Agent-Based Optimization of Maintenance Planning for Railway Vehicles



Published Jul 22, 2020
Julian Franzen Udo Pinders Bernd Kuhlenkötter


In practice, maintenance of rail vehicles is based on reactive and preventive maintenance strategies. Condition-oriented maintenance approaches are only slowly finding their way into the market. When researching the state of the art, it is noticeable that the majority of the approaches presented is considering individual components - the system focus necessary for maintenance optimization is not taken into account.

Depending on the target system (number of components) and planning period, a complex optimization problem (OP) results. The OP is an NP-heavy problem for which the use of Genetic Algortihms can deliver suitable solutions for small search spaces. When applying it on a complex system with a larger solution space, this heuristical approach alone is not sufficient for the analytical optimization of a system representing a locomotive.

Therefore, in this paper agent-based distributed problem solving is applied to analytically optimize the maintenance of the target system. Therefore, a multi-agent system (MAS) based on the O-MaSE-model will be developed, which captures the configuration of a target system and formulates the overall OP using the fictional data from a drivetrain of a shunting locomotive as an example. Following the principle of co-evolutionary problem solving, the overall problem is divided into smaller subproblems (SP). These SP have the right size to be solved by an own agent using genetic algorithms. In addition to, the solution focuses on the autonomous negotiation of an acceptable solution for the entire system by the SP agents.

How to Cite

Franzen, J., Pinders, U., & Kuhlenkötter, B. (2020). Agent-Based Optimization of Maintenance Planning for Railway Vehicles. PHM Society European Conference, 5(1), 9.
Abstract 597 | PDF Downloads 280



Optimization, Multi-Agent System, Railway Maintenance

Technical Papers

Similar Articles

1 2 3 4 5 6 7 8 9 10 > >> 

You may also start an advanced similarity search for this article.