A Framework for Data-Driven Scehduling of Resource-Constrained Processes Based on Event Logs

 

Student: Philipp Alexander Heisenberger

Title: A Framework for Data-Driven Scehduling of Resource-Constrained Processes Based on Event Logs

Supervisor: Dr. Sebastiaan van Zelst

1st Examiner: Prof. Dr. Wil van der Aalst

2nd Examiner: Prof. Dr. Wolfgang Prinz

Summary:

Efficient scheduling of manufacturing processes is crucial for optimizing productivity and meeting customer demands. Traditionally, scheduling in manufacturing relies on spreadsheets and domain knowledge. However, these conventional approaches often fail to capture the intricate dynamics inherent in manufacturing processes, resulting in schedule deviations and inefficiencies. To address this gap, this research proposes using event data and process simulation as a tool for identifying improvement opportunities in scheduling scenarios within manufacturing processes. By employing process simulation techniques, the complex dynamics of manufacturing processes can be better understood and incorporated into the scheduling process. Successful implementation of process simulation requires careful consideration of the inherent complexities, necessitating a balance between accuracy and feasibility through appropriate assumptions and simplifications. This study addresses the limitations of existing research on simulation model building, which often overlooks the challenges inherent in manufacturing processes. We extend current approaches by explicitly focusing on aspects such as data discovery, data quality, and the assessment of manufacturing-specific control-flow and performance aspects. Further, our enhanced approach is applied to a real-world data set obtained from a <<Identity Undisclosed>> company, serving as a comprehensive case study. We expect that increased transparency on process characteristics, particularly process performance, facilitates a better identification of scheduling heuristics that lead to increased throughput. Our results demonstrate that simulation models provide reliable improvement insights for scheduling scenarios. In particular, in our case study, we identified product and capacity dependent scheduling constraints that can be incorporated into the scheduling process. Despite the increased understanding, capturing the complex dynamics of manufacturing processes remains a challenging task. Consequently, several assumptions and simplifications are made to tackle this complexity.

---

This thesis was performed under an NDA. For inquiries regarding the thesis, please contact the thesis supervisor.