Fluid Flow and Heat Transfer in Chevron-type Plate Heat Exchangers

Abstract

The unsteady local flow characteristics in the chevron-type plate heat exchangers (CPHEs) were investigated by using a large-eddy simulation (LES). Correlations of the friction factor and Colburn factor were derived, and the design optimization was carried out via meta-modeling. The flow in the CPHE consisted of the streamwise direction component and the upper and lower furrow direction components. The repetitive impingement caused by the flow in the streamwise direction leads to an increased the heat transfer coefficient, but it also accompanies an increased pressure drop. In the laminar regime, the fluid flow was steady with no swirling flow. In the turbulent regime, however, the swirling flow occurred in the furrow direction, and the flow was oscillatory. Moreover, a vortex pair occurred at the rear of the contact points, which increased the form drag in the channel. The Strouhal number was investigated according to the geometric parameters. The Strouhal number was almost constant with the chevron angle, whereas the Strouhal number varied according to the ratio of the chevron pitch to the chevron height p/h. The critical Reynolds number depended on p/h, and was found to be 170 ~ 330. Correlations of friction factor f and Colburn factor j were derived in terms of geometric parameters (the chevron angle β and ratio of the chevron pitch to the chevron height p/h) for a variety of working fluids, including water, aqueous solution 50% of ethylene glycol, and diesel fuel. The application range of the correlations was 15° ≤ β ≤ 75°, 2.0 ≤ p/h ≤ 4.4, 200 < Re < 10000, and 2 < Pr < 50. Design optimization was carried out via meta-modeling. The multi-objective optimization was performed by using a weighted sum method. The optimal points were investigated according to the weighting factors, and a Pareto front was generated. Optimization was also carried out by using the JF factor as a single-objective function. The optimal point was β = 66.5° and p/h = 2.73, which was coincident with that for ω1 = ω2 = 0.5 in the multi-objective optimization. The optimal point was independent of the Reynolds number.