Structural damage detection using a weighted Bayesian fusion model

Structural damage detection plays a vital role in structural health monitoring, and various vibration-based methods have been actively developed in recent years. However, inherent uncertainties in measurements and modelling often compromise damage identification accuracy, limiting practical applications. This paper presents a decision-level data fusion framework designed to improve detection reliability by strategically combining existing algorithms, rather than introducing new damage indices. This study employs two conventional dynamics-based methods: element modal strain energy (EMSE) method and generalised flexibility sensitivity (GFS) method. Using identical modal parameters, damage detection results are first obtained independently from both methods. These results are then processed as complementary information sources with confidence-based weighting, where an optimisation procedure determines the optimal weight distribution. Comprehensive validation through numerical simulations of a plane truss structure and experimental tests on a steel cantilever beam confirms that the proposed fusion approach achieves significantly higher accuracy in damage identification compared to single-method implementations, demonstrating strong potential for real-world structural health monitoring applications.