A Wavelet Decomposition Method for Tuning Thermal Models to Aperiodic Transient Test Data
- View All
Although it is generally agreed that thermal analysis models are improved through parameter tuning to empirical test data, efficient methods to improve the correlation of models to test data are not in general use, especially when dealing with aperiodic transient simulations and multiple model configurations. If model tuning is performed, it is often accomplished through a manual process of changing parameter values by trial and error. This process is time-consuming and highly inefficient. Even when automatic methods are utilized, the model response is often compared to the empirical data using direct comparisons such as the sum of the root mean square (RMS) difference, which has significant limitations. ATA Engineering, Inc., (ATA) has developed a novel method for tuning thermal models to aperiodic transient test data using wavelet decompositions of the time domain model response and test data. This method compactly describes the transient functions while retaining sufficient information to correlate the model to test data at multiple time scales. The method has been demonstrated by automatically correlating models of two-phase vapor compression systems (VCSs) developed at Texas A&M University. While the RMS error between model response and test data was not used as part of the correlation algorithm, it serves as an easily understood correlation metric. The average RMS error for all functions studied was reduced by at least 95% after the model was correlated using the automatic algorithm. This paper presents details of the algorithm developed and results from a correlation study, as well as a general overview of the VCS models used. The work was funded by the U.S. Air Force Research Laboratory under a Small Business Innovation Research grant. © 2012 by ATA Engineering, Inc. Published by the American Institute of Aeronautics and Astronautics, Inc.
author list (cited authors)
Kaplan, M., Garrett, M., & Rasmussen, B.