Heat Transfer Correction Modelling for RANS Simulations on Rough Surfaces

Heat Transfer Correction Modelling for RANS Simulations on Rough Surfaces

Beitler, Jari (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Energy Technology)

Peeters, J.W.R. (mentor)
Pecnik, Rene (graduation committee)
Hooman, K. (graduation committee)
Boersma, B.J. (graduation committee)

Delft University of Technology

Mechanical Engineering | Energy, Flow and Process Technology

2022-09-21

The Nuclear Research and consultancy Group (NRG) performs CFD simulations to improve the safety of nuclear installations and conducts research leading to new computational methods. One of the problems encountered in nuclear installations is that vital parts of the reactor, such as the fuel rod bundle, corrode over time. Corrosion leads to roughness elements on the surface which affect the heat transfer. To increase safety and reduce costs, it is of paramount importance to predict the heat transfer accurately when the surface is rough. The effect of roughness on a flow is an increase in both skin friction and heat transfer. Current models rely on the Reynolds analogy to calculate the heat transfer. Because in most applications pressure has no effect on heat transfer, it is often overestimated. In this thesis the accuracy of RANS modelling with respect to predicting heat transfer rates from or to a rough wall has been investigated.

It was found that a model for Low Reynolds Number meshes showed promising results. A downside of a RANS simulation on a Low Reynolds Number mesh is the increased amount of computational time due to fully resolving the velocity and temperature profiles compared to one on a High Reynolds Number mesh where wall functions are used. Therefore the model has been applied on several High Reynolds Number meshes and was compared to DNS results from literature. The results clearly showed that the mesh size was of influence on the predictions. However, a possibility was identified to reduce the mesh size dependency. The damping function was found to be responsible for this dependency and was reformulated using DNS data, resulting in a fitted equation for two different Prandtl numbers. The calibrated damping function which was made for a Prandtl number of 0.7 was validated with experimental results. It was found that the adjusted model could predict the Stanton number accurately and that the mesh size dependency was greatly reduced.

For future research it is recommended that the damping function should be calibrated for other Prandtl numbers as well, so the different damping functions could be combined in a single Prandtl dependent equation.

Heat Transfer
RANS
Roughness
CFD

http://resolver.tudelft.nl/uuid:61aca476-9485-49f3-b928-605aa5bc3d07

2023-09-21

Student theses

master thesis

© 2022 Jari Beitler