基于深度迁移学习的深孔镗削刀具状态监测  

Tool Condition Monitoring for Deep Hole Boring Based on Deep Transfer Learning

作　　者：崔益铭 牛蒙蒙 田芝豪 杨顼 刘阔 王永青[1,2] CUI Yiming;NIU Mengmeng;TIAN Zhihao;YANG Xu;LIU Kuo;WANG Yongqing(State Key Laboratory of High Performance Precision Manufacturing,Dalian University of Technology,Dalian 116024,China;Intelligent Manufacturing Longcheng Laboratory,Changzhou 213164,China;不详)

机构地区：[1]大连理工大学高性能精密制造全国重点实验室,大连116024 [2]智能制造龙城实验室,常州213164 [3]华晨宝马汽车有限公司,沈阳110044

出　　处：《组合机床与自动化加工技术》2024年第12期164-168,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：辽宁省科技重大专项资助项目(2020JH1/10100016,2021JH1/10400102)。

摘　　要：基于深度学习的刀具状态监测方法可以准确地预测深孔镗削过程中刀具的破损情况。然而,由于装配误差、加工环境差异等因素的影响,一台镗床上训练的刀具状态监测模型往往不能直接应用于同类镗床。针对这一问题,提出了基于深度迁移学习的刀具状态监测方法。首先,基于深度置信网络,利用第一台机床的数据训练模型。在模型隐藏层顶层引入多核最大均值差异衡量源域与目标域的差异,并利用无标签的第二台机床的数据微调模型。刀具状态监测的实例表明,经过深度迁移后,加工状态监测模型针对目标域的准确率提高了28.42%,证明了此方法的有效性。Deep learning-based tool condition monitoring methods can accurately predict the breakage of the cutting tools in deephole boring process.However,due to factors such as assembly errors and differences in processing machining environments,the tool condition monitoring model trained on one particular boring machine tool cannot be directly applied to similar boring machine tools.A tool condition monitoring method based on deep transfer learning is proposed to solve this problem.First,monitoring model is trained based on deep belief network using data obtained from the first machine tool.Multi-kernel maximum mean discrepancy is introduced at the top hidden layer of the monitoring model to measure the difference between the source domain and the target domain.Then unlabeled data obtained from the second machine tool is used to finetune the monitoring model.The example of tool condition monitoring shows that after deep transfer learning,the accuracy of the tool condition monitoring model for the target domain has been improved by 28.42%,which proves the effectiveness of the proposed method.

关 键 词：迁移学习 深度学习 刀具状态监测 

分 类 号：TH16[机械工程—机械制造及自动化] TG53[金属学及工艺—金属切削加工及机床]