Optimal swarm decomposition with whale optimization algorithm for weak feature extraction from multicomponent modulation signal

Since multicomponent modulation and complicated interference simultaneously exist in the vibration signals caused by the bearing compound fault, the fault feature becomes rather weak and is hard to be extracted. Therefore, the diagnosis of bearing compound fault is always considered as the bottle neck issue of machinery condition monitoring. The decomposition methods which can separate the individual component into different modes with appropriate criterion provide an alternative for solving this issue. However, various weaknesses, including mode mixing, depending on predefined mode number and poor noise suppression, seriously restrict their application range. To overcome the disadvantages, a novel decomposition method, swarm decomposition (SWD), is initially introduced in the diagnosis of the machinery fault. Inspired by the swarm intelligence, SWD can intelligently decompose the signal by using the swarm filtering and iterative algorithm. A series of analysis, including an impulse responses test and a numerical validation, verifies SWD has a remarkable performance in the decomposition of multicomponent modulation signal. Yet, the superiority also relies on the choice of thresholds. Therefore, a novel nature-inspired meta-heuristic optimization algorithm, whale optimization algorithm (WOA), is applied to solve the problem. Benefiting from the virtues of SWD and WOA, the proposed optimal swarm decomposition (OSWD) is more suitable for the weak feature extraction from multicomponent modulation signal. Finally, to further highlight its superiority in the diagnosis of the bearing compound fault, a number of simulation and real datasets from the axle box bearings of locomotive are applied. Results show that OSWD can be considered as an alternative to address the bottle neck issue from the bearing compound fault.