circulation device. The draft-tube baffled and fluidised-bed crystallisers of Figures 8.4
and 8.5, respectively, are especially suited for the production of large crystals, as
secondary nucleation is reduced by fluidising the crystals.
Process ControlThe two main objectives for control of a dissolution-reaction-crystallisation process are
to meet: 1) the product specifications, and 2) the manufacturer’s requirements for
economic and trouble-free operation. In large-scale continuous industrial crystallisers,
this generally means that control is focused on preventing excessive nucleation by
making sufficient crystal surface area available in the bulk of the suspension. For
appropriate control, a model predicting the concentrations and crystal sizes of all
components at any place in the vessel as a function of measurable process variables and
other relevant processes, must be available.
In general, a model for a heterogeneous reaction crystalliser can be obtained by
integrating a kinetic model, containing the dissolution, reaction, and crystallisation
kinetics, and a hydrodynamic model, as mixing could cause dissolution, reaction, or
crystallisation to be rate limiting. Our kinetic model for solid salt-to-solid salt
bioconversions in a batch stirred bioreactor seeded with product crystals (Michielsen et
al., 1999a) can in principle also be applied for continuous solid-to-solid bioconversions.
For that, the model should be extended with an expression for the secondary nucleation
rate. Application of our crystal growth model (Michielsen et al., 1999b), which is based
on an exponential rate law derived by Nielsen and Toft (1984), revealed that it fails at
low supersaturation, predicting a lower growth rate then is observed experimentally
(Michielsen et al., 1999a). Such a shortcoming of exponential growth rate models was
also reported by Myerson and Ginde (1993). Our experimental data (Michielsen et al.,
1999a) could be predicted well by assuming that mixing was not rate limiting (see Figure
8.3). For that reason, it was unnecessary to develop a hydrodynamic model.
The measurable process variables used in the control of continuous crystallisers might
be temperature, flow rate, pressure, the residence time of different size ranges of crystals,
the total volume of the suspension in the crystalliser, and the volumetric ratio of the clear
liquor flow and the product removal rate. The relevance of processes like mixing and
reactants and seed addition with respect to appropriate control of a crystalliser is
described above. With respect to seeding, a narrow distribution of prewashed product
crystals is recommended. By prewashing the product crystals with supersaturated product
solution (comparable with the solution in the crystalliser), breeding due to the adherence
of small crystals to the surface of seed crystals can be avoided completely. Other
processes often applied in the control of the crystal-size distribution of continuous
crystallisers are classified product removal and fines dissolution. Classification relies on
the relative settling of crystals of different sizes. The classifying device may be a
hydrocyclone, a wet screen, a fluidised bed, or a centrifuge. It separates the suspension
flow withdrawn from the crystalliser in at least two fractions, containing crystals smaller
or larger than the separation size; the former fraction is generally recycled to the
crystalliser (Mersmann, 1995b). Formation of some fines by nucleation is almost
inevitable. These fines can serve as nucleation sites. For that, they are commonly
Solid-to-solid bioconversions 249