Schedule-based execution bottleneck identification in a job shop

• We identify schedule-based execution bottlenecks at the execution level in a job shop.
• We present an identification method and evaluate it by computational studies.
• We show that execution bottlenecks often differ from planning bottlenecks.
• We evaluate the impacts of different indicators on the bottleneck identification.
• Case studies show execution bottlenecks converge and stabilise when improving schedules.

This paper aims to identify execution bottlenecks based on a specific schedule in a job shop. An execution bottleneck refers to a machine that dominates the scheduling performance of production systems in the strongest manner at the execution level. To identify such bottlenecks, a two-layer framework is proposed, in which a job shop scheduling problem is solved using a modified immune algorithm (IA_ADO), and then a multi-attribute bottleneck identification (MABI) method is introduced to identify the execution bottleneck based on the obtained schedule. The framework is implemented and tested on 24 job shop scheduling benchmarks. We show that IA_ADO is able to return optimal or near optimal schedules. The bottleneck identification results demonstrate that the average uninterrupted active duration plays a dominant role amongst the three bottleneck attributes. Furthermore, our results show that the execution bottleneck often differs from the planning bottleneck. This finding indicates that the current practice of using a planning bottleneck to produce a schedule might be inadequate for shop-floor control. In addition, case studies show that the execution bottlenecks converge quickly into specific machines when the schedules returned by IA_ADO move towards the optimal solutions. This finding has a high practical value, as the optimal schedules are often difficult to find in practice.