Development of a framework for adaptive operation and maintenance strategies by means of artificial intelligence

• Applicant: Dr.-Ing. Martin Dazer
• Subject Area: Engineering Design, Machine Elements, Product Development
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 532373244

For manufacturers and operators of technical systems, topics such as resource efficiency, sustainability, cost efficiency, safety, but also availability and customer satisfaction are of great importance. Current operation and maintenance strategies require a definition of the operation and maintenance conditions before the system is deployed in order to act as efficiently as possible. However, these conditions are highly variable and vary widely between different markets. In addition, information or data on the conditions is often sparse or incomplete. An operation and maintenance strategy that adapts as ideally as possible to the prevailing situation and its boundary conditions provides a promising approach. However, the current state of research usually considers operation and maintenance strategies separately and only optimizes them individually and for a defined application scenario. As a result, considerable potential remains untapped. This can manifest itself, for example, in overloading of maintenance capacities, uneconomical spare parts management, incorrect maintenance intervals, incorrect maintenance strategy, etc. Possible consequences of the unused potential are, among others, too low availabilities and a worse environmental impact of the developed product. From this, the necessity of a systematic approach can be derived, which includes approaches and methods of adaptive operation strategies using the fundamentals of prognostics as well as adaptive maintenance strategies using the fundamentals of maintenance and their interaction through reinforcement learning. The synergistic use of these methods is summarized in this application as a framework. The overall goal is therefore the development of such a framework for adaptive operation and maintenance strategies, realized in a hybrid network by means of a reinforcement learning agent. The agent forms the central link to unlock the synergies. This allows the two strategies to be modeled multivariate and optimized, achieving the global optimum for a given solution space. The remaining service life can thus be optimized synergistically with the maintenance strategy (intervals, spare parts inventory, etc.). In the global optimum, the ideal construct of operation and maintenance strategy is present, which maximizes the target value (e.g. availability) for the solution space. In addition to high availability, low maintenance costs and low negative environmental impacts can be used as additional constraints in the optimization.

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr.-Ing. Bernd Bertsche