基于神经网络的数控铣削变形预测  

Deformation Forecast in Numerical Control Milling Based on Artificial Neural Networks

作　　者：王立涛[1] 柯映林[1] 黄志刚[2] 程耀东[2] 

机构地区：[1]安徽工程科技学院机械工程系,芜湖241000 [2]浙江大学机械与能源工程学院,杭州310027

出　　处：《机械科学与技术》2004年第2期209-211,共3页Mechanical Science and Technology for Aerospace Engineering

基　　金：军工项目"型号工程数控零件变形控制与校形技术";国家自然科学基金"大型整体结构零件变形控制的关键技术研究"(5 0 175 10 2 )资助

摘　　要：数控铣削加工变形问题一直是自动化制造领域的瓶颈问题。铣削过程的复杂性及引起变形的多因素性使加工变形问题很难得到精确的解析解。本文在相关课题研究的基础上 ,将自动控制领域的前沿科学———人工神经网络引入该问题的研究进程之中 ,采用三层反向传播BP网络模拟铣削参数与变形间的非线性关系 ,为加工变形预测及进一步实现变形控制提供科学依据。Processing deformation in the course of numericalcontrolmilling is a key problem in the roboticized manufacturefield. The complicacy of milling processing and other factors of causing deformation make it difficulty to obtain anexact analytic solution from the processing deformation. In this paper, artificial neural networks were used to study this problem, and a three-layer BP neural network was adopted to find the nonlinear connection between milling parameters and parts deformation. Simulation results show that our method is feasible and effective.

关 键 词：加工变形 铣削参数 神经网络 BP算法 

分 类 号：TH391.72[机械工程—机械制造及自动化] TH16