Learning Forum

[za36kuzo]

I am using DEM collision model where I predefined few particles with density of 2650 Kg/m^3 inside the pipe, The pipe diameter = 0.0042m, length L= 1m. nParticle initial conditions can be read from an external file. at time = 0, (334) particles are placed at random position within the pipe. The velocity of the particles in radial direction is chosen randomly from -v to v= 0.01, while the particle velocity in axial direction is set to zero. nThe collision parameters for the collision pairs was Herzian_dashpot for normal contact force and friction_dshf_rolling for tangential force. nPeriodic boundary condition is applied in the axial direction with specified air mass flow in same direction of zero Kg/s.nTransient model, with unsteady particle tracking, and the gravity was activated, so the particles are falling in the axial direction.nI have run 334 particles for 10000-time steps, with 0.0001s time step size. The solution was successfully converged, but the calculation time took extremely long time (more than 24 hrs. for 40% Iteration).nAccording to the calculation of collision time and martial properties mentioned above, I should use time step size about 10^-5, with 100000-time steps, which supposed to have longer iteration time.I appreciate any advice on setup and/ or parameter that can help in reducing the iteration time. Thanks in advance.nnAbinn

[Keyur Kanade]

Can you also provide some mesh details like cell count, min orthogonal quality. How may cores you are using etc..nRegards,nKeyurnHow to access Ansys Online Help DocumentnHow to show full resolution imagenGuidelines on the Student CommunitynHow to use Google to search within Ansys Student Communityn

[za36kuzo]

Thanks for replaying, I am running it in parallel ,with 4 cores . I have 109152 elements , and the elements size is 0.005m as shown in the attached photos. nRegards,nAbinn

[Rob]

We're not permitted to open the attachments. How many particles? DEM can get expensive as the particle count goes up, and the time step needs to resolve the collision. Check the cpu load on each core too, is it balanced?n

[za36kuzo]

I have run 334 particles for 10000-time steps, with 0.0001s time step size, the remote computer which I perform my simulation on has processor: nInlet (R) Xeon (R) CPU E3 1280 v5 @3.70 GHz.nI solved the continuous phase as first with periodic boundary condition and specified air mass flow of zero kg/s, and then the DPM simulation as next step.nWhat make me confused about Residual, because the residual monitor for the first step (mass air flow = 0) showed no velocity and continuity, whereas for the DPM simulation, the monitor shows velocity and continuity iterations.nRegardsn

[Rob]

Once the particles start moving they will (if coupled) drag air with them. Given all of the particles will fall at about the same rate what are you trying to model? n

[za36kuzo]

I am trying to find a correlation by varying the pipe diameter and the time needed to occurance of the particles blockage in the pipe due to the collision between the particle-particle and particle-wall. Thus, for an accurate calculation of collision time, the simulation should run in time step size of 10^-6, for minimum 5-10 second. That means long iteration time.nNow I started with a large time step of 10^-4 for 1 second, which means 10000-time steps, and it was really longer than expected. I wonder if I have made wrong in the setup which I have sent in the first post (above).nRegardsn

[Rob]

If gravity and periodic conditions are in the axial direction the pipe can never block? n

[za36kuzo]

Collisions between the particles and between the particles and the wall lead to deceleration of the particles and a decrease in the kinetic energy of the system.nIn the regime of small pipe to particle diameter ratios, the distribution of the flow along the pipe may be strongly affected by geometric effects as the number of possible arrangements of the particles along the cross section of the pipe affects the probability of the formation of a stable arch. the jamming probability increases with the probability that a stable arch along the cross section of the pipe is formed. nThis already done in many previous studies with different commercial software?s but not with Ansys. The problem is, to detect the collision time, the time step size must be very short within 10^-6 for a time-period of 5-10 sec. Is it possible to perform that with Ansys? I have run a simulation withen this short time step size, and it was very very slow like 3 % iteration for 24 hrs. nRegardsn

[Amine Ben Hadj Ali]

DEM runs are expensive. Do you require to have the flow resolved? If not switch off flow and run pure particles!n

[za36kuzo]

DEM runs are expensive. Do you require to have the flow resolved? If not switch off flow and run pure particles!/forum/discussion/comment/105915#Comment_105915

Thanks for replaying, actually, because of the periodic boundary condition, I solved the continuous phase as first with periodic boundary condition and specified air mass flow as of zero Kg/s, in another word I switch off the air flow and then I have run the DPM phase as next as pure particle, but it still very longs time Iteration. I really surprised, because algorithmically, it should not be a problem, the particle dynamics to simulate 300 particles on a CPU.nRegardsnn

[Amine Ben Hadj Ali]

Yeah should be straight away: You will just to resolve particle collision and that is it. Are you running on a single machine? How many partitions?n

[za36kuzo]

Yeah should be straight away: You will just to resolve particle collision and that is it. Are you running on a single machine? How many partitions?/forum/discussion/comment/105927#Comment_105927

The remote computer which I perform my simulation on has processor Inlet (R) Xeon (R) CPU E3 1280 v5 @3.70 GHz, with 8 cores. Should this be enough for the calculation? nI am running it in parallel, with 4 cores. I have 109152 elements, the element size is 0.005m as shown as in the photos.nWhat do you mean partitions? should I enable parallel meshing, before I run with parallel mode? I just thought that Ansys fluent will automatically partition.nRegardsnnn

[Amine Ben Hadj Ali]

No parallel meshing but parallel run. Can you issue /parallel/timer usage in your Fluent session? Is it possible to directly run the case on the machine without being remotely connected and avoid using directly Fluent Standalone?n

[za36kuzo]

No parallel meshing but parallel run. Can you issue /parallel/timer usage in your Fluent session? Is it possible to directly run the case on the machine without being remotely connected and avoid using directly Fluent Standalone?/forum/discussion/comment/105974#Comment_105974

I perform my simulation on workbench 2020 R2 via remote computer, but not directly on my machine. The attached photo shows the calculation time ,I have run paralleln in 2/ 8 core.nRegardsnn

[za36kuzo]

No parallel meshing but parallel run. Can you issue /parallel/timer usage in your Fluent session? Is it possible to directly run the case on the machine without being remotely connected and avoid using directly Fluent Standalone?/forum/discussion/comment/105974#Comment_105974

I am sorry, I forget to send the CPU time. Here I send all info again, nI perform my simulation on workbench 2020 R2 via remote computer, but not directly on my machine. The attached photo shows the calculation time, please let me know, if you need more details, thanks in advance. nRegardsnn

[Amine Ben Hadj Ali]

Please first stop marking you are replies as answers to the question.n

[Amine Ben Hadj Ali]

Can you please try to run the case directly on the machine and not via Workbench? I see a huge amount of wall clock time compared to DPM solve time. Moreover you ran it for 212 iterations? Are the numbers reported here reasonable for you?n

[Amine Ben Hadj Ali]

Okay other post shows 25260 iterations:5 hours of run with DPM solve time less than 1%. Are the numbers reasonable? n

[za36kuzo]

Thanks for replaying, With the time step of 10^3 and 1000 time step, I didn?t face any problems with workbench, but since I stated reducing the time step to 10^-5, delay started occurring. nRegarding to the photos I attached, the old one refers to an old calculation I did but I did not send you CPU time. nI sent you the new photo shows the CPU time, and it concern to a new calculation, that I have run early today.