Advancing Functional Safety: Improving Failure Mode Analysis and Fault Injection Using Automation and GNN Algorithms

High-Reliability systems need protective measure to ensure functional safety. Traditionally, Failure Mode Analysis (FMA) and Fault Injection (FI) methodologies have been key components in ensuring the functional safety of these systems. However, the ever-increasing complexity of electronic systems poses significant challenges in the efficient and consistent implementation of FMA and FI processes. The challenges include identifying susceptible modules in a short time during FMA, high simulation time, manual effort and resource requirement during FI. To address these challenges, this paper presents two innovative solutions. First, we introduce a FMA automation system that ensures consistency across modules, while significantly reducing effort. Second, we propose a machine learning-based solution to enhance the FI process. Specifically, we leverage Graph Neural Networks (GNNs) to accelerate the detection of critical nodes. Our novel application of the GCNII model overcomes the limitations of traditional Graph Convolutional Networks (GCNs) in handling complex designs with higher logic depth. The employed technique showed increased accuracy and capabilities in forecasting fault criticality. Consequently, this paper facilitates the development of reliable and functionally safe electronic systems.