Exchange experiences in 3D simulation
Is there anyone who wants to exchange his or her experiences in 3D simulation such as airflow around buildings with me? I think it's a widely application for Phoenics in engineering. From my tests, I think the range of the max and min values of P1,V1, U1 and W1 is important, which could speed up a lot of time. Also, the value of SARAH and the initial value of P1,V1, U1 and W1 is also very important. Sometimes i'll try to adjust the relaxation factors or false time step every now and then, but the effect is not so good. Hope a deep communication here and any suggestion is welcome.
Bryan |
Re: Exchange experiences in 3D simulation
If there are a lot of buildings, then use of the Conjugate Gradient Residuals Solver for P1 (pressure correction) may be very beneficial. This is activated by setting ENDIT(P1)=GRND1.
You can also try setting CSG3=CNGR, which activates the CGR solver for all solved-for variables. The use of the CGR solver does involve an increase in computer time per sweep, but usually less sweeps are required for convergence. Inertial relaxation based on wind speed and typical cell length should be adequate for convergence, and the use of linear relexation on KE and EP with KELIN=3 is recommended. No linear relaxation on pressure should be necessary. I do not recommend using VARMIN and VARMAX on the velocities and pressure. In general, this is a dangerous practice, but it can be useful when applied to the density and temperature in buoyant flows, but then the introduction of buoyancy involves other considerations on the setting of numerical parameters which I will not enter into here. SARAH can be useful, but the optimum SARAH factor usually has to be determined by experimentation. The initial values of pressure, velocity and turbulence are not usually of importance, although initialising the appropriate velocity to the wind speed must help a little. |
Re: Exchange experiences in 3D simulation
dear Mike
Thank you very much for your good suggestion. But sometimes when i do a simulation for airflow around buildings, there are some strange breaking down. I used KE-chen model to do my 3d simulation and used the Conjugate Gradient Residuals Solver for P1 and set CSG3=CNGR, the mesh is 150x110*50, but when i ran the earth program, the screen showed some strange information as follows: EARDAT has been read from IRUN=1 LIBREF=302 PROPERTIES READ FROM PROPS PLOTS OF RESIDUAL & SPOT DATA REDUCED THEREFORE: 3D VELOCITY COEFFS STORED OUT OF CORE THEREFORE: DEPENDENT VARIABLES STORED OUT OF CORE THEREFORE: LU CONJUGATE GRADIENT SOLVER CUT OUT and the simulation just ran 1 step and stopped. Why? I think maybe the array exceeded the limitation of f array, but i am not sure. Waiting for your answers. Best regards. Bryan |
Re: Exchange experiences in 3D simulation
The EARTH F-array limit has been exceeded in main.for. Increase the dimension of the F array, recompile main, and build a new earexe.
|
Re: Exchange experiences in 3D simulation
It seems to me that the PHOENICS installation embodies two or three "earexe" executables correponding to various sizes of the F-array (ear128.exe, ear256.exe or sthing like that).
You may not have to rebuilt anything yourself therefore. Just make a safe copy of the existing "earexe.exe" and then copy "ear256" to "earexe.exe". If problems persue, you'd better contact CHAM's support directly. |
Re: Exchange experiences in 3D simulation
Quite right, Herve. A lot will depend on just how much memory this case requires.
|
Re: Exchange experiences in 3D simulation
Many thanks. And then how many memory is the default earexe.exe in the 3.3 version corresponding?
|
Re: Exchange experiences in 3D simulation
The default F-array size in PHOENICS V3.3 is 21 million which corresponds to a memory requirement of 80 MB.
|
All times are GMT -4. The time now is 11:18. |