特征融合与GA-SVM在刀具状态监测中的应用研究  

Study on the application of feature fusion and GA- SVM in cutting tool wear monitoring

作　　者：李顺才[1] 李巍[1] 吴明明[1] 

机构地区：[1]江苏师范大学机电工程学院,江苏徐州221116

出　　处：《制造技术与机床》2015年第4期145-148,共4页Manufacturing Technology & Machine Tool

基　　金：国家自然科学基金项目(50974107);江苏省"333工程人才培养"资助项目(333GC2011001);江苏省研究生科研创新计划项目(kylx_1433)

摘　　要：为保证监测的准确性,刀具监控系统往往采用多个传感器进行监测并采集数据,导致监测成本的增加。通过对一个传感器的信号数据建立多个信号处理模型,将多个模型的特征进行融合,深度挖掘信号中所蕴含有关刀具磨损的敏感特征,提高监测的准确性,降低成本;同时针对将融合所得特征输入支持向量机(SVM)进行刀具磨损状态的识别时,常需要反复调整惩罚参数c和核函数g的问题,采用遗传算法(GA)对惩罚参数c和核函数g进行优化,减少了SVM模型的搭建时间,提高了SVM的识别准确率。In order to ensure the monitoring accuracy of tool monitor,researchers usually collect signals with several sensors,which result in the increasing of tool monitor cost. In this article,we have set up several signal- processing models based on the signal from a single sensor and fused the characters of these models,moreover,dug deeply the sensitive feature relating with the tool wear to improve the accuracy of monitoring and reduce the cost. Aiming at the problem that the penalty parameter c and the kernel function g must be adjusted repeatedly when the support vector machine( SVM) is chosen to identify the tool wear state,optimizes c and g by the genetic algorithm,which reduces the time of setting up SVM model and increases the recognition accuracy.

关 键 词：刀具状态监控 特征融合 遗传算法 支持向量机 模式识别 

分 类 号：TG71[金属学及工艺—刀具与模具]