【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。更多还原

【Abstract】 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.更多还原