|
|
In addition I would like to ask a question. In phasePair.C the methods EoH1() and EoH2() call the method EoH() for computing the Eotvos number. Both methods compute the hydraulic diameter by dividing the diameter of the dispersed phase (dispersed_.d()) by the cubic-root of the aspect ratio. My question is: what is the rationale behind using the cubic-root of the aspect ratio? Sorry, for abusing the notes on the bug for an off-topic question. Keep up the great work! |
|
|
Thanks for checking the implementation against the original paper. I only have subseqent papers in which the coefficients are slightly different. Resolved by commit 1f187d6f9e58c232a1363aafe4ffe46e6f20f876 I don't know the rationale behind using the cubic-root of the aspect ratio, I would need to dig-up the papers to see if it stated there. Someone else may be able to provide you with more information on this but this bug reporting system is not the right forum for these discussions. |
|
|
Reopened to allow others to comment |
|
|
The difference between the EoH1() and EoH2() is that the 2 applies user selected aspect ratio model to calculate the hydraulic diameter, while 1 uses a hard coded Wellek model for the aspect ratio. Another slight variation of the Tomiyama lift model is used in CFX where they have modified the coefficients to make the correlation continuous: http://www.drthfrank.de/publications/2004/Shi_Frank_ICMF_2004_final.pdf neg(EoH - scalar(4))*min(0.288*tanh(0.121*pair_.Re()), f) + pos(EoH - scalar(4))*neg(EoH - scalar(10))*f + pos(EoH - scalar(10))*(-0.27); As shown in the attached image, other continuous variations more similar to the original paper could be: neg(EoH - scalar(4))*min(0.288*tanh(0.121*pair_.Re()), f) + pos(EoH - scalar(4))*neg(EoH - scalar(10))*f + pos(EoH - scalar(10.7))*(-0.2784); or volScalarField f(0.0010422*pow3(EoH) - 0.0159*sqr(EoH) - 0.0204*EoH + 0.474); neg(EoH - scalar(4))*min(0.288*tanh(0.121*pair_.Re()), f) + pos(EoH - scalar(4))*neg(EoH - scalar(10))*f + pos(EoH - scalar(10.7))*(-0.288); I would perhaps prefer the last one, but there is no reference for the last two variations. |
|
|
TomiyamaLift.png (5,784 bytes) |
|
|
I am happy to put in the form of the model your propose but it would be useful to add some text to justify the changes to the coefficients from the original paper. Could you provide the rationale that could be added to the header? |
|
|
Note that there's a typo in the variations I posted above. The upper limit mid EoH part should of course match the lower limit of the high EoH part i.e. volScalarField f(0.0010422*pow3(EoH) - 0.0159*sqr(EoH) - 0.0204*EoH + 0.474); neg(EoH - scalar(4))*min(0.288*tanh(0.121*pair_.Re()), f) + pos(EoH - scalar(4))*neg(EoH - scalar(10.7))*f + pos(EoH - scalar(10.7))*(-0.288); The rationale for the header could be something like: "The coefficient for pow3(EoH) proposed by Tomiyama (2002) has been modified to make the model continuous at EoH = 10.7 while maintaining the lift coefficient proposed by Tomiyama (2002) when EoH >= 10.7." |
|
|
Thanks Juho, I have applied your correction. |
- Anonymous
