A new way to detect problems in complex processes involving many interacting parts
Dominique Sommers developed methods to detect hidden problems in complex processes with many interacting parts, helping organizations improve performance.
In many real world systems, such as package delivery or patient care, processes involve multiple people, tasks, and objects working together. These interactions make it difficult to detect when and why things go wrong. Traditional methods often miss problems that arise between connected parts of the process.
PhD researcher Dominique Sommers developed a new approach that looks at the full system rather than individual parts, making it possible to uncover hidden problems and better understand how complex processes can be improved. He defended his thesis on Thursday, May 15.
Understanding how things go wrong
Organizations use process models to describe how their systems should work, and event logs to record what actually happens. But these two often do not match.
Process models are simplified and may miss unusual situations. Event logs may contain mistakes due to human error or technical issues. Comparing the two is important but difficult, especially when many people or systems are involved.
approach focused on the full picture. Instead of analyzing each person or object on its own, hos method studied how everything interacts.
This made it easier to see where the system failed, what caused the failure, and which parts were affected. For example, if one courier changed their route, it might delay other deliveries. His method made it possible to detect this kind of impact.
Creating realistic test data
To test these ideas, Sommers developed a new way to create realistic synthetic data. This data included both unusual behavior and typical recording mistakes, which are common in real life but missing from most test data.
With this, he was able to test his methods in realistic situations and show that they revealed problems that traditional tools often missed.
Sommers' data generation method used patterns that could be adjusted or expanded, so that different kinds of errors or behavior could be added depending on what was needed.
This made it a useful tool not only for testing, but also for training and improving other process analysis techniques.
A clearer view of complex systems
By looking at the entire system and how its parts affected one another, ³§´Ç³¾³¾±ð°ù²õ’ research provided a deeper and more accurate way to detect problems in complex processes.
His methods could be used to understand issues in both the process design and the way it was recorded, leading to better decisions and more reliable systems.
PhD researcher Dominique Sommers. Photo: Angeline Swinkels
Bringing Clarity to Complicated Workflows
Dominique ³§´Ç³¾³¾±ð°ù²õ’ work provides a strong foundation for developing advanced tools that help analyze and improve the way tasks are performed in complex and realistic environments.
By examining how different parts of a system interact with one another, his approach supports better decision making, smoother workflows, and greater overall reliability.
This has valuable applications in fields such as logistics, health care, and any area where many interdependent steps must function together in a coordinated and efficient manner.
-
Supervisors
Boudewijn van Dongen, Natalia Sidorova
Written by
More on AI and Data Science
Latest news
