Addition of MWCNT-Al₂O₃ nanopowders to water- ethylene glycol (EG) base fluid for enhancing the thermal characteristics: Design an optimum feed-forward neural network

Abstract

Prediction the thermal conductivity of nanofluids has been subject of many researches. Artificial Neural Networks are used to obtain thermal conductivity of NAnofluids because not only this method is fast and acurate but also it can reduce the Lab costs. To predict the thermal conductivity of water- EG/MWCNT-Al₂O₃ hybrid nanofluid ($k_{nf}$) a feed-forward neural network with different neuron numbers has been tested and the best network based on the performance is selected. The Levenberg Marquardt algorithm is used for training the network, which is one of the best algorithms in machine learning. Also, using a fitting method, a surface is used to illustrate the behavior of nanofluids based on the volume fraction of nanoparticles ($\varphi$) and temperature (T). $\varphi =$0, 0.001, 0.002, 0.004, 0.008, 0.0016 and T = 25, 30, 35, 40, 45, 50 °C are used.. The obtained results show that the ANN and Fitting results are close to the experimental datapoints, and both methods can predict $k_{nf}$ accurately. As the results of these methods are very close, but the ANN method is better in predicting the behavior of this nanofluid.