CFD-informed machine learning surrogate dataset for thermo-hydraulic prediction in partially porous wavy channels

This dataset supports the manuscript “CFD-Informed Machine Learning Surrogate Modeling for Thermo-Hydraulic Prediction in Partially Porous Wavy Channels for Heat Sink Applications.” It contains the cleaned long-form CFD-derived dataset, fixed train-test split information, trained-model outputs, prediction files, validation tables, plot data, and one final reproducibility script used to train Random Forest surrogate models for predicting average Nusselt number and pressure drop.

The dataset includes 4,608 CFD-derived samples generated from 18 geometric configurations and 256 operating-condition combinations. The input features are Reynolds number, Prandtl number, Darcy number, porosity, porous slab thickness, wave amplitude, and wavelength. The target outputs are average Nusselt number and pressure drop.