• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2021, Vol. 57 ›› Issue (13): 262-272.doi: 10.3901/JME.2021.13.262

• 制造工艺与装备 • 上一篇    

扫码分享

车削表面粗糙度解析模型与DDQN-SVR预测模型研究

陈超逸1, 鲁娟1,2, 陈楷1, 黎宇嘉1, 马俊燕1, 廖小平1   

  1. 1. 广西大学制造系统与先进制造技术重点实验室 南宁 530004;
    2. 北部湾大学机械与船舶海洋工程学院 钦州 535011
  • 收稿日期:2020-07-07 修回日期:2021-01-26 出版日期:2021-08-31 发布日期:2021-08-31
  • 通讯作者: 廖小平(通信作者),男,1965年出生,博士,教授,博士研究生导师。主要研究方向为数字化设计、智能制造。E-mail:xpfeng@gxu.edu.cn
  • 作者简介:陈超逸,男,1996年出生。主要研究方向为智能制造。E-mail:244368715@qq.com
  • 基金资助:
    国家自然科学基金(51665005)、广西自然科学基金(2019JJB160048、2020JJD160004)和广西高校中青年教师基础能力提升(2020KY10014)资助项目

Research on Analytical Model and DDQN-SVR Prediction Model of Turning Surface Roughness

CHEN Chaoyi1, LU Juan1,2, CHEN Kai1, LI Yujia1, MA Junyan1, LIAO Xiaoping1   

  1. 1. Guangxi Key Laboratory of Manufacturing Systems and Advanced Manufacturing Technology, Guangxi University, Nanning 530004;
    2. Department of Mechanical and Marine Engineering, Beibu Gulf University, Qinzhou 535011
  • Received:2020-07-07 Revised:2021-01-26 Online:2021-08-31 Published:2021-08-31

摘要: 在车削加工中,零件质量是生产者需要密切关注的问题。表面粗糙度作为评价零件质量的一项重要指标,选择满意的切削参数来提高表面粗糙度可有效提高零件质量。为提高表面粗糙度的预测精度,在现有研究基础之上提出一种分段的表面粗糙度理论解析模型对表面粗糙度进行预测。同时尝试采用双深度Q网络(DDQN)优化支持向量回归(SVR)提高数据驱动模型的预测性能,探寻DDQN优化SVR内部参数的环境设计,并且与其他算法对比了其优化效果与稳定性。基于45钢的车削试验,验证分段的表面粗糙度理论模型和DDQN-SVR预测模型的有效性,为基于表面粗糙度的切削参数选择提供了较好的技术支持。

关键词: 车削加工, 表面粗糙度, 理论模型, 支持向量回归, 深度强化学习

Abstract: In turning process, the quality of parts is a problem that producers need to focus on. The surface roughness is an important index to evaluate the quality of parts. Selecting satisfactory cutting parameters to improve surface roughness can effectively improve part quality. In order to improve the prediction accuracy of surface roughness, a segmented analytical model of surface roughness theory is proposed based on the previous research. At the same time, double deep Q network (DDQN) is used to optimize the Support vector regression (SVR) to improve the predictive performance of the data-driven model. The environment design of DDQN to optimize the internal parameters of SVR is explored, and its optimization effect and stability are compared with other algorithms. Based on the turning experiment of 45 steel, the validity of segmented surface roughness theory model and DDQN-SVR prediction model is verified, which provides better technical support for the selection of cutting parameters based on surface roughness.

Key words: turning, surface roughness, theoretical model, support vector regression, deep reinforcement learning

中图分类号: