Development of a New Correlation Model for Heat Transfer in Solar Air Heater with Corrugated Absorber Plates Using Swarm Optimization

Solar air heaters play a crucial role in distributing heated air at low to medium temperatures. The heart of these systems lies in the absorber plate, which directly absorbs solar heat energy and then efficiently transfers it to the flowing air. However, the challenge lies in achieving optimal thermal efficiency by modifying the absorber plate roughness. Traditional smooth absorber plates have a limited contact area for heat transfer, leading to suboptimal performance. Using passive techniques, such as corrugation on the absorber plate, may increase thermal efficiency by creating turbulence in the laminar sublayer. In this study, corrugated solar air heaters are considered, with the absorber plates roughened into square, semicircular, and triangular ribs. The flow characteristics of heat due to the roughness of the absorber plate is simulated using the computational fluid dynamic (CFD) technique. The ANSYS Fluent 2019R3 is used to investigate the turbulent air flow in the absorber plate. The simulation analyses are performed over a Reynolds number range of 4000–18,000 using three different pitches. As the main contribution, two different model equations, namely M1 and M2, for the correlation model of the solar heat transfer were proposed to predict the Nusselt number, where M2 is an original model and M1 is newly established with the fine-tuned coefficients. Also, for the first time in the literature, the recent swarm optimization algorithm, namely Honey Formation Optimization with Single Component (HFO-1), is used to optimize the solar air heater and a comparison study with a popular non-swarm optimization, namely the Generalized Reduced Gradient (GRG) optimization, is provided. The Nusselt number was obtained using HFO-1 with a percentage error (MAPE) of 2.15% and 0.86% for the models M1 and M2, respectively. Moreover, the average achievement of HFO-1 on the proposed correlation models is 50% better than that of the GRG optimization, with respect to the RMSE.