Documente Academic
Documente Profesional
Documente Cultură
Feature
Cover Story
Report
Mixing Systems:
Design and Scale Up
Many options are available to meet
the mixing challenges confronted by the CPI.
Both theoretical and empirical methods can help
the engineer to fit the right system to the task
Werner Himmelsbach, David Houlton,
Wolfgang Keller, Mark Lovallo
EKATO Mixing Technology
F
undamental to the chemical pro- subject in itself and has had
cess industries (CPI) — whether many significant advances in re-
specialty or bulk chemicals, cent years. Rather than give a very ab- FIGURE 1. This 1900 kW agitator-drive
pharmaceuticals, food products, breviated summary here, we refer the unit is for a continuously operated bulk-
minerals processing, environ- reader to Reference [2] for this topic. chemicals reactor
mental protection or other products or In addition to agitator design and
activities — is the need for mixing. The power requirements, which are fun- where c is the concentration after an
wide variety and complexity of mix- damental to mixing systems, many infinite period of time. The same rela-
ing tasks encountered in industrial other considerations also play a part tion also holds for temperature homo-
applications require careful design in maximizing performance. These geneity. A standard target homogene-
and scale up to ensure that effective considerations include mechanical ity is 95%. The blend time needed to
mixing is achieved efficiently. Designs aspects, seal selection, materials of achieve a higher homogeneity can be
based on a small range of traditional construction and surface finishes to calculated from the equation for stan-
agitators are no longer economically prevent fouling or aid cleaning. Figure dard blending time since inhomogene-
acceptable. Modern impellers and the 2 outlines the process and factors in- ities generally decline exponentially:
use of physical or computer modeling volved in designing a suitable system.
tm, x = tm,95 ⋅ ln(1 − x) / ln(0.05)
can greatly enhance performance and (2)
reduce costs. BLENDING OF
Mixing tasks fall into six main cat- MISCIBLE LIQUIDS where x is the desired degree of ho-
egories: 1. blending of miscible liquids; During processing, inhomogeneities mogeneity. When comparing the per-
2. blending of mixtures with “difficult” of concentration or temperature often formance of blending equipment, it
rheologies (such as non-Newtonian arise. This typically happens dur- is particularly important to compare
properties); 3. suspension of solids; ing process steps such as addition of both the blending time and the degree
4. liquid-liquid dispersions; 5. heat chemicals, mass transfer, heat trans- of homogeneity achieved.
transfer; and 6. gas-liquid dispersion. fer and chemical reaction. Inhomoge-
Different mixing behaviors and rules neities lead to non-uniform processing Turbulent blending
govern each basic mixing task. To op- and can negatively impact product The dimensionless blend time charac-
timize a design, or to scale-up reliably, quality. The objective of blending is to teristic (Ntm) is constant for geometri-
these behaviors and rules need to be maintain the required degree of homo- cally similar, agitated, baffled vessels
understood and defined. Complex geneity. that operate in the turbulent-flow re-
tasks that involve two or more of the gime. Appropriate baffling is always
above categories require special at- Degree of homogeneity required in the turbulent regime,
tention. The controlling task must be The homogeneity at a given time, M(t), not only to achieve efficient blend-
identified to determine the design and is given by the change in concentra- ing, but for all mixing tasks. Without
scale-up rules to be applied. tion of a component from c0 to c(t) with proper baffling, fluids will tend to
This article addresses the first five time: rotate in the vessel and blend times
mixing tasks that are listed above. c0 − c(t) will increase. The value of Ntm de-
M (t) =
Gas-liquid dispersion is a complex c0 − c∞ (1) pends mainly on the type of impeller
46 CHEMICAL ENGINEERING WWW.CHE.COM APRIL 2006
����������������������
�������������������
NOMENCLATURE
A m² heat transfer area Nu - Nusselt number
�������������
���������
cw kg/kg concentration Nu = hPT/kP
������� ��������
�������� (mass fraction) P W impeller power (transmit-
cv m³/m³ concentration ted to fluid)
�����������������������������
(volume fraction) Po - impeller power number,
C - impeller specific factor Po = P / ( ρ N3 D5 )
����������� used in Equ. (12) Pr - Prandtl number Pr = µc/k
�������������������������� ����� and (17)
��������� Psettle W settling power
D m impeller diameter
q m³/s impeller pumping rate
dp m particle size
������������������ ���������� Q W rate of heat
fc - correction factor for hin-
dered-settling velocity Re - impeller Reynolds number,
Re = N D²ρ / µ
g m/s² acceleration due to
���������� s - Zwietering constant, impel-
�������� ���� gravity
H m vessel filling height ler specific
�����������
��������������
hP W/m²K heat-transfer coefficient, T m vessel internal diameter
�����������
������������������������
���������������� process side tm s mixing time
���������������
��������������
��������������
kMO - Metzner-Otto constant V m3 volume of fluid
������������� k, kp W/mK thermal conductivity, pro- vs,sh m/s settling / hindered settling
��������������� ���� cess side velocity
K Pa·sm viscosity “consistency v(r) m/s local velocity at radius r
����������� factor” x - degree of homogeneity
���
���������������������� ��������������� m - viscosity “flow index”
�����������
z - number of recirculations to
���������
Mt N·m shaft torque achieve homogeneity
���� M(t) - homogeneity (time) µ, µW Pa·s mixture viscosity, at wall
����������������������� njs s-1 minimum shaft speed for
������������� m²/s kinematic viscosity
suspension
ρl,s kg/m³ density liquid, solid
FIGURE 2. Many factors are involved N s-1 agitator speed
γ s-1 local shear rate
with designing an agitator system for a Ntm - dimensionless
given application mixing time τ0, w Pa yield stress, stress at wall
and the diameter ratio. Mersmann [1] necessary to circulate the vessel vol- is a function of the viscosity of the
evaluated the measurement results ume to achieve a given homogeneity. liquid and the average specific power
of various authors and determined a The blend time is calculated from input. Turbulence cannot help to mix
simple correlation that characterized the assumption that vessel contents the reactants on a scale smaller than
the performance of many mixers, with will be homogeneous after they have the micro-vortex. Mixing down to the
H/T = 1 and a single-stage impeller, been recirculated z times by the mixer, molecular scale therefore relies on
in terms of the impeller power number where z is usually taken to be 4 or 5: molecular diffusion of the chemicals
and ratio of impeller to vessel diam- within the micro-vortices.
tm = z . V (5)
eter. In dimensionless form his equa- q The time for micro-mixing to occur
tion becomes: where V is the total liquid volume is, in most cases, not significant com-
−5 / 3 in the vessel, and q is the impeller pared to the macro-blend time. For
N tm = 6.7 ⋅ D ⋅ Po −1/ 3 (3) discharge or pumping rate, which example, an aqueous solution might
T
can be measured or estimated by cal- typically have micro-vortices of 36 µm
Our own measurements show that culation. This theoretical method is and a diffusion rate of approximately
the above equation forms a good basis not as reliable as the dimensionless 2 x 10-9 m2/s, so the micro-mixing time
for design calculations for impellers mixing time method, embodied by would be about 0.08 s.
with large diameter ratios (D/T > Equations (3) and (4), which is based Engineers should, however, always
0.5), whereas the following equation on direct measurements and scale- consider whether micro-mixing might
is more accurate for predicting blend up rules. The discrepancy arises be- be an issue. In some cases it can be
times for axial-pumping impellers cause mixing is not just a function of decisive, such as when competing con-
with diameter ratios in the range of the main impeller discharge flow, but secutive reactions occur extremely
0.1 to 0.5. Accuracy is about ±10% [2]: also of the flow patterns generated. quickly.
−5 / 3
N tm = 5.5 ⋅ D ⋅ Po −1/ 3 (4) Mixing for chemical reactions VISCOUS AND
T
For a chemical reaction to proceed, it NON-NEWTONIAN FLUIDS
is necessary to intimately mix the re- For high-viscosity liquids, the mixing
Circulation rate actants down to the molecular scale. regime changes from one in which
When comparing the mixing perfor- According to turbulence theory [3], turbulence dominates to one in which
mance of different agitators, many however, there is a theoretical mini- viscous drag forces dominate, and
users and suppliers discuss the con- mum size of vortex, called a micro-vor- agitation throughout the bulk is by no
cept of “circulation rate,” which ex- tex, that is generated in a turbulent means uniform. This non-uniformity
presses the number of times that it is liquid. The size of the micro-vortices can be made significantly worse if the
CHEMICAL ENGINEERING WWW.CHE.COM APRIL 2006 47
Cover Story
� �����������
high-viscosity mixture also exhibits tional to the shaft speed. �
the anomalous flow properties of non- γ =k MO⋅ N (9)
Newtonian rheology.
���� ����
The most frequently encountered where kMO, the Metzner-Otto
anomaly is shear thinning, where vis- constant, depends on the im- ����
•
cosity decreases with increasing shear peller design. Using the im- ��
•
��
•
rate according to: peller shear rate, the local vis- �������������������
cosity can be calculated from ��������������������
m− 1 μ ��� μ ���
Local viscosity, µ = K ⋅ γ or (6) Equation (6) or (8). The local ������������γ������
•
���� ���������������������������������
����������
������
�������������������������
����������������������������������
���������������������������������
�����
���
��
� �� ��� ����� ������ ������� ��
FIGURE 5. The blending time is dependent on agitator type FIGURE 6. Pilot plant testing of high-viscosity slurries,
such as this mineral-ore slurry with a yield stress, is highly
recommended for good mixer-design performance
ture is stretched and folded, and is also both turbulent- and laminar-flow re- powers can be correspondingly very
sheared local to the impeller blades. gimes, the amount of liquid displaced large. It is not possible to accurately
Once sufficient stretching and folding by the agitator is proportional to the predict flow behavior in such a system
has taken place, local molecular diffu- number of times it turns and the circu- by using theoretical correlations based
sion completes the micromixing. lation rate of the impeller. Hence, the on solids concentration and particle-
The optimized ribbon has a simpler blending time for a large-scale plant is size data. For high-viscosity slurries,
construction than the helical ribbon calculated from the number of impel- tests with original product are indis-
and is suitable for a wider range of ler revolutions found to be necessary pensable to ensure good performance
viscosities due to its flexibility. This in the pilot scale. In fact, larger ves- without overdesign (see Figure 6).
optimized impeller can be fitted with sels will generally have higher Reyn-
inner baffles that allow it to maintain olds numbers than smaller vessels SUSPENSION OF SOLIDS
good mixing performance even at high containing the same mixture. The rule There are various objectives when
Reynolds numbers (low viscosities). is, therefore, somewhat conservative. using an agitator for suspension tasks.
Helical ribbons are poor mixers in In washing processes or for solids dis-
non-laminar-flow conditions because High-viscosity slurry processing solution, off-bottom suspension can be
the viscous drag of the mixture on the Slurries that contain a high-solids sufficient. In a process with continu-
walls becomes insufficient to prevent concentration of small particles can ous operation or where several vessels
bulk rotation. Figure 5 compares the exhibit non-Newtonian flow behav- are arranged in series, more-stringent
dimensionless-blending-time, Ntm, ior. Such slurries are commonly en- specifications must be made for the
performance of several open and near- countered in the mineral-processing homogeneity of the suspension.
wall impellers over a wide range of industry, where typical solids volume Figure 7 illustrates three degrees of
Reynolds numbers. concentrations are 35 to 50%. In these suspension. To progress from one de-
The rule that the dimensionless applications, the settling velocities gree to the next requires an increase
mixing time is constant in geometri- of the suspended solids are very low in power input via higher agitator
cally similar systems applies not only and the key mixing task becomes the speeds. In the first case, “on-bottom
for turbulent but also for laminar blending of the highly-viscous, non- motion,” there may be temporary local
systems, when a positive-displace- Newtonian mixture, rather than solids deposition of solids on the base, with
ment impeller is used. To explain this suspension. Tanks for such ore slurries occasional movement of the solids
phenomenon, one can consider that in can be up to 10,000 m³, and agitator — large clear zones can occur in the
CHEMICAL ENGINEERING WWW.CHE.COM APRIL 2006 49
Cover Story
FIGURE 7.
upper part of the vessel. In the second The degree of
case of “off-bottom suspension,” no suspension
desired for a
particle comes to rest for longer than given applica-
one second on the bottom of the ves- tion depends
sel. This criterion, often used for large- on the objec-
scale, mineral-processing applications, tive. Here,
three levels of
defines the lowest mixing power at suspension are
which the entire surface area of the shown, each
particles is exposed to the liquid phase requiring more
for chemical reaction or mass transfer. power input
Zwietering [5] produced the following than the one
before
well-known correlation for minimum
shaft speeds, njs, at which “off-bottom
suspension” occurs:
0.45
g ⋅ ( ρ s − ρl ) FIGURE 8. To
��������������������������������������
��
n js = s ⋅ v0.1 ⋅ ⋅ d p0.2 ⋅ cw0.13 ⋅ D −0.85 �� �� ��������������
ρl describe the
���������������������
�� complex nature
(13) ��� ���������������������
of suspension
The third case is “visually uniform �
� tasks, suspen-
�� ����������
suspension,” which is defined as hav- � sion duties
� � �������� �� �����
ing no large clear zones. Settling of �������� are classified
� � � �� into four broad
coarse particles may still occur for sol- � ��� groups
�
ids with a wide range of particle sizes. ��� � ���
Occasionally, for very demanding ��� � � ���
��� � ������� � �����
duties such as continuous-overflow op- � �������� � �����
eration, or manufacture of dispersions ��� ��������
�
� �����
as end products, “uniform suspension” ����
��� � � � ��� � � � ��� �
is required. In this case, particle con-
��
����������������������������������������
centration and size distribution are
uniform throughout the vessel. This
is very difficult to achieve unless the If it is assumed that all solid particles value where between 80 and 90% pass
particle settling velocity is very low. in the liquid are distributed uniformly, through the mesh size.
and all simultaneously begin to settle
Design fundamentals under the effect of gravity, they re- Scale up
In addition to agitator parameters and lease a “settling power”, which can be Scale up of suspension duties can be
the vessel geometry, the properties of quantified by the relation: very complex. Various scale up criteria
both the liquid and the solid particles have been proposed based on the type
influence the fluid-particle hydrody- Psettle = vsh ⋅ cv ⋅ ∆ρ ⋅ g ⋅ V (15) of suspension needed, as discussed
namics and, thus, the suspension. The earlier. Specific process or product re-
important physical properties for agi- where is the difference in density quirements can impose additional cri-
tator design are: the liquid density, the between solid and liquid. In order to teria for consideration. Some common
density difference between solids and maintain a defined degree of unifor- complications include these:
liquid, the liquid viscosity, the average mity in the suspension, the agitator • Solids with extremely wide particle
particle size and the volumetric con- must provide a power input to the size distributions — the fine par-
centration of the solids. liquid that counteracts this settling ticles affect the suspension of the
A single particle’s free-settling veloc- power. The agitator power always large particles
ity, vs, is calculated by methods given amounts to a multiple of the settling • Very high-solids concentrations —
in the relevant literature. The hinder- power. particle interactions affect the ap-
ing effect on the settling process due When one is using the above Equa- parent rheology
to the presence of several particles is tions (14) and (15), the choice of par- • Presence of small amounts of ex-
quantified by the following relation, ticle size that is used to calculate the tremely large particles — impos-
where the exponent m is a function free-settling velocity, vs, is very impor- sible to suspend but must be moved
of the particle Reynolds number, and tant. In powders or slurries, the indi- around on the base of the vessel
varies between 2.33 and 4.65: vidual particles vary in size and shape. • The presence of significant quanti-
Choosing the largest particle size can ties of extremely small particles
vsh = vs (1 − cv ) m (14) result in a much higher agitator power — these essentially behave as part
where vsh is the hindered settling ve- than is required. From experience, re- of the fluid
locity, vs is the free-settling velocity liable results are obtained with a de- To accommodate these considerations,
and cv is the volume fraction of solids. sign particle size that corresponds to a solid-suspension duties are generally
50 CHEMICAL ENGINEERING WWW.CHE.COM APRIL 2006
To achieve homogeneous suspension separation of the phases afterwards.
of true suspensions, high-efficiency, Mixer-settler systems can generally
axial-pumping impellers are gener- use finer dispersions than solvent-ex-
ally used because of the lower power traction columns. The former systems
inputs required. are generally better suited to slower
extractions, and the latter to fast ex-
Special considerations tractions. Typical industrial applica-
Some considerations for specific sus- tions of solvent extraction are refining
pension applications are as follows: of metals from acid leach liquor and
• In many solids-suspension tasks, es- purification of pharmaceutical prod-
pecially in the minerals-processing ucts.
industries, abrasion can be a signifi- Surfactants are often used to aid
cant issue. In this case, lower veloci- formation of the correct dispersion
FIGURE 9. High viscosities can hin- ties may be required to limit abra- size and to hinder recoalescence, in
der heat transfer. Wall scrapers added sion. An increased impeller size can both temporary and stable emulsions.
to an agitator, such as this optimized compensate for the lower velocities. The interaction of the various factors
ribbon, can help increase heat transfer • During power failures, sediments affecting dispersion processes is very
substantially
can quickly build up in suspension complex and nearly always necessi-
classified into four broad categories on tanks. Impellers are often designed tates mixing tests. The main factors
the basis of hindered-settling velocity to withstand attempted restarts, affecting dispersion include these:
(see Figure 8). while submerged in a densely settled 1. Physical properties of the liquids
Type I tasks are simple suspend- slurry. In some instances, “restart in (especially interfacial tension)
ing duties that are readily predicted slurry” becomes the key design cri- 2. Effects due to mutual solubility or
because the liquid flows around the terion. reaction
particles in simple, laminar flow. Type • Solids suspension and gas dispersion 3. The presence of surfactants, and
II are demanding suspension tasks commonly occur simultaneously in 4. The turbulence and shear created
where the fluid flow is more complex the chemical- and minerals-process- by the agitator.
but predictable from empirical corre- ing industries. The presence of gas Most liquid-liquid dispersions are
lations — this category covers the ma- affects the performance of the impel- formed in the turbulent regime and
jority of industrial applications. Type ler and the ability of the fluid to sus- droplet sizes of between 5 and 500 m
III are difficult or "heavy" suspension pend solids. Likewise, the presence can be achieved. Typically, for a spe-
tasks, which probably involving large of solids affects gas dispersion. Both cific power input of 1 W/kg, droplet
or heavy particles. In this category, must be taken under consideration. sizes between 100 and 150 m can be
scale-up is usually based on pilot- Additional factors that can have a expected.
scale tests. For type IV tasks, uniform severe impact on the agitator design Droplets break when shear stresses
suspension is no longer attainable, as include: the shape of the base of the induced in the droplets by turbulence
they require very high liquid velocities vessel (dished, conical or flat); the lo- cause sufficient deformation to over-
that cannot be achieved economically. cation of the draw-off point; and the come the stabilizing effect of surface
In general, for solids suspension, the aspect ratio (H/T) of the tank. Con- tension. Minimum droplet size is gov-
agitator power requirement is scaled tinuous processes are most sensitive erned by the size of the micro-vorti-
up as a function of tank diameter ac- to such factors. ces, and is generally between 3 and
cording to the following equation: 5 times the size of the micro-vortices.
P ∝ DX
LIQUID-LIQUID When finer dispersions are required,
V (16) DISPERSIONS rotor-stator devices that produce a
Many industrial processes require the zone of very high energy dissipation
where X can vary from 0 to -1 depend- dispersion of one liquid into another, can be used. Stabilizers can be added
ing on the type of suspension duty. For immiscible liquid. Unstable disper- to ensure that fine droplets produced
example, if particles with a high hin- sions are created in processes such as in this high-energy zone do not re-
dered-settling velocity must be kept in solvent extraction (to provide a large coalesce in the bulk. For even finer
visually uniform suspension (Type III surface area for mass transfer), and in emulsions (from 5 m to 50 nanome-
in Fig. 8), a criterion close to constant dispersion polymerizations (to create ters), high-pressure homogenizers
specific power input P/V must be used the required size distribution of poly- with local energy dissipations of over
(X = 0). For solids with low hindered- mer particles or provide a heat sink). 500 W/kg are used.
settling velocities, a criterion closer to In the case of solvent extraction,
constant tip speed (X = -1) can be used the dispersion must be fine enough to Scale up
(Type I in Fig. 8). These examples il- allow rapid mass transfer and keep Due to the complexity of the interact-
lustrate how the specific power inputs the size of the extractor down, while at ing factors that affect droplet forma-
required to achieve “suspension” can the same time, the dispersion should tion, actual performance is usually
vary hugely on an industrial scale. be coarse enough to enable rapid measured in pilot trials. Because of
CHEMICAL ENGINEERING WWW.CHE.COM APRIL 2006 51
Cover Story
����������
������������
������������
�����
��������������
�������� � �� �� �� � � �� �� � � �� �� �
���������� � �� � �� �� �
����������
� �
����������
����������
�� � �� � �� �� �
�������������
����������
�� � �� � �� �� �
�������������
������������� � � � �� �� �� � �� � �� � �� �� �
������������
����������������������������� ��������������������������
���������
����������������������������������������� ����������������������������������
FIGURE 10. Modern-day impeller systems offer a great range of choice to help meet the mixing challenges facing the CPI
the sensitivity of surface properties blending and cooling during the reac- Design
to minor levels of contaminants, pilot tion to remove the reaction heat, fur- Heat-transfer coefficients in jackets,
trials should use actual chemicals ther heating to evaporate a solvent, coils and plate heat exchangers can be
from the process plant rather than and finally cooling down to near ambi- predicted from well established corre-
laboratory-quality reagents. Scale ent temperature before the product is lations [2]. Since the range of utility
up to production should follow rules discharged. fluids encountered is quite small and
that are based on the mechanism Heat transfer performance is gov- their physical and thermodynamic
(bulk or local turbulence) which was erned by: properties are generally well docu-
used to achieve the required perfor- • The flowrate and temperature of the mented, the accuracy of predicted film
mance. utilities, and heating/cooling me- coefficients is often very good.
dium On the process side, however, pre-
HEAT TRANSFER • The heat-transfer coefficients on the diction of the heat transfer coefficient
Stirred vessels are rarely used purely product side and the utility side is based on a general equation of the
for heat transfer because equipment • The type and contact area of heat form:
with much more-efficient heat ex- exchange surfaces hP ⋅ T µ 0.14
Nu = = C ⋅ Re 2 / 3 ⋅ Pr 1/ 3 ⋅ ( )
change is available. Heat transfer is, Some of these factors may need to be kP µW (17)
however, a critical unit operation that modified in order to achieve required
stirred vessels must be capable of per- heat fluxes. In many reactors, the ves- where the constant C, which depends
forming. A typical batch process, for sel, itself, provides insufficient surface on the impeller type and size, can
example, could comprise the heating area and it is necessary to install ad- be found in the literature or derived
of the bulk to reaction temperature, ditional heat-transfer surfaces. by measurements. The viscosity