Problem: Convergence in unsteady EBU combustion model

Defend · April 21, 2014, 06:33

Hello!

For monthes I've tried to solve the problem with the combustion and then decided to ask for a help on this forum =)

I used non-premixed Standart EBU combustion model for chemical reaction C2H5OH+3O2=2CO2+3H2O to verify the experimental data. In experiment I have a burning liquid C2H5OH in closed volume room. To exclude multiphasing I used wall with normal mass flow of gas C2H5OH from its surface (not mass flow rate on inlet). Also I added small pressure outlet bondary condition.

I tried to use different turbulent models, time-steps, Courant number, other solver parameters and all the time I have the same problem: residuals of CO2, H2O at every time-step becomes flat without any value decreasing. First time at value 10e-4, then 10e-2... and after many time steps at value 10e+7... I tried to increase the number of cells - I made 100000 polyhedral cells for simple 2d geometry. Also I found, that Green-Gauss gradient method instead of Hybrid Gauss-LSQ gives me better convergence. But again and again I have nondecreasing value of CO2 and H2O residual 10e+7.
It would be the great luck if anyone of you could explain me, how to get the residual convergence 10e-5 at any time-step of such task.

Thank you in advance.

me3840 · April 21, 2014, 14:01

Are those the normalized or unnormalized residuals?

Defend · April 21, 2014, 14:07

what does it mean - normalize/unnormalize residual?..

me3840 · April 21, 2014, 14:49

The residuals are normalized by default to the average value of the given residual over the first few iterations in order to group them closer together. You can disable this and view their true values by turning the normalization option to off in the monitors folder. If the values you're looking at are very low, you may be hitting a precision limit or just seeing noise.

Defend · April 21, 2014, 15:39

Hmm... Ok, thank you, I'll try tomorrow!

Defend · April 22, 2014, 05:02

me3840, thank you! I had normalized values of residuals. I changed them to unnormalized and the true value is 1e-8!..

Maybe you could answer to my question about coefficient A of EBU combustion model:

Everywhere in literature I see, that this coefficient for modeling must be A=4. And in my task to reach the temperature values that accords to experimental data this coeff must be A=0.1. Is it normal to change this coeff or I have some mistakes in solving?..
Thank you in advance.

me3840 · April 22, 2014, 11:25

I don't really know anything about the EBU models to say, sorry. From what I've read the coefficients are tuned for most cases, but can be changed in some.

Defend · April 22, 2014, 15:48

Anyway, thank you for solving my problem with residuals!
I hope someone else know more about EBU model. =)

April 21, 2014, 06:33	Problem: Convergence in unsteady EBU combustion model	#1
Defend New Member Alexey Join Date: Apr 2014 Posts: 5 Rep Power: 12	Hello! For monthes I've tried to solve the problem with the combustion and then decided to ask for a help on this forum =) I used non-premixed Standart EBU combustion model for chemical reaction C2H5OH+3O2=2CO2+3H2O to verify the experimental data. In experiment I have a burning liquid C2H5OH in closed volume room. To exclude multiphasing I used wall with normal mass flow of gas C2H5OH from its surface (not mass flow rate on inlet). Also I added small pressure outlet bondary condition. I tried to use different turbulent models, time-steps, Courant number, other solver parameters and all the time I have the same problem: residuals of CO2, H2O at every time-step becomes flat without any value decreasing. First time at value 10e-4, then 10e-2... and after many time steps at value 10e+7... I tried to increase the number of cells - I made 100000 polyhedral cells for simple 2d geometry. Also I found, that Green-Gauss gradient method instead of Hybrid Gauss-LSQ gives me better convergence. But again and again I have nondecreasing value of CO2 and H2O residual 10e+7. It would be the great luck if anyone of you could explain me, how to get the residual convergence 10e-5 at any time-step of such task. Thank you in advance.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
unsteady heat transfer model problem	Beryl	FLUENT	0	July 13, 2011 00:26
convergence problem in RNG k-e model	pj21aug	FLUENT	0	October 14, 2010 12:43
Convergence problem for P1 & Energy	HP	FLUENT	5	May 21, 2005 16:01
Unsteady Problem Convergence	Sham	FLUENT	2	March 17, 2005 05:34
Convergence Problem in k-e model	Rajeev Kumar Singh	FLUENT	3	April 15, 2002 08:13

April 21, 2014, 14:01		#2
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 25	Are those the normalized or unnormalized residuals?

April 21, 2014, 14:07		#3
Defend New Member Alexey Join Date: Apr 2014 Posts: 5 Rep Power: 12	what does it mean - normalize/unnormalize residual?..

April 21, 2014, 14:49		#4
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 25	The residuals are normalized by default to the average value of the given residual over the first few iterations in order to group them closer together. You can disable this and view their true values by turning the normalization option to off in the monitors folder. If the values you're looking at are very low, you may be hitting a precision limit or just seeing noise.

April 21, 2014, 15:39		#5
Defend New Member Alexey Join Date: Apr 2014 Posts: 5 Rep Power: 12	Hmm... Ok, thank you, I'll try tomorrow!

April 22, 2014, 05:02		#6
Defend New Member Alexey Join Date: Apr 2014 Posts: 5 Rep Power: 12	me3840, thank you! I had normalized values of residuals. I changed them to unnormalized and the true value is 1e-8!.. Maybe you could answer to my question about coefficient A of EBU combustion model: Everywhere in literature I see, that this coefficient for modeling must be A=4. And in my task to reach the temperature values that accords to experimental data this coeff must be A=0.1. Is it normal to change this coeff or I have some mistakes in solving?.. Thank you in advance.

April 22, 2014, 11:25		#7
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 25	I don't really know anything about the EBU models to say, sorry. From what I've read the coefficients are tuned for most cases, but can be changed in some.

April 22, 2014, 15:48		#8
Defend New Member Alexey Join Date: Apr 2014 Posts: 5 Rep Power: 12	Anyway, thank you for solving my problem with residuals! I hope someone else know more about EBU model. =)