Multimodal data fusion damage identification method in noisy environments: A case study of cable-stayed bridges

To enhance the accuracy of damage detection and prevent misjudgments when applying a single damage index, we have developed a method for multivariate data fusion damage detection based on signal denoising. This approach involves fusing two modal indicators and two vehicle excitation response indicators, ultimately performing a secondary fusion of the combined indicators. A statistical noise reduction method was applied to minimize noise in the fundamental indices. A modebased fusion index was created based on the curvature and displacement modes, whereas a fusion index based on the vehicle excitation response was generated by using the acceleration energy difference and acceleration energy-curvature difference. The Dempster-Shafer evidence theory was utilized for the secondary fusion of multiple fusion indices, leading to higher damage detection accuracy. The effectiveness of the damage identification method was confirmed by the ratio of the sub-peak value to the peak value. Moreover, numerical simulation data from a cable-stayed bridge further validated the damage detection method, showing a significant decrease in the ratio of the sub-peak value to the peak value (i.e., a reduction of 16-99%) after secondary fusion. These results demonstrate the feasibility of this method.