Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
D
source
Dt
Mass transfer
A r
(u A ) ( DAB A ) S
1t4 4 2 4 4 3
14 2 43 {A
diffusion flux. source term due to evaporation
compare with the continuity D AB is coefficient kgA
r or chemical reaction [ 3 ]
of binary diffusion of A
equation g( u ) 0 m s
t in mixture A+B [m2 / s ]
D r
Using identity g(u ) it is possible to write the transport equation as
Dt t
D A A r
( u g A ) ( DAB A ) S A
Dt t
MHMT14
Mass Transfer – transport eqs.
So far everything seems to be as usual. For example exactly the same equation
(Fourier Kirchhoff) holds if you substitute T for the mass fraction A. However,
what does it mean the velocity u in the case of mixture? Individual components
are moving with different velocities and resulting mass flux (kg/(m2.s)) is the sum
of the component fluxes r r r r r r
u {
Au A BuB u Au A BuB
r r r
jA A (u A u )
r
n mass
A
flux of A
Dy A y A r ( m)
c c( u{ gy A ) (cDABy A ) S A( m)
Dt t mean molar velocity
MHMT14
1D – steady diffusion
General transport equation for steady state, constant density and DAB, without
source term reduces to
r 2
A
u g A DAB A
2
uz A
DAB
for 1D case
transport in z z 2
z-direction
For gases it is more suitable to assume constant overall pressure and to use
molar fractions (yA) instead mass fractions
r (m) y 2
yA
u gy A DAB y A
2
uz
( m) A
DAB
for 1D case
transport in z z 2
z-direction
L
MHMT14
1D – steady diffusion
Let us consider the case that u(m)=0, i.e. the same number of
molecules A is moving in one direction as the number of B molecules
in the opposite direction. uz( m) 0 z
This is so called equimolar diffusion and concentration profile is linear yA(z)
L
2 yA y A ( z ) y A1 z
0 DAB yA1
z 2 y A2 y A1 L
Different case is for example evaporation of water vapors (component A) into air
(component B). Air cannot be absorbed in water and therefore mass or molar flux
is zero (uB=0) and mean molar velocity is determined by the velocity of vapors uA
N zA
uz( m ) uzA y A (how to calculate the molar flux NzA will be shown in the next slide)
c
Equation of transport (steady state, without sources and unidirectional diffusion)
u (m) yA2
1 exp( z
z) u z( m) 0
y A yA
2
y A ( z ) y A1 DAB
u z( m ) DAB z yA(z) L
z z 2 y A2 y A1 uz( m )
1 exp( L)
DAB yA1
MHMT14
1D – steady diffusion
The mean molar velocity (convective velocity) for uB=0 and the resulting molar
flux NA follow from the definition of molar flux and the Fick´s law
J zA N zA c Au z( m ) N zA y A N zA N zA (1 y A )
y A
J zA cDAB
z
cD y
N zA AB A
1 y A z
Substituting for yA the previously calculated exponential yA profile gives
cDAB y A cDAB 1 y A2
N zA ln
1 y A z L 1 y A1
Flow around a
sphere Nu 2 (0.4 Re 0.06 Re2/3 ) Pr 0.4 Sh 2 (0.4 Re 0.06 Re2/3 )Sc0.4
Flow around
cylinder Nu (0.4 Re 0.06 Re2/3 ) Pr 0.4 Sh (0.4 Re 0.06 Re2/3 )Sc0.4
MHMT14
RTD – axial dispersion
t
L
The mean concentration in a cross-section cmA can be defined either as the area
average 1
cmA 2 c A (r )rdr
0 1
or the mass average cmA 2( u (r )cA (r )rdr ) / U
0
(both definitions are the same for the case
when velocity and/or concentrations are
uniform at cross section of pipe)
MHMT14
RTD – axial dispersion
Distribution of tracer concentration is described by the transport equation
cA cA 1 cA 2cA
u (r ) DAB ( 2 (r ) 2 )
t x R r r r x this term, axial diffusion, is in fact negligibly
small when compared with the radial diffusion
The solution cA(r,x,t) obtained in the lecture 8 assumed zero diffusion (DAB=0),
therefore purely convective transport of tracer cA 2 c A
2U (1 r ) 0
t x
5
1
L2
cmA (t ) 4(1 r )cA (r , L, t )rdr
3
2
c
2 3
0
2U t 2
Validity of this convective solution is restricted to very short times, so short that the
penetration depth of tracer is less than the radius of pipe D t R AB
MHMT14
RTD – axial dispersion
Transport equation taking into account radial velocity profile is very difficult to
solve (see later). However, for example at turbulent flow regime the velocity
profile is almost uniform and the transport equation is simplified
Colour
cA cA 2cA
injection
Ut U DAB
t x x 2
And this is called axial dispersion model ADM
x R c (x,t)
A
Initial condition at time t=0: There is Zero concentration everywhere with the
exception of origin where a unit amount of tracer was injected c A ( x, t 0) ( x)
Dirac delta
function
MHMT14
RTD – axial dispersion
It is not necessary but useful to simplify the diffusion equation by using the
transformation to a convected coordinate system (,t) moving with fluid
cA 2cA
x Ut DAB
t 2
This is a linear parabolic equation, exactly the same as the equation for unsteady
heat transfer (distribution of temperature in a plate). Only the boundary and initial
conditions are different. The analytical solution based upon infinite series of terms
Fi(t)Gi() is suitable for the case of bounded regions (e.g. a plate of finite
thickness), while in infinite regions (-,) integral transforms are preferred. In
our case we try to use the Laplace transform of time ts
f%( s) f (t )e st dt
0
c%A c% s 1
| 0 A | 0 2 A( s)
D D
1 4 2 4 3AB 1 4 2 AB4 3
derivative of general solution integral of right hand side (delta function)
MHMT14
RTD – axial dispersion
Thus determined coefficient A(s) completes the solution (expressed in the
Laplace domain), satisfying open/open boundary conditions and also initial
condition
1 DAB s
c%A ( s, ) exp( | |)
2 DAB s DAB
Next step must be the inverse Laplace transform usually performed by using
tables of Laplace transforms. In this reference you find 1 e s 1 2
L exp( )
s
t 4t
and this is almost our case, the only difference is scaling of the s-variable by a
constant DAB. There is simple rule for scaling L f ( D t ) 1 f%( s )
AB (Prove!)
DAB DAB
Using the formula for inverse transform we obtain final result, concentration at
a distance x and time t
1 2
cA (t , x Ut ) exp( )
4 DABt 4 DABt
MHMT14
RTD – axial dispersion
Comparison of impulse response of the convective model (parabolic velocity profile)
and the axial diffusion model (constant velocity U) for open/open case
DAB=0.05 m2/s 4
U=1 m/s cA
3 x2
L=1 m cmA (t )
2U 2t 3
Please, note the fact, that the
value of diffusion coefficient 2
0.05 is absolutely unrealistic, 1 ( x Ut ) 2
even at gases the molecular cA exp( )
diffusion is of the order 10-5. 4 DABt 4 DABt
1
Much greater values (e.g. 0.05)
are effective dispersion
coefficients (see later)
0
0 0,5 1 1,5 t 2
1 ( L Ut ) 2
giving similar result cA (t , x Ut ) exp( )
4 DABt 3 4 DABt
Both models of axial dispersion (open/open, closed/closed) are frequently used
in practice for modeling RTD in apparatuses like tubular reactors, packed beds,
extruders, fluidised beds, bubble columns and many others. However to match
experimental results it is necessary to use experimentally determined coefficient
De which is usually much greater than the molecular diffusion coefficient DAB
evaluated from tables or correlations. This is the same situation like with the
turbulent viscosity which is much greater than the molecular viscosity. And the
same reason: effective diffusion coefficient (called dispersion coefficient) is not a
material parameter and its value is affected by macroscopic motion - convection.
Laminar flow in pipe with nonuniform velocity profile is a good example: axial
dispersion is determined first of all by convection (by the parabolic velocity
profile). This problem was first solved by G.I.Taylor, see next slides…
MHMT14
Axial dispersion model ADM
G.I.Taylor (do you remember his analysis of large bubbles?) performed experiment
with injection of colour tracer into laminar flow in pipe.
Taylor G.I.: Dispersion of soluble matter in solvent flowing slowly through a tube. Proc.Roy.Soc. A, 219, pp.186-203 (1953)
Experimental setup consists in a long glass tube with small boring (alternatively 0.5 and 1 mm). Water flows inside the tube
very slowly (U1 mm/s) and thus perfect fully stabilised parabolic velocity profile exists in the whole tube. Diluted potassium
permanganate was used as a tracer and its concentration was evaluated visually, comparing colour in the test tube A with
color of prepared samples with precisely determined concentrations in the tube B. Flowrate was controlled by the valve N
and measured from the motion of meniscus it the tube T.
Dispersion increases
Comparison tube with the decreasing
diffusion coefficient!!!
KMnO4 Meniscus velocity
MHMT14
ADM laminar/turbulent flow
Theoretical models for axial dispersion in a pipe developed by Taylor 1953 (very
slow laminar flow), 1954 (turbulent flow) can be summarized in this way
tracer injection
Ut
x R
cA(x,t)
R 2U 2
laminar De [m2/s]
cA cA cA 2
48DAB
u De 2
t x x w
turbulent De 10.1Rv 10.1R
*
Example (corresponding to the Taylor’s experiment) R=0.0005 m, U=0.001 m/s, D AB=10-9 m2/s.
De=5e-6 (dispersion coefficient is 500 times greater than diffusion coefficient), Re=1 (laminar flow, stabilisation of
parabolic velocity profile almost immediately, at a distance from inlet less than 0.1 mm), minimum time corresponding
to equilibrations of radial concentration profile according to penetration theory 80 s (therefore axial dispersion model
can be used only for times longer than 80 seconds)..
MHMT14
ADM - restrictions
Axial dispersion model can be applied either at very high flowrates (at turbulent
flow regime) or at very small flowrates, when radial diffusion has got enough
time to equilibrate transversal concentration profile. There is a gap for
intermediate flowrates, where a numerical solution is still necessary.
Q [m3/s]
L [m]
De
R 2U 2 w Q
De 10.1R
48DAB
2s 4s 6s 8s 10s
DAB =10-6m2/s
2s 4s 6s 8s 10s
DAB =10-7m2/s R
Pipe axis
cA
DAB =10-6 m2/s DAB =10-7 m2/s
L=1m, Numerical solution
R=0.01m,
U=0.1 m/s ADM (Taylor)
Numerical solution
ADM (Taylor)
t t
MHMT14 1886–1975
Diffraction-quant. m.(24 years old)
Motion of shocks (25 years old)
Instabilities T.C. (38 years old)
…
T T 1 T (1)
Pe(1 2r 2 ) (r )
r r r
x Pe
Mass average Tm for parabolic velocity profile is the integral
1
Tm ( , ) 4 r (1 r 2 )T (r , , )dr (2)
0
Tm 1 ! !
0 0
1
Pe 2 r 4 Pe
0
2
r (1 r )(h1 ( r )) dr 0 h1
4 2 16
MHMT14
ADM - derivation
In a similar way the function g(r) is derived
g (r ) A Br 2 Cr 4 Dr 6 Er 8
The five coefficients A,B,C,D,E are selected so that the radial dependence
will be eliminated (3 equations for coefficients at r2, r4, r6),
then the normalisation condition 0 r (1 r 2 ) g (r )dr
0
24 8k x 3 k 640 3 k x 2 Tm ( , ) 4 r (1 r 2 )T (r , , )dr
0
If we repeat the whole procedure but now using the area average Tm
Tm 12 4k Tm 8k Pe2 60 17k 2Tm
Pe Tm 1
The both equations reduces to the ADM for insulated (impermeable) wall (k=0)
Tm Tm Pe2 2Tm
Pe
x 48 x 2
and this is the result
obtained by Taylor
MHMT14
ADM - derivation
Mass transfer
MHMT14 What is important (at least for exam)
Transport equation, written either in concentrations (mass or molar) or in
fractions
D A A r
( u g A ) ( DAB A ) S A
Dt t
Fick law
r
jA DAB A DAB A
r r r
jA A (u A u )
Penetration depth
DAB t
MHMT14 What is important (at least for exam)
Analogy and corresponding dimensionless criteria
D
Nu Nusselt number Pr Prandtl number
a
D
Sh Sherwood number Sc Schmidt number
DAB DAB