|
|||||||||||||||||
43901-AC9
|
|
where fDi is the total fluid-particle drag per unit volume of suspension on type i particles, and ¦¤Ui is the average velocity of type i particles relative to the fluid. The friction coefficient bij, owning to the linear characteristics of the low-Re flows, is a function of particle volume fraction fi, particle size di, and a length scale ¦Ë on which the lubrication force between particles begins to saturate.
As bij is independent of the velocities ¦¤Ui, one can set ¦¤U1 = 0 and ¦¤U2 °Ù 0 to determine b12 and b22 from the numerically obtained drag forces fDi, and then ¦¤U1 °Ù 0 and ¦¤U2 = 0 to determine b11 and b12. bij is symmetric due to the principle of acting and reaction between the two particle species; furthermore, the row summation of bij yields the friction coefficients bi = fDi-fixed/¦¤U for bidisperse fixed beds, because Eqn. (1) must be able to recover the fixed-bed drag when ¦¤U1 = ¦¤U2.
We first revisited the drag in bidisperse fixed beds. Because this problem has been studied by van der Hoef et al. (2005, 2007), it is an excellent opportunity for comparison and validation. Moreover, accurate determination of bi is important for the development of a general drag law for suspensions with particle-particle relative motions because of the connections between bij and bi. Our simulation data, obtained in the total volume fraction range of 0.1 < f < 0.4, with volume fraction ratio f1:f2 varied from 1:1 to 1:7 and particle size ratio d1:d2 varied from 1:1 to 1:4, agree very well with the existing data in van der Hoef et al. (2005, 2007). The extended parameter range in our simulations enables us to confirm the deficiencies of the existing drag law. A new drag law for fixed beds containing spheres of different sizes is proposed
|
where yi is the dimensionless particle size normalized by the Sauter mean <d>
|
with ni the number density of type i particles, and a is given by the following cubic polynomial
|
The dimensionless drag force is related to bi by
|
where µ is the viscosity of the fluid. This drag law, with an average error of 3.9% and maximum error of 9.4%, is more accurate than current drag laws for bidisperse fixed beds.
We then characterized the drag forces in suspensions with particle-particle relative motions. As bij is symmetric and the row summation of bij yields bi, it can be simplified to
,
where b12, the off-diagonal components representing the hydrodynamic particle-particle drag, is the only free parameter to be characterized. Our simulations show that b12 is always an important contribution to the net fluid-particle drag, and can be well approximated by a linear function of the harmonic mean of b1 and b2
|
where ¦Á12 is a logarithmic function of the lubrication cutoff
(assuming d1
< d2).
This logarithmic dependence indicates that the primary contribution to the hydrodynamic particle-particle drag is made by the normal lubrication force between dissimilar particles. Eqns. (6) and (7) have an average error of 13% and a maximum error of 31%. Both Eqns. (2-5) and Eqns. (5-6) can easily be generalized to polydisperse suspensions containing more than two particle species. In the next year, we will extend these drag laws to suspensions with finite Reynolds numbers.
References: