CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   An absolute begginner - ADVICE NEEDED... (http://www.cfd-online.com/Forums/main/15324-absolute-begginner-advice-needed.html)

Milos Stanic June 20, 2008 05:38

An absolute begginner - ADVICE NEEDED...
 
Hi to whoever reads this and thnx in advance.

I am planning to go for PhD title in CFD area next summer and I would like to know some basics about serious dealing with CFD.

I am currently reading J.Andersons book and I have bought T.J.Chungs book too (I'll read that when I'm done with Anderson). I am familiar with main concepts of CFD (fluid mechanics, basic algorithms and some elementary math behind it). The things that I would like to know are:

1. - What exact level of math is neccessary to start working with CFD?

2. - At what point of your research do you actually get involved with serious programming (Which code do you suggest: C++, Matlab...?)?

3. - As a PhD student - do you work only with the codes or do you actually use software packages (Fluent, CFX...)?

Thank you once more and hopefully I'll be asking some more challenging questions soon. :)

cfd.newbie June 22, 2008 09:12

Re: An absolute begginner - ADVICE NEEDED...
 
Although I am new to CFD as well but I think I am in a position to answer your questions. Anderson is good but you dont need to read it from cover to cover. It depends what you are planning to do in CFD. If you can explain what yo will be doing in CFD, we can advice you accordingly. Also, read veersteeg's book, I think its excellent book for beginners.

1. As long as you know some calculs and understand its physical meaning you should be ok. Understand the meaning of each term in Navier-stokes equations and make sure you know what it does.

2. Well, it depends, if you will be working on code development or be user. Both fortran and C are useful for code development work.

3. You should ask and clarify this with your supervisor. Traditionally people use CFX or Fluent at the start and then later do some code work.

Good Luck


shantanu June 22, 2008 22:44

Re: An absolute begginner - ADVICE NEEDED...
 
Hi Milos Stanic,

first of all thats a good decision to do PhD in CFD. Here are few tips, which may be helpful to you.

1. Along with Anderson, you may also study a book by "Laney titled computational gasdynamics" but only the portion related to waves and PDE's. and study fundamental of mechanics (absolutely essential)

2. Try to understand and get a feel of different PDE's and their classification and their behaviour according to their classification. This is ABSOLUTELY important while working in CFD.

3. According to me, maths knowledge required is related to ODE, PDE(MUST), linear algebra, vector calculus (like greens' theorem, gradient, divergence concepts) and some computational knowledge like Finite difference method, finite volume method, finite element method (for fluid structure interaction)

4. I think serious programming may involved at the end of your first year, when you have defined your problem, understood its physics and devised a numerical scheme to solve it. I think c++ is more reliable and fast and you will have more control of everything BUT MATLAB is easy to work with and it is fast to develop code with MATLAB and lot of functions are readymade.

5. It depends whether you want to develop a code or use software. if you are working on something problem, which necessary features are not available in commercial softwares, then you need to develop ur own code else use some commercial software. But according to my experience, it is good to develop ur own code, as lot of concepts get clear when we work by ourselves.

Feel free to ask anything.

with best regards, Shantanu S. Mulay.

Milos Stanic June 23, 2008 07:25

Re: An absolute begginner - ADVICE NEEDED...
 
Thank you for your advice, it's just as I suspected.

I am quite familiar with Navier-Stokes equations and it's terms as well with Reynolds's equations. As for the math, I am fine with elementary ODE and PDE, linear algebra, FDM, so I guess I'll have to work on the rest.

I haven't yet decided which area of CFD would I go for, but I am interested in large variety of fields: automotive aerodynamics, high Mach number flows, turbomachinery and combustion... Those I would prefer, but almost anything would do since I believe it's relatively easy to transfer from one area to another.

Thank you once more for your answers and if you have some additional advice for me, please do not hesitate to write to me. It would mean a lot to me. Thanks again!

Vinayender June 23, 2008 08:51

Re: An absolute begginner - ADVICE NEEDED...
 
Hi,

I wont recomend you to use Matlab for your coding, instead use C++ or FORTRAN. Initially, you may be get tempted to use mabtlab as it is very friendly (eassy to porgram, debug and contains many inbuild functions) but, Matlab is very slow for computations (with compared with C,C++ and FORTRAN)and in later stage of your work you may feel that your simulations eats much of your time and may think of converting those codes in to C++ or FORTRAN. So be cautious at the begining of starting of your coding.

ALL THE BEST for your PhD...

Thanks, Vinayender.

andy June 23, 2008 11:37

Re: An absolute begginner - ADVICE NEEDED...
 
> 1. - What exact level of math is neccessary to start working with CFD?

The maths taught in engineering courses is usually fine to get going and then you will start to learn what is needed for your PhD.

> 2. - At what point of your research do you actually get involved with
: serious programming (Which code do you suggest: C++, Matlab...?)?

Normally your supervisor will either give you a research CFD code for you to modify in some way or ask you write to a simple code from scratch. The choice of language usually follows from what the CFD group is using. If the group uses language X then it would be wise to start with that.

> 3. - As a PhD student - do you work only with the codes or do you actually
: use software packages (Fluent, CFX...)?

The methods in commercial CFD codes are standard and therefore not suitable subjects for research in themselves. To study new CFD methods you will need to write or adapt CFD research codes. However, commercial codes can be useful as a reference to compare against and, sometimes, as a means to generate part of a CFD solution when the focus of your research is another.



All times are GMT -4. The time now is 11:13.