Neural Network-based Fault Diagnosis of Satellites Formation Flight

The main objective of this thesis is to develop a methodology for detecting and isolating faults (i.e. fault diagnosis) in any of multiple reaction wheels that are commonly employed as actuators in a consensus-based virtual structure controlled formation of satellites. In order to accomplish this objective, a two-level fault diagnosis system is developed based on Dynamic Neural Networks (DNNs). In the lower-level of the formation flight system hierarchy, a local fault diagnosis module is available in each individual satellite. In this level, the fault diagnosis system may consist of a dynamic neural network that is trained by using absolute measurements and states of each
\nsingle satellite. Unfortunately, a local fault diagnosis system may fail to detect the presence of low severity faults. In an individual satellite these low severity faults
\nmay not cause any serious complications with the specifications of the overall mission, however they can cause significant impact on the satellite’s attitude or rates in a given precision formation flight of a network of satellites. Consequently, in order to detect these low severity faults a fault detection system is required to be designed and developed at the higher-level or the formation-level of the mission hierarchy. Towards this end, the highly nonlinear dynamics of the formation flight and the reaction
\nwheels are modeled by using dynamic multilayer perceptron neural networks. The proposed formation-level DNNs invoke the extended back propagation learning algorithm and are trained based on sets of input/output data that are collected from the relative attitude determination sensors of the 3-axis attitude control subsystems of the satellites. The DNN parameters are adjusted to minimize certain performance indices (representing the output estimation errors).
\nThe capabilities of the proposed DNNs are investigated under various faulty situations, including single and multiple actuator fault scenarios and under high severity and low severity faulty situations. Using a Confusion Matrix evaluation method, it is demonstrated that by using the proposed fault detection and isolation (FDI) scheme, one can achieve a high level of accuracy and precision in detecting faults. The proposed formation-level FDI system has capabilities in efficiently detecting and isolating
\nactuator low severity faults simultaneously.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.