Background. Consistent cross-disciplinary engineering data models have become increasingly important for engineers and project managers to validate system designs or implement new features in existing systems. However, discipline-speci c designs (mechanical, electrical, automation engineering etc.) in isolated data models and proprietary software tools often create information silos. Similar to information systems, the challenges in Cyber-Physical Production System (CPPS) are a high amount of heterogeneous data that needs to be analysed and accessible for stakeholders and systems. Aim. The goal of the Flexible Multiaspect Model Integration project is to support the integration of local engineering views and artefacts using the de nition of common concepts across di erent disciplines. Therefore, the thesis project will provide capabilities to integrate and validate multi-aspect models more e ciently to increase the data quality. Method. The project will follow Design Science methodology to design and evaluate i) a method for collecting and de ning common concepts across engineering disciplines, ii) a modularised software system design that enables exible model integration processes in a CPPS context, and iii) an exemplary model integration process that supports data integration needs in the planning, operation, and analytics phase. The model integration processes are evaluated with real-world uses cases from industry. Conclusion. The information systems community will gain insight into the requirements in engineering and a method for agreeing on an inter-disciplinary common understanding from this research.
Information System Engineering (ISE) for CPPS use cases, like digital
shadows, require techniques for (i) automated aggregation and reduction of data,
(ii) data analysis methods, (iii) data accessibility to stakeholder and systems
and (iv) feedback for decision support and system controlling [
To enable these CPPS aspects the discipline-speci c views, local data sources
and stakeholder concerns need to be successful integrated in a combined model
for further validation and veri cation of the system design [
The real-world use case from our industry partner data exchange towards
production system simulation [
Engineer Squirrel-Cage Motor
Electrical Engineer Engineering Data Exchange
C1 Motor
C2
Plant Planner Three-Phase Induction Motor
Updates
Automation Engineer Simulation Data Delivery
C3
Simulation
Engineer Three-Phase Induction Motor
Control Backflow
Engineering Data Artifact Fig. 1: Data Exchange Towards Production System Simulation Use Case and Challenges.
The plant planner designs the initial CPPS functions and structure. Subsequently, the mechanical engineer builds up the system tree based on the basic plan with mechanical functions, system parts, properties, and location of parts. Next, the electrical engineer adds electrical components with interfaces to the mechanical parts and views (electrical system parameters, such as voltage or energy supply). Finally, the automation engineer adds control code that automates the previous engineered parts. At a later stage, the simulation expert integrates the engineering artefacts from basic, mechanical, electrical and automation engineering to design the simulation model.
Ideally, these tasks would be completed sequentially. Yet, engineers design in parallel regularly triggering follow-up changes across disciplines due to increased competition and complex dependencies between hard- and software. Hence, they need to synchronise engineering artefacts during the engineering phase. These data are often synchronised and integrated manually or with poor tool support that takes additional e ort and tends to be error-prone, inducing avoidable project risks.
Challenges to Data Integration in CPPS Engineering. Information
systems and computer-assisted engineering aim at supporting model integration
processes to achieve a common view on engineering objects and accessing the
combined model. However, we have identi ed the following challenges (depicted
in Figure 1) from the industrial use case and literature [
C1 Information loss due to artefact-based data exchange. Data
exchange in a multi-disciplinary environment is usually based on artefact-based
transactions across several workgroups [
C2 High e ort of repetitive tasks and con guration of domain data.
Data integration in a multi-disciplinary process often requires repetitive tasks
that are usually performed manually or require high con guration e ort [
C3 Insu cient common understanding of system boundaries. From
the engineers perspective, foreign views and data formats from other disciplines
are, in general, not visible [
CPPS architectures. We aim at proposing an information system design that will improve the data exchange and integration in Multi-Disciplinary Engineering (MDE) (see Figure 2).
Domain Experts
*.pdf Import Engineers *.csv Import
The multi-aspect information system consists of three parts: (1) Data
integration will handle the import, transformation and integration to Common
Concepts (CCs) of engineering artefacts coming from data providers, such as
engineers or domain experts. (2) Engineering data logistics will handle the
common uni ed model. The data curator will be responsible for the management of
discipline-speci c concepts, their relations to CCs and semantic links between
engineering views. (3) Data delivery will handle speci c data consumer requests.
Data consumers will be able to request data deliveries (a) in their domain-speci c
hierarchy (e.g., a simulation view) or (b) as domain-agnostic networks (e.g., for
analysis tasks across several engineering views) [
The advantage of this information system design will be the exible adaption to di erent application environments and delivery needs. This will improve the work process of multi-disciplinary environments and facilitate more e cient data provision that will likely lead to better data quality for consumers. 2
The challenges to data integration (cf. Section 1) lead to an error-prone and ine cient knowledge creation and quality assurance in information system engineering in Industry 4.0, and can lead to information loss or silos.
System modelling in CPPS engineering is challenging due to
disciplinespeci c views, tools, languages and legacy system environments [
Inconsistency management and consistency checking in systems
engineering is crucial to organise collaborative organisation networks. Egyed et al. [
The thesis will explore methods that build on the strengths of these modelling and inconsistency management approaches and mitigate the impact of their shortcomings on the data exchange process.
Engineering data logistics [
Open challenges mentioned in this context are the lack of use case scenarios
and the need for an adaptable architecture to cover enterprise model evolution.
Tunjic et al. [
In this work, we build on the SUM concept and engineering data logistics to enable a common uni ed model with processes for distributed data integration and model evolution. 3
To address the challenges to data integration introduced in Section 1, we raise the following research questions (RQs).
RQ1: What are requirements and capabilities to enable multi-view model integration towards a common model view? RQ1 investigates the multi-view capabilities needed to integrate di erent viewpoints in CPPS engineering. We will elicit requirements from relevant industrial use cases for integrating heterogeneous views. We will also explore di erent data integration approaches and methods to nd common attributes and concepts among di erent system contexts. We will elicit these capabilities from literature and industrial use cases.
RQ2: What methods can address the multi-aspect model integration in CPPS engineering? This research question aims at developing and evaluating methods for multi-aspect model integration to address the requirements coming from RQ1. First, we will de ne a method for collecting engineering concepts and for de ning common concepts, such as products, production processes, or production resources, which link the discipline-speci c engineering views. Second, we will develop a method for designing a data integration pipeline consisting of multi-aspect model integration operations and process ow speci cation as a foundation for exibly designing and con guring data transformation capabilities for data integration and delivery (see Figure 2). Third, we will design a method for operating pipelines for data integration and delivery based on DevOps approaches.
RQ3: What information system design can automate multi-aspect model integration in CPPS engineering? RQ3 aims at designing and evaluating a system that automates tasks for exible multi-aspect model integration based on the methods coming from RQ2. We will develop a system design that will include (a) tool support for common concept de nition; (b) a Domainspeci c language (DSL) for data integration work ow speci cation considering approaches such as DevOps and business process modelling; (c) data integration operators that can be exible orchestrated to data integration pipelines. In this context, we de ne exibility as the ease to adapt the system design to di erent application environments, such as the number of engineering disciplines, data sources and stakeholder perspective. We will conduct a workshop with domain experts to evaluate the exibility of our approach concerning the e ort needed to conduct adaption tasks.
We follow the Design Science approach [
In the problem investigation phase, we will conduct a domain analysis to investigate what kind of data integration is required to support the CPPS engineering process and data exchange. Speci cally, we will focus on the key stakeholders' needs and processes and analyse their data models and existing standards and solutions. Consequently, we will develop a conceptual problem framework as a guiding use case for further system design and evaluation.
In the treatment design phase we will derive requirements for data integration needs from the conceptual problem framework. Candidate treatments from model-driven engineering, semantic web, and software engineering will be evaluated and investigated. Formal concepts from model-driven engineering, such as the Meta Object Facility (MOF), will be explored to derive a DSL for objectoriented meta-models. We will research semantic web-based methods on how to construct and combine discipline-speci c taxonomies and other structures. Also, linking and integrating these discipline-speci c taxonomies to a common model will be a topic of interest. Further, we will explore software engineering design patterns for modularising software system design.
In the treatment validation phase we will derive typical use cases for stakeholder goals. Typical aspects include (i) data integration during CPPS engineering, operation, e.g., integration of sensor data, (ii) connection to open interfaces, such as OPCUA, and (iii) query aspects of the CPPS for non-experts in information systems methods to analyse the process or to perform data quality tasks.
In the treatment implementation phase we will develop a prototype that realises the identi ed capabilities. 5
We will build on recent research of the Christian Doppler Laboratory on Security
and Quality Improvement in the Production System Lifecycle [
We have developed an engineering data exchange approach that focuses on
the data consumer's needs and requirements [
To support domain experts in their analysis tasks, we have explored
graphbased visualisation methods that support domain experts to review multi-view
engineering data [
We designed and prototypically evaluated an initial approach of a multi-view
model transformation pipeline using a common underlying model and automated
by a Continuous Integration (CI) server [
This thesis' aim is to overcome gaps and challenges in engineering data
integration by combining semantic and model-based approaches to facilitate a lossless,
transparent and comprehensive data exchange and transformation. However,
major challenges of these techniques are a steep learning curve, high setup costs,
scarce expert knowledge required to gain the expected bene t [
This doctoral thesis project is supervised by Stefan Bi . The nancial support by the Christian Doppler Research Association, the Austrian Federal Ministry for Digital & Economic A airs and the National Foundation for Research, Technology and Development is gratefully acknowledged.
Rinker et al.