关键词: 不确定性分析, 空间科学任务分析, 蒙特卡洛仿真, 神经网络模型

Abstract: Uncertainty analysis plays an important role in early phases of a space mission. The uncertainty of parameters, such as the accuracy of positioning, timing and error in payloads, broadly exists in space mission. When it comes to a distributed satellite system, this problem turns to be more complicated. Traditional method is Monte Carlo simulation, which is time consuming and hard to recognize the coupling relationship between parameters. In this paper, we propose a method which uses a neural network as the model to replace the MCS process. By training the network model using small number of samples, this method can build a model and have advantages in time consuming. We use a multi-spacecraft astronomical observation mission as the use case, And the comparison results between the two methods verify the feasibility of the proposed method.

Key words: Uncertainty analysis, Space science mission analysis, Monte Carlo simulation, Artificial neural network