Integrated Health Manager for Satellite Space Segment

Abstract

The steady increase in the complexity of systems and structures is creating new depend-ability challenges across many industries. This includes the space domain, where growing demands on autonomy and onboard computing power has led to utilization of commercial-off-the-shelf (COTS) hardware. While providing more computing power than hardware usually employed in space, COTS hardware is not space-qualified, hence the dependability suffers. However, the demands on dependability are simultaneously growing, in particular for increasingly autonomous systems designed for deep space missions and that cannot depend on human intervention or physical maintenance. The utilization of less reliable COTS hardware combined with increased demands on dependability necessitate a means to increase the dependability of a spacecraft's onboard computer (OBC) and hence avoid corruption of critical data and/or onboard software (OBSW) failure, while simultaneously retaining the high computing power offered by COTS hardware. Integrated Health Manage- ment (IHM), a framework for realtime monitoring of system health and determination of Remaining Useful Life (RUL), presents a promising solution, and is currently the focus of much research in many different technical domains. IHM can increase the dependability of an OBC by continuously tracking its degradation state, providing key system health information, and alerting the OBSW/operator of impending critical damage or system failure. In contrast to Fault Detection, Isolation, and Recovery (FDIR), IHM tracks longterm degradation that develops gradually and will eventually lead to critical damage/system failure. This property is of considerable value to any mission, as the OBSW/operator can be informed of catastrophic events long before they occur, providing sufficient time to take action. This is crucial to avoiding critical data loss or even mission failure. This thesis presents a design and implementation of an IHM applied to a distributed OBSW, capable of tracking longterm software degradation and predicting when the OBSW will crash. The predictive algorithm is based on least-squares polynomial regression (LSPR), which was evaluated in a diverse set of error injection scenarios (EIS) wherein numerous types of failure-inducing degradation were introduced into the OBSW. The results indicate that LSPR-based IHM is capable of longterm predictions and modeling many types of degradation while simultaneously remaining computationally inexpensive and straight- forward to implement. Thus, LSPR-based IHM is recommended for missions with limited processing power. By contrast, distinguishing between real and false degradation signals, i.e. degradation that certainly leads to system failure and that which does not, necessitates the application of more sophisticated predictive algorithms.