基于BP神经网络的钢丝绳精密传动误差预测  

Prediction of Transmission Error of Cable Precision Drive based on BP Neural Network

作　　者：谢宏伟[1] 陶忠[1] 周晓光 侯军占[1] 张卫国[1] 王宏浩[1] Xie Hongwei Tao Zhong Zhou Xiaoguang Hou Junzhan Zhang Weiguo Wang Honghao(Xi＇an Institute of Applied Optics, Xi＇an 710065, China Xi＇an Military Representative Office of Army Aviation Military Representative Agency, Xi＇an 710065, China)

机构地区：[1]西安应用光学研究所,陕西西安710065 [2]陆军航空兵军事代表局驻西安地区军代室,陕西西安710065

出　　处：《机械传动》2017年第9期130-134,共5页Journal of Mechanical Transmission

基　　金："十二五"兵器预研支撑基金项目(62201070139);中国兵器工业集团高等院校协同创新合作专项项目(KH201504;KH2016006)

摘　　要：针对不同绳组数的钢丝绳精密传动方式,研究了基于神经网络的传动误差预测方法。以预紧力、负载、转角为输入,传动误差为输出,建立了双隐含层BP神经网络模型。在不同绳组数的传动条件下,测试了不同预紧力、负载和转角下的传动误差。将传动误差测试结果分为训练组和预测组,对网络进行训练和检验。测试结果显示,从动轮在0°~180°范围内转动时,传动误差随转角的增大而增大,最大值为5.6 mrad,随绳组数的增加而降低。在该转角范围内,模型预测误差小于0.294 mrad。表明采用双隐含层BP神经网络模型对传动误差进行预测是有效的。For cable drive with different rope number, the transmission error prediction method is explored based on BP neural network. A prediction model for transmission error is developed by a double hidden layer BP neural network. The input parameters of the model are preload, load and rotation angle respectively. The output parameter is the transmission error. The transmission errors under different conditions are tested firstly. And then the results are divided to training samples and prediction samples, which are used to train the network and verify the performance of the trained network. The experiment results show that the transmission error in- creases with the increase of rotation angle at the range of 0 to 180. And the biggest value is 5.6 mrad, the pre- diction error is less than O. 294 mrad. In conclusion, the double hidden layer BP neural network is effective to predict transmission error.

关 键 词：钢丝绳传动 传动误差预测 BP 神经网络 预紧力 负载 

分 类 号：TH132.3[机械工程—机械制造及自动化]