Analysis and evaluation of discrete reliability growth models with and without failure discounting

摘要

A survey of some evaluation work on discrete reliability growth models is presented. Extension of an accurate exponential growth model is provided that uses regression analysis to fit the natural logarithm of the failure rate 1-p in the geometric distribution. Some useful theorems and relationships are developed that provide estimates of reliability which have better properties than the usual maximum likelihood estimates. The effect of discounting is portrayed with graphs that allow comparison among different failure discounting methods and their affect on different models