The Application Of Support Vector Machine In Cutting Tool's Fault Diagnosis

In-process tool wear has a profound influence on the precision and roughness of work piece, and even results in discarded product and interrupted machine. Therefore, cutting tool's fault diagnosis has become one of the most significant parts in the whole manufacturing process.First, the data of the tool's cutting condition can be acquired by using AE technology. Power characters are picked up by wavelet packet analysis to realize feature extraction. Secondly, SVM classifier is built based on extracted features by using the theory of Support Vector Machines(SVMs), it is trained and tested using different data, the method has good performance on the classification of the tool's cutting condition.The dissertation is outlined below.In chapter 1, the purpose and significance of the tool"s state recognition are summarized. The developments and research status of Acoustic Emission (AE) and wavelet package analysis technology are analyzed. The applications of SLT (Statistical Learning Theory) and SVMs are described. The goal and main contents of this dissertation are presented.Chapter 2 is devoted to introduce the principle and characteristic of AE technology, at the same time, the extraction method of AE signal is analyzed.Chapter 3 introduces the characteristic of wavelet package analysis technology. According to the characteristics of AE of cutting tool, wavelet packet analysis can deal with time-frequency signals, wavelet packet decompose has proposed to extract the information related to tool wear from the data, and the analysis results show that wavelet packet decompose is more suitable to process AE signal, as the extracted information rightly reflects the tool cutting condition.In chapter 4, the main idea and characteristic of SVMs are briefly introduced. The AE signals of cutting tool condition are processed with wavelet packet analysis, the power of every frequency are picked up to built SVMs classifier. Through the experiments, it is proved that SVMs are feasible and available for identifying the tool's cutting situation. The results show that this method has good performance on identification of the tool's cutting condition.In chapter 5, the main contributions and conclusions of this dissertation are given, and some problems which need further research are set forward.