Efficient Multi-view Change Management
in Agile Production Systems Engineering
Felix Rinker
1,2 a
, Sebastian Kropatschek
3 b
, Thorsten Steuer
3 c
, Kristof Meixner
1,2 d
,
Elmar Kiesling
4 e
, Arndt L
¨
uder
3,5 f
, Dietmar Winkler
1,2 g
and Stefan Bif
1,3 h
1
Institute of Information Systems Engineering, TU Wien, Vienna, Austria
2
CDL for Security & Quality Improvement in the Production System Lifecycle, TU Wien, Vienna, Austria
3
Center for Digital Production, Vienna, Austria
4
Institute of Data, Process, and Knowledge Engineering, WU Wien, Vienna, Austria
5
Institute of Ergonomics, Manufacturing Systems and Automation, OVGU, Magdeburg, Germany
Keywords:
Industry 4.0, Change Management, Multi-view Modeling, Multi-aspect Information System, Production
Systems Engineering.
Abstract:
Agile Production Systems Engineering (PSE) is a complex, collaborative, and knowledge-intensive process.
PSE requires expert knowledge from various disciplines and the integration of discipline-specific perspectives
and workflows. This integration is a major challenge due to fragmented views on the production system with
a difficult a priori coordination of changes. Hence, proper tracking and management of changes to heteroge-
neous engineering artifacts across disciplines is key for successful collaboration in such environments. This
paper explores effective and efficient multi-view change management for PSE. Therefore, we elicit require-
ments for multi-view change management. We design the agile Multi-view Change Management (MvCM)
workflow by adapting the well-established Git Workflow with pull requests with a multi-view coordination
artifact to improve over traditional document-based change management in PSE. We design an information
system architecture to automate MvCM workflow steps. We evaluate the MvCM workflow in the context of
a welding robot work cell for car parts, using a typical set of changes. The findings indicate that the MvCM
workflow is feasible, effective, and efficient for changes of production asset properties in agile PSE.
1 INTRODUCTION
Flexible production and shorter product and produc-
tion development cycles are major goals of the Indus-
try 4.0 (I4.0) vision (Galati and Bigliardi, 2019). Con-
sequently, Production Systems Engineering (PSE) has
to become increasingly agile, requiring engineers
from several disciplines to work iteratively and in par-
allel (Eisentr
¨
ager et al., 2018). PSE exhibits char-
acteristics of knowledge-intensive processes (Di Cic-
cio et al., 2015), as design activities typically involve
a
https://orcid.org/0000-0002-6409-8639
b
https://orcid.org/0000-0002-9049-8038
c
https://orcid.org/0000-0002-4847-5182
d
https://orcid.org/0000-0001-7286-1393
e
https://orcid.org/0000-0002-7856-2113
f
https://orcid.org/0000-0001-6537-9742
g
https://orcid.org/0000-0002-4743-3124
h
https://orcid.org/0000-0002-3413-7780
considerable tacit knowledge. However, in agile set-
tings, knowledge-intensive aspects are even more pro-
nounced in PSE due to the dynamics coming from
various domains after the completion of an activity.
This setting results in an unpredictable flow of ac-
tivities (dos Santos Franc¸a et al., 2015). Effective
and efficient PSE is covered in guidelines like the
VDI 3695 (VDI, 2009) that discusses the maturity
of PSE organizations. One aspect of the VDI 3695
is Change Management (CM) targeting the multidis-
ciplinary nature of PSE. A method increasingly ap-
plied to support PSE maturity is the Product-Process-
Resource (PPR) approach (Schleipen et al., 2015). It
aims to organize the heterogeneous aspects of prod-
uct designs, production process models, and produc-
tion resources across disciplines. To understand the
impact of collaborative design updates on production
system properties (e.g., on performance, cost, and
risk), changes across disciplines need to be managed
134
Rinker, F., Kropatschek, S., Steuer, T., Meixner, K., Kiesling, E., Lüder, A., Winkler, D. and Biffl, S.
Efficient Multi-view Change Management in Agile Production Systems Engineering.
DOI: 10.5220/0011074000003179
In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 2, pages 134-141
ISBN: 978-989-758-569-2; ISSN: 2184-4992
Copyright
c
2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
and integrated while supporting their heterogeneous
perspectives (Passow and Passow, 2017). Views on a
PPR properties (i) are often semantically similar, al-
though the engineering artifacts or measurement units
differ; and (ii) concern implicit domain knowledge
that may be difficult to express. These dependencies
introduce challenges for efficient Multi-view Change
Management (MvCM) in agile PSE:
Challenge 1: Scattered and Heterogeneous Domain
Knowledge. Scattered engineering views make it
difficult to track changes’ origin and manage them
across stakeholder views (Wohlrab et al., 2020) prop-
erly. Consequently, MvCM requires comparison ca-
pabilities that go beyond best-practice text compari-
son for program source code (Toulm
´
e, 2006).
Challenge 2: Insufficient Multi-view Change Man-
agement. The use of change management in PSE
has matured within individual disciplines. How-
ever, mechanisms for multi-view change management
across disciplines are currently not well understood.
PSE, typically organized along several disciplines, re-
quires engineers of different disciplines and organiza-
tions to cooperate, making change management much
more difficult than for a single discipline (VDI, 2009).
To tackle these challenges, we raise the research
question: What process and information system de-
sign can improve the effectiveness and efficiency of
multi-view change management in agile PSE?
Aim. We follow the Design Science methodology
to address the research question aiming at improving
the effectiveness and efficiency of multi-view change
management in PSE by (i) adapting and evaluating the
Git workflow with pull request (Krusche et al., 2016);
(ii) building on a Multi-Domain Engineering Graph
(MDEG) as coordination artifact (Biffl et al., 2021;
Rinker et al., 2021a) for change management.
Remainder of This Paper. Section 2 summarizes re-
lated work. Section 3 introduces an illustrative use
case and requirements for efficient MvCM in agile
PSE. Section 4 introduces the MvCM process and
techniques to automate MvCM activities. Section 5
reports on the results of a feasibility study on a con-
ceptual MvCM prototype. The requirements and data
for the study come from use cases and domain experts
from automotive manufacturing. Section 6 discusses
the research results and limitations. Section 7 con-
cludes and raises directions for future research.
2 RELATED WORK
This section summarizes related work on PSE, on the
management of knowledge-intensive processes, and
on change management in Software Engineering.
Multi-view System Modeling in PSE. In PSE, in-
formation is encapsulated in discipline- and tool-
specific artifacts (Strahilov and H
¨
ammerle, 2017).
The VDI 3695 guideline (VDI, 2009) describes ma-
turity levels for improving the capabilities of PSE or-
ganizations. In particular, it describes change man-
agement regarding the iterative exchange of engineer-
ing models/data, from single-discipline management
(maturity level B) towards tool-assisted management
across several disciplines (maturity level D). Change
management concerns version/change management
of systems engineering artifacts.
The I4.0 vision requires seamless and traceable
information exchange across disciplines (Biffl et al.,
2017) Model-based Systems Engineering (MBSE)
provides the basis for domain-specific modeling and
setting up cooperative processes (Huldt and Stenius,
2019). I4.0 assets (Heidel et al., 2017) represent
the three main aspects of PSE: (i) products with
their properties, (ii) processes producing the products,
and (iii) resources that execute production processes.
Such assets can be defined for the I4.0 Asset Admin-
istration Shell (AAS) (Plattform Industrie 4.0, 2020)
using PPR modeling (Schleipen et al., 2015). PPR-
based multi-view engineering networks seem promis-
ing to coordinate changes in artifacts and models, us-
ing markers according to a coordination policy (Biffl
et al., 2021; Rinker et al., 2021a).
Knowledge-intensive Processes. A key character-
istic of I4.0 is to support highly configurable prod-
ucts with a low overhead for production system re-
configuration (Gilchrist, 2016). Therefore, agile PSE
and production require agile and flexible workflows.
Agile PSE exhibits characteristics of Knowledge-
Intensive Processes (KIPs) (Di Ciccio et al., 2015)
as PSE processes are not tightly framed and do
not follow predefined, deterministic workflows. In-
stead, their execution depends heavily on knowledge-
intensive decision-making by experts coming from
multiple disciplines. To manage KIPs, it is neces-
sary to focus not only on tasks and workflows but to
capture a variety of interrelated elements (Di Ciccio
et al., 2015). In the agile PSE context, these elements
include data, actions, rules, processes, production re-
sources, and experts from various domains along all
phases of the PSE life cycle.
Change Management in Agile Software Engineer-
ing. In agile software engineering, the Git Work-
flow With Pull Requests (Krusche et al., 2016) is a
best practice for quality assurance and coordination
of changes to source code, resulting from the parallel
work of several contributors in a distributed project.
This workflow consists of three phases: (i) scop-
ing of source code management by defining main and
Efficient Multi-view Change Management in Agile Production Systems Engineering
135
feature branches as context for the change manage-
ment workflow, including explicit dependencies of
code elements; (ii) pull requests by contributors to
trigger merging changes from feature branches into
the main branch, defining the changescope to assess;
and (iii) a review process for assessing the change-
set regarding its impact on the main branch, including
dependencies to other code parts. In particular, this
concerns parallel changes, which may warrant con-
sidering the changesets of several pull requests that
concern related/dependent code.
The Git Workflow with Pull Requests has been
successfully used for source code, with text-based
identification of changes. Popular code management
platforms, such as GitHub
1
or Bitbucket
2
, implement
this workflow. Code reviewing systems, such as Ger-
rit
3
, also incorporate the Pull Request concept. While
this approach could be useful in agile PSE, the follow-
ing limitations prevent its direct application: PSE re-
quires change management for changesets in hetero-
geneous multi-model, multi-domain engineering arti-
facts, not homogeneous source code.
In this paper, we build on a PPR based multi-view
engineering graph as integrated multi-view coordina-
tion artifact, to coordinate multi-view change man-
agement in agile PSE following the Git Workflow With
Pull Requests by mapping key concepts from soft-
ware engineering to PSE.
3 USE CASE AND
REQUIREMENTS
This section describes the use case laser welding
with a robot and derives requirements from a do-
main analysis at an industry partner from car man-
ufacturing. Fig. 1 illustrates the production process
laser welding automated by a robot with a welding
head (Kropatschek et al., 2021).
Multi-view Change Scenario 1: Change propaga-
tion between semantically similar properties. Engi-
neers work on semantically similar asset properties.
For instance, the welding head temperature proper-
ties in the process expert and the operator views (cf.
Fig. 1, violet circles with letter Y) are considered sim-
ilar. However, the properties might be represented in
different formats or units, e.g., degrees C or K, in arti-
facts, e.g., a specification, plan, or log file. Therefore,
a change of the welding head temperature by the pro-
cess expert is a candidate for efficient propagation and
1
https://github.com
2
https://bitbucket.org
3
https://www.gerritcodereview.com
notification of the Operator.
Therefore, stakeholders require means to effi-
ciently mark property changes (cf. Fig. 1, red dia-
monds) to semantically similar properties (cf. Fig. 1,
dependencies represented as violet circles). The prop-
agation typically occurs between stakeholder views of
a single PPR asset, e.g., welding speed, but could also
occur between property views of different PPR as-
sets. If there is a transformation function to represent
the semantic similarity relationship, the transforma-
tion can (i) notify the owner of the dependent prop-
erty of the required change; and (ii) provide a suitable
property value for a valid update. efficient
Multi-view Change Scenario 2: Multi-view re-
validation of properties after changes. Engineers
change asset properties that relate to other properties
that then require re-validation. For example, welding
speed, temperature, and performance are (i) related
among each other and (ii) across stakeholder views
(cf. Fig. 1). For instance, the process expert, automa-
tion engineer, and operator work with these proper-
ties, but in different method and technology contexts.
Hence, stakeholders require means to efficiently
identify asset properties that need re-validation after
such changes (cf. Fig. 1, red diamond markers). Fig. 1
shows these multi-lateral change dependencies as cir-
cles in orange with a letter. For instance, the letter
F represents the dependency between welding speed,
temperature, performance, and seam quality. On a
property change, related properties could be marked
(cf. Fig. 1, yellow diamond markers). These change
markers can be used to indicate (i) a required re-
validation to the property’s “owner”; and (ii) possible
change conflicts if several markers are present, e.g.,
the welding robot’s PE.welding performance. Change
conflicts may occur if stakeholders change asset prop-
erties with dependencies in parallel, in particular, if
they span over engineering artifacts in several views.
For instance, the operator’s change to the welding
speed (cf. Fig. 1) needs to be re-validated and re-
leased by the quality manager, the process expert, and
the automation engineer. The quality manager needs
to approve the achieved quality, the process expert has
to check dependencies to connected processes, and
the automation engineer may have to adjust a program
to address new speed tolerances.
Requirements for Multi-view Change Manage-
ment. Based on the use case, we identified the fol-
lowing requirements (Rx) for an efficient MvCM ap-
proach to achieve maturity level D (VDI, 2009) re-
garding changes to PPR asset properties. R1. Multi-
view configuration management. The approach shall
represent domain knowledge for multi-view configu-
ration management of PPR assets and properties. R2.
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
136
Process Resource
Product
Asset
Types:
Link
Types:
Functional
Asset Property
Process-Resource
Product-Process
Assembly
QM.Cycle_time
QM.Process_
Quality
QM.Welding_
Seam_Quality
Metal
Profiles
Welding
Robot
PE.Welding_Speed.min
PE.Welding_Speed.max
PE.Welding_Performance
PE.Welding_Direction
PE.Angle_of_Attack
PE.Lateral_Angle
Welding
Head
Quality
Manager
(QM)
Process
Expert (PE)
Automation
Engineer (AE)
Operator
(O)
Protective
Gas
Metal
Sheet
Welding
Filler
Material
PE.Cycle_time
PE.Process_
Quality
AE.Cycle_time
O.Cycle_time
Laser
Welding
PE.Temperature
_Welding_Head
PE.Welding_
Seam_Quality
O.Temperature
_Welding_Head
AE.Welding_Speed.min
AE.Welding_Speed.max
AE.Welding_Speed.default
AE.Welding_Direction
AE.Angle_of_Attack
AE.Lateral_Angle
O.Welding_Speed
O.Welding_Performance
V
Y
Y
F
D
E
D
E
D
F
F
F
Products &
Process
Production
Resources
Multi-lateral
Change Dependencies
(markers represent
arrow ends)
Re-validation
Dependency
Semantic
Similarity
D
V
U
U
W
W
F
F
GF
F
G
A
V V
A
Property value changed
Property to re-validate
Stakeholders
Property value progated
x
x
x
x
x
#
#
#
#
#
#
#
#
o
o
o
x
x
Figure 1: Stakeholder view properties for (i) the Laser Welding process (and products); and (ii) the Robot and Welding Head
as main resources of a work cell, based on the Formalised Process Description (FPD) notation (Biffl et al., 2021).
Change tracing. The approach shall facilitate trac-
ing change requests and changes in engineering arti-
facts across stakeholder views. R3. Change coordi-
nation. The approach shall facilitate representing and
manipulating the changestate of PPR assets and prop-
erties as a basis for defining and executing change
coordination policies. In particular, change propaga-
tion for semantically similar properties, and change
re-validation analysis, to efficiently determine the po-
tential impact of a change on dependent PPR prop-
erties. R4. Efficient multi-view change management
process. The approach shall provide an efficient pro-
cess for change management across PSE stakeholder
views, building on domain knowledge represented ac-
cording to requirements R1 to R3.
4 MULTI-VIEW CHANGE
MANAGEMENT
This section introduces the MvCM workflow and the
MvCM system design to automate workflow tasks.
We adapted the Code Review Workflow introduced in
(Krusche et al., 2016) from the software engineering
domain to equivalent phases in multi-view manage-
ment in PSE.
MvCM Phase 0: Team Workspace Setup. Be-
fore the MvCM workflow can be conducted, the team
workspace, which is comparable to a Git repository,
needs to be set up.
Setting up the team workspace concerns: (i) De-
fine the local concepts of each discipline; (ii) Ne-
gotiate Common Concepts (CCs) between the disci-
plines; (iii) Instantiate a Multi-Domain Engineering
Graph (MDEG) using Common Concepts (CCs); and
(iv) Define semantic links and dependencies between
properties in the graph (Rinker et al., 2021a).
The MDEG contains the multi-view domain
knowledge required for MvCM as a foundation for
efficient analysis in a graph database. To facilitate
describing dependencies among a group of proper-
ties, we adapt the PPR Asset Network (PAN) meta
model (Biffl et al., 2021) to represent multi-lateral de-
pendencies (cf. Fig. 1, dependency F), going beyond
bi-lateral dependencies, between PPR asset proper-
ties. Further, we define structured properties, such
as welding speed.min, to represent dependencies on
property welding speed, including property substruc-
tures, such as required, minimal, or maximal welding
Efficient Multi-view Change Management in Agile Production Systems Engineering
137
Change
Welding_Speed
Calculate Diff to
unified model
Create Pull Request to
integrate changeset to
unified model
Test changeset to unified model
and calculate impact on
dependencies by following
previous defined semantic links
Process
Expert
Analysis related engineers and
assign review tasks
Automation
Engineer
Quality
Manager
Engineers review
assigned changes
Common
Integration
Operator
Changes
ok?
Request
Improvement
Rework
Operator
no
Request
justified?
yes
Escalate
Operator
has
impact ?
yes
no
yes
need improvement
Start
1 1 1
2
2
3
3
4
5
no
End
Escalate
Figure 2: Multi-view Change Management (MvCM) workflow tasks and decisions, based on (Krusche et al., 2016).
speed. Fig. 2 depicts the MvCM workflow to address
changes in a MDEG.
MvCM Phase 1: Local Preparation. In this phase,
an engineer changes a property in a local, domain-
specific view. The local view is comparable to a fork
in the Git workflow in software engineering. For in-
stance, Figure 1 shows the property Welding Speed of
the operator view (O) to be changed, e.g., from 10 to
20 mm/s.
To initiate the integration of the local view
changes to the common view, the engineer creates
a Pull Request, which includes the origin and target
models and the set of change candidates. For instance,
the property O.Welding Speed in the common view
has to be changed to 20. The Pull Request is placed
into the main repository that holds the common view,
assigning the local view engineer as initial reviewer.
MvCM Phase 2: Multidisciplinary Change Anal-
ysis. In this phase, the Semantic Analyzer service
analyzes the impact of the changeset on dependent
properties following the previously defined semantic
links in the MDEG. For instance, semantic links oc-
cur if properties are semantically equal, requiring the
propagation of a property value, or if a semantic re-
lation requires the re-validation of a property value
by an engineer of the corresponding discipline. If the
changeset has no impact on other properties, the next
phase is Common Integration. If the changeset con-
cerns other properties, the Semantic Analyzer service
marks these properties (i) changed due to an update of
a stakeholder artifact (cf. Fig. 1, red diamonds); (ii)
changed due to change propagation between seman-
tically similar properties (cf. Fig. 1, violet diamond);
or (iii) to re-validate due to a change dependency to
a changed property (cf. Fig. 1, yellow diamonds).
These marked changes are assigned for review to en-
gineers of the disciplines concerning marked proper-
ties, e.g., the Quality Manager, Process Expert, and
Automation Engineer.
MvCM Phase 3: Multidisciplinary Examination.
In this phase, the assigned engineers review the
multi-view changes. Review activities concern (i)
the validation of a changed property value due to a
value propagation initialized by the origin pull re-
quest changeset or (ii) the re-validation of a changed
property value due to a calculation, dependent on the
origin change set values. These changes can be (i)
marked as correct, which leads to the phase Common
Integration, (ii) marked as incorrect, which leads to
the task Request Improvement, or (iii) declined, which
terminates the pull request process and escalates the
workflow that requires the involved disciplines to dis-
cuss solution options.
MvCM Phase 4: Local Rework. This phase is
initialized by the Request Improvement as a result
of the Multidisciplinary Examination phase. A Re-
quest Improvement is filed, if the initial change-
set results in implausible values in other disciplines,
e.g., after a change by the Operator the calculated
QM.Welding
Seam Quality is not in the range of the
required quality. The engineer, who submitted the
original change in our case the Operator, decides
whether (i) she can address the improvement request
or (ii) rework is impossible, and the pull request needs
to be escalated to project management. If rework is
possible, the engineer updates the open pull request
and triggers phase 2 Multidisciplinary Change Anal-
ysis.
MvCM Phase 5: Common Integration. In
this phase, the pull request applicant integrates the
changeset into the common view. The changeset con-
sists of a list of model differences, i.e., deltas of the
original modification in the local view and the prop-
agated and calculated values. These model deltas are
displayed for a final review and consistency check-
ing, following the Git Workflow staging step. If the
list of model differences is feasible, the pull request
applicant commits the changes to the common view
repository and adds a change description. Finally, the
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
138
common view version is incremented, similar to the
Git Workflow using a semantic versioning number.
Multi-view Change System Design. Based on the
elicited requirements (cf. Section 3) we infer the
need for a lightweight and easy-to-manage system de-
sign that incorporates low-code solutions (Bucaioni
et al., 2022). The system design is inspired by the
techniques of the well-established Eclipse Modeling
Framework (EMF) for advanced meta-modeling and
model comparison and merging operations. However,
EMF is closely coupled to Eclipse
4
, which hinders a
custom-tailored integration into other system archi-
tectures (Batory and Altoyan, 2020). Also, the Git
5
change tracking system and workflow is the foun-
dation for the multi-view model change management
system.
Comparator Merger
Converter
Transformer Semantic Analyser
Multi-view Model
Framework
Graph Database Backend
External Tools
Git Model Backend
Change Management UI
Project Graph
Generator
Figure 3: MvCM System Design, based on MvMF (Rinker
et al., 2021b).
Fig. 3 depicts our design of the Multi-view Change
System architecture building on the Multi-view Mod-
eling Framework (MvMF) implementation (Rinker
et al., 2021b). The system consists of a Git Model
Backend, Graph Database Backend, a Multi-view
Modeling Framework adapted by the Semantic Ana-
lyzer and Change Management UI.
5 EVALUATION IN A
FEASIBILITY STUDY
This section reports on the evaluation of the MvCM
approach. The evaluation environment consists of a
MDEG, i.e., a PAN (Biffl et al., 2021), for the laser
welding process (cf. Section 3). A Neo4J
6
instance
can be found online.
7
The case focuses on three PPR
assets and 30 to 40 properties from four stakeholder
views, 12 change dependencies, and a set of three
changes to asset properties. The input data for the
4
Eclipse IDE: https://www.eclipse.org
5
Git SCM: https://git-scm.com/
6
Neo4J: https://neo4j.com
7
Graph instance: https://github.com/tuw-qse/use-cases/
tree/main/laser welding
MvCM process are XML files from the use case, in-
cluding quality dependencies from an FMEA tool
8
.
The symbols in the diamond-shaped change markers
(cf. Fig. 1) represent a property changeset. The assets,
properties, and links in the PAN provide the basis to
specify graph queries answering questions concern-
ing the change workflow, e.g., which properties de-
pend on a changed property? The following Cypher
9
query retrieves a PAN sub-graph for marking PAN el-
ements that depend on a changed PAN element (cf.
Fig. 1, yellow diamonds).
MATCH (a:Attribute)
-[:has_PPRDependency *]
-(b:Attribute)
WHERE a.name="O.Welding_Speed" AND
a.ChangeState="Changed"
SET b.ChangeState="To Validate"
Listing 1: Cypher query for marking dependent properties
of a changed property for re-validation.
MvCM Efficiency in Comparison to Document-
based Change Management. The conceptual pro-
cess comparison estimates the effort for each MvCM
phase, analyzing for which scenarios MvCM is likely
to be more efficient than the traditional approach.
Phase 0. Team Workspace Setup. The engineering
organization in the study context operates on maturity
level CM-B, with engineering toolchains within the
disciplines and data exchange among the disciplines.
Modeling the PAN took only a small extra effort, as
the data can be efficiently derived input data.
In the study context, the PAN of the robot cell can
be defined as an output artifact of the data exchange.
For a typical robot cell, defining the PAN requires up
to three workdays. If the PAN is filled automatically,
it should only take seconds. For the traditional ap-
proach, this task is not required.
Phases 1. Local Preparation and 2. Multidisciplinary
Change Analysis. For the MvCM approach, the con-
ceptual MvCM information system enables the an-
alyzing of PPR asset property value updates. The
conceptual Semantic Analyzer service uses Neo4J for
the change impact evaluation and sets change mark-
ers based on the dependency definitions coming from
Phase 1 (cf. Listing 1). Therefore, this analysis can
be conducted efficiently after each change to the team
workspace, taking a few seconds, to raise the aware-
ness of the involved domain experts.
For the traditional approach, change impact anal-
ysis on the level of PPR asset properties is time-
consuming and error-prone as the domain experts
8
APIS FMEA: https://www.apis-iq.com
9
Cypher language: https://www.opencypher.org
Efficient Multi-view Change Management in Agile Production Systems Engineering
139
have to consider the impact of their local changes
to related disciplines. This is often tacit knowledge
or can only be decided by all involved domain ex-
perts. In the study context, domain experts from the
involved disciplines have to come together in a time-
consuming meeting to identify the changes in the use
case.
Phases 3. Multidisciplinary Examination, 4. Local
Rework, and 5. Common Integration. The review re-
quires an overview of changes and their impacts to
accept or reject a pull request and identify follow-up
tasks to address the change impact. All properties
with change dependencies receive markers to guide
the review on these changes (cf. Fig. 1, yellow di-
amonds). To inform planning a focused review, the
PAN provides an integrated view to identify which
stakeholders are required to discuss a specific set of
changes and to decide on reworking and integrating
the changes, or to reject the pull request.
For the traditional approach, all potentially in-
volved stakeholders have to meet to determine which
views could be affected by a change. Therefore, the
review meetings tend to become large and inefficient.
Further, in a large meeting, it is easy for stakeholders
to overlook a dependency in their view, leading to the
risk of late and costly rework.
6 DISCUSSION
The MvCM process and system design introduced in
this paper facilitate dynamic cross-domain collabora-
tion of experts in KIPs (Di Ciccio et al., 2015) that
are typical in agile PSE. Specifically, we investigated
how the MvCM approach can improve multi-view
change management by tracking and organizing par-
allel changes by several stakeholders. The evaluation
in the context of a typical industrial joining process
showed the MvCM process to be feasible, effective,
and efficient (cf. Section 5).
The traditional document-based approach works
well for treating isolated, sequential changes in small,
co-located project settings. However, distributed or
large projects in agile PSE, require fast, precise, and
automated analysis of changes on the level of model
property values, considering dependencies between
individual domains, i.e., maturity level CM-D (cf.
Section 2). Therefore, we expect the MvCM approach
to make up the one-time effort required for model-
ing the PAN within a typical large automation project
by reducing the effort for recurring change impact
analysis and avoidable rework due to uncoordinated
changes.
The results go beyond the state of the art in
the area of integrating data from and coordinating
multi-disciplinary changes in PSE (Kattner et al.,
2019; Meixner et al., 2021), in particular MvCM re-
validation, (i) by defining a sufficiently fine-grained
MDEG for analyzing changes of property values,
(ii) by considering multi-lateral dependencies, rather
than just bi-lateral dependencies between PPR as-
set properties, (iii) by efficiently describing seman-
tic constraints among a group of properties, and
(iv) by demonstrating the feasibility based on data
from changes to PSE artifacts.
Limitations. The following limitations require fur-
ther investigation. Feasibility study. The study fo-
cused on a use case abstracted from a large PSE com-
pany. This may introduce bias due to the specific se-
lection of the production process, stakeholder views,
or individual preferences of the domain experts. To
overcome these limitations, we plan case studies in a
wider variety of application contexts. Change Cases
and Workflow. The selected change cases may in-
troduce bias due to the specific selection of the data
sets and assumptions of parallel changes. Further, the
research focused on the Git Workflow with Pull Re-
quests, a best practice in software engineering, while
there is a wide variety of other engineering change
management approaches. Therefore, we plan to ex-
plore further change cases and change management
approaches to compare their benefits and limitations
in diverse application settings.
7 CONCLUSION AND FUTURE
WORK
This paper investigated effective and efficient Multi-
view Change Management (MvCM) capabilities to
integrate discipline-specific aspects in Production
Systems Engineering (PSE) into a unified view. The
MvCM process definition and evaluation builds on the
Software Engineering best-practice process Git Work-
flow with pull request (Krusche et al., 2016). The
MvCM process and system design provide capabili-
ties for tracking the origin of changes and properly
managing changes in parallel PSE to achieve multi-
view maturity levels according to the PSE guideline
VDI 3695-3 (VDI, 2009).
Results of the feasibility on a typical scope of
changes in PSE to a robot work cell for welding car
parts indicate the MvCM process to be feasible, ef-
fective regarding requirements derived from the PSE
guideline VDI 3695, and efficient in comparison to
traditional document-based change coordination.
Future Work. We plan to investigate approaches that
are semantically more expressive than Neo4J, such
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
140
as Semantic Web technologies and domain-specific
model management options, to represent and query
larger change management use cases.
ACKNOWLEDGEMENT
The financial support by the Christian Doppler Re-
search Association, the Austrian Federal Ministry for
Digital & Economic Affairs and the National Foun-
dation for Research, Technology and Development is
gratefully acknowledged. This work has been par-
tially supported and funded by the Austrian Research
Promotion Agency (FFG) via Austrian Competence
Center for Digital Production” (CDP) under contract
nr. 881843. This work has received funding from the
Teaming.AI project in the European Union’s Horizon
2020 research and innovation program under grant
agreement No 95740.
REFERENCES
Batory, D. S. and Altoyan, N. (2020). Aocl : A Pure-Java
Constraint and Transformation Language for MDE. In
MODELSWARD 2020, pages 319–327, Set
´
ubal, Por-
tugal. SCITEPRESS.
Biffl, S., L
¨
uder, A., and Gerhard, D., editors (2017). Multi-
Disciplinary Engineering for Cyber-Physical Produc-
tion Systems, Data Models and Software Solutions for
Handling Complex Engineering Projects. Springer.
Biffl, S., Musil, J., Musil, A., Meixner, K., L
¨
uder, A.,
Rinker, F., Weyns, D., and Winkler, D. (2021). An In-
dustry 4.0 Asset-Based Coordination Artifact for Pro-
duction Systems Engineering. In 23rd IEEE Int. Conf.
on Business Informatics. IEEE.
Bucaioni, A., Cicchetti, A., and Ciccozzi, F. (2022). Mod-
elling in low-code development: a multi-vocal sys-
tematic review. Software and Systems Modeling.
Di Ciccio, C., Marrella, A., and Russo, A. (2015).
Knowledge-intensive processes: characteristics, re-
quirements and analysis of contemporary approaches.
Journal on Data Semantics, 4(1):29–57.
dos Santos Franc¸a, J. B., Netto, J. M., do ES Carvalho,
J., Santoro, F. M., Bai
˜
ao, F. A., and Pimentel, M.
(2015). KIPO: the knowledge-intensive process ontol-
ogy. Software & Systems Modeling, 14(3):1127–1157.
Eisentr
¨
ager, M., Adler, S., Kennel, M., and M
¨
oser, S.
(2018). Changeability in engineering. In 2018 IEEE
International Conference on Engineering, Technology
and Innovation (ICE/ITMC), pages 1–8.
Galati, F. and Bigliardi, B. (2019). Industry 4.0: Emerging
themes and future research avenues using a text min-
ing approach. Computers in Industry, 109:100–113.
Gilchrist, A. (2016). Introducing industry 4.0. In Industry
4.0, pages 195–215. Springer.
Heidel, R., Hankel, M., D
¨
obrich, U., and Hoffmeister, M.
(2017). Basiswissen RAMI 4.0: Referenzarchitektur-
modell und Industrie 4.0-Komponente Industrie 4.0.
Beuth Verlag.
Huldt, T. and Stenius, I. (2019). State-of-practice survey of
model-based systems engineering. Systems engineer-
ing, 22(2):134–145.
Kattner, N., Bauer, H., Basirati, M. R., Zou, M., Brandl, F.,
Vogel-Heuser, B., B
¨
ohm, M., Krcmar, H., Reinhart,
G., and Lindemann, U. (2019). Inconsistency man-
agement in heterogeneous models. In Proc. Design
Society: Int. Conf. Eng. Design, pages 3661–3670.
Cambridge Univ.
Kropatschek, S., Steuer, T., Kiesling, E., Meixner, K.,
Fr
¨
uhwirth, T., Sommer, P., Schachinger, D., and Biffl,
S. (2021). Towards the representation of cross-domain
quality knowledge for efficient data analytics. In
ETFA 2021, pages 1–4.
Krusche, S., Berisha, M., and Bruegge, B. (2016). Teaching
code review management using branch based work-
flows. In Proceedings of the 38th International Con-
ference on Software Engineering Companion, pages
384–393.
Meixner, K., L
¨
uder, A., Herzog, J., Winkler, D., and Biffl,
S. (2021). Patterns For Reuse In Production Systems
Engineering. International Journal of Software En-
gineering and Knowledge Engineering, pages 1623–
1659.
Passow, H. J. and Passow, C. H. (2017). What competen-
cies should undergraduate engineering programs em-
phasize? a systematic review. Journal on Engineering
Education, 106(3):475–526.
Plattform Industrie 4.0 (2020). Part 1 - The exchange of
information between partners in the value chain of In-
dustrie 4.0 (Version 3.0RC01 Review). Standard, Ger-
man BMWI. https://bit.ly/37A002I.
Rinker, F., Meixner, K., Waltersdorfer, L., Winkler, D.,
L
¨
uder, A., and Biffl, S. (2021a). Towards efficient
generation of a multi-domain engineering graph with
common concepts. In ETFA 2021, pages 1–4. IEEE.
Rinker, F., Waltersdorfer, L., Meixner, K., Winkler, D.,
L
¨
uder, A., and Biffl, S. (2021b). Continuous Integra-
tion in Multi-view Modeling: A Model Transforma-
tion Pipeline Architecture for Production Systems En-
gineering. In MODELSWARD 2021, pages 286–293.
SCITEPRESS.
Schleipen, M., L
¨
uder, A., Sauer, O., Flatt, H., and
Jasperneite, J. (2015). Requirements and con-
cept for plug-and-work. at-Automatisierungstechnik,
63(10):801–820.
Strahilov, A. and H
¨
ammerle, H. (2017). Engineering work-
flow and software tool chains of automated produc-
tion systems. In Multi-Disciplinary Engineering for
Cyber-Physical Production Systems. Springer.
Toulm
´
e, A. (2006). Presentation of EMF Compare Utility.
In Eclipse Modeling Symposium, pages 1–8.
VDI (2009). VDI Guideline 3695: Engineering of industrial
plants - Evaluation and optimization. Standard, VDI-
Verlag, D
¨
usseldorf, DE.
Wohlrab, R., Knauss, E., Stegh
¨
ofer, J.-P., Maro, S., Anjorin,
A., and Pelliccione, P. (2020). Collaborative trace-
ability management: a multiple case study from the
perspectives of organization, process, and culture. Re-
quirements Engineering, 25(1):21–45.
Efficient Multi-view Change Management in Agile Production Systems Engineering
141