Experimental study and neural network-based prediction of thermal performance of applying baffles and nanofluid in the double-pipe heat exchangers

Amirhossein Khadang,Mousa Nazari,Heydar Maddah,Mohammad Hossein Ahmadi,Mohsen Sharifpur
DOI: https://doi.org/10.1007/s10973-024-12969-0
IF: 4.755
2024-03-17
Journal of Thermal Analysis and Calorimetry
Abstract:The present study's goal was to investigate the efficiency of applying carbon/water Nanofluid and triangular baffles in the double-tube heat exchangers. The Reynolds number was in the range of 4000 to 14,000, and the carbon/water Nanofluid, which was with various volume fractions ranged between 0.1 and 0.3 volume percentages, was applied. Moreover, triangular baffles were implemented as turbulators with the numbers of 4, 8, and 12, as well as angles of 70 and 40 . Using the response surface method, the experiments design was carried out. The prediction process was eventually conducted using Perceptron artificial neural network that included the input data of Reynolds number, Nanofluids volume fraction, baffles number and angle, as well as Nusselt number target. Results showed that the simultaneous use of 12 triangular baffles at the 40 angle and a Nanofluid with a 0.3 volume fraction led to the 35% increase in Nusselt number compared to applying smooth tubes and water-based fluid. The coefficient of friction with a 0.3% volume fraction of Nanofluid associated with 12 triangular baffles at a 40 angle was, respectively, 5–10% and 1.5 times higher than that of the water flow within a smooth tube without the presence of baffles and Nanofluids. Moreover, results of Perceptron artificial neural network showed that applying a 1-32-4 topology and Levenberg–Marquardt algorithm would derive 0.98828 and 3.6275 values for the mean squared error (MSE) and coefficient of correlation, respectively.
chemistry, physical, analytical,thermodynamics
What problem does this paper attempt to address?