Statistical analysis and modelling of small satellite reliability

• A statistical analysis of small satellite reliability using in-orbit failure data.
• Bayesian/Markov Chain Monte Carlo is proposed for satellite reliability modelling.
• Results show the infant mortality of small satellites, especially with TTC.
• Bayesian/MCMC results in better fitting Weibull models than the classical MLE method.
• Bayesian/MCMC has good performance when very limited in-orbit failures are observed.

This paper attempts to characterize failure behaviour of small satellites through statistical analysis of actual in-orbit failures. A unique Small Satellite Anomalies Database comprising empirical failure data of 222 small satellites has been developed. A nonparametric analysis of the failure data has been implemented by means of a Kaplan–Meier estimation. An innovative modelling method, i.e. Bayesian theory in combination with Markov Chain Monte Carlo (MCMC) simulations, has been proposed to model the reliability of small satellites. An extensive parametric analysis using the Bayesian/MCMC method has been performed to fit a Weibull distribution to the data. The influence of several characteristics such as the design lifetime, mass, launch year, mission type and the type of satellite developers on the reliability has been analyzed. The results clearly show the infant mortality of small satellites. Compared with the classical maximum-likelihood estimation methods, the proposed Bayesian/MCMC method results in better fitting Weibull models and is especially suitable for reliability modelling where only very limited failures are observed.