Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FsimNNs: An Open-Source Graph Neural Network Platform for SEU Simulation-based Fault Injection

Li Lu, Jianan Wen, Milos Krstic

TL;DR

FsimNNs tackles the high computational cost of simulation-based SEU fault injection by leveraging Spatio-temporal Graph Neural Networks to predict SEU outcomes from circuit graphs. The authors propose three STGNN architectures (STGCN, ASTGCN, ASTGAT) with ASPP and attention mechanisms, and build open-source datasets from six circuits, including gate-level netlists and VCD-based temporal features, to train and evaluate the models. Key findings show that all architectures achieve high accuracy on circuit benchmarks (often >$93\%$) with strong precision and recall, while generalization to new test cases (e.g., Ibex workloads) remains feasible but with some degradation (approximately $89\%$ accuracy). The work provides open-source code and data, enabling reproducible reliability analysis and faster SEU vulnerability assessment across diverse workloads.

Abstract

Simulation-based fault injection is a widely adopted methodology for assessing circuit vulnerability to Single Event Upsets (SEUs); however, its computational cost grows significantly with circuit complexity. To address this limitation, this work introduces an open-source platform that exploits Spatio-Temporal Graph Neural Networks (STGNNs) to accelerate SEU fault simulation. The platform includes three STGNN architectures incorporating advanced components such as Atrous Spatial Pyramid Pooling (ASPP) and attention mechanisms, thereby improving spatio-temporal feature extraction. In addition, SEU fault simulation datasets are constructed from six open-source circuits with varying levels of complexity, providing a comprehensive benchmark for performance evaluation. The predictive capability of the STGNN models is analyzed and compared on these datasets. Moreover, to further investigate the efficiency of the approach, we evaluate the predictive capability of STGNNs across multiple test cases and discuss their generalization capability. The developed platform and datasets are released as open-source to support reproducibility and further research on https://github.com/luli2021/FsimNNs.

FsimNNs: An Open-Source Graph Neural Network Platform for SEU Simulation-based Fault Injection

TL;DR

FsimNNs tackles the high computational cost of simulation-based SEU fault injection by leveraging Spatio-temporal Graph Neural Networks to predict SEU outcomes from circuit graphs. The authors propose three STGNN architectures (STGCN, ASTGCN, ASTGAT) with ASPP and attention mechanisms, and build open-source datasets from six circuits, including gate-level netlists and VCD-based temporal features, to train and evaluate the models. Key findings show that all architectures achieve high accuracy on circuit benchmarks (often >) with strong precision and recall, while generalization to new test cases (e.g., Ibex workloads) remains feasible but with some degradation (approximately accuracy). The work provides open-source code and data, enabling reproducible reliability analysis and faster SEU vulnerability assessment across diverse workloads.

Abstract

Simulation-based fault injection is a widely adopted methodology for assessing circuit vulnerability to Single Event Upsets (SEUs); however, its computational cost grows significantly with circuit complexity. To address this limitation, this work introduces an open-source platform that exploits Spatio-Temporal Graph Neural Networks (STGNNs) to accelerate SEU fault simulation. The platform includes three STGNN architectures incorporating advanced components such as Atrous Spatial Pyramid Pooling (ASPP) and attention mechanisms, thereby improving spatio-temporal feature extraction. In addition, SEU fault simulation datasets are constructed from six open-source circuits with varying levels of complexity, providing a comprehensive benchmark for performance evaluation. The predictive capability of the STGNN models is analyzed and compared on these datasets. Moreover, to further investigate the efficiency of the approach, we evaluate the predictive capability of STGNNs across multiple test cases and discuss their generalization capability. The developed platform and datasets are released as open-source to support reproducibility and further research on https://github.com/luli2021/FsimNNs.

Paper Structure

This paper contains 19 sections, 3 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Datasets
  3. Raw data of SEU fault simulation
  4. Spatio-temporal graph generation
  5. Sampling and training
  6. STGNN models
  7. Spatial feature learning
  8. Graph Convolutional Layer
  9. Graph Attention Layer
  10. Temporal feature learning
  11. Temporal Convolutional Layer
  12. Atrous Spatial Pyramid Pooling (ASPP)
  13. Models
  14. Experiments
  15. Experiment pipeline
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Overview of FsimNNs within the STGNN approach framework.
  • Figure 2: The process of feature extraction and STGNN training.
  • Figure 3: The architectures of STGNNs.
  • Figure 4: The generalization gap in accuracy on validation and test datasets. The box extends from the 25th percentile to the 75th percentile, and lines extend from the box to the smallest and largest values. Wider boxes or longer lines indicate greater variability in generalization. The red line inside the box represents the median of the data.