基于长短时记忆网络的深孔镗削刀具状态监测  

Deep hole boring tools condition monitoring based on LSTM network

作　　者：厉大维 沈明瑞 张贺清 王书利 陈锋 刘阔[1] 王永青[1] Li Dawei;Shen Mingrui;Zhang Heqing;Wang Shuli;Chen Feng;Liu Kuo;Wang Yongqing(College of Mechanical Engineering,Dalian University of Technology,Dalian 116024,Liaoning,China;Inner Mongolia North Heavy Industries Group Co.,Ltd.,Baotou 014033,Inner Mongolia,China)

机构地区：[1]大连理工大学机械工程学院,大连116024 [2]内蒙古北方重工业集团有限公司,包头014033

出　　处：《现代制造工程》2020年第8期92-96,共5页Modern Manufacturing Engineering

基　　金：国家科技重大专项项目(2017ZX04011009);NSFC-辽宁联合基金项目(U1608251);大连市高层次人才创新支持计划项目(2018RD05)。

摘　　要：镗削加工是机械加工领域中非常重要的一种加工手段,被广泛应用于大型零件的深孔加工过程中。但由于镗削加工的切削区域位于深孔内部,所以机床操作者难以对刀具状态做出准确的判别。针对这一问题,提出了基于深度长短时记忆(Long Short-Term Memory,LSTM)网络的镗削刀具状态监测方法。通过对镗削过程的振动和声音信号采集,利用振动和声音信号的频域数据训练深度长短时记忆网络,建立了振动和声音信号与镗削刀具状态的映射模型。在深孔镗床上进行了模型测试试验。试验表明:深度长短时记忆网络模型对刀具状态有着较好的预测准确度。Boring is a very important processing method in the field of machining.It is widely used in deep hole machining of large parts.However,since the cutting area of the boring process is located inside the deep hole,it is difficult to accurately recognize the tool state by machine operator.In response to this problem,it proposed a method for condition monitoring of boring tools based on deep Long Short-Term Memory network.Through the vibration and sound signal acquisition of the boring process,the frequency domain data of the vibration and sound signals was used to train the deep LSTM network,and the mapping model of the vibration and sound signals and the state of the boring tool was established.Model test experiments were performed on a deep hole boring machine tool.Experiments showed that the deep LSTM network model has good prediction accuracy for tool state.

关 键 词：深孔镗削 状态监测 长短时记忆网络 

分 类 号：TG156[金属学及工艺—热处理]