Process vessels

Chemical engineers who understand the fundamentals of bulk solids handling should be adept at designing moving
bed processors. After all, moving bed process vessels are basically hoppers, bins, or silos that have been modified to
allow processing of a bulk solid, e.g., heating or cooling, drying, conditioning, or providing time and space for a
chemical reaction to occur. Frequently a gas is injected either countercurrent or perpendicular to the flow of the bulk
solid. The gas may be used to remove volatiles or it may be a reactant itself. A schematic of a process vessel with
lots of bells and whistles is shown in Figure 1.

Gas Out
(Countercurrent)

Bulk Solids In

Gas In Gas Out

(Cross-Flow) (Cross-Flow)

Heat-Transfer
Gas In Plates
(Countercurrent)

Cylinder Sec?on

Hopper Sec?on

Bulk Solids Out

Figure 1. Process vessel schematic.

The key to the performance of a moving bed process vessel is uniform solids and gas flow. For the solid phase, the
usual features, i.e., mass flow and a large enough outlet, is desired. If the process requires injection of a gas, it must
flow uniformly and in a manner that does not cause channeling or other bed instabilities.
The gas can be either counter-flow or cross-flow. In a counter-flow system, the gas injection system must avoid high
gas velocities as localized fluidization can cause channeling. Hence, nozzles or perforated plates are not
recommended. Two examples of properly-designed gas distributors, in which the gas is injected into the moving
solids bed via an annulus and a set of crossbeams located near the intersection of the cylinder and hopper sections,
are shown in Figure 2 [Mehos, Chem. Engr.116, 5, 34 (2009)].
When gas flows countercurrent to the flow of solids, the gas slip velocity (i.e., the difference between the velocity of
the gas and solids) must be low enough to prevent fluidization in the cylindrical section of the processor. A good
rule of thumb is that the vessel’s cross-sectional area at the level the gas is introduced should be large enough to keep
the superficial gas velocity less than approximately one-third the bulk material’s minimum fluidization velocity.

1
Figure 2. Gas distributors with crossbeams.
The gas pressure profile is determined by adding a pressure gradient term to the differential form of the Janssen
equation (Equation 1) and applying Darcy’s law (Equation 2) to evaluate the pressure gradient:

dσ v k tan φ ʹ dP
+ σ v = ρb g + (1)
dz RH dz

dP uρ g
=− s b (2)
dz K
where σv is the solids stress, k is the Janssen coefficient, ϕ’ is the wall friction angle, ρb is the bulk density, g is
acceleration due to gravity, P is the gas pressure, K is the permeability, us is the gas slip velocity, and z is the distance
from the top of the solids bed. These equations must be solved numerically using a known value of the solids stress
(generally equal to zero at the top of the bed) and the gas pressure (not necessarily zero at the top of the column)
since both the bulk density and the permeability are functions of solids stress. Because the gas density changes with
pressure, the slip velocity will also change within the column. Equations 8.14 and 8.15 can be used to determine the
required gas pressure at the inlet of the processor and the solids stress at that level.
If an annulus and crossbeams are used for gas injection, they should be sized to ensure that the solids stress on the
components is non-zero. It is best to have Jenike & Johanson size these components. On the other hand, this may
not be as great an issue if only an annulus is used to inject the gas into the solids (i.e., no crossbeams – this is
Zeppelin’s design).
If the gas flow rate is too low, a driving force for mass transfer will not exist in a portion of the column. The
minimum gas rate can be found from a mass balance in which the concentration of the volatile species in the gas
leaving the top of the column is in equilibrium with the incoming solids. This is analogous to specifying a high
enough gas rate to avoid pinching in strippers that handle liquids and gases. A filter is recommended at the top of the
column to capture any entrained solids.
Cross-flow designs are preferred if the required gas flow rate is high because the pressure drop will be lower. Cross-
flow processors are fabricated with permeable walls, such as those manufactured by Young Industries through which
the gas enters and exits the processing vessel perpendicular to the flow of solids. Moving bed cross-flow processors
having a circular cylinder section are designed with two (or more) permeable annuli. Gas is typically fed into a
permeable inner cylinder, travels through a bed of solids that are flowing inside the annulus, and then exits through
an outer permeable wall. Figure 3 is a schematic of a radial gas flow moving bed processor (see
http://jenike.com/13778-2/).
For cross-flow designs, the gas velocity must be low enough to prevent cavities or pinning. A cavity can develop if
the pressure gradient that develops when gas is injected into the bed causes a gap to form between the bulk solid and
the wall from which the gas is introduced. If a cavity forms, gas will flow preferably upward rather than across the
bed. Introducing solids through a rotary valve may reduce cavity formation.
2
 Solids in

Gas out
Gas
permeable
walls

Gas in

Solids out

Figure 3. Radial flow moving bed reactor.

Pinning can occur on the opposite wall. Pinning happens when the friction between the bulk solid and the permeable
wall through which the gas exits is high enough to prevent the particles from flowing downward along the wall. A
periodic pulse of reverse air can be used to prevent material from building up on the exit walls.
Moving bed processors generally have residence time requirements. The residence time of a bulk solid in a process
vessel is related to its feed rate by

z ρb A
t= (3)
m! s

! s is the solids mass flow rate, z is the distance traveled from the top of the
where t is the time in the processor, m
column, A is its cross-sectional area, and ρb is the bulk density of the solids. The product Az is equal to the volume of
the moving bed in the cylinder. For some applications, the volume of the hopper section should be included in the
residence time calculations.
Purge columns and gravity dryers
In a purge column or gravity dryer, the mass transfer driving force must be known. The driving force is equal to the
difference between the volatiles content of the solid particles and that which is in equilibrium with the bulk gas.
Phase equilibrium data are therefore needed to design purge vessels and gravity dryers. Usually, determining the
phase equilibria is much less challenging than gathering kinetics information.
Kinetics within solid particles are frequently described by Fick’s law. Fick’s second law for spherical particles is
given by

∂x 1 ∂ ⎛ ∂x ⎞
= 2 ⎜ Deff r 2 ⎟ (4)
∂t r ∂r ⎝ ∂r ⎠

where x is the mass fraction of the volatile component in the solid, Deff is the effective diffusion coefficient of the
volatile species in the solid particle, and r is the radial coordinate. Often, the effective diffusivity is expressed as the
product of the molecular diffusivity, which is usually known, and what’s called the porosity-tortuosity ratio, which is
3
rarely known and cannot even be estimated with confidence. It’s best just to determine the effective diffusivity by
experiment.
Initially, the mass fraction of the volatile component is equal to xinit, i.e.,

x(r,0) = xinit (5)

Boundary conditions are:

∂x(0,t)
=0 (6)
∂r
x(ro ,t) = x∞ (7)

where x∞ is the mass fraction of the volatile component in the solid in equilibrium with the bulk gas. Equation 6
describes symmetry at the center of the particle. The mass fraction at the solid surface is only equal to x∞ if there is
no resistance to mass transfer in the gas phase, which is not always the case. If the gas-phase resistance cannot be
dismissed, Equation 8 is used to describe the boundary condition at the surface:

∂[x(ro ,t)]
−Deff = k x (x − x∞ ) (8)
∂r
where kx is the convective mass transfer coefficient. In principle, expressions for mass transfer in the gas phase
should be based on a gas-phase driving force, but for convenience, the difference in solids mass fractions is often
used. If a linear relationship exists, i.e., y = mx, where m is the slope of the equilibrium line1 and y is the mass
fraction of the volatile species in the gas phase,

k y ( y − y∞ ) = k y (mx − mx∞ ) (9)

and therefore kx = mky, where ky is the mass transfer coefficient based on the gas phase driving force (y - y∞).
Whether or not gas-phase resistance is important can be inferred from the Biot number Bi, which reveals the relative
contributions of the resistances to mass transfer due to diffusion in the solid particle and convection in the gas phase.
The Biot number is a dimensionless group given by

k x ro
Bi = (10)
Deff

A Biot number significantly greater than 1 signifies that mass transfer is limited by diffusion in the solid phase. Low
values, i.e., Biot numbers less than approximately 0.1, indicate that the rate is limited by convection in the gas phase.
For systems having small Biot numbers, the concentration of the diffusing species is nearly constant and the kinetics
can be described by

dx
= −k ʹ(x − x∞ ) (11)
dt
where x is the mean volatiles content (mass fraction) of the particle, and

3k x (1− ε )
kʹ = (12)
ro

1
m is related to the Henry’s law constant (or the Hank’s law constant if we want to be less formal).
4
The 3/ro factor arises from a surface area per unit volume calculation. Analytical expressions for the solution to the
diffusion equation for Bi appreciably greater than 1 or less than 0.1 respectively are:

x − x∞ ⎛ −n 2π 2 D t ⎞
6 ∞ 1
= ∑ exp ⎜⎜ r 2 eff ⎟⎟
xinit − x∞ π 2 n=1 n 2
Bi >> 1 (13)
⎝ o ⎠
and

x − x∞
= exp(−k ʹt) Bi < 0.1 (14)
xinit − x∞

For the dreaded scenario in which the Biot number is neither very large nor very small, the following approximate
solution can be used:

x (t) − x∞ 3A1 ⎛ D t⎞
= 3 (sin λ1 − λ1 cos λ1 )exp ⎜⎜ −λ12 eff2 ⎟⎟ (15)
xinit − x∞ λ1 ⎝ ro ⎠

where

4sin( λ1 ) − λ1 cos( λ1 )
A1 = (16)
2λ1 − sin(2λ1 )

and λ1 is the first eigenvalue, which for spheres is the root of the equation

λ1
1− = Bi (17)
tan λ1

Equation 815 is the truncated form of an infinite series solution to the diffusion equation with a mass transfer type
boundary condition. It holds for residence times greater than ca. 0.2ro2/Deff.
In a purge column, the surface volatiles content is not constant but instead varies with position inside the cylinder.
Hence, Equations 13 through 15 cannot be applied to moving beds. However, the equations can be used to regress
the parameters Deff, k’, and kx from batch stripping data, which can be collected by passing a gas stream through a
short bed of powder and measuring its volatiles content over time. The Sauter mean particle size should be used to
calculate ro. The tests should be conducted over a range of bulk densities to match the range expected in the column.
Figures 4 through 5 illustrate how the parameters are determined by regressing batch stripping data. When data are
plotted on a semi-logarithmic scale, regression of the linear portion of the data will yield a line that intersects the
vertical axis at 6/π2 if mass transfer is limited by diffusion, 1 if mass transfer is limited by convection in the gas
phase, and a value somewhere in between if resistances due to diffusion and convection are both significant. Hence,
inspection of the intercept allows insight to which mass transfer resistances dominate.

5
 1
6

π2



π 2 Deff t
0.1
Slope = −
ro2


x − x∞

x0 − x∞

0.01






0.001
0 0.1 0.2 0.3 0.4 0.5
Time

Figure 4. Typical batch devolatilization data when Biot Number >>1.

3A1 1
(sin λ1 − λ1 cos λ1 )
λ13





0.1


x − x∞

x0 − x∞
λ12 Deff t
0.01 Slope = −
ro2





0.001
0 0.1 0.2 0.3 0.4 0.5
Time

Figure 5. Typical batch devolatilization data when Biot Number > 0.1 and < 1.

1






Slope = -k’
0.1


x − x∞

x0 − x∞

0.01






0.001
0 0.1 0.2 0.3 0.4 0.5
Time

Figure 6. Typical batch devolatilization data when Biot Number < 0.1.
The parameters obtained by regression of batch data can then be used to model continuous columns. Once the
kinetics parameters have been determined, the transfer equations are solved numerically using a mass balance to
track the volatiles in the gas stream. The column is partitioned, each segment having gas and solids streams entering
6
from or exiting to adjacent segments. Unless Bi < 0.1, the diffusion equation is integrated numerically over the
residence time of each segment to determine the concentration profile of the particles leaving the segment, and a
material balance is used to determine the volatiles content of the gas entering the segment. The solids concentration
profile is then used as the initial condition for solving the diffusion equation for the next segment. The required
column height or total residence time is determined iteratively until the volatiles content of the solids leaving the
column is equal to the target.
Note that the above analysis assumes spherical particles, which is often not the case. Solids fed into a process vessel
can have a variety of shapes, and engineering judgment must be used to assess the results of analyses that assume a
spherical geometry. Other shapes can be assumed, but a characteristic length or diameter may not be immediately
apparent. Given the choice between calculating the volume and area of an ellipsoid or waterboarding, enemy
combatants residing in Gitmo would probably chose the latter. In addition, the characteristic radius may be that of a
collection of particles. For this reason, it is often convenient to use Deff/ro2 as a lumped term.
When the Biot number is less than 1, a purge column can be designed by methods analogous to those used to design
packed columns for mass transfer between gases and liquids. The column height Z is equal to the height of a transfer
unit Hs times the number of transfer units Ns:

Z = Hs Ns (18)

The height of a transfer unit is calculated from

m! s
Hs = (19)
ρb k x as Ax
where Ax is the cross-sectional area of the column and as is the specific surface area of the bulk solid (i.e., the surface
area per unit volume of the solids bed). For spherical particles, the specific surface area is given by

3(1− ε )
as = (20)
ro

where ε is the void fraction of the solids bed. Average values of ε and kx are used.
The number of transfer units Ns is given by
xout
dx
Ns = ∫ (21)
xin x *−x

where x * is the mass fraction of the volatile compound in the solid phase that is in equilibrium with bulk gas. The
subscripts in and out denote the top and the bottom of the column, respectively.
If the gas fed into a purge column is free of volatiles, as is typical, and the phase equilibrium relationship is linear,
the number of transfer units can be calculated from

⎡⎛ x ⎞ ⎤
ln ⎢⎜ in ⎟ 1− A
⎢⎝ xout ⎠
( )+A ⎥
⎥
NS = (22)
(
1− A )
where the absorption factor A is given by

7
 m! s
A = (23)
mm! g

! g is the gas mass flow rate.

where m

The analysis may feel nostalgic for chemical engineers who have designed strippers or absorbers. For purge
columns, the stream flowing downwards is a bulk solid rather than a liquid, and the column cross-sectional area is set
by bed stability rather than hydraulic considerations. For both liquids and solids, the gas requirement is set by phase
equilibria.
The analysis above applies only to purge columns where trace amounts of volatiles are removed from the solids. For
moving bed dryers where considerable drying is required, both heat and mass transfer must be considered, which can
make the analysis quite challenging.
In general, drying initially proceeds at a constant rate, but once the moisture content reaches a critical value, the
drying rate begins to decline. This critical moisture content is a function of the wet bulb temperature of the drying
gas if the gas is air.
A much simpler method to model drying kinetics is a semi-empirical approach in which the constant rate period is
described by a convective heat transfer model and an empirical fudge factor is applied when the moisture content of
the solid falls below its critical value [Satija, “A Scale-up Study of Nozzle Spray Dryers”, Drying Techn., 5, 1, 63
(1987)]:

dX
Δ H lv ρb = −hasv (T − Twb ) • f (24)
dt
where ΔHlv is the latent heat of vaporization, X is the moisture content (dry basis), T is the temperature of the drying
gas, Twb is its wet bulb temperature, and f is a function defined as:

f = 1 for X ≥ X eq (25a)

q
⎛ X−X ⎞
eq
f = ⎜⎜ ⎟ for X > X cr
⎟ (25b)
X
⎝ cr − X eq ⎠

where Xeq and Xcr are the equilibrium and critical moisture contents, respectively, and q is an empirical parameter.
The parameters Xcr and q are determined from batch drying tests. The equilibrium moisture content Xeq is determined
by exposing a sample of dry bulk material to an environment with a controlled relative humidity and measuring its
moisture uptake or from moisture desorption isotherm data.
Moving bed heat exchangers
The analysis of direct contact moving bed dryers, in which a gas is injected into a moving bed of solids to change the
streams’ temperatures, is similar to purge columns except that heat transfer is modeled instead of mass transfer. The
Biot number for heat transfer is

hro
Bi = (26)
k

8
where h is the heat transfer coefficient, and k is the thermal conductivity of the particle.
If Bi < 0.1, which is often the case, the solid particle temperature profile is nearly uniform, and volume requirements
for direct contact bulk solids heat exchangers can be readily determined. For countercurrent heat exchangers, a gas
rate that gives a suitable approach temperature, i.e., the difference between the temperatures of the solids entering
and the gas leaving the column, is specified. The volume V needed to provide the required heat transfer can then be
calculated from:

Tgout − Tsin ⎛ 1 1 ⎞⎟
ln = −hassv ⎜⎜ − ⎟V (27)
Tgin − Tsout ⎝ m! g C Pg m! sC Ps ⎠

where CPg is the specific heat of the gas, Tgin and Tgout are the average inlet and outlet gas temperatures,
respectively, and Tsin and Tsout are the average solids inlet and outlet temperatures, respectively.

A rather painful analytical expression can be used to calculate the temperature profile of the solids leaving the cooler
[Almendros-Ibáñez et al., App. Therm. Engr., 31, 1200 (2011)]:
j k
⎛ A xha ⎞ ⎛ A Hha ⎞
⎜ x s⎟
⎜ z s
⎟ (28)
⎛ ⎞ ∞ ⎜ m! C ⎟
⎝ m! sC Ps ⎠
j
Ts − Tsin A Hhas Ax xhas
= 1− exp ⎜⎜ − z − ⎟ ∑ ⎝ g Pg ⎠ ∑ k!
Tgin − Tsin ⎝ m! sC Ps m! g C Pg ⎟⎠ j=0 j! k=0

where Ax and Az are the side and plan cross-sectional areas, respectively, H is the height of the heat transfer section of
the cylinder, and x is the horizontal distance from the gas entry.
An indirect contact moving bed bulk solids heat exchanger usually consists of a rectangular cylinder section
containing plate-and-frame heat-transfer plates above a mass flow hopper. Solex Thermal Science is the authority of
bulk solids heat exchangers.
If the heat exchanger acts as a cooler, a sweeping gas is sometimes introduced into the bed to prevent condensation
by keeping the vapor above its dew point. Condensation can lead to material adhering onto the plates and plugging
up the heat exchanger.
A modified version of Fourier’s second law is used to describe heat transfer in an indirect contact bulk solids heat
exchanger:

∂T ∂2T
ρbusC Ps = −keff 2 (29)
∂z ∂x
with boundary conditions

Ts (x,0) = Tsin (30)

b
Ts ( , z) = Tw (31)
2
∂[T (0, z)]
=0 (32)
∂x
where us is the solids velocity, Ts is the solids bed temperature, Tw is the wall temperature, keff is the effective thermal
conductivity of the bulk material, x is the distance from the centerline between adjacent plates, b is the plate spacing,

9
and z is the vertical distance from the top of the heat exchanger. If there is additional heat transfer resistance at the
heat exchanger walls, the boundary conditions given by Equation 31 can be replaced with

∂[Ts ( b2 , z)
−keff = h[T ( b2 , z) − T f (z)] (33)
∂x
where Tf is the temperature of the heat transfer medium.
The solids velocity can be calculated from the solids feed rate by

m! s
us = (34)
N ρb Azb

where is the plan cross-sectional area between adjacent plates, and N is the number of channels.
Note that the thermal conductivity used in the analysis is an effective conductivity. Be wary of mixing rules that
some investigators use. It is best to measure thermal conductivity directly. The specific heat can be assumed to be
that of the pure solid since the thermal mass of the gas in the voids of the bulk material is paltry.
Because the wall temperature is variable, the system of equations must be solved numerically to determine the height
of the heat exchanger that gives the desired solids exit temperature. The fluid side heat transfer coefficient must be
known, and the calculated height depends on the width, spacing, required duty, and number of heat transfer plates.
The temperature of the bulk solids exiting the heat transfer plates will not be uniform as it will have a parabola-like
profile. The profile can be flattened by using banks of heat exchanger plates in which the downstream plates are
offset from the upstream plates as shown in Figure 7. Because each stream leaving a channel will be split and mixed
with a stream exiting an adjacent channel, the parabolic temperature profile will be converted into one that represents
a lower case omega. This is not surprising considering all the Greek letters in the energy balance equations.
b
-b/2 b/2
x

H z

T
Tf

Figure 7. Banks of heat-exchanger plates.

10
Moving bed reactors
Moving bed reactors are reactors in which a bulk solid is continuously fed into the top of column and removed from
the bottom. A gas is added such that it flows either co-current, countercurrent, or radially through a moving bed of
solids. Often, the solids are used as a catalyst.
If the reaction is exothermic, heat transfer plates positioned radially can be used. The hopper section beneath the
reactor must be designed for mass flow to prevent the formation of stagnant regions.
Fixed bed reactors are more common than moving beds, especially for catalytic reactors. A fluid is passed either
axially or radially through the catalytic bed, and once the catalyst has become spent or fouled, the reactor is shut
down, and its contents are replaced with fresh catalyst.

11