made by humans and restoring correctness of the re-
sulting process. We show the implementation of the
architecture for instantiation of a reference process of
a particular business unit of Siemens AG.
The paper is structured as follows: Section 2 dis-
cusses related work. Section 3 describes the devel-
oped instantiation approach and its architecture. A
case study describing the implementation of the ap-
proach at Siemens AG follows in Section 4. In Sec-
tion 5 we evaluate our approach and draw some con-
clusions in Section 6.
2 RELATED WORK
Instantiation of processes to project specific needs
has been subject to intensive research in recent years.
In early software process approaches it was thought
that a perfect process can be developed which fits
all software developing organizations and all types of
projects (Boehm and Belz, 1990). It was soon rec-
ognized that no such process exists (Basili and Rom-
bach, 1991), (Osterweil, 1987). Due to the dynamics
of software development, every project has individual
needs. Therefore, the description of a general process
(i.e. the reference process) has to be adapted.
Early approaches to overcome this problem have
been developed e.g. by (Boehm and Belz, 1990)
and (Alexander and Davis, 1991). Many different
approaches have been proposed since then. For ex-
ample, the situational method engineering approach
(ME) (Brinkkemper,1996) emphasizes bottom-up as-
sembly of project specific methods from fragments,
requiring very detailed fragment specifications. In
contrary to ME, approaches like SLANG (Bandinelli
and Fuggetta, 1993) regard processes as programs,
enacted by machines (Feiler and Humphrey, 1993).
Although much effort has been put into improving
the adaption of software processes to project specific
needs, none has become accepted in industry as the
standard so far. An important reason is the variety
of processes used. For instance, (Yoon et al., 2001)
developed an approach for adapting processes in the
form of Activity-Artifact-Graphs. Since the process
is composed of activities and artifacts, only the oper-
ations ”addition” and ”deletion” of activities and ar-
tifacts are supported as well as ”split” and ”merge”
of activities. Another example is the V-Model (BMI,
2004), a process model developed for the German
public sector. It offers a toolbox of process mod-
ules and execution strategies. The approach for de-
veloping a project specific software process is to se-
lect required process modules and an execution strat-
egy. Due to these dependencies on the Meta mod-
els, none of the existing approaches offers a complete
and semi-automated method for instantiating Siemens
processes.
Since the main purpose of Siemens software pro-
cesses is to offer high level guidance for humans,
adaptation approaches for business processes are also
of interest. Approaches for processes and workflows
of higher complexity are often restricted to only a
subset of adaptation operations. For instance, config-
urable EPCs (C-EPCs) (Rosemann and van der Aalst,
2007) enable customization of reference processes.
However, the approach only allows activities to be
switched on/off, the replacement of gateways and the
definition of dependencies of adaptation decisions.
(Armbrust et al., 2008) developed an approach for
the management of process variants. A process is
split up in stable parts and variant parts. The pro-
cess is adapted by choosing one variant at the start of
a project. Although the need for further adaptations
during the execution of the process has been identi-
fied, no standardization or tool support is provided.
(Allerbach et al., 2008) developed an approach
called Provop. Processes are adapted by using change
operations which are grouped in Options. Options
have to be predefined and can be used to adapt pro-
cesses, but they do not guarantee correctness.
Existing tools (e.g. Rational’s Method Composer
(RMC) (IBM, 2008)) provide only minimal support
in instantiation. Decisions made by humans have to
be executed mostly manually. Approaches have been
developed making the user aware of inconsistencies
(Kabbaj et al., 2008), but the actual correction of the
process is still left to humans.
3 META MODEL BASED
ARCHITECTURE
In order to improve current situation, we propose an
architecture for instantiation systems that execute in-
stantiation decisions made by humans and automati-
cally restore correctness of the resulting process.
The actual changes in the process are executed
by elemental operations on the process. Examples
are ”Deleting an Activity” and ”Association of a Re-
source with an Activity”. Existing tools allow to per-
form those changes but they do not restore correctness
of the resulting process.
For example: An activity a
1
is connect by an con-
trol flow to an activity a
2
which in turn is connected
by an control flow to an activity a
3
. If a project man-
ager wants to delete a
2
in most existing tools he re-
moves a
2
from the process . Additionally, he has to
take care of restoring correctness e.g. establish the
METHOD MANUAL BASED PROCESS GENERATION AND VALIDATION
257