Fault diagnosis of horizontal centrifugal pump orifice ring wear and blade fracture based on complete ensemble empirical mode decomposition with adaptive noise-singular value decomposition algorithm

Abstract

Horizontal centrifugal pump orifice ring wear and blade fracture failure will not only affect the hydraulic performance but also affect the safety and stability of the whole unit. In this paper, the horizontal centrifugal pump orifice ring wear and blade fracture failure are studied, and carry out condition monitoring and fault identification through the vibration signal under the failure. Combined with the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Singular Value Decomposition algorithm of adaptive noise, a vibration feature extraction method of horizontal centrifugal pump based on intrinsic mode singular value is proposed. Through the BP neural network, based on the time domain, frequency domain, wavelet packet-AR spectrum, and intrinsic mode singular value characteristics of single-point and double-point vibration, the identification model is constructed and the identification effect is compared. The research shows that the vibration feature recognition effect of CEEMDAN-SVD decomposition is verified based on BP neural network model, and the BP neural network is improved by Particle Swarm Optimization to further improve the recognition effect and speed, which provides the diagnosis model for the design of subsequent diagnosis system.