Artificial neural network to predict the power number of agitated tanks fed by <scp>CFD</scp> simulations

The power consumption of the agitator is a critical variable to consider in design mixing system. It generally evaluated through dimensionless number known as N p $$ {N}_p . Multiple empirical equations exist calculate based on Reynolds Re \operatorname{Re} and geometrical variables that characterize tank, impeller, height fluid. However, correlations perform poorly outside conditions which they were established. We create rich database 100 k computational fluid dynamics (CFD) simulations. simulate paddle pitched blade turbines unbaffled tanks from 1 use an artificial neural network (ANN) robust accurate predictor number. mesh sensitivity analysis verify precision values given by CFD To sample mixers their physical properties, we Latin hypercube sampling (LHS) method. then normalize data with MinMax transformation put all features same scale thus avoid bias during ANN's training. Using grid search cross-validation, find best architecture ANN prevents overfitting underfitting. Finally, quantify performance extracting 30% database, predicting using ANN, evaluating mean absolute percentage error. error prediction 0.5%, its accuracy surpasses even for untrained geometries.