Comparing Probabilistic Methods for the Asset Management of Distributed Items

Markov models have met with widespread success when used to determine asset management strategies for infrastructure systems such as pavements, bridges, and electricity and water networks. However other probabilistic models could be chosen. This paper compares the performance of the Markov model with two of these, the semi-Markov model and the delay time model. Both these models let the transition probabilities between states depend on the time already spent in the state. This is an attractive feature as many degradation situations have it. The three models are compared on two data sets derived from measurements carried out on 11 kV transformers and switchgear. As full condition histories are required to know how the predicted costs of a proposed asset management policy compare with the actual costs that would be obtained in practice, a method for simulating these condition histories was developed. All the models performed well, but the semi-Markov model was generally significantly better. Particularly noteworthy was the fact that this was true even when there were a limited number of observations.