Problem of convergence- simulation of flow wind around cylinder

activo · October 19, 2015, 12:51

Hi all,

I want realise a 2D simulation of a turbulent flow of a fluide, wind for example around a cylinder.
So the wind flow is turbulent. The turbulence model chosen (that i think is the best), is the Shear Stress Transport (SST) ; Reynolds-averaged NaviereStokes (RANS) model.
* the dimensions are :
- The solution domain => L=90 m ; H= 30 m
- the Cylindre has a diameter of D= 10 m ; and it is situated in midle of the domain
The problem is that I have no convergence amthough I give a big number of iterations
Is there a simulation that can never converge and that we can trust the resuts given by it ?

Thanks.

RodriguezFatz · October 20, 2015, 10:02

From what you write, most probably you have a problem without a steady state solution. Geomtries like this produce large vorticies, google for "von karman vortex street". That's why you don't get a converged solution with a steady state solver.

Of course there might be 1000 other possible solutions but you need to post some more information, such as a picture of the residuals, the latest flow (velocity) field and a picture of the grid.

highorder_cfd · October 20, 2015, 10:45

It is possible that is unsteady, but for flow past a cylinder there are few unsteady cases based on the flow regime, thus you should also give details of the fluid and boundary conditions.

activo · October 21, 2015, 20:42

Thanks guys for replying,

In fact my main objective is that simulate a geometry partially cylindrical not a right cylinder (see pics below) I just begin by the cylinder the geometry that I thinked easy to simulate.
I remove actually to may main gemetry and as always no convergence !

RodriguezFatz I dont think is the problem of unsteady state you can see it in the velocity contour (pic below )
also I post a residual picture and a meshing picture.
- the fluide is the Air, the initial velocity imposed is of 17 m/s.
- the left boundary is the inlet .. the right and the high is defined as Pressure-outlet

thanks !

RodriguezFatz · October 22, 2015, 02:20

I think you can see from velocity contour that it is unsteady (most probably). It even looks like a beginning von Karman vortex street. Just the solver is not time-dependent so it tries to damp the vorticies as much as it can but it doesn't succeed. That's why your residuals don't decrease.

activo · October 22, 2015, 23:08

Hi, Thanks very much

So what can I do to fix this ? what's exactly the parameters that can be behind this kind of problem ?

I just checked also the effect of meshing on the convergence by using different fomrs of meshing but without any good results !

Far · October 23, 2015, 02:57

use transient simulation.

You can use strouhal number for the cylinder for equivalent Reynolds number and guess the time step. Then half the time step and see the effects.

compare results for different time steps and when they stop to change, that is your required time step.

Now you use these settings to run your required simulation and get data for analysis.

October 19, 2015, 12:51	Problem of convergence- simulation of flow wind around cylinder	#1
activo New Member Badr badro Join Date: Oct 2015 Posts: 6 Rep Power: 11	Hi all, I want realise a 2D simulation of a turbulent flow of a fluide, wind for example around a cylinder. So the wind flow is turbulent. The turbulence model chosen (that i think is the best), is the Shear Stress Transport (SST) ; Reynolds-averaged NaviereStokes (RANS) model. * the dimensions are : - The solution domain => L=90 m ; H= 30 m - the Cylindre has a diameter of D= 10 m ; and it is situated in midle of the domain The problem is that I have no convergence amthough I give a big number of iterations Is there a simulation that can never converge and that we can trust the resuts given by it ? Thanks.

October 20, 2015, 10:02		#2
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	From what you write, most probably you have a problem without a steady state solution. Geomtries like this produce large vorticies, google for "von karman vortex street". That's why you don't get a converged solution with a steady state solver. Of course there might be 1000 other possible solutions but you need to post some more information, such as a picture of the residuals, the latest flow (velocity) field and a picture of the grid. __________________ The skeleton ran out of shampoo in the shower.

October 22, 2015, 02:20		#5
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	I think you can see from velocity contour that it is unsteady (most probably). It even looks like a beginning von Karman vortex street. Just the solver is not time-dependent so it tries to damp the vorticies as much as it can but it doesn't succeed. That's why your residuals don't decrease. __________________ The skeleton ran out of shampoo in the shower.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Wind Tunnel Flow Simulation Mass Flow BC Issue	ledelman	SU2	4	August 3, 2014 19:38
Convergence Problem in Axisymmetric Periodic Flow	atheresia	FLUENT	3	February 10, 2014 04:00
3D Compressible Nozzle Flow Convergence Problem	mep10jl	FLUENT	2	July 30, 2013 18:09
Low pressure de Laval simulation convergence problem	heksel8i	FLUENT	3	July 22, 2013 11:28
DPM simulation convergence problem	gemini	FLUENT	2	May 22, 2012 02:41

October 20, 2015, 10:45		#3
highorder_cfd Member Join Date: Jan 2015 Posts: 63 Rep Power: 11	It is possible that is unsteady, but for flow past a cylinder there are few unsteady cases based on the flow regime, thus you should also give details of the fluid and boundary conditions.

October 22, 2015, 23:08		#6
activo New Member Badr badro Join Date: Oct 2015 Posts: 6 Rep Power: 11	Hi, Thanks very much So what can I do to fix this ? what's exactly the parameters that can be behind this kind of problem ? I just checked also the effect of meshing on the convergence by using different fomrs of meshing but without any good results !

October 23, 2015, 02:57		#7
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	use transient simulation. You can use strouhal number for the cylinder for equivalent Reynolds number and guess the time step. Then half the time step and see the effects. compare results for different time steps and when they stop to change, that is your required time step. Now you use these settings to run your required simulation and get data for analysis.