Imperial College London
Portrait Picture

DrMichaelBluck

Faculty of Engineering, Department of Mechanical Engineering

Reader in Nuclear Engineering
 
 
 
//

Contact

 

+44 (0)20 7594 7055m.bluck

 
 
//

Location

 

658City and Guilds BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Duan:2021:10.1016/j.jcp.2021.110243,
author = {Duan, Y and Eaton, MD and Bluck, MJ},
doi = {10.1016/j.jcp.2021.110243},
journal = {Journal of Computational Physics},
pages = {1--14},
title = {Fixed inducing points online Bayesian calibration for computer models with an application to a scale-resolving CFD simulation},
url = {http://dx.doi.org/10.1016/j.jcp.2021.110243},
volume = {434},
year = {2021}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - This paper proposes a novel fixed inducing points online Bayesian calibration (FIPO-BC) algorithm to efficiently learn the model parameters using a benchmark database. The standard Bayesian calibration (STD-BC) algorithm provides a statistical method to calibrate the parameters of computationally expensive models. However, the STD-BC algorithm does not scale well with regard to the number of data points and also it lacks an online learning capability. The proposed FIPO-BC algorithm greatly improves the computational efficiency of the algorithm and, in addition, enables online calibration to be performed by executing the calibration on a set of predefined inducing points.To demonstrate the procedure of the FIPO-BC algorithm, two tests are performed, finding the optimal value and exploring the posterior distribution of 1) the parameter in a simple function, and 2) the high-wave number damping factor in a scale-resolving turbulence model (scale adaptive simulation shear-stress transport model/SAS-SST). The results (such as the calibrated model parameter and its posterior distribution) of FIPO-BC with different inducing points are compared to those of STD-BC. It is found that FIPO-BC and STD-BC can provide very similar results, once the predefined set of inducing points in FIPO-BC is sufficiently fine. Given that fewer datapoints are needed in the proposed FIPO-BC algorithm, compared to the STD-BC algorithm, it will be a more computational efficient algorithm. In our demonstration test cases, the proposed FIPO-BC algorithm is at least ten times faster than the STD-BC algorithm. Meanwhile, the online feature of the FIPO-BC allows continuous updating of the calibration outputs and potentially reduces the workload on generating the database.
AU  - Duan,Y
AU  - Eaton,MD
AU  - Bluck,MJ
DO  - 10.1016/j.jcp.2021.110243
EP  - 14
PY  - 2021///
SN  - 0021-9991
SP  - 1
TI  - Fixed inducing points online Bayesian calibration for computer models with an application to a scale-resolving CFD simulation
T2  - Journal of Computational Physics
UR  - http://dx.doi.org/10.1016/j.jcp.2021.110243
UR  - https://www.sciencedirect.com/science/article/pii/S0021999121001388?via%3Dihub
UR  - http://hdl.handle.net/10044/1/87371
VL  - 434
ER  -
Download