Bayesian Estimation of Constant Failure Rate and Unavailability

The paper presents a simple procedure for estimating and updating the time-independent failure rate and the constant unavailability on demand of an item. The procedure is Bayesian with binomial likelihood and beta prior. The binomial distribution models the underlying failure process. The beta distribution as prior is rationally justified. The Bayesian method is compared with the s-confidence interval and fiducial distribution methods. A numerical example concerning failure data of nuclear plants is discussed briefly.