A non-parametric approach for modelling reliability growth of complex repairable systems

For a complex repairable system, the identified failure modes during a reliability growth test program are not always fixable. It is a challenging issue to estimate the reliability of the system based on the test observations and corrective actions. This paper presents a non-parametric approach for evaluating the system reliability. The proposed approach incorporates partial information about the effectiveness of corrective actions. Using the partial information, the observed data set is revised, which is then fitted into a distribution model. The MTBF is estimated based on the fitted model. The proposed approach is illustrated by an example.