Structural optimization design of dimple plate heat exchanger based on machine learning

Abstract

This study employs computational fluid dynamics (CFD) to investigate the thermal-hydraulic performance within the channels of dimple plate heat exchangers (DPHE). A parametric simulation approach is utilized to evaluate the effects of dimple height (H = 4 mm, 6 mm, 8 mm, 10 mm and 12 mm), dimple pitch (P = 10 mm, 15 mm, 20 mm and 30 mm), and channel width (W = 8 mm, 10 mm and 12 mm) on the overall performance of the DPHE. To decouple the interactions among these structural parameters, an artificial neural network (ANN) combined with an optimization algorithm is employed for multi-objective optimization based on the results of parametric simulation. The complex channel geometry and the presence of welded points significantly alter the flow path and fluid dynamics within the exchanger. The findings indicate that the Nusselt number (Nu) reaches a maximum value of 67.67 at a flow velocity of 0.3 m/s when H = 12 mm, while variations in P exhibit the most substantial influence on the overall performance of the DPHE. The optimal structural parameters determined through optimization are H = 4.0031 mm, P = 25.716 mm, and W = 8.001 mm. Under these conditions, the j-factor (JF) of the DPHE attains a maximum value of 0.125.