Thermal, frictional and total entropy generation rates are numerically investigated for the water TiO2 nanofluid flow in a circular minichannel considering particle migration. Then, an Artificial Neural Network (ANN) model is developed for prediction of the entropy generation rates in terms of Reynolds number, particle concentration and particle size using the obtained numerical data. The entropy generation rates are examined both locally and globally. The particle migration causes non-uniform distribution of the concentration and thermophysical properties of the nanofluid. This non-uniformity becomes more intense by increasing each of the parameters of concentration, particle size and Reynolds number, such that for mean concentration of 4% and particle size of 80 nm, the values of concentration increment from the wall to the pipe center are about 32% and 66% for Reynolds numbers of 1000 and 2000, respectively. The results indicate that the particle migration can have significant effects on the entropy generation rates, especially at great particle sizes and high concentrations. To address the contribution of each of the factors in the total entropy generation, Bejan number is calculated. The Bejan number reduces by increasing the concentration and Reynolds number, but increases by the particles enlargement. At all concentrations, the Bejan number is greater than 0.8 which indicates that more than 80% of the generated entropy comes from heat transfer. The developed ANN model can predict the outputs with an acceptable accuracy, such that the values of MAE for rates of thermal, frictional and total entropy generation are respectively 4.265 × 10−6, 1.606 × 10−6 and 4.305 × 10−6 based on the test data.
