Weak Local Fault Diagnosis of Gearboxes Based on Adaptive Inertia Factor Particle Swarm Independent Component Analysis

Due to the interference of surrounding noise when collecting the vibration signal of a fixed shaft gearbox, it is impossible to extract the fault features contained in the vibration signal with a high degree of accuracy and this reduces the accuracy of fault diagnosis of the gearbox. Aiming at this problem, this paper proposes a method for local weak fault diagnosis of gears based on improved independent component analysis (ICA). Firstly, for the shortcomings of ICA, such as high requirements for initial value selection, ease of falling into local extrema and the need to derive formulae in advance, this paper proposes to improve the separation performance of the algorithm by combining ICA with particle swarm optimisation (PSO). Also aiming at the shortcomings of slow convergence of PSO and decreased searchability in the later iteration, this paper proposes an adaptive inertia weight particle swarm optimisation (AIWPSO) algorithm by introducing the roulette idea into PSO. Then, combining ICA with AIWPSO, an independent component analysis method for adaptive inertia weight particle swarm optimisation (AIWPSO-ICA) is proposed to improve the signal separation performance. Finally, based on AIWPSO-ICA, a method for diagnosing weak local faults of gears is offered. The simulation signals and the real data experimental results verify the effectiveness and superiority over conventional AIWPSO-ICA.