摘要
采用齿面网格化处理和粗糙度测量技术,得到4个主要因素影响的粗糙度预测的训练集和测试集序列,考虑飞秒激光精修面齿轮实验的小样本特点,选择支持向量机回归SVR模型方法进行齿面粗糙度预测。通过对83组实验数据进行预处理,包括48组训练集和35组测试集,比较分析得出采用多项式核函数进行训练的总体预测误差较小,对测试集进行齿面粗糙度预测的评价参数(E_(MSE)、E_(RMSE)、E_(MAE)、E_(MAPE))均保持较低水平,实现了较高精度的预测,说明该预测模型适用于飞秒激光精修齿面粗糙度的预测,特别是对粗糙度在0.2~0.5μm区间的预测效果佳,为面齿轮齿面粗糙度预测提供了有效的技术参考。
By using tooth surface meshing processing and roughness measurement techniques,the training sets and test sets sequences of roughness prediction influenced by four main factors are obtained.Considering the small sample characteristics of face gear experiments of femtosecond laser finishing,the support vector machine regression SVR model is selected for tooth surface roughness prediction.By preprocessing 83 experimental data,which were divided into 48 training sets and 35 test sets,the comparative analysis shows that the overall prediction error of training with polynomial kernel function is relatively small,and the evaluation parameters(E_(MSE),E_(RMSE),E_(MAE),E_(MAPE))of tooth surface roughness prediction for the test set are kept at a low level,thus achieving a relatively high precision prediction.It shows that the prediction model is suitable for predicting tooth surface roughness in precision finishing of femtosecond laser,especially for roughness in the range of 0.2~0.5μm,which provides an effective technical reference for predicting tooth surface roughness of face gear.
作者
李湾
杨向东
明兴祖
伍昆军
吴陶
LI Wan;YANG Xiangdong;MING Xingzu;WU Kunjun;WU Tao(School of Mechanical and Electrical Engineering,Hunan Automotive Engineering Vocational University,Zhuzhou 412000,China;School of Mechanical and Electrical Engineering,Guangzhou Huali College,Guangzhou 511325,China;School of Mechanical Engineering,Hunan University of Technology,Zhuzhou 412007,China)
出处
《机械与电子》
2025年第8期23-29,共7页
Machinery & Electronics
基金
湖南省自然科学基金资助项目(2023JJ50207)
广东省普通高校重点领域专项(2023ZDZX3050,2023ZDZX3051)
湖南省教育厅科学研究优秀青年项目(22B0994)
国家自然科学基金资助项目(51975192)。
关键词
飞秒激光精修
面齿轮
支持向量机回归
齿面粗糙度
预测
femtosecond laser finishing
face gear
support vector regression
tooth surface roughness
prediction
