The thesis is embedded in the Versioned Planning initiative at Mercedes-Benz Mercedes-Benz Manufacturing Engineering (MO/ET). In our center, we contribute to the digital transformation with initiatives such as the MO360 platform or the digital twin inside the omniverse. Furthermore, we integrate engineering processes with these new and AI-driven capabilities. Your thesis in the team “MO360 Engineering AI & Data Management” contributes directly to the long-term vision of a business end of the Semantic Layer for MO/E as the foundation for AI Native Engineering and Agent2Agent orchestration.
Object-Centric Process Mining (OCPM) in Celonis provides a powerful, quantitative view on planning processes — it reveals how often, how long, and in which variants activities are executed across multiple object types. However, in complex automotive production planning (e.g., Mercedes-Benz MO/E), the resulting Process Intelligence Graph remains largely descriptive: it answers "how much?" but not "why?". The semantic context — which scenario, premise, milestone, or review order (Prüfauftrag) triggered a given planning iteration — is not natively captured in event logs.
Our approach currently being rolled out as the methodological backbone, explicitly models scenarios and specifications. It therefore provides exactly the semantic information that OCPM lacks and is envisioned as the foundation of a Semantic Layer for MO/E.
The thesis investigates how we can complement OCPM by adding a semantic "why" layer on top of the quantitative "how much" delivered by Celonis. The goal is to design, prototype and evaluate a concept that links object-centric event data from Celonis with the version- and scenario-semantics from our software, enabling explainable, scenario-aware process intelligence in production planning.
Possible Research Questions (to be discussed and aligned your university)
Which structural and semantic gaps exist in the Celonis OCPM representation of the current Mercedes-Benz planning process?
Which semantic concepts of our approach can be formalized as a Semantic Layer (e.g., as an ontology / knowledge graph)?
How can this Semantic Layer be technically integrated with Celonis OCPM (e.g., via the Process Intelligence Graph, AI Annotation Builder, or external graph alignment) to enrich object-centric events with planning rationale?
To what extent does the enriched representation improve explainability, scenario awareness and impact analysis compared to a baseline OCPM model?
Expected Contribution
A formalized Semantic Layer concept for versioned planning, bridging OCPM and engineering semantics
A prototype demonstrating the integration of eVMS semantics with Celonis OCPM
Empirical insights into the added value of semantic enrichment for impact analysis, scenario steering and autonomous planning agents in the MO/E context. In simple words, extraction of some useful KPIs and steering concepts for management
The activity can begin from September (or October).
The final thesis selection is made in close consultation with you, the university and us.
