CAJ | 학술논문

电刷镀制备Al2O3-13%TiO2(AT13)复合陶瓷涂层是1个多参数耦合的非线性过程。在分析工艺参数对涂层厚度影响的基础上，通过实验采集样本，建立预测涂层厚度的误差反向传播(back propagation, BP)人工神经网络模型。为验证人工神经网络预测模型的准确性，将该模型的预测结果与多元线性回归模型(multiple linear regression model, MLR)的预测结果进行对比。结果表明：与传统多元线性回归模型相比，人工神经网络模型能捕捉工艺参数的非线性规律，能更好地预测涂层厚度，拟合优度R2达到0.86，模型具有较强的泛化能力和自适应能力，为实现电刷镀制作过程中涂层厚度的实时预测与控制提供参考。
전쇄도제비Al2O3-13%TiO2(AT13)복합도자도층시1개다삼수우합적비선성과정。재분석공예삼수대도층후도영향적기출상，통과실험채집양본，건립예측도층후도적오차반향전파(back propagation, BP)인공신경망락모형。위험증인공신경망락예측모형적준학성，장해모형적예측결과여다원선성회귀모형(multiple linear regression model, MLR)적예측결과진행대비。결과표명：여전통다원선성회귀모형상비，인공신경망락모형능포착공예삼수적비선성규률，능경호지예측도층후도，의합우도R2체도0.86，모형구유교강적범화능력화자괄응능력，위실현전쇄도제작과정중도층후도적실시예측여공제제공삼고。
The forming process of Al2O3-13TiO2 (mass fraction,%) composite ceramic coating by electro-brush plating is a non-linear process with multi-parameters coupling. Based on the analysis of the influences of process parameters on coating thickness, the back propagation (BP) artificial neural network prediction model was carried out through the samples acquired by electro-brush plating experiment. In order to verify the accuracy of coating thickness prediction based on BP artificial neural network model, multiple linear regression model (MLR) and BP artificial neural network model were compared. The results show that BP artificial neural network model can be trained to model the highly non-linear relationships between coating thickness and process parameters, and to provide better results than the traditional multiple linear regression models with R2 of 0.86; meanwhile, the BP artificial neural network model has strong generalization ability and adaptive capacity, it referes for the theoretical foundation for real-time coating thickness prediction and control.