Home >
Community >
Difference between Steady and Unsteady Tracking in the DPM of Fluent?
Upvote
30
Downvote
+ Particle
Posted by
Ken Giuffre
Difference between Steady and Unsteady Tracking in the DPM of Fluent?
Hi Eduard,
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I know it is quite late to answer your questions, but perhaps my answer would be helpful for everybody who works on Eulerian/Lagrangian modelling.
In general, we have 3 different E/L modellings:
steady, unsteady and semi-unsteady modelling.
In steady modelling, first, run the eulerian model until it reached the convergence criteria and just freeze it as suggested. Then run Lagrangian solver just for one Euler iteration. Of course, it does not mean that the lagrangian solver works as steady, but in one eulerian iteration, it solves PDE until the particle reaches the final destination. This is why you only can have the particle track line at the end.
In unsteady modelling, you have to select a time step for both solvers (dt_E~dt_L). Thus you can track the particle location, velocity, etc at each time step. However, in general, the time step needed for an Eulerian solver is quite bigger than a lagrangian solver and using a small-time step for both solvers would be time-consuming. Evidently, this idea is suitable for DNS modelling and not for all cases.
In most simulations, we can use a semi-unsteady model. In this approach, you have to run an eulerian solver with a bigger time step, then after each time step transfer data to lagrangian solver with smaller time step, run it and again transfer data to Euler solver if needed (like two-way coupling) and continue until reach final time needed for modelling. As the consequence, the simulation would be quite faster (dt_E>>dt_L).
Finally, to the best of my knowledge, all of the options explained above exist in Fluent, OpenFOAM and StarCCM+.
I know it is quite late to answer your questions, but perhaps my answer would be helpful for everybody who works on Eulerian/Lagrangian modelling.
In general, we have 3 different E/L modellings:
steady, unsteady and semi-unsteady modelling.
In steady modelling, first, run the eulerian model until it reached the convergence criteria and just freeze it as suggested. Then run Lagrangian solver just for one Euler iteration. Of course, it does not mean that the lagrangian solver works as steady, but in one eulerian iteration, it solves PDE until the particle reaches the final destination. This is why you only can have the particle track line at the end.
In unsteady modelling, you have to select a time step for both solvers (dt_E~dt_L). Thus you can track the particle location, velocity, etc at each time step. However, in general, the time step needed for an Eulerian solver is quite bigger than a lagrangian solver and using a small-time step for both solvers would be time-consuming. Evidently, this idea is suitable for DNS modelling and not for all cases.
In most simulations, we can use a semi-unsteady model. In this approach, you have to run an eulerian solver with a bigger time step, then after each time step transfer data to lagrangian solver with smaller time step, run it and again transfer data to Euler solver if needed (like two-way coupling) and continue until reach final time needed for modelling. As the consequence, the simulation would be quite faster (dt_E>>dt_L).
Finally, to the best of my knowledge, all of the options explained above exist in Fluent, OpenFOAM and StarCCM+.
My opinion is that you can "frozen" the continuous field to integrate the particle motion over a subset of time-steps. The time step for particles refers to the governing equation in Lagrangian form as detailed here http://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node243.htm
My opinion is that you can "frozen" the continuous field to integrate the particle motion over a subset of time-steps. The time step for particles refers to the governing equation in Lagrangian form as detailed here http://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node243.htm
Hi Eduard,
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I hope this helps!
Hi Eduard,
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I hope this helps!
More
VOTE
Dear Eduard
The DPM method is well described in
1. http://www.afs.enea.it/fluent/Public/Fluent-Doc/PDF/chp19.pdf
2. https://www.sharcnet.ca/Software/Fluent6/html/ug/node835.htm
All the best
Dear Eduard
The DPM method is well described in
1. http://www.afs.enea.it/fluent/Public/Fluent-Doc/PDF/chp19.pdf
2. https://www.sharcnet.ca/Software/Fluent6/html/ug/node835.htm
All the best
More
VOTE
Hi Eduard,
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I hope this helps!
Hi Eduard,
I am no expert in Fluent, but in DPM methods (better known as particle trajectory methods or particle-in-cell methods), Prof. Denaro put the answer very nicely.
For unsteady flow-unsteady particle coupling, the flow time step could be made different from the particle time step, especially if the Stokes number is very small. That means that during a single Eulerian (flow) time step, multiple smaller Lagrangian (particle) time steps can take place.
I hope this helps!
More
VOTE
Hi Eduard Vinter
I know it is quite late to answer your questions, but perhaps my answer would be helpful for everybody who works on Eulerian/Lagrangian modelling.
In general, we have 3 different E/L modellings:
steady, unsteady and semi-unsteady modelling.
In steady modelling, first, run the eulerian model until it reached the convergence criteria and just freeze it as suggested. Then run Lagrangian solver just for one Euler iteration. Of course, it does not mean that the lagrangian solver works as steady, but in one eulerian iteration, it solves PDE until the particle reaches the final destination. This is why you only can have the particle track line at the end.
In unsteady modelling, you have to select a time step for both solvers (dt_E~dt_L). Thus you can track the particle location, velocity, etc at each time step. However, in general, the time step needed for an Eulerian solver is quite bigger than a lagrangian solver and using a small-time step for both solvers would be time-consuming. Evidently, this idea is suitable for DNS modelling and not for all cases.
In most simulations, we can use a semi-unsteady model. In this approach, you have to run an eulerian solver with a bigger time step, then after each time step transfer data to lagrangian solver with smaller time step, run it and again transfer data to Euler solver if needed (like two-way coupling) and continue until reach final time needed for modelling. As the consequence, the simulation would be quite faster (dt_E>>dt_L).
Finally, to the best of my knowledge, all of the options explained above exist in Fluent, OpenFOAM and StarCCM+.
I hope you find it useful.
Hi Eduard Vinter
I know it is quite late to answer your questions, but perhaps my answer would be helpful for everybody who works on Eulerian/Lagrangian modelling.
In general, we have 3 different E/L modellings:
steady, unsteady and semi-unsteady modelling.
In steady modelling, first, run the eulerian model until it reached the convergence criteria and just freeze it as suggested. Then run Lagrangian solver just for one Euler iteration. Of course, it does not mean that the lagrangian solver works as steady, but in one eulerian iteration, it solves PDE until the particle reaches the final destination. This is why you only can have the particle track line at the end.
In unsteady modelling, you have to select a time step for both solvers (dt_E~dt_L). Thus you can track the particle location, velocity, etc at each time step. However, in general, the time step needed for an Eulerian solver is quite bigger than a lagrangian solver and using a small-time step for both solvers would be time-consuming. Evidently, this idea is suitable for DNS modelling and not for all cases.
In most simulations, we can use a semi-unsteady model. In this approach, you have to run an eulerian solver with a bigger time step, then after each time step transfer data to lagrangian solver with smaller time step, run it and again transfer data to Euler solver if needed (like two-way coupling) and continue until reach final time needed for modelling. As the consequence, the simulation would be quite faster (dt_E>>dt_L).
Finally, to the best of my knowledge, all of the options explained above exist in Fluent, OpenFOAM and StarCCM+.
I hope you find it useful.
More
VOTE
My opinion is that you can "frozen" the continuous field to integrate the particle motion over a subset of time-steps. The time step for particles refers to the governing equation in Lagrangian form as detailed here
http://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node243.htm
My opinion is that you can "frozen" the continuous field to integrate the particle motion over a subset of time-steps. The time step for particles refers to the governing equation in Lagrangian form as detailed here
http://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node243.htm
More
VOTE