Centrifugal Pump Impeller Health Diagnosis Based on Improved Particle Filter and BP Neural Network

Abstract

This paper proposes an improved particle filter (PF) algorithm for the denoising of fault signals to reduce the impact of noise on the centrifugal pump impeller fault diagnosis. This method is combined with BP (back propagation) neural network to propose a trouble diagnosis method for impeller of centrifugal pump. Selecting the normal impeller and three centrifugal pumps with different fault impellers as experimental models. The improved PF algorithm is used to denoise the experimental data, then the principal component analysis (PCA) method is used for optimizing and selecting the eigenvalues. Finally, the constructed BP neural network model is used for fault identification. The accuracy of the model was verified by a four-fold cross test. In order to objectively compare the advantages of the proposed BP neural network diagnosis method based on improved PF. In this paper, the experimental results are compared with the experimental results of BP neural network based on traditional PF and particle swarm optimization particle filter (PSO-PF) algorithm. The experiment results indicate that the BP neural network diagnosis method based on the improved PF algorithm is effective for the centrifugal pump impeller fault diagnosis and has higher diagnostic accuracy. This method has certain significance for the research of centrifugal pump impeller fault diagnosis method.