Study of sound based tool wear monitoring in micro milling using Self-Organizing Map

• Main Authors: Ming-Hsing Lee, 李明興
• Other Authors: 盧銘詮
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/76531349084698269668

碩士 === 國立中興大學 === 機械工程學系所 === 97 === With the fast development in technology. The demand of the product is getting smaller and smaller. Therefore, the development of micro-cutting technology draws a much more attention. In the micro-cutting process, the effect of tool wear on the product quality is much more serious than in the conventional cutting. It is necessary to establish tool condition monitoring. In this reserch, the purpose to establish tool wear monitoring system for the micro milling was based on the sound signal obtained by the microphone sensor in the cutting process. The way of feature signal processing was using fast Fourier Transform to get frequency spectrum. After class scatter criterion, the feature signal was putting on self-organizing Map to reducing variance. The experiment was setup with SK2 workpiece milled by the micro endmill of 700 in diameter. For the classifier design. Learning Vector Quantization(LVQ) network and Fisher Linear Discriminant was used to classify the tool condition. The result shows that the performance of classification was getting batter withing the increasing bandwidth size of feature reducing noise of sound signal. After saturation of increasing bandwidth size of feature the sound signal corresponding with tool wear condition was getting weaker. Putting the chosen features into the LVQ and FLD classifiers. It shows that when bandwidth size of feature reaching 8KHz, the performance for sharp and worn tool testing were 100% probability of classification.