基于特征优选和狮群优化支持向量机的铣刀磨损状态识别

State Recognition of Milling Tool Wear Using Feature Optimization with SVM Parameterized by LSO

在数控机床加工过程中对其铣刀的磨损状态进行识别与监测是实现智能制造的重要课题之一.以往的人工识别法会造成过早或过晚更换刀具的问题,使用机器学习对刀具磨损状态进行识别已成为主流.本文对采集到的刀具信号,选择对切削力信号和振动信号进行特征提取.针对特征子集的优化选择,提出了多层Filter式特征选择算法PP_mRMR(Pearson+PCA+mRMR),逐层降低特征维数,减少每层计算量,得到维数低、冗余度小的优选特征子集.最后再使用狮群优化算法(LSO)优化支持向量机(SVM)的参数,并利用优化好的LSO-SVM分类器实现刀具磨损状态识别.计算机仿真实验结果表明,对优选的特征子集,采用经过LSO优化后的SVM算法能够准确识别铣刀的磨损状态.相比于现有的其他算法,识别准确率提升了2.4%,可达98.13%,并且由于算法迭代速度更快,识别时间也更少.

During the machining process of CNC,state monitoring and recognition of milling tool wear is an important topic for realizing intelligent manufacturing.Manual recognition methods may cause the problem of replacing the tool at an inopportune time.Nowadays machine learning technologies are widely proposed to achieve the state recognition in various field.In this paper,the cutting force signals and the vibration signals are extracted at first.Then,in order to obtain the subset of optimal features,a novel feature selection algorithm,PP_mRMR(Pearson+PCA+mRMR),is proposed to reduce the feature dimension layer-by-layer and the calculation of each layer.Here the proposed PP_mRMR is composed of a multi-layer filter based on principal component analysis.Finally,the support vector machine(SVM)is applied as a classifier to achieve the state recognition of milling tool wear based on the selected optimal features,while the lion swarm optimization(LSO)algorithm is employed to properly initialize the parameters of SVM.Compared with other existing algorithms,the recognition accuracy is improved by 2.4%,up to 98.13%,and efficiently decrease the state recognition time with faster iteration speed compared to the existing schemes.