Modeling of change over time
In conventional methodologies, modeling of change over time is often difficult or even ‘not supported’. Semantic modeling provides easy modeling of change over time. This is caused by the following difference: when the life of a relation in a conventional model is terminated (the relation is deleted) the consequence is that the entity (and the attributes) do not exist anymore. However, when the life of a relation in a semantic model is terminated, then the related things remain existent. As a consequence, in a semantic model every binary relation can be terminated and replaced by another binary relation as and when required, exept for binary classification and specialization relations. The modeling of change over time might include replacing one component by another as is modeled by binary relations as follows:
For example, the composition of a product P1 may change over time when one of its components C1 is replaced by another component C2. As the existence of an individual product nor of a component relies on their relations, except that their existence depends on their classification relation, the modeling of such a change is natural for semantic modeling as follows:
|
Name of left hand object |
Name of kind of relation |
Name of right hand object |
From |
To |
|
P1 |
is classified as a |
pump |
t1 |
|
|
C1 |
is classified as a |
bearing |
t1 |
|
|
C2 |
is classified as a |
bearing |
t1 |
|
|
P1 |
has as part |
C1 |
t1 |
t2 |
|
P1 |
has as part |
C2 |
t2 |
The flexibility to record changes over time also holds for changes in higher order relations. Semantic modeling enables to express that some elements in a relation stay as they were, whereas others are terminated or newly added, without changing the data structure. Conventionally such a requirement implies a change of the data model.
This can be illustrated by a separation of two objects by a separating medium that is replaced by another medium. Such a separation is a typical example of a ternary relation. For example, assume that A is separated from B by S1. Conventionally this separation could be modeled for example as a ternary relation R3, for example as the first line in the following table.
|
Ternary relation |
Kind of relation |
Role-1 |
Role-2 |
Role-3 |
From date-time |
To |
|
R3 |
separation |
A |
B |
S1 |
t1 |
t2 |
|
R4 |
separation |
A |
B |
S2 |
t2 |
The above table enables to express that the separator S1 is replaced at time t2 by another separator S2. However, this requires that the separation relation R3 is terminated and replaced by another separation relation R4. If the requirement arises that it should become possible to record that the separation R3 is not terminated, while separator (S1) still could be replaced one or more times (by S2, S3, etc), this would be impossible with the above data structure and may be only becomes possible after a change of the data model.
With semantic modeling the natural way to model this implies already that the relation R3 can remain, also when S1 is replaced by S2 because usually such a change is modeled as follows:
|
Name of left hand object |
Name of kind of relation |
Name of right hand object |
From |
To |
|
R3 |
is classified as a |
separation |
t1 |
|
|
R3 |
involves as separated |
A |
t1 |
|
|
R3 |
involves as separated from |
B |
t1 |
|
|
R3 |
involves as separator |
S1 |
t1 |
t2 |
|
R3 |
involves as separator |
S2 |
t2 |
Another example is the modeling of a changing property value, such as a measured temperature of some object. In conventional modeling there is usually no distinction between a temperature and its values. As a consequence, a possessing object will have as many temperatures as it will have (discrete) temperature values. For example, the temperature T1 of product P1 is recorded as a function of time. This is for example modeled as follows:
|
Relation |
Product |
Kind of property |
Value |
UoM |
From date-time |
To |
|
R5 |
P1 |
temperature |
20 |
deg C |
t1 |
t2 |
|
R6 |
P1 |
temperature |
21 |
deg C |
t2 |
t3 |
|
R7 |
P1 |
temperature |
22 |
deg C |
t3 |
t4 |
Note that T1 (the temperature of P1) does not appear in this table and is not represented, but instead there are in fact three temperatures of P1, each represented on their own line by a numeric value and a unit of measure. This illustrates that the temperature of P1 is split into a number of discrete points, depending on the number of measurements, whereas the property essentially is a continuous function.
In semantic modeling the aspects of individual things are always explicitly modeled and their quantifications are always expressed as separate binary relations, with their own validity period. Thus in the above example ‘the temperature of P1’ will be explicitly modeled as follows:
|
Name of left hand object |
Name of kind of relation |
Name of right hand object |
UoM |
From |
To |
|
P1 |
has as aspect |
T1 |
t1 |
||
|
T1 |
is classified as a |
temperature |
t1 |
||
|
T1 |
has on scale a value equal to |
20 |
deg C |
t1 |
t2 |
|
T1 |
has on scale a value equal to |
21 |
deg C |
t2 |
t3 |
|
T1 |
has on scale a value equal to |
22 |
deg C |
t3 |
t4 |
Note that T1 is explicit in the model, even if there would have been no measured values (yet). Furthermore, the numbers 20, 21 and 22 are just numbers, and only have a role as temperature values because of the relations in which they are used.
The explicit kinds of relations also allow to add other quantifications (or qualifications) to the aspects. For example, it could be stated that T1 is greater than or equal to 20, or to state that its average value is 21 deg C, without the need to create extra ‘attributes’ or parameters, such as minimum temperature or average temperature. It remains explicit that the values are about T1 as follows:
|
Name of left hand object |
Name of kind of relation |
Name of right hand object |
UoM |
From |
To |
|
T1 |
has on scale a value greater than or equal to |
20 |
deg C |
t1 |
t4 |
|
T1 |
has on scale a mean value equal to |
21 |
deg C |
t1 |
t4 |
Thus, modeling of change is one of the natural features of semantic modeling.
The document ”Modeling of Measurements and Observations”, part 9 of the Gellish Modeling Methodology, is based on the above principles.
