The reliable and safe operation of rolling stock requires huge investment in maintenance. To protect the investment and the sustainability of rolling stocks, it is vital to maintain the assets as economically and efficiently as possible. We use business analytics to help identify the optimal maintenance strategy with the least whole life cost while maintaining safety.

Historical data is needed in defining asset relationships; the timings of different interventions are required for optimizing the maintenance strategy, together with the capital and operational expenditure costs for the range of interventions. Ideally, this information should be based on a rolling stock’s historical failure and maintenance data. As failure data is rarely available, we put focus on the maintenance data from the asset management information systems. However, our client has a low level of maturity in asset management; the historical maintenance records are not stored electronically, which is extremely time consuming and costly to interpret. This gives great difficulty to obtain the required information. 

Given the limited availability of the information, we have decided to use a statistical estimation of costs. We investigate a small sample of rolling stocks with detailed information as well as use structured elicitation techniques to extract experts’ knowledge (both general and local) based on the experiences they have gained from routine data collection practices. We then adopt Bayes linear methods to combine the sample data with the experts’ knowledge, to generate the estimation. Finally, we build an asset cost model based on the estimation, optimize the maintenance strategy and estimate the penalty for delaying maintenance when funding is not sufficient.