Iterative Dispersive Vold-Kalman Filter With Local Adaptive Bandwidth for Dispersive Signal Decomposition in Structural Health Monitoring

Dispersive signals, with frequency-dependent propagation velocities, are omnipresent in structural health monitoring (SHM) and nondestructive testing. However, their multimodal nature, overlapping group delays (GDs), and bandwidth variations bring substantial challenges to conventional signal processing techniques. This paper proposes a novel decomposition algorithm – iterative dispersive Vold-Kalman filter (IDVKF) – for accurate GD estimation and robust damage feature detection from multicomponent broadband dispersive signals. Specifically, a dispersive Vold-Kalman filter framework is developed by formulating the decomposition problem in the frequency domain, enabling effective demodulation of highly dispersive signals. To further enhance performance in scenarios with overlapping components and nonuniform bandwidths, variable filtering bandwidths and a joint decomposition scheme are integrated. In addition, a GD refinement strategy is introduced to iteratively update GD estimates via reconstructed envelopes, improving time-frequency resolution and estimation accuracy. Finally, an iterative local bandwidth adaptation mechanism, inspired by windowed signal orthogonality, is proposed to improve algorithm adaptivity against noise and mitigate feature leakage under frequency-dependent bandwidth variations. Numerical simulations and two experimental validations in SHM demonstrate that IDVKF achieves superior accuracy and robustness in GD tracking and damage detection compared to existing methods.