基于物理模型和神经网络融合的表面粗糙度预测方法  

Surface Roughness Prediction Method Based on Fusion of Physical Model and Neural Network

作　　者：钱思瑜 张晟玮 官威 王成瀚 苏沛源 沈彬[1] QIAN Siyu;ZHANG Shengwei;GUAN Wei;WANG Chenghan;SU Peiyuan;SHEN Bin(School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Aero Engine Corporation of China(Chengdu),Chengdu 610503,China;HUDONG HEAVY MACHINERY CO.,LTD.,Shanghai 200129,China)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240 [2]中国航发成都发动机有限公司,成都610503 [3]沪东重机有限公司,上海200129

出　　处：《机械设计与研究》2023年第3期156-160,166,共6页Machine Design And Research

基　　金：“高档数控机床与基础制造装备”科技重大专项子课题(2018ZX04011001);国防科工局基础产品创新科研项目—XXX研究(DE0904);上海航天科技创新基金项目(SAST2018-055)。

摘　　要：当前表面粗糙度预测的单一建模方法都存在一定的局限性,物理建模方法无法表征实际加工动态过程,机器学习模型需要大量训练数据且解释性较差。提出了一种物理模型与神经网络深度耦合的融合模型,通过训练卷积自编码器作为特征提取器构建数据集,然后训练融合模型,实现对表面粗糙度的精确预测,通过高温合金侧铣实验建立的数据集进行了验证,上述模型在训练集上预测相对误差为4.48%,测试集上的平均预测相对误差为5.67%。以10%为允差范围,则预测的准确率为84.29%,有较高的精准度。The current single modeling methods for surface roughness prediction have certain limitations.Physical modeling methods cannot represent the actual machining dynamic process.Machine learning models require a large amount of training data and are poorly explained.In this paper,a fusion model of deep coupling between the physical model and the neural network is proposed.The data set is constructed by training the convolutional autoencoder as a feature extractor,and then the fusion model is trained to achieve accurate prediction of surface roughness.Through the side milling experiment of superalloy the established data set is verified,and the relative prediction error of the above model is 4.48%on the training set and 5.67%on the test set.Taking 10%as the tolerance range,the accuracy of the prediction is 84.29%,which has a high accuracy.

关 键 词：表面粗糙度预测 神经网络 融合模型 

分 类 号：TH166[机械工程—机械制造及自动化]