基于异类信息融合的砂轮磨损状态监测  

Grinding Wheel Wear Condition Monitoring Based on Heterogeneous Information Fusion

作　　者：胡天荣 郑红伟 戴士杰[1] Hu Tianrong;Zheng Hongwei;Dai Shijie(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300401,China;不详)

机构地区：[1]河北工业大学机械工程学院,天津市300401

出　　处：《工具技术》2023年第2期126-131,共6页Tool Engineering

基　　金：国家重点研发计划(2019YFB1311104)。

摘　　要：Inconel 718作为难加工材料,磨削时砂轮磨损较快,影响加工后工件表面质量。提出了基于磨削力信号和磨削振动信号两种异类信息的砂轮磨损状态识别模型,研究了不同磨损状态下磨削力和磨削振动的时域和频域特征,并提取71种与磨损状态相关的特征。采用核主成分分析法对原始特征集进行特征提取,选取方差累计贡献率为96%的前13个主元作为融合特征集,基于最小二乘支持向量机(LS-SVM)算法建立了砂轮磨损状态识别模型,并采用受试者工作特征曲线(ROC)和准确率两个指标对模型进行了评估,受试者工作曲线面积(AUG)为0.93,准确率可达93.6%。As a difficult-to-machine material,Inconel 718 has a high grinding wheel wear during its grinding process,which affects the surface quality of workpiece processing.In this regard,a grinding wheel wear state identification model based on two heterogeneous information of grinding force signal and grinding vibration signal is proposed.The time-domain and frequency-domain features of grinding force and grinding vibration under different wear states are investigated,and 71 features related to wear states are extracted.Kernel principal component analysis is used for feature extraction of the original feature set,and the top 13 principal elements with a cumulative contribution of 96%of the variance are selected as the fused feature set.The least squares support vector machine(LS-SVM)algorithm is used to establish a grinding wheel wear state recognition model,and the model is evaluated using two indicators:the subject′s working characteristic curve and the accuracy rate,and the area under the subject′s working curve(AUG)is 0.93,with an accuracy rate of 93.6%.

关 键 词：刀具状态监控 机器人磨削 异类信息 最小二乘支持向量机 核主元素分析 

分 类 号：TG743[金属学及工艺—刀具与模具]