现代制造工程 ›› 2026, Vol. 553 ›› Issue (10): 9-21.doi: 10.16731/j.cnki.1671-3133.2026.10.002

• 先进制造系统管理运作 • 上一篇    下一篇

考虑多扰动因素的航空发动机试验装配车间滚动调度*

蒋明杰1, 卢聪2, 王艺伟1, 唐文杰2, 刘杰2, 熊子涵1, 王峻峰1   

  1. 1 华中科技大学机械科学与工程学院,武汉 430074;
    2 中国航发四川燃气涡轮研究院,成都 610500
  • 收稿日期:2025-10-09 出版日期:2026-10-18 发布日期:2026-11-05
  • 通讯作者: 王峻峰,教授,博士,主要研究方向为生产系统工程、数字化智能化装配和人机协作与交互。E-mail:wangjf@hust.edu.cn
  • 作者简介:蒋明杰,博士研究生,主要研究方向为智能制造系统。
  • 基金资助:
    *中国航发四川燃气涡轮研究院稳定支持项目(GJCZ-0302-04)

Rolling scheduling for areo-engine test assembly workshops considering multi-disturbance factors

Jiang Mingjie1, Lu Cong2, Wang Yiwei1, Tang Wenjie2, Liu Jie2, Xiong Zihan1, Wang Junfeng1   

  1. 1 School of Mechanical Science and Engineering,Huazhong University of Science and Technology,Wuhan 430074,China;
    2 Sichuan Gas Turbine Research Institute,Aero Engine Corporation of China, Chengdu 610500,China
  • Received:2025-10-09 Online:2026-10-18 Published:2026-11-05

摘要: 航空发动机试验装配车间具有装配工艺多变、产品型号多样和扰动因素多发等特点。针对航空发动机试验装配中半成品发动机占用工位资源和识别扰动因素对车间的影响这2个难点,以最小化最大完工时间为优化目标,设计了一种考虑多扰动因素和车间实时状态的滚动调度框架。定义了多种扰动因素对车间状态的影响,提出了基于拓扑排序和改进的近端策略优化的深度强化学习调度方法。同时,引入了状态正则化、学习率衰减以及奖励缩放+Sigmoid平滑这3个改进点以增强调度方法的性能。围绕多种不同规模的航空发动机试验装配车间滚动调度场景,验证了所提方法和改进点的有效性。

关键词: 车间调度, 航空发动机, 多扰动因素, 试验装配, 深度强化学习

Abstract: Aero-engine test assembly workshops are characterized by variable assembly processes,diverse product models,and frequent disturbances. To address the two primary challenges of workstation resource occupation by semi-finished engines and identification of impact of disturbances on workshop operations in aero-engine test assembly,a rolling-horizon scheduling framework that incorporates disturbance factors and real-time workshop status was proposed,with the objective of minimizing the maximum completion time. The influence of various disturbances factors on the workshop state was systematically defined,and a deep reinforcement learning scheduling method based on topological sorting and improved proximal policy optimization was introduced. To enhance the performance of the scheduling approach,three improvements were implemented:state regularization,learning rate decay,and reward scaling combined with Sigmoid smoothing. The effectiveness of the proposed method and its enhancements were validated across multiple rolling-horizon scheduling scenarios of varying scales in aero-engine test assembly environments.

Key words: workshop scheduling, aero-engine, multiple disturbance factors, test assembly, deep reinforcement learning

中图分类号: 

版权所有 © 《现代制造工程》编辑部 
地址:北京市东城区东四块玉南街28号 邮编:100061 电话:010-67126028 电子信箱:2645173083@qq.com
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn