthquake. The Corridors-Supported Optimization is shown to decrease road network disruption more than other approaches based on ranking bridges according to their traffic capacity or their location in the network.
Deep-learning based retrofitting and seismic risk assessment of road networks
Rodrigo Silva-Lopez, Jack W. Baker and Alan Poulos
Seismic risk assessment of road systems involves computationally expensive traffic simulations to evaluate the performance of the system. To accelerate this process, we developed a neural network surrogate model that allows rapid and accurate estimation of changes in traffic performance metrics due to bridge damage. Some of the challenges explored on the calibration of this neural network are the definition of sampling protocols, selection of adequate hyperparameters, and developing a validation that incorporates practical considerations of the model. In addition to the neural network, a modified version of the Local Interpretable Model-Agnostic Explanation (LIME) is being explored as retrofitting strategy that minimizes the impact of earthquakes on the system. The modified version (LIME-TI) uses Traffic Impacts and rates of occurrence to aggregate the importance of individual damage realizations during the computation of variable importance. The study we are developing uses the San Francisco Bay Area road network as a testbed. As a conclusion of this study, the neural networks accurately predict the system's performance while taking almost 100,000 times less to compute traffic metrics, allowing decision-makers to quickly evaluate the impact of retrofitting bridges in the system. It is also observed that the proposed LIME-TI retrofitting strategy is superior to other metrics (such as traffic volume or vulnerability) in identifying bridges whose retrofit effectively improves network performance.
Integration of people-centric metrics on the seismic risk analysis of road networks
Rodrigo Silva-Lopez, Gitanjali Bhattacharjee, Alan Poulos and Jack W. Baker
This study integrates welfare loss, a measure of the impact of post-earthquake road network disruption on individual commuters' well-being previously formulated by Mackie 2001 with a probabilistic seismic risk assessment framework in a computationally tractable way. As a summary statistic of network performance, welfare takes into account that the same change in commute time can impact commuters with different characteristics to different degrees; traditional measures of road network performance such as drivers' delay, the number of infeasible trips, or combinations thereof obscure potentially disparate impacts. For a probabilistic case study of the San Francisco Bay Area, welfare loss is computed by augmenting an origin-destination matrix with publicly available information about commuters' income levels, residences, and workplace locations for 1980 earthquake scenarios. While all commuters have a similar risk of drivers' delay (i.e., increased travel time) due to post-earthquake road network disruption, commuters with low incomes have substantially higher risk of welfare loss than commuters with high incomes. Aggregate welfare loss underestimates the welfare loss of commuters with low incomes while overestimating the welfare loss of commuters with high incomes. Disaggregation of welfare loss -- or other measures of road network disruption -- is therefore necessary to devise equitable risk reduction policies, as a probabilistic comparison of two bridge retrofit prioritization policies demonstrates. While the retrofit policy determined using drivers' delay reduces the expected drivers' delay for the region, it increases the disparity in the per-capita welfare loss of commuters with low and high incomes relative to the network's baseline state. In contrast, the retrofit policy determined on the basis of the proportion of commuters that use a bridge and have low incomes reduces the difference in the per-capita welfare loss of commuters with low and high incomes compared to the baseline network state.
Community-centric lifeline risk management