Analysis of interval censored data from highly fractionated experiments

In most cases, a highly fractionated life testing experiment involves a lar ge number of factors for a small number of runs. Typical ANOVA assumes that the observed lifetimes in a highly fractionated experiment (HFE) are norma lly distributed and are not subject to censoring. This paper uses the desig n of experiment (DOE) technique to design and analyze HFE with interval-cen sored response data for reliability improvement. The paper also present an alternative method for analyzing interval-censored data from HFE based on r egression imputation technique. After imputing censored data, the mean and the rank transformation of the response data are analyzed to deal with the interval-censoring problem. This paper provides experimental plan in analyz ing interval-censored data for reliability improvement. The choice of exper imental design, interpretation and the role of interactions are discussed. The method is demonstrated by examining the beat exchanger experiment data where lifetimes are measured in several time intervals. Significance: Many engineering experiments may encounter unforeseen events, which could be terminated prior to their predetermined termination. This p aper presents a methodology for imputing censored data in a highly fraction ated experiment.