Abstract: CubeSat formations or constellations can form distributed space sensor networks, enhancing the ability of the CubeSat to execute sophisticated space missions. However, CubeSats are prone to malfunction, and the uncertainty of their failure time causes the performance instability of the sensor network, highlighting the importance of on-orbit maintenance of CubeSat networks. In view of the function maintenance of space CubeSat sensor networks, a maintenance architecture is described. It can improve the rapid responsiveness and recovery capability to a single CubeSat fault event of the network by launching CubeSats regularly, making backup CubeSats on the orbit and replacing damaged CubeSats in time. The operation cost model of this architecture is built, involving fixed costs, storage costs and shortage costs. The practical CubeSat lifetime data is collected to acquire the optimal CubeSat lifetime stochastic model with the parameter estimation optimization method of maximizing the coefficient of determination. The supply time and quantity of spare CubeSats are optimized by a Monte-Carlo-simulation-based genetic algorithm. This solution is a trade-off between the costs of backups and the losses caused by the system performance decline, finally achieving an optimal comprehensive income.

Key words: CubeSat failure data, space sensor networks, on-orbit maintenance architecture, on-orbit service, Weibull distribution