A Framework for Self-Diagnosis and Condition Monitoring for Embedded Systems Using a SOM-Based Classifier

This paper presents a system level framework for System-on-Chip (SoC) based embedded devices that may include adaptive and reconfigurable elements. Current development support and debugging solutions are highly dependant on off-line post-mortem style inspection, and even those that utilise tracing for real-time and schedule-critical systems rely on external development tools and environments. This new framework introduces an AI-lead infrastructure that has the potential to reduce much of the development effort while complementing existing debugging circuits. Specifically this paper investigates how to use a Kohonen self-organising map (SOM) as a classifier, and shows a preliminary investigation into how to determine the quality of a map after training. This classifier is a first step in diagnosing failure, degradation and anomalies (i.e. provides condition monitoring) in an embedded system from a system level point of view, and in the larger task of self-diagnosis of an embedded system. © 2008 IEEE.