基于PCC-GWO-SVM算法的刀具磨损预测  

Tool Wear Prediction Based on PCC-GWO-SVM Algorithm

作　　者：蒋忞源 罗敏[1] 刘翰林 夏弋涵 Jiang Minyuan;Luo Min;Liu Hanlin;Xia Gehan

机构地区：[1]湖北汽车工业学院,湖北省十堰市442002

出　　处：《工具技术》2024年第11期131-138,共8页Tool Engineering

摘　　要：针对在刀具磨损实时监测过程中受外界噪声影响而导致预测准确度较低问题,提出一种基于皮尔逊相关系数(Pearson Correlation Coefficient,PCC)和灰狼优化支持向量机(Grey Wolf Optimization Support Vector Machine,GWO-SVM)的刀具磨损量预测模型。该模型采用时域、频域和时频联合域上的特征提取方法,能有效捕捉刀具磨损过程中不同方面的信息;通过PCC优化方法筛选与刀具磨损高度相关的特征数据,提高模型的特征提取能力;利用灰狼算法获取搜索狼群中具有最佳适应度值的位置,即对应的SVM惩罚因子C和核函数参数σ作为SVM的最优参数进行构建和训练,提高预测精度。实验结果表明,PCC-GWO-SVM模型在球头铣刀磨损预测任务中的均方误差MSE为0.0181mm^(2),平均相对误差MAPE为0.187%,决定系数R^(2)为0.9827,均优于预测模型GA-SVM和BES-LSSVM,验证了该模型的有效性和可行性。A tool wear prediction model based on Pearson Correlation Coefficient(PCC)and Grey Wolf Optimization Support Vector Machine(GWO-SVM)is proposed to address the issue of low prediction accuracy caused by external noise during real-time monitoring of tool wear.This model adopts feature extraction methods in the time-domain,frequency-domain,and time-frequency joint domains to effectively capture different aspects of information during tool wear process.Filter feature data highly related to tool wear through PCC optimization method to improve the feature extraction ability of the model.Use the Grey Wolf algorithm to obtain the location with the best fitness value in the search wolf pack,namely the corresponding SVM penalty factor C and kernel function parametersσ.Construct and train SVM as the optimal parameters to improve prediction accuracy.The experimental results show that the mean squared error MSE of PCC-GWO-SVM model in ball end milling cutter wear prediction task is 0.0181mm^(2),the average relative error MAPE is 0.187%,and the coefficient of determination R^(2) is 0.9827,which are better than the prediction models GA-SVM and BES-LSSVM,and verify the effectiveness and feasibility of the model.

关 键 词：皮尔逊相关系数 灰狼优化算法 支持向量机 刀具磨损预测 

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