The Investigation Of Intelligent Tool Wear Monitoring Techniques For Metal Cutting Process

Tool wear and tool breakage are inevitably phenomenon occurring in the process of metal cutting, the transformation of tool-state directly cause product quality declining, production cost increasing, and gradually affect the ability of market competition of products. Therefore it is necessary to do researches on the tool wear monitoring technique in NC machining process.This thesis deals with the following research efforts:1. The effect of experimental method on recognition precision of neural network is investigated; the research method of adopting full factors combination to study tool wear monitoring technique is given out; Full data that network modeling need and establish the solid foundation for research success are obtained.2. The change rule of signal features with different factors is analyzed, two kinds of tool wear monitoring methods based separately on standard module and "dynamic tree" theory are given out. The two methods completely eliminate machining parameter effect on features, and improve the recognition precision of monitoring system, and also, it can build the tool wear monitoring system under arbitrary machining conditions.3. Based on the detail analysis of cutting forces signals, vibration signals and acoustic emission signals in time domain, frequency domain as well as wavelet packet transformation, a new method of monitoring tool wear through variable features is presented and a new approach named "synthesis coefficient" of selecting monitoring features are given out. These methods can select features automatically, improve monitoring precision and be advantageous to build a self-adjusting tool monitoring system.4. The features of B-spline neural network applied to tool wear monitoring are analyzed and a new method of using B-spline neural network to build a character model is given out. This method improves the recognition speed and multiplication of monitoring system, therefore it is suitable for on-line monitoring.5. Aimed at the difficult problem of poor reliability of monitoring system, the allusive relationship between tool wear and signal features is build by means of integrated neural network. And an improved model for pattern recognization is proposed. That can eliminate the signal aberrance's effect on classifying results by monitoring tool state in different time, which makes the system reliability greatly enhanced.6. The practical model of tool wear monitoring is creatively presents on the basis of the proposed monitoring theories, which is applied to milling operation and turning operation. The research results indicate that the proposed model is beneficial to industrial popularization of research finding.The researches on the experiment design, signal analysis, monitoring strategy, feature extraction and feature selection and pattern recognition and etc are exploratory in this thesis. A series of problems for the practicality progress of tool wear monitoring technique are solved, the precision and reliability of monitoring system are improved, and a new idea for the practicality of monitoring system, which enriches and develop the tool wear monitoring technique, is explored. The research achievements can be applied to all sorts of machining, and it has a good application foreground.