| |
Areas of interest and related main topics
The Process Systems Engineering Group
at ISR-Porto recognizes its work in a number of clusters of interest:
|
Dynamic modelling and simulation of chemical and
biochemical processes - exploring first principles models, knowledge
engineering methods and hybrid approaches to batch and semi-batch (fed-batch) processes
|
|
Monitoring and control of chemical and biochemical
processes - development and implementation of model-based state observers
(software sensors) and of model-based and supervisory control methods for industrial batch
and semi-batch (fed-batch) processes.
|
|
Optimisation strategies for Non-Linear Programming
(NLP) and Mixed-Integer Non-Linear Programming (MINLP) problems -
exploring the coupling of simulated annealling with direct search methods for robust
global optimisation. Development of a robust optimized equation-oriented simulator for
global NLP and MINLP optimization.
|
|
Optimisation strategies for gas cleaning devices
- exploring numerical optimization for the development of efficient gas cleaning devices,
such as VHE (very high efficiency) cyclones and electrostatic precipitators.
|
|
Methods of knowledge engineering for data mining -
use of neural networks, fuzzy systems and hybrid methods for extracting hidden knowledge
from process data, acquired during process operation and stored in process databases. |
|
Image analysis methods for characterisation of
particle morphology - application to crystallisation and precipitation
processes. |
Current projects
For the 3-year period of
1999-2001, eight projects or
topics of work are in progress. They involve the development of theory, of methods and of
tools, within a frame of postgraduate education and training, with a concern of
cooperation at international level and with industrial partners.
The following most relevant aspects should be mentioned (project
codes as in the official report of ISR):
|
PSE_1 - Knowledge engineering methods for process
modeling and control (S. Feyo de Azevedo)
|
The academic research area for developing models,
methods and basic technologies for representing and processing knowledge and for building
intelligent knowledge-based systems, is called knowledge engineering. This is seen by a
school of thought ( Kasabov, 1996; Kosko, 1992) as a part of the AI area, directed more
towards applications. Knowledge Engineering includes both symbolic and numerical
frameworks as well as both structured and unstructured forms of knowledge representation.
The objective in this project is to develop and apply
hybrid networks, concerning numerical frameworks (fuzzy and neural network solutions) in
conjunction with classical forms of knowledge representation (based on the so-called
'first-principles methodologies), to the modelling, optimisation and control of chemical
and biochemical processes. It is known that 'first-principles' knowledge alone cannot
model all existing knowledge of a process and it is believed that through hybrid
structures significant improvement can be achieved in process knowledge representation.
From a theoretical point of view, the group will focus
on problems of controllability and stability, and on how to define adaptive (dynamic)
hybrid systems.
From the applications point of view, hybrid structures
will be applied to the modeling and operation of both chemical (crystallisation) and
biochemical (fermentations) industrial processes.
The following is relevant literature concerning this
line of work -
- Kasabov, N.K., Foundations of Neural Networks, Fuzzy
Systems and Knowledge Engineering, The MIT Press, Cambridge, London, 1996
- Kosko, B., Neural Networks and Fuzzy Systems - a
Dynamical Systems Approach to Machine Intelligence, Prentice Hall, Englewood Cliffs, NJ,
USA, 1992
- Thomson, M.L., M.A. Kramer, Modeling chemical processes
using prior knowledge and neural networks. AIChE J., 40, pp. 1328 - 1340, 1994
- Schubert, J., R. Simutis, M. Dors, I. Havlik, A.
L�bbert, Hybrid Modelling of Yeast Production Processes Combination of a priori
knowledge on Different levels of Sophistication. Chem Eng. Technol., 17, pp. 10-20, 1994
- Simutis, R., R. Oliveira, M. Manikowski, S. Feyo de
Azevedo, A. L�bbert, How to increase the performance of models for process control and
optimization. J. Biotechnol., 59, pp. 73-89, 1997
- Feyo de Azevedo, S., B. Dahm, F.R. Oliveira, Hybrid
Modelling of Biochemical Processes: A comparison with the conventional approach. Computers
chem. Engn, 21, Suppl., pp. 751-756, 1997
|
PSE_3 - Characterisation of agglomeration and
crystallisation kinetics in industrial processes through image analysis (S. Feyo de
Azevedo), and
PSE_7 - Modelling and optimisation of precipitation processes in the chemical industry (F.
Rocha)
|
This is a long term cooperation between the two
scientific responsibles, which aims at detailed characterisation of basic mechanisms of
crystallisation processes. It further includes the most valuable cooperation of Dr. Marie
Noelle Pons, from ENSIC-Nancy. Specifically, the characterisation of nucleation and growth
rates and the degree of agglomeration represent major scientific information with direct
application to the understanding of industrial operation. The work includes a significant
experimental component (at laboratory level) and a complex step of data interpretation
through image analysis.
An automated method is being developed for the
morphology classification of crystals. It combines the analysis of images obtained from an
optical microscope and the analysis of the resulting shape descriptors by a discriminant
factorial method. Most relevant is the information concerning the influence of impurities
on kinetic rates.
The method is being applied to such industrial processes
as sugar crystallisation and dicalcium phosphate production.
|
PSE_4 - Computer-based Lab for operator training and
R&D in process control (S. Feyo de Azevedo)
|
This is essentially a background action, rather than a
scientific project, which is being developed since 1991. The group has developed a process
simulator which can run synchronised with real-time, including A/D and D/A interfacing
with the environment. The simulator is physically connected to external control systems,
serving as a process laboratory for R&D developments in process control. A model-bank
keeps being constructed by adding new case-studies. Throughout the years this tool has
proved most useful in several M.Sc. and Ph.D theses concluded and in training of control
systems.
|
PSE_5 - Modular simulation and integrated optimisation
of chemical processes (R. Salcedo)
|
The simulation and optimisation of complex industrial
processes is a subject where a great research effort has been made. During the project,
which will clearly extend beyond the three year period, we intend to develop tools for
equation oriented optimization, where the decision variables are automatically chosen,
removing much of the burden from the designer. Such will require interfacing robust MINLP
solvers developed within our research group (the MSGA and MSIMPSA algorithms) with
appropriate tearing-partitioning algorithms. Interfacing of these solvers with the ASPEN
sequential modular simulator is also a possibility. We will apply these tools to
biotechnology processes, whereby the system topology of an industrially important process
will be optimised through a MINLP formulation. We will also study the optimum synthesis of
reactive distillation columns, whereby the number of theoretical trays and feed location
(reaction zones) will be optimised together with the continuous variables. It is expected
that this work will result in a viable MINLP simulator/optimiser package that will have a
wide field of applicability in the chemical process industries.
|
PSE_6 - Optimisation of cyclone design for fine
particle collection and gas cleaning (R. Salcedo)
|
The objective of this work is to develop numerically
optimized optimized geometries for reverse-flow cyclones with recirculation capabilities,
that may be used both for dedusting and acid gas cleaning. A reverse-flow cyclone
numerically designed to retain very small particles (< 2 �m), which has been subject
to a European Patent Pending process (PT B01D045/12, PE No 99670006 - High
Efficiency Ciclones) will receive dry injection of an
adequate sorbent (Ca(OH)2), which will react with HCl to produce innocuous solid residues
(A.M. Fonseca, J.M. �rf�o and R.L. Salcedo, "Kinetic modeling of the reaction of
HCl with solid lime at low temperatures", Ind. Eng. Chem. Res., 37(12), 4570-4576,
1998). The very high collection efficiencies coupled with an adequate residence time of
the reacting particles within the gas stream can in principle be achieved by the adoption
of a recirculating scheme with the use of a venturi/straight-through secondary cyclone.
The experimental laboratory-scale facility will be designed to promote a smooth transition
towards a pilot-scale unit for testing under typical industrial conditions. The
recirculating cyclones have been recently subject of a Portuguese Patent Pending Process
102392M dated 99.12.13.
|
PSE_8 - Re-use of carbonation sludge in the sugar
refining process and in the treatment of flue gases (F. Rocha)
|
This is a project of chemical engineering, resulting
from the group's long-standing cooperation with RAR sugar refinery in Porto. It is
essentially a detailed experimental work where the process sensitivity to basically all
its input variables will be tested and characterised.
|
PSE_9 - Monitoring and multivariable control of
industrial ovens (S. Feyo de Azevedo)
|
This activity corresponds to a well defined contract
work with an industrial company. It is to be developed in cooperation with the Control
Systems and Technology Group of the Institute for Systems and Robotics-Porto. The main
objectives are the development and implementation (on the industrial ovens) of a robust
control strategy for multivariable systems based on a state-space representation of the
process. Of no less importance are the developments concerning off-line and on-line model
identification algorithms.
|