The dataset comprises 760 CFD simulations generated from the Latin hypercube sampling (LHS) design-of-experiments. The motivation for generating this dataset was to train supervised AI\ML models to predict spatial concentration gradients in an industrial reactor under varying operating conditions. We split the data into a development set (560 simulations) for model selection and hyperparameter tuning, and an unseen test set (200 simulations) for final evaluation.
For transparency and reproducibility, the complete case lists - including all sampled input parameters for each simulation - are provided in the companion files development_set_cases.csv and test_set_cases.csv. Each row corresponds to a unique simulation (case ID) and can be used to reproduce or subset the experiments reported in this work.
Further details about the meshing workflow, models implemented, and their parameterization can be found in the README.md file.
This dataset is part of the BioBrAIn collection - A CFD database of bioreactors.
%I 4TU.ResearchData