Enabling Prognostics of Robust Design with Interpretable Machine Learning

Design of robust systems needs to fully account for reliability physics, operational stresses and interactions thereof - while accommodating range of stresses from qualification to field. This research demonstrates the method of empirically analyzing system-internal parametric data of Solid-State Storage devices (SSD) with Machine Learning (ML). ML is shown to be a necessary, effective and novel means of proactively assessing and interpreting prognostics of the resilient system design. The methodologies and results also bear strong relevance to assessment of current and future designs for evolving usage models and new application areas.