Sf. Deazevedo et al., MONITORING CRYSTALLIZATION .2. ON LINE MONITORING OF WHITE SUGAR CRYSTALLIZATION THROUGH SOFTWARE SENSORS, International Sugar Journal, 96(1141), 1994, pp. 18-26
In this paper we present the theoretical bases and experimental implem
entation of a methodology which allows, on-line and in real-time, the
characterization of the semi-batch evaporative crystallization process
in cane sugar refining. This methodology does not require the use or
installation of 'sensitive' sensors, such as consistency and total den
sity meters, or the theoretical knowledge of 'less accurately known' q
uantities such as crystal growth rates and evaporation rates. The algo
rithm proposed is based on a theoretical model of the crystallization
process which includes growth rate dispersion mechanisms and the chara
cterization of crystal size distribution (CSD) by its first six linear
moments. A transformation is performed which eliminates the growth ra
te and leaves the state estimation dependent only on measures availabl
e on-line. The latter are readings of the 'input' variables to the sys
tem and of three internal variables characteristic of the crystallizat
ion state. The 'input' readings are those related to - (i) liquors/syr
up feed: flowrate, brix, temperature and purity; (ii) steam; flowrate,
temperature and pressure; (iii) stirrer power: current. The 'internal
variables' required are the vacuum pressure, the brix of solution in
the massecuite and its temperature. These are all known to be reliable
measurements. The time trajectories of the following properties and i
ndices of performance are obtained, viz.-(i) supersaturation; (ii) tot
al mass and volume; (iii) mass fraction of crystals; (iv) CSD characte
rised by an average crystal size and a coefficient of variation; and (
v) energy consumption indices. As subsidiary results from the practica
l point of view, but most relevant for theoretical studies are the est
imates simultaneously obtained of the growth rate and evaporation rate
throughout the operation. The concept is now implemented at industria
l scale at RAR, undergoing experimental tests for robustness to noises
and sensor failure. Based upon this data-bank on the process operatio
n, which is now available, studies are being carried out both on alter
native, more fundamental, descriptions of the process behaviour and on
alternative forms of controlled operation.