Convergence Study

asda3D · March 24, 2019, 12:07

Hey everyone I am quite new to CFD and I am currently undergoing a project where I am required to carry out a convergence study. I want to make sure that parameters (heat transfer, velocity) are steady within my domain.

I have a questions and that is, does my choice of convergence points in my model have to be located at a node or can it be any point of interest inside the domain?

For example to ensure heat transfer near the wall is steady over several hundred iterations, I can pick my convergence point 5mm away from the wall on a node in my first mesh. Redesigning the mesh to save on computational time, this point no longer is represented by a node. Would this have any impact?

I am a little unsure about this as i believe the whole domain is calculated by interpolation techniques between nodes, so this shouldn't have any impact on my mesh convergence study.

If you have any knowledge you are willing to share much appreciated.

Thank you in advance,
asda3D

FMDenaro · March 24, 2019, 12:42

The convergence to the steady state requires that time derivatives in all the computational points tend to vanish. You need to check this condition (numerical time derivatives< threshold) everywhere in the cells.

asda3D · March 24, 2019, 17:02

Thank you for the reply, I do understand that. Overall convergence is easily identifiable on CFD software, however even if the global parameters have converged there are points of interest where the physics of the flow is very disruptive and may not have converged. This is why points of interest need to be accounted for as well.
My questions is, do these specific monitoring points have to be located on the mesh in order to show convergence or is it OK for the points to lie off the mesh. Would this impact the points convergence?

FMDenaro · March 24, 2019, 17:26

Quote:

Originally Posted by asda3D

Thank you for the reply, I do understand that. Overall convergence is easily identifiable on CFD software, however even if the global parameters have converged there are points of interest where the physics of the flow is very disruptive and may not have converged. This is why points of interest need to be accounted for as well.
My questions is, do these specific monitoring points have to be located on the mesh in order to show convergence or is it OK for the points to lie off the mesh. Would this impact the points convergence?

It makes no sense to use interpolation to monitor points that are not those wherein the equations are numerically solved. If the derivatives in the computational nodes are telling you that the steady state is reached you get the converged solution.

Eifoehn4 · March 25, 2019, 08:36

Quote:

Originally Posted by asda3D

Hey everyone I am quite new to CFD and I am currently undergoing a project where I am required to carry out a convergence study. I want to make sure that parameters (heat transfer, velocity) are steady within my domain.

I have a questions and that is, does my choice of convergence points in my model have to be located at a node or can it be any point of interest inside the domain?

For example to ensure heat transfer near the wall is steady over several hundred iterations, I can pick my convergence point 5mm away from the wall on a node in my first mesh. Redesigning the mesh to save on computational time, this point no longer is represented by a node. Would this have any impact?

I am a little unsure about this as i believe the whole domain is calculated by interpolation techniques between nodes, so this shouldn't have any impact on my mesh convergence study.

If you have any knowledge you are willing to share much appreciated.

Thank you in advance,
asda3D

In my opinion it doesn't matter, if all points of your convergence study on different meshes are collocated. Each sampled point in the computational domain is part of the solution, collocated or not. Therefore it shouldn't have any impact on your study.

LuckyTran · March 25, 2019, 09:18

Convergence is not a binary converged/not-converged decision nor is it an objective decision.

There is the direct way of checking that time derivatives is less than a threshold in every cell. There is also the indirect way of checking other calculated parameters (like interpolated values). Then someone makes the subjective value call or whether it is good enough or not.

March 24, 2019, 12:07	Convergence Study	#1
asda3D New Member anonymous Join Date: Apr 2018 Posts: 26 Rep Power: 8	Hey everyone I am quite new to CFD and I am currently undergoing a project where I am required to carry out a convergence study. I want to make sure that parameters (heat transfer, velocity) are steady within my domain. I have a questions and that is, does my choice of convergence points in my model have to be located at a node or can it be any point of interest inside the domain? For example to ensure heat transfer near the wall is steady over several hundred iterations, I can pick my convergence point 5mm away from the wall on a node in my first mesh. Redesigning the mesh to save on computational time, this point no longer is represented by a node. Would this have any impact? I am a little unsure about this as i believe the whole domain is calculated by interpolation techniques between nodes, so this shouldn't have any impact on my mesh convergence study. If you have any knowledge you are willing to share much appreciated. Thank you in advance, asda3D

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Grid Convergence Index (GCI) for study involving five grids	RobV	Visualization & Post-Processing	0	December 30, 2016 07:22
[ANSYS Meshing] Grid convergence study	AS_Aero	ANSYS Meshing & Geometry	1	June 7, 2013 05:29
[ICEM] Grid Convergence Index study for a hybrid mesh	j_h_86	ANSYS Meshing & Geometry	2	December 6, 2012 12:45
increasing mesh quality is leading to poor convergence	tippo	CFX	2	May 5, 2009 11:55
convergence/ advise for permeability study	CFDnewbie	CFX	0	August 21, 2005 03:50

March 24, 2019, 12:42		#2
FMDenaro Senior Member Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,865 Rep Power: 73	The convergence to the steady state requires that time derivatives in all the computational points tend to vanish. You need to check this condition (numerical time derivatives< threshold) everywhere in the cells.

March 24, 2019, 17:02		#3
asda3D New Member anonymous Join Date: Apr 2018 Posts: 26 Rep Power: 8	Thank you for the reply, I do understand that. Overall convergence is easily identifiable on CFD software, however even if the global parameters have converged there are points of interest where the physics of the flow is very disruptive and may not have converged. This is why points of interest need to be accounted for as well. My questions is, do these specific monitoring points have to be located on the mesh in order to show convergence or is it OK for the points to lie off the mesh. Would this impact the points convergence?

March 25, 2019, 09:18		#6
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,750 Rep Power: 66	Convergence is not a binary converged/not-converged decision nor is it an objective decision. There is the direct way of checking that time derivatives is less than a threshold in every cell. There is also the indirect way of checking other calculated parameters (like interpolated values). Then someone makes the subjective value call or whether it is good enough or not.