volume flow-rate calculation.....

edited October 2014 in Old versions
In the last question.

Mr. Alex said that maybe paraview can calculate the flow-rate.

But I have been searching all sites, I cannot find it.

So I stuck in same state..Could somebody suggest the ideas?

For calculating volume or mass flow-rate....

Thanks.

Comments

  • Try the pv-meshless plugin for Paraview

    Regards
  • Thank you for respond.

    I used pv-meshless. tried many options. sph probe, image resample...etc.

    But also that cannot calculate. so just possible option is making code?

    Regards
  • Yes, the good option is implementing your own code.
    You can use as example ToVTK.sln and implement an algorithm where you count the particles that are travelling into a region and the number of particles that leave that region for each time step.

    Regards
  • "count the particles that are travelling into a region and the number of particles that leave that region for each time step", does that mean we can calculate the flow rate by paticles mutiply its unit volume? If so, however, how do you deal with the volumes that among paticles??
  • And how to count the particles, by id? Can you tell how does one particle set to an id?
  • Has there been any progress in implementing a mass flow rate algorithm?
  • As mentioned, this is an easy task:
    You can use as example ToVTK.sln and implement an algorithm where you count the particles that are travelling into a region and the number of particles that leave that region for each time step.

    We will try to do an effort and we can include sthg like that in the next release

    Regards
  • Hi Alex, which specific files should I look to be analysing and modifying in ToVTK? Sorry, I'm just new to working with such a big code base.

    Adam
  • edited December 2015
    Yes, DualPhysics code is not that easy (for me at least, even tough it is fairly well documented), especially since it might envolve CUDA concepts and such (not TOVTK specifically, but the simulation code do).
    If you have trouble working with such a big code why not use the Measure Tool and process the output of it? You could interpolate velocities on a plane-grid and multiply by the effective area. If this plane is parallel to any of the axis it would be very simple, even if not it's still doable. You can even do it in Excel!
    It sure ain't pretty but it would work.
  • edited December 2016
    Hi alexmees, I'm starting to implement the general idea you mentioned.

    I think the only problem with the idea is when considering the flow of "clumped" and unsteady flow through a grided surface - specifically when the temporal discretization is large compared to the acceleration of the flow - which results in the MeasureTool missing the dynamics of the flow in between time steps.

    I am doing an initial validation of this technique on a cube of water experiencing free fall, and have set up a plane grid of points (approx 3600 points equispaced to the order of the kernel spacing) to feed into MeasureTool. As you would expect, the y velocities are uniform across the grid, and the MeasureTool only records data for 3 time steps - 0.045s (which is physical reasonable). The only problem with this is that the velocities measured step up suddenly from 0 to around 1ms, and then step down again after the 3 time
    steps. The velocity data at each point basically looks like this:

    0
    -1.0289
    -1.17748
    -1.3263
    0

    I have used trapezoidal integration to integrate the velocity flow over the time steps and then multiplied by the effective area of each grid and summed to get the total volume that has flown through the surface over the total time.

    This form of approximation of the dynamics in-between time steps severely overestimates the total volume that has flown through however ~ 0.00019m^3 compared to the initial volume of 0.000125m^3 (it was a 0.05m cube).

    When I perform trap. integration only in-between points 3-2 and 4-3, i get a much more reasonable volume calculation of 0.0001276m^3. (performing no interpolation and just multiplying each velocity by it's time step gives 0.0001324m^3).

    I'm not happy with seemingly performing an arbitrary truncation of the data set to get calculations closer to what they should be. I have a gigantic simulation (already performed) that I wish to eventually apply these processing techniques to; and it will be very unpractical and maybe even incorrect to do such truncation on results that are almost impossible to validate against. It seems that the solution to this problem lies primarily in making the output time step smaller to accurately capture the dynamics of the flow.

    If interested, I made a python script I can share that generates a grid of points set to a user defined plane to feed into MeasureTool. I also can comment up and share a general excel file if interested as I will have to do this anyway for future simulations.

    Best,

    Adam
  • I don't think this technique would work for (very) unsteady flow (the larger the flow variations the larger the error). I'm working with 6000 time steps (1/100s output = 60s) and measuring flow in a relatively non-turbulent region of the simulation and still the results vary quiet a bit around the mean (which is correct). My guess you would have to sample the results at a higher rate or a longer time to get meaningful results (depending on what would work for your specific case).
    This is not a method limitation but exactly what would happen in a physical model/real world application. When I worked with magnetic flowmeters we used to account for the full range of discharge variations during test runs (about 10 min). Even though those flowmeters are relatively accurate the variations where of a few percent during the length of the measurements.
Sign In or Register to comment.