基于指令域大数据的零件尺寸预测方法  

Research on Part Size Evaluation Method Based on Instruction Domain Data

作　　者：向华[1,2] 孙金伟 周浩 周会成 陈国华[2] XIANG Hua;SUN Jin-wei;ZHOU Hao;ZHOU Hui-cheng;CHEN Guo-hua(National Numerical Control System Engineering Research Center,Huazhong Uniersity of Science and Tecnology,Wuhan 430074,China;Xiangyang Advanced Manufacturing Engineering Research Institute of Huazhong University of Science and Technology,Xiangyang Hubei 441053,China)

机构地区：[1]华中科技大学国家数控工程技术研究中心,武汉430074 [2]襄阳华中科技大学先进制造工程研究院,湖北襄阳441053

出　　处：《组合机床与自动化加工技术》2019年第2期90-93,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家科技重大(04)专项(2016ZX04003003;2017ZX04011006-005;5157052041;2015ZX04002202)

摘　　要：针对机械加工零件的尺寸误差人工检测效率低和在机检测成本过高的问题,提出了一种尺寸在线预测方法。基于指令域分析方法,通过在加工过程中获得的数控系统内部指令域大数据,结合实际加工参数与加工后测量的尺寸误差建立起非线性映射模型,该模型基于LM-BP神经网络与RBF神经网络学习实现零件尺寸误差的预测。最后进行对比研究,两种建模方法均能达到很好的预测效果。该方法适应加工参数发生变化的生产环境,能够对被加工零件尺寸误差进行自适应预测。For the low efficiency of manual measurement of dimension error of machined parts and the high cost of OMI(On Machine Inspection),this paper presented a method of on-line prediction of dimension.The method uses instruction domain analysis method in NC machining process to study the changes in the size of the processed parts.This method combines the big data of instruction domain obtained during processing and actual machining parameters with size error of measurement and then build nonlinear model with these data.The model based on LM-BP neural network and RBF neural network learning can finish the prediction of parts size error.The methods are suitable for size error prediction under the changing processing environment.

关 键 词：尺寸预测 指令域 LM-BP RBF 

分 类 号：TH391[机械工程—机械制造及自动化] TG801[金属学及工艺—公差测量技术]