Bridge Deterioration Modeling by Markov Chain Monte Carlo (MCMC) Simulation Method

Abstract

There are over 10,000 rail bridges in Australia that were made of different materials and constructed at different years. Managing thousands of bridges has become a real challenge for rail bridge engineers without having a systematic approach for decision making. Developing best suitable deterioration models is essential in order to implement a comprehensive Bridge Management System (BMS). In State Based Markov Deterioration (SBMD) modeling, the main task is to estimate Transition Probability Matrixes (TPMs). In this study, Markov Chain Monte Carlo (MCMC) simulation method is utilized to estimate TPMs of railway bridge elements by overcoming some limitations of conventional and nonlinear optimization-based TPM estimation methods. The bridge inventory data over 15 years of 1,000 Australian railway bridges were reviewed and contribution factors for railway bridge deterioration were identified. MCMC simulation models were applied at bridge network level. Results show that TPMs corresponding to critical bridge elements can be obtained by Metropolis-Hasting Algorithm (MHA) coded in MATLAB program until it converges to stationary transition probability distributions. The predicted condition state distributions of selected bridge element group were tested by statistical hypothesis tests to validate the suitability of bridge deterioration models developed.