Wednesday, 29 June 2016: 9:45 AM-10:45 AM
Fundamentals and Applications
Chair:
Nick Gangemi, Northern Air Systems
Technical Committee: 09.09 Mission Critical Facilities, Technology Spaces and Electronic Equipment
CoSponsor: 04.10 Indoor Environmental Modeling
CFD modeling is a powerful tool to simulate and analyze an existing data center or explore the multiple design parameters of a new data center. The challenge is the extreme run time required. Often times this necessitates making choices on what to leave in and what to take out. This session explores several ways to speed up run time for multiple options while preserving accuracy and not sacrificing the physics.The session also takes a look at stanchions and why they should not be excluded in any raised floor model and validated from actual measurements.
1.00 Improving Model Calculation Time without Sacrificing Physics
Data center CFD is often criticized for being too slow. In practice the speed rarely causes an issue since construction and hardware changes are normally planned in advance. However the common approach is to simplify the physics or excessively simplify the geometry resulting inaccurate or, perhaps worse still, uncertainty in whether the results are representative. This discusses two options for increasing productivity while retaining the full physics: Unstructured grid to reduce solution time while retaining the ability to model details where they are important and localized simulations to analyze the impact of small operational changes without a full simulation
2.00 The Compact Modeling of Raised-Floor Stanchions
Stanchions (also referred to as jacks, posts, or pedestals) support the raised floor system common to most data centers. These small-scale objects are often omitted from CFD models of data centers because they are thought to have negligible impact on airflow or because of the increased computational effort required to model them explicitly. This presentation discusses how and why stanchions should be included in data center simulations. A compact modeling approach is proposed which is based on a combination of experimental data and a new analytical model. The model is validated by detailed CFD simulation and actual data center measurements.
3.00 Data Center Modeling Using Response Surface Methodology
In the early design of new data centers or modification of existing ones, the need for an optimal solution of many design parameters such as inlet air flow rate, inlet temperature, and server heat load is of paramount importance. The traditional way using computational fluid dynamics (CFD) to create alternative designs options can be problematic because of their intensive running time. This presentation introduces the Response Surface Methodology (RSM), which is based on CFD data but allows the study of many design parameters in data centers more feasible and economical in terms of modeling time while preserving reasonable accuracy.