[liang11]

Hello, everyone!

I have another question. I have run one case twice with different maximum Mach CFL. One corresponds to a maximum Mach CFL of 50, the other is 3.
However, the maximum in-cylinder pressure and combustion phase are totally different, just as the picture in the attachment. The change of TKE of this two calacualtions are similar.
How to keep the mean in-cylinder presure unchange?

Thank you very much!

[MCoil]

Hi -
It seems like you might have some cycle-to-cycle variation. Could you run 3-4 cycles of this to see what the results look like?

[MCoil]

I want to add a comment. When you are using the cfl_mach at 3, you are taking smaller timesteps, which is more accurate. On the cfl_mach=50, you may not be getting convergence every timestep. Take a look at time.out for both cases and see what you can learn from that.

[liang11]

Originally Posted by MCoil

I want to add a comment. When you are using the cfl_mach at 3, you are taking smaller timesteps, which is more accurate. On the cfl_mach=50, you may not be getting convergence every timestep. Take a look at time.out for both cases and see what you can learn from that.

Hi

Thanks for your help.
The multi-cycle simulation is quite expensive, it might need a lot of time to check cycle-to-cycle variation.

Comparing to the Maximum Mach CFL=50, there are much fewer recoveries with maximum CFL at 3. Although CFL=50 causes a lot of recoveries, the real time step is still larger than the minimum time step (1e-08). I think that means convergence.

My work focus on knock simulation. I try to compare pressure oscillation under different maximum mach CFL. However, decreasing mach CFL from 50 to 3 also lowers down the mean in-cylinder pressure, which generating a similar effect of regarding the spark advance.
So, it hard for me to do that.

Thank you again for your help. I really appreciate that.

[MCoil]

Hi -


Just because your timestep is not decreasing does not mean that the solution is converging at each timestep. It merely means you're not triggering one of the timestep limiters.



You should view the simulation from the max_cfl_mach case = 3 as more correct than the max_cfl_mach = 50 case, since it's using a smaller timestep.



If you want to look at knock, you'll need the lower timstep, and you'll need to look at more than one cycle. This will be a little expensive, but you can't resolve the knock with the longer timestep.



There are ways to perturb simulations to estimate the cycle to cycle variation by running several simulations at once instead of running a single simulation for several cycles. If this option appeals to you, I can explain more.


Thanks,

[liang11]

Hi

How to judge whether the solution is converging or not?
With longer timestep, there are also many pressre-oscillations being captured.

can you please explain more of cycle-cycle variation?

Thanks!

Originally Posted by MCoil

Hi -


Just because your timestep is not decreasing does not mean that the solution is converging at each timestep. It merely means you're not triggering one of the timestep limiters.



You should view the simulation from the max_cfl_mach case = 3 as more correct than the max_cfl_mach = 50 case, since it's using a smaller timestep.



If you want to look at knock, you'll need the lower timstep, and you'll need to look at more than one cycle. This will be a little expensive, but you can't resolve the knock with the longer timestep.



There are ways to perturb simulations to estimate the cycle to cycle variation by running several simulations at once instead of running a single simulation for several cycles. If this option appeals to you, I can explain more.


Thanks,

[MCoil]

Hi -


You should look in your log file and see if the various transport equations are converging. Set the screen print level in inputs.in to 3 to enable printing of more information.



For information on cycle-to-cycle variation, you might consult our training materials on internal combustion engines. If you register on our downloads page, you will have access to them. In addition, you might look at the paper below. This is an LES simulation, but I think it will give some background on cycle-to-cycle variation and also talk about estimating that variation by running concurrent cycles instead of a series of cycles.



Daniel M. Probst,Sameera Wijeyakulasuriya,Eric Pomraning,Janardhan Kodavasal,Riccardo Scarcelli,Sibendu Som, "Predicting Cycle-to-Cycle Variation with Concurrent Cycles in a Gasoline Direct Injected Engine with Large Eddy Simulations", J. Energy Resour. Technol. Apr 2020, 142(4): 042202



I hope this gets you started.

[liang11]

Originally Posted by MCoil

Hi -


You should look in your log file and see if the various transport equations are converging. Set the screen print level in inputs.in to 3 to enable printing of more information.



For information on cycle-to-cycle variation, you might consult our training materials on internal combustion engines. If you register on our downloads page, you will have access to them. In addition, you might look at the paper below. This is an LES simulation, but I think it will give some background on cycle-to-cycle variation and also talk about estimating that variation by running concurrent cycles instead of a series of cycles.



Daniel M. Probst,Sameera Wijeyakulasuriya,Eric Pomraning,Janardhan Kodavasal,Riccardo Scarcelli,Sibendu Som, "Predicting Cycle-to-Cycle Variation with Concurrent Cycles in a Gasoline Direct Injected Engine with Large Eddy Simulations", J. Energy Resour. Technol. Apr 2020, 142(4): 042202



I hope this gets you started.

Thanks again. I will try it.

[liang11]

Originally Posted by MCoil

Hi -


You should look in your log file and see if the various transport equations are converging. Set the screen print level in inputs.in to 3 to enable printing of more information.



For information on cycle-to-cycle variation, you might consult our training materials on internal combustion engines. If you register on our downloads page, you will have access to them. In addition, you might look at the paper below. This is an LES simulation, but I think it will give some background on cycle-to-cycle variation and also talk about estimating that variation by running concurrent cycles instead of a series of cycles.



Daniel M. Probst,Sameera Wijeyakulasuriya,Eric Pomraning,Janardhan Kodavasal,Riccardo Scarcelli,Sibendu Som, "Predicting Cycle-to-Cycle Variation with Concurrent Cycles in a Gasoline Direct Injected Engine with Large Eddy Simulations", J. Energy Resour. Technol. Apr 2020, 142(4): 042202



I hope this gets you started.

Dear Millicent Coil,

I am focusing on spark ignition combustion engine and the related knock, using RANS. If RANS is replaced by LES, how to do that? such as grid control and turbulence modeling.
Is there any LES example?

Thanks!

[MCoil]

Hi -

If you want to run using LES, you'll have to resolve the flame thickness in the combustion zone, and that will be quite expensive. What do you hope to achieve with an LES simulation?

We don't currently have a published engine example using LES, because it is so expensive.

[liang11]

Hi,
I want to know the different simulation results of knock, using RANS or LES.
And I know LES is expensive, but I don't know how expensive it is?

Also, if other diluents, like argon, water or CO2, are added to the intake, how can I change the gas transport data and global transport parameters?

Thanks!

[MCoil]

LES...it's pretty expensive. You could always send your case in and we could help you gauge the cell count.

When you are setting up your case in CONVERGE Studio, go to Materials > Gas simulation. That brings you to a panel that has a place where you can change your gas.dat to reflect the transport properties of your mixture. Then, go to Materials > global transport parameters, and you may change your turbulent Schmidt and Prandtl numbers there.

Best,
Millicent