Research On Fault Diagnosis Method Of Roller Bearing Based On Model Optimizing VPMCD

Roller bearing is mainly used to carry and transmit load, it is also one of the most easily damaged components in mechanical equipment. Once the roller bearing failure is likely to make the whole mechanical equipment can’t run normally, even cause paralysis of the whole production chain, inestimable enormous losses and serious casualties accidents when it serious. As a result, the running status real-time monitoring and diagnosis has exceptional value and significance.In essence, the fault diagnosis of roller bearing is a pattern recognition process. Based on the inner relationship between the extracted feature parameters, Raghuraj et al. put forward the pattern recognition method based on Variable predict mode-based class discriminate(VPMCD). The method using mathematical regression mode to quantitatively describe the internal relationship between the extracted feature parameters, selecting the appropriate mathematical model and order to establish the best prediction model by training samples, forecasting the samples with the established optimal forecasting model and identifying the classification with the minimal criterion of the prediction error sum of squares. Aimed the defect that the accuracy of the model established in VPMCD method is low when the relationship between characteristic values is complex, the improvement methods based on the optimization model, the substitution model and the optimization substitution model was presented. The experiment showed that the improved VPMCD method can be effectively applied to fault diagnosis of roller bearing.The main research content of this paper is as follows:1. The comparison analysis of VPMCD artificial neural network and support vector machine which were widely used was carried on. The experimental results through the UCI standard data of these three kinds of pattern recognition method showed the effectiveness of VPMCD method.2. Aimed the defect that the best model established by VPMCD method only choose a single mathematical regression model which can’t fully described the relationship between the extracted characteristic parameters and lacked of the prediction accuracy, the Quantum Genetic intelligent Algorithm(QGA) was used to weight all four types of regression model and optimize the weights, a weighted comprehensive model which can forecast samples better was established.3. For the problem that the four mathematical regression models provided by VPMCD the ability of fitting and prediction was short in the face of the high complex degree relationship between the extracted eigenvalue, two alternative models with strong prediction ability and two improved methods based on the alternative models were put forward.4. Artificial fish intelligence algorithm was applied to weight and fuse the several related Kriging models,the AKVPMCD method was proposed. A new method of noise reduction-ASTFA correlation criterion noise reduction method was proposed and combined with AKVPMCD applied to the strong background noise of roller bearing fault diagnosis. Firstly, ASTFA noise reduction method was used for filtering out the background noise of rolling bearing vibration signal effectively, the characteristic value was extracted, then the AKVPMCD method with several related Kriging models’ weight optimized by the artificial fish swarm algorithm was used to diagnose classification recognition. For information redundancy and “dimension disaster” problems by increasing characteristic dimension and number of features to gain a large number of fault information, the dimensionality reduction method Autoencoder Network(AN) manifold learning method was proposed, after dimensionality reduction the AKVPMCD method was applied to fault diagnosis of rolling bearing has obtained remarkable effect.