摘要
实际工程中数控机床刀具磨损状态识别时,作为训练样本的刀具和待识别磨损状态的刀具必然不同,不同刀具间磨损状态准确识别成为必须解决的问题。受走刀次数、安装误差等众多因素影响,同类型同磨损状态的不同刀具切削信号不可避免的存在差异。针对不同刀具间磨损状态准确识别的问题,提出了深度特征联合匹配的刀具磨损状态识别方法。首先,利用堆栈加噪自编码网络对样本频谱进行特征自提取以获得表征磨损状态的深度特征;然后,通过迁移联合匹配算法对所获得的深度特征进行联合匹配以缩小不同刀具磨损状态深度特征间的差异,解决同类型同磨损状态的不同刀具间特征差异问题;最后将联合匹配后的深度特征输入K最近邻分类器实现刀具磨损状态识别。铣刀与车刀磨损状态识别实验结果显示,所提方法识别准确率最高可达到97.04%,证明了该方法能够有效识别刀具磨损状态,并具有较好的泛化能力和稳健性。
During the recognition process of wear state of numerical control machine tools in practical engineering, the tools used as training and recognition samples are different. The accurate recognition of wear state among different tools is a necessary problem to be solved. The influence factors include times of cutting installation error, etc. The signals of the same type and wear state of different cutting tools are inevitably different. To achieve accurate recognition of wear state among different tools, a method of tool wear state recognition is proposed, which is based on the joint matching of depth characteristics. Firstly, the sample spectrum is extracted by stacked denoising auto encoder to acquire depth characteristics demonstrating wear patterns. Then, the obtained depth characteristics are matched by transfer joint matching to reduce the difference among tool depth characteristics of wear state. In this way, the problem of the difference among tools in the same type and wear state is solved. Finally, the depth features after joint matching are utilized as the input of K-nearest neighbor classifier to realize tool wear state recognition. Experimental results show that the identification accuracy of the proposed method can reach 97.04%, which proves that this method can effectively identify the wear status of the cutter, and has good generalization ability and robustness.
作者
陈仁祥
吴志元
胡小林
杨钢
赵玲
Chen Renxiang;Wu Zhiyuan;Hu Xiaolin;Yang Gang;Zhao Ling(Chongqing Engineering laboratory for Transporation Eninering Aplication Robot,Chongqing Jiotong Uniersiy,Chongqing 400074,China;Chongqing Innoration Center of Industrial Big-Data Co.,Id.,Chongqing 400056,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第12期138-145,共8页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金项目(51975079)
重庆市技术创新与应用示范项目(cstc2018jscx-msybX0012)
重庆市教育委员会科学技术研究项目(KJQN201900721)
交通工程应用机器人重庆市工程实验室开放基金项目(CELTEAR-KFKT-202002)资助。
关键词
刀具
磨损状态识别
深度特征
联合匹配
tool
wear state recognition
depth characteristics
joint matching
