Reliability prediction model of a multi-state system subject to general repair and maintenance with limited failure data

Effective maintenance management is essential to reduce the adverse effect of equipment failure to operation. This can be accomplished through accurately predicting the equipment failure such that appropriate actions can be planned and taken in order to minimize the impact of equipment failure to operation. This paper presents a development of model based on discrete time Markov process for a degraded multi-state system to assess the system reliability. The system degradation was quantified by discrete level of system's performance from perfect functioning state to complete failure. The optimum number of states is estimated using the combination of k-means cluster analysis and silhouette plot. At any point in time, the system can experience Poisson failure from any state upon which a minimal repair will be performed. The results prove the applicability of discrete time Markov in assessing the reliability of multi-state systems using the system performance data when there is very limited time to failure data.