Research On Fault Diagnosis And Health State Assessment Of Rolling Bearings In Automotive Production Lines

Due to the complex working environment of rolling bearings in automotive production lines,the non-linear and non-smooth vibration signals obtained from rolling bearings are often mixed with a lot of noise,which seriously affects the accuracy of rolling bearings fault diagnosis and health status assessment.In this paper,the rolling bearings in the automobile production line is studied,the vibration signal under its operating condition is obtained,and the vibration signal is analyzed by the data-driven method to complete the fault diagnosis and health state assessment of the rolling bearings.The specific content of the work is as follows:Aiming at the fault diagnosis of rolling bearings in automobile production lines,a fault diagnosis method based on whale optimized swarm decomposition combined with kernel principal component analysis and K-nearest neighbor classification algorithm was proposed.Firstly,the time-frequency domain features of the signal were extracted by the whale optimized swarm decomposition method,and other time-domain and frequency-domain features were combined to construct a high dimensional feature set;secondly,the high dimensional feature set was subjected to dimensionality reduction processing by kernel principal component analysis;finally,the K-nearest neighbor classification algorithm was used for pattern recognition and classification on the low dimensional feature set.For the health status assessment of rolling bearings in automobile production lines,the root mean square value and peak-to-peak value in the signal time domain characteristics were used to establish bearings performance degradation indicators.And the support vector machine in machine learning was used to evaluate the health status.Through the analysis of simulation signals and experimental signals,the superiority of the optimized swarm decomposition compared with the empirical mode decomposition and the unoptimized swarm decomposition in the time-frequency domain feature extraction of signals was verified.In the subsequent dimensionality reduction process,the experimental data proved that compared with principal component analysis,kernel principal component analysis could retain more characteristics of the signal while reducing the data dimension.Finally,the optimized swarm decomposition combined with kernel principal component analysis and K-nearest neighbor classification algorithm was used to diagnose the rolling bearings of automobile production line.In the study of health state assessment,the state classification of rolling bearings whole life cycle data was carried out by performance degradation index.Then the classification assessment was carried out using back propagation neural network and support vector machine,and the advantages of support vector machine in the health status evaluation was verified.Finally,the health state assessment of rolling bearings of automobile production line was carried out by support vector machine.After experimental verification,the swarm decomposition based on whale optimization combined with kernel principal component analysis and K-nearest neighbor classification algorithm form a fault diagnosis method and support vector machine health state assessment method proposed in this paper can effectively carry out fault diagnosis and health state assessment for rolling bearings in automotive production lines.