中国机械工程 ›› 2023, Vol. 34 ›› Issue (15): 1820-1831.DOI: 10.3969/j.issn.1004-132X.2023.15.007

• 可持续制造 • 上一篇    下一篇

共享制造下基于数据迁移的滚齿加工碳耗预测方法

易茜1,2;罗雨松2;胡春晖2;卓俊康2;李聪波1,2;易树平2   

  1. 1.重庆大学机械传动国家重点实验室,重庆,400044
    2.重庆大学机械与运载工程学院,重庆,400044
  • 出版日期:2023-08-10 发布日期:2023-08-14
  • 作者简介:易茜,女,1986年生,副教授、博士。研究方向为绿色制造、生产系统与管理、智能制造。E-mail:yiqian@cqu.edu.cn。
  • 基金资助:
    国家自然科学基金(52005062);国家重点研发计划(2018YFB1701205);中央高校基本科研业务费(2016CDJXY)

Carbon Consumption Prediction Method of Gear Hobbing Based on Data Migration in Shared Manufacturing

YI Qian1,2;LUO Yusong2;HU Chunhui2;ZHUO Junkang2;LI Congbo1,2;YI Shuping2   

  1. 1.State Key Laboratory of Mechanical Transmission,Chongqing University,Chongqing,400044
    2.College of Mechanical and Vehicle Engineering,Chongqing University,Chongqing,400044
  • Online:2023-08-10 Published:2023-08-14

摘要: 针对一般制造企业生产智能化程度不高、单个企业难以收集足够加工数据的问题,提出了一种共享制造下企业间数据迁移的滚齿加工碳耗预测方法。分析滚齿加工数据特征,在共享制造环境下提出用改进TrAdaBoost算法的数据迁移方法融合企业间的滚齿加工碳耗数据,形成跨企业联合数据集;利用蜻蜓算法优化支持向量回归,构造了跨企业滚齿加工碳耗预测模型。通过案例分析验证了所提方法的有效性,其预测性能在数据量少、数据关联性差的情况下具有优势,平均相对误差和决定系数分别比传统算法平均提高59.23%和16.56%。

关键词: 共享制造, 滚齿加工, 迁移学习, 碳耗预测

Abstract: Aiming at the problems that general manufacturing enterprises were not intelligent enough and it was difficult for a single enterprise to collect enough processing data, a prediction method of carbon consumption in gear hobbing processes was proposed based on data transfer between enterprises under shared manufacturing. The data characteristics of gear hobbing processes were analyzed, and the data migration method of improved TrAdaBoost algorithm was proposed to integrate the carbon consumption data of gear hobbing processes among enterprises under the data sharing manufacturing to form a cross-enterprise joint data set. The Dragonfly algorithm was used to optimize the support vector regression to construct a cross-enterprise carbon prediction model for gear hobbing processes. The effectiveness of the proposed method was verified by case analysis. Predictive performance has advantages when the data volume is small and the data correlation is low. Compared with the traditional algorithm, mean absolute percentage errors and coefficient of determination are improved by 59.23% and 16.56% respectively. 

Key words: shared manufacturing, gear hobbing, transfer learning, carbon consumption prediction

中图分类号: