基于RBF神经网络的连杆衬套强力旋压轴线直线度预测  

Prediction on axial straightness of connecting rod bushing in the power spinning based on RBF neural network

作　　者：吉梦雯 樊文欣[1] 尹馨妍 王瑞瑞 郭芳 Ji Mengwen,Fan Wenxin,Yin Xinyan,Wang Ruirui,Guo Fang(College of Mechanical Engineering, North University of China, Taiyuan 030051, Chin)

机构地区：[1]中北大学机械工程学院

出　　处：《锻压技术》2018年第3期67-71,共5页Forging & Stamping Technology

基　　金：山西省自然科学基金资助项目(2012011023-2);山西省高校高新技术产业化项目(20120021);中北大学第十届研究生科技基金项目(20131018)

摘　　要：为了实现对连杆衬套强力旋压轴线直线度误差的预测,从而改善连杆衬套的性能,基于MATLAB平台,建立了减薄率、进给比、首轮压下比与轴线直线度误差之间的RBF神经网络模型。用仿真数据对其进行训练,然后预测内、外轴线的直线度误差。并将预测值与仿真值比较,得出RBF神经网络预测误差百分比,与实测值进行比较,验证RBF神经网络在实际生产中的预测性能。再与同样条件下所建立的BP神经网络预测误差百分比对比。发现RBF神经网络可以用来预测连杆衬套强力旋压轴线的直线度误差,并且比BP神经网络收敛速度及学习速率更高,训练过程更稳定,预测精度更高。In order to realize the prediction of axial straightness error of the connecting rod bushing during the power spinning and improve the performance of connecting rod bushing,RBF neural network model was established based on the MATLAB platform among the thinning ratio,feeding ratio,the first pressure ratio and axis straightness error. Then,it is trained by the simulation data,and straightness error of the inside and outside axis was predicted. Next,comparing the prediction value with simulation value,the prediction error percentage of RBF neural network was obtained,and the prediction performance of the RBF neural network model in the actual production was verified by comparing with the measured values. Furthermore,the prediction error percentage was compared with that of the BP neural network built under the same conditions,and RBF neural network can predict the axial straightness error of the connecting rod bushing during power spinning. Thus,compared with BP neural network,RBF neural network can obtain higher convergence rate,better learning rate,more stable training process and higher prediction accuracy.

关 键 词：连杆衬套 强力旋压 轴线直线度误差 RBF神经网络 BP神经网络 

分 类 号：TG146.11[一般工业技术—材料科学与工程] TP311[金属学及工艺—金属材料]