2.6 Integration of Models 73
especially with respect to multi-purpose models – the modeller can choose to
provide several representations, so that the user of the model is able to choose the
most appropriate for the moment or for the respective task representation.
We shall distinguish between two different types of model representation:
internal and external.
The
internal representation concerns how a given model is represented within a
software system or within a computer, and is implementation dependent. For
instance, the internal representation of the simplified model of a circle from Figure
2.17 can be as short as three numbers, connected with the knowledge that they
represent the radius and the coordinates of the centre point, respectively.
The
external (or observable) representation of a software model is usually a
dynamic representation, depending on the values of the model data at the moment.
Figure 2.17 itself is an external representation of the (parameterized) circle. It
should be noted that the external representation is usually based on the internal
one.
Within the internal representation we distinguish between
model data (or
parameters) and model invariance or model knowledge.
The model data is different among the instances of the modelled class of
entities and is used to create distinguishable representatives of the class. It is
almost always included in the model saved on a medium to guarantee its
persistence.
Model invariance can be of two subtypes:
programs and metadata. A program
can be viewed as data, describing one or more algorithms. It can describe
operations on real data or on placeholders. At runtime the placeholders are
replaced with the actual values of the parameters.
The metadata describes the relation among the data elements at the lowest level
(the parameters) and is usually implemented by means of data structures.
On the next level the relations among the metadata can be described by means
of meta-metadata –
cf. Figure 2.17 once more. Since we can always describe the
relations among elements of one level by means of
meta
x
-data on the next higher
level, we can speak about metadata of different degree (cf. Section 2.4.2.1.2.3.1
Levels of Hierarchical Structure above).
2.6 Integration of Models
The majority of devices, machines and other products are actually complex
systems built up from separately produced components. These components can be
of pure mechanical, electrical, or electronic nature, or they can also be intermixed.
When users observe and use them as a whole –
i.e. the product – there is no need to
speak about integration from the user's point of view. From the manufacturer's
point of view, however, all components have to be assembled or built together; this
process can be viewed as integration. Therefore, when a compound product is
modelled, depending on the purpose of the model, it could be natural and useful
first to model each component alone and after that to integrate all these models in a
compound model of the product. But what is integration? According to Lutters
(2001), there are countless definitions of integration. One more reason to define
again what should be understood under integration in the present work is that none
of the known definitions is perfect, including this in Lutters (2001): “
the