Neural Network And Hidden Markov Mixture Model In The Mechanical Machining Flutter Prediction

Based on the "Application on cutting vibration online monitor and control in Hidden Markov Models"(National Nature Science Fund Project, No: 50375070), the Hidden Markov Models (HMMs) and Artificial Neural Network (ANN) theories and methods are studied, then proposed the application of cutting vibration forecast system by HMM model methods and developed the forecast software based on HMM+ANN fixed model. The details are studied as follows:Firstly, the paper briefly introduced the meanings of studying the forecast in cutting vibration and summarized the present conditions of it, then, analyzed the feasibility to use HMM in cutting vibration's forecast.intoduced the base theory of ANN and the progressing for HMM model. Finally, given the research substance of this paper basing on this National Nature Science Fund Project, and proposed the total frame and innovations of this paper.Secondly, introduces the basic ideas of Markov Chain theories briefly, and then extends it to Hidden Markov Models. Then, studying the theories and algorithms of Hidden Markov Models in detail, and discusses some issues in actual applications and gives corresponsive solution means to them. Give a way how to select the data for training. At last the functions of HMM used in cutting vibration's forecast are given.Thirdly, the article introduces two important model of ANN: Self-Organizing feature Map( SOM)and Multilayer Perceptrons(MLP). Study and practice the tool of vertified in HMM with SOM. Combined MLP and HMM to improve forecasting rate.Fourthly , introduces the way of HMM's original model by Genetic Algorithms (Gas).Fifthly, this paper gives a Digital filter method by using the Viterbi algorithm, and discusses the ways of Feature extraction of signals and the Scalar quantization technology of amplitude spectrum vectors. At last brings up the FFT-HMM method to predict the cutting vibration.Sixthly, introduces the way to design the forecast software based on the ANN-HMM theory, the development environment of software system, and development tools and the hybrid method between them. Finally, introduces the functions of this forecast software.Seventhly, verifies the effect of this forecast method experiment upon a CA6140 horizontal ordinary lather.At last, all of work in this dissertation is summed up, and the future researches on applications of HMMs are prospected.