Density based solver

Techies · September 30, 2022, 04:42

Hi guys,
I'm trying to run steady simulation with density-based solver. But my case diverged instantly after few iteration. I tried running pressure-based first and switch to densiy-based but divegence still occur.
So i want to know why it's hard to converge with density based solver and how to reach convergence with that ?
Thanks.

Roithamer · September 30, 2022, 05:09

You should explain your case physically. Also you need to give details about mesh and specific solver inputs.

Techies · September 30, 2022, 05:13

Hi, my case operates at M=2, just flow over one object. I use icem meshing with hexa element. Min orthogonal 0.05 and it works well with pressure based solver.

Roithamer · September 30, 2022, 06:54

If your case is operating at Mach number 2, you must use compressible solvers. Your given information about case is still not enough but you can try running unsteady compressible solvers with Courant Number lower than 0.4.

Techies · September 30, 2022, 07:55

My case is just pure aerodynamic, it's flow over the aircraft, i use pressure far field for outer domain, non slip and adiabatic wall for object. I calculate Cl, Cd and Cm. With pressure based solver, Cl and Cm are quite resonable but Cm is over predict. I think it's due to the pressure based solver. That's why i need to run density based solver. Is it enough ?

Roithamer · September 30, 2022, 08:30

For physical explanation yes but for numerical convergence you should clarify inputs of solver such as solver type (implicit or explicit), numerical schemes used, relaxation factors and if your analysis solver is explicit formulation time step or courant number. When you face divergence problem in solver, you should try to understand problem is related with physical model or numerical model.

Due to the Ma > 1 condition, your case mostly influenced with shock effects. That's why you should need to run case with density based transient solver with low Courant numbers. It may solve your divergence problem.

Techies · September 30, 2022, 08:59

I use state state with implicit density based, all qualities are discreted with second order scheme. I modify relaxation factor a little bit.
For shockwave, i see many publication used steady state. That's why i just want to use steady state for low cost and time

Roithamer · September 30, 2022, 09:44

Probably steady state analysis' are for weak shock effects. If your geometry is blunt or mesh is coarse you will observe huge shocks near the walls. That's why I recommend you to solve with low time step explicit solver.

Also you can try start running your solver with first order numerical schemes and low under-relaxation factors. After results are settled you can change numerical schemes first order to second order.

At this point, these are the recommendations that come to my mind for solution of your divergence problem.

September 30, 2022, 04:42	Density based solver	#1
Techies Member Mercurial Join Date: Mar 2021 Posts: 72 Rep Power: 5	Hi guys, I'm trying to run steady simulation with density-based solver. But my case diverged instantly after few iteration. I tried running pressure-based first and switch to densiy-based but divegence still occur. So i want to know why it's hard to converge with density based solver and how to reach convergence with that ? Thanks.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Combustion chamber with a sonic fuel injector, density or pressure based solver?	Ryan T	FLUENT	19	February 18, 2021 18:23
Pressure gradient in UDF for density based solver	matzb	FLUENT	0	February 22, 2010 06:34
Regarding Density based solver	Eswar	Main CFD Forum	2	June 6, 2007 11:00
Density based solver	Eswar	Main CFD Forum	2	May 29, 2007 07:29
Pressure based and Density based Solver	Xobile	Siemens	1	November 30, 2004 21:13

September 30, 2022, 05:09		#2
Roithamer New Member Roithamer Join Date: Feb 2021 Posts: 7 Rep Power: 5	You should explain your case physically. Also you need to give details about mesh and specific solver inputs.

September 30, 2022, 05:13		#3
Techies Member Mercurial Join Date: Mar 2021 Posts: 72 Rep Power: 5	Hi, my case operates at M=2, just flow over one object. I use icem meshing with hexa element. Min orthogonal 0.05 and it works well with pressure based solver.

September 30, 2022, 06:54		#4
Roithamer New Member Roithamer Join Date: Feb 2021 Posts: 7 Rep Power: 5	If your case is operating at Mach number 2, you must use compressible solvers. Your given information about case is still not enough but you can try running unsteady compressible solvers with Courant Number lower than 0.4.

September 30, 2022, 07:55		#5
Techies Member Mercurial Join Date: Mar 2021 Posts: 72 Rep Power: 5	My case is just pure aerodynamic, it's flow over the aircraft, i use pressure far field for outer domain, non slip and adiabatic wall for object. I calculate Cl, Cd and Cm. With pressure based solver, Cl and Cm are quite resonable but Cm is over predict. I think it's due to the pressure based solver. That's why i need to run density based solver. Is it enough ?

September 30, 2022, 08:30		#6
Roithamer New Member Roithamer Join Date: Feb 2021 Posts: 7 Rep Power: 5	For physical explanation yes but for numerical convergence you should clarify inputs of solver such as solver type (implicit or explicit), numerical schemes used, relaxation factors and if your analysis solver is explicit formulation time step or courant number. When you face divergence problem in solver, you should try to understand problem is related with physical model or numerical model. Due to the Ma > 1 condition, your case mostly influenced with shock effects. That's why you should need to run case with density based transient solver with low Courant numbers. It may solve your divergence problem.

September 30, 2022, 08:59		#7
Techies Member Mercurial Join Date: Mar 2021 Posts: 72 Rep Power: 5	I use state state with implicit density based, all qualities are discreted with second order scheme. I modify relaxation factor a little bit. For shockwave, i see many publication used steady state. That's why i just want to use steady state for low cost and time

September 30, 2022, 09:44		#8
Roithamer New Member Roithamer Join Date: Feb 2021 Posts: 7 Rep Power: 5	Probably steady state analysis' are for weak shock effects. If your geometry is blunt or mesh is coarse you will observe huge shocks near the walls. That's why I recommend you to solve with low time step explicit solver. Also you can try start running your solver with first order numerical schemes and low under-relaxation factors. After results are settled you can change numerical schemes first order to second order. At this point, these are the recommendations that come to my mind for solution of your divergence problem.