You are on page 1of 8

Reactions and Separations

From Bench to Plant:

Scale Up Specialty Chemical


Processes Directly
Ronald B. Leng
The Dow Chemical Co.

Many new chemical, particularly batch operations,


can be scaled up directly from the bench to the
plant by developing the process and performing
lab testing with the scaleup in mind.

ANY COSTLY AND TIME-CONSUMING


startup problems can be avoided if key scaleup
issues are understood and resolved during the
development of a new chemical process. Processes are
often scaled up in stages from the lab to the pilot plant or
semi-works scale to obtain engineering data for commercial plant design. However, this staged scaleup strategy is
not always practical for specialty chemicals, which are
often characterized by multi-step batch syntheses and relatively low volume, and where speed to market and rapid
ramp-up are essential for commercial success.
This article explains how a direct scaleup strategy can be
used to successfully move a new process directly from the
bench to the commercial scale without demonstration in a
pilot plant. This approach involves conducting process development research in 18-L, geometrically similar mini-plants,
with a focus on simulating expected manufacturing conditions and testing the operating boundaries. It emphasizes
understanding particle processing, heat management, agitation, trace chemistry and other scale-sensitive issues.

Choosing a scaleup strategy


Scaleup is defined as The successful startup and operation of a commercial size unit whose design and operating procedures are in part based upon experimentation and
demonstration at a smaller scale of operation (1). Many
factors must be considered when selecting the scaleup
strategy. Answering a few process-specific and businessrelated questions early is key to a successful startup.

Process factors
What are the critical factors of the new chemistry
and process? Are extreme temperatures, pressures or
other conditions required? Are operating instructions
complicated?
Does the process involve a single reaction, or is it a
multi-step synthesis? If the last step in a multi-step
process will be piloted, will it be necessary to also make
intermediates at the pilot-plant scale, or are they commercially available?
Are new chemical technologies, unit operations or
equipment being considered?
How novel is the new process? Have similar reactions
or processing steps been successfully scaled up?
Will the new process be run in batch, semi-batch or
continuous mode?
Business factors
Does the commercial success of the project depend on
a flawless initial production campaign?
Is there an alternative supply of material in case startup problems limit the production rate?
Are project economics sensitive to yield or to the ability to recover and recycle some of the streams at relatively
high levels?
What is the commercial timeline? Is there enough
time to design, build and operate a pilot plant to generate scaleup data and still meet the planned commercial
launch?

CEP

November 2004

www.cepmagazine.org

37

Reactions and Separations

If the startup is delayed, what is the impact on the


product launch strategy and project economics?
Are significant quantities needed for the launch of the
product, or will it be introduced into the market slowly?
Are development samples needed over a period of
time leading up to the launch?
If a pilot-plant campaign is being considered, will the
business support the cost and human resources needed to
perform this activity?

Benefits and risks


Direct scaleup avoids the costs for pilot-plant design,
construction and operation. Fewer resources are needed to
develop the process entirely at the bench scale.
Development timelines can be compressed by eliminating
the pilot-plant stage.
However, surprises that dont appear until the larger
scale can be costly. More resources may be needed during
the startup phase. The physical form, purity or performance of the product may change as the process moves
from the lab to the plant. Certain scale-sensitive parameters cannot be fully tested at the lab scale.
If these risks are unacceptable, it is good to realize this
early so they can be addressed. A vital goal of process
development, either at the bench or pilot scale, should be
to understand the fundamentals prior to scaleup.
Scaleup issues
Some of the most common and difficult types of problems encountered during scaleup are particle formation
and isolation, liquid/liquid separation, agitation, heat history and trace impurities. (Reaction scaleup is widely discussed in the literature and will not be covered here, and it
is assumed that a sound chemical route has already been
selected.) Often, scaleup problems are a combination of
several of these factors (2).
Particle formation and isolation
Solids can form as a result of precipitation, often duirng
a reaction, or be produced intentionally, such as by crystallization. Generally, the goal is to form large, uniform particles, which will be filtered, washed and dried more efficiently, and are of higher purity, than fine particles. In
almost all cases, understanding and controlling the particle
growth environment will result in better particles (3).
Many reactions are run in a semi-batch or continuousaddition mode, where one of the reactants is metered into
the reactor and the product formed is a solid. The order of
addition, rate of addition and feed location, as well as the
intensity and design of the agitation system, can all affect
the particle formation process. It is also important to con38

www.cepmagazine.org

November 2004

CEP

sider the physical aspects in addition to the chemical


aspects of the reaction, and how these affect the particle
growth environment.
Crystallization processes involve creating a state of
supersaturation, typically by cooling, evaporation, chemical reaction or anti-solvent addition, which drives nucleation and particle growth. These processes are governed
by the conditions of the environment immediately next to
the particle. A basic understanding of the solubility curve
and supersaturation limit is quite helpful. Changing the
solvent phase composition can have a significant effect on
the solubility curve. Tools such as Fourier transform
infrared (FTIR) spectroscopy, optical density probes, and
microscopes are very useful for studying and optimizing
crystallization processes. It is a good idea to determine the
crystal size distribution (CSD), shape, strength and
whether multiple polymorphs exist. The latter is particularly important in the pharmaceutical industry.
Particles are usually isolated in the lab by filtration.
Scaleup to a pressure or vacuum filter can be predicted reasonably well from lab data (4). Vertical basket centrifuges
are often used in batch fine-chemical plants; scaleup prediction for these is more difficult because they use centrifugal
force to deliquor the crystals. A specialized test using a filter bucket centrifuge can be used to obtain scaleup data (5).
Data can also be obtained from a 12-in.-dia. test unit, but up
to 20 L of slurry may be required for one set of runs.
Comparison of lab filtration performance for a new application and a similar one currently operating in a plant centrifuge can be helpful. If particles are sensitive to attrition,
this should be incorporated into the design of the slurry
transfer equipment. Measuring the filtration rate as a function of cake depth is very useful for estimating filtration
cycle times, which can be the rate-limiting step in the plant.
Cake washing evaluations should also be included in the
lab experimental plan. Often, wash ratios in the plant are significantly higher than expected, which leads to higher costs
and waste disposal challenges. Wash media can be introduced
either as a flood or a surface spray, so both should be evaluated. If the cake has a tendency to crack and cause wash channeling, maintaining a liquid pool above the surface of the
cake is desirable. Feed maldistribution, particularly on a centrifuge, can have a detrimental effect on wash effectiveness.
Online filtrate conductivity can provide a relative indication
of the impurity level in the wash liquor, and can be used to
optimize the wash procedure.
If the solid is to be isolated as a dry product, drying
data should be obtained. Larger particles tend to dry faster
and more completely. A thorough understanding of the
thermal stability of the product to be dried is essential.
Knowing the dust explosion potential and toxicity of the

solid is also extremely important. Some solids go through


a viscous, pasty phase that can damage the dryers drive
unit if it is not designed to handle the added power
requirement. Finally, quality and toxicity should be considered in the design of the packaging system.
Example 1: Switching the order of reagent addition
improves particle characteristics. A two-step, reactive precipitation illustrates the significance of reagent addition order:

the cause and developing a method to prevent or break


the emulsion.
For coalescence, and thus phase separation, to occur,
small droplets of the dispersed phase must get close
enough to each other that the liquid continuous phase
between the drops drains, allowing the drops to unite.
Viscosity and surface tension are important variables
affecting the coalescence rate.
It is useful to test the sensitivity of the phase separation to
R-Na+ + H+ R-H
a variety of factors that simulate expected plant conditions,
2R-H + 0.5 H2O2 R-R
including the expected agitation intensity and the vessels
materials of construction. Agitator tip speed will normally be
First, acetic acid was added rapidly to a sodium-organic
higher in the plant, which can lead to smaller drop size and
salt in a 1,000-gal agitated reactor. This was followed by a
slower coalescence rates. Sometimes corrosion salts, even at
1-h metered addition of hydrogen peroxide to form the
relatively low levels, can impede the coalescence rate of tiny
product. The resulting product R-R particles were very fine
drops. Reagent or reaction impurities can accumulate at the
and had poor centrifugation and washing performance.
liquid interface and reduce surface tension or prevent drops
A follow-up solubility study showed that the concentrafrom getting close enough to coalesce. Look for impurities
tion of R-H at the end of the acid addition exceeded the
that have both hydrophilic and hydrophobic characteristics,
solubility limit by a factor of three. When the supersaturatwhich are attracted to both phases.
ed R-H crashed out of solution, it was impossible to form
Even an empirical understanding of factors that improve
large R-R particles.
or retard coalescence is helpful. Density difference is a key
The reagent-addition scheme was modified so the Na-R
variable in gravity or centrifugal separation. If it is much
solution was added simultaneously with the peroxide into
less than 0.1 specific gravity units, the difference may be
a 2-gal lab reactor containing the acid and solvent. The
increased by changing the temperature, changing the solresulting particles were large (50100 mm) spherical
vent, or adding salt to the aqueous layer. The density of the
agglomerates with much improved filtration and washing
organic phase will usually decrease more with an increase
characteristics (Figure 1). This modified approach was
in temperature than will the aqueous phase density.
successfully implemented in the plant.
Adjusting the pH of the system may be beneficial.
Changing the phase ratio can affect which is the dispersed
Liquid/liquid systems
phase and which is the continuous phase. Testing extremes
Two aspects of scaling up multi-phase reactions and
to determine whether phase inversion can occur is wise.
solute extractions are:
Consider whether the two-phase mixture will be subject to
if you dont see an emulsion in the lab, you will
additional shear, such as through a recirculation pump and
likely see one in the plant
loop. Often solid impurities can cause emulsions. Passing
if you see an emulsion in the lab, it will likely be
the mixture through a syringe filter is a simple test that can
worse in the plant.
usually identify if solids are part of the problem.
Process research should be focused on understanding
Evaluation of interface detection methods, such as conductivity, capacitance or density, is useful
for plant design. Examination of an
emulsion under a microscope may
reveal causes of emulsions, such as
solids or phase inversions.
Before scaling up, decide on a
process strategy to deal with emulsions. Processing methods can be either
equipment- or process-related. If there
is a small density gradient between the
phases, a decanting centrifuge can be
used. If gravity settling is to be
Figure 1. Adding reagents simultaneously produced the higher-quality particles on the right
employed, make sure to account for the
(Example 1).

CEP

November 2004

www.cepmagazine.org

39

Reactions and Separations

additional cycle time needed for the rag layer to coalesce.


Consider including an additional vessel to segregate the
rag layer for further settling. Sometimes a two-stage countercurrent system can be employed where the rag layer
with one of the phases can be passed through a filter to
remove solids as it is passed from one extractor to the
next. Also, there are a variety of coalescing filter designs
to choose from (6). If you plan to use either a centrifuge
or a coalescer, test its operation in the pilot plant.
Emulsions can sometimes be broken chemically. Consider
changing the solvent, although the options may be limited by
the reaction or extraction involved. Salt can be added to an
aqueous phase to change the density or alter the surface properties; this can sometimes be accomplished by recycling a
portion of a purged brine stream. Another effective approach
is to add a small amount of a surface-active agent to modify
the physical properties at the liquid interface.
Example 2: Changing the step sequence solves an
emulsion problem through the pH/salt effect. A post-distillation slurry containing a crude organic product and a partially soluble HCl salt of an organic base was quenched in
water in a 2,000-gal vessel. The original procedure was to
decant the organic product phase from the aqueous layer.
Caustic was added to the remaining aqueous layer to neutralize the HCl salt, liberating the free base, which was
then azeotropically distilled from the alkaline brine and
recycled. Upon scaleup, a fairly severe emulsion limited
capacity and caused operational difficulty.
Excellent phase separation was achieved when the
organic product was decanted from the alkaline brine following the base recovery operation. As a side benefit, one
of the impurities in the organic layer was hydrolyzed and
removed in the alkaline brine, which increased the purity
of the product.

Speed,
rpm
30
50
75
100
140
180
250
370

40

Agitation
The stirred tank reactor is the workhorse in many
chemical processes. Unfortunately, mixing (or agitation) in
a stirred tank is often overlooked during the scale up of
new processes. Successful scaleup involves three steps:
classifying the mixing requirement, determining the most
critical mechanism for scaleup, and then choosing the
design of the mixing system (7).
The first question to consider is what kind of mixing is
needed. Is the goal to blend two liquids together, or to
bring two or more different phases together such that a
chemical reaction involving mass transfer can occur at the
interface? Are solids being formed where the goal is to
suspend them gently without creating fines? Heat removal
is an important consideration, particularly for highly
exothermic reactions. Consider the method of reactant
incorporation, and whether feed location can have an
effect on side reactions; this can be important if competing
reactions are very fast and can be affected by localized
concentration gradients (8). The pumping direction created
by the agitator can be important, particularly in reactions
involving gas incorporation or removal.
Systems that are flow-dependent, such as the blending
of miscible liquids, can be scaled up based on impeller tip
speed. Some crystallizations can, too, provided the crystals
are suspended and moved through the region where the
supersaturation is created.
Reactions, crystallizations and extractions that are turbulence-dominated are often scaled up by keeping the
power per unit volume (P/V) constant. This is normally a
reasonable starting point for many new processes.
For reaction systems characterized by extremely fast
kinetics with competing reactions, scaleup based on constant mixing time may be optimal, although this is typically difficult and expensive. To illustrate this point, Table 1
presents a scale-down comparison of a
1,000-gal plant reactor and the geometrically
Table 1. Mixing scale-down example.
similar 2-gal reactor used in Example 1, both
fitted with dual pitched-blade turbines:
Plant Scale: 1,000 gal
Laboratory Scale: 2 gal
1. The agitation rate in the plant reactor
Mixing
Tip
Mixing
Tip
was 50 rpm (green). In testing the proposed
Time,
P/V,
Speed,
Time,
P/V,
Speed,
new addition order, P/V was kept constant at
min
(ft-lbf/s)/ft3
ft/s
min
(ft-lbf/s)/ft3
ft/s
0.58 (ft-lbf/s)/ft3. To achieve the P/V ratio in
3.9
0.12
3.7
3.2
0.003
0.5
the 2-gal lab reactor, a speed of 180 rpm was
2.3
0.58
6.1
1.9
0.01
0.8
necessary. In this case, the results achieved
1.6
1.95
9.1
1.3
0.04
1.2
in the lab and the plant were very similar.
1.2
4.62
12.2
1.0
0.10
1.6
2. However, if particles were sensitive to
0.8
12.67
17.1
0.7
0.27
2.3
agitator tip speed, this could not be simulat0.7
26.93
21.9
0.5
0.58
2.9
ed in the lab independent of P/V. At P/V =
0.5
72.16
30.5
0.4
1.57
4.1
0.58 (ft-lbf/s)/ft3, the lab tip speed is 2.9 ft/s
0.3
5.08
6.1
(red) only about half the tip speed of the

www.cepmagazine.org

November 2004

CEP

plant system (6.1 ft/s). To test at 6 ft/s, the lab reactor


would have to be run at 370 rpm, which would increase
P/V nearly tenfold, to 5.08 (ft-lbf/s)/ft3.
3. Finally, if a reaction were run at 180 rpm in the lab
and needed to be scaled up based on mixing time, to
achieve mixing in 0.5 min, the plant agitation rate would
need to be 250 rpm (blue). This would increase P/V more
than a hundredfold (to 72.15 (ft-lbf/s)/ft3). This would be
very difficult to accomplish in a stirred tank reactor, so
alternative contacting should be evaluated.
When scaling up agitation, pay attention to the impeller
or turbine design, and to the tank baffling configuration.
Many styles of impellers are tailored to the specific mixing requirement. Styles range from foil-type impellers that
can deliver high axial flow with low shear, to high-shear,
radial-flow, flat- or cupped-blade turbines designed for gas
dispersion applications. Impeller physical dimensions,
number and the power of the drive motor can be specified.
Keep in mind that the new process may be run in an
existing agitated vessel, so it is important to understand
the type of system available and its limitations. For
instance, if the equipment is glass-lined steel, the reactor
will usually have a large-diameter, crow-foot agitator
and an h-type or beaver-tail baffle that can create
turbulence but not axial flow. In a metal vessel, baffles
can be welded close to the reactor wall, which will
change the flow direction produced by the agitator from
radial to axial flow.
Process development research should test very high and
very low agitation rates to determine sensitivity to mixing. If
the results are mixing-sensitive, consult with a mixing expert
and consider testing in a larger vessel. Finally, consider other
sources of turbulence and mixing, such as would be encountered in a recirculation loop or other equipment.
Example 3: Shear-induced phase inversion produced a
high-viscosity emulsion. A product isolation process
required extraction of the solute from an organic solvent
using dilute aqueous acid. In normal operation, the waterin-oil dispersion of the two similar-density liquids was fed
to a stacked-disc centrifuge to separate the phases (Figure
2). Upon scaleup, an occasional high-viscosity, puddinglike emulsion formed, which choked off the centrifuge and
severely limited plant throughput.
It was determined that the emulsion was caused by a
phase inversion phenomenon. The aqueous phase became
the continuous phase, which was surprising because the
aqueous-to-organic-phase ratio was very low (0.16 to 1).
Phase inversions often occurred during centrifuge restarts
or upsets. Lab studies showed phase inversions resulted
when high shear was introduced in the quiescent aqueous
phase that resulted during upsets.

1% H2SO4 (0.16 of feed)

Feed

Extract
Stacked Disc
Centrifugal
Separator

4,000-gal
Vessel with
two Radial
Flow
Turbines

Raffinate

Figure 2. The stacked-disc centrifuge in this product extraction process


occasionally was choked by a high-viscosity emulsion (Example 3).

Analysis of the plant operation revealed that the shearinducing turbine design in the centrifuge feed tank (reused
equipment), the piping configuration, and the operating
procedures all contributed to the problem. No phase inversions could be created at phase ratios < 0.14, or when a
small amount of a specific surfactant was added to the system. Several subsequent modifications to the process prevented phase inversions from occurring in the plant.

Heat management considerations


A common scaleup axiom is Everything takes longer
and runs hotter in the plant. The key scaleup consideration is that the volume-to-surface-area ratio increases in
proportion to vessel diameter. A 1,000-gal reactor has 10%
of the relative heat-removal surface area of a 1-gal reactor.
It is essential to understand the heat involved in the
process the heat of reaction and heats of vaporization
for all desired processing. An accelerating rate calorimeter
is a useful tool for determining the point of onset, potential rate and magnitude of energy release from unplanned
reactions or thermal runaways.
Often, a highly exothermic reaction needs to be scaled
up. Typically, adding one of the reactants continuously can
control the exotherm. Be sure to starve the reaction and
dont build up potential energy by adding the reactant too
fast, which could lead to a thermal runaway situation. A
dynamic heat balance can confirm that the reaction is proceeding normally before all of the reactant is added. This
is especially important if a minor component or catalyst is
mistakenly omitted or if agitation stops.
Determine the heat removal capability of equipment to
be used in the plant. Be aware that the heat-transfer coefficient for a glass-lined steel reactor may be 50% or less
than that of an alloy vessel. If an existing vessel is to be
used, the jacket may have some fouling, which will limit
heat transfer. This can usually be remedied by cleaning
the jacket periodically and by using clean or treated heattransfer fluid.

CEP

November 2004

www.cepmagazine.org

41

Reactions and Separations

Perform heat balance calculations to determine how


long heat removal steps (i.e., exothermic reactions or distillation steps) will take in the plant, then simulate these in
the lab. If there are no adverse effects on yield, quality or
cycle time, then you should be able to remove or add heat
directly through the jacket.
If greater heat removal capability is required, this
can be accomplished by adding a heat exchanger in a
recirculation loop, although this can introduce other
problems. An option may be to remove heat via a
reflux condenser, sometimes by the addition of a lowboiling liquid.

Desired:

A+BC

fast, product formation

Competing:

A+CD

slow, dimer formation

During process development studies, time and temperature were tested independently, but worst-case plant conditions were not studied. In all tests, dimer levels were consistently less than 1%.
In a follow-up lab study, the cause of the higher dimer
level was confirmed, and temperature was found to be the
more important variable. When the reaction was conducted
at 05C for 9 h, which would be achievable in the plant,
the dimer level was < 1%. Furthermore, when the concentration of Reactant A was kept low by metered addition, the
dimer level was very low (< 0.5%), even at 20C.

Inverse Drain Rate, 1/(g/s-cm2

Distillation considerations
Batch distillation is usually a good choice for relatively
low volume production. It integrates well with other batch
unit operations, is flexible for making multiple distillation
Trace chemistry
cuts, and handles mixtures with solids and variable feed
Consider all sources of impurities, including raw matecomposition. Quite often, vacuum distillation is needed to
rials, side reactions and accumulation in recycle loops.
achieve reduced temperature for heat-sensitive applicaThen determine the fate of all impurities in the process.
tions. If a distillation column is used, lower-pressure operModeling impurity buildup in recycle loops can also be
ation will require a larger-diameter column to achieve
valuable. If the reaction conversion or the amount of
comparable rates. As the column diameter is increased, the
excess reactant(s) can vary from batch to batch, evaluate
height of a theoretical stage may also increase, so a comthe impact of such variation on the impurity spectrum.
promise must be struck.
Processes with recovery and recycle loops should include
Modeling the distillation is recommended. It is also
provisions for purging impurities. Finally, consider the
prudent to conduct the lab distillation at the anticipated
effects of impurities on product quality.
time/temperature profile in the plant, using a column
In planning process development research, evaluate
designed to achieve the desired separation. Be aware
commercial-grade raw materials. Analyze for trace
that even if the time is simulated in the lab, the wall
chemistry in all process streams, including vents, as this
temperature in the plant vessel will be higher. This can
can be a source of reactive chemistry concern. Spike lab
be tested by limiting the heat-transfer area in a specialruns with high levels of impurities to determine the
ly designed lab reactor to mimic the area/volume ratio
robustness of the process. Run up to 10 cycles in the lab
in the plant vessel. If the heat history is excessive, conusing recycle streams to identify potential problems at
sider alternative separation technology, such as shortsteady state.
path or multi-pass continuous distillation, solvent or melt crystallization, or
3.0
solvent extraction.
Projected Rate
due to loss in cake permeability
2.5
Example 4: Side reaction is four times
based on 16 cm data
higher in the plant due to heat history. A
2.0
highly exothermic, semi-batch reaction
was scaled up from a 1-gal lab vessel to
1.5
a 750-gal plant reactor. Yield loss to a
dimer byproduct was three to four times
1.0
higher in the plant than in the lab. It was
Extrapolated Performance
suspected that the cause was from run0.5
based on 10 cm data for
ning the reaction about 510C hotter in
normal cake resistance
0.0
the plant in order to meet cycle time
0
5
10
15
20
25
requirements.
Cake Depth, cm
The chemistry involved two competing
reactions in series:
Figure 3. Cake resistance increases sharply at depths greater than 15 cm (Example 5).
42

www.cepmagazine.org

November 2004

CEP

Table 2. Example of success parameters.


Example 5: Feed impurity caused
a filtration problem. A semi-batch
Needs
crystallization process was scaled up
Units
Work
Excellent
Actual
from a 1.5-L lab crystallizer to a
2,000-gal vessel with a 4-m2 filter.
Reactant Drying
Upon scaleup, excellent product
Evaporation Time
h
>18
<12
6
crystals, very similar to those proAzeotrope Time
h
>12
<6
8
Azeotrope Ratio
lb/lb
>1
<0.4
1
duced in the lab, were formed.
Water
Level
wt%
>4
<1
0.5
to 1
The filtration rate, however, was
Intermediate
Preparation
very slow, with some batches requirAcid Addition Time
h
>12
<6
6
ing up to three days to complete. In
Temperature Control
C
10
5
1
addition, the plant experienced poor
Acid Addition Control
% of setpoint
2
1
1
cake washing and slow drying. Slurry
Product Reaction
samples from the plant were filtered
Base Addition Time
h
>4
<1
<0.5
in the lab to confirm the problem.
Volatiles Removed
% of theoretical
<40
>55
50
For normal filtrations, a linear
Overall Reaction Time
h
>24
<18
12
relationship exists between the
Product Precipitation
inverse of the filtration rate and the
Particle Size/Shape
/micron
plates/<50
cubic/>100
cubic/>200
Final pH
pH
4.0 0.5
4.0 0.1
4.0 0.2
cake resistance or thickness. The data
Product
Isolation
in Figure 3 show a linear relationship
Total Filtration Time
h
>36
<24
>48
at cake depths below 10 cm, but a
Wash Rate
lb/lb
>4
<2
3.5
steep increase in cake resistance
Water in Dry Cake
wt%
>0.5
<0.2
<0.2
above 15 cm. It was determined that
Overall Performance
a floc-like solid clogged the voids in
Product Assay
wt%
<90
>95
95
the cake, decreasing permeability.
Main Impurity Content
wt%
>6
<4
4
The floc was traced to an impurity in
Overall Yield
%
<75
>82
6570
a raw material, which was present at
roughly twice the level tested in the
lab during process development. Also, the maximum cake
Model, simulate and rigorously test expected plant condepth tested in the lab was only 10 cm, while the bed
ditions, particularly materials of construction, agitation,
depth in the plant was 25 cm.
heat removal and separations.
To solve the problem, the plant shut off the agitator and
Acquire and test commercial-grade raw materials.
allowed the crystals to settle. The mother liquor containing
Perform insult studies to test process robustness.
the floc was decanted to waste treatment, and the product
Brainstorm with your manufacturing colleagues to devise
crystals were diluted with wash water and fed to the filter,
insult scenarios. Insult variables include time, temperature,
resulting in cycle times less than 24 h.
pH, agitation, reagent quality and stoichiometry, and materials of construction.
Recommended approaches
Establish success parameters for your scaleup, such as
those in Table 2. Select two or three performance metrics
for process development and scaleup
The single most important thing you can do in the
per unit operation, such as yield or cycle time. Once stable
process development phase to ensure a successful scaleup
operation is achieved, compare the actual values with lab
is to promote and ensure early-stage, ongoing teamwork
or expected results to identify gaps. These may become
between chemists and engineers. Unfortunately, engineers
the focus of follow-up research and optimization.
are often brought in too late and chemists are pulled from
development projects too early. Management should proTools for process development and scaleup
vide incentives to foster this desirable, collaborative enviUse geometrically scaled lab reactors and separation
ronment from the outset of the project.
systems made from the materials of construction to be
A holistic methodology that considers cost of manufacused in the plant. Computer control and data collection
ture, capital, complexity, robustness, environmental,
capabilities are useful.
health, safety and scaleability issues should be used to
Use in situ analysis to gain as much information from
evaluate possible process routes.
your new process as possible.

CEP

November 2004

www.cepmagazine.org

43

Reactions and Separations

Understand the energy of the system for both planned


and unplanned chemistries. Use calorimeters to obtain heats
of reaction. Perform other reactive chemicals testing, such as
differential scanning calorimetry (DSC), accelerated rate
calorimetry (ARC), and dust explosion testing.
Use computer models to evaluate expected plant conditions.

Closing thoughts
Many new chemical processes, particularly those involving batch operations, can be scaled up directly from the
bench to the plant. However, not all processes should be
scaled up without pilot plant demonstration. As a process
development engineer or chemist, you need to evaluate the
risks and benefits of each scaleup strategy. Be aware of and
test scale-sensitive parameters, including solids operations,
multi-phase systems, agitation, heat transfer and history, and
the effects of trace chemistry. Develop the process and perform lab testing with the scaleup in mind. Liberally consult
with subject matter experts. Finally, plan for a successful
CEP
start up and evaluate the results.

Literature Cited
1.
2.
3.
4.
5.
6.
7.
8.

Bisio, A., and R. L. Kabel, Scaleup of Chemical Process,


Wiley, Hoboken, NJ, p. 3 (1985).
Anderson, N. G., Practical Process Research and
Development, Academic Press, San Diego, CA (2000).
Myerson, A. S., Handbook of Industrial Crystallization,
Butterworth-Heinemann, Newton, MA, pp. 1519 (1993).
Perry, R. H. and D. W. Green, eds., Perrys Chemical
Engineers Handbook, 6th ed., Chapter 19, pp. 65103,
McGraw-Hill, New York, NY (1984).
Purchas, D. B., ed., Solid-Liquid Separation Equipment
Scaleup, Uplands Press, London, pp. 493553 (1977).
Osmonics, Inc., Liquid/Liquid and Gas/Liquid Coalescing
Handbook, Ninnetoka, MN (1991).
Paul, E. L., et al, eds., Handbook of Industrial Mixing,
Wiley, Hoboken, NJ (2004).
Fasano, J. B., and W. R. Penney, Cut Reaction
Byproducts by Proper Feed Blending, Chem. Eng. Progress,
87 (12), pp. 4652 (Dec. 1991).

Further Reading
Sharnatt, P. N., Pilot Plants and Scale-up of Chemical
Processes, Hoyle, W., ed., Royal Society of Chemistry,
Cambridge, UK, pp. 1321, 130, 655690 (1997).

RONALD B. LENG is currently a process development leader in the


Chemical Sciences Dept. of The Dow Chemical Co. (1710 Building,
Midland, MI 48674; Phone: (989) 636-6158; E-mail: rbleng@dow.com).
During his 23-year career, he has held a variety of positions both in
research and development and manufacturing. His chief interests lie in
bridging the gap from the lab to the plant, and he has worked on a
number of process development teams to bring new processes for
agricultural chemical products to realization. Particular interests include
understanding key scaleup fundamentals and their practical application
to the plant environment. He has a BS in chemical engineering from the
Michigan Technological Univ.

44

www.cepmagazine.org

November 2004

CEP

www.cepmagazine.org or Circle No.123

You might also like