Data visualization for reliability analysis of repairable systems.

The objective of this paper is to find out possible patterns of failure occurrences in a repairable system. We develop a graphical exploratory tool and perform visual inference considering non-parametric local linear kernel estimators for the rate of occurrence of failures (ROCOF) and its first derivative. The shape characteristics of the ROCOF function are distinguished from those which are merely an artefact of the sampling variability of the data through the construction of confidence intervals for the first derivative. The proposal is illustrated with several real data sets, and its performance is evaluated through an extensive simulation study.