Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Combining Low-dimensional Models with High-fidelity Data: A Multi-fidelity Approach to Transient Heat Transfer Problems

Pengdi, Zhang (2021) Combining Low-dimensional Models with High-fidelity Data: A Multi-fidelity Approach to Transient Heat Transfer Problems. Master's Thesis, University of Pittsburgh. (Unpublished)

Download (4MB) | Preview


We present a generic multi-fidelity approach for combining constructing multi-fidelity surrogate models for transient heat transfer applications. We apply our developed methodology to build various surrogate models for mixing the temperature of two different fluids. This is a classical heat transfer problem with numerous applications in diverse industries and it is considered in this work as a template problem for our methodology. In the presented framework, data from various levels of fidelity can be combined in a principled manner. More broadly, our target applications are problems where relying on high-fidelity data alone is not sufficient to build satisfactory surrogate models due to the high expense often associated with high-fidelity data acquisition. On the other hand, it may be possible to build reduced-order models that can be sampled at a high rate with insignificant computational cost. However, the ROM may be inaccurate due to physics deficiency of the full-dimensional model as well as the reduction errors. To this end, we utilize reduced-order models of heat transfer mixing, i.e. low fidelity model, and high-fidelity measurements, which are obtained by performing direct numerical simulations. We will combine these two data sources using Gaussian process regression (GPR) and auto-regressive stochastic strategy. GPR is a non-parametric Bayesian regression technique that has a fully probabilistic workflow, in which the prediction uncertainties can be quantified in a principled manner. In the following research, we will successively verify the accuracy and computational effectiveness of multi-fidelity results for different quantities of parameters and the different sizes of training data.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Pengdi, Zhangpez37@pitt.edupez370000-0002-9104-6703
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Thesis AdvisorBabaee,
Committee ChairBabaee,
Committee CoChairPeyman,
Committee CoChairSammak,
Date: 3 September 2021
Date Type: Publication
Defense Date: 5 April 2021
Approval Date: 3 September 2021
Submission Date: 4 May 2021
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 58
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Mechanical Engineering and Materials Science
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: Multi-fidelity Gaussian process regression; Spectral/hp element method; Proper orthogonal decomposition.
Date Deposited: 03 Sep 2021 18:50
Last Modified: 03 Sep 2021 18:50


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item