Use of machine learning in predicting heat transfer and entropy generation in a flat plate solar collector with twisted tape turbulator and ferrofluid under the influence of an external uniform magnetic field: A numerical study

The flat-plate solar collector is the most common and widely used model of solar collectors. This model of solar collectors is more popular than other solar collectors due to its easy installation and lower cost. It is possible to further reduce the cost and size of these systems by improving their performance. For this purpose, simultaneous application of external uniform magnetic field (MF), water-Fe3O4 ferrofluid, and twisted tape turbulator along with changing the cross-section of the serpentine absorber tube of a collector is proposed in the present research. The outlet temperature of working fluid, pressure drop of working fluid, the rate of thermal energy transferred to the flowing fluid in the collector along with the entropy production rate in the system due to fluid friction and heat transfer were considered as the performance parameters of the system. The effect of cross-section of absorber tube (circular, square, triangular), Reynolds number (Re = 500, 1000, 1500 and 2000), pitch distance of turbulator (delta = 150 mm, 200 mm and 250 mm), ferrofluid volume fraction (sigma = 0, 1 %, 2 % and 4 %) and magnetic field intensity (0, 600, 900 and 1200 Gauss) on these parameters was investigated numerically using the finite volume method. In the investigations conducted in the absence of MF, collectors equipped with circular and triangular absorber tubes showed the best and worst thermal performance, respectively, while the lowest and highest total entropy production rate was associated with collectors equipped with circular and rectangular tubes, respectively. Moreover, the escalation in sigma from 0 % to 4 % and under the MF effect showed an increase of 0.14 %, 8.62 %, 0.45 %, and 4.55 % in the outlet temperature, pressure drop, useful heat, and thermal entropy generation rate, respectively. While the frictional entropy generation diminished by 8.73 %. In addition, the intensification of MF intensity from 0 G to 1200 G led to enhance the outlet temperature, useful heat, and pressure drop, by almost 0.0014 %, 3.96 %, and 0.013 %, respectively at three different sigma values. While total entropy generation rate reduces by 0.195 % as MF intensity increases from 0 G to 1200G. Meanwhile, the results of neural network modeling were presented as two second-order polynomial functions to predict the total entropy generation rate and useful heat under the MF effects.