Graph Sequence-Based Prediction of Electrothermal Stress Evolution for Power Delivery Networks

With the advancement of integrated circuit technology, electrothermal migration (EM-TM) in power delivery networks (PDNs) has become a key reliability challenge. Existing machine learning methods are unable to capture the evolution trajectory of EM-TM stress and actively predict the lifetime of PDNs. We propose a spatiotemporal graph-based prediction framework that integrates spatial structural features and captures early stress accumulation patterns from input sequence, enabling long-term forecasting of EM-TM coupling evolution. The experimental results show that our approach achieves higher prediction accuracy and stability on PDNs datasets spanning up to 1,500 interconnect segments, delivering near 5× speedup in prediction compared to the single-timestep model and 6896× acceleration over the commercial software COMSOL. Furthermore, the framework enables reliable PDNs lifetime estimation with millisecond computation time.