Hello,

I am new to the field of simulating the flow around moving bodies. I have so far thought about using Chimera (overset grids) since it is a nice method.

However, I am also thinking of coupling a cfd solver with a mesher. The scenario would be something like: 1) Generate an unstructured mesh 2) Compute the flow and extract the forces acting on the body. 3) Move the body corresponding to an ODE or to a given trajektory 4) Remesh the domain and compute again the flow

Here are my questions: 1) Is this approach feasible? 2) Has anybody experiences with coupling a CFD solver with a mesher? I mean, the solver knows only the mesh. Yet, the mesher needs geomtries (like CAD surfaces) in order to do the meshing. Is there a method to go back from the CFD mesh to the CAD surfaces in order to feed the mesher with the new position of the body? Or are there any better approaches to coupling a CFD solver and a mesher?

Thank you very much for any hints, explanations, and infos.

Alan

If you seek to remesh the entire domain you will run into issues of grid quality - i.e. can you insure that the mesher will always give you a good a grid each time thru. In addition, the regridding becomes a real time sink when you go to 3D. Most of the approaches that I have seen that are not Chimera/overset involve "stretch and break" where a very small region around the body is allowed to deform to a point and then that smaller grid region is re-gridded. The deformation is typically handled via the solution of a "spring analogy" differential equation to reposition the grid points at each time step. With stretch and break, the grid immediately adjacent to the body can be held fixed and the regridding is focused on regions where the mesh generator is usually better-behaved. In addition the smaller domain of remeshing means less time required by the mesher.

Having said all that, overset is more flexible but limited to structured grids at present. If you want to use an unstructured solver then stretch and break may be your best option.

SInce the problem is associated with moving bodies, obviously it is an unsteady flow, which means you need information from the previous timestep. This means you need a robust method to incorporate (interpolate) information from the previous mesh onto the mesh of the current timestep. In principle, what you're suggesting is doable, but there are a lot of little details you need to be concerned with (if you're new to CFD, this will be very challenging).

The choice of strategy also depends on the type of problem you're trying to solve. Is the body experiencing perturbative motion (so that just one original mesh distribution would be fine), or does it go through very large changes? And even then, are there multiple bodies with relative motion/rotation or just one body in motion?

Problems of this kind are ideal for particle-based, "grid-free" solution methods if the flow is incompressible. In this case, you essentially only need to mesh the surfaces and not worry about meshing the fluid domain. Efficient implementation would take time, but it depends on your objective. Do you want to solve a problem or develop methodology? If the former, you should look around for available technologies.

Adrin Gharakhani

Thank you both for you suggestions and explanation.

I plan to use readily available tools for my simulations. I have heard that, e.g. Fluent and StarCD, can do remeshing and I was wondering how does it work in general and what are the different approaches with advantages and disadvantages?

Currently, I am putting together a survey on methods (and software tools) for computing the flow around moving bodies. The problem I want to compute is a closing and opening valve. The valve body is typically a sphere that will not remain in the center once the valve has openend but it will "stick" to one of the surrounding walls... It would be nice to be able to simulate all this.

I am not new to CFD but I have absolutely no background on CFD for moving meshes (bodies).

Please do not hesitate to post further comments, hints or references. I appreciate any info I can get.

Thank you very much, Alan

Have a look at

http://www.adapco-online.com

There are some articles about moving mesh and engine modeling.

"Currently, I am putting together a survey on methods..."

Excellent. I was going to say that this is exactly where you should start when you're getting into terrain that is new to you but has been out there for the longest time. There is a wide variety of techniques to treat this problem as you can sense from reading the responses you got at this forum. Make sure you know all of them and understand the advantages and disadvantages of commonly used methods, before trying to re-invent the wheel. A survey involves tedious work, but in the end it will save you a huge amount of time. It's good to know that your problem is not new. I am sure there actually are surveys on this problem available (why don't you search the Annual Review of Fluid Mechanics). Your own survey should start by reading one of the older or more recent surveys on the topic.

You have any references or links to "stretch and break" approach?

Hi, Alan,

As the developer for general moving object model for FLOW-3D, I welcome you to check our website. If you need more information, please let me know your personal email address and I will send you some valve simulation results.

http://www.flow3d.com/GMO.htm

George Wei

Hi, Alan,

As the developer for general moving object model for FLOW-3D, I welcome you to check our website http://www.flow3d.com/GMO.htm

If you need more information, please let me know your personal email address and I will send you some valve simulation results. You may also contact me at george@flow3d.com

George Wei