Predictive Digital Twins for Thermal Management Using Machine Learning and Reduced-Order Models

TL;DR

This work addresses the challenge of real-time thermal management in automotive systems by integrating physics-based reduced-order models with supervised learning to form predictive digital twins for a headlamp heatsink. A POD-based, component-wise ROM library captures essential thermal dynamics with around $k\approx 10$ modes, reducing CFD cost by roughly 100×, while ML models, especially a neural network, predict ROM configurations to update the digital twin under changing conditions. Using 1000 CFD simulations, the neural network achieves a mean absolute error of $MAE=54.240$ across three outputs: heat transfer coefficient, maximum heatsink temperature, and total heat transfer, outperforming other models like SVR, k-NN, and decision trees. The framework yields real-time, scalable digital twins with transferability to other thermal-management domains (batteries, aerospace, electronics) and supports predictive maintenance and design optimization with substantial resource savings.

Abstract

Digital twins enable real-time simulation and prediction in engineering systems. This paper presents a novel framework for predictive digital twins of a headlamp heatsink, integrating physics-based reduced-order models (ROMs) from computational fluid dynamics (CFD) with supervised machine learning. A component-based ROM library, derived via proper orthogonal decomposition (POD), captures thermal dynamics efficiently. Machine learning models, including Decision Trees, k-Nearest Neighbors, Support Vector Regression (SVR), and Neural Networks, predict optimal ROM configurations, enabling rapid digital twin updates. The Neural Network achieves a mean absolute error (MAE) of 54.240, outperforming other models. Quantitative comparisons of predicted and original values demonstrate high accuracy. This scalable, interpretable framework advances thermal management in automotive systems, supporting robust design and predictive maintenance.