TensorFlow框架下的车削工件表面粗糙度预测方法  

Prediction Method of Surface Roughness of Turning Workpiece Based on Tensor Flow Framework

作　　者：田景海 陈江义[1] 陈瑛琳 杨布尧 TIAN Jing-hai;CHEN Jiang-yi;CHEN Ying-lin;YANG Bu-yao(School of Mechanical Engineering,Zhengzhou University,Henan Zhengzhou 450001,China)

机构地区：[1]郑州大学机械工程学院,河南郑州450001

出　　处：《机械设计与制造》2021年第5期82-84,共3页Machinery Design & Manufacture

基　　金：河南省重大科技专项资助(171100210300-01)。

摘　　要：利用TensorFlow机器学习框架建立了前馈神经网络模型,以三个切削参数作为输入变量,分别是刀具切削深度ap、切削速度vc和进给量f,输出变量是表征工件表面粗糙度的三个指标,即轮廓算数平均偏差Ra、轮廓最大高度Ry或微观不平度十点高度Rz。利用数控车床加工数据对神经网络进行训练,训练好的网络可以用来预测工件的表面粗糙度。预测结果表明基于TensorFlow框架的表面粗糙度预测方法具有建模方便和精度高的特点,因此提出的方法对车削工艺的智能化编制有一定的参考价值。A feed-forward neural network is constructed to predict the surface roughness of turning workpiece based on TensorFlow,a machine learning framework.Three cutting parameters consisting of depth of cutting ap,cutting speed vc and feed rate f are taken as the input arguments of the network,and surface roughness consisting of arithmetical mean deviation of the assessed profile Ra,maximum height of profile Ry or the microscopic point height of irregularities Rz as the output.Some data measured from the CNC lathe is adopted to train the neural network.After that the trained network can be used to predict the surface roughness of the workpiece.The prediction results show that the presented method based on TensorFlow framework has higher prediction accuracy and convenient modelling.Therefore,the method presented in this paper is helpful for smartly programing the turning process.

关 键 词：TensorFlow框架 机器学习 神经网络 表面粗糙度 预测方法 数控车床 

分 类 号：TH16[机械工程—机械制造及自动化] TG519.1[金属学及工艺—金属切削加工及机床]