Evaluation of a discrete modification of the continuous AMSAA reliability growth model

Abstract

A new discrete reliability growth model is created by modifying the often used Army Material System Analysis Activity (AMSAA) continuous reliability growth model. The new model is labeled the AMSAA-D model. Its accuracy is evaluated and compared with three other existing discrete reliability growth models. In particular it is more accurate than an AMSAA discrete model which requires computer supported numerical methods to calculate the reliability estimates from test data. The AMSAA-D reliability estimates can be made with a hand-held calculator. Computer simulations were used to generate test data needed for the evaluation. The simulated test plan assumes that repeated tests on a system are performed until a predetermined number of failures occur, at which time a design change is made to the system to improve its reliability. Ten reliability values are specified to define a reliability growth pattern. Five hundred replications of each growth pattern are simulated. For each replication, reliability estimates are calculated for each of the ten sets of generated test data using equations from each of the ten sets of generated test data using equations from each of the four growth models. Averages and sample mean square error values across the 500 replications are used to determine accuracy. Sensitivity of the AMSAA-D model to the number of failures before system modification and to the number of possible failure causes is system modification and to the number of possible failure causes is also evaluated. Theses. (SDW)