Instance-Specific Production Control in Circular Manufacturing Using Gaussian Mixture Models

Circular manufacturing systems are characterized by high uncertainty, as the condition of each returned product is individual, resulting in stochastic pro-cess behavior that directly affects production control decisions. Deterministic routing rules and fixed process parameters, which are common in linear pro-duction, are therefore insufficient. This paper proposes a probabilistic pro-duction control framework for circular factories based on Gaussian Mixture Models (GMMs). Products and their components are represented as probabil-istic entities, with their process-related characteristics modeled using GMMs trained on historical shopfloor data, prior process steps for the specific in-stance, and inspection results. For each product and component instance, the models predict the process duration, the probability of successful execution, and the risk of failure or damage. These probabilistic predictions enable in-stance-specific decisions on process and station selection, as well as the adaptive choice between direct processing and intermediate inspection steps. By explicitly capturing multi-modal condition states and uncertainty, the framework enables risk-aware scheduling and routing decisions and improves robustness and responsiveness compared to deterministic control approaches. ... mehrThe proposed method thus operationalizes data-driven uncertainty modeling for adaptive production control in circular manufacturing systems.