A hybrid PSO-DV based intelligent method for fault diagnosis of gear-box

Abstract

The gear box fault occur can lead to the fatal breakdown of mechanical system. Back propagation neural network (BPNN) have been proved to be of widespread utility for identifying and classifying gear box faults to prevent serious damage in a mechanical system. Some researchers have used particle swarm optimization (PSO) to train BPNN. However, because the PSO algorithm has several parameters to be adjusted by empirical approach, if these parameters are not appropriately set, the search will become very slow near the global optimum and even trap into local minima. In this paper, a novel hybrid intelligent method for classifying gear box faults based on vibration signal using the particle swarm optimization (PSO) algorithm, differential evolution (DE) algorithm and BPNN named PSO-DV based BP is presented. The proposed PSO-DV includes both faster convergence of PSO and capability escape from local optima of DE. Experiments were performed on a gear-box fault simulator. The fault samples are obtained by simulating corresponding fault on experiment gear-box. In presented work, a classical PSO based BP neural network and PSO-DV based BP neural network are used for gear box fault classification, their relative effectiveness in fault diagnosis is compared. The experimental results verified that proposed hybrid PSO-DV intelligent method can escape from local minima, so has better convergence than BP neural network and classical PSO based BP neural network. Meanwhile, it achieves also very high accuracy rate of recognition and thus provides decision support in fault classification.