Project Details
Projekt
Prognostic Model Selection for Predictive Maintenance and an Integrated Reinforcement Learning-based Production Scheduling in Dynamic Manufacturing Systems
Applicant
Professor Dr.-Ing. Michael Freitag
Subject Area
Production Systems, Operations Management, Quality Management and Factory Planning
Term
since 2025
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 547349768
In the course of digital transformation, approaches for data-based and integrated production and maintenance planning are becoming increasingly important. Building on the findings of a recently completed joint project and the current literature, this new research project aims to develop a self-adaptive selection procedure for forecasting models for condition-based maintenance and combine it with reinforcement learning-based production planning in dynamic production environments. Previous work has identified discrepancies between production logistic metrics and conventional error-based metrics, which complicate the selection of suitable prediction models in this application area. The aim of the project is therefore to develop a method that takes these constraints into account in the automated selection of forecasting models. In addition, the selection method for unsupervised multivariate anomaly detection in sensor data is to be extended in such a way that labels (labelling the true state of a machine), which in practice have to be procured at considerable expense, can be dispensed with. Furthermore, the suitability of logistical KPIs as criteria for improving the selection process will be analyzed. In addition, the development of a method for integrated production and maintenance planning based on reinforcement learning represents a promising approach for improving decision-making and meeting the requirements of dynamic production environments.
DFG Programme
Research Grants
International Connection
Brazil
