Research On Mechanical Fault Diagnosis Method Based On Optimized Particle Filter Algorithm

With the development of artificial intelligence and the popularization of Internet technology,the era of Industry 4.0 is coming.Machinery and equipment will develop in the direction of high integration,high precision,and intelligence.The correlation and complexity between machinery and equipment will be greatly improved.The fault diagnosis of machinery will also usher in the era of big data.Under this background of mechanical big data,researching a mechanical fault diagnosis method suitable for real-time data processing is of great significance to realize the online health monitoring of the entire equipment.Particle filter(PF)is a non-linear filtering algorithm that is theoretically applicable to any nonlinear,non-Gaussian,and time-varying complex systems.However,the traditional PF algorithm has problems such as sample dilution,poor real-time filtering,and low filtering accuracy.Therefore,this paper selects the PF algorithm as the research object,and introduces an improved particle swarm optimization algorithm(NPSO)into the sampling process of the PF algorithm to obtain an optimized particle filtering algorithm(NPSO-PF).The excellent search and optimization capabilities of the NPSO algorithm are used to improve the defects of the PF algorithm,and the NPSO-PF algorithm is applied to the centrifugal pump fault diagnosis under the background of mechanical big data.Simulation experiments prove that the NPSO-PF algorithm has good state estimation performance and stable filtering and noise reduction performance,and is more suitable for noise reduction processing of non-linear data signals.In view of the important status of centrifugal pumps in the national economy and the characteristics of huge noise interference when centrifugal pumps fail.In this paper,the centrifugal pump is selected as the experimental object,and a normal and three impeller-damaged centrifugal pumps are set for experiments.A centrifugal pump fault diagnosis method based on optimized particle filtering(NPSO-PF-BP)proposed in this paper is used to identify and classify the fault types.The diagnosis process is as follows:(1)The proposed NPSO-PF algorithm is used to denoise the original vibration signal collected during the experiment.(2)The time domain analysis method is used to extract the time domain feature parameters of the denoised data.(3)The dimensionality reduction and selection of the extracted feature parameters were performed by the principal component analysis method.(4)The constructed BP neural network model is used to identify and classify the selected feature parameters,and the four-fold cross-validation is used to verify the diagnostic accuracy.In order to verify the effectiveness of the proposed method,two diagnostic models based on different denoising algorithms are set as control groups.The experimental comparison results show that the proposed diagnosis method is effective for the centrifugal pump fault diagnosis,and has faster speed and higher diagnosis accuracy,which also shows that the method is suitable for mechanical faults diagnosis in the context of mechanical big data.