A fault diagnosis algorithm using probabilistic neural network with particle fish swarm algorithm

Traditional probabilistic neural network (PNN) uses identical smooth factor, which easily leads to low recognition rate and misclassification. When the number of training samples increases the number of pattern layer neurons is large, which will lead to complex network structure. Due to these shortcomings, this paper proposes an algorithm using PNN with particle fish swarm algorithm (PFSA-PNN). The advantages of PSO are fused to AFSA, and the improved AFSA is used to select the smoothing vector and optimize the network structure of PNN. Then the optimized PNN is used to the fault diagnosis of bearings. The comparative experiments show that the proposed algorithm not only keeps the classification accuracy but also reduces the complexity of the network, and it has better universality.