### Post by anurag on Nov 11, 2017 7:14:59 GMT -6

Hi Alberto,

I would like to thank you for the OpenQBMM library. I have been working with it in the recent months for my simulations in a multiphase separation system. I use the library to model droplet coalescence using custom coalescence kernels. I have a couple of questions regarding this:

1. The initial moments in the library seem to be calculated from a number-based distribution. I am using q0 (1/m) and mean diameter d (m) values to calculate the initial fields. The way I do it is this (V_total is total volume of the geometry) :

However, in one of your comments (http://openqbmm.freeforums.net/post/53/thread) you mention that moments should be calculated in the following way:

This is a bit confusing as the units don't seem to match!

2. The second question is related to advection schemes. As far as I understand, the code uses the mean phase velocity to transport moments. The following link, however, mentions one can specify the nature of advection (including noAdvection).

www.openqbmm.org/moment-advection-schemes/

I have not explicity defined any advection scheme in the quadratureProperties dictionary of my case (I have both batch as well as continuous separation cases). Does this mean the advection is set to zero when not defined?

3. I've found that my simulations run fine with first-order discretization for all quantities. However, I encounter divergence when I use second-order discretization schemes for velocity or phase fractions. I've tried using limitedLinear/limitedLinearV schemes but it does not help much (zero-order moment becomes negative). I also tried playing around with other settings in the fvSchemes dictionary or reducing the time step (1E-3) but it has very little effect. Another thing I noticed was that improving the mesh quality (non-orthogonality 60+ -> 40+) had a huge impact as my simulation can now run for 5 seconds of flow time as compared to less than 1 second before. I am using the realizable ODE solver settings, as mentioned here as I am using a custom coalescence kernel. Do you have any suggestions to get around this problem?

Thanks a lot for your time!

Anurag

I would like to thank you for the OpenQBMM library. I have been working with it in the recent months for my simulations in a multiphase separation system. I use the library to model droplet coalescence using custom coalescence kernels. I have a couple of questions regarding this:

1. The initial moments in the library seem to be calculated from a number-based distribution. I am using q0 (1/m) and mean diameter d (m) values to calculate the initial fields. The way I do it is this (V_total is total volume of the geometry) :

`m_k = sum_i [(q0_i/V_total) * (d_i)^k)*(d_i - d_i-1)]`

However, in one of your comments (http://openqbmm.freeforums.net/post/53/thread) you mention that moments should be calculated in the following way:

`m_k = sum_i (w_i * (xi_i)^k)`

This is a bit confusing as the units don't seem to match!

2. The second question is related to advection schemes. As far as I understand, the code uses the mean phase velocity to transport moments. The following link, however, mentions one can specify the nature of advection (including noAdvection).

www.openqbmm.org/moment-advection-schemes/

I have not explicity defined any advection scheme in the quadratureProperties dictionary of my case (I have both batch as well as continuous separation cases). Does this mean the advection is set to zero when not defined?

3. I've found that my simulations run fine with first-order discretization for all quantities. However, I encounter divergence when I use second-order discretization schemes for velocity or phase fractions. I've tried using limitedLinear/limitedLinearV schemes but it does not help much (zero-order moment becomes negative). I also tried playing around with other settings in the fvSchemes dictionary or reducing the time step (1E-3) but it has very little effect. Another thing I noticed was that improving the mesh quality (non-orthogonality 60+ -> 40+) had a huge impact as my simulation can now run for 5 seconds of flow time as compared to less than 1 second before. I am using the realizable ODE solver settings, as mentioned here as I am using a custom coalescence kernel. Do you have any suggestions to get around this problem?

Thanks a lot for your time!

Anurag