Scholarlink

Accurately identifying the true drivers of production stoppages is critical for improving availability, maintenance prioritization, and throughput in discrete multi-stage manufacturing systems. In serial lines with finite buffers, disturbances propagate with time delays, often causing downstream blocking far from their origin. Conventional diagnostic approaches, such as OEE-based classification and bottleneck detection heuristics, do not account for this lag and therefore misattribute symptoms to non-causal stations. This study introduces a proportional multi-source root-cause attribution method that distributes blocked time among all feasible upstream internal disturbances occurring within buffer-dependent propagation windows. Contributions are weighted by disturbance duration and historical downtime, reflecting cumulative production loss rather than event proximity. The method is evaluated using a FlexSim discrete-event simulation of cosmetic packaging. Compared to a simplified single-source backtracking approach, the proposed method significantly improves causal resolution. In the studied case, blocked time attributed to the filler via the cartoner decreased from 9.56 min to 8.27 min, with reductions exceeding 75% in selected scenarios. The method successfully captures the influence of remote and short-duration micro-stoppages ignored by simpler logic. These results demonstrate that accounting for distributed causality yields more realistic diagnostic output suitable for integration with MES/SCADA and digital-twin systems.

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

Modeling the Root Causes of Production Line Stoppages by Leveraging Downtime History and Buffer Dynamics: Simulation-Based Evaluation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tomasz Bartkowiak,
  • Jędrzej Gembiak,
  • Jan Dąbrowski,
  • Dariusz Dąbrowski

摘要

Accurately identifying the true drivers of production stoppages is critical for improving availability, maintenance prioritization, and throughput in discrete multi-stage manufacturing systems. In serial lines with finite buffers, disturbances propagate with time delays, often causing downstream blocking far from their origin. Conventional diagnostic approaches, such as OEE-based classification and bottleneck detection heuristics, do not account for this lag and therefore misattribute symptoms to non-causal stations. This study introduces a proportional multi-source root-cause attribution method that distributes blocked time among all feasible upstream internal disturbances occurring within buffer-dependent propagation windows. Contributions are weighted by disturbance duration and historical downtime, reflecting cumulative production loss rather than event proximity. The method is evaluated using a FlexSim discrete-event simulation of cosmetic packaging. Compared to a simplified single-source backtracking approach, the proposed method significantly improves causal resolution. In the studied case, blocked time attributed to the filler via the cartoner decreased from 9.56 min to 8.27 min, with reductions exceeding 75% in selected scenarios. The method successfully captures the influence of remote and short-duration micro-stoppages ignored by simpler logic. These results demonstrate that accounting for distributed causality yields more realistic diagnostic output suitable for integration with MES/SCADA and digital-twin systems.