reversed flow in a tank

Comi · May 9, 2019, 04:15

Hi, I'm trying to simulate temperature stratification of water inside a tank, via transient simulation. THe initial temperature of the fluid inside the tank is 35 °C and the inlet (Mass-flow inlet) flow has a temperature of 41.3 °C. The output is placed on the cilindrical wall, perpendicular to the ground. I set the output as pressure outlet.
My problem is that I can't remove the reversed flow on the pressure outlet. Morevoer the temperature in the tank takes too much physical time to reach the expected value.

I'm using a pressure based solver with a k-epsilon turbulent model.

Hope that somebody has some good ideas.
Thanks everybdoy

MKuhn · May 9, 2019, 05:42

A good way to prevent the reversed flow is to model an additional pipe to the outlet of your tank.

Another hint: First perform a steady state solution until convergence at constant temperature of 35°C. Than switch to transient mode and change the temperature at your inlet. The time step size depends what you would like to see. Just the mixing of the water? Than the outflow conditions could be sufficient and you will have no problems with reversed flows. If you are interested in buoyancy effects, than swith buoyancy on and define a temperature dependent density for water in the material properties. In this case you need smaller time steps.

Comi · May 9, 2019, 05:48

Quote:

Originally Posted by MKuhn

A good way to prevent the reversed flow is to model an additional pipe to the outlet of your tank.

Do you mean to add another pipe?

Quote:

Originally Posted by MKuhn

Another hint: First perform a steady state solution until convergence at constant temperature of 35°C. Than switch to transient mode and change the temperature at your inlet. The time step size depends what you would like to see. Just the mixing of the water? Than the outflow conditions could be sufficient and you will have no problems with reversed flows. If you are interested in buoyancy effects, than swith buoyancy on and define a temperature dependent density for water in the material properties. In this case you need smaller time steps.

My goal is to study the temperature variation in function of the time inside the tank. Since the tank volume is of 25 liters, my problem is htat it takes too many liters from the input to reach the new temperature value inside the domain.

MKuhn · May 9, 2019, 08:11

Yes I mean adding another pipe.

How small is your flow? 1 l/min is not so much and means it takes 25 minutes to fill up the tank.
If your flow is so small, than are other effects like the heat losses much more important.

Indeed you need smaller time steps and more iterations per time step until the flow is well-established.

Comi · May 9, 2019, 08:17

My flow is actually 10 l/min. o it should take more or less 2-3 minutes. This is confirmed by the tests performed by my colleagues

Comi · May 16, 2019, 08:32

The reversed flow was fixed using outflow thanks. I've got another question, though. By comparing the results with the experimental data, it looks like the process described by the simulation takes 10 times more time than the prototype, even if the volumes and the mass flow are the same. Is there maybe a scale factor in the transient algoritm that I'm not taking into consideration?

MKuhn · May 16, 2019, 09:40

Take smaller time steps (to divide in half) and check if the time is the same. If not you have to take ever smaller time steps.
Did you consider buoyancy with a temperature dependent density? This could have an additional mixing effect.

Comi · May 16, 2019, 09:43

Thanks, I will try that

MKuhn · May 16, 2019, 09:51

But not both hints at once!

Comi · May 16, 2019, 10:49

Yes, of course. It would be unproductive otherwise. Thanks again!

Comi · June 11, 2019, 06:29

Hi, I wasn't able to fix the problem. So I diecided to make something simpler , in order to understand how the time dependent slution works.

My idea was to study the flow inside a L = 1 meter tube. The tube has a 1 inch diameter. The fluid (water) temperature at the beginning is 35 °C, I've selected an outflow as output. I've set a mass flow inlet with a flow of that gives me a velocity of v = 0.27335 m/s ( more or less, depens on the approxiamtion of the mass flow), with a temperature of 41.3 °C. Is it right to suppose that the heat front should be able to reach the outflow in a period of time (T) given by
T = L/v ?

Thanks

MKuhn · June 11, 2019, 06:58

Yes, try it with a time step of 0,1 s. And stop the solution after every time step and see how it looks.

If the flow is not etablished (like open a water-tap) you will need some hundred iterations for the first step.

If you have only a temperature jump at time=0 and the flow is etablished (from a previous steady state solution for example) than you need less iterations.

May 9, 2019, 04:15	reversed flow in a tank	#1
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	Hi, I'm trying to simulate temperature stratification of water inside a tank, via transient simulation. THe initial temperature of the fluid inside the tank is 35 °C and the inlet (Mass-flow inlet) flow has a temperature of 41.3 °C. The output is placed on the cilindrical wall, perpendicular to the ground. I set the output as pressure outlet. My problem is that I can't remove the reversed flow on the pressure outlet. Morevoer the temperature in the tank takes too much physical time to reach the expected value. I'm using a pressure based solver with a k-epsilon turbulent model. Hope that somebody has some good ideas. Thanks everybdoy

May 16, 2019, 09:43		#8
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	Thanks, I will try that MKuhn likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PEM fuel cell: reversed flow at high current density	pchoopanya	FLUENT	3	March 23, 2019 10:58
Does fluent calculate reversed flow situations more simply than it should?	heksel8i	FLUENT	0	August 8, 2013 08:03
Reversed flow at pressure outlet	Seeker Phil	FLUENT	9	January 2, 2010 05:21
reversed flow	Sastry	FLUENT	5	March 22, 2007 08:05
Flow in a circular tank	PK	FLUENT	0	November 23, 2006 05:13

May 9, 2019, 05:42		#2
MKuhn Senior Member Moritz Kuhn Join Date: Apr 2010 Location: Germany, Dresden Posts: 207 Rep Power: 17	A good way to prevent the reversed flow is to model an additional pipe to the outlet of your tank. Another hint: First perform a steady state solution until convergence at constant temperature of 35°C. Than switch to transient mode and change the temperature at your inlet. The time step size depends what you would like to see. Just the mixing of the water? Than the outflow conditions could be sufficient and you will have no problems with reversed flows. If you are interested in buoyancy effects, than swith buoyancy on and define a temperature dependent density for water in the material properties. In this case you need smaller time steps.

May 9, 2019, 08:11		#4
MKuhn Senior Member Moritz Kuhn Join Date: Apr 2010 Location: Germany, Dresden Posts: 207 Rep Power: 17	Yes I mean adding another pipe. How small is your flow? 1 l/min is not so much and means it takes 25 minutes to fill up the tank. If your flow is so small, than are other effects like the heat losses much more important. Indeed you need smaller time steps and more iterations per time step until the flow is well-established.

May 9, 2019, 08:17		#5
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	My flow is actually 10 l/min. o it should take more or less 2-3 minutes. This is confirmed by the tests performed by my colleagues

May 16, 2019, 08:32		#6
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	The reversed flow was fixed using outflow thanks. I've got another question, though. By comparing the results with the experimental data, it looks like the process described by the simulation takes 10 times more time than the prototype, even if the volumes and the mass flow are the same. Is there maybe a scale factor in the transient algoritm that I'm not taking into consideration?

May 16, 2019, 09:40		#7
MKuhn Senior Member Moritz Kuhn Join Date: Apr 2010 Location: Germany, Dresden Posts: 207 Rep Power: 17	Take smaller time steps (to divide in half) and check if the time is the same. If not you have to take ever smaller time steps. Did you consider buoyancy with a temperature dependent density? This could have an additional mixing effect.

May 16, 2019, 09:51		#9
MKuhn Senior Member Moritz Kuhn Join Date: Apr 2010 Location: Germany, Dresden Posts: 207 Rep Power: 17	But not both hints at once!

May 16, 2019, 10:49		#10
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	Yes, of course. It would be unproductive otherwise. Thanks again!

June 11, 2019, 06:29		#11
Comi Member Mattia Join Date: Mar 2018 Posts: 45 Rep Power: 8	Hi, I wasn't able to fix the problem. So I diecided to make something simpler , in order to understand how the time dependent slution works. My idea was to study the flow inside a L = 1 meter tube. The tube has a 1 inch diameter. The fluid (water) temperature at the beginning is 35 °C, I've selected an outflow as output. I've set a mass flow inlet with a flow of that gives me a velocity of v = 0.27335 m/s ( more or less, depens on the approxiamtion of the mass flow), with a temperature of 41.3 °C. Is it right to suppose that the heat front should be able to reach the outflow in a period of time (T) given by T = L/v ? Thanks

June 11, 2019, 06:58		#12
MKuhn Senior Member Moritz Kuhn Join Date: Apr 2010 Location: Germany, Dresden Posts: 207 Rep Power: 17	Yes, try it with a time step of 0,1 s. And stop the solution after every time step and see how it looks. If the flow is not etablished (like open a water-tap) you will need some hundred iterations for the first step. If you have only a temperature jump at time=0 and the flow is etablished (from a previous steady state solution for example) than you need less iterations.