Dynamic Modeling of a Refrigerant-Based Cross-flow Heat Exchanger for Close-Coupled Hybrid Cooled Data Centers

Close-coupled hybrid cooling systems (CCHCS) are an efficient solution for cooling of modern high-density data centers (DC). Such systems are installed in proximity to the IT equipment and offer advantages over centralized DC in terms of managing hot spots and providing demand-based cooling while eliminating raised floors. By their on-demand nature, close-coupled cooling systems must operate in dynamic modes in which their capacity, controlled by mass flow rate and operating states are in response to load variations. CCHCS feature a cross-flow heat exchanger (HX) with refrigerant or chilled water as the working fluid. The present work implements a numerical model to predict the transient response of a refrigerant-based, crossflow HX as part of a CCHCS. The use of refrigerants enhances the thermal performance of these systems by taking advantage of the high heat-transfer coefficients and high latent heat absorption obtained during the boiling process. The model assumes the refrigerant enters the HX as a sub-cooled liquid, with perhaps a few to 10 degrees C of subcooling. A system of partial differential equations for conservation of mass, energy, and momentum representing each of the flow phases for the refrigerant, HX tube wall and fins, and air is solved in MATLAB. A trapping method is implemented to capture the phase change locations (i.e., a ‘front tracking’ method) inside the HX, including the transition from single to two-phase and v.v. The temperature of the air entering the HX is varied over time to simulate the air leaving the servers in a DC rack. The model is validated against preliminary experimental data obtained in our laboratory for chilled water flowing in single phase. Understanding and modeling the dynamic behavior of a crossflow HX will provide insight regarding the effectiveness of CCHCS systems with subcooled inlet conditions. This can augment the control strategies implemented for cooling of data centers.