# wall, pressure outlet or outflow ??

 Register Blogs Members List Search Today's Posts Mark Forums Read

 June 11, 2017, 14:28 wall, pressure outlet or outflow ?? #1 Senior Member   dilaw meda Join Date: Jun 2017 Location: algeria Posts: 128 Rep Power: 2 hello every one , i am modeling an air jet in a room, and i get confused how to set the boundary conditions because i have convergence problems, well the room is 2.5 widh, 3m lenght, and 2m heigh , the blower of the air jet is placed in ceiling. what are your suggestions for the ceiling , the floor and the walls of the room? initial velocity of the jet is 8m/s ; velociti near to the walls and the floor is about 0.1 m/s or less

 June 12, 2017, 03:28 #2 Senior Member   Join Date: Aug 2014 Location: N/A Posts: 157 Rep Power: 5 For walls, one uses the wall BC. For an outlet one uses usually pressure Outlet BC. "Convergence issues" is not specific enough problem. Also it can depend on different things like mesh, time step, BC's... So a more detailed description of your setup is needed to help.

June 12, 2017, 08:28
#3
Senior Member

dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
Quote:
 Originally Posted by BlnPhoenix For walls, one uses the wall BC. For an outlet one uses usually pressure Outlet BC. "Convergence issues" is not specific enough problem. Also it can depend on different things like mesh, time step, BC's... So a more detailed description of your setup is needed to help.
Yes actually i think the issues was because the time step, in the begining i was using k epsilon model with a steady state, then i sitched to transient , when i set the time step to 0.0001 the solution converges rapidaly and every 20 itération a message displays "solution converged" then i increased the time step to 0.01 and until now i am waiting for that message �� . Can you explain to me the role of the time step , and why it converged in the transient state ?

 June 12, 2017, 08:46 #4 Senior Member   Join Date: Aug 2014 Location: N/A Posts: 157 Rep Power: 5 In numerical fluid dynamics you need to have a timestep because the continously coupled partial differential equations need to be disrcetized in order to be solved iteratively. If you look at the N-S equations you will have a variable t (=time). This needs to be defined as discrete time intervals -> the time step. What you see now in your simulations is that when you have a too large time step it can be a) that you simply need more subiterations per time step to reach the defined convergence level or be b) your solution diverges. For a) i suggest you to increase the number of subiterations in the solution window. You will reach the same level of convergence as with a lower time step, it simply takes longer. For b) it is necessary that you decrease your time step, because the Courant Criterium is not satisfied. But this seems not be the case here for you. So for a), ideally you want to find the transient time step, that you reach desired level of convergence in each time step in about 5-10 subiterations or so.

June 12, 2017, 08:53
#5
Senior Member

dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
Quote:
 Originally Posted by BlnPhoenix In numerical fluid dynamics you need to have a timestep because the continous coupled partial differential equations need to be disrcetized in order to be solved iteratively. If you look at the N-S equations you will have a variable t (=time). This needs to be defined as a disrete time intervals -> the time step. What you see now in your simulations is that when you have a too large time step it can be a) that you simply need more subiterations per time step to reach the defined convergence level or be b) your solution diverges. For a) i suggest you increase the number of subiterations in the solution window. You will reach the same level of convergence as with a lower time step, it simply takes longer. For b) it is necessary that you decrease your time step, because the Courant Criterium is not satisfied. But this seems not be the case here for you. Ideally you want to find the transient time step, that you reach desired level of convergence in each time step in 5-10 subiterations.
Ok, lets say the 0.0001 is good for me, the convergence was every 20 iterations and the residuals was acceptable. when should i stop the calculations because when i tooke a look to contours i found the jet was not so developped ?

 June 12, 2017, 09:04 #6 Senior Member   Join Date: Aug 2014 Location: N/A Posts: 157 Rep Power: 5 As long as you think it needs to be representative to your problem. I can speak for my simulations: I let them run at least 2 theoretical residence times. Meaning the domain volume needs to be changed 2 times by inlet flow rate. BTW: Are you doing single or multiphase? Have you tried steady state solver already?

June 12, 2017, 09:11
#7
Senior Member

dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
Quote:
 Originally Posted by BlnPhoenix As long as you think it needs to be representative to your problem. I can speak for my simulations: I let them run at least 2 theoretical residence times. Meaning the domain volume needs to be changed 2 times by inlet flow rate. BTW: Are you doing single phase or multiphase? Have you tried steady state solver already?
I am using a single phase and yes i tried the steafy state but the residuals were not so low as in the transient, also the magic message that i love "solution converged " didnt displays in steady state، the residuals keep decaying until some valu the it become stable. How should i choose steady or transient for my case ? I choose the one wich gives me convergence or it depends on my case ?

 June 12, 2017, 09:20 #8 Senior Member   Join Date: Aug 2014 Location: N/A Posts: 157 Rep Power: 5 Depends on your case. If residuals ocillate around, it usually indicates steady state solver has problems finding the solution and something transient happens in your model, so it's better modeled as transient. But i would not rely on what Fluent says if its converged or not. You decide if it's converged or not. Also, please don't open several threads for the same problem. I'm assuming that you have different simulations all the time..

June 12, 2017, 10:24
#9
Senior Member

dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
Quote:
 Originally Posted by BlnPhoenix Depends on your case. If residuals ocillate around, it usually indicates steady state solver has problems finding the solution and something transient happens in your model, so it's better modeled as transient. But i would not rely on what Fluent says if its converged or not. You decide if it's converged or not. Also, please don't open several threads for the same problem. I'm assuming that you have different simulations all the time..
thank you , because i have a big domaine i have a question about the mesh, how to qualify a mesh ? i am setting convergence criteria 10^-4 for velocities and continuity , and 10^-6 for energie. if i get thouse results with a not so refine mesh with small number of elements does it means that every thing is ok and i don't have to refine it or increase the number of elements ?

 June 12, 2017, 10:50 #10 Senior Member   Join Date: Aug 2014 Location: N/A Posts: 157 Rep Power: 5 Everything is OK, when you reached convergence for the values of interest!!! The Residuals are not the values of interest! Values of interest are velocities, pressure etc. If these values do not change within a certain tolerance, your solution is converged! Not if the residuals are 10e-4, 10e-8...or whatever. Residuals are at best an indicator that your simulation is iterating stable. So if you reached true convergence (see above what that means), then you compare to experiment. Then you do a mesh independence study. Then you do a time step independence study. And then you are finished and can make certain statements about the quality of your CFD model. Everything before that is just playing around... Sorry.

June 12, 2017, 10:59
#11
Senior Member

dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
Quote:
 Originally Posted by BlnPhoenix Everything is OK, when you reached convergence for the values of interest!!! The Residuals are not the values of interest! Values of interest are velocities, pressure etc. If these values do not change within a certain tolerance, your solution is converged! Not if the residuals are 10e-4, 10e-8...or whatever. Residuals are at best an indicator that your simulation is iterating stable. So if you reached true convergence (see above what that means), then you compare to experiment. Then you do a mesh independence study. Then you do a time step independence study. And then you are finished and can make certain statements about the quality of your CFD model. Everything before that is just playing around... Sorry.

 Thread Tools Display Modes Linear Mode

 Posting Rules You may not post new threads You may not post replies You may not post attachments You may not edit your posts BB code is On Smilies are On [IMG] code is On HTML code is OffTrackbacks are On Pingbacks are On Refbacks are On Forum Rules

 Similar Threads Thread Thread Starter Forum Replies Last Post Mike_Tom CFX 6 September 29, 2016 01:27 auf dem feld FLUENT 17 February 26, 2016 14:04 j0hnny CFX 12 January 7, 2016 03:44 Kwong FLUENT 1 April 11, 2007 05:04 chong chee nan FLUENT 0 December 29, 2001 06:13

All times are GMT -4. The time now is 19:02.