Object_Oriented Modeling as an Instrument for analyzing a Film Processing Plant
From Proceeding of European Simulation Multiconference,ESS94, Barcelona, June 1_3
1994
P. Giribone
MODELING TOOL
In developing this model, it was always necessary to
consider that it would probably be interfaced in the
future with a scheduling system based on artificial
intelligence techniques. Therefore, special attention
was focused on the modular design and expandability
of the model. This is why there was a tendency from
the very beginning to develop a simulation based on an
object_oriented approach.
The author is currently developing different scheduling
applications using AI and operates in general with
KBS. These systems normally refer to C++ as the
interface tool.
The OS/2 release of MODSIM was selected as the
modeling tool for the specific problem. Theoretically,
this language offers special advantages for our specific
case because it can be interfaced with subsequent
scheduling modules and can produce file objects
directly in C++.
Since the simulation product utilized is oriented
towards industrial applications, the powerful graphic
features offered by MODSIM can be used to streamline
the data entry phase, to evaluate results and to screen
dynamic processes; thus the simulator can be used
easily and directly by plant personnel.
An alternative was also considered involving the direct
use of C++ through libraries which had already been
developed in the past by the author's work team for this
type of problem. In fact, the author is part of a research
group which normally uses general purpose languages
(C++, C, Pascal) to model complex plants; this general
line of action is possible and economically feasible
thanks to the previous experience acquired from
numerous concrete industrial applications.
Among other advantages, the use of general purpose
languages offers improved running speed (and
therefore reduced experimentation time), enhanced
portability in different operating systems and greater
useful service life of the modules developed for the
specific applications. In fact, an analysis of the sector
indicates that general purpose languages can maintain
themselves on the market, guaranteeing complete
compatibility with the previous releases for at least
twice the time period relative to the languages
developed for the simulation.
However, the desire to verify MODSIM capacities to
interface with the C++ modules acquired from
advanced applications (KBS or even Neural Networks)
made it necessary to choose the latter. For the specific
case, the delivery terms to develop the model, as
established by the industrial representative, were not
very demanding.
The decision to use the OS/2 operating environment
was dictated only by the contingent availability of the
internal hardware resources. This meant that the most
recent generation of personal computers, each with a
12 MB RAM, could be used to develop the work.
In fact, the potentials of the simulation tool used remain
partially limited by the hardware equipment used, thus
leading to increased running time; in spite of this, the
fact that the code can be transferred into a UNIX
environment means that a final release can be offered
in a more interesting industrial environment.
From this point of view, the structure of the graphic
module obtained is very important. In fact, this tool is
used to condense the entire graphic part relative to
drawings, windows, switches, etc., into an ASCII file;
therefore, this system guarantees immediate portability
from the OS/2 operating system to the UNIX
environment.
Thus, the entire model could be developed within a
specific low_cost environment. Furthermore, it was also
possible to perform the experimental part within such
an environment, while supplying an industrially
powerful tool at the same time that can be immediately
transported to workstations.
Obviously, the graphic potential offered by this tool can
be implemented to perform a logical check on the first
phase of the model; however, this must be followed by
a statistical analysis of the data to validate the model.
A comparison with other languages oriented towards
simulations also provides interesting insight. Over the
last few years, the author has developed simulation
models using SLAM and SLAM II as well as Simscript
(the Modsim predecessor).
In reality, the problem of languages oriented towards
simulations may be expressed in terms of: model
development speed, duration of the service life of the
product and the capacity to go into detail during
modeling.
While all the dedicated languages concentrate on
simplifying and accelerating the modeling phase,
unfortunately, it is often quite a complex task to adapt
industrial situations to models built with these tools.
The new generation of these languages operates within
a slightly less modular framework compared to tools
such as SLAM, Simfactory, etc.; however, it ends up
reproducing the structure of existing multipurpose
languages (Modula 2, C++, Pascal Object Oriented) in
a very faithful way, usually adding a set of interesting
graphic and statistical utilities which are certainly quite
useful from a simulation point of view.
In fact, the author also agrees that it does not make
much sense to talk about object_oriented languages, but
rather a programming philosophy aimed at objects;
while the OOD is undoubtedly the modeling technique
for the future, because of its ability to produce highly
flexible tools, it is very difficult to imagine that the
many different tools currently being developed wil offer
major improvements in creating simulators for
industrial applications.
On the other hand, it can be said that the situation is
totally different for graphic interfaces, which are
evolving very rapidly in terms of interactions between
users and model and also in terms of the possibility
(though to a lesser degree up to now), of developing
complex animation with relatively simple procedures.
In this particular application the use of a pre_
determined language has certainly been very
rewarding.
Another very important element to be considered when
developing this type of application for industrial
situations refers to products which are very compatible
with company policies in terms of the type of language
and the type of available hardware instruments.
In fact, for companies which consider the simulation to
be an important instrument, there is a tendency to
standardize the model development tools to guarantee
maintenance and updating. Over the years, this has led
to the continued use of obsolete tools also in advanced
applications where it would have been much more
effective to use languages equipped with modern
features.
External routines in C are linked to high_level
programming languages (like G2: expert system),
through the construction of data conversion procedures
which guarantee an exchange of information. In this
type of application, we limited ourselves to checking
the possibility of transferring the information in the
simulator to the main system, since the transposition of
knowledge is still in the initial phase.
Return to the Title