Diverged caused by MUSCL in turb equations

ShiheJia · March 17, 2022, 08:34

hi,

I am using SU2 for quite a long time and in my experience, I find the case is easily to diverge when the muscl interpolation in turb equation is used.

It is rare in 2-D cases and common in 3-D case, no matter how much the CFL number is, whether the turb model is SA or SST, which MUSCL reconstruction method is. Even if the case can meet convergence criteria with MUSCL_TURB= NO

So I wonder if I miss sth. important for the robustness. The below options are what I usually set.

NUM_METHOD_GRAD= GREEN_GAUSS
NUM_METHOD_GRAD_RECON = GREEN_GAUSS
CFL_NUMBER= 5
MUSCL_FLOW= YES
SLOPE_LIMITER_FLOW= VENKATAKRISHNAN
MUSCL_TURB= YES
SLOPE_LIMITER_TURB= VENKATAKRISHNAN

Any help would be appreciated.

pcg · March 17, 2022, 14:48

I think VENKATAKRISHNAN_WANG would be a better choice because the range of the turbulence variables is very different from the flow variables.
The typical value of the VENKAT_LIMITER_COEFF (0.05) basically means that the SA model would be unlimited (since nu-tilde is of the order of a viscosity).
You can also try using SA_NEG instead of SA.

For SST I think it is going to be harder... Omega has very large gradients close to walls.

ShiheJia · March 18, 2022, 00:57

Thanks for your help.

I have tried to lower the value of k and omega (It seems k is more easily to get a negative value using muscl than omega in my current case) in SST with using REF_DIMENSIONALIZATION= FREESTREAM_PRESS_EQ_ONE option, and it runs well with BARTH_JESPERSEN limiter in a 3-D case.

However, it is not easy to judge the convergence by the performance of RMS when the non_dim turns on. So I wonder if the non_dim of k and omega can be apart from the euler equation.

Also, I notice that the non-dim method of k/omega/mu_l/mu_t in su2 is as below:
mu_l_ref = rho_ref * vel_ref * l_ref
mu_t_ref = mu_l_ref
k_ref = vel_ref^2
omega_ref = k_ref * rho_ref /mu_t_ref

As a matter of fact, CFL3D shows as below:
mu_l_ref = mu_l_farfield
mu_t_ref = mu_l_ref
k_ref = vel_ref^2
omega_ref = Re * vel_ref/ l_ref = k_ref * rho_ref /mu_t_ref

The latter method could get a lower omega and k nearwall by lower mu_l_ref and may be better. Actually I am thinking to do this job but I am not fully aware of the influence to other part of codes of this change. Shall I do this change?

March 17, 2022, 08:34	Diverged caused by MUSCL in turb equations	#1
ShiheJia New Member Join Date: Feb 2020 Posts: 6 Rep Power: 7	hi, I am using SU2 for quite a long time and in my experience, I find the case is easily to diverge when the muscl interpolation in turb equation is used. It is rare in 2-D cases and common in 3-D case, no matter how much the CFL number is, whether the turb model is SA or SST, which MUSCL reconstruction method is. Even if the case can meet convergence criteria with MUSCL_TURB= NO So I wonder if I miss sth. important for the robustness. The below options are what I usually set. NUM_METHOD_GRAD= GREEN_GAUSS NUM_METHOD_GRAD_RECON = GREEN_GAUSS CFL_NUMBER= 5 MUSCL_FLOW= YES SLOPE_LIMITER_FLOW= VENKATAKRISHNAN MUSCL_TURB= YES SLOPE_LIMITER_TURB= VENKATAKRISHNAN Any help would be appreciated.

March 17, 2022, 14:48		#2
pcg Senior Member Pedro Gomes Join Date: Dec 2017 Posts: 467 Rep Power: 14	I think VENKATAKRISHNAN_WANG would be a better choice because the range of the turbulence variables is very different from the flow variables. The typical value of the VENKAT_LIMITER_COEFF (0.05) basically means that the SA model would be unlimited (since nu-tilde is of the order of a viscosity). You can also try using SA_NEG instead of SA. For SST I think it is going to be harder... Omega has very large gradients close to walls. ugurtan666 likes this.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Guide: Writing Equations in LaTeX on the CFD Online Forums	pete	Site Help, Feedback & Discussions	27	May 19, 2022 04:19
Calculation of the Governing Equations	Mihail	CFX	7	September 7, 2014 07:27
Riemann invariants of adjoint equations of shallow water equations	zqb0929	Main CFD Forum	0	March 15, 2012 01:54
CFD governing equations	m.gos	Main CFD Forum	0	April 30, 2011 15:21
? fluctuating equations for homogenous shear turb.	ff_fan	Main CFD Forum	1	September 20, 2002 08:39

March 18, 2022, 00:57		#3
ShiheJia New Member Join Date: Feb 2020 Posts: 6 Rep Power: 7	Thanks for your help. I have tried to lower the value of k and omega (It seems k is more easily to get a negative value using muscl than omega in my current case) in SST with using REF_DIMENSIONALIZATION= FREESTREAM_PRESS_EQ_ONE option, and it runs well with BARTH_JESPERSEN limiter in a 3-D case. However, it is not easy to judge the convergence by the performance of RMS when the non_dim turns on. So I wonder if the non_dim of k and omega can be apart from the euler equation. Also, I notice that the non-dim method of k/omega/mu_l/mu_t in su2 is as below: mu_l_ref = rho_ref * vel_ref * l_ref mu_t_ref = mu_l_ref k_ref = vel_ref^2 omega_ref = k_ref * rho_ref /mu_t_ref As a matter of fact, CFL3D shows as below: mu_l_ref = mu_l_farfield mu_t_ref = mu_l_ref k_ref = vel_ref^2 omega_ref = Re * vel_ref/ l_ref = k_ref * rho_ref /mu_t_ref The latter method could get a lower omega and k nearwall by lower mu_l_ref and may be better. Actually I am thinking to do this job but I am not fully aware of the influence to other part of codes of this change. Shall I do this change?