Upwind scheme - convergence problem

MissCFD · March 31, 2015, 23:47

Hello,

I simulate a passage of stator and a passage of rotor for a hydraulic turbine so I use water. I use a mass flow rate condition for inlet and pressure condition for outlet.

I don't understand why when I use an upwind advection scheme I have a wall which take place in outlet and my calculation differ, crash. This doesn't happen when I use a high resolution or blend factor = 1 for advection scheme.

With an upwind advection scheme, I introduce numerical diffusion so my calculation should converge more easily, isn't it ?

What can I do to avoid that a wall which take place in outlet of my domain ? Reduce the time scale ?

Thank you for your help

Thomas MADELEINE · April 1, 2015, 05:23

For what I understood of the advection scheme :
upwind => blend factor of 0 => first order
HR => blend factor vary between 0.5 and 1 (normally close to 0.9) => relaxation term between first order and second order
blend factor of 1 => second order

The equations will be modeled at each integration by a first/second order function.
first order functions are easier to calculate and converge faster but they are less precise and can't bend so much.
So if you have big gradient the first order will be unable to converge and your simulation will crash.

solution :
stay at an order bigger than 1 (maybe 1.5 => blend factor at 0.5) or refine your mesh.

If I said something wrong please correct me, it is what I understood but I am not 100% sure i get it all.

March 31, 2015, 23:47	Upwind scheme - convergence problem	#1
MissCFD Senior Member Join Date: Aug 2014 Posts: 160 Rep Power: 12	Hello, I simulate a passage of stator and a passage of rotor for a hydraulic turbine so I use water. I use a mass flow rate condition for inlet and pressure condition for outlet. I don't understand why when I use an upwind advection scheme I have a wall which take place in outlet and my calculation differ, crash. This doesn't happen when I use a high resolution or blend factor = 1 for advection scheme. With an upwind advection scheme, I introduce numerical diffusion so my calculation should converge more easily, isn't it ? What can I do to avoid that a wall which take place in outlet of my domain ? Reduce the time scale ? Thank you for your help

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence problem in Fluent for quenching process	kaeran	FLUENT	4	December 1, 2014 03:14
Accuracy problem of HO schemes on unstructured mesh, HO scheme gives 1st order result	gemini	Main CFD Forum	12	December 27, 2011 22:01
Force can not converge	colopolo	CFX	13	October 4, 2011 23:03
convergence problem	limseokmin	FLUENT	3	November 14, 2004 13:43
Help!explicit multistage scheme with upwind scheme	D .T.	Main CFD Forum	1	June 4, 2003 04:42

April 1, 2015, 05:23		#2
Thomas MADELEINE Senior Member Thomas MADELEINE Join Date: Oct 2014 Posts: 126 Rep Power: 12	For what I understood of the advection scheme : upwind => blend factor of 0 => first order HR => blend factor vary between 0.5 and 1 (normally close to 0.9) => relaxation term between first order and second order blend factor of 1 => second order The equations will be modeled at each integration by a first/second order function. first order functions are easier to calculate and converge faster but they are less precise and can't bend so much. So if you have big gradient the first order will be unable to converge and your simulation will crash. solution : stay at an order bigger than 1 (maybe 1.5 => blend factor at 0.5) or refine your mesh. If I said something wrong please correct me, it is what I understood but I am not 100% sure i get it all.