[Bartosz]

I have come because I have been spending three weeks trying to work out a mesh issue.

I am modelling a jet engine outlet for a thesis. I have a strange issue with my mesh. I have a 100% mesh and then I have made a 20% mesh from this same one. When I run the 20% it converges, however, if I run a more fine mesh it crashes.
I am unable to get the higher percentages to converge though.
I am using the SpalartAllmaras turbulence modelling with an outlet velocity of 200 m/s.

I have a model where there is the
- Ground (modelled as a wall)
- an outlet (BC fixed value 200 m/s for the jet engine)
- The rest of the walls are all atmosphere and their BC are pressureInletOutletVelocity and totalPressure as in the tutorials. I found these BC to be very stable at a smaller grid up to 300 m/s.
- nut and nuTilda BC is freestream 0.125 for both.

However, I need to model the jet blast profile of a jet up to 100 m behind the engine outlet. I found at the outlet circled in the diagram, this issue is always occuring. In the diagram is a contour of nuTilda at 0.4, however, this sudden drop is always being caused. I am unable to solve this issue and have come to these forums.

I have tried everything from clustering that wall (which makes it unstable and crash), changing the turbulence, changing the relaxation factors, making the ceiling much higher, and even making the top and sides walls. But this is always occuring. Does anyone have any ideas why this is occuring? Any help would be greatly appreciated!

Regards,
Bartosz
nuTila 0.4 Walls 2.jpg


nuTila 0.4 with 100 m ceiling.jpg
nuTila 0.4 with 100 m ceiling.jpg

[piu58]

Spalart-Allmaras tends to introduce too much turbulence in the flow and tend therefore to crashing. If you use a finer mesh you have to lower the time step / maxCo at the same expanse. That means you should at least start with a 5 time lower time step.

I recommend experimenting with dynamicKEqn, which is more stable.

[Bartosz]

I forgot to mention I am using simpleFoam for solving this, which I was under the impression the timestep is always 1 and it just needs to do more iterations.

[piu58]

You may set the time step and the time to finish in controlDict. If you want to have ore than one time step before changing anything, do it here. Probably you have to modify controlDict in your script too.

[Bartosz]

What do you mean having one more time step before changing it? As I talked to several people and they said for simpleFoam the time step is always 1 and then you change the number of iterations. Rather than changing the value of time step, you change the relaxation factors in the fvSolutions files

[piu58]

If you need considering a changing inlet velocity you cannot use a steady state solver. Your case has no steady state solution.

[Bartosz]

Ahhhh I have a single inlet velocity of 200 m/s which is is uniform on the boundary condition and does not change.

[piu58]

In simpleFoam you have to play with the relaxation factor. This factor allows to take only a fraction of the change to the next iteration step.