Towards SEU Fault Propagation Prediction with Spatio-Temporal Graph Convolutional Networks

Assessing Single Event Upset (SEU) sensitivity in complex circuits is increasingly important but challenging. This paper proposes an efficient approach using Spatio-temporal Graph Convolutional Networks (STGCN) to predict the results of SEU simulation-based fault injection. Representing circuit structures as graphs and integrating temporal data from the workload's waveform into these graphs, STGCN achieves a 94-96% prediction accuracy on four test circuits.

I. Introduction

Soft errors induced by high-energy particles, notably Single Event Upset (SEU), pose increasing concerns in semiconductor devices due to growing complexity and reduced dimensions. SEU, a prevalent soft error, particularly affects space applications, causing temporary bit slips in storage components like flip-flops. Despite being non-permanent, SEU faults can lead to unforeseen data corruption or system failures. Simulation-based fault injection, a widely used technique for circuit reliability analysis in the early stages of the design flow, can be excessively time-consuming, especially with complex circuits. The time needed for fault simulation may exponentially increase as circuit complexity grows, making exhaustive SEU fault injections on large circuits practically unfeasible.