China Mechanical Engineering ›› 2022, Vol. 33 ›› Issue (08): 977-985.DOI: 10.3969/j.issn.1004-132X.2022.08.013

Previous Articles     Next Articles

Analysis for Roll-bending Forming Quality of Spaceflight Thin-walled Cylindrical Workpieces Based on PointCPP-LSF Method#br#

AI Qingbo1;ZHANG Jie1;CHENG Hui2;LYU Youlong1;ZUO Liling1;HU Lan2   

  1. 1.College of Mechanical Engineering,Donghua University,Shanghai,201620
    2.Shanghai Aerospace Equipments Manufacturer Co.,Ltd.,Shanghai,200245
  • Online:2022-04-25 Published:2022-05-19



  1. 1.东华大学机械工程学院,上海,201620
  • 通讯作者: 张洁(通信作者),女,1963年生,教授、博士研究生导师。研究方向为智能制造与大数据。。
  • 作者简介:艾青波,男,1996年生,硕士研究生。研究方向为工业智能。。
  • 基金资助:

Abstract: In order to solve the problems that the accuracy and speed requirements of the roll-bending forming quality detection for spaceflight thin-walled cylindrical workpieces might not be met due to the high dependence of the traditional cylindrical fitting method on the initial values of parameters, a PointCPP-LSF method was proposed to realize the analysis of spaceflight thin-walled cylindrical workpiece roll-bending forming quality. Based on point cloud deep learning, a point network for cylindrical parameter prediction (PointCPP) model was established to obtain reliable initial values, and then the cylindrical parameters were iteratively optimized based on the improved LSF method, and combined with the gross error elimination mechanism, the robust curvature radius calculation results were finally obtained. The experimental results show that the proposed method may effectively improve the accuracy and speed of roll-bending forming quality detection for spaceflight thin-walled cylindrical workpieces. 

Key words:  , cylindrical fitting, point cloud, deep learning, parameter prediction, least square fitting(LSF)

摘要: 为解决传统圆柱拟合方法对参数初值具有较高依赖性而导致无法满足航天筒段薄壁件滚弯成形质量检测精度及速度要求的问题,提出一种PointCPP-LSF方法,以实现航天筒段薄壁件滚弯成形质量分析。基于点云深度学习建立面向圆柱参数预测的点云网络(PointCPP)模型以获得可靠的圆柱参数初值,然后基于改进的最小二乘拟合(LSF)方法对圆柱参数进行迭代优化,并结合粗差点剔除机制,最终获取稳健的曲率半径计算结果。实验结果表明,所提方法能有效提高航天筒段薄壁件滚弯成形质量检测的准确度和速度。

关键词: 圆柱拟合, 点云, 深度学习, 参数预测, 最小二乘拟合

CLC Number: