Control oriented modelling of an integrated attitude and vibration suppression architecture for large space structures

This thesis is divided into two parts. The main focus of the research, namely active vibration control for large flexible spacecraft, is exposed in Part I and, in parallel, the topic of machine learning techniques for modern space applications is described in Part II. In particular, this thesis aims at proposing an end-to-end general architecture for an integrated attitude-vibration control system, starting from the design of structural models to the synthesis of the control laws. To this purpose, large space structures based on realistic missions are investigated as study cases, in accordance with the tendency of increasing the size of the scientific instruments to improve their sensitivity, being the drawback an increase of its overall flexibility. An active control method is therefore investigated to guarantee satisfactory pointing and maximum deformation by avoiding classical stiffening methods. Therefore, the instrument is designed to be supported by an active deployable frame hosting an optimal minimum set of collocated smart actuators and sensors. Different spatial configurations for the placement of the distributed network of active devices are investigated, both at closed-loop and open-loop levels. Concerning closed-loop techniques, a method to optimally place the poles of the system via a Direct Velocity Feedback (DVF) controller is proposed to identify simultaneously the location and number of active devices for vibration control with an in-cascade optimization technique. Then, two general and computationally efficient open-loop placement techniques, namely Gramian and Modal Strain Energy (MSE)-based methods, are adopted as opposed to heuristic algorithms, which imply high computational costs and are generally not suitable for high-dimensional systems, to propose a placement architecture for generically shaped tridimensional space structures. Then, an integrated robust control architecture for the spacecraft is presented as composed of both an attitude control scheme and a vibration control system. To conclude the study, attitude manoeuvres are performed to excite main flexible modes and prove the efficacy of both attitude and vibration control architectures. Moreover, Part II is dedicated to address the problem of improving autonomy and self-awareness of modern spacecraft, by using machine-learning based techniques to carry out Failure Identification for large space structures and improving the pointing performance of spacecraft (both flexible satellite with sloshing models and small rigid platforms) when performing repetitive Earth Observation manoeuvres.