CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Using LogNormal distribution with DPM spray model (

joshkemp March 27, 2007 14:39

Using LogNormal distribution with DPM spray model
I am trying to model spray from a nozzle for which I have experimental data. The distribution is fit very well with a log normal curve, so I would like to use this to initialize the spray from the nozzle. I also would like to have to droplet size randomly distributed across the nozzle area.

What is the best way to do this? I know there is the read from file option or possibly a UDF, but what is the best approach?

Allan Walsh March 28, 2007 17:36

Re: Using LogNormal distribution with DPM spray mo
In my experience, it is better to write a stand alone program to make a file to read into Fluent. In version 4 of Fluent, we wrote our code in Fortran that was integrated into Fluent to generate the sprays. Incidentally, one of the reasons we wrote our spray routine was to integrate a lognormal diameter relationship and to have each diameter class represent an equal fraction of the spray.

In version 6, we have used Excel with a Visual Basic macro to generate the spray file. This has been much simpler as you don't have to do a lot of work generating on-the-fly forms to input spray parameters. And, you don't have to recompile all the time to debug your code.

Good luck.

joshkemp March 28, 2007 18:20

Re: Using LogNormal distribution with DPM spray mo
Thanks for the help Allan. I am new to spray modeling.

I have another question relating to mesh requirements for DPM. I recall reading somewhere that your mesh must be coarse enough so that your biggest droplet does not occupy more than one cell. I searched through the Fluent documentation and cannot find any mention of this. I think I may have read this in a multiphase book I have. Do you know if this is required for Fluent? If so, I may have a hard time keeping the correct y+ values near the wall for the k-epsilon turbulence model.

Allan Walsh March 29, 2007 13:23

Re: Using LogNormal distribution with DPM spray mo
I don't think it is a requirement - and after modeling hundreds of cases like this in Fluent - it hasn't been an issue. In a steady-state calculation, you can tweak the step size or length scale to get nominally one discrete exchange event in each cell for each time step. This would spread out the source exchange terms.

All times are GMT -4. The time now is 12:51.