Mesh Generation

[b1ck0]

Hello,
I am intending to develope an application that creates planar mesh ( triangles ) over an existing plane and over an existing wave grid. But I don't have much background knowlege about mesh requirements, I mean what kind of triangles I have to create? Is there any requirement about area ratio of near triangles, or about the orientation of normal vectors and etc. I will look forward your reply

Thanks in advance!

[Tim B]

Your intended purpose will indicate requirements for the mesh. Why do you want to generate a triangle surface mesh?

There are quite a few open-source surface grid generation tools, mostly written for FEA work. THey might be worht a look.

Hope asking obvious questions helps,

Tim B.

[b1ck0]

Well, I want to make tessellated boundary surfaces of my fluid domain. In matter of fact I'm intending to make an application which automaticlly generates these tessellated surfaces if I have an existing hull form. For example HullSpeed and nu-Shallo have such fatures ...

[Tim B]

Ok, let me re-phrase that... what is the problem you're trying to solve?

Resistance, dynamics, local flow conditions wake improvement, wave system analysis? There are a lot of possible problems, and only you know which one. Which means that only you can decide which quality metrics are important.

From a software engineering point of view, you should capture the requirements first. They may change, but at least you have an idea of what you're starting with.

Tim B.
Tim B.

[b1ck0]

Mainly resistance and dynamics. I already have wave grid from potential flow analysis and now I just want to run RANSE ... I want to make the tessellated boundary surfaces and after that I will create volume mesh of that domain with numeca or gambit etc ...

[Tim B]

I thought GAMBIT had a surface mesher, but I haven't used it so I may be wrong. There are loads of surface mesh generators around, and you might even be able to use a render mesh for the first pass if you are careful about cell sizes and aspect ratios.

Dynamics (as in ship motions) usually deals with a hull-only mesh, or hullforms described by offsets.

With resistance calculations, be sure to check your results against at least one empirical method.

Good luck,

Tim B.

[b1ck0]

Thanks for advises but I want to make my own surface mesher. Gambit have surface and volume meshers, but in order to use the volume mesher I have to make surface mesh on boundary planes ( manually ) so I want to make this process automatic.

[Leo Lazauskas]

Originally Posted by b1ck0

Thanks for advises but I want to make my own surface mesher. Gambit have surface and volume meshers, but in order to use the volume mesher I have to make surface mesh on boundary planes ( manually ) so I want to make this process automatic.

Suppose that you want to iterate the attitude of a vessel until it is in equilibrium with hydrostatic and hydrodynamic forces and moments (e.g. squat).

Will you have to remesh the hull at each step?
If so, that seems like a very lengthy process in CFD, apart from the need to re-calculate the flow field at each step.

I'm asking because I have not seen many CFD results for those sort of calculations.

All the best,
Leo.

[DCockey]

Originally Posted by Leo Lazauskas

Suppose that you want to iterate the attitude of a vessel until it is in equilibrium with hydrostatic and hydrodynamic forces and moments (e.g. squat).

Will you have to remesh the hull at each step?
If so, that seems like a very lengthy process in CFD, apart from the need to re-calculate the flow field at each step.

I'm asking because I have not seen many CFD results for those sort of calculations.

All the best,
Leo.

One possible approach for the hull surface only:

Mesh the surface of the hull higher than the waves will ever be. Then at each step "trim" the mesh at the water surface. Triming a triangular mesh will result in both quadralaterals at the new edge. The quadralaterals can be easily split into triangles. If it's a structured mesh to begin then there may need to be some special handling of the mesh adjacent to the edge. That takes care of the hull surface.

A volume mesh with an arbitrary changing surface is more difficult.

[Leo Lazauskas]

Originally Posted by DCockey

One possible approach for the hull surface only:

Mesh the surface of the hull higher than the waves will ever be. Then at each step "trim" the mesh at the water surface. Triming a triangular mesh will result in both quadralaterals at the new edge. The quadralaterals can be easily split into triangles. If it's a structured mesh to begin then there may need to be some special handling of the mesh adjacent to the edge. That takes care of the hull surface.

A volume mesh with an arbitrary changing surface is more difficult.

Over-hanging transoms are often messy things to handle, particularly when the hull trims so the underside of the transom just touches the water, or becomes immersed during one iteration, and emerged during another.
That will keep you busy, B1ck0!

[DCockey]

Originally Posted by Leo Lazauskas

Over-hanging transoms are often messy things to handle, particularly when the hull trims so the underside of the transom just touches the water, or becomes immersed during one iteration, and emerged during another.
That will keep you busy, B1ck0!

Presumable a hard chine which passes through the surface would be similar. Can you elaborate on what the problems/challenges would be? It's not obvious to me.

[Leo Lazauskas]

Originally Posted by DCockey

Presumable a hard chine which passes through the surface would be similar. Can you elaborate on what the problems/challenges would be? It's not obvious to me.

An over-hanging transom can present problems for non-linear codes because it might, or might not, contribute to the hydrodynamic drag.

For example, in the 1st iteration (usually a linear calculation) the overhang is out of the water and does not contribute. During subsequent iterations, when the free-surface is included, the overhang might be wet, and therefore it contributes. On the next iteration, the free-surface changes, and the overhang might be above the wave at that station.

It is possible that the program goes into an infinite loop or, if the program is stopped at some pre-determined iteration, the hydrodynamic calculations might be in error. I remember Bob Beck of the Uni of Michigan grumbling about this annoying, pathological behaviour.

[DCockey]

Thanks for the explaination. I thought you were refering to problems with mesh generation itself.

[Leo Lazauskas]

Originally Posted by DCockey

Thanks for the explaination. I thought you were refering to problems with mesh generation itself.

No, the method you proposed is good, but there are potential problems with where to cut it at the real waterline.
Of course, there are a host of other difficulties for computer codes that attempt to include the boundary layer.

[Tim B]

Multiphase moving mesh solutions have been done, and are starting to become more viable. It is worth pointing out, of course, that the computing resource required is only really available to big business and research organisations. Generally these are done by deforming the mesh around the hull, but you could potentially achieve the same result by simply moving the whole mesh in space, if boundary conditions allow.

I agree that immersed transoms are a problem, and this requires significant care in the mesh and solution. You should also consider how to apply refinements within your mesh. You will need these, particularly around the transom, where the surface-mesh alone will not be sufficient.

Hope this is of some help,

Tim B.