A joint sparse reconstruction method for mechanical vibration signals based on multi-measurement vector model

Aiming at mechanical equipment becoming more and more intelligent, high-speed, integrated and complicated, The traditional compressed sensing model (single measurement vector) obtains single measurement information, it needs measuring at different monitoring points to obtain multiple signal data, it wastes time and ignores signal correlation between different monitoring points on the same machine. Here, to make full use of inter-signals and intra-signal correlations and further reduce redundancy and sampling time, a joint sparse reconstruction method for mechanical vibration signals based on the multi-measurement vector model was proposed. The design of the reconstruction method was studied emphatically. Firstly, based on the particle swarm optimization algorithm, the initial solution was solved with the time sparse Bayesian algorithm. Then, the pruning technique of the greedy algorithm combined with the adaptive particle activation mechanism was used to do position updating and search the optimal solution. Finally, the vibration signal was reconstructed accurately. Test results showed that compared with other methods, this method can effectively recover mechanical vibration signal and the reconstruction error is relatively smaller. © 2021, Editorial Office of Journal of Vibration and Shock. All right reserved.