Dear Reza,

There are standard formats for CFD data which are quite efficient for general purpose CFD codes some include HDF, HDF5, CGNS etc ... which most of the post-processors and grid generators can read and write. Another thing to consider is that whether your grid is structured or not

However if you are just coding a specific problem it would probably be much simpler to decide which mesh generator and post processing software you're going to use and just write a code for that specific format which are usually very simple to code specially for structured and block-structured data.

To my experience for post-processing VTK and tecplot binaries are fast and I had no problems before. if you need to read a grid from a grid generator save it in CGNS which is efficient and simple.

Good luck

Sina

Hi Sina,

thanks for the answer but that is not what I am looking for. I am looking for the methodology to save the grid points and their connectivity since I am going to write a subroutine to load the point and hold them for the solver to ask for. what I am looking for is how the grid data structure looks like to give a fast and efficient access to the solver. the method that explain the nods connectivity, boundary nodes, cells info and so on.

Hi Reza,

You need then to write in your output file both the triangles and the nodes, and that for each object at every time step of the simulation. The file will then be read using e.g. Matlab in order to visualize the sequence. You can for example define a function called WriteTriangles_ObjectSequence which writes into file the node indexes that define every triangle for a given object (3 integers per line corresponding to the nodes defining each triangle).

The second function that you can define is WriteNodes_ObjectSequence. It will write into the same file the 3 coordinates of each node (3 doubles per line).

Hope this will help you a bit to answer to your question.

Bests

a

My recommendation would be to look into the PATRAN/NASTRAN datatypes (if you're dealing with finite element or similar). Then look into an open source code like Nek5000 for how they handle connectivity and storage.

You're completely correct that this really isn't discussed much in CFD books. I had a similar problem at the beginning of my research. The bulk of help in this area came from contacting others directly about how they handle data and getting first hand experience with their codes. It allowed me to see the strengths and weaknesses of various storage schemes.

These types of conversations tend to come up more in applied math fields. But, you should look into 'linked list' type storage as well as adjacency matrices for storage and access of data.

For example, for easy storage:

Create an array of struct/type elements in an arbitrarily ordered one-dimensional list. Traverse the list and create pointers to bounding elements, therefore 6 faces would be 6 pointers. You then have a linked list of connectivity, where you can traverse the connection of any element to any other element. Costly, but simple from a 'thinking about it' point of view.

try with the Gambit software

create a sample mesh

read the journal file which include all creation commands

The file format depends only in your goals. If you want to write software you have complete freedom. If you want to use 3th party or commercial software you should understand the file formats they used. Additionally, you may want to use a specific mesh generator. Assuming that you are developing a solver you make look at the gmsh file format (http://geuz.org/gmsh/). I have been using it for a while and I can use gmsh for pre and post-processing. Gmsh is free and runs in several platforms.

If your question is what efficient data structure can be used for CFD then ithink that you must read this book :

Applied Computational Fluid Dynamics Techniques: An Introduction Based on Finite Element Methods, 2nd Edition  Wiley

Rainald Löhner 

you can fou a chapter on GRID GENERATION - EDGE BASED DATA - ADAPTIVE MESH REFINEMENT - EFFICIENT USE OF COMPUTER HARDWARE ...

You can found also many interesting solver on the homepage of Rainald Lohner

http://www.imr.sandia.gov/papers/authors/lohner.html

Dear  Reza,

What I understand  from your post is that you would like to make a connection between your CFD code and a [1.mesh generation or a 2.generated mesh] output.

1. If you also are developing a mesh generation code which for simple problems is very easy, therefore you won't need to have two different codes (CFD and mesh generation) . You may simply write a subroutine in your main code for mesh generation.

2. However, if you would like to use the output of a mesh which is generated by another software or your mesh generation is too complicated to place it into the main code, then you would need first to choose the graphical environment you would like to present your results. Different graphical environments like Tecplot, Surfer, MATLAB etc. have different formatting. Once you decide on this, then you need to transform the format of the mesh into the format that your specific software requires. Hope this is helpful.