刘涛,张文超,张文帅.变截面涡旋盘齿面粗糙度的双预测模型[J].表面技术,2019,48(8):323-329.
LIU Tao,ZHANG Wen-chao,ZHANG Wen-shuai.Double-predictive Model of Tooth Surface Roughness of Variable-section Scroll[J].Surface Technology,2019,48(8):323-329 |
| 变截面涡旋盘齿面粗糙度的双预测模型 |
| Double-predictive Model of Tooth Surface Roughness of Variable-section Scroll |
| 投稿时间：2018-03-26 修订日期：2019-08-20 |
| DOI：10.16490/j.cnki.issn.1001-3660.2019.08.043 |
| 中文关键词: 变截面涡旋盘 粗糙度 正交试验 多元回归模型 BP神经网络 双预测模型 铣削参数 |
| 英文关键词:variable cross section scroll disk roughness orthogonal test multiple regression model BP neural network double prediction model milling parameters |
| 基金项目:国家自然科学基金项目（51665053） |
|
|
|
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 精确预测三段基圆变截面涡旋盘齿面粗糙度，确定合理的铣削参数，提高变截面涡旋盘齿面的加工质量。方法 首先在正交试验的铣削参数条件下，用XK714数控铣床对毛坯件进行铣削加工，获得三段基圆变截面涡旋盘，用SJ-210表面粗糙度测量仪测量已加工涡旋齿侧面的粗糙度值。然后利用铣削参数和测量的粗糙度值，建立齿面粗糙度的多元回归预测模型和改进的BP神经网络预测模型及双预测模型，并验证该三种模型的精确度。最后对单一因素条件下的粗糙度进行预测、分析。结果 经过计算可得，齿面粗糙度的多元回归预测模型的平均误差为1.43%，最大误差为3.09%。改进的BP神经网络预测模型的平均误差为1.33%，最大误差为3.22%。两种模型的预测平均值作为双预测模型时，预测平均误差为0.627%，最大误差为1.51%。结论 齿面粗糙度的双预测模型的平均误差明显降低，同时可以避免单一预测模型产生主观预测误差。各铣削因素对粗糙度的影响程度不同，进给量fz>吃刀深度ap>刀具转速n>侧吃刀量ae。随着进给量、吃刀深度、侧吃刀量的增加，齿面粗糙度值增加；随着刀具转速升高，齿面粗糙度值降低。 |
| 英文摘要: |
| The work aims to accurately predict the tooth surface roughness of the three-section base-circular variable-section scroll, determine the reasonable milling parameters and improve the machining quality of the variable-section scroll. Firstly, under the orthogonal test milling parameters, the blanks were milled by XK714 CNC milling machine to obtain three-section circular-variable scrolls, and the side roughness of the processed scrolls was measured by SJ-210 surface roughness measuring instrument. Then, milling parameters and measured roughness values were adopted to establish a multivariate regression prediction model of the tooth surface roughness, the improved BP neural network prediction model and bi-predictive model and verify the accuracy of the three models. Finally, the roughness under single factor was predicted and analyzed. Through calculation, the average error of the multivariate regression prediction model for tooth surface roughness was 1.43% and the maximum error was 3.09%, while the average error of the improved BP neural network prediction model was 1.33%, and the maximum error was 3.22%. When the average value was used as the double prediction model, the predicted average error was 0.627% and the maximum error was 1.51%. Therefore, the average error of the double prediction model of the tooth surface roughness is significantly reduced, and the subjective prediction error can be avoided by the single prediction model. Influence degree of each milling factor on the roughness is different, i.e. the feed rate fz > the knife depth ap > the tool speed n>side knife amount ae. As the feed amount, knife depth, and amount of side knife increase, the tooth surface roughness value increases, but the tooth surface roughness value decreases as the tool speed increases. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
渝公网安备 50010702501715号