A turbulent heat flux prediction framework based on tensor representation theory and machine learning. (July 2018)

• Record Type: Journal Article
• Authors: Sotgiu, C.; Weigand, B.; Semmler, K.
• Is Part Of: International communications in heat and mass transfer. Volume 95(2018:Jul.)
• Page Start: 74
• Page End: 79
• Publication Date: 2018-07
• Subjects: DNS Direct Numerical Simulation -- EBRSM Elliptic Blending Reynolds Stress Model -- LES Large Eddy Simulation -- RANS Reynolds-Averaged Navier-Stokes -- ReLU Rectified Linear Unit -- YSC Younis-Speziale-Clark scalar flux model; Heat -- Transmission -- Periodicals; Mass transfer -- Periodicals; Chaleur -- Transmission -- Périodiques; Transfert de masse -- Périodiques; Heat -- Transmission; Mass transfer; Periodicals; 621.4022
• DOI: 10.1016/j.icheatmasstransfer.2018.04.005
• Languages: English
• ISSNs: 0735-1933

Abstract

Abstract: The need of more sophisticated cooling schemes for gas turbine blades is continuously increasing, since the hot gas maximum temperature in gas turbines has a direct influence on the engine thermal efficiency. Here, predictions of the heat transfer by means of Computational Fluid Dynamics (CFD) can complement or even reduce the number of experimental investigations. However, the modelling of the turbulent heat fluxes in the Reynolds-averaged Navier-Stokes equations heavily relies on empirical approaches. We propose a new framework for the prediction of the turbulent heat fluxes based on machine learning and tensor representation theory. A data-driven model is constructed based on the tensor description of Younis et al. (2005) and implemented in OpenFOAM. Its validation for Poiseuille flow at different Reynolds numbers shows very good agreement with reference data.