Analysis and optimization on flow and heat transfer characteristics of twisted oval shell-and-tube heat exchangers based on numerical simulation and model prediction

Conventional shell-and-tube heat exchangers face substantial limitations in improving thermal efficiency. To address this, a novel shell-and-tube heat exchanger design incorporating twisted oval tubes in place of conventional round tubes is proposed. This study examines the effects of twisted pitch length and shell-side Reynolds number on the heat transfer and flow characteristics of the twisted oval tube shell-and-tube heat exchanger with baffles. The optimization of input variables was performed using a quadratic polynomial model and an artificial neural network model. The findings demonstrate that compared to the round shell-and-tube heat exchanger with baffles, the heat transfer coefficient of the twisted oval shell-and-tube heat exchanger with baffle increases by 15.9 % and the pressure drop decreases by 28.9 %. The heat transfer coefficient decreases with increasing twisted pitch length, with the most significant reduction of 5.7 % observed as twisted pitch length increased from 60 mm to 90 mm. Conversely, the heat transfer coefficient increases markedly with higher Reynolds number. The pressure drop rises substantially with increasing Reynolds number but decreases with higher twisted pitch length. The ratio of heat transfer coefficient to pressure drop declines with increasing Reynolds number, albeit at a diminishing rate, while the influence of twisted pitch length on the ratio of heat transfer coefficient to pressure drop is less pronounced compared to Reynolds number. When the twisted pitch length was 60 mm and the Reynolds number increased from 3167 to 5588, the ratio of heat transfer coefficient to pressure drop decreased from 1.61 to 0.46 W/m²·K·Pa, representing the maximum decline. The interactions between twisted pitch length and Reynolds number significantly affect heat transfer coefficient and pressure drop. Both predictive models demonstrate high accuracy, with the artificial neural network model outperforming the quadratic polynomial model, particularly in predicting pressure drop. The artificial neural network and quadratic polynomial models also exhibit greater sensitivity to heat transfer coefficient than to pressure drop, with discrepancies in predictive accuracy being more evident for pressure drop.