Abstract:
In this work, Response Surface Methodology (RSM) an
d Artificial Neural Networks
(ANN) were employed for the data reduction of a num
erically simulated extraction
process of sugar in an industrial RT2 extractor. Th
e numerical model developed in
OpenFOAM library was first validated using actual p
lant data and its stability and
sensitivity to the processing variables was tested.
Then, the model was used to generate
data of juice and pulp sugar concentrations as affe
cted by the main processing
parameters including draft, Silin number, and capac
ity. The data were modelled using
RSM and ANN. Both RSM and ANN were able to predict
the data accurately, however,
R
2
values obtained for ANN were slightly higher. Sinc
e the numerical model can be time
consuming to be solved for all data ranges, the reg
ression equation obtained by the RSM
method or the network created according to the ANN
model can be utilized as fast and
ready to use tools to optimize the extractor.
Keywords:
ANN, CFD, Mass transfer, Open FOAM, RSM.
