An improved black box model and the details of its numerical treatments for rack in data center simulation

The energy conservation for cooling system has attracted widespread interest in a data center design and operation. Black box assumption is a conventional simplification approach for a rack. However, there is no complete description about how to deal with the boundary conditions for the rack model in the existing literatures. In this paper, the boundary conditions and their numerical treatments for the conventional black box model of a rack are clearly presented, seeming firstly in the literature, and realized by self-coding. The developed model is validated compared with the simulation results using ANSYS Icepak for a typical raised floor data center and good agreements are observed. Furthermore, the conventional black box model in ANSYS Icepak is improved in two aspects of (1) the determination method of server flow rate, (2) the determination of the temperatures of the rack outlet surfaces by point-point correspondence method. Under the studied condition, the conventional black box model will underestimate the maximum temperature in the studied data center by 6.21 °C, and the mixing length zero equation turbulence model shows 2–3 orders of magnitude of time saving for computation and higher convergence and numerical stability compared with the standard k-ε model with an acceptable accuracy.