Reliability growth analysis of randomly censored data

Abstract

Randomly censored data is often met in reliability assessment, since individuals withdraw from test for some reason or haven't failed at the end of the test. Current methods don't make full use of censor information, and only use the positions or sequence of censors, not the exact times. In the paper, a modified method has been used to combine non-parametric and parametric features, and made fully use of the censor information. Thus, more accurate and practical results were obtained. An engineering calculating example shows the fine performance of the method and the results are practical.