The SVM-based Fault Diagnosis Of Electric Spindle Of CNC Machine Tools

Electric spindle,which affects the machining accuracy of machine tool directly,is one of the most core components for the CNC machine tool,and the reliability of CNC machine tools can be improved by fault diagnosis of electric spindle.Therefore,fault diagnosis of electric spindle is of great significance.Setting the electric spindle as the research object,the thesis studied the methods of fault diagnosis and established the fault diagnosis platform by using the methods of empirical mode analysis(EMD)and support vector machine(SVM).The main work is as follows:First,this thesis studied the structure of the motorized spindle in detail,then collected and analyzed the faults based on the data of maintenance history from an enterprise.The bearing faults were identified as the main object,which appeared frequently and brought worse influence.Second,the mechanism of bearing faults was studied which includes the mechanism of vibration,failure modes and characteristic frequency,and the features of the typical faults were concluded.Characteristic frequency was extracted successfully from experimental data by the method of combining wavelet denoising,EMD,kurtosis and correlation index.The validity of the signal processing method is verified.Third,the generating method for eigenvector of SVM was optimized by EMD and autoregression model(AR);the test accuracy of experimental data was increased from 87.8% to 95.6% by data pre-processing(normalization and dimensionality reduction)and parameter optimization(method of exhaustion,genetic algorithm and particle swarm algorithm),which proved that data pre-processing and parameter optimization can improve the performance of diagnostic model.Fourth,a platform of fault diagnosis for an electric spindle of a CNC milling machine from a company was established to verify the validity of the diagnostic scheme.To sum up,the fault diagnosis method of electric spindle based on SVM in this paper has good practicability.