TY - GEN
T1 - Deep gaussian process enabled surrogate models for aerodynamic flows
AU - Rajaram, Dushhyanth
AU - Puranik, Tejas G.
AU - Renganathan, S. Ashwin
AU - Sung, Woongje
AU - Fischer, Olivia Pinon
AU - Mavris, Dimitri N.
AU - Ramamurthy, Arun
N1 - Funding Information:
The authors would like to acknowledge the support of Siemens Corporate Technology for the work performed in this paper. In particular, we would like to thank Dr. Sanjeev Srivastava and Dr. Wei Xia from Siemens for their valuable technical feedback and support. This material is based upon work partially supported by the U.S. Department of Energy, Office of Science, under contract number DE-AC02-06CH11357.
Publisher Copyright:
© 2020, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.
PY - 2020
Y1 - 2020
N2 - Deep Gaussian process (DGP) models are multi-layered hierarchical generalizations of the well-known Gaussian process (GP) models widely used to construct surrogate models of aerodynamic quantities of interest. Combining the desirable features of GP models and deep neural networks (DNN), DGP models are known to perform well when training data is scarce and the behavior of the system response is highly non-stationary. In this paper, the performance of DGP models is evaluated against GP models. Detailed comparisons are made and conclusions are drawn in terms of training time, data requirements, predictive error, and robustness to choice of training design of experiments, among other metrics. Additionally, sensitivity and scalability analyses are conducted for the DGP models to evaluate their usability. Finally, an adaptive construction of both models is presented, where the models are built sequentially by selecting points that maximize posterior variance. Several experiments are conducted with canonical test functions at varying input dimensions and a viscous transonic airfoil test case at 42 input dimensions. The experiments suggest that DGP models outperform traditional GP models in terms of accuracy but incur higher computational costs for training.
AB - Deep Gaussian process (DGP) models are multi-layered hierarchical generalizations of the well-known Gaussian process (GP) models widely used to construct surrogate models of aerodynamic quantities of interest. Combining the desirable features of GP models and deep neural networks (DNN), DGP models are known to perform well when training data is scarce and the behavior of the system response is highly non-stationary. In this paper, the performance of DGP models is evaluated against GP models. Detailed comparisons are made and conclusions are drawn in terms of training time, data requirements, predictive error, and robustness to choice of training design of experiments, among other metrics. Additionally, sensitivity and scalability analyses are conducted for the DGP models to evaluate their usability. Finally, an adaptive construction of both models is presented, where the models are built sequentially by selecting points that maximize posterior variance. Several experiments are conducted with canonical test functions at varying input dimensions and a viscous transonic airfoil test case at 42 input dimensions. The experiments suggest that DGP models outperform traditional GP models in terms of accuracy but incur higher computational costs for training.
UR - http://www.scopus.com/inward/record.url?scp=85091899305&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85091899305&partnerID=8YFLogxK
U2 - 10.2514/6.2020-1640
DO - 10.2514/6.2020-1640
M3 - Conference contribution
AN - SCOPUS:85091899305
SN - 9781624105951
T3 - AIAA Scitech 2020 Forum
BT - AIAA Scitech 2020 Forum
PB - American Institute of Aeronautics and Astronautics Inc, AIAA
T2 - AIAA Scitech Forum, 2020
Y2 - 6 January 2020 through 10 January 2020
ER -