| 引用本文: | 陆志强,朱宏伟,廖怡娜.考虑质量预测的前摄调度问题模型与算法[J].哈尔滨工业大学学报,2020,52(7):96.DOI:10.11918/201908116 |
| LU Zhiqiang,ZHU Hongwei,LIAO Yina.Modeling and algorithm for proactive scheduling problem considering quality prediction[J].Journal of Harbin Institute of Technology,2020,52(7):96.DOI:10.11918/201908116 |
|
| 摘要: |
| 为有效描述装配人员水平、工件质量等因素对飞机装配作业质量的影响,并为飞机装配过程建立合理的前摄调度计划,提出支持向量回归（SVR）预测模型和两阶段循环迭代搜索算法. 采集相关历史质量数据,以装配人员水平和工件质量等数据为输入,作业质量为输出训练SVR预测模型. 基于已训练的SVR预测模型,采用基于作业列表禁忌搜索框架对作业列表进行邻域搜索,并通过内嵌人员分配搜索模块的串行调度实现人员配置的优化. 数值实验结果表明:采用SVR预测模型求得的作业质量预测值相较实测值的误差能够控制在5%以内,预测精度最高达到97.38%；两阶段循环迭代搜索算法求解所得模板计划与CPLEX偏差均值保持在9.99%~ 27.54%,与其他前摄调度生成方法相比偏差最小；在不确定性环境中,右移算法在两阶段循环迭代搜索算法所得模板计划中能取得最优或次优的平均装配工期和平均计划偏差. SVR预测模型能够对飞机装配作业质量进行有效预测,而两阶段循环迭代搜索算法则能满足构建飞机装配前摄调度计划的需求. |
| 关键词: 飞机装配 质量预测 人员配置 支持向量回归 禁忌搜索 |
| DOI：10.11918/201908116 |
| 分类号:F273 |
| 文献标识码:A |
| 基金项目:国家自然科学基金(61473211) |
|
| Modeling and algorithm for proactive scheduling problem considering quality prediction |
|
LU Zhiqiang,ZHU Hongwei,LIAO Yina
|
|
(School of Mechanical and Energy Engineering, Tongji University, Shanghai 201804, China)
|
| Abstract: |
| To effectively describe the influence of factors such as the assembly personnel level and the workpiece quality on the job quality of aircraft assembly and establish a reasonable proactive schedule for the assembly process, a support vector regression (SVR) prediction model and a two-level iterative search algorithm are developed. Firstly, with collecting relevant historical quality data, a SVR prediction model is trained by taking the data of assembly personnel level, workpiece quality and so on as input and job quality as output. On the basis of the trained SVR prediction model, a job list-based tabu search framework is adopted to search the neighborhood of job list, and the optimization of personnel allocation is achieved through the serial scheduling generation scheme with embedded personnel assignment search module. The results of numerical experiments show that the predicted value of job quality obtained by the SVR prediction model can be controlled within 5% in comparison with the measured value, and the highest prediction accuracy is 97.38%. The mean deviation between the two-level iterative search algorithm and CPLEX is between 9.99% and 27.54%, which is the smallest among proactive scheduling generation methods. In the uncertain environment, the right shift algorithm can obtain the optimal or sub-optimal mean makespan and mean deviation in the baseline schedules obtained by the two-level iterative search algorithm. The SVR prediction model can effectively predict the job quality of aircraft assembly, and the two-level iterative search algorithm can meet the requirement of constructing proactive scheduling for aircraft assembly. |
| Key words: aircraft assembly quality predication personnel arrangement support vector regression tabu search |
