In the study, the optimal distribution of discrete heat sources in a two-dimensional data centre was investigated. The optimal placement of the cool supply air inlet and outlet was also investigated. The governing equations were solved by using the finite volume method. The computational fluid dynamics code Fluent was used to solve the governing equations. Optimisation was achieved using a goal-driven optimisation approach and a response surface methodology.
The numerical model was validated using past experimental work and the results were in good agreement with each other, showing an error of less than 6%. The realisable k-? turbulence model was used as closure equations to solve the Reynolds-averaged Navier-Stokes equations. Additionally, the viscosity affected near-wall regions were modelled using a wall treatment method.
The optimum distribution of constant height (42 U) server racks was established for three different configurations of inlet and outlet locations. After these optimal placements were established, the effect of varying the height of the server racks was investigated for the same inlet and outlet placements and the optimum locations were determined. By means of a sensitivity analysis, it was found that the placement of the first and last servers as well as their respective heights had the most influence on the heat transfer between the server panels and the ambient surroundings.
It was concluded that the inlet and outlet should be placed on opposing walls of the data centre and variable server rack heights should be used in order to achieve maximum heat transfer.
Dissertation (MEng)--University of Pretoria, 2016.