Process mining analyzes event data to improve real-world processes. The use of various process models, each with distinct representational biases, necessitates transformations between them to leverage unique strengths and tool support. The emerging field of object-centric process mining predominantly lacks such methods. This thesis addresses this gap by proposing bidirectional transformations between object-centric causal nets, a newly proposed formalism, and object-centric Petri nets, a well-established process model. This work contributes formal definitions of the transformations, formal proofs of their correctness, a publicly available implementation, a qualitative evaluation, and play-out and replay procedures for both process models. An object-centric Petri net is transformed into a behaviorally equivalent object-centric causal net. Conversely, an object-centric causal net is transformed into an object-centric Petri net that underfits the initial net. Our qualitative evaluation shows that the degree of underfitting is directly related to the structure of the initial object-centric causal net, particularly the number of activities having markers allowed to consume a variable amount of obligations.
Ole Kuhlmann (Wed,) studied this question.