基于FS和GA的特征选择方法及其刀具状态监测  

FS and GA Based Feature Selection Method and Its Application in Tool Condition Monitoring

作　　者：黄称意 朱锟鹏[2] HUANG Chen-yi;ZHU Kun-peng(School of Information Science and Technology,University of Science and Technology of China,Hefei 230026,China;Institute of Advanced Manufacturing Technology,Hefei Institutes of Physical Science,Chinese Academy of Sciences,Changzhou Jiangsu 213164,China)

机构地区：[1]中国科学技术大学信息科学技术学院,合肥230026 [2]中国科学院合肥物质科学研究院先进制造技术研究所,江苏常州213164

出　　处：《组合机床与自动化加工技术》2021年第12期92-96,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家重点研发计划项目(2018YFB1703200);安徽省自然科学基金资助项目(1808085ME119)。

摘　　要：在现代精密数控机床加工过程中,从原始多源信号中提取并选择出对刀具状态变化敏感的特征子集是实现刀具状态监测的重要环节,对保证加工质量、提高加工效率具有重要意义。针对Fisher score(FS)特征选择方法在多分类问题中无法区分出样本分布不均匀以及选择的特征子集存在信息冗余等不足,提出了一种改进的FS结合遗传算法(GA)的两步特征选择方法,根据特征判别性得分以概率的形式对种群进行初始化,同时考虑特征维数和信息冗余,保证了特征子集的综合性能。最后,利用高速铣削加工实验中收集的多传感器数据,从不同方面的定量分析与比较验证了提出的方法的有效性。In the process of modern precision CNC machine tool processing,extracting discriminative features and selecting an appropriate feature subset sensitive to tool condition variations from the original multi-sensor signal is believed as a significant step,which ensure product processing quality and improve processing efficiency.In order to solve the problem that the Fisher score(FS)feature selection method cannot distinguish the uneven sample distribution in the multi-classification problem and to reduce the information redundancy in the selected feature subset,an improved FS combined with genetic algorithm(GA)two-step feature selection method is proposed.All feature discriminative scores are used to initialize the population in the form of probability and the dimension of features and information redundancy are considered,which can effectively ensure the overall performance of feature subset.Finally,the multi-sensor signals were collected in the high speed milling experiment for quantitative and comparative analysis from different perspectives,and the result verified the effectiveness of the proposed method.

关 键 词：特征选择 Fisher score 遗传算法 刀具状态监测 

分 类 号：TH162[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]