Cost modelling in maintenance strategy optimisation for infrastructure assets with limited data

Our paper reports on the use of cost modelling in maintenance strategy optimisation for infrastructure assets. We present an original approach: the possibility of modelling even when the data and information usually required are not sufficient in quantity and quality. Our method makes use of subjective expert knowledge, and requires information gathered for only a small sample of assets to start with. Bayes linear methods are adopted to combine the subjective expert knowledge with the sample data to estimate the unknown model parameters of the cost model. When new information becomes available, Bayes linear methods also prove useful in updating these estimates. We use a case study from the rail industry to demonstrate our methods. The optimal maintenance strategy is obtained via simulation based on the estimated model parameters and the strategy with the least unit time cost is identified. When the optimal strategy is not followed due to insufficient funding, the future costs of recovering the degraded asset condition are estimated.