COSY is a European research program founded by the European Science
Foundation (ESF). Its focus is Control of Complex Systems. This is a brief
report from a Workshop on Modeling of Complex Systems held in Lund Sept.
5-7 highlighting some of the problems. There were 23 participants from 11
European countries with background from control, chemical engineering,
mechanical systems, software development and data bases, and numerical
analysis.
Kaj Justin from VTT, Finland presented a platform for modeling and simulation
of large scale systems. Keywords are efficiency in constructing large network
models, numerical solution methods, and an object-oriented specification
language.
Large models (>10^4 states) had been developed for nuclear power plants,
pulp and paper mills, and energy distribution networks.
Wolfgang Marquard from RWTH Aachen presented an interesting view on the
modeling process itself, projecting it along 3 axes; Representation,
specification (or model granularity), and agreement (ie. the social dimension
establishing consensus on the final product). Further, he presented the latest
version of their modeling tool in which features to support the modeling
process and model maintenance had been included.
Costas Pantehdes from Imperial College presented gPROMS, a comprehensive
language for modeling mixed continuous and discrete systems. Systems include
both a plant and procedures related to plant operation.
My own talk presented activities modeling methods for first principles and
semi-empirical models. I got good feedback on our proposed graphical language
for model development.
Jan Willems, Gronningen gave an interesting talk on Representation of
Dynamical
Systems from a mathematical viewpoint. He takes a very general approach not
distinguishing system inputs and outputs, and discusses issues like "When are
to models equivalent" and "how choose the state-space". This is similar to
some of Erik Weyers thesis work.
Hilding Elmqvist (Dymola) reported on language standardization for modeling,
the idea being to define a common syntax and semantics for modeling of
continuous systems. The status on mixed continuous/discrete systems is rather
unclear at the moment both wrt. representation as well as numerical solution
methods, see also http://www.Dynasim.se
Discussion
-----------
7 The use of object-oriented software structure was (of course)
promoted. It was, however, a feeling that this concept limits model
development if strictly enforced. Hence, changing a (copied) object for
prototyping purposes should be allowed even if this new object does not
conform with any of the class definitions.
7 Automatic conversion of equations derived from some symbolic or
graphical representation to a form that is efficient for numerical simulation.
There were different opinions on "how automated" this conversion should be.
7 There exists different approaches to handle mixed continuous/discrete
systems as exemplified by gPROMS and OMSIM. There is no clear understanding
today as to how these systems should be represented and integrated.
There were demos of Dymola and OMSIM. In Dymola they have a feature to export
results (like animations) directly to Netscape for animation (vrml-format).
Dymola is specialized toward mechatronic system.
This was the first of a series of meeting within the COSY program. More
information on the Workshop can be obtained from me.
The participants had a diversifies background. The workshop, however, had a
clearly defined focus. This made it a very valuable meeting for me.
More information can be obtained from me.
Bjarne Foss