42CrMo钢精密切削的刀具磨损量预测研究  

Research on Tool Wear Prediction of 42CrMo Steel in Precision Cutting

作　　者：成钢 唐昆[2] 刘庞中 刘子聪 袁剑平 胡永乐[2] 毛聪[2] Cheng Gang;Tang Kun;Liu Pangzhong;Liu Zicong;Yuan Jianping;Hu Yongle;Mao Cong(Jianglu Electromechanical Group Co.,Ltd,Xiangtan,Hunan 411100,China;Hunan Provincial Key Laboratory of Intelligent Manufacturing Technology for High-performance Mechanical Equipment,Changsha University of Science and Technology,Changsha 410114,China;不详)

机构地区：[1]江麓机电集团有限公司 [2]长沙理工大学机械装备高性能智能制造关键技术湖南省重点实验室

出　　处：《工具技术》2024年第3期138-143,共6页Tool Engineering

基　　金：湖南省高新技术产业科技创新引领计划(2022GK4027);江麓机电集团有限公司技术质量攻关类项目(220029)。

摘　　要：针对42CrMo钢精密切削刀具磨损量预测研究小样本、非线性的特点,将量子粒子群算法(QPSO)、卷积神经网络(CNN)及长短期神经网络(LSTM)相结合,构建了QPSO-CNN-LSTM组合预测模型。采用QPSO算法对CNN-LSTM模型的隐藏层单元数、学习率、卷积核等进行优化,结合CNN网络特征提取能力强、LSTM网络具备记忆能力的特点,对实际加工实验的刀具磨损量进行预测,并通过误差评价指标分析,与CNN、LSTM、BP等单一模型以及PSO-GRNN组合模型进行预测效果对比研究。研究结果表明,本文构建的组合预测模型相对于单一预测模型,其预测值与真实值吻合程度更高;相对于PSO-GRNN组合模型,三种误差评价指标的误差值至少降低了27%,其泛化性和稳定性较好,预测精度与非线性拟合能力更强。Focusing on the small samples and nonlinear characteristics of tool wear prediction of 42CrMo steel in precision cutting,the QPSO-CNN-LSTM prediction model based on quantum particle swarm optimization(QPSO),convolutional neural networks(CNN)and long short-term memory network(LSTM)are proposed.QPSO algorithm is used to optimize the number of hidden layer units,learning rate and convolution kernel of CNN-LSTM model.Combined with the strong feature extraction and memory ability of CNN and LSTM network,the tool wear amounts of actual machining experiment are predicted.By the analysis of error evaluation indexes,it is compared with the single model of CNN,LSTM and BP,and PSO-GRNN combined model for the prediction effect.The results show that the proposes has a higher degree of coincidence between the predicted and true values compared with the single prediction model.The error value of the three error evaluation indexes is reduced by 27% at least compared with PSO-GRNN combined model.It has better generalization and stability,and greater prediction accuracy and nonlinear fitting ability.

关 键 词：刀具磨损量 组合预测模型 量子粒子群算法优化 卷积神经网络 长短期神经网络 

分 类 号：TH71[机械工程—测试计量技术及仪器] TH164[机械工程—仪器科学与技术]