数控机床热误差的最优线性组合建模  

Optimal Linear Combination Modeling for CNC Machine Thermal Error

作　　者：闫嘉钰[1] 杨建国[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2009年第4期633-637,共5页Journal of Shanghai Jiaotong University

基　　金：全国优秀博士学位论文作者专项资金资助项目(200131);上海市引进技术吸收与创新年度计划项目(05-12)

摘　　要：提出数控机床热误差的最优线性组合建模方法及其相关算法.该方法通过线性和的方式对基于不同数学理论所建立的热误差模型进行综合,并以不同拓扑结构及训练算法的反向传播神经网络为例,建立了最优线性组合神经网络.通过对一台CNC机床的实际加工数据进行分析,对该建模方法进行验证,并探讨了该方法的最佳使用条件及其原因.建模结果表明,所提出的方法能够在节省建模时间的同时大幅提高所建立模型的预测精度,是一种高性价比的建模方法.The optimal linear combination modeling method and associated algorithms were presented. This method can be expected to improve the accuracy of the constructed model by forming linear sum of thermal error models based on different math theories. As an example, optimal linear combination of neural network （OLCNN） model was constructed adopting back propagation neural network with different topology structures and training algorithms. The OLCNN model was verified by using real cutting data of a CNC turning machine, the best condition for adopting this method and its reason were discussed. The modeling results show that OLC method is cost-effective for the model accuracy and its generalization ability is significantly improved while the modeling time is saved.

关 键 词：数控机床 热误差 线性组合 建模 神经网络 

分 类 号：TH161[机械工程—机械制造及自动化]