Application of artificial neural network–genetic algorithm (ANN–GA) to correlation of density in nanofluids

This paper applies the hybrid model including back-propagation network (BPN) and genetic algorithm (GA) to estimate the nanofluids density. GA was coupled with BPN to optimize the BPN's parameters and improve the accuracy of proposed model. The experimental density of four nanofluids in the temperature range of 273–323 K with the nanoparticle volume fraction up to 10% was examined. The obtained results by BPN–GA model have good agreement with the experimental data with absolute deviation 0.13% and high correlation coefficient (R≥0.98)(R≥0.98). The results also reveal that BPN–GA model outperforms to radial base function net and Pak and Cho model for predicting of the density of nanofluids with the overall improvement of 64% and 95% respectively.

► The authors employ the ANN-GA for prediction of nanofluid density.
► The advantage of this technique compared to conceptual models is its high speed, simplicity and generalization.
► Combine genetic algorithm with BPNN to avoid local minima in training phase for limited available experiment data.
► The GA algorithm employs to achieve global convergence quickly and correctly with optimizing the weights and thresholds of the neural network.