Approximating the full-field temperature evolution in 3D electronic systems from randomized "Minecraft" systems

TL;DR

This paper introduces a method using fully convolutional neural networks trained on randomized voxel-based systems to accurately predict temperature evolution in large 3D electronic systems, demonstrating strong generalization capabilities.

Contribution

The study presents a novel training approach with randomized voxel systems for neural networks, enabling effective generalization to larger, unseen electronic systems.

Findings

Achieved 0.07% one-step prediction error on large systems

Demonstrated effective generalization from small to larger systems

Proposed a workflow for generating training data with randomized systems

Abstract

Neural Networks as fast physics simulators have a large potential for many engineering design tasks. Prerequisites for a wide-spread application are an easy-to-use workflow for generating training datasets in a reasonable time, and the capability of the network to generalize to unseen systems. In contrast to most previous works where training systems are similar to the evaluation dataset, we propose to adapt the type of training system to the network architecture. Specifically, we apply a fully convolutional network and, thus, design 3D systems of randomly located voxels with randomly assigned physical properties. The idea is tested for the transient heat diffusion in electronic systems. Training only on random "Minecraft" systems, we obtain good generalization to electronic systems four times as large as the training systems (one-step prediction error of 0.07% vs 0.8%).