Sensitivity analyses for preventive-maintenance models

Models for determining preventive-maintenance strategies usually make assumptions of constancy regarding lifetime distributions, cost functions, temporal effects, down-time durations, and preventive-maintenance intervals. In practice, extraordinary sources of variability arise, leading to inaccurate information about parameters, probabilities, predictions, and decisions. This paper examines these errors, using simulation studies for delayed renewal processes with Weibull lifetime distributions. These sensitivity analyses demonstrate the importance of allowing for extraordinary sources of variability when deciding on optimal strategies for preventive maintenance.