摘要
为了更加准确地进行电火花小孔加工工艺目标性能预测,针对电火花小孔加工时非线性和非平稳的动态特性,提出了利用LSTM循环神经网络将电火花小孔加工工艺的一系列参数作为序列问题进行处理,同时考虑电火花小孔加工工艺各参数之间的内在关系,建立了电火花小孔加工工艺目标性能预测模型。通过使用直径为0.3~1.0 mm的8种规格电极,分别对不同厚度的304不锈钢材料设计了正交试验,获得大量训练样本,使用建立的预测模型分别对加工时间、电极损耗量和孔径尺寸3种目标性能进行预测。试验结果表明,该模型能够准确地映射出电火花小孔加工工艺参数之间以及工艺参数与目标性能之间的复杂关系。对比BPNN网络模型,利用LSTM循环神经网络建立的预测模型的3种目标性能预测决定系数R 2的值分别提高了1.54%、1.34%和0.85%。
In order to predict the target performance of Electrical Discharge Machining(EDM)process more accurately,in view of the non-linear and non-stationary dynamic characteristics of small hole EDM,a series of parameters of EDM process were treated as a sequence problem by using LSTM recurrent neural network,and the internal relationship among the parameters of EDM process was taken into account.The prediction model of the target performance of the small hole machining process was established.Through the use of 8 kinds of electrode with diameter of 0.3~1.0 mm,the orthogonal test was designed for 304 stainless steel with different thickness,and a large number of training samples were obtained.The prediction model was used to predict 3 kinds of target performance,including processing time,electrode loss and aperture size.The experimental results show that the model can accurately map the complex relationship between the process parameters and the target performance.Compared with the BPNN model,the value of three kinds of target performance prediction decision coefficients(R 2)of the cyclic neural network model increased by 1.54%,1.34%and 0.85%,respectively.
作者
王民
高晓东
刘建勇
Wang Min;Gao Xiaodong;Liu Jianyong(Key Laboratory of Advanced Manufacturing Technology Beijing Municipal,College of Mechanical Engineering and Applied Electronics Technology,Beijing University of Technology,Beijing 100124,China;Beijing Institute of Electro-Machining,Beijing 100191,China;Beijing Municipal Commission of Science and Technology,Beijing 100744,China;Beijing Municipal Key Lab of EDM Technology,Beijing 100191,China)
出处
《现代制造工程》
CSCD
北大核心
2020年第12期75-82,共8页
Modern Manufacturing Engineering
基金
北京市科技新星计划项目(Z181100006218078)。
关键词
电火花加工
孔
正交试验
LSTM循环神经网络
加工工艺
目标预测
Electrical Discharge Machining(EDM)
hole
orthogonal test
LSTM recurrent neural network
processing technology
target prediction
