[Anna Tian]

We suppose that we have two grids with different grids topology for the same geometry. These two grids match the requirements listed as below.

1. Converge to the same and very small residual value, e.g. 10^(-6), both globally and locally. The resulting factor we care about won’t change with more iterations anymore.

2. Both of them are tested to have very good grids refinement independence.

3. They all have very good grids qualities value, e.g. skewness, aspect ratio, grids expansion rate.

4. The solver and boundary settings are all the same.

5. Both of the grids are structured grids.

I’m wondering, is it possible that they still give very different CFD simulation results? Does grids topology really matters to the results if the simulation matches all the theoretical requirements?

[mnabi]

Originally Posted by Anna Tian

We suppose that we have two grids with different grids topology for the same geometry. These two grids match the requirements listed as below.

1. Converge to the same and very small residual value, e.g. 10^(-6), both globally and locally. The resulting factor we care about won’t change with more iterations anymore.

2. Both of them are tested to have very good grids refinement independence.

3. They all have very good grids qualities value, e.g. skewness, aspect ratio, grids expansion rate.

4. The solver and boundary settings are all the same.

I’m wondering, is it possible that they still give very different CFD simulation results? Does grids topology really matters to the results if the simulation matches all the theoretical requirements?

As long as the results are tested to be grid independent, they must have no big difference.

[Anna Tian]

Originally Posted by mnabi

As long as the results are tested to be grid independent, they must have no big difference.

Grids independence includes grids refinements independence and grids topology independence. Grids refinements independence is one of requirements as listed above. My questions is about grids topology independence.

[mnabi]

Originally Posted by Anna Tian

Grids independence includes grids refinements independence and grids topology independence. Grids refinements independence is one of requirements as listed above. My questions is about grids topology independence.

The grid topology doesn't affect the results so much as long as you have grid independence solutions.

[Anna Tian]

Originally Posted by mnabi

The grid topology doesn't affect the results so much as long as you have grid independence solutions.

What do you mean by your 'grid independence' here?

[mnabi]

Originally Posted by Anna Tian

What do you mean by your 'grid independence' here?

I mean, your grid is fine enough, and refining the grid does not affect your solution.

In the coarse grid, the topology of grid may have big effect. For example, if you solve on two different grid, let us say, one Cartesian, the other curvilinear, the solution is dependent of several factors, such as the cell size, the location of nodes etc., But as your grids are fine enough, and the solution is not affected by the grid size, you have to get similar solutions.

[Ford Prefect]

Originally Posted by mnabi

I mean, your grid is fine enough, and refining the grid does not affect your solution.

In the coarse grid, the topology of grid may have big effect. For example, if you solve on two different grid, let us say, one Cartesian, the other curvilinear, the solution is dependent of several factors, such as the cell size, the location of nodes etc., But as your grids are fine enough, and the solution is not affected by the grid size, you have to get similar solutions.

This is an interesting point. What about the boundary topology in the case of the Cartesian mesh? For example; if you have a circle then the refined curvilinear mesh would approach a curve length of (2*r*pi), but the Cartesian mesh would stay at a curve length of (4*r) regardless of refinement.

[mnabi]

Originally Posted by Ford Prefect

This is an interesting point. What about the boundary topology in the case of the Cartesian mesh? For example; if you have a circle then the refined curvilinear mesh would approach a curve length of (2*r*pi), but the Cartesian mesh would stay at a curve length of (4*r) regardless of refinement.

I didn't get your point. Do you mean, the Cartesian grid doesn't fit the complex geometries? In this case, you need to force the boundaries by immersed boundary methods, either cut-cell or ghost-cell.

[Ford Prefect]

Originally Posted by mnabi

I didn't get your point. Do you mean, the Cartesian grid doesn't fit the complex geometries?

Yes. Perhaps it doen't matter either, if the cells are small enough and we are careful when calculating surface fluxes?

Originally Posted by mnabi

In this case, you need to force the boundaries by immersed boundary methods, either cut-cell or ghost-cell.

Fair enough, and if properly implemented then I fully agree.

[gafoorcp]

hii..
i am doing diesel engine closed simulation by using AVL FIRE. For the grid independency test purpose,i have taken different average cell size with different number of cells for the same problem. after running some simulation with different cell size. i am getting similar results for some properties(such as pressure, temperature etc..) bt for the same case am getting different values for some other properties.
so please suggest me the important aspects and care we should have to take before a grid independence test in AVL fire ESE diesel simulator.
thanking you all
Abdul gafoor


please give me some suggestions to perform a grid independencey test

[leflix]

Originally Posted by Anna Tian

I’m wondering, is it possible that they still give very different CFD simulation results? Does grids topology really matters to the results if the simulation matches all the theoretical requirements?

The result depends on the grid quality. If your solver is not bugged you should find exactly the same results for equivalent grids.

[gafoorcp]

Originally Posted by leflix

The result depends on the grid quality. If your solver is not bugged you should find exactly the same results for equivalent grids.

sir...
after completing some run with different number of cells....am getting a matching result within a range of number of cells....and a different result after that range.....so can we take the range at which we r getting same solution for our actual testing....is that a valid method......
please reply....
regards
gafoor