Bi-fidelity UQ with combination of co-Kriging and arbitrary polynomial chaos: Film cooling with back facing step using RANS and DES. (March 2019)

• Record Type: Journal Article
• Authors: Sakai, Eiji; Bai, Meng; Ahlfeld, Richard; Klemmer, Kerry; Montomoli, Francesco
• Is Part Of: International journal of heat and mass transfer. Volume 131(2019)
• Page Start: 261
• Page End: 272
• Publication Date: 2019-03
• Subjects: Uncertainty quantification -- Film cooling -- Multi-fidelity CFD -- Back facing step; Heat -- Transmission -- Periodicals; Mass transfer -- Periodicals; Chaleur -- Transmission -- Périodiques; Transfert de masse -- Périodiques; Electronic journals; 621.4022
• DOI: 10.1016/j.ijheatmasstransfer.2018.10.071
• Languages: English
• ISSNs: 0017-9310

Abstract

Highlights: The improvement in the film effectiveness by the back facing step is shown. The co-Kriging can predict high-fidelity DES results from low-fidelity RANS results. The proposed method can perform higher order UQ with less computational cost. Abstract: Uncertainty quantification analysis of a film cooled flat plate is performed by Sparse Approximation of Moment Based Arbitrary polynomial chaos (SAMBA). To combine simulations of different fidelities, such as DES and RANS, a co-Kriging formulation is used and combined with the Non-Intrusive Polynomial Chaos (NIPC) framework. The co-Kriging allowed to have the same accuracy in the prediction as all the collocation points were obtained from DES simulation, leveraging low fidelity RANS cases. As test case, a film cooled flat plate is used. Such configuration is representative of coolant systems of gas turbines. The configuration used present a back facing step in front of the coolant hole. This design has been recently suggested to improve the cooling performance but is strongly affected by the geometrical variations of the step. This work uses a UQ formulation to quantify these effects. The co-Kriging formulation successfully estimates the DES results from the RANS results, and the probability density function of film cooling effectiveness obtained with mixed RANS/DES cases adequately matches with that obtained DES only simulations. This work shows the advantage of combining the SAMBA and co-Kriging methods to perform higher order UQ analysis with less computational cost. Such approach is particularly useful in industrial applications where a limited number of high fidelity simulations is available.