Research On Tool Wear Condition Monitoring In Titanium Alloy Milling Processing

The titanium alloy has advantages of high specific strength, superb corrosion resistanceand high temperature-resistant, so it has been widely used in the structure and engine ofairplane. But because of poor machinability, tool wear has become the prominent problem inthe milling of titanium alloy. During the milling processing, the cutting tool gradually wearsout and the blunt tool will not only inevitably affect the machining accuracy and work-piecesurface quality, but also decrease the processing efficiency and increase the cost of production.Therefore, it is essential to monitor the tool wear condition in the milling of titanium alloys.This paper discusses our work on the tool of the milling of titanium alloys, and theresearches are about how to achieve an effective monitoring system of tool wear in the millingof titanium alloys. Acoustic emission signals which were selected as the monitoring signalsunder different cutting conditions were collected and analyzed. The signal analysis resultsshow that acoustic emission signals are difficult to extract features from only purely timedomain or frequency domain processing. Combining parametric analysis and local meandecomposition together to extract signals ring-down count, root mean square, energy countand the average energy value of each band under local mean decomposition, and then we putthem together as the main features to reflect tool wear states. Meanwhile, Cutting parametersaffect the tool states to some extent, so the four factors of cutting (spindle speed, feed rate,axial cutting depth, radial cutting depth) were put as the additional features to reflect toolwear states. The principal component analysis (PCA) not only realizes the feature vectorsdimension reduction, but also eliminates the correlation between feature vectors. So PCA wasused to process the feature vectors which constitute the main features and auxiliary features,and the principal components were seen input vector as BP neural network and support vectormachine.The methods of BP neural network and support vector machine are proposed for buildingtool wear states monitoring model. The results show that the training error, test error ofsupport vector machine is less than BP neural network, and the learning times that requests the same error of support vector machine is significantly lower than BP network. The systemis established under support vector machine, which not only provides highly sort-able, quickrecognition speed, but also it can be more accurate for tool states monitoring.