DeepOHeat-v1: Efficient Operator Learning for Fast and Trustworthy Thermal Simulation and Optimization in 3D-IC Design

2025-04-04

AI Summary

To address key bottlenecks in 3D-IC thermal simulation—namely, challenges in multiscale modeling, prohibitive training costs, and low prediction reliability—this paper proposes a physics-informed operator learning framework. Methodologically, it innovatively employs a Kolmogorov–Arnold Network (KAN) with learnable activation functions to represent multiscale thermal fields; introduces an axis-separable coordinate training strategy to significantly reduce computational and memory overhead; and develops a confidence-score-guided GMRES-accelerated finite-difference refinement optimization pipeline. Experimental results demonstrate that the framework achieves 1.25–6.29× lower thermal simulation error, 62× faster training speed, 31× reduced GPU memory consumption, and 70.6× higher optimization efficiency. Moreover, its peak temperature minimization accuracy matches that of high-fidelity finite-difference solvers, confirming both fidelity and practicality for industrial-scale 3D-IC thermal design.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Hardware-aware ML
• Computer Vision: Learning & Optimization for CV

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Thermal analysis is crucial in three-dimensional integrated circuit (3D-IC) design due to increased power density and complex heat dissipation paths. Although operator learning frameworks such as DeepOHeat have demonstrated promising preliminary results in accelerating thermal simulation, they face critical limitations in prediction capability for multi-scale thermal patterns, training efficiency, and trustworthiness of results during design optimization. This paper presents DeepOHeat-v1, an enhanced physics-informed operator learning framework that addresses these challenges through three key innovations. First, we integrate Kolmogorov-Arnold Networks with learnable activation functions as trunk networks, enabling an adaptive representation of multi-scale thermal patterns. This approach achieves a $1.25 imes$ and $6.29 imes$ reduction in error in two representative test cases. Second, we introduce a separable training method that decomposes the basis function along the coordinate axes, achieving $62 imes$ training speedup and $31 imes$ GPU memory reduction in our baseline case, and enabling thermal analysis at resolutions previously infeasible due to GPU memory constraints. Third, we propose a confidence score to evaluate the trustworthiness of the predicted results, and further develop a hybrid optimization workflow that combines operator learning with finite difference (FD) using Generalized Minimal Residual (GMRES) method for incremental solution refinement, enabling efficient and trustworthy thermal optimization. Experimental results demonstrate that DeepOHeat-v1 achieves accuracy comparable to optimization using high-fidelity finite difference solvers, while speeding up the entire optimization process by $70.6 imes$ in our test cases, effectively minimizing the peak temperature through optimal placement of heat-generating components.

Problem

• Enhance multi-scale thermal pattern prediction in 3D-IC design
• Improve training efficiency and reduce GPU memory usage
• Ensure trustworthy thermal optimization with hybrid workflow

Innovation

• Kolmogorov-Arnold Networks with learnable activation functions
• Separable training method for efficiency
• Confidence score and hybrid optimization workflow