Basically I'm validating a research paper of heat exchanger with ansys fluent. when i gather the data in the experimental paper and so do some calculation for checking of nature of flow by evaluating the hydrodynamic and thermal entry lengths based on the data in the paper. im getting a non-fully developed (developing flow) and the entry lengths exceeding the geometry dimensions. i want to ask whether fluent models handles a developing flow or not? if No then Is there's a way to run developing flow?
Hi,
there might be two aspects of this question:
1) Generally the flow conditions are such, that they could be named fully turbulent. I.e. the Reynolds number of the flow is sufficiently high and the geometrical configuration is such, that a pronounced level of turbulence is developed due to shear layers, boundary layers, flow around objects etc.
Under such circumstances the flow length along a surface could be not sufficiently long to allow for the development of a fully developed boundary layer, because some L/D>20-50 might be required for this to occur. The usage of logarithmic law of the wall with a larger grid y+>32 would make the assumption, that the boundary layer is fully developed in all near-wall mesh points, which is under these circumstances obviously not true for the grid points being located in the region, where the turbulent boundary layer is still in developing stage. This would result in an error in the CFD solution. The error can be mitigated or reduced, if a mesh with a y+~1-2 is used. In that case the flow solver is usig the linear relationship:
u+ ~ y+
in order to determine the boundary conditions for the turbulence model equations and it allows for the prediction of a turbulent boundary layer from a not yet fully developed stage to a fully developed turbulent boundary layer.
But there might be conditions of the case 2):
Let's assume an aircraft in not disturbed atmospheric conditions. At the leading edge of the aircraft the almost non-disturbed flow hits the surface of the aircraft and over a certain length the developing boundary layer is not a turbulent boundary layer, but a laminar boundary layer (with correspondingly different BL profile). Only after a certain flow length this laminar BL is transitioning into a turbulent BL. Obviously a 2-equation turbulence model, which is made for turbulent flow conditions, will not be able to predict the laminar BL profile in the first section of the aircraft surface before the transition point occurs.
For the more accurate prediction of such transitional flows (transitional from laminar to fully turbulent flow conditions) the so-called transition models have been developed. One of them is e.g. the Gamma-Theta transition model. Other models exist. Some of them solve an additional intermittency transport equation in order to account for the laminar-turbulent transition.
https://en.wikipedia.org/wiki/Gamma-Re_Transition_Model
See the corresponding documentation pages of ANSYS CFX / Fluent for more details. Since the process of laminar-turbulent transition is even harder to be predicted then full turbulent flow conditions, the meshing requirements are even more severe. Some of the transition models require meshes with a near-wall y+ being less then 0.1 in order to work properly.
Regards,
Dr. Th. Frank.
Hi,
there might be two aspects of this question:
1) Generally the flow conditions are such, that they could be named fully turbulent. I.e. the Reynolds number of the flow is sufficiently high and the geometrical configuration is such, that a pronounced level of turbulence is developed due to shear layers, boundary layers, flow around objects etc.
Under such circumstances the flow length along a surface could be not sufficiently long to allow for the development of a fully developed boundary layer, because some L/D>20-50 might be required for this to occur. The usage of logarithmic law of the wall with a larger grid y+>32 would make the assumption, that the boundary layer is fully developed in all near-wall mesh points, which is under these circumstances obviously not true for the grid points being located in the region, where the turbulent boundary layer is still in developing stage. This would result in an error in the CFD solution. The error can be mitigated or reduced, if a mesh with a y+~1-2 is used. In that case the flow solver is usig the linear relationship:
u+ ~ y+
in order to determine the boundary conditions for the turbulence model equations and it allows for the prediction of a turbulent boundary layer from a not yet fully developed stage to a fully developed turbulent boundary layer.
But there might be conditions of the case 2):
Let's assume an aircraft in not disturbed atmospheric conditions. At the leading edge of the aircraft the almost non-disturbed flow hits the surface of the aircraft and over a certain length the developing boundary layer is not a turbulent boundary layer, but a laminar boundary layer (with correspondingly different BL profile). Only after a certain flow length this laminar BL is transitioning into a turbulent BL. Obviously a 2-equation turbulence model, which is made for turbulent flow conditions, will not be able to predict the laminar BL profile in the first section of the aircraft surface before the transition point occurs.
For the more accurate prediction of such transitional flows (transitional from laminar to fully turbulent flow conditions) the so-called transition models have been developed. One of them is e.g. the Gamma-Theta transition model. Other models exist. Some of them solve an additional intermittency transport equation in order to account for the laminar-turbulent transition.
https://en.wikipedia.org/wiki/Gamma-Re_Transition_Model
See the corresponding documentation pages of ANSYS CFX / Fluent for more details. Since the process of laminar-turbulent transition is even harder to be predicted then full turbulent flow conditions, the meshing requirements are even more severe. Some of the transition models require meshes with a near-wall y+ being less then 0.1 in order to work properly.
Regards,
Dr. Th. Frank.
Thomas Frank already addressed many points, I would just add that, no matter the CFD code is used, RANS is by definition a formulation for solving only fully developed and statistically steady turbulent flows. Even if URANS has the appearence of an unsteady term, it is debated the real physical meaning of describing a time-evolving problem with such a formulation.
In conclusion, solving transitional conditions is possible in both LES and DNS formulations.
- Badges
- Science topic