A Numerical Study of Flow and Heat Transfer in Compact Heat Exchangers

Abstract

An accurate computational method for the calculations of flow and heat transfer in compact heat exchangers is developed in collaboration with the National Center for Supercomputing Applications. In this method, the unsteady Navier-Stokes and energy equations are solved. In the simulations of flow and heat transfer over relatively simple heat exchanger fin geometries, a linearly scalable performance of the code is achieved on the massively parallel CMS, demonstrating the capability of this method to solve large scale heat transfer problems. Using this code, the heat transfer enhancement mechanisms and performance of parallel-plate fin heat exchangers are studied extensively. Geometry effects such as finite fin thickness and different fin arrangements (inline and staggered) have been investigated. The time-dependent flow behavior due to vortex shedding has been taken into consideration by solving the unsteady Navier-Stokes and energy equations. In the unsteady regime, in addition to the time-dependent calculations, companion steady symmetrized flow calculations have also been performed to clearly identify the effect of vortex shedding on heat transfer and frictional loss. Additional comparisons have been made to the theoretical results for fully developed flow between uninterrupted continuous parallel plates and those of restarted boundary layers with negligible fin thickness, in order to quantify the role of boundary layer restart mechanism as well as the geometry effects of finite fin thickness and fin arrangement. It is shown in the current study that at higher Reynolds numbers, the additional effect introduced by intrinsic three-dimensionality of the flow also plays an important role in determining the overall heat exchanger performance. At sufficiently high Reynolds numbers, when the actual flow is threedimensional, corresponding two-dimensional models overpredict overall heat transfer efficiency by as much as 25%, while the overprediction of frictional loss is much less. More importantly, the overprediction of rms fluctuations in heat transfer and frictional loss in two-dimensional models is much larger, where the amplitude of fluctuations from two-dimensional models can be as much as 2 and 5 times of those from corresponding three-dimensional models for the heat transfer efficiency (Colburn j factor) and frictional loss (friction factor), respectively. These differences are attributed to the strong coherence of spanwise vortices in two-dimensional simulations 'and the weakening of spanwise vortices in the corresponding three-dimensional simulations due to the presence of streamwise vortices. In two-dimensional simulations, the coherent spanwise vortices enhance mixing and result in higher heat transfer efficiency. These span wise vortices at the same time lowers skin friction on the fin surface. On the other hand, it has been well established that two-dimensional simulations overpredict form drag due to higher Reynolds stresses in the wake. In current two-dimensional simulations of flow over parallel-plate fins, the overprediction of form drag is nearly counter-balanced by the underprediction of skin friction. Such mechanisms also shed light on enhancing heat transfer while avoiding the normally associated increased pumping power penalty. In the simulations of flow and heat transfer in more complex louvered fin geometries, current numerical results clearly show the different flow regimes as the Reynolds number is increased, which are generally in agreement with those observed in experimental flow visualizations. However, at low Reynolds numbers, current interpretation of the flow characteristics is somewhat different. At higher Reynolds numbers, the effect of flow unsteadiness is to increase overall heat transfer and associated frictional loss.