Block-wise recursive sliding variational mode decomposition method and its application on online separating of bridge vehicle-induced strain monitoring signals

Abstract

A critical element of structural intelligence monitoring is to separate the vehicle-induced response from the original monitoring signal in real time. To separate the vehicle-induced strain components from the strain monitoring signals mixed with various load effects, an online recursive sliding variational mode decomposition method is proposed in this paper. A real-time online signal separation framework is first proposed; then four technical measures are taken to modify the classical variational mode decomposition (VMD) to address the problems of its time-consuming nature and unsuitable for online use. These measures include: by establishing the block recursive Fourier transform results between frames in the sliding, which significantly reducing calculation time and achieving the best balance between calculation efficiency and real-time performance; during iteration, the center frequency is initialized to avoid possible traps; the convergence criterion is tightened; displacement technology eliminates boundary effects during sliding. Simulations show that the proposed method has a better separation effect than VMD, and its operation speed is much higher than VMD, which meets the real-time requirements of online signal separation. Based on the real-time data collected by Daishan Second Bridge Health Monitoring System strain sensors, the proposed recursive sliding variational mode decomposition (RSVMD) realizes the online real-time separation of vehicle-induced strain signals, proving its capability and potential in actual bridge health monitoring.