Bayesian model updating method and probabilistic damage identification based on an improved differential evolution adaptive Metropolis algorithm

Abstract

The Bayesian finite element model updating (FEMU) method is widely used in structural health monitoring. Traditional Bayesian FEMU methods face challenges such as dimensional limitations, slow convergence, and low computational efficiency. To improve the convergence speed and computational efficiency of the Bayesian FEMU method, this paper proposes a Bayesian FEMU method based on an improved Differential Evolution Adaptive Metropolis (DREAM) algorithm, named the DREAM_ZC algorithm, and constructs a probabilistic damage identification framework based on this method. The ZC strategies represent the centroid update and sampling difference vectors from past states. The effectiveness of the DREAM_ZC algorithm in FEMU is verified through numerical examples of a simply supported beam and experimental examples of a three-story frame structure. The updated model can serve as a baseline model for probabilistic damage identification. The results show that the proposed DREAM_ZC algorithm has high updating accuracy and fast convergence speed. Using the updated model as the baseline model for probabilistic damage identification can effectively locate the structural damage position and quantify the degree of structural damage, thereby improving the reliability of the damage identification results.