Comparison of different approaches for numerical modeling of nanofluid subcooled flow boiling and proposing predictive models using artificial neural network

Nanofluids subcooled flow boiling was numerically studied using the Two-fluid model and Three-phase model. The Eulerian approach simulated the contraction of the base fluid and vapor, and the interaction between base fluid and nanoparticles was simulated by the Euler-Lagrange approach; the two Euler-Euler and Euler-Lagrange methods were solved together. The simulations were done in three volume fractions of nanoparticles, 0.0935%, 0.28%, and 0.561%. In the low concentration of nanoparticles, the difference between the results of these approaches was negligible, but by increasing the nanoparticles concentration, the difference increased. When the nanoparticles' volume fraction was 0.28%, the difference between average volume fractions of vapor at the outlet was 16.13%. When the nanoparticle volume fraction was increased up to 0.561%, the difference increased to 28.3%. Also, a comparative study of the two approaches is presented. The Three-phase modeling seems to be the more accurate model. Consequently, a large number of simulations have been carried out to propose predictive models for heat transfer coefficient and vapor volume fraction based on artificial neural networks. •Subcooled flow boiling of nanofluids is numerically simulated using the Two-fluid model and Three-phase model.•The results show that the Three-phase model performs more accurately in higher nanoparticle concentrations.•The Two-fluid model proved to be as accurate as the Three-phase model in lower nanoparticle concentrations.•The ANN models are used to propose predictive models for heat transfer coefficient and vapor volume fraction.•The ANN models had an MAE of lower than 2.3% and R_squared higher than 0.96.