Statistical Reliability Estimation of Earth-Orbiting Satellites Operating from 1991 to 2020

Reliability analysis using satellite failure data for satellites launched in the years 1991–2020 is presented. The analysis is conducted using nonparametric as well as parametric methods. In order to derive a nonparametric reliability estimate from the raw failure data, the Kaplan–Meier estimator is utilized. In order to determine a correct distribution to parameterize the nonparametric estimate, a novel distribution identification framework is applied. The Weibull distribution is consistently identified as best and is then utilized to parameterize the behaviors seen in the nonparametric results. A novel robust regression parameterization method is also applied to improve the traditional regression methodology. Data are split into three decade-long groups with respect to launch dates to investigate time-based trends. Results from this analysis show the most recent decade (2011–2020) to be the worst performing in terms of reliability. This decade in particular is found to be plagued by early or failed reentry cases.