基于深度学习的刀具磨损监测研究现状  

Research Status of Tool Wear Monitoring Based on Deep Learning

作　　者：王大春 李国和[1] 王丰[1] 闫冬 范建勋 Wang Dachun;Li Guohe;Wang Feng;Yan Dong;Fan Jianxu(Tianjin University of Technology and Education,Tianjin 300222,China)

机构地区：[1]天津职业技术师范大学机械工程学院,天津市300222

出　　处：《工具技术》2022年第6期3-13,共11页Tool Engineering

基　　金：国家自然科学基金(51875409);天津市教委重点项目(2020ZD08);天津市“项目+团队”培养专项(XC202051)。

摘　　要：基于传统机器学习的刀具磨损监测模型监测精度低且计算复杂度高,难以满足智能制造的发展需求。而基于深度学习的刀具磨损监测模型数据处理和特征提取能力较强,可明显提高监测精度,使加工过程更智能化,因此广泛应用于刀具磨损监测。根据采用的模型,将基于深度学习的刀具磨损监测分为基于卷积神经网络的刀具磨损监测、基于稀疏自编码网络的刀具磨损监测、基于深度置信网络的刀具磨损监测、基于长短时记忆网络的刀具磨损监测和基于混合模型的刀具磨损监测,介绍了各种深度学习模型的基础理论及基本结构,总结了国内外基于深度学习模型的刀具磨损监测方法,分析了存在的问题,并指出了未来的发展方向。The traditional tool wear monitoring model based on machine learning is difficult to meet the development needs of intelligent manufacturing due to its low monitoring accuracy and high computational complexity.The tool wear monitoring model based on deep learning, with its powerful data processing and feature extraction capabilities, improves the monitoring accuracy significantly, and the machining process is more intelligent, and it is widely used in tool wear monitoring.According to the model used in this paper, the tool wear monitoring based on deep learning is divided into tool wear monitoring based on convolutional neural network, tool wear monitoring based on sparse auto-encoding network, tool wear monitoring based on deep belief network, long and short-term memory network-tool wear monitoring and tool wear monitoring based on a hybrid model.The basic theories and basic structures of various deep learning models are introduced.The current research status of tool wear monitoring methods based on various deep learning models at home and abroad are summarized.The existing problems are analyzed and the future development direction is pointed out.

关 键 词：刀具磨损监测 卷积神经网络 自编码神经网络 深度置信网络 长短时记忆神经网络 混合模型 

分 类 号：TG71[金属学及工艺—刀具与模具] TH117.1[机械工程—机械设计及理论]