A Multiple Attribute Evaluation of Bayesian Availability Estimators

Three system availability estimation methods, maximum likelihood, traditional Bayes, and Brender´s Bayes were used to obtain estimates for two exponentially distributed data sets. From the simulation results, a best estimation method was chosen through the use of five multiple attribute decision making (MADM) techniques: dominance, simple additive weighting, linear assignment, ELECTRE, and TOPSIS. In terms of the five criteria which are: closeness to steady state, variability between samples, computer execution time, ease of programming, and ease of understanding; and their importance (weight) on the final decision, Brender´s Bayes estimation was superior to the other two.