This paper investigates reliability analysis of a repairable unit at fleet level. In fleet, multiple similar systems are running in different operating environments. In fact, due to the highly censored data as well as high variations in failure rate within the fleet, merging the datasets becomes a challenge. The non-parametric analysis is used to capture the failure trend of repairable units at fleet level. Consequently, fleet data have been aggregated and the actual number of failure has been compared with the expected total number of failures. In addition, parametric models are used to model the reliability trend obtained through a non-parametric approach to identify the reliability parameters at fleet level. Real data are used to demonstrate the applicability and validity of the proposed method. Results shows the accuracy of log-linear process is reasonably acceptable. © 2015 Taylor & Francis Group, London, UK.