摘要
为了对切削加工件的表面粗糙度进行预测,避免原材料浪费,提出一种基于敏感因素选择与残差网络(ResNet)的表面粗糙度预测方法。该方法首先分析切削系统中不同采样通道的振动信号与表面粗糙度之间的相关性确定敏感信号,然后利用小波包分解将敏感信号分解为不同频段的小波包系数并经过相关性分析选择敏感频段,最后融合各敏感频段的小波包系数构成系数矩阵作为ResNet的输入参数。结果表明,基于敏感因素选择与ResNet的预测方法的相对百分比误差不超过5.8%,均方根误差为0.0159,平均绝对误差为0.0133,决定系数为0.9148。通过与多层前馈网络、支持向量机、卷积神经网络对比证明,所提方法的预测精度具有优越性。
To predict the surface roughness of machined parts and avoid the waste of raw materials,a surface roughness prediction method based on sensitive factor selection and Residual Network(ResNet)was proposed.The correlation between the vibration signals of different sampling channels in the cutting system and the surface roughness was analyzed to determine the sensitive signal.Then,the sensitive signal was decomposed into wavelet packet coefficients of different frequency bands by Wavelet Packet Decomposition(WPD),and the sensitive frequency band was selected by correlation analysis.The wavelet packet coefficients of each sensitive frequency band were fused to form a coefficient matrix as the input parameters of ResNet.The results showed that the relative percentage error of the prediction method based on sensitive factor selection and ResNet was not more than 5.8%,the root mean square error was 0.0159,the average absolute error was 0.0133,and the determination coefficient was 0.9148.By comparing with Back Propagation neural network(BP),Support Vector Machine(SVM)and Convolutional Neural Network(CNN),the prediction accuracy of surface roughness prediction method based on sensitive factor selection and ResNet was improved.
作者
史丽晨
邵献忠
王海涛
豆卫涛
SHI Lichen;SHAO Xianzhong;WANG Haitao;DOU Weitao(College of Mechanical and Electrical Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China;College of Aviation Manufacturing Engineering,Xi’an Aeronautical Polytechnic Institute,Xi’an 710089,China)
出处
《计算机集成制造系统》
北大核心
2025年第2期512-523,共12页
Computer Integrated Manufacturing Systems
基金
西安航空职业技术学院院级课题资助项目(23XHZK-14)
陕西省重点研发计划资助项目(2020GY-104)。
关键词
残差网络
小波包分解
相关性分析
敏感频段
表面粗糙度
预测
residual network
wavelet packet decomposition
correlation analysis
sensitive frequency band
surface roughness prediction
