Bachelor Thesis - Visualization of Production Lines: Making the Digital Shadow Actionable



Tobias Brockhoff

Scientific Assistant


+49 241 80 21910




Today's businesses record an increasing volume of data about their internal processes with event data being a very common type of data. Event data is the digital footprint of individual business cases passing through multi-step processes. For example, consider a flight booking which, inter alia, requires the steps 'check booking' and 'confirm booking'. Each booking leaves a trace in the system constituting an event log, i.e., a collection of traces. Process mining is concerned with leveraging knowledge from these event logs to create process transparency and, finally, to improve the process.

Traditional process mining thereby focuses on processes that deal with virtual, non-physical objects, such as a booking or a fine. In this context, common topics are process discovery, which is concerned with data-driven model discovery, conformance checking, which asses the compliance between recorded and modeled behavior, or process model enhancement, for example enhancing models by performance metrics.

Recently, in the context of the excellence cluster Internet of Production at RWTH Aachen, a process structure that is specifically encountered in manufacturing came into focus. Backed by an increasing process data availability in manufacturing through the digital shadow, operational processes become an interesting field for applying process mining. Operational processes are characterized by a single main assembly process which sets the pace and has additional dependencies to subassembly processes. These subassembly processes are concurrent to the main line and their output is required for certain main assembly steps. For example, consider a car manufacturing process where the main assembly line starts with a raw chassis and ends with complete car. During the production, pre-built components, such as doors, built in separate subassembly lines, are assembled to the chassis. In contrast to the other business processes, ground truth models exist, and it can be assumed that the conformance between the model and the event log is high. For instance, subsequent assembly steps might strongly depend on each other, it is physically impossible to skip stations, or there are already control tools implemented, e.g., automatically counting the number of mounted bolts.

To monitor and optimize the process, a visualization of the current state of the production and how it potentially evolves with respect to preceding periods of time is of high value. Unfortunately, existing process metric visualization techniques do not properly support and respect the structure of operational processes. On the one hand, concurrency poses a big challenge for commercial tools that depend on directly-follows graphs. On the other hand, Petri net-based visualization quickly becomes cluttered due to the number of stations, and, moreover, it natively does not consider the difference between the main and subassembly line. For example, waiting times between the main and subassembly lines, caused by pre-build components being prepared long before they are assembled to the chassis, are generally not a problem and, therefore, should not dominate the visualization.

In this thesis, you will investigate and implement a visualization for operational processes which considers their specific properties. Given an event log, it should show the important Key Performance Indicators (KPIs) and their evolution. To this end, you will investigate methods for comparing KPIs at different time windows. Your final visualization technique should be implemented as a web service using PM4Py for processing the event data.


  • Knowledge of basic computer science concepts and interest in process mining
  • Good programming skills in Python
  • Optional: experience in using PM4Py
  • Interest in front end development (Angular)



Prof. Dr. Ir. Wil van der Aalst


Tobias Brockhoff (primary daily advisor)

Dr. Seran Uysal

For more Information

Send an e-mail to . Make sure to include detailed information about your background and scores for completed courses.