A well-trained artificial neural network (ANN) using the trainlm algorithm for predicting the rheological behavior of water – Ethylene glycol/WO3 – MWCNTs nanofluid

Abstract

In this study, the influence of volume fraction of nanoparticle (φ) and temperatures on the dynamic viscosity (μnf) of water – ethylene glycol/WO3 – MWCNTs hybrid nanofluid was analyzed. For this reason, the μnf of water – ethylene glycol/WO3 – MWCNTs nanofluid has derived for 42 various experiments through a series of experimental tests, including a combination of 7 different φ and 6 various temperatures. These data were then used to train an Artificial Neural Network (ANN) to generalize results in the predefined ranges for two input parameters. For this reason, a feed-forward Perceptron ANN with two inputs (T and φ) and one output (μnf) were used. The best topology of the network was determined by trial and error, and two-layer with 10 neurons in the hidden layer with the tansig function had the best performance. Also, to analyze the effect of various training algorithms on the performance of μnf prediction, 10 different training functions were used for this reason, and the best ANN was obtained when the trainbr is used as a training function. The trained ANN roles as a predicting function of μnf in every combination of temperature and φ. The obtained results show that a well-trained ANN is created using the trainlm algorithm and showed an MSE value of 4.2e-4 along 0.998 as a correlation coefficient for predicting μnf. Also, the temperature has an inverse effect on the output parameter (μnf). By increasing the temperature, the μnf decreases for all φ. At the same time, this decrement is more noticeable at higher φ. For example, they increase the temperature from 25 to 50 °C changes the dynamic viscosity of the pure fluid by only about 15%. In contrast, the same temperature changes in φ= 0.6% cause a 35% drop in μnf.