DISTRIBUTION LAW PARAMETERS EVALUATION CONSIDERING LEFT-TRUNCATED AND RIGHT-CENSORED DATA

Abstract

At the present stage of power industry development there is a need to improve nuclear power unit safety, reliability, and efficiency. To solve these problems, the mathematical and statistical reliability theory has been under development for the last couple of decades. The mathematical models representing the system behavior vs. time are being created. The reliability indicators are assessed with the information available through special tests or obtained in actual operation. The most objective information to determine the component reliability is the performance data since they reflect the actual unit operation, the factors it is affected to, and other features. The paper considers various statistical data obtained in service, their features, and modeling approaches. As an example, a complete, left-truncated and right-censored data modeling method is used because it is quite common in real life. A number of likelihood functions for the exponential, gamma, and Weilbull distribution are presented. A case study is included. The max likehood method is applied to evaluate the exponential distribution law parameters for a test sample that contains complete, left-truncated and right-censored data. The changes in the exponential distribution law parameter and its accuracy vs. the share of truncated/censored data have been studied.