Abstract
Machine Learning (ML) algorithms have become popular in many fields, including applications related to turbomachinery and heat transfer. The key properties of ML are the capability to partially tackle the problem of slowing down of Moore’s law and to dig-out correlations within large datasets like those available on turbomachinery. Data come from experiments and simulations with different degree of accuracy, according to the test-rig or the CFD approach. When dealing with modelling of turbulent flows in turbomachinery there is a constant trade-off between accuracy and computational costs, but starting from the large amount of data on turbomachinery performance, with ML it is possible to train a learner to correct and improve CFD. The aim of this work is to investigate an innovative data-driven approach that could lead to a significant improvement in the analysis of heat transfer in turbulent flows. The effects of Reynolds number and wall temperature on heat transfer for a double forward-facing step with two squared obstacles were investigated by numerical simulations carried out in OpenFOAM. Then a machine-learnt model was derived using a regression algorithm. The results of regressor showed that a data-driven approach can effectively predict results of the RANS model.
Highlights
Machine Learning (ML) algorithms have become popular in many fields, including applications related to turbomachinery and heat transfer
Modelling of heat transfer is a key issue in many engineering applications, especially when dealing with Computational Fluid Dynamics (CFD) of turbulent flows with heat transfer
In the following we discuss the possibility of deriving the distributions of turbulent viscosity and diffusivity in a low-Reynolds RANS approach using machine learning, and in particular to apply this strategy to a complex flow configuration with multiple flow features, such as geometry-induced separation, impingement, reattachment, wakes using data from multiple Reynolds numbers
Summary
Modelling of heat transfer is a key issue in many engineering applications, especially when dealing with Computational Fluid Dynamics (CFD) of turbulent flows with heat transfer. Most comes from the Boussinesq approximation and in particular to the limit of having Reynolds fluxes aligned to Reynolds stresses [1]. This condition is nonphysical, especially in strongly non-isotropic turbulence regions and in particular along solid walls. In the following we discuss the possibility of deriving the distributions of turbulent viscosity and diffusivity in a low-Reynolds RANS approach using machine learning, and in particular to apply this strategy to a complex flow configuration with multiple flow features, such as geometry-induced separation, impingement, reattachment, wakes using data from multiple Reynolds numbers.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
