Performance-Based Risk Assessment for Large-Scale Transportation Networks Using the Transitional Markov Chain Monte Carlo Method

Abstract

Accurately assessing failure risk due to asset deterioration and/or extreme events is essential for efficient transportation asset management. Traditional risk assessment is conducted for individual assets by either focusing on the economic risk to asset owners or relying on empirical proxies of systemwide consequences. Risk assessment directly based on system performance (e.g., network capacity) is largely limited due to (1) an exponentially increasing number of system states for accurate performance evaluation, (2) potential contribution of system states with low likelihood yet high consequences (i.e., gray swan events) to system risk, and (3) lack of actionable information for asset management from risk assessment results. To address these challenges, this paper introduces a novel approach to performance-based risk assessment for large-scale transportation networks. The new approach is underpinned by the transitional Markov chain Monte Carlo (TMCMC) method, a sequential sampling technique originally developed for Bayesian updating. The risk assessment problem is reformulated such that (1) the system risk becomes the normalizing term (i.e., evidence) of a high-dimensional posterior distribution, and (2) the final posterior samples from TMCMC yield risk-based importance measures for different assets. Two types of analytical examples are developed to demonstrate the effectiveness and efficiency of the proposed approach as the number of assets increases and the influence of gray swan events grows. The new approach is further applied in a case study on the Oregon highway network, serving as a real-world example of large-scale transportation networks.