A modified blind source separation algorithm for underdetermined structural modal analysis

To improve the mode decomposition capacity for underdetermined and unclear modes, a modified blind source separation (MBSS) method has been proposed, where a multi-synchroextracting transform algorithm with a sliding window is proposed for a higher sparsity time-frequency spectrum. The proposed transform algorithm incorporates an iterative formula of the instantaneous frequency with a sliding window. Then, it is embedded into the existing novel blind source separation (NBSS) method to highly improve the modal decomposition accuracy. The feasibility of the proposed method has been verified against a numerical 3DOF mass-spring-damper system, a numerical three-story frame structure, and an experimental five-story steel frame. The analysis results demonstrate that the proposed MBSS method can well decompose the acceleration signals, providing better precisions than the NBSS method under the circumstance of unclear and underdetermined modes. Moreover, the proposed method has higher decomposition accuracy for close modes.