Discrepancy btw. two different Fluent Versions_convergence problem

ooz · November 1, 2022, 11:53

Hello Everyone,

I am dealing with 2D CFD simulation of a vertical axis wind turbine. I simulated the same problem with the same mesh in two different Fluent versions(2019 and 2021). The convergence criteria (Sca. Res. 1E-5 for all variables) provided by Fluent 2019 at each time step. On the other hand, I got 4E-4 for continuity for the simulations run by Fluent 2021. The convergence criteria were succeded by Fluent-2021, only after an initialization for 10-20 periods by Fluent-2019. I double-checked the configuration and numerical setup which are the same for the simulations run by two versions of the software. There is a mesh interface btw. stationary zone and rotational zone where the turbine adjacent. I used kwSST as a turbulence model.

Mesh quality: ortho: min. 0.37, skew: min 0.57, max. AR ~700

Any ideas??

Nawaf_Ar · November 2, 2022, 06:00

Those "convergence criteria" are honestly useless, and meaningless. At least alone.
You don't wait for your solution to hit that.
You create multiple (at least two) report definitions that you want to measure, the most common one for all is mass flow rate from your inlets/outlets or pressure far-fields.
These two report-definitions alongside your residuals are what tell you if your solution "converged".
Usually wait for continuity to be 10e-3 at least (for some applications it might need to be higher, or lower).
As long as your residuals, and controls (report-definitions) are straight, and not highly turbulent/oscillating heavily, then your solution converged. Regardless of which version tells you what to go for.
Physics, and Maths are the same. The solution should be the same, regardless of what you get in your continuity (scaled residuals) plot since initializations differ from one machine to another, let alone versions.

November 1, 2022, 11:53	Discrepancy btw. two different Fluent Versions_convergence problem	#1
ooz New Member ooz Join Date: May 2022 Posts: 2 Rep Power: 0	Hello Everyone, I am dealing with 2D CFD simulation of a vertical axis wind turbine. I simulated the same problem with the same mesh in two different Fluent versions(2019 and 2021). The convergence criteria (Sca. Res. 1E-5 for all variables) provided by Fluent 2019 at each time step. On the other hand, I got 4E-4 for continuity for the simulations run by Fluent 2021. The convergence criteria were succeded by Fluent-2021, only after an initialization for 10-20 periods by Fluent-2019. I double-checked the configuration and numerical setup which are the same for the simulations run by two versions of the software. There is a mesh interface btw. stationary zone and rotational zone where the turbine adjacent. I used kwSST as a turbulence model. Mesh quality: ortho: min. 0.37, skew: min 0.57, max. AR ~700 Any ideas??

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Problem with Volumetric Reactions in Fluent	Kashif Rashid	Fluent UDF and Scheme Programming	2	December 28, 2017 00:36
Running UDF with Supercomputer	roi247	FLUENT	4	October 15, 2015 13:41
The fluent stopped and errors with "Emergency: received SIGHUP signal"	yuyuxuan	FLUENT	0	December 3, 2013 22:56
fluent parallel problem in win7 x64 system	dunga82	FLUENT	8	April 19, 2012 20:23
fluent convergence problem	josip76	FLUENT	0	May 26, 2011 20:08

November 2, 2022, 06:00		#2
Nawaf_Ar New Member Nawaf Alrajeh Join Date: Aug 2022 Posts: 9 Rep Power: 3	Those "convergence criteria" are honestly useless, and meaningless. At least alone. You don't wait for your solution to hit that. You create multiple (at least two) report definitions that you want to measure, the most common one for all is mass flow rate from your inlets/outlets or pressure far-fields. These two report-definitions alongside your residuals are what tell you if your solution "converged". Usually wait for continuity to be 10e-3 at least (for some applications it might need to be higher, or lower). As long as your residuals, and controls (report-definitions) are straight, and not highly turbulent/oscillating heavily, then your solution converged. Regardless of which version tells you what to go for. Physics, and Maths are the same. The solution should be the same, regardless of what you get in your continuity (scaled residuals) plot since initializations differ from one machine to another, let alone versions.