Information Technology Reference
In-Depth Information
LEGEND
- Solid black line = normal execution
- Dashed red line = too restrictive
preconditions
- Dotted blue line = unreachable state
s1
s2
s0
s3
s5
s4
Fig. 7.
Object life cycle of data object
D1
of the business process in Fig. 6
and then process it. For too restrictive preconditions, this re-construction has 4
steps. The resulting Petri net is kept while processing the rest of warnings.
1. Find the transition at which the deadlock takes place.
2. Identify the data place that is input of the blocked transition.
3. Identify the data place at which there is a token.
4. As we now know the transition that could not be triggered (step 1) and the
actual current state of the data object (step 3), to re-construct the Petri
net we have to duplicate the blocked transition and replace the arrow from
the data place without token (step 2) by an arrow from the data place with
token (step 3), leaving the rest of inputs and outputs like they are.
Unreachable states are reflected in a Petri net in the form of unfired transitions.
To detect them and fix them, we have to find the transitions that were never
triggered and set the markup of the net with a token in each of their input
places. Then a new reachability graph with this configuration is obtained and
we can examine the rest of the net from that point by processing it.
In order to warn the modeller/analyst about the presence of data anomalies
in the BP model, new elements emerged from dealing with them are marked
in a different way in the OLC. In Figure 7, transitions and states generated
from too restrictive preconditions are shown with dashed red lines, and those
corresponding to unreachable states have dotted blue lines.
5 Related Work
The importance of complementing the activity-centered view of BP models with
an object-oriented view has been described by Snoeck et al. [9]. We are generating
such a view from the data objects that appear in a BP model.
The work most related to our approach is the one of Ryndina et al. In [10]
they present an ad-hoc approach for the automatic generation of OLCs from a
BP model and propose some techniques to analyse the consistency of BPs and
OLCs on the basis of the concepts of
conformance
and
coverage
between OLCs.
