基于BP和RBF神经网络预测硬质阳极氧化膜的硬度  

Prediction of Hardness of Hard Anodic Oxidation Film Based on BP and RBF Neural Network

作　　者：邢翀[1] 王坤昊 XING Chong;WANG Kunhao(Department of Information and Technology,Changchun Finance College,Changchun 130028,China;Computer College,Jilin Normal University,Siping 136000,China)

机构地区：[1]长春金融高等专科学校信息技术学院,吉林长春130028 [2]吉林师范大学计算机学院,吉林四平136000

出　　处：《电镀与精饰》2021年第7期25-29,共5页Plating & Finishing

基　　金：吉林省科技厅自然科学基金(20190201191JC);吉林省教育厅科学技术研究基金(JJKH20210787KJ)。

摘　　要：通过两种不同方式分别构建NEWRB函数RBF神经网络和K-均值聚类RBF神经网络,同时构建BP神经网络。采用正交实验数据对不同神经网络进行训练,然后用训练完成的不同神经网络预测硬质阳极氧化膜的硬度,并将预测结果与实测值进行对比。结果表明:与BP神经网络相比,NEWRB函数RBF神经网络和K-均值聚类RBF神经网络的平均相对误差和最大相对误差均较低。通过两种不同方式构建的RBF神经网络都具有较高的预测精度,并且K-均值聚类RBF神经网络具有更高的预测精度,更适用于预测硬质阳极氧化膜的硬度。NEWRB function RBF neural network and K-mean clustering RBF neural network were es‐tablished in two different ways,and BP neural network was also established.Different neural networks were trained using orthogonal experimental data,and then the trained different neural networks were used to predict the hardness of hard anodic oxidation film,and the predicted value was compared with the measured value.The results showed that compared with BP neural network,the average relative er‐ror and maximum relative error of NEWRB function RBF neural network and K-mean clustering RBF neural network were lower.The RBF neural network established by two different methods has higher prediction accuracy,and the K-mean clustering RBF neural network has much higher prediction accura‐cy,which was more suitable for predicting the hardness of hard anodic oxidation film.

关 键 词：硬质阳极氧化膜 硬度 BP神经网络 RBF神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]