Physically interpretable surrogate modeling of thermal fields in electronics cooling using combined proper orthogonal decomposition and neural networks

As advances in semiconductor technology drive higher power densities, thermal management is becoming increasingly crucial to maintaining the performance and reliability of devices. In recent decades, computational fluid dynamics (CFD) has been widely used for thermal management system design. However, the computational requirement of CFD limits its use for comprehensive system optimization and digital twin active control systems, which require quick, high-fidelity thermal field prediction. Data-driven surrogate models address the speed limitations of CFD but are often treated as black boxes whose predictions are physically uninterpretable. Neural networks (NN) combined with proper orthogonal decomposition (POD) maintain the speedup of other surrogate models while offering a physically interpretable latent space. The interpretability of POD-NN learned representations remains largely unexplored. This work develops and demonstrates a POD-NN surrogate model framework to predict 2D thermal f..., , # Data from: Physically interpretable surrogate modeling of thermal fields in electronics cooling using combined proper orthogonal decomposition and neural networks

Dataset DOI: [10.5061/dryad.k0p2ngfp5](https://doi.org/10.5061/dryad.k0p2ngfp5)

## Description of the data and file structure

This dataset accompanies Curl & Hu (2026). It contains 2000 steady-state CFD simulations of a liquid-cooled, dual-chip, pin-fin cold plate run in ANSYS Fluent, the POD-NN surrogate trained on those simulations, and the source code that reproduces every figure in the paper from the published data. The DOE varies six boundary conditions (chip 1 / chip 2 surface heat fluxes, inlet 1 / inlet 2 mass flow rates, and inlet 1 / inlet 2 total temperatures) across the operational space of the cold plate using Latin Hypercube Sampling. The CFD model solves the steady, incompressible Navier-Stokes equations coupled with the energy equation and Menter's k-ω SST turbulence model on a ~50,000-element 3D grid; onl..., ,