Satellite reliability analysis using mathematical prediction models: a case study of satellite anomalies

The importance of satellites in delivering critical services including communication systems, navigation, and earth observation heavily depends on system reliability. Satellite reliability is defined as the probability that a system is failurefree under specified conditions to ensure the satellites remain safe, efficient and long-lasting. Previous studies highlighted the impact of satellite anomalies as significant contributors to objective failures, cost inefficiencies, and satellite system degradation. However, these studies often lack a comprehensive analysis of reliability using structured mathematical models. Hence, this research addresses this gap by analyzing 87 satellite data from the Seradata database using different mathematical models, categorized into parametric and non-parametric models. The parametric models are Exponential, Weibull and Poisson, whereas the non-parametric models are Kaplan-Meir and Monte Carlo. A case study of satellite anomalies is conducted focusing on satellite communication subsystems of failed satellites due to anomalies. The detected anomalies vary from antenna, transponder, amplifier and battery subsystems. Then, the mathematical models are developed and compared to evaluate the study’s objective in obtaining the best mathematical model based on the lowest Root Mean Square Error (RMSE). Based on the five mathematical models, the results reveal that the Exponential model exhibits the best overall performance with the lowest RMSE values for antenna, transponder, amplifier and battery of 0.22, 0.27, 0.30 and 0.30 respectively. The findings of this study demonstrate the advantages of having multiple mathematical model evaluations that enhance the selection of models in satellite reliability analysis for the development of satellite design and operation.