Data Fusion with Genetic Algorithm Based Lifetime Prediction for Dependable Multi-Processor System-on-Chips

With the prevalence of big-data technology, intricate, nanoscale Multi-Processor System-on-Chips (MP-SoCs) have been used in various safety-critical applications. However, with no extra countermeasures taken, this widespread use of MP-SoCs can lead to an undesirable decrease in their dependability. This study presents a promising approach using a group of Embedded Instruments (EIs) inside a processor core for health monitoring. Multiple health monitoring datasets obtained from the employed EIs are sampled and collated via the implemented experiment and thereafter used for conducting its remaining useful lifetime prognostics. This enables MP-SoCs to undertake preventive self-repair, thus realizing a zero mean downtime system and ensuring improved dependability. In addition, a principal component analysis based algorithm is designed for realizing the EI data fusion. Subsequently, a genetic algorithm based degradation optimization is employed to create a lifetime prediction model with respect to the processor.