By Topic

Estimation of turbulence closure coefficients for data centers using machine learning algorithms

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
S. Yarlanki ; IBM Research, 1101 Kitchwan Road, Yorktown Heights, NY, USA, 10598 ; B. Rajendran ; H. Hamann

CFD models of data centers often use two equation turbulence models such as the k-ε model. These models are based on closure coefficients or turbulence model constants determined from a combination of scaling/dimensional analysis and experimental measurements of flows in simple configurations. The simple configurations used to derive the turbulence model constants are often two dimensional and do not have many of the complex flow characteristics found in engineering flows. Such models perform poorly, especially in flows with large pressure gradients, swirl and strong three dimensionality, as in the case of data centers. This study attempts to use machine learning algorithms to optimize the model constants of the k-ε turbulence model for a data center by comparing simulated data with experimentally measured temperature values. For a given set of turbulence constants, we determine the Root Mean Square `error' in the model, defined as the difference between experimentally measured temperature from a data center test cell and CFD calculations using the k-ε model. An artificial neural network (ANN) based method for parameter identification is then used to find the optimal values for turbulence constants such that the error is minimized. The optimum turbulence model constants obtained by our study results in lowering the RMS error by 25% and absolute average error by 35% compared to the error obtained by using standard k-ε model constants.

Published in:

Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), 2012 13th IEEE Intersociety Conference on

Date of Conference:

May 30 2012-June 1 2012