支持向量机在刀具磨损多状态监测中的应用  

Application of Support-Vector-Machine in Tool Wear of Multi-Stage Monitoring

作　　者：王国锋[1] 李启铭[1] 秦旭达[1] 喻秀[1] 崔银虎[1] 彭东彪[1] 

机构地区：[1]天津大学机械工程学院,天津300072

出　　处：《天津大学学报》2011年第1期35-39,共5页Journal of Tianjin University(Science and Technology)

基　　金：国家自然科学基金资助项目(50805100);国家科技支撑计划资助项目(2008BAF32B11)

摘　　要：基于多传感器信号、采用多分类支持向量机(support-vector-machine,SVM)实现了刀具监测的多状态辨识.通过对切削过程中的多向切削力和振动信号等多传感器信息进行分析,分别获得时域、频域和小波域的信息作为磨损分类特征;同时,运用基于一对多(one-versus-all,OVA)的多分类支持向量机对刀具不同磨损状态下的特征数据样本进行训练和识别.对切削过程中不同磨损状态的分类结果表明,多分类支持向量机具有出色的学习能力,能够实现在小样本情况下的不同磨损阶段分类,并具有较高的识别精度.Based on signals of multi-sensor,the recognition of multi-state of tool wear was realized by support-vector-machine（SVM） of multi-classification.The cutting force and vibration signals were analyzed to draw informa-tion of the multi-sensor in time domain,frequency domain and wavelet domain respectively as recognized features.Meanwhile the feature samples of various wear extents were trained and recognized by support-vector-machine of multi-classification based on one-versus-all（OVA）,which indicates that SVM is of excellent learning ability,able to realize classification of different wear extents for small samples and of high recognition precision.

关 键 词：刀具磨损监测 支持向量机 一对多 多状态识别 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]