Scholarlink

This study investigates the impact of variable sub-lot sizing on hybrid flow shop (HFS) scheduling systems, addressing the dual objectives of minimizing makespan (production time) and tardy jobs (enhancing customer satisfaction). A mixed-integer nonlinear programming model is developed that consistently outperforms benchmark approaches, particularly for scenarios with low job counts and high maximum batch sizes. The relative percentage improvement in makespan ranges from 0% to 239%, with superior machine utilization demonstrated across all test scenarios. Through comprehensive statistical analysis, four key factors are identified as significant: number of jobs, setup times, maximum sub-lots, and weighting preference. Machine learning analysis using XGBoost (achieving 98.8% prediction accuracy) confirms that all factors significantly impact performance, with job count and maximum sub-lots being the most critical parameters. Optimal operating conditions are determined to be 3 jobs, 14-min setup time, 4 sub-lots, and a 0.1 weighting factor, resulting in zero tardy jobs and a 162-min makespan. This configuration effectively balances production efficiency, resource utilization, and customer satisfaction in manufacturing environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning Analysis of Lot Streaming Hybrid Flow Shop Scheduling with Variable Sub-lot Sizes: Impacts on Makespan and Tardiness

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Heba I. Elkhouly,
  • Enas Ahmed Zaky,
  • Ali Saeed Almuflih,
  • Mohamed Mahmoud Samy,
  • Shimaa Barakat

摘要

This study investigates the impact of variable sub-lot sizing on hybrid flow shop (HFS) scheduling systems, addressing the dual objectives of minimizing makespan (production time) and tardy jobs (enhancing customer satisfaction). A mixed-integer nonlinear programming model is developed that consistently outperforms benchmark approaches, particularly for scenarios with low job counts and high maximum batch sizes. The relative percentage improvement in makespan ranges from 0% to 239%, with superior machine utilization demonstrated across all test scenarios. Through comprehensive statistical analysis, four key factors are identified as significant: number of jobs, setup times, maximum sub-lots, and weighting preference. Machine learning analysis using XGBoost (achieving 98.8% prediction accuracy) confirms that all factors significantly impact performance, with job count and maximum sub-lots being the most critical parameters. Optimal operating conditions are determined to be 3 jobs, 14-min setup time, 4 sub-lots, and a 0.1 weighting factor, resulting in zero tardy jobs and a 162-min makespan. This configuration effectively balances production efficiency, resource utilization, and customer satisfaction in manufacturing environments.