Learning heat: High-fidelity experimental and Koo–Kleinstreuer–Li thermal conductivity predictions in nanofluids via advanced data augmentation and metaheuristic search

The precise forecasting of thermal conductivity in nanofluids is essential for enhancement of thermal management systems within industrial contexts. This research establishes a robust framework that combines machine learning (ML), deep learning (DL), advanced augmentation, and optimization strategies to predict two types of thermal conductivity: experimental thermal conductivity (Exp-TC) and the effective thermal conductivity based on the Koo-Kleinstreuer-Li model (KKL-TC). The study curates a comprehensive dataset consisting of 278 samples, which includes a variety of nanoparticle materials (such as Cu , CuO , Al , and Al 2 O 3 ) and base fluids (including water , ethylene glycol , and transformer oil ). Critical parameters, including particle size, temperature, volume fraction, thermal conductivity of particle and liquid medium, were enhanced through advanced data augmentation techniques—specifically, polynomial and fourier expansion inspired augmentation and conditional variational autoencoders to improve data diversity and the robustness of the models. The approach utilized fourteen ML and DL models in two configurations: standalone and ensemble stacking with linear regression. Optimization techniques, such as grey wolf optimization and particle swarm optimization, were employed to refine hyperparameters and enhance predictive accuracy. The results reveal that both augmentation and optimization significantly boost predictive accuracy. Specifically, CatBoost delivers the best Exp-TC (R 2 = 0.99964, RMSE = 0.00464) and KKL-TC performance (R 2 = 0.99782, RMSE = 0.00391). These insights establish a highly scalable and adaptable framework that extends beyond thermal system design to a broad range of engineering and industrial applications. This framework drives innovation in energy management, facilitating the development of more efficient, sustainable, and intelligent thermal systems.