Abstract

Health monitoring and fault diagnosis in traditional single spacecraft missions are mostly accomplished by human operators on ground through around-the-clock monitoring and trend analysis on huge amount of telemetry data. Future multiplatform space missions, commonly known as the formation flight missions, will utilize multiple inexpensive spacecraft in formation by distributing the functionalities of a single platform among the miniature inexpensive platforms. Current spacecraft diagnosis practices do not scale up well for multiple space platforms due to an increasing need to make the long-duration missions cost-effective by limiting the size of the operations team which will be large if traditional diagnosis is employed. An ideal solution to this problem is to incorporate an autonomous fault detection, isolation, and recovery (FDIR) mechanism. However, the effectiveness of spacecraft autonomy is yet to be demonstrated and due to the existence of perceived risks, it is often desired that the expert human operators be involved in the spacecraft operations and diagnosis processes i.e., the autonomous spacecraft actions be understandable by the human operators on ground so that intervention may be made, if necessary.

To address the above problems and requirements, in this research a systematic and transparent fault diagnosis methodology for ground-based operations of multi-platform space systems is developed. First, novel hierarchical fault diagnosis concepts and framework are developed. Within this framework, a multi-platform space system is decomposed hierarchically into multiple levels. The decomposition is driven by the need for supporting the development of the components/subsystems of the overall system by a number of design teams and performing integration at the end. A multi-platform system is considered to be a set of interacting components where components at different levels correspond to formation, system, sub-system, etc. depending on the location of the node in the hierarchy. Two directed graph based fault diagnosis models are developed namely, fuzzy rule based hierarchical fault diagnosis model (HFDM), and Bayesian networks (BN)-based component dependency model (CDM).

In HFDM, fault diagnosis of different components in the formation flight is investigated. Fuzzy rules are developed for fault diagnosis at different levels in the hierarchy by taking into account the uncertainties in the fault manifestations in a given component. In this model, the component interactions are quantified without taking the uncertainties in the component health state dependencies into account. Next, a component dependency model (CDM) based on Bayesian networks (BN) models is developed in order to take the uncertainties in component dependencies into account. A novel methodology for identifying CDM parameters is proposed. Fault evidences are introduced to the CDM when the fault modes of a component are observed via fuzzy rule activations. Advantages and limitations associated with the proposed HFDM and the CDM are also discussed. Finally, the verification and validation (V&V) of the hierarchical diagnosis models are investigated via a sensitivity analysis approach.

It should be noted that the proposed methodology and the fault diagnosis strategies and algorithms that are developed in this research are generic in a sense that they can be applied to any hierarchically decomposable complex systems. However, the system and domain specific knowledge they require, especially for modeling component dependencies, are mostly available in the aerospace industry where extensive system design and integration-related analysis are common due to high system building cost and failure risks involved.