0263–8762/04/$30.00+0.

00
# 2004 Institution of Chemical Engineers
www.ingentaselect.com=titles=02638762.htm Trans IChemE, Part A, July 2004
Chemical Engineering Research and Design, 82(A7): 865–880

DESIGN OF A MULTI-FUNCTIONAL CRYSTALLIZER

FOR RESEARCH PURPOSES
G. M. WESTHOFF1,*, H. J. M. KRAMER1, P. J. JANSENS1 and J. GRIEVINK2
1
Laboratory for Process Equipment, Delft University of Technology, Delft, The Netherlands
2
Process System Engineering, Delft University of Technology, Delft, The Netherlands

T
he design procedure and the specifications for the design process and the resulting
experimental set-up of a multi-functional 75 l pilot plant crystallizer are described. The
crystallizer can operate in different operational (batch=continuous) and crystallization
modes (cooling=flash–cooling=evaporative). This flexibility enables the generation of experi-
mental information on the relationship between process conditions and product quality for
different model compound systems using one crystallizer. The design is realized using a
hierarchical design procedure described previously, which is extended with an additional level
for the design of instrumentation and control. The design procedure is useful to structure the
design process of a pilot plant crystallizer and to identify design problems at an early stage. A
description of the pilot plant crystallizer, a design synthesis model, a flowsheet design and an
investment cost estimation of the experimental set-up and the cost of this infrastructure per
typical experiment are presented. The designed pilot plant was actually built and successful
crystallization experiments have been performed. However, the deviation seen between the
median crystal sizes as specified as design target and the measured one in the realized
crystallizer illustrates the need for a rigorous predictive crystallizer model to further analyse
and optimize a design.

Keywords: hierarchical design procedure; multi-functional pilot plant; crystallizer; crystal-

lization; process design.

INTRODUCTION certain solvent–solute system, crystallizer geometry and

process conditions, it is necessary to take into account the
Crystallization involves the formation of one or more non-uniform distributions of the process variables (e.g.
structured phases from a fluid phase or an amorphous crystal size distribution, energy dissipation, solids density,
solid phase. It is one of the older unit operations in the etc.) thorough the crystallizer. Kramer et al. (1999) and
chemical industry. Crystallization enables the production of Bermingham et al. (2000) have presented more comprehen-
solid particles with well-defined properties, such as optical sive design procedures for crystallizers using rigorous
properties and bulk density. The produced crystals have a crystallization models. However, these design procedures
characteristic crystal size distribution, purity, shape and have not been validated so for by building and testing the
morphology. These product properties are dependent on designs.
the geometry and the type of the crystallizer, physical The objective of this paper is to validate the usefulness of
properties of the solvent and the crystallizing substance, a comprehensive design procedure for the design of a
process conditions during the crystallization process and the (continuous) multi-functional pilot plant crystallizer. The
downstream processing. Traditionally scale-up design purpose of the pilot plant crystallizer is to generate experi-
procedures for crystallizers have been developed in the mental data for the validation and development of rigorous
last decade (Mullin, 1993; Benett, 1993; Mersmann, 1995). crystallization models. Therefore the crystallizer must have
These procedures contain many useful elements with respect the flexibility to operate in different operational (batch=
to process and equipment design, but pay little attention to continuous) and crystallization modes (cooling=flash–
product design. In order to predict product quality for a cooling=evaporative). This flexibility enables the generation
of experimental information on the relationship between
process conditions and product quality for different model
*Correspondence to: Dr G.M. Westhoff, Laboratory for Process Equipment,
Delft University of Technology, Leeghwaterstraat 44, 2628 CA Delft, The
compound systems using one crystallizer. An advantage of
Netherlands. using a continuous producing pilot plant crystallizer is that
E-mail: [email protected] the crystallization phenomena (e.g. nucleation, growth,

865
866 WESTHOFF et al.

attrition, dissolution and agglomeration) occurring in a are presented. Finally, the experimental median crystal size
continuous industrial crystallization process can be studied. measured during crystallization experiments in the realized
The estimation of a phenomenon as attrition in an industrial experimental set-up is compared with specified median
crystallization process is more reliable in a pilot plant crystal size used in the design procedure to test the relia-
crystallizer than in a 500 ml beaker. An additional advantage bility of the crystallization model used.
of a continuous producing pilot plant crystallizer is that the
process is more stable than with an intermittent product
INITIAL DESIGN SPECIFICATIONS
removal.
The design of a multi-functional pilot plant crystalliser is Traditionally, in an industrial environment, crystallization
more complicated than the design of a production crystal- processes are designed to optimize manufacturing perfor-
lizer, because the design is more focussed on the flexibility mance for one or a few products. Manufacturing perfor-
of the crystallizer, rather than the production of one specific mance is related to both product quality aspects and to the
crystalline material. process efficiency, such as economics, materials and energy
In this paper, the design process of a multi-functional demand. During the design of a crystallization process for a
pilot plant crystallizer is described and applied. The design specific product the crystallization method, the operation
follows a hierarchical design procedure proposed by mode and the crystallizer type need to be selected. The
Bermingham et al. (2000), which consists of five levels crystallization method, such as evaporation and (flash-)
(levels 0–IV). The design procedure is extended with an cooling, is mainly determined by the temperature depen-
additional level V, ‘design of instrumentation and control’ dence of the saturation concentration. The choice for a
(see Table 1). continuous or (semi) batch process is dependent of criteria
The design process starts with the formulation of the such as the production capacity, scaling tendency and width
initial design specifications followed by the subsequent of crystal size distribution. The crystallizer type (i.e. draft
levels. These specifications are translated to crystallization tube baffled, draft tube, forced circulation, Oslo) influences
tasks, flowsheet design, design of crystallization stage and to a great extent the maximum obtainable crystal size
design of instrumentation and control. distribution. Overviews of different crystallizer types are
At each level the same step-by-step procedure is followed. given in Perry and Green (1998) and Benett (1993).
First the applicable design specifications are listed and In industrial manufacturing the crystallizer would be
the design space is identified. Secondly, the accompanying optimized for a given specific crystalline material. However,
domain knowledge (experimental data, heuristics and the purpose of this multi-functional pilot plant crystallizer is
models) is gathered. The third step at each level is the syn- different. The crystallizer will not be designed for the
thesis task, where possible design alternatives are created. In production for one or a few products. The main purpose is
the fourth step the design alternatives are analysed. The to generate experimental data over a sufficiently wide range
results of steps 3 and 4 are evaluated in step 5. A few of the of operating conditions, using different model materials and
design alternatives (if evaluated positively) are propagated different operation modes. The multi-functional crystallizer
to the next level. must have the flexibility to crystallize different model
The result of the design, the flowsheet of the pilot plant materials in different operation (batch=continuous) and
crystallizer and the costs of the realized experimental set-up, crystallization (cooling=evaporative=flash-cooling) modes.
Thus in this typical crystallizer design the product perfor-
mance is of less importance than the process performance
Table 1. Level description of the systematic design procedure.
requirements (flexibility regarding the crystallisation and
Level 0 Initial design specifications operation modes).
Level I Design of the crystallization product The multi-functional crystallizer must enable the crystal-
Level II Physical=chemical design of the crystallization task lization of the model systems adipic acid–water, potassium
Level III Flow sheet design of the crystallization process
Level IV Design of a crystallization stage
nitrate–water and ammonium–sulphate water under varying
Level V Design of instrumentation and control conditions. In addition, the crystallizer must have the
possibility of operating in batch and continuous modes.

Table 2. Level 0—initial design specifications [(þ) ¼ relevant, (
) ¼ irrelevant].

Product performance requirements

In the crystallizer No flotation þ Suspendability þ
Downstream handling Filterability þ Pneumatic handling

Washability
Freedom from dust

Dryability þ Flowability

Dissolution rate
Mechanical strength

Customer application No caking in storage
Freedom from dust

Dissolution rate
Bulk density or porosity

Mechanical strength
Aesthetic appearance

Process requirements Production capacity þ Availability þ
Feed composition þ SHE considerations þ
Yield
Battery limits þ
Energy consumption
Resiliency þ
Controllability þ Research requirements þ
Design process requirements Design budget
Time to market

In-house or licensed technology
Availability skilled design staff

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 867

The economics and energy demand of the crystallizer are Jansens et al. (1995) presented an integral model frame-
less important. work, which allowed the prediction of the slurry density,
In industrial cases the production capacity will be speci- crystal size distribution, porosity (e) and permeability (Pm)
fied, but in this typical case the production capacity is of a cake of the crystalline product from a cascade
minimized (because of the availability of laboratory space) cooling crystallizer based on the actual operating conditions.
and determined by the requirement of a continuous instead MacDonald and Kozeny-Carman equations were used to
of a discontinuous product removal flow. This is related to calculate the permeability. The porosity was calculated
the minimum diameter of the product line and the minimum using the model of Ouchiyama and Tanaka (1984).
allowable product flow velocity. In general, models that describe the relation between the
The relevance (
=þ) of the list of items of the initial product performance and product composition are scarce.
design specifications (Bermingham et al., 2000) is listed in Therefore empirical knowledge is often used, obtained from
Table 2. An additional process specification (research requi- literature or determined by laboratory experiments. The
rement) is added. Short motivations for the relevant design crystal size distribution of the crystal slurry and the crystal
specifications, marked with ‘þ’, are listed in Table A1. morphology has a strong influence on the filterability and
dryability. A narrow crystal size distribution will give better
filterability and dryability characteristics than a broad crys-
DESIGN PROCESS tal size distribution. The type of crystallizer and the resi-
Levels I–IV of the design procedure aim to find design dence time also influences the size of the crystals. Impurities
alternatives for the requirements specified in level 0 (Table 2). like metal ions have a strong influence on the crystallization
kinetics; therefore the composition of the feed stream and
accumulation of the impurities should be taken into account.
Design of Crystalline Product (Level I) No flotation is expected for the model systems due to the
At this stage the product, composition of the crystalline density difference of the crystals and the saturated solutions
product is determined, which will result in a product which in the range between 0 and 65 C.
meets the product performance criteria. The product perfor- It has to be noted, that by crystallizing a polymorphic
mance criteria specified in level 0 are: no flotation, suspend- system, the crystallization process can result in the produc-
ability, filterability and dryability. It has to be noted that tion of more than one crystal form, in fact a mixture of
these product performance criteria are important for indus- polymorphs. Therefore, knowledge of the process that
trial application. Poor filterability of a crystalline product governs nucleation, crystal growth and phase transformation
will lead to a reduced capacity in the solid–liquid sepa- in these systems is essential for process control. An over-
ration step. The dryability of the crystalline product will view of concominant polymorphism is given by Bernstein
be dependent of the characteristics of the solid. A solid et al. (1999).
may be porous, non-porous and can also be hygroscopic or Dry KNO3 could give explosion risk by mixing with
non-hygroscopic. The drying rate of the solids is dependent organic material. Therefore special attention should be given
on the internal mechanism of liquid flow in the crystal to the storage of dry KNO3.
(e.g. diffusion), and on the effect of external conditions of
temperature, humidity, airflow, etc.
The relevant design aspects are marked in Table 3. The
Physical=chemical Design of the Crystallization
first column of Table 3 gives the design specifications, the
Task (Level II)
second column the design variables and the third column
the domain knowledge. In the hierarchical design procedure The design specifications at this design level result
the design variables are determined and appear as design from the product composition characteristics from level I
specifications in the next design level. Motivation of design and the initial design specifications from level 0. In Table 4
choices is given in Table A2. the relevant design specifications, design and operational

Table 3. Design level I—design of the crystalline product.

Design specifications Design variables Domain knowledge

Filterability þ Discrete Filterability tests and models: þ
permeability and compressibility
No caking in storage
Polymorphism þ Shear tests

No flotation þ Indentation tests

Suspendability þ Continuous Caking tests

Washability
Morphology þ Flowability tests

Dryability þ Crystal size distribution þ Safety aspects þ
Dissolution rate
Purity þ
SHE considerations þ Maximum inclusion content

Pneumatic handling

Freedom from dust

Flowability

Mechanical strength

Abrasion resistance

Bulk density or porosity

Aesthetic appearance

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
868 WESTHOFF et al.
Table 4. Design level II—physical=chemical design of the crystallization task.

Design specifications
objectives and constraints Design and operational variables Domain knowledge
Process Discrete Thermodynamic properties of species=components þ
Production capacity þ Crystallization mode þ in solid, liquid and vapour phase
Feed composition þ Feed purification þ Adsorption (energy) of components=species on the

Yield
Recrystallization step
various crystal faces
Energy consumption
Solvent(s)
Scaling or encrustation tendency of component species þ
Availability þ Additive(s)
Meta-stable zone with respect to primary nucleation þ
SHE considerations þ Material of construction þ Safety aspects þ
Corrosion þ
Product Continuous
Polymorphism þ Pressure range þ
Morphology þ Temperature range þ
Crystal size distribution þ Concentration solvent (s)

Purity þ Concentration additive (s)

Maximum inclusion content

variables and domain knowledge are marked. The motiva- Table 5. Tasks to be realized in the multi-functional crystallizer.
tion is given in Table A3.
The crystallization methods are selected according to a Tasks Description
procedure presented by Kramer et al. (1999). The solubility Heating Heat transfer through wall to heat
lines of the specified model compound systems determine mother liquor
the crystallization mode, which are shown in Figure 1. Evaporation Flashing solvent to concentrate solute
Indirect cooling Heat transfer from mother liquor to
The solubilities of the adipic acid–water system and the cooling circuit
potassium nitrate–water system both show a strong depen- Flash-cooling Flash solvent to cool another liquor
dency on the temperature. Therefore cooling crystallization and to create supersaturation
or flash-cooling crystallization is the most efficient method Condensation vapour Heat sink by external condensing unit
(not part of crystallizer)
of crystallization for these two model compound systems. Feed supply Continuous=semi-batch
Flash-cooling crystallization has the advantage of the (Iso-kinetic) product removal Continuous=semi-batch
absence of a heat exchanger and thus eliminates scaling Suspension Keep crystals in suspension
problems. The ammonium sulphate–water system has a flat (sufficient mixing)
solubility line. Evaporative crystallization is therefore the
most appropriate crystallization method for this model
compound system. Thus the crystallizer must have the
possibility of operating at a certain pressure (0.1–1 bar) and addition, the material of construction of the crystallizer is
temperature range (25–65 C), and operation (batch=contin- important, because limited release of metal ions can affect
continuous) modes. The tasks, which should be achievable in the crystallization process. Also, the construction material
the crystallizer, are summarised in Table 5. The physical tasks must be corrosion-resistant [must resist solutions with a
are considered from the viewpoint of mass, energy and low pH (adipic acid–water solution)] and must be easy to
impulse transfer between the system and surrounding. modify. Therefore Stainless Steel 316 is selected as
Encrustation problems can be expected by the crystal- construction material for the crystalliser. The use of Stain-
lization of adipic acid from water. The impurity level in the less Steel as construction material enables small modifica-
solution can have an influence on the crystallization tions to be made to the crystallizer easily, which is difficult
kinetics; therefore a pure feed stream is necessary. In if glass is used as the construction material. Sometimes the
scaling tendency can be lowered by coating the construction
material. No safety risks are expected in the temperature
and pressure range during the crystallization of the model
compound systems.

Flowsheet Design of the Crystallization

Process (Level III)
The outcome of the design decisions at level II is propa-
gated as design specifications, objectives and constraints to
design level III. The points of attention are listed in Table 6.
The motivation of the marked constraints and objectives
of this level are listed in Table A4. Despite the fact that a
multiple stage operation will produce a more narrow crystal
size distribution than a single stage, a single stage is chosen.
A single stage operation is simple, robust and sufficient to
determine the relation between process conditions and
Figure 1. Solubility lines of model compound systems (Mullin, 1993). product quality. In addition, the investment costs for a

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 869
Table 6. Design level III—flowsheet design of the crystallization processes.

Design specifications Design and operational

objectives and constraints variables Domain knowledge
Process Discrete Thermodynamic activity of species=components þ
Production capacity þ Operation mode þ in solid, liquid and vapour phase
Feed composition þ Number of stages þ Crystallization kinetics, i.e. rate þ
Crystallization method þ Feed configuration þ expressions for the nucleation,
Yield
Recycle structure þ growth attrition, agglomeration and
Pressure range þ Location purge stream(s) þ breakage of crystals
Temperature range þ Fouling kinetics þ
Supersaturation range þ Continuous Shape factors of the crystalline components

Energy consumption
Residence time in each þ Fire and explosion index þ
Availability, controllability þ stage or batch time
and resiliency Recycle flow rates þ
SHE considerations þ Purge flow rate

Pressure and=or temperature
Product in each stage þ
Crystal size distribution þ Heating=cooling duty or trajectory þ
Heat exchange rates þ

multiple stage operation will be high. The different modes different crystallization modes. The crystallization tempera-
of crystallization during continuous operation require a ture is chosen to be approximately 50 C, which enables
more flexible flowsheet design. In the case of cooling comparison with other crystallization experiments perfor-
crystallization the crystallizer must be cooled with cooling med in this research programme. The flows will be based on
water. In case of evaporative crystallization, heating of the the calculated ‘worst case’ crystallization mode.
solution=suspension in the crystallizer and condensation of
the vapour stream is necessary. Condensation of the vapour
stream is also necessary during flash-cooling crystallization. Design of a Crystallizer Stage (Level IV)
Batch-wise operation of the crystallizer requires a special The design specifications and constraints will now be
arrangement for the realization of the supersaturation trajec- extended with equipment related specifications. These are
tory (by cooling and evaporative crystallization), which is marked in Table 7. The motivation is given in Table A5.
discussed in level V. The first selection at this level involves the crystallizer
The pilot plant must be a closed system with no waste type. The crystallizer type influences the growth and attri-
streams (SHE requirement, see Table A1). Recycling of the tion phenomena. The choice of crystallizer type will deter-
condensed vapour, product (relevant for continuous mode) mine the kinetic model, which has to be used for the sizing
and cooling water streams (cooling of crystallizer and of the crystallizer unit. Several crystallizer types are possi-
cooling water for condenser) is required. Therefore a ble, such as forced circulation (FC), DTB, draft tube (DT)
buffer vessel for the product and condensed vapour is crystallizer, Oslo crystallizer, etc. As already specified, the
added to the flow sheet together with a feed preparation crystallizer must be simple, robust, easy to operate and
vessel. The hot=cold water vessel, buffer and feed prepara- flexible in operation and crystallization mode. In addition,
tion vessel must be electrically heated (availability the relation between process conditions and product quality
constraint, see Table A1). Purge streams are not necessary must be determined (such as impeller-crystal collisions).
in this pilot plant, because of the use of technical pure Both a DT and DTB crystallizer are equipped with an
solutions. However, in industrial cases this will be necessary impeller. Practical experience has shown that a fines removal
to avoid accumulation of impurities. The flow sheet diagram system is possible by crystallizer volumes >1 m3, which
is given in Figure 2.
At this design level the vapour, product and feed flows are
calculated with use of energy and mass balances. The
outcome of these calculations is listed in Table A9. In
industrial cases the product flow will be determined by the
required production capacity and the chosen crystal fraction.
However, in this case the product flow is determined by the
requirement of the lowest possible continuous product
removal (minimal velocity in product line, vmin) and
the chosen crystal fraction. Typically, industrial draft tube
baffled (DTB) crystallizers are designed to operate at
magma densities up to 25 wt%. However, operating
problems with the pilot plant crystallizer (e.g. blockage
product pipe and encrustation) are expected at magma
densities up to 25 wt%. Therefore the crystal fraction is
chosen to be 10% (e ¼ 0.9). This crystal fraction is equal to
the product solid concentration of an 1100 l DTB crystal-
lizer as described by Neumann (2001). The calculated flows
of the pilot plant crystallizer are listed in Table A9 for the Figure 2. Flow sheet diagram of the crystallization process.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
870 WESTHOFF et al.
Table 7. Design level IV—design of a crystallization stage.

Design specifications objectives Design and operational

and constraints variables Domain knowledge
Process Discrete
Production capacity þ Crystallizer type þ Equipment characteristics þ
Feed composition þ Circulation device þ Hydrodynamics þ
Crystallization method þ Fines classification and þ Fouling behaviour þ
Yield
dissolution=clear liquor advance Thermodynamic properties of species=component þ
Operation mode þ Product classification þ in solid, liquid and vapour phase
Pressure and temperature range þ Heat exchanger type þ Crystallization kinetics þ
Supersaturation range þ Shape factors of the crystalline compounds

Residence time (distribution)
Continuous
Heat exchange rates þ Equipment dimensions þ
No boiling in heat exchanger þ Feed location þ
No entrainment of þ Solids concentration þ
droplets by vapour Product removal location þ
Suspension criteria þ Circulation flow rate þ
SHE considerations þ Operating conditions of

Availability, controllability þ classification devices
and resiliency Flow rate through heat exchanger þ

Product
Crystal size distribution þ

were operated under vacuum. A 22 l crystallizer equipped surface area in the crystallizer, the draft tube can function
with a fines removal system showed practical problems as an internal heat exchanger and a jacket is mounted around
(Neumann, 1998, Private communication). To avoid these the crystallizer (external heat exchanger). The final crystal-
practical problems and to have a flexible, simple and robust lizer dimensions are summarized in Table 8.
crystallizer, a DT crystallizer was selected. At this stage, since the dimensions and conditions are
Equipment related equations or rules of a DT crystallizer known, a more sophisticated rigorous model could have
are listed in Table A7. At this stage, a design model for the been applied to predict and analyse:
sizing of the crystallizer unit must be chosen in which a
 product quality as function of processing conditions;
relation must be used which describes the dependence of the
 dynamic behaviour and stability;
kinetics and product quality on the process conditions in the
crystallizer. This relation is, however, far from established  controllability aspects.
(Kramer et al., 1999). Therefore here a very simple model However, a validated rigorous crystallization model was
was chosen, in which the growth rate was estimated based not available during the design period. It has to be noted that
on available kinetic data in literature. The consequence of one of the objectives of this multi-functional pilot plant
this choice is that the product quality is poorly predicted.
Recently, Bermingham (2003) has applied, in a first attempt,
a rigorous crystallizer model to find an optimal design for a
continuous process.
The constraints for the design are listed in Table A6. The
crystallizer volume and dimensions are calculated with
design equations listed in Table A7 and the additional
design equations (Table A8). Note: the flows are already
calculated at level III (flow sheet design; for the calculation
results see Table A9).
Some additional design equations are used to represent
the constraints (such as Pinlet, heat exchanger area, etc.). To
simplify the design calculations, the crystallizer is consid-
ered to be an ideally mixed vessel. Different iterative
calculations for the model systems are performed manually.
The procedure of the calculations is schematically given in
Figure 3.
The calculations are performed for different crystalliza-
tion modes; evaporative crystallization of (NH4)2SO4,
evaporative crystallization of KNO3, flash cooling crystal-
lization of KNO3 and adipic acid. The final results of these
calculations are listed in Table A9. The minimal required
hydrostatic pressure in the feed inlet during flash-cooling
crystallization of adipic acid mainly determined the height
of the crystallizer (the H=D0 ratio therefore changed from
typically 3=1 to 4=1). To create enough heat exchange Figure 3. Procedure of design calculations.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 871
Table 8. Final dimensions pilot plant crystallizer.

Variable Value
3
Volume (liquid), m 75  10
3
Internal diameter crystallizer, m 0.30
Internal diameter draft tube, m 0.20
Height draft tube, m 1.15
Diameter product removal pipe, m 8  10
3
Diameter marine-impeller, m 0.19

crystallizer is to generate experimental data for the valida-

tion and development of rigorous crystallization models.
The final crystallizer dimensions and the flowsheet design
Figure 4. The CCD camera and a laser-diffraction measurement device
are propagated to the detailed engineering phase of the (Helos) implemented onto the product line of the 75 l DT crystallizer.
crystallizer in the original procedure as proposed by
Bermingham et al. (2000). Since no attention was given to
the structure of the instrumentation and control of the
crystallization process in this procedure, a new design diffraction measurement device. By changing the flow of the
level V ‘Design of instrumentation and control’ was added crystal-free saturated solution (mother liquor) the solid
to structure the design. concentration can be changed. This system is schematically
shown in Figure 4.
The separate measurement loop has to be added to the
Design of Instrumentation and Control (Level V)
flowsheet design. If the crystallizer temperature equals the
At this level the control objectives, the possible measure- feed temperature, the feed flow can be used as a dilution
ments and available manipulated variables have to be stream. During flash and continuous cooling crystallization
identified. In addition the control structure has to be defined the feed flow has a higher temperature than the product flow
(see Table 9). and therefore the mother liquor from the feed vessel cannot be
The motivation for the marked specifications, variables used for the dilution flow for the product stream. In this case
and domain knowledge at this design level is given another dilution stream was created by the use of a buffer
in Table A10. The control loops are briefly described in vessel, which contained a saturated solution at the same
Table A10. A Honeywell S9000 DCS system using a temperature as the product stream. To avoid accumulation
SCAN3000 PC-based visualization tool is used to monitor of crystals in this buffer vessel, a solid–liquid separation is
and control the process variables (e.g. temperature, flow necessary before the diluted product stream returns to the
rates, pressure). The product stream contains approximately buffer vessel. The solids must flow back to the original buffer
10% crystals (e ¼ 0.9). For measurement of the crystal size vessel and the liquid flows back to the buffer vessel, where it is
distribution by a laser diffraction device, dilution of the used for the dilution stream. A schematic drawing of the
product flow was needed. A dilution system described by modified flowsheet design is given in Figure 5. (Note: some
Neumann et al. (1999) was implemented. The laser diffrac- detailed engineering information has already been included in
tion measurement device was situated in a separate measure- this drawing, it is not within the scope of the paper to present
ment loop. At regular time intervals the pneumatic valves here the full piping and instrumentation diagram, P&ID.)
were switched and part of the product stream was injected in The motivation for the solid–liquid separator and pumps
a crystal-free saturated solution. The diluted product sample is given in Table A11. The final flowsheet and crystallizer
flows through the optical cells of a CCD camera and a laser unit design (e.g. type of flow controllers, pressure valves,

Table 9. Design level V—design of instrumentation and control.

Design specifications objectives

and constraints Design and operational variables Domain knowledge
Process conditions Discrete Measurement principles þ
Pressure þ Controlled variables þ Actuators þ
Temperature þ Sensors (type) (includes measurement devices) þ Feed back control þ
Liquid levels þ Sensor locations þ Process dynamics þ
Concentration (supersaturation) þ Manipulated variables þ
Actuator type þ
Product and feed stream Control structure þ
Flow (slurry) þ Loop pairing þ
Slurry density þ
Temperature þ Continuous
Control tuning þ
Crystalline product Set point=ranges þ
CSD þ
Filterability þ
Inclusions

Dryability

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
872 WESTHOFF et al.

Figure 5. Flow scheme of the 75 l DT crystallizer.

A=D converters, etc.) is propagated to the detailed engineer- instrumentation, measurement devices, etc. A schematic
ing phase. This will not be discussed in this paper. drawing of the experimental set-up is shown in Figure 5.
A Watson-Marlow1 hose pump pumped a continuous
product stream out the crystallizer. The pump was placed
DESCRIPTION OF THE PILOT-PLANT 3 m below the liquid level of the crystallizer to obtain
CRYSTALLIZER SET-UP
Description of the 75 l DT Crystallizer
A schematic drawing of the 75 l DT crystallizer is given
in Figure 6. The bottom of the DT crystallizer has a
geometry that increases the suspension flow speed by 10%
when it passes the bottom (see Figure 6). This reduces
possible sedimentation of crystals at the bottom. The draft
tube has a slide with five holes (which can be individually
closed) at different heights. At the same heights are also five
holes in the ‘body’ of the crystallizer. This construction
enables the profiles to be mesasured (for instance super-
saturation and suspension velocity) inside the draft tube of
the crystallizer. To obtain enough heat exchange surface
area in the crystallizer the draft tube performed as a heat
exchanger (internal heat exchanger) and a jacket was
mounted around the crystallizer (external heat exchanger).
The effective height of both heat exchangers was 850 mm.
The feed point of the crystallizer was located under the
marine impeller to enable proper mixing of the inlet stream
with the suspension in the crystallizer. Four baffles were
mounted in the draft-tube (length ¼ 1013 mm, diame-
ter ¼ 25 mm). The centre of the product removal pipe was
located in the downstream zone, 20 mm from the wall of the
body of the crystallizer.

Plant Description
The crystallizer was integrated in an experimental set-
up with pumps, recycle streams, buffer vessels, piping, Figure 6. Seventy-five litre batch=continuous DT crystallizer.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 873

enough suction pressure. The dilution stream (saturated and gives an approximation of the infrastructure-related costs of
crystal free) and the product stream were collected in a 200 l one typical 8 h experiment.
buffer vessel. A pipe connected the buffer vessel and the
200 l feed preparation vessel. The dilution solution and the
feed flow were pumped out of the feed vessel during
EXPERIMENTAL VALIDATION OF THE
crystallization (without flashing).
DESIGN (MODEL)
The feed and dilution pumps were respectively a mem-
brane pump and a Watson-Marlow1 hose pump. Millipore1 The design has been based on a rather simple crystallizer
filters (pore size 5 mm) were built in the feed line, to create a model. Therefore the experimental median crystal sizes
solid=dust-free flow for both streams. The filter system was measured during crystallization experiments in the realized
mounted twice, so that during continuous operation the experimental set-up were compared with the specified
filters could be changed. The possibility of rinsing every median crystal size to test the reliability of the design
part of the system independently was included in the design. procedure used in the crystallization model. During commis-
This enabled blockages in the lines to be removed during sioning, several implementation errors were corrected. After
the crystallization process. All the rinsing points were commissioning, several successful crystallization experi-
connected with a rinsing system, which contained water at ments were performed in this experimental set-up.
a controlled temperature. For illustration purposes only, the evaporative crystal-
lization of potassium nitrate–water system is briefly
discussed. Figure 7 shows the measured quantiles of a
OPERATION COSTS selected experiment.
The crystal size distribution became stable after 2 h
The cost prices of the different components are sum- operation. In the first hour of operation the laser-diffraction
marized in Table 10. The total construction costs of the measurement device gave some errors (due to air bubbles
described experimental set-up were E291,500. The main- in the dilution loop) resulting in a 20 min gap in the data.
tenance costs (replacement defect components etc.) for The crystals produced during this crystallization experiment
each year are expected to be around E5000. Normally, are rounded-off (see Figure 8). The rounding-off of the
one would make an economical evaluation by estimating crystals was caused by attrition.
the annualized capital cost and the operational cost and
comparing these with the income generated from product
sales. In this case the objective is to gain information.
Therefore, it seems more relevant to estimate the cost per
typical experiment.
A rough estimate is made of the price of a typical experi-
ment of 8 h. Assuming 12 weeks of operation during a year
and a total crystallizer lifetime of the crystallizer of 10 years,
the total working hours of the crystallizer is 4800 h. The cost
of a typical 8 h experiment can be estimated by summing the
costs of building and maintenance over the lifetime of the
crystallizer and is E2846. Of course, the real cost of an
experiment is higher, because the costs of operators, chemi-
cals and energy were not taken into account. However, it

Table 10. Cost estimation of experimental set-up of 75 l pilot Figure 7. Measured quantiles (L10, L50 and L90) during continuous crystal-
plant crystallizer. lization of potassium nitrate-water system (process conditions: Q ¼ 6.8 kW;
stirrer rate ¼ 600 rpm, t ¼ 4000 s).
Components Price (E)
Extension of existing data acquisition 11,500
and control system
Crystallizer body 16,100
Storage vessels þ condensor 6400
Piping þ valves 28,000
Pumps 12,000
Stirrers þ impellers 3200
Solid–liquid separator 2500
Level and pressure control valves 3400
Flow meters 5500
Density and pressure meters 7300
Level indicators in storage vessels 2700
Electricity components (PLC, frequency 8900
controllers, cables etc.)
Labour costs of mechanical and electronic 180,000
workshop þ design costs
Other costs 4000
Total costs 291,500
Figure 8. SEM image of KNO3 crystals at t ¼ 3 h.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
874 WESTHOFF et al.

The median crystal size (Lmedian  600 mm) used in the V crystallizer volume, m3
v velocity, m s
1
design calculation was much higher than the measured
vsup superficial velocity, m s
1
steady-state Lmedian of approximately 300 mm. This could vtip impeller tip velocity, m s
1
be due to a number of reasons: use of a very simple
crystallizer model (simple kinetic relations); assumption of Greek symbols
an ideally mixed crystallizer at the design calculations; or a heat transfer coefficient, J (m2 s
1)
D difference
sensitivity of potassium nitrate for attrition (crystal–impeller e liquid fraction
collisions). This deviation between the predicted and r density, kg m
3
measured steady-state median crystal size emphasizes the t residence time, s
need for rigorous crystallization models. Predictive model- fv volumetric flow, m3 s
1
ling by implementation of the final crystallizer dimensions FHvapour energy of vapour flow, kW
in a rigorous crystallization model at design level IV would Subscripts
have given more precise information about the expected circ circulation
median crystal size. crys crystallizer
imp impeller
max maximum
CONCLUSION min minimum
prod product
A multi-functional pilot plant crystallizer has been vap vapour
designed, constructed and tested. The systematic design vol volume
procedure according to Bermingham et al. (2000) for the
design of this crystallizer was found useful.
This procedure structures the design process by delivering REFERENCES
a systematic list of functional requirements, which are Bennett, R.C., 1993, Handbook of Industrial Crystallization, Meyerson,
subsequently translated into a structure and size for the A.S. (ed.) Series in Chemical Engineering (Butterworth-Heinemann,
crystallization unit, using a heuristics-based and computa- London, UK).
tionally straightforward design model. Furthermore, the Bermingham, S., 2003, A design procedure and predictive models for
solution crystallisation processes, Development and application, PhD
procedure creates awareness of the areas where more Thesis, Delft University of Technology, The Netherlands.
domain knowledge has to be obtained and allows for early Bermingham, S.K., Neumann, A.M., Kramer, H.J.M., Verheijen, P.J.T.,
inclusion of operational aspects, like safety and controll- van Rosmalen, G.M. and Grievink, J., 2000, A design procedure and
ability, at an early stage. Since any plant requires control, a predictive models for solution crystallisation processes, AIChE Symp Ser,
new design level V, ‘Design of instrumentation and control’, 323(96): 250–264.
Bernstein, J., Davey, R.J. and Henck, J.-O., 1999, Concominant polymorphs,
has been added to the design procedure developed by Angew Chem Int Edn, 38: 3440–3461.
Bermingham et al. (2000). At this level the control objec- Jansens, P.J., Bruinsma, O.S.L. and van Rosmalen, G.M., 1995, Optimization
tives, the possible measurements and available manipulated and modelling of the crystallization process in a cascade with backmixing,
variables are identified, and control structures are selected. AIChE J, 41(4), 828–837.
Kramer, H.J.M., Bermingham, S.K. and van Rosmalen, G.M., 1999, Design
The designed pilot plant was actually built and put into of industrial crystallisers for a given product quality, J Crystal Growth,
effective operation for crystallization experiments. However, 198=199: 729–737.
the deviation between the median crystal size, specified in Mersmann, A., 1995, Crystallisation Technology Handbook (Marcel
the design, and the measured one in this pilot crystallizer Dekkers, New York, USA).
emphasizes the need for better predictive crystallizer Mullin, J.W., 1993, Crystallization, 3rd edn (Butterworth-Heinemann,
Oxford, UK).
models, to exploit the potential for further analysis and Neumann, A.M., 2001, Characterizing industrial crystallizers of dif-
optimization of design. ferent scale and type, Ph.D. thesis, Delft University of Technology,
The Netherlands.
Neumann, A.M., Kramer, H.J.M., Zhenhua, M. and Scarlet, B., 1999,
NOMENCLATURE On-line measurement techniques for industrial crystallization processes,
in Crystallisation Conference, Cambridge.
A cross sectional crystallizer area, m2
Ó Meadhra, R., 1995, Modelling the kinetics of suspension crystallizers, a
Ahex,DT internal heat exchanger area, m2
new model for secondary nucleation, Ph.D. thesis, Delft University of
DDT draft tube diameter, m
Technology, The Netherlands.
DDTclearance distance between draft tube wall and impeller, m
Ouchiyama, N. and Tanaka, T., 1984, Porosity estimation for random
Dimp impeller diameter, m
packings of spherical particles, Ind Eng Chem Fundam, 23: 490–493.
D0 draft tube outer diameter, m
Perry, R.H. and Green, D.W., 1998, Perry’s Chemical Engineers’ Handbook,
Dprod.pipe diameter product pipe, m
7th edn (McGraw-Hill, New York, USA), pp 18.44–18.51.
g gravitation constant, m s
2
Reid, R.C., Prausnitz, J.M. and Poling, B.E., 1998, The Properties of Gases
G growth rate, m s
1
and Liquids, 4th edn (McGraw-Hill, New York, USA).
h height of liquid level in crystallizer, m
Standiford, F.C., 1998, Perry’s Chemical Engineers’ Handbook, 7th edn
hDT height of draft tube, m
(McGraw-Hill, New York, USA), pp 11–111.
L crystal size, m
N impeller frequency, s
1
Nv pump discharge coefficient
P pressure, bar ACKNOWLEDGEMENTS
Pinlet inlet pressure, Pa
ProdRate crystal production rate, kg s
1 The authors would like to acknowledge the Dutch Foundation of
Pm permeability, m Technology (S.T.W.), Akzo Nobel, BASF, Bayer, Dow Chemicals, DSM,
Pvap vapour pressure, Pa E.I. Dupont de Nemours and Purac Biochem for supporting the UNIAK-III
Q heat input, kW research program.
T temperature, K The manuscript was received 2 April 2003 and accepted for publication
t time, s after revision 14 April 2004.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 875

APPENDIX: MOTIVATION OF CHOICE AND RESULTS OF THE DESIGN LEVELS

Table A1. Level 0—Initial design specifications.

Product performance requirements

Product=model compound selection:
 multi-purpose—crystallization of potassium nitrate–water system, ammonium sulphate–water system, and adipic acid–water system
In the crystallizer:
 no flotation
 suspendability (off-bottom criterion satisfied)
Downstream handling:
 filterability (research subject, no product filterability range specified)
Process requirements
 Production capacity—lowest possible continuous product removal [minimal product line diameter: 8 mm, minimum product velocity: 0.5 m=s (to avoid
settling, heuristic knowledge), crystal content  10 vol%]
 Controllability—easy to operate by one person, easy start-up and shut down procedures, no fines removal system (fines removal on small scale will give
operation problems)
 Availability—cooling water, electricity, high pressure air, no steam supply available
 Safety Health Environment—open laboratory space, which limits temperature, non-toxic, no explosion risk, low noise level, closed loop for all process
streams, emergency stop
 Resilience—the process must be robust and reliable
 Research requirements—multi-purpose, different operation modes, batch=continuous.
Battery limits: excluded utility systems. Maximal flexibility concerning pH-range, pressure, temperature and residence time. The I=O diagram of the
crystallizer is given in Figure A.1.

Figure A1. I=O diagram of the crystallizer.

Table A2. Design choices of the crystalline product and their motivation (design level I).

Model systems
Design
specifications Adipic acid–H2O solution (NH4)2SO4–H2O solution KNO3–H2O solution
No flotation Satisfied (rcrystal > rsaturated solution) Satisfied (rcrystal > rsaturated solution) Satisfied (rcrystal > rsaturated solution)
Filterability Slurry good filterable Slurry good filterable (dependent of CSD), Slurry good filterable (dependent of
(dependent on CSD) no needles expected CSD), no needles expected
SHE: explosion No No Yes, mixtures of organic material
risk with dry KNO3
SHE: environment No disposal to sewage (closed loops) No disposal to sewage (closed loops) No disposal to sewage (closed loops)
Discrete
Crystal size Narrow crystal size distribution relevant Narrow crystal size distribution relevant Narrow crystal size distribution relevant
distribution for filterability and dryability for filterability and dryability for filterability and dryability
(Lmedian  350 mm, Lmax  3  Lmedian) (Lmedian  550 mm, Lmax  3  Lmedian) (Lmedian  600 mm, Lmax  3  Lmedian)
Polymorphs No No No
Continuous
Crystal purity — Metal ions can have an influence on Metal ions can have an influence on
crystallization kinetics crystallization kinetics
Morphology Monoclinic prims Rhombic Orthorhombic

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
876 WESTHOFF et al.
Table A3. Motivation of the physical=chemical design of the crystallization task (design level II).

Model systems
Design specifications
objectives and constraints Adipic acid–H2O solution (NH4)2SO4–H2O solution KNO3–H2O solution
Process
production capacity See Table A1 See Table A1 See Table A1
feed composition No dust, impurity free (use of No dust, impurity-free (use of No dust, impurity-free (use of
technical pure solutions), technical pure solutions), technical pure solutions),
saturated feed saturated feed saturated feed
yield See Table A1 See Table A1 See Table A1
availability See Table A1 See Table A1 See Table A1
SHE considerations No No See Table A2
Product
polymorphism See Table A2 See Table A2 See Table A2
morphology See Table A2 See Table A2 See Table A2
purity See Table A2 See Table A2 See Table A2
Design and operational variables
Crystallisation method Evaporative=flash cooling= Only evaporative crystallization Evaporative=flash cooling=batch
cooling crystallization useful due to flat solubility line cooling crystallization
Feed purification Filtration necessary to remove Filtration necessary to remove Filtration necessary to remove
dust, crystals etc. dust, crystals etc. dust, crystals etc.
Material of Easy to modify and resistant Easy to modify: Easy to modify: SS 304 sufficient
construction against solution with pH ¼ 2.5: SS 304 sufficient
stainless steel 316 necessary
Continuous
pressure range 0.1–1 bar (dependent on Vacuum (0.1 bar) 0.1–1 bar (dependent on
crystallization method) crystallization method)
temperature range Tmax ¼ 65 C (safety constraint), Tprocess ¼ 50 C Tmax ¼ 65 C (safety constraint),
Tcrystallizer, min ¼ 25 C (Tevaporation of model Tcrystallizer, min ¼ 25 C
(Tcrystallizer min ¼ Tmax, cooling water compound system (Tcrystallizer, min ¼ Tmax, cooling water
(maximum cooling water at 0.1 bar) (maximum cooling water
temperature in summer) temperature in summer)
Domain knowledge
Thermodynamic Solubility curve, heat of Solubility curve, heat of Solubility curve, heat of
properties crystallization, heat crystallization, heat crystallization, heat
capacity solution capacity solution capacity solution
Scaling or High tendency, crystallizer measures Low tendency Low tendency, in case of
encrustation needed: rinsing system at boiling cooling crystallisation:
tendency of zone level. In case of cooling DTheat exchangers  2 C
component species crystallization, DTheat exchangers  2 C to avoid encrustation
to avoid encrustation
Meta-stable zone — 1.5% supersaturation 1.5% supersaturation
Safety aspects No dangerous situation expected in No dangerous situation No dangerous situation
T and P ranges expected in T and P ranges expected in T and P ranges
Corrosion See material of construction See material of construction See material of construction

Table A4. Motivation of the flow sheet design of the crystallization process (design level III).

Process Motivation and criteria

Design specifications objectives and constraints
Production capacity See Table A1
Feed composition See Table A3
Crystallization method See Table A3
Pressure range 0.1–1 bar (dependent on crystallization mode)
Temperature range 25 C < Tcrystallizer < 65 C (see Table A3) (dependent on crystallization mode)
Supersaturation range See Table A3
Availability, controllability and resiliency See Table A1
SHE considerations See Table A3
Product
crystal size distribution See Table A2
Design and Operational variables
Discrete
operation mode See Table A3
number of stages Single stage chosen. Multiple stage is not necessary and has higher investment costs.
feed configuration Saturated solution of the model compound, impurity-free (see Table A3)
recycle structure No recycling of mother liquor. Note: no dry crystals are produced, so recycling of crystal slurry
into buffer vessel together with the condense stream of the evaporated solvent. Crystals can
dissolve and a saturated feed solution can be prepared again
Recycle structure of cooling water: cooling water back into cooling water system, towards
cooling towers
(continued)

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 877
Table A4. Continued.

Process Motivation and criteria

Continuous
residence time in each stage or batch time Residence time is dependent on different aspects: growth rate of the used model system, the
minimal desired crystal content concentration in the crystallizer and the desired median
crystal size
recycle flow rates Minimal product flow velocity is 0.5 m s
1 to avoid settling in the pipes (heuristic knowledge).
(Note: in this application the product flow is recycled.) Maximum cooling water flow rate is
2 m s
1 (heuristic knowledge)
pressure and=or temperature in each stage Evaporative crystallization: T ¼ 50 C, P ¼ 100 mbar
Cooling crystallization: P ¼ atmospheric pressure, 25 C < T < 60 C
Flash cooling crystallization: T ¼ 50 C, P ¼ 100 mbar
heating=cooling duty or trajectory Cooling duty trajectory is dependent of required production capacity and heat exchangers
constraints (maximal DTheat exchanger ¼ 2 C, to avoid encrustation). Maximum cooling
capacity for crystallizer approximately 3 kW (value chosen based on assumption if heat
in case of flash cooling crystallization is not removed via condensers, but via heat
exchangers in the crystallizer)
heat exchange rates Maximal velocity through heat exchanger is 2 m s
1. The heat transfer coefficient dependent
on model compound system and the construction material of heat exchanger (first estimate:
a ¼ 1000 W m
2 K
1)
Domain knowledge
crystallization kinetics, i.e. rate O’Meadhra (1995) gives an expression for the growth in 22 l DT crystallizer operated at 50 C
expressions for growth of crystals for adipic acid–H2O system and (NH4)2SO4–H2O system
fouling kinetics Relevant in case of crystallization of adipic acid. Use of PFA tubing will reduce fouling=
encrustation of product line
fire and explosion index See Table A2

Table A5. Motivation of the design of a crystallization stage (design level IV).

Process Motivation
Design specifications objectives and constraints
Production capacity See Table A4
Feed composition See Table A4
Crystallization method See Table A4
Yield See Table A4
Operation mode See Table A1
Pressure and temperature range See Table A4
Heat exchange rates See Table A4
No boiling in heat exchanger Maximum temperature water <100 C
No entrainment of droplets by vapour Maximum vapour velocity dependent on solution density (see also Table A6, for equation for vvap,max)
Suspension criteria See Table A6
SHE considerations See Table A4
Availability, controllability and resiliency See Table A4
Product
crystal size distribution See Table A4
Design and operational variables
Discrete
crystallizer type See notes at design level IV
circulation device Circulation device will be an impeller with a high pumping number (four blade marine-impeller).
The model compound KNO3 is sensitive especially for attrition.
fines classification and dissolution= No fines classification. On small scale a fines removal will lead to an
clear liquor advance difficult controllable crystallizer (heuristic knowledge)
product classification In industrial cases, product classification can be useful to obtain required
product specifications
heat exchanger type Condensor: plate heat exchanger (high capacity, low volume,
low investment costs, easy expandable)
Crystallizer: crystallizer has to be equipped with a
double wall for heat input or application cooling profiles.
Continuous
feed location Near the stirrer to create good mixing with the suspension
solids concentration See Table A1
product removal location In the down-coming zone of the draft tube
circulation flow rate High enough to create MSMPR conditions
flow rate through heat exchanger See Table A4
Domain knowledge
equipment characteristics DT design constraints are given in Table A6
crystallization kinetics See Table A4.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
878 WESTHOFF et al.
Table A6. Constraints of conceptual pilot plant crystallizer design.

Inequality constraint Requirements

(1 7 eprod,max) (0.3) (1 7 eprod) (1 7 eprod,min) (0.1) Comparable with industrial conditions
Lprod,max (2000 mm) Lprod Lprod,min (50 mm) Well measurable
s  smax (1.5%) Avoid primary nucleation, avoid kinetic roughening
Tcrys,max (60 C) Tcrys Tcrys,min (25 C) Safety, operability
Tfeed,max (60 C) Tfeed Tfeed,min (25 C) Operability, flashing
vtip,max vtip Prevention of excessive attrition
Ahex Ahex,min Prevention of scaling
DDT,max DDT D0 Suspension criterion in DT
D0 D0,min Prevent entrainment of liquid droplets (vvap,max vvap)
Dprod.pipe,max Dprod.pipe Dprod.pipe,min Upper bound: suspension in product line (vsup,prod vsup,min)
Lower bound: no blockage
Pinlet > Pflash To avoid flashing in feed line
(D02 7 DDT2) DDT2 Geometric constraint
Additional equations
p for constraints
Dprod.pipe,max ¼ p (4  fv,prod)=(p  vsup,min)(m) Geometric consideration
DDT ¼ 0.5  D0  2 (m) Geometric consideration
DTclearance ¼p4  Lmedian (m) Geometric consideration
DDT,max ¼p (4  fv,DT)=(p  vsup,min) (m) Geometric consideration
D0,min ¼ (4  fv,prod=(p  vvap,max)) (m) Geometric consideration
Dprod.pipe,min ¼ p1  Lmedian,max (m) p1 ¼ 4 (avoid blockage)
vsup,min ¼ p2 (m s
1p ) p2 ¼ 0.5 (suspension criteria)
vvap,max ¼ 0.0244  ((rliquid 7 rvap)=rvap) (m s
1) Souders and Brown equation (Standiford, 1998)

1
vtip,max ¼ p4 (m s ) p4 ¼ 7 (avoid excessive attrition, heuristic knowledge)
Lmedian,max ¼ p3  Lmedian (m) p3 ¼ 3 (narrow CSD)
Tcrys,min ¼ f (P) lowest feasible pressure (K) —
Tfeed,max highest temperature to avoid boiling at inlet point (K) —
Ahex,min ¼ Q=(amin  DTmin) (m2) a ¼ 1000 W m
2 K
1
Pflash ¼ P(T) P(T) ¼ antoine equation

Table A7. Equipment-related design equations for a DT crystallizer.

Equations Dimensions Nature of equations

Lmedian ¼ 3.67  G  t m Benett (1993)
G ¼ f (T,s) m s
1 Literature, see Table A4
Vcrystallizer ¼ fv,prod  t m3 —
ProdRate ¼ fv,prod p  (1 7 e)  rcrystal kg s
1 —
DDT ¼ 0.5  D0  2 m Geometrical considerations
DTclearance ¼ 5  Lmedian m Empirical knowledge
DDT 7 Dimp ¼ 2  DDtclearance m Heuristic knowledge
vsup,prod ¼ vsup,min m s
1 —
h ¼ 4  D0 (typically draft tube: h ¼ 3  D0) m Empirical knowledge
fv,DT ¼ Nv Dimp3  N m s
1 —
vsup,DT ¼ fv,DT=[(p=4)  DDT2] (informative) m s
1 —
vtip ¼ p  Dimp  N m s
1 —
fv,prod ¼ vsup,prod  Aprod.pipe m3 s
1 Geometrical consideration
Dprod.pipe ¼ 8  10
3 m See Table A1

Table A8. Additional design equations.

Design equations Dimensions Nature of equations

2 2
A ¼ (p=4)D0 m —
Ahex,DT ¼ p DDT  hDT  2 m2 —
tcirc ¼ V=fv,DT s —
Aprod.pipe ¼ (p=4)Dprod.pipe2 m2 —
Pinlet ¼ Pvap þ rgDh Pa —
T ¼ 45.15 þ 1668.21= K Reid et al.
{7.96681 7 log[P(760=1.013)]} (1988)
(Antoine equation)

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
 MULTI-FUNCTIONAL CRYSTALLIZER 879
Table A9. Calculated values for the different crystallization modes.

Evaporation crystallization Flash cooling Evaporation crystallization Flash cooling of

of KNO3 of KNO3 of (NH4)2SO4 adipic acid
Vcryst (m3) (liquid) 72  10
3 72  10
3 90  10
3 76  10
3
hcryst (m) (liquid) 1.14 1.14 1.23 1.2
D0 (m) 0.284 0.284 0.306 0.289
DDT (m) 0.201 0.201 0.217 0.204
Dimp (m) 0.195 0.195 0.211 0.201
Dprod.pipe (m) 7.1  10
3 7.1  10
3 7.7  10
3 8.9  10
3
DTclearance (m) 0.003 0.003 2.75  10
3 1.75  10
3
Tcryst ( C) 50.2 50 50 47.3
Tfeed ( C) 50 62 50 63
Qheatexch.(kW) 11.5 0 11.2 0
e 0.9 0.9 0.9 0.9
Prod Rate (kg s) 4.2  10
3 4.2  10
3 4.2  10
3 4.2  10
3
Lmedian (m) 600  10
6 600  10
6 550  10
6 350  10
6
G (m s
1) 4.1  10
8 4.1  10
8 3.6  10
8 3.6  10
8
Fv,feed (m3 s
1) 2.5  10
5 2.15  10
5 2.9  10
5 3.3  10
5
Fv,product (m3 s
1) 2  10
5 1.98  10
5 2.36  10
5 3.1  10
5
Fv,vap (m3 s
1) 0.043 6.8  10
3 0.042 1.0  10
4
FHvapour (kW) 13.1 2.1 12.8 3.0
t (s) 3.7  103 3.7  103 3.8  103 2.4  103
tcirc (s) 2.4 2.4 2.6 2.4
N 685 685 633 666

Table A10. Design of instrumentation and control (design level V).

Design specifications objectives and constraints

Process Pressure is used: to calculate the liquid level [pressure in vapour zone (Pvap) þ pressure at certain place below
pressure liquid level]; to control the crystallizer temperature (see control structure and loop pairing)
temperature Measurement of suspension temperature in the crystallizer at several locations (check for presence of temperature
profiles, control of crystallizer temperature, determination solubility concentration) measurement of tempera-
tures in hot=cold water, buffer, feed preparation vessel
liquid level Pressure is used to calculate the liquid level in the crystallizer (see pressure)
Liquid levels are measured in buffer, cold=hot water, feed preparation vessels
concentration (supersaturation) Solution concentration is measured to determine supersaturation level in crystallizer
Product=feed stream
flow Measurement of product flow needed for control of production rate
slurry density Density measurement of product flow to determine the liquid fraction
temperature Measurement of feed, crystallizer and product temperature
Crystalline product
crystal size distribution Research objective (determine relation between process conditions and product quality (e.g. crystal size
distribution and filterability)
Measurement with laser-diffraction measurement device
Laser-diffraction measurement device can measure until 1% crystal content; therefore dilution of product stream
is necessary.
filterability Research objective (relate product quality-process conditions), in-line filterability measurement device has to be
developed. Off-line measurement of filterability.
Design and operational variables
controlled variables Feed temperature, crystallizer temperature, crystallizer liquid level, temperature cold=hot water vessel, tempera-
ture of feed preparation and buffer vessel, hot=cold water flow to external=internal heat exchanger, suspension
criteria (sufficient mixing)
sensors (type) Pressure measurement, temperature measurements with PT-100, density meter and flow meters, data acquisition
(of flows, temperatures, pressure and density measurements), in-line measurement device for concentration
determination, off-line measurement of crystal slurry filterability, in-line video images of crystal slurry product
stream with a CCD camera
sensor locations Dependent of heuristic knowledge, important for accurate measurements (e.g., temperature probes at inlet
and outlet of internal and external heat exchanger; pressure sensors below liquid level in crystallizer and
in vapour zone)
manipulated variables Feed flow, product flow, hot=cold water flow towards internal=external heat exchanger; pressure; electrical heat
input in hot=cold water vessel; feed preparation vessel and buffer vessel; crystallizer impeller frequency
control structure and loop pairing Here the connections between the controlled variable and manipulated variable have to be described
Pressure and temperature control loop of the crystallizer: at evaporative and flash cooling crystallisation the
temperature of the crystalliser is controlled by a master–slave-loop from the measured pressure in the top part
of the crystallizer and the solution=suspensions temperature in the system controlling the valve near the vacuum
pomp
In case of cooling crystallization the temperature of the crystallizer slurry is controlled by the valve in the coolant
flow towards the internal and external heat exchanger in the crystallizer
During batch cooling crystallization, the set-point temperature profile is controlled according to a certain profile
Level control loop: the liquid level in the crystallizer is controlled by the feed flow. The liquid level is determined
from the pressure difference, which is measured with two pressure sensors placed respectively in the top part
and halfway the crystallizer
(continued)

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880
880 WESTHOFF et al.
Table A10. Continued.

Heat input control loop: the heat input in the crystallizer during evaporative crystallization is controlled with the
valve of the hot water flow towards the internal and external heat exchanger
Product flow loop: product flow controlled by product pump rate
control tuning Tentative empirical settings, but relevant for process dynamics
set point=ranges Description of initial setpoints and ranges of manipulated variables (e.g. position of valves, ranges of valves,
pumping rates, stirrer frequency etc.)
Domain knowledge
measurement principles Thermocouples, manometers, conductivity measurement, laser diffraction
actuators Pump rates (product flow), valves (feed flow, cold=hot water flow to internal=external heat exchangers), electrical
heat input (storage vessels) pressure (vacuum release valve)
feed-back control Use of feedback control (PID) for liquid level, product flow, crystallizer temperature, pressure
process dynamics Analyses of process dynamics with a process model (beyond the scope of this paper)

Table A11. Motivation of auxiliary equipment.

Device
Membrane pump No damage to crystals and no damage to pump by blockage of product line
Hose pump No damage to crystals, easy to replace and to remove blockages. Disadvantage is the damage
to the hose by blockage of flow
Solid–liquid separator There are different options for the solid–liquid separation, such as hydro cyclones, centrifuges
and settling tanks. The settling tank is chosen for solid–liquid separation, because the
investment costs for the hydro cyclone and the centrifuge are too high. Also higher slurry
flows are needed for a proper operation of the hydro cyclone and the centrifuge

Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 865–880