基于BP神经网络的工作液电导率与粗糙度模型研究  

Study on Model of Electrical Conductivity of Working Fluid and Roughness Based on BP Neural Network

作　　者：李贵东[1] 郭翠霞[1,2] 邓嫄媛[3] LI Guidong GUO Cuixia DENG Yuanyuan(School of Mechanical Engineering, Sichuan University of Science ＆ Engineering, Zigong Sichuan 643000, China Key Laboratory of Artificial Intelligence of Sichuan Province, Zigong Sichuan 643000, China School of Mechanical Engineering, Chengdu University, Chengdu Sichuan 610106, China)

机构地区：[1]四川理工学院机械工程学院,四川自贡643000 [2]人工智能四川省重点实验室,四川自贡643000 [3]成都学院机械工程学院,四川成都610106

出　　处：《机床与液压》2017年第15期118-121,共4页Machine Tool & Hydraulics

基　　金：过程装备与控制工程四川省高校重点实验室开放基金科研项目(GK201409);人工智能四川省重点实验室开放基金项目(2014RYY04);2014-2016学年大学生创新创业训练计划项目(201510622006)

摘　　要：研究电火花线切割工作液电导率与工件表面粗糙度的关系模型,因试验数据过少,不能可靠建模,现提出在原始数据基础上运用BP神经网络扩充数据建立预测模型。结果表明,预测值与期望值的平均误差为-0.001 7,绝对误差为0.008 5,均方差为8.169 5×10^(-5)、标准差为0.009 0,该方法建立的模型预测精度较高。应用BP神经网络建模可解决试验数据过少而不能可靠建模的问题,为电火花线切割试验数据处理、建模等研究提供一种可靠的方法。The model of the conductivity of the working fluid and the surface roughness of the work piece of the Wire cut Electrical Discharge Machining （WEDM） is studied. Because of the test data too few, it cannot be creating a reliable model. Based on the original data, a way is proposed that using BP neural network to expand the data and to establish a prediction model. The results show that the average error between the predicted value and expected value is -0. 001 7, the absolute error is 0. 008 5, the mean square er- ror is 8. 169 5×10-5 and the standard deviation is 0. 0090, which is of high accuracy. The application of the BP neural network model can solve the problem that the test data was too few to be creating a reliable model, which can provide a reliable method for the process- ing test data and modeling of the WEDM.

关 键 词：电火花线切割机床 电导率 粗糙度 BP神经网络 

分 类 号：TH161[机械工程—机械制造及自动化]