Researches On Fault Intelligent Diagnosis Technology For The Key Mechanical Parts Of High-Grade CNC Machine Tools

CNC machine tools being of high precision and efficiency are the necessary infrastructure in the development of modern machinery manufacturing. High-level CNC machine tools have been the major machine manufacturing equipment in the field of aerospace, automotive and other high-tech industry. The economic losses could be caused due to CNC machine tool failure. CNC machine tool accidents can be prevented and its normal production order can be maintained if intelligent fault diagnosis technology is employed. So that, the machine tool maintenances quantity can be reduced and maintenance costs can also be derated significantly. There is strong utility value, therefore, to research intelligent fault diagnosis technology and apply it in CNC machine tool.The essence of intelligent fault diagnosis of CNC machine tools is to monitor running status of mechanical components in real-time. By means of detecting and analyzing the relevant informations and running statu data of mechanical components, it extracts a series of fault-sensitive eigenvalues are picked up, complex mapping relationship between feature values and the parts failure can be drafed out by the way of artificial intelligence, and then the fault diagnosis mathematical model that accords with the component actual performance is built to predict the failure in different operating conditions.In practice, mechanical part faults are complex, the distribution of data sample detected by sensors is not uniform usually, and once failure occurs,the machine tool must be stopped to do repairements as soon as possible, the total failure samples, therefore, are limited. Tranditional manual indetects are very difficult to classify these samples accurately.Hypersphere support vector machine is used to construct fault diagnosis model in this thesis. The data collected by sensor are decomposed by wavelet packet, and then energy feature values are piched up as feature vectors. One part of the fault data is used for training, and another part will be used to diagnosis the fault of CNC machine tools. Compared with BP neural network and RBF neural network, the results show that higher diagnostic accuracy is of by this method, especially, strong superiority is shown in the diagnosis of small sample size.Finally, fault intelligent diagnosis system is developed by VC++ and MATLAB, sensors are installed on Changzheng 718 machine tools to collect data and the diagnostic system is run and tested in scence. The result shows that the system can diagnose fault of CNC machine tools effectively. Because of the similarity of CNC machine tools, the method used in this paper also can be used in other machine tools and flexible line to monitor and diagnose fault.