基于LSTM的刀具磨损预测模型  

Tool Wear Prediction Model Based on LSTM

作　　者：汤玉红 赖朝安[1] TANG Yuhong;LAI Chaoan(School of Business Administration,South China University of Technology,Guangzhou 510006,China)

机构地区：[1]华南理工大学工商管理学院,广东广州510006

出　　处：《机械制造与自动化》2023年第4期101-104,139,共5页Machine Building & Automation

基　　金：中国高校产学研创新基金项目(2021ALA04002);广东省软科学项目(2019A101002006)。

摘　　要：对刀具每次走刀后的磨损量与走刀过程中的传感器信号进行研究,提出基于长短时记忆神经网络(LSTM)的刀具磨损预测模型。使用LSTM模型自主提取时间序列数据特征,将多个传感器信号作为模型输入自动化处理,避免了人工提取特征出现的信息丢失,充分把握了实验数据的长时依赖性和多维度、多长度特征。经实验对比分析表明:LSTM模型在刀具磨损预测结果上具有显著优势。Based on the study of wear amount of the tool after each pass and the sensor signal during the pass,a tool wear prediction model upon long short-term memory(LSTM)is proposed.The LSTM model is used to autonomously extract time series data features,and input multiple sensor signals as models for automatic processing,which avoids information loss caused by manual feature extraction and fully grasps the long-term dependence and multi-dimensional and multi-length features of experimental data.Comparative analysis results show that the LSTM model has significant advantages in tool wear prediction achievements.

关 键 词：神经网络 长短时记忆神经网络 刀具磨损预测 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]