Author |
Message |
Reken Patel (reken)
New member Username: reken
Post Number: 1 Registered: 08-2004
| Posted on Saturday, August 07, 2004 - 03:27 pm: | |
Hello I am using the student version of Flexpde and am trying to solve time dependent fluid flow through a pipe. I am having a little trouble with the boundary conditions though. Ultimately, I want to vary the viscosity distribution in the pipe and see how it changes with time. The boundary condition I want to setup would take what ever viscosity distribution that came out the end of the pipe and replace it as the incoming viscosity distribution. I was wondering if this was possible using Flexpde, and if so how can I implement this boundary condition. Thanks Reken
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 203 Registered: 06-2003
| Posted on Saturday, August 07, 2004 - 10:29 pm: | |
Assuming that viscosity is one of the variables of the problem, you can use the PERIODIC boundary command to match values on two sides of your figure. This will lock all the variables on the selected boundaries; you cannot make one variable periodic and another not. See Periodic Boundaries in the Help Index. |
Reken Patel (reken)
New member Username: reken
Post Number: 2 Registered: 08-2004
| Posted on Monday, August 09, 2004 - 01:35 pm: | |
I tried using this condition and found it to work if the problem was not time dependent. However, if I add time into the equation Flexpde fails to solve the problem. Is there something else I have to specify if the problem is time dependent? Thanks Reken
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 210 Registered: 06-2003
| Posted on Wednesday, August 18, 2004 - 03:05 pm: | |
There does indeed seem to be an error in handling periodic boundaries in time dependent problems. I will look into this and post a corrected version. However, I'm not sure what periodicity means in a pipe flow problem. Requiring that the pressures be equal at the ends of the pipe seems weird. |
gflo Member Username: gflo
Post Number: 5 Registered: 12-2009
| Posted on Wednesday, July 21, 2010 - 11:12 am: | |
Dear Mr. Nelson, We recently acquired our FlexPDE 6.13 professional version. On trying to run the attached program (which is the very basic form of a much more complex piping system), I found out that ver. 6.13 takes about 50 min. of running time, to reach a simulation time of 2.56 sec. The time step that is used in the runs is very small (about 0.003 to 0.005). With version 5 (student), the program ends in 1.3 min. in all and the time step is about 0.03. How can I make ver. 6 professional work in a logical time? Changing the increment when specifying the time range has no effect. Looking forward to hearing from you, George Florides
|
rgnelson Moderator Username: rgnelson
Post Number: 1387 Registered: 06-2003
| Posted on Wednesday, July 21, 2010 - 12:59 pm: | |
You sent the wrong file. We need the ".pde" script. |
gflo Member Username: gflo
Post Number: 6 Registered: 12-2009
| Posted on Thursday, July 22, 2010 - 02:00 am: | |
Apologies. I attach the '.pde' file. |
mgnelson Moderator Username: mgnelson
Post Number: 201 Registered: 07-2007
| Posted on Thursday, July 22, 2010 - 03:19 pm: | |
You have a steep temperature gradient at the inner surface of the pipe. The Professional version creates more nodes (~9000) and keeps a shorter time step in order to track the solution accurately. The node limit of the Student version (1600 in 3D) does not allow FlexPDE to do this, so it gets a poorer answer faster. If you do not care to keep the default accuracy of the Professional version (2e-3), you can request less accuracy by lowering the error limit. i.e. SELECT ERRLIM=1e-2. You can also limit the number of nodes, i.e. SELECT NODELIMIT=1600. If you add these to your script file, you will see behavior similar to the Student version. But the answer will not be as good as letting FlexPDE use the default values. There is always a trade-off between accuracy and speed. |
gflo Member Username: gflo
Post Number: 7 Registered: 12-2009
| Posted on Friday, July 23, 2010 - 03:06 am: | |
The run time can be controlled as you indicated. Thank you Mr. Nelson. |
gflo Member Username: gflo
Post Number: 8 Registered: 12-2009
| Posted on Sunday, August 08, 2010 - 06:04 am: | |
Dear Mr. Nelson, In the attached file, I have modeled a u-tube borehole heat exchanger which works fine. I want to extend the simulation to a real case scenario at which the only difference will be the depth of the borehole (bd) that will extend from 0.4 that is now, to 100. On my pc I can extend bd to 5 at which case Flexpde produces 756000 notes and 559000 cells using 2.5GB of memory out of 3 and about 25% of CPU. Extending the depth to 10, no mesh is produced after a 6 hour run. Please inform me on what I can do to lower the mesh size, at least on the z direction and at what type of computer I can run such simulations. Is there any other way to overcome the problem?
|
gflo Member Username: gflo
Post Number: 9 Registered: 12-2009
| Posted on Sunday, August 08, 2010 - 10:30 am: | |
correct file |
rgnelson Moderator Username: rgnelson
Post Number: 1393 Registered: 06-2003
| Posted on Sunday, August 08, 2010 - 03:58 pm: | |
FlexPDE tries to make mesh cells as nearly equilateral as possible. Severely distorted cells increase the numerical difficulty of the matrix solution. This is especially true for thin cylindrical shells. At your target length of 100, you have two cylinders, each with a length 1000 times the radius and 50,000 times the thickness of the cylindrical shell. This requires an enormous number of cells for adequate modeling. You should consider using coordinate scaling to reduce the length of the tube to a manageable size. See "Help->Technical Notes->Coordinate Scaling" for a discussion of this technique. You could reduce your pipe length by a factor of 100 and bring it back to the mesh size of your simpler test.
|
gflo Member Username: gflo
Post Number: 10 Registered: 12-2009
| Posted on Saturday, October 30, 2010 - 01:09 pm: | |
Dear Mr. Nelson, For the attached program, I would like to have your help in extracting the values of Taver, Tsp1 and Tsp2, every 30 seconds, in a table form (one variable in every column). The nearest I could get is as indicated in the output file, which obviously is not correct since it repeats the same numbers. Thanking you in advance, George ( Output file "Temps_2.txt" removed due to size ) |
rgnelson Moderator Username: rgnelson
Post Number: 1422 Registered: 06-2003
| Posted on Monday, November 01, 2010 - 02:55 pm: | |
TABLE assumes it must span the domain, so you get an enormous file with duplicated numbers. Since your tsp1, tsp2 and taver are functions of Z along a line, you can use an ELEVATION with EXPORT to output your table. See attached script. Also, by requesting monitors at 0.1 intervals, you hold the timestep to 0.1, and will need 500,000 timesteps to complete your run, no matter how stable the solution might be. You should not do this, because it prevents FlexPDE from completing the run as quickly as the conditions allow. Use FOR CYCLE=10 or similar control to allow the timestep to grow.
|