[Disco_Broccoli]

Hi CFD community,

I am new to CFD but took some university level classes. I have an beginner understanding on the numerical methods such as how to discretize a simple scheme, validate with MEA, review stability using von Neumann method, also learned about the RK time stepping. However, applying my own 1D/2D solver for FDM and FEM (DGM) in the classes helped most (Isentropic vortex and lid driven cavity).

I am now in a master degree and I am reading about turbulence, inevitably RANS is mentionned in Wilcox and Pope. Except for derivation found in books I have a fundamental problem with it: How does one write a RANS solver?

From my limited background, I have trouble picturing where/how the time-stepping is involved. Mainly, from google images I see results that are not time dependent so if I am correct we can't see the eddies at a given time.

Does anyone know any good books teaching about implementing RANS? Is going into the OpenFOAM source code a good idea?

Thanks!

[FMDenaro]

Originally Posted by Disco_Broccoli

Hi CFD community,

I am new to CFD but took some university level classes. I have an beginner understanding on the numerical methods such as how to discretize a simple scheme, validate with MEA, review stability using von Neumann method, also learned about the RK time stepping. However, applying my own 1D/2D solver for FDM and FEM (DGM) in the classes helped most (Isentropic vortex and lid driven cavity).

I am now in a master degree and I am reading about turbulence, inevitably RANS is mentionned in Wilcox and Pope. Except for derivation found in books I have a fundamental problem with it: How does one write a RANS solver?

From my limited background, I have trouble picturing where/how the time-stepping is involved. Mainly, from google images I see results that are not time dependent so if I am correct we can't see the eddies at a given time.

Does anyone know any good books teaching about implementing RANS? Is going into the OpenFOAM source code a good idea?

Thanks!

RANS formulation solves the steady state equations, the time derivatives vanish.
The solver is as a same as for a laminar steady flow the difference being in the additional term taking into account the turbulence model.

[Disco_Broccoli]

Originally Posted by FMDenaro

RANS formulation solves the steady state equations, the time derivatives vanish.
The solver is as a same as for a laminar steady flow the difference being in the additional term taking into account the turbulence model.

Thank you for the feedback. Just to clarify, when one talks about laminar steady flow. Are they referring to solving algebraic equations? Also, if the time derivative vanishes does that mean there is no time-stepping?

Thank you!

[FMDenaro]

- You always get a final algebric system to solve.

- The solver can be also time-dependent, you run the simulation until the time derivatives numerical vanish. In case the equations are for the steady state there is not a physical time step, you have only a residual to drive to be under certain threshold.

[Disco_Broccoli]

Originally Posted by FMDenaro

- You always get a final algebric system to solve.

- The solver can be also time-dependent, you run the simulation until the time derivatives numerical vanish. In case the equations are for the steady state there is not a physical time step, you have only a residual to drive to be under certain threshold.

Do you have any recommendation for readings on the subject of RANS implementation? Thanks for the fast response!

[arjun]

Originally Posted by Disco_Broccoli

Do you have any recommendation for readings on the subject of RANS implementation? Thanks for the fast response!

http://ftp.demec.ufpr.br/disciplinas...sekera_2ed.pdf

[t.teschner]

There are a few useful books in my opinion which you may want to consult. The book of Pope "Turbulent Flows" and Davidson "Turbulence" may be a good starting point about turbulent flow in general, but probably don't go into that much depth as you would like, especially if you are interested in more RANS modelling. Here is a list of my recommendation, in no particular order:

An Introduction to Computational Fluid Dynamics: The Finite Volume Method by Versteeg and Malalasekera: While this book is more about the general discretisation of the governing equations used in CFD, they have an excellent chapter on Turbulence (its physics but also its modelling). It is relatively compact but gives you all you need to get started.

Statistical Turbulence Modelling For Fluid Dynamics - Demystified: An Introductory Text For Graduate Engineering Students by Michael Leschziner: This is probably the book you want to be reading if you want to understand more about the different RANS models and how they are constructed. It doesn't necessarily give you the discretised equations, but RANS modelling by itself is complicated enough and this book is a welcomed guide that will help you with RANS specific questions.

Turbulence Modeling for CFD by David C. Wilcox: You mentioned this book but have a look in the later chapters, there are specific hints about the implementation, though maybe not as much as you would like. Still it is a useful resource.

Computational Fluid Dynamics by Jiri Blazek: This book basically gives you all the equations in a discretised form which is very helpful when writing your RANS code. This book reads like a manual for the TAU CFD solver (developed by DLR (german aerospace center)) and hence is very close to how the equations are actually discretised. The focus of this book is on unstructured grids, which may help to write a generic solver.

The Finite Volume Method in Computational Fluid Dynamics by Moukalled, F., Mangani, L., Darwish, M.: I have some reservations recommending this book speciffically for turbulence. Don't get me wrong, this is an excellent book and I use it for almost all of my discretisation related questions / reading and I always find something new in there, I just feel that the RANS chapter could have been a bit more thorough to live up to the excellent quality in the rest of the book. Still, it may be very useful to you as they write out the equations exactly as they are discretised, which is probably what you want. I just feel some additional background reading will help you here.

A New Hypothesis on the Anisotropic Reynolds Stress Tensor for Turbulent Flows: Volume I and II by Laszlo Konozsy: This is a relatively new book and mainly concerned with a new anisotropic RANS model (based on the k-w model), however, parts of the book are dedicated for RANS implementation specific issues (and full source codes are provided) which may be yet another useful resource.

Mathematics, Numerics, Derivations and OpenFOAM by Tobias Holzmann: This may not be directly related to your question but a very useful reference in general for the derivation of the scalar turbulence equations, which may be of use for other people stumbling upon this thread. Tobias is a quite active member in this forum and I thought he deserved a shout out, I just bought a copy and find it very useful as a reference (there is also a free version of the book available).

Those are my suggestions, hope these will help to get started

[Disco_Broccoli]

Originally Posted by t.teschner

There are a few useful books in my opinion which you may want to consult. The book of Pope "Turbulent Flows" and Davidson "Turbulence" may be a good starting point about turbulent flow in general, but probably don't go into that much depth as you would like, especially if you are interested in more RANS modelling. Here is a list of my recommendation, in no particular order:

An Introduction to Computational Fluid Dynamics: The Finite Volume Method by Versteeg and Malalasekera: While this book is more about the general discretisation of the governing equations used in CFD, they have an excellent chapter on Turbulence (its physics but also its modelling). It is relatively compact but gives you all you need to get started.

Statistical Turbulence Modelling For Fluid Dynamics - Demystified: An Introductory Text For Graduate Engineering Students by Michael Leschziner: This is probably the book you want to be reading if you want to understand more about the different RANS models and how they are constructed. It doesn't necessarily give you the discretised equations, but RANS modelling by itself is complicated enough and this book is a welcomed guide that will help you with RANS specific questions.

Turbulence Modeling for CFD by David C. Wilcox: You mentioned this book but have a look in the later chapters, there are specific hints about the implementation, though maybe not as much as you would like. Still it is a useful resource.

Computational Fluid Dynamics by Jiri Blazek: This book basically gives you all the equations in a discretised form which is very helpful when writing your RANS code. This book reads like a manual for the TAU CFD solver (developed by DLR (german aerospace center)) and hence is very close to how the equations are actually discretised. The focus of this book is on unstructured grids, which may help to write a generic solver.

The Finite Volume Method in Computational Fluid Dynamics by Moukalled, F., Mangani, L., Darwish, M.: I have some reservations recommending this book speciffically for turbulence. Don't get me wrong, this is an excellent book and I use it for almost all of my discretisation related questions / reading and I always find something new in there, I just feel that the RANS chapter could have been a bit more thorough to live up to the excellent quality in the rest of the book. Still, it may be very useful to you as they write out the equations exactly as they are discretised, which is probably what you want. I just feel some additional background reading will help you here.

A New Hypothesis on the Anisotropic Reynolds Stress Tensor for Turbulent Flows: Volume I and II by Laszlo Konozsy: This is a relatively new book and mainly concerned with a new anisotropic RANS model (based on the k-w model), however, parts of the book are dedicated for RANS implementation specific issues which may be yet another useful resource.

Those are my suggestions, hope these will help to get started

Thank you for sharing your suggestion. I will take a closer look at each of them!

[sbaffini]

A good source of material for a modern FV solver (including its turbulence part) are the Fluent references of the 90s (basically, everything written by Mathur, Murthy, Weiss, Smith, Caraeni and Kim)

[arjun]

Originally Posted by sbaffini

A good source of material for a modern FV solver (including its turbulence part) are the Fluent references of the 90s (basically, everything written by Mathur, Murthy, Weiss, Smith, Caraeni and Kim)

Send me email if someone wants them :-D

PS: You are right. I learned a lot from them and refer to them still (in 2020).

[FMDenaro]

Originally Posted by sbaffini

A good source of material for a modern FV solver (including its turbulence part) are the Fluent references of the 90s (basically, everything written by Mathur, Murthy, Weiss, Smith, Caraeni and Kim)

The progress is no longer a progress...

[sbaffini]

Originally Posted by FMDenaro

The progress is no longer a progress...

Well, there has certainly been progress in some aspects, but for what concerns the solvers and their workflows, the basic schemes and the particular choices that actually lead to a robust implementation, there is nothing really new, at least nothing to be meant as an actual replacement for a solid base recipe useful to novices.

For example, there are still people working on gradients and limiters, but wouldn't consider this material for someone just starting.

In contrast, I think that the Fluent material is among the most useful, self-contained ones, if not the only one at all. And the fact that it is the most used commercial CFD code, I think, adds to the relevance.

Also, if you ever tried to read other CFD manuals, in comparison, they look as black magic (which is bad)

[q__]

Of course, there is a great book which perfectly fits to your query as previously mentioned. I think it should serve as a standard textbook at universities for the beginners IMO.


Here it is:


https://www.google.com/url?sa=t&rct=...wzmhnVjH-bmyYm


BR,
Jakub

[arjun]

I would love to know what progress that is made in CFD after year 2000 though.



Most of the things that are used published before 2000 and all the major help is in terms of hardware. (GPU etc). Its only now that AI has started to influence the simulation aspects of CFD but that is far far from mainstream.



Off top of my head pressure based coupled solver's development may be something that could be classified that became prominent after 2000.

[sbaffini]

Originally Posted by arjun

I would love to know what progress that is made in CFD after year 2000 though.



Most of the things that are used published before 2000 and all the major help is in terms of hardware. (GPU etc). Its only now that AI has started to influence the simulation aspects of CFD but that is far far from mainstream.



Off top of my head pressure based coupled solver's development may be something that could be classified that became prominent after 2000.

I was mostly just saying, just for the sake of the argument. However, if we exclude the algebraic solver side (which I don't know), every now and then you can still find people coming out with simple, nice things that fit well.

For example, in my code I use a mixed, Least squares/Green Gauss approach for gradients, where the GG part is based on a 2018 work. Nothing fancy, but it overcomes a serious problem of the classical GG implementation that was unsolved to date.

Also, pressure based coupled solvers are, from what I can see, still in a sort of infancy and new works keep coming out

[JBeilke]

Originally Posted by arjun

I would love to know what progress that is made in CFD after year 2000 though.

QGD is completely new - at least outside Russia :-)

QGDsolver - OpenFOAM computational framework for fluid flows based on regularized equ

[arjun]

Originally Posted by JBeilke

QGD is completely new - at least outside Russia :-)

QGDsolver - OpenFOAM computational framework for fluid flows based on regularized equ

This is indeed very interesting and could be very useful if used properly. The thing is that classic method decouples at very very small time step size and it has been long standing problem of CFD. This could be very useful there.
I would try and see if i understand enough to get it in Wildkatze as an option to the users. So I hope i have understood by first reading but things could be different when actually implemented.

(interestingly at the core is the method that i have been using for gradients in Wildkatze for 3 years now. That is calculating the gradients at the face).

Originally Posted by sbaffini

For example, in my code I use a mixed, Least squares/Green Gauss approach for gradients, where the GG part is based on a 2018 work. Nothing fancy, but it overcomes a serious problem of the classical GG implementation that was unsolved to date.

Also, pressure based coupled solvers are, from what I can see, still in a sort of infancy and new works keep coming out

Could you share what problem was that and how it was overcome.

I know that green gauss are affected by skew. So to mitigate it I used to have an option where the least square method was used at the cell face to get the gradient and using this gradient cell face values were computed. This would have taken cared of skew mostly. (this was costly so it is removed and replaced)

Now I provide option to the user to calculate least square gradient from extended stencil so that the skew does not affect the results much.

[sbaffini]

This is the work I was referring to:

https://www.researchgate.net/publica...reconstruction

And, instead of using iterstively, I use it after the least squares

[arjun]

Originally Posted by sbaffini

This is the work I was referring to:

https://www.researchgate.net/publica...reconstruction

And, instead of using iterstively, I use it after the least squares

Paolo


Thank you very much for pointing to the paper. Looks very interesting work indeed.

I will give it a serious trial very very soon. Most probably will share on linkedin so you could see what comes out of the study.



Arjun

[sbaffini]

Originally Posted by arjun

Paolo


Thank you very much for pointing to the paper. Looks very interesting work indeed.

I will give it a serious trial very very soon. Most probably will share on linkedin so you could see what comes out of the study.



Arjun

Also, I now see, there is an additional work on RG citing it that also seems interesting.

So, little pieces but, at least for gradient reconstruction there seems to be stuff in people's pipelines.

I also find interesting several of the works of Nishikawa (which are all post 2000)