基于深度学习与特征后处理的支持向量机铣刀磨损预测模型  

Support vector machine milling wear prediction model based on deep learning and feature re-processing

作　　者：戴稳 张超勇[1] 孟磊磊 李晋航[2] 肖鹏飞[1] DAI Wen;ZHANG Chaoyong;MENG Leilei;Li Jinhang;XIAO Pengfei(State Key Laboratory of Digital Manufacturing Equipment and Technology,Huazhong University of Science and Technology,Wuhan 430074,China;R&D Center,Dongfang Electric Corporation,Chengdu 611731,China;School of Computer Science,Liaocheng University,Liaocheng 252059,China)

机构地区：[1]华中科技大学数字制造装备与技术国家重点实验室,湖北武汉430074 [2]中国东方电气集团有限公司中央研究院,四川成都611731 [3]聊城大学计算机学院,山东聊城252059

出　　处：《计算机集成制造系统》2020年第9期2331-2343,共13页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金面上资助项目(51575211,51805330,51705263);国家自然科学基金国际(地区)合作与交流资助项目(51561125002);吉林省自然科学基金资助项目(20180101058JC)。

摘　　要：为提高机械加工过程中的刀具磨损预测精度,建立了一种基于深度学习特征降维及特征后处理的布谷鸟优化参数的最小二乘支持向量机预测模型。该模型利用堆叠稀疏自动编码网络将时域、频域及时频域3方面提取的特征向量进行降维处理,然后利用特征后处理确保降维向量单调不递减及平滑趋势,最后采用自适应步长布谷鸟算法优化参数的最小二乘支持向量机模型预测铣刀磨损量。通过试验测试比较所提方法与其他预测方法,表明了所提模型能更有效表征铣刀磨损量,大幅降低预测误差。To improve the accuracy of tool wear prediction during machining,a least squares support vector machine prediction model based on cuckoo optimization parameters of deep learning feature dimension reduction and feature RE-Processing was established.The model used Stacked Sparse Auto-Encoder Network(SSAEN)to reduce the feature vectors extracted from the time domain,frequency domain and frequency domain,then used the feature Re-Processing to guarantees the monotonous non-decreasing and smoothing trend of the dimensionality reduction vector.The Least Squares Support Vector Regression(LSSVR)model with Self-Adaptive Step Cuckoo Search(ASCS)optimization parameter was used to predict the wear of milling cutter.The comparison between the proposed method and other prediction methods showed that the proposed model could more effectively characterize the wear of milling cutter and greatly reduce the prediction error.

关 键 词：刀具磨损 自动编码器 特征提取 特征后处理 布谷鸟搜索算法 最小二乘支持向量机回归算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TH164[自动化与计算机技术—计算机科学与技术]