Unsteady Implicit / Piso

FluentStarter · August 7, 2017, 04:15

Hey you,

I just wanted to start a small discussion about the algorithms available for unsteady calculations in Star. Maybe someone is interested in this topic as well and we can share some experiences.
I work on LES simulations and as the combustion side of Star recommends "Piso" and the acoustics side "Unsteady Implicit" I made some generic tests.

During those Piso showed a strong dissipative behavior but was way faster.

It is easy to find some information about Piso but I'm not able to find anything about Unsteady Implicit. Support told me that it is based on Simple? But wasn't Simple for steady flows? Do they mean Pimple? Does someone have informations to this? I really would like to know why UI is less dissipative in my simulations...

me3840 · August 8, 2017, 10:18

"pimple" is the OpenFOAM jargon for the transient simple algorithm - yes, it exists. Pretty much the same thing.

What was the timestep difference between the runs? How did you ensure the simple-scheme's timesteps converged? Those are likely the source of the differences.

FluentStarter · August 9, 2017, 06:09

-The timestep was exactly the same.

- For UI I monitored values like velocity and pressure over the inner iterations and defined maximum iterations according to the findings I took from the observations (8 inner it)

- For PISO I used the default settings at first and afterwards played around with the residual reduction and the maximum corrector steps. Improvements were marginal thereby.

- Computations times were approximately 1.5 times higher with UI

Especially the convection of vortices and waves in an inviscid flow were totally different for the two algorithms.

me3840 · August 10, 2017, 19:04

For a sufficiently small timestep they should give pretty similar answers. PISO is usually faster than SIMPLE for small timesteps as it doesn't have all the inner iterations.

What was the maximum CFL in both cases? For PISO it's not really wise to use anything above 5 or so, but 1 is best. SIMPLE doesn't have any timestep limitations really except the timestep chosen can affect the answer.

FluentStarter · August 11, 2017, 05:36

That's what I was suspecting as well. However I saw different results.
But I found the reason therefore:

In StarCCM+ the only temporal discretization is first order BWD for the PISO algorithm. For Unsteady Implicit I used second order. Using 1st order BWD with UI resulted in the same dissipation.

CFL numbers were below 0.7.

August 7, 2017, 04:15	Unsteady Implicit / Piso	#1
FluentStarter Member Join Date: Apr 2014 Location: N/A Posts: 47 Rep Power: 12	Hey you, I just wanted to start a small discussion about the algorithms available for unsteady calculations in Star. Maybe someone is interested in this topic as well and we can share some experiences. I work on LES simulations and as the combustion side of Star recommends "Piso" and the acoustics side "Unsteady Implicit" I made some generic tests. During those Piso showed a strong dissipative behavior but was way faster. It is easy to find some information about Piso but I'm not able to find anything about Unsteady Implicit. Support told me that it is based on Simple? But wasn't Simple for steady flows? Do they mean Pimple? Does someone have informations to this? I really would like to know why UI is less dissipative in my simulations... Last edited by FluentStarter; August 7, 2017 at 05:22.

August 11, 2017, 05:36		#5
FluentStarter Member Join Date: Apr 2014 Location: N/A Posts: 47 Rep Power: 12	That's what I was suspecting as well. However I saw different results. But I found the reason therefore: In StarCCM+ the only temporal discretization is first order BWD for the PISO algorithm. For Unsteady Implicit I used second order. Using 1st order BWD with UI resulted in the same dissipation. CFL numbers were below 0.7. PhilippCFD likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Current number and time step for an implicit unsteady simulation	Clémence	STAR-CCM+	2	September 30, 2016 09:30
Question about Implicit Unsteady Solver and Maximum Inner Iterations	Awesomo	STAR-CCM+	4	June 28, 2016 08:57
Implicit Unsteady Tending to Steady Solution	Muzz	STAR-CCM+	2	September 21, 2015 11:56
Problem with implicit unsteady solver	CCMuser	STAR-CCM+	2	March 3, 2010 11:20
why implicit isn't suited for unsteady case?	ravi	Main CFD Forum	4	April 16, 2007 13:43

August 8, 2017, 10:18		#2
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 24	"pimple" is the OpenFOAM jargon for the transient simple algorithm - yes, it exists. Pretty much the same thing. What was the timestep difference between the runs? How did you ensure the simple-scheme's timesteps converged? Those are likely the source of the differences.

August 9, 2017, 06:09		#3
FluentStarter Member Join Date: Apr 2014 Location: N/A Posts: 47 Rep Power: 12	-The timestep was exactly the same. - For UI I monitored values like velocity and pressure over the inner iterations and defined maximum iterations according to the findings I took from the observations (8 inner it) - For PISO I used the default settings at first and afterwards played around with the residual reduction and the maximum corrector steps. Improvements were marginal thereby. - Computations times were approximately 1.5 times higher with UI Especially the convection of vortices and waves in an inviscid flow were totally different for the two algorithms.

August 10, 2017, 19:04		#4
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 24	For a sufficiently small timestep they should give pretty similar answers. PISO is usually faster than SIMPLE for small timesteps as it doesn't have all the inner iterations. What was the maximum CFL in both cases? For PISO it's not really wise to use anything above 5 or so, but 1 is best. SIMPLE doesn't have any timestep limitations really except the timestep chosen can affect the answer.