A novel approach to in-field, in-mission reliability monitoring based on Deep Data

Abstract

This paper describes a deep data approach to reliability monitoring in advanced electronics, based on degradation as a precursor for failure. By applying machine learning algorithms and analytics to data created by on-chip monitoring IPs (Agents), IC/system health and performance can be continuously monitored, at all stages of the product lifecycle. Realtime degradation analysis of critical parameters and failure mechanisms, under field conditions and application environments, points to the underlying Physics of Failure, which in turn allows to estimate the time to failure. Users are alerted on faults in advance, via a cloud-based analytics platform, and can take corrective action to prevent failures. The future of reliability physics and engineering is fundamentally shifting from accelerated lifetime tests to in-field failure prediction.