Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Particle tracking with freeze condition - how to not track frozen particles?
Posted 3 apr 2013, 08:00 GMT-4 Charged Particle Tracing, Particle Tracing for Fluid Flow Version 4.3a 3 Replies
Please login with a confirmed email address before reporting spam
Hi all,
I have a fairly simple model consisting of purely axial, laminar flow in a rectangular channel. These simulations are in the Stokes regime, and the flow is modeled as steady state. I would like to track the time-dependence of a large number of particles through this domain (Np > 50,000), where three of the channel sidewalls are inactive towards the particles (bounce condition) and the fourth wall is active towards particle capture (freeze or stick condition).
I am implementing both the drag and Brownian force on each particle, and I would like to measure either (1) the position of interaction of each particle with the active wall or (2) the location at which the particle flows through the channel outlet.
My problem: Because of the brownian motion condition, I need to use very small time step (Delta_t ) values in order to ensure the motion of the collective particle set resembles that of a continuum. Coupling this with the large number of particles used gives simulation times that are somewhat large.
I have noticed that the overall simulation time is independent of the boundary condition used for the capture of particles (freeze, stick, or even disappear). The simulation times for low Pe flows (where all of the particles should be captured before the outlet) are identical to the simulation times for very high Pe flows (where very few particles should be captured).
Is there a convenient way to set up these simulations such that after a particle is "captured," its location on the boundary is noted, and removed from the particle tracking algorithm? This should drastically speed up the simulation times for low Pe flows.
Thanks in advance,
Nicholas
I have a fairly simple model consisting of purely axial, laminar flow in a rectangular channel. These simulations are in the Stokes regime, and the flow is modeled as steady state. I would like to track the time-dependence of a large number of particles through this domain (Np > 50,000), where three of the channel sidewalls are inactive towards the particles (bounce condition) and the fourth wall is active towards particle capture (freeze or stick condition).
I am implementing both the drag and Brownian force on each particle, and I would like to measure either (1) the position of interaction of each particle with the active wall or (2) the location at which the particle flows through the channel outlet.
My problem: Because of the brownian motion condition, I need to use very small time step (Delta_t ) values in order to ensure the motion of the collective particle set resembles that of a continuum. Coupling this with the large number of particles used gives simulation times that are somewhat large.
I have noticed that the overall simulation time is independent of the boundary condition used for the capture of particles (freeze, stick, or even disappear). The simulation times for low Pe flows (where all of the particles should be captured before the outlet) are identical to the simulation times for very high Pe flows (where very few particles should be captured).
Is there a convenient way to set up these simulations such that after a particle is "captured," its location on the boundary is noted, and removed from the particle tracking algorithm? This should drastically speed up the simulation times for low Pe flows.
Thanks in advance,
Nicholas
3 Replies Last Post 3 apr 2013, 13:01 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Nicholas, I don't think your suggestion will make the model solve any faster. The timesteps taken by the solver are determined by the error tolerances specified. The default solver tolerances can be too strict when including Brownian motion in the model.
I suggest you look at the model Particle Tracing Module Model Library>Tutorial Models>Brownian Motion. This model relaxes the relative and absolute tolerances to 1E-3 and the model solves hundreds of times faster and the results still agree with the theoretical values.
I suggest you look at the model Particle Tracing Module Model Library>Tutorial Models>Brownian Motion. This model relaxes the relative and absolute tolerances to 1E-3 and the model solves hundreds of times faster and the results still agree with the theoretical values.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Daniel,
Thanks for your reply.
I found that the absolute tolerances used by the Brownian motion model (1e-3) are too low. Although the solution pertaining to the particle tracing model gives similar results to the solution to the diffusion equation, those solutions only match for the diffusivity (particle size) and final time given by the module. The results for the two solution methods do not agree well over a variety of simulation conditions (the particle size, the overall size of the domain, or the final time).
I am at home now so I don't have access to these particular results, but I will try to post them in the morning.
Nicholas
Thanks for your reply.
I found that the absolute tolerances used by the Brownian motion model (1e-3) are too low. Although the solution pertaining to the particle tracing model gives similar results to the solution to the diffusion equation, those solutions only match for the diffusivity (particle size) and final time given by the module. The results for the two solution methods do not agree well over a variety of simulation conditions (the particle size, the overall size of the domain, or the final time).
I am at home now so I don't have access to these particular results, but I will try to post them in the morning.
Nicholas
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi, the model was created assuming that none of the initial mass (for the concentration equation) or particles (for the Particle tracing equations) reach the outer walls. If, for example, you drop the particle radius by a factor of 2, particles will reach the outer walls before the last simulation time. In this case, you need to set the boundary condition for the particle tracing interface to "Bounce" and remove the "Concentration" boundary condition in the Diluted Species interface.
~ Dan
~ Dan