Predicting the accuracy of nanofluid heat transfer coefficient's computational fluid dynamics simulations using neural networks

This research presents a neural network algorithm to identify the best modeling and simulation methods assumptions for most widespread nanofluid combinations. The is trained using data from earlier experiments. A multilayer perceptron with one hidden layer was employed in investigation. set were created Python Keras module forecast average percentage error heat transfer coefficient of models. Integer encoding used encode category variables. total 200 trials different networks taken into consideration. worst-case bound chosen architecture then calculated after 100 runs. Among eight models examined single-phase, discrete-phase, Eulerian, mixture, mixed model discrete mixture phases, fluid volume, dispersion, Buongiorno's model. We discover that broad range configurations accurately covered by Buongiorno, discrete-phase They accurate particle sizes (10–100 nm), Reynolds numbers (100–15,000), volume fractions (2%–3.5%). accuracy evaluated root mean square (RMSE), absolute (MAE), R2 performance metrics. algorithm's value 0.80, MAE 0.77, RMSE 2.6.