Estudio de la fiabilidad de los sistemas reparables y desarrollo de un procedimiento de análisis multivariante

Abstract

For the modeling and analysis of the reliability of repairable systems the scientific community provides various statistical methodologies developed in the last few decades. The methods of greater acceptance and employment have been collected by the International Electrotechnical Commission (IEC) in its Technical Committee TC56 Dependability, through the issuance of standards for the European space, taking into account the military handbooks (MIL-HDBK) issued by the U.S. Department of Defense. IEC 60300-3-5 collects statistical methods and analysis procedures applicable. The failure intensity of a repairable system can be estimated using the successive times between failures, using a Stochastic Point Process. If there is no trend and the times between failures are distributed exponentially, then failure intensity is constant and in this case can be modeled by a Homogeneous Poisson Process. In cases where there is a tendency in the failure intensity, is modeled with a Non-Homogeneous Poisson Process (NHPP) and the standard includes applying the Power Law Process. IEC 60605-6 develops the contrasts to verify the existence or not of trend in the failure intensity by means of the statistic U, both for a repairable system, such as for an identical set of repairable systems. In the event that there are no trend, is the point estimate of parameters and confidence intervals, alleged an exponential distribution with the standard IEC 60605-4. If there is a trend in the failure intensity the estimate is made of the model parameters of the Power Law Process with IEC 61710. It has been applied the methodology proposed by IEC to the electric traction system in 3 series of railway trains of Metro de Madrid, on the available data of operation for a period exceeding 10 years, and in order to compare the results obtained, has been made on the same study of a group of 40 escalators of Metro de Madrid with 10 years of operating data, propose below a complementary and multivariate analysis to characterize the failure intensity obtained and the influence of repetitive failures.