Modern space mission concepts are increasingly dependent on the robust and reliable deployment of spacecraft with large appendages, such as antennas, booms or solar panels. Such deployment requires the ability to properly capture and control the coupled system dynamics, which requires accurate in-orbit system identification of the mass and structural properties. This paper utilises dual Unscented Kalman filters (DUKF) to develop an online system identification strategy that captures both the structural and mass properties, and the attitude and orbit dynamics. The dynamics of the flexible multibody problem are derived from the Lagrangian equations, with the flexible body characteristics modelled with finite element software. A genetic algorithm is used to optimise the accelerometer placement, and hence improve the DUKF performance. We demonstrate that this approach can accurately capture the coupled attitude, orbit, and structural dynamics, as well as being able to provide in-orbit updates for mass properties such as the moment of inertia. The methodology is explored for two illustrative cases: one in which the initial moment of inertia is incorrectly characterised, one in which the moment of inertia changes with time. In both cases, the DUKF approach captures both the system dynamics and the mass properties, which are captured with an error of less than 1%.

中文翻译：

---

使用双无迹卡尔曼滤波器对质量可变的柔性卫星进行在轨系统识别


现代太空任务概念越来越依赖于具有大型附件（如天线、吊杆或太阳能电池板）的航天器的稳健可靠部署。这种部署需要能够正确捕获和控制耦合系统动力学，这需要对质量和结构特性进行准确的在轨系统识别。本文利用双无迹卡尔曼滤波器 （DUKF） 开发了一种在线系统识别策略，该策略可以捕获结构和质量特性，以及姿态和轨道动力学。柔性多体问题的动力学源自拉格朗日方程，柔性体特性使用有限元软件建模。使用遗传算法来优化加速度计的放置，从而提高 DUKF 性能。我们证明，这种方法可以准确捕获耦合的姿态、轨道和结构动力学，并且能够为质量属性（如转动惯量）提供在轨更新。该方法针对两种说明性情况进行了探索：一种是初始惯性矩被错误地表征，另一种是惯性矩随时间变化。在这两种情况下，DUKF 方法都捕获了系统动力学和质量特性，这些特性的捕获误差小于 1%。