Analysis of heavily interval-censored data from- A fleet of repairable systems

Abstract

In repairable system reliability analysis, it is quite common that only partial failure history of a system is available. Very often failure data are interval censored (i.e. left and right censored) in that only failures within particular periods are recorded. This poses a special challenge to engineers in data analysis and decision making. In this paper, faced with a large collection of heavily interval censored field data of a fleet of repairable systems, we propose a two-phase analysis (TPA) to perform the reliability analysis. The analysis is done through two critical ratios, the failure dispersion ratio (FDR) and the average repair per system (ARS). Compare to the well-established graphical method based on the mean cumulative repair function (MCRF) plot, the proposed TPA enables the management to quickly capture the key characteristics of the facilities under their charge. In particular, we take into account in our analysis the heterogeneity among systems as well as the ineffective maintenance, which are two important issues in practice.