A Multi-Objective Approach for Railway Maintenance Planning Based on Combined Pruning Methods

Maintenance planning for railway networks is an exacting task, as it directly affects a wide range of stakeholders and entails multiple objectives. Further complications arise when maintenance strategies have to be identified at asset-type level but their corresponding performance is determined at network level. Although the problem can be formulated as a multi-objective optimization model, it usually leaves the decision maker (DM) with many solutions on a Pareto front. To overcome this issue, this paper proposes a three-step approach. In the first step, NSGA-II is adopted to determine the Pareto optimal set. The latent profile analysis (LPA) is then employed to group similar solutions together in the second step. To help the DM finalize the decision, the data envelopment analysis (DEA) method is employed in the final step to rank the solutions in a selected group by their relative efficiency. Application of our model to a real-world case study allows us to demonstrate the capability of the proposed approach. The major advantage of LPA resides in its flexibility in dealing with various shapes of Pareto front, while the use of the combined pruning method can mitigate the risk of obtaining an undesirable solution in extreme regions of the Pareto front