A novel virtual age reliability model for Time-to-Failure prediction

Abstract

Mean Time To Failure (MTTF) is widely accepted as the main reliability metric in industry. However, several works indicate that MTTF does not accurately capture the reliability characteristics of Integrated Circuits (ICs) and systems given their relatively short operating lifetime. To overcome the MTTF weakness, this paper proposes a novel virtual age based reliability model, which is able to predict the electronic systems Time-To-Failure (TTF). The aging and degrading factors that have an influence on the system's reliability are modeled as cumulative increments of the system's virtual age, and the system's failure point is defined as a proper cut-off cumulative failure rate during its operating lifetime. Thus, system's TTF can be easily estimated based on its virtual age, which reflects the current and historical reliability status of the system. The proposed model is computationally recursive and provides real-time reliability status and prediction, which are critical requirements for enabling reliability-aware resource management and computing.