A New Approach for The Estimation of Reliabilities Based on Repair Times Using Simulation Technique

Abstract

This paper concentrates on a formal procedure for testing the goodness of fit of underlying theoretical probability distribution and evaluation of reliability indices for complicated systems like power generators. It is assumed that the repair times of these systems follow an exponential failure model with probability distribution function F (x). In view of the assessing goodness of it for the underlying distribution, directly on the repair hours, we find large chi-squire values and hence is not a good fit. To avoid this situation, a new procedure is introduced by using simulation technique. Failures are divided into two types of failures based on the repair time duration namely, (1) minor repairs and (2) major repairs based on a threshold value T.. A binomial probability model is considered for above two types of failures and carried out Chi-square test for goodness of it for the collected data. Based on the chi-square values, determined the Ideal threshold value which has smallest chi-square calculated value. Based on these identified ideal threshold value, reliabilities were calculated for each unit. Finally, reliability indices are calculated and presented graphically for the units under consideration and conclusions were drawn based on the results obtained.