薄壁细长轴自适应校直技术  

Self-adaptive straightening technology of thin-walled slender shaft

作　　者：韩宾[1] 王肖笛 滕朝斌 李颖慧 王聚存 张琦[1] Han Bin;Wang Xiaodi;Teng Chaobin;Li Yinghui;Wang Jucun;Zhang Qi(School of Mechanical Engineering,Xi′an Jiaotong University,Xi′an 710049,China;AECC South Industy Company Limited,Zhuzhou 412002,China)

机构地区：[1]西安交通大学机械工程学院,陕西西安710049 [2]中国航发南方工业有限公司,湖南株洲412002

出　　处：《锻压技术》2022年第2期100-105,共6页Forging & Stamping Technology

基　　金：西安交通大学校企合作科研项目(N-20010394)。

摘　　要：为提高现有直线度校直设备的检测精度,提高校直参数计算的准确性,通过对现有轴类零件校直方法的优、缺点进行分析,针对常见细长轴零件的直线度校直加工,设计出校直加工精度为0.1 mm·m^(-1)的薄壁细长轴零件自适应校直设备总体结构。校直过程基于机器学习的BP神经网络算法和数据库积累,基于三点弯曲校直的基本原理,此设备确定了一个适用于此校直工艺的BP神经网络结构,其结构为:输入层为7个节点、输出层为1个节点、单隐含层为6个节点。通过对该神经网络结构的精度验证可得:当数据库包含800组实验数据时,经过1次校直加工,即可满足精度要求,此设备可大幅地减少校直加工次数。In order to improve the detection accuracy of existing straightness straightening equipment and improve the calculation accuracy of straightening parameters. The advantages and disadvantages of the existing straightening methods for shaft parts were analyzed. Aiming at the straightness straightening processing of common slender shaft parts, the over structure of a self-adaptive straightening equipment for thin-walled slender shaft parts with the straightening accuracy of 0.1 mm·m^(-1) was designed. The straightening process was established based on machine learning consisting of BP neural network algorithm and database accumulation. Based on the basic principle of three-point bending and straightening, a BP neural network structure suitable for this straightening process of this equipment was determined. Its structure is that the input layer has seven nodes, the output layer has one node, and the single hidden layer has six nodes. By verifying the accuracy of the neural network structure, it can be concluded that when the database containes 800 sets of experimental data, the accuracy requirements can be met after one straightening process. This equipment can greatly reduce the numbers of straightening.

关 键 词：轴类零件 自适应校直技术 BP神经网络算法 三点弯曲校直 智能化校直 

分 类 号：TH69[机械工程—机械制造及自动化]