Estudo comparativo de testes de tendência na confiabilidade de sistemas

Abstract

In the development and validation of new products, statistical analyzes are often considered to evaluate the existence of a tendency in the process of failure and realiability of prodution. These analyzis aim to determine if, and how, the failure pattern is evolving over time, and its results are decisive for the success of the product design phase. On the other hand, the knowledge of the temporal behavior of the failure occurrence pattern of a repairable equipment is a central aspect in the maintenance policy perspective. This implies the need for statistical techniques to evaluate the existence of a tendency in the process of failure. There are, described in the literature, several trend evaluation tests. Each one with its application specificities. In this work we present a review on a set of statistical tests of reliability trend evaluation. A large Monte Carlo simulation study, involving single systems, is carried out to compare the performance of the different tests reviewed, both in several reliability scenarios and in the presence of different types of samples. The results seem to indicate that the Anderson & Darling test is indicated for general use because it is robust for most of the considered scenarios and is not sensitive to the sample size