Chatter Online Vibration Monitoring, Based On Hidden Markov Models And Support Vector Machine, Machine Tool Cutting

During the cutting process transiting from normal to chatter state, there is a large amout of information embeded in the system, and the signals of this duration are always high nonstability and poor reappearance. In view of these characteristics, Hidden Markov Models (HMMs), as a basic modeling tool in this paper, are introduced to model the cutting process and recognize the cutting state. Besides, considering the strong ability to classification of neural networks (NNs), generalization of Support Vector Machine (SVMs) and model the dynamic temporal process of HMMs, hybrid models are constructed to monitoring the cutting process on the basis of HMMs. The later experiments testified the validation of the models, the robust of the system, real time forecasting and stable monitoring.The main achievements and innovation in this paper includes:(1) Methods about how to extract features from cutting signal has been intensively studied and a new feature extraction method is proposed based on the thought of short-time Fourier Transform.(2) Two monitoring modes are constructed respectively based on DHMM and CDHMM.(3) Structure and performance of Multi-Layer Perceptron has been studied, and a new process monitoring method is proposed to integrate multi-source information in cutting stage.(4) Hybrid system is constructed based on MLP and HMM and applied to monitor the cutting process.(5) Combinating HMM and SVM for monitoring cutting chatter in cutting process for the first time and achieves a very good results.(6) A chatter monitoring system has been developed and gives a very kind performance in experiments.This research can promote the progress of on-line monitoring and control technology of cutting chatters, open out the characteristics of cutting chatters, and advance products quality and productivity.