A Computational Intelligence Approach to Railway Track Maintenance

Railway track intervention planning is the processes of specifying the location and time of required maintenance and renewal activities. However, intervention planning may be a complex process due to the wide variety of track components used and their complex interactions which can result in highly complex component deterioration patterns. To facilitate intervention planning decision support tools have been developed to aid the permanent way engineer. Existing systems typically use an expert system based approach, with rules specified by track maintenance engineers. However, due to the complex, interrelated, nature of railway track component deterioration, using a rule based approach it is problematic for an engineer to consider all combinations of possible deterioration. Further, the rules may contain errors or omissions, which a static solution such as an expert system is unable to correct automatically.

To address these issues, this chapter describes, an approach to railway track intervention planning which uses a variety of computational techniques. The proposed system learns rules for maintenance from historical data and incorporates future data as it becomes available thus evolving over time to a more optimal state.