Improved Computational Framework for Efficient Bayesian Probabilistic Inference of Damage in Truss Structures Based on Vibration Measurements

Abstract

This paper presents a computationally efficient Bayesian inference framework and its application to infer damage in truss bridges. The novelty of the proposed framework for reducing the computational burden of Bayesian inference of structural damage lies in (a) the revised transitional Markov chain Monte Carlo algorithm for efficiently drawing statistical samples of damage indexes and (b) the likelihood function in the frequency domain based on modal properties that could be derived from the stochastic subspace identification. Other special features of the proposed framework include (a) explicitly modeling complexities of truss joints for allowing identification of truss joint damage and the assessment of its associated uncertainties and (b) incorporating the structural finite element model into the probabilistic inference computational environment through the program application interface. Those improvements make new contributions to the field of damage identification for truss bridges, leading to the practical application of Bayesian inference of the large amounts of potential damage not only in truss members but also in truss joints. The applicability and efficiency of the proposed framework are illustrated and examined by using numerical simulation of a damaged prototype truss structure. Simulation results indicate that the proposed approach has the potential capacity for determining the extent and location of large amounts of truss damage and the probabilistic characteristics. Finally, the limitation of this study and the future research need for practical application of the proposed probabilistic framework are discussed.