A machine learning approach for choosing component level conditions for prognostics of AMS systems

Ageing of components is a predominant concern for the reliability of systems which are expected to be in use over a long period of time. Unlike digital circuits where ageing causes logical errors, the ageing of analog components is often manifested in terms of performance degradation. When analog components are used inside large integrated circuits, the performance degradation of individual components may not show up in the visible output. In this early position paper we propose a methodology for choosing the conditions to be monitored on-chip for the component to determine that it is no longer fit to work in the context of the overall function of the integrated circuit.