ET-GD与K近邻相结合的刀具磨损监测方法

Research on Tool Wear Monitoring Based on ET-GD and K-nearest Neighbor Algorithm

针对铣削过程中刀具磨损量监测问题,提出一种基于极端随机树和高斯分布,与K近邻相结合的刀具磨损监测方法。该方法选用截断法和hampel滤波法剔除力、振动和声发射信号中的异常值和奇异点。其次通过极端随机树和高斯分布的偏离情况对特征集进行优选,降低数据矩阵的复杂性。分别对比分析了两次优选前后三种K近邻模型的拟合度和评估度量。利用优选后的特征对逻辑回归、极端随机树、支持向量回归和K近邻算法模型进行训练,并利用十折交叉验证法和测试集进行验证。最终得出,基于极端随机树和高斯分布与K近邻的刀具磨损监测模型的拟合度达到99.17%,均方误差和平均绝对误差分别为13.0688、1.8241。结果表明该方法能够实现对铣刀磨损的有效监测,从而提高工件加工质量。

Aiming at the problem of tool wear monitoring in milling process,a tool wear monitoring method based on extreme random tree and Gaussian distribution and K-nearest neighbor is proposed.In this method,truncation method and Hampel filtering method are used to eliminate outliers and singular points in force,vibration and acoustic emission signals.Secondly,the feature set is optimized by the deviation of extreme random tree and Gaussian distribution to reduce the complexity of data matrix.The fitting degree and evaluation measure of the three K-nearest neighbor models before and after the two optimization are compared and analyzed.The optimized features are used to train logical regression,extreme random tree,support vector regression and K-nearest neighbor algorithm models,and verified by ten fold cross validation method and test set.Finally,the fitting degree between the tool wear monitoring model based on extreme random tree and Gaussian distribution and K-nearest neighbor is 99.17%,and the mean square error and mean absolute error are 13.0688 and 1.8241 respectively.The results show that this method can effectively monitor the wear of milling cutter,so as to improve the machining quality of workpiece.