Time step - DPM steady tracking

  • Last Post 08 March 2019
liliana_augusto posted this 07 March 2019

Hello all.


I have a simulation with the following characteristics: Two concentric cylinders where the flow of a non newtonian fluid occurs in the annular region. The inner cylinder is swirling at 200 rpm. The outer cylinder is stationary. 

I would like to track a particle inside the annular region using DPM. The flow can be considered as unchanged during the particle track. So I am using the steady tracking for DPM particles. Basically, the particle also swirls with the fluid.

However, I am having some issues by defining the time step parameters for the steady tracking (maximum number of step and length scale). How would be the right way to choose these values? Is there any kind of test (like a grid independence test) to define the time step by decreasing the value of the length scale (but keeping the max. number of steps) until the solution is unchanged? Is this correct to do?

I have observed that by changing these parameters, the solution also changes.


Order By: Standard | Newest | Votes
rwoolhou posted this 08 March 2019

You're not setting time step parameters as it's steady: you're telling the solver how often to check the particle trajectory. Sorry for being pedantic but mixing terms causes a lot of confusion. 

Step length & factor are defined in the manual: read that and then follow that advice. The maximum number of steps should be enough to get the particles through the domain but not too high as it'll use lots of cpu if the particle gets stuck somewhere (corner or recirculation zone). A max value of 5000-10000 is reasonable in many applications, but again you'll need to check as it's cell & domain size dependent. 

liliana_augusto posted this 08 March 2019

Hi, thanks for reply.

I just said "time step" because it is on Fluent's manual:

Length Scale

controls the integration time step size used to integrate the equations of motion for the particle. The integration time step is computed by ANSYS Fluent based on a specified length scale , and the velocity of the particle () and of the continuous phase ():

The problem is, as I said, I have two cylinders and one of them is swirling. So, I have a recirculation zone. That is why is no so easy to follow the manual, because all recommendations to define the value of these parameters is made regarding the time/distance to cross the domain. That is why I asked if there is an independence test that I could do.

Best regards.

liliana_augusto posted this 08 March 2019

Do you have any recommendation about the use of "accuracy control", in DPM numerics options?

rwoolhou posted this 08 March 2019

Leave it alone: it should be on. If the flow gradient is smooth the solver will reduce the number of checks & speed up the solution; if it's not smooth it'll do more checks.  

The independence check for max steps is whether the number of aborted particles keeps dropping with extra steps. If you've got particles stuck in a recirculation zone they may not leave so don't set too high a value or you'll be waiting a long time for anything to happen. 

liliana_augusto posted this 08 March 2019

Ok, I asked because the solution with the accuracy control activated is very different from the solution with this option disabled.