Christian Doppler Laboratory „Software Engineering Integration for Flexible Automation Systems“ (CDL-Flex)

Research scope and mission. The general scope of research in the CDL-Flex is the analysis, automation, and improvement of software and systems engineering processes for complex software-intensive automation systems (AS), such as industrial production plants. Software and systems engineering projects in the context of the CDL-Flex aim at producing software-intensive automation systems and involve several engineering disciplines, such as automation process, mechanical, electrical, and (automation) software engineering. In such a software and systems engineering environment, the work of software engineers depends on the inputs from other engineering disciplines, e.g., requirements, process specification, and design constraints incorporated in a range of engineer-ing models.

Figure 1: Overview on the research challenges in the CDL-Flex.

Figure 1 illustrates the gap in automation systems engineering (ASE) between the engineers, who use a wide range of domain-specific tools to provide engineering data, and the need for effective and efficient access to the engineering tool data on the project level.

Major challenges for analyzing, automating, and improving software and systems engineering processes at the system level are (see the red numbered circles in Figure 1): (1) The heterogeneity and weak integration of soft-ware tools, often loosely coupled systems-of-systems engineering environments, which evolve in parallel with-out central control, make it hard to provide an engineering environment that routinely supports organizational policies and engineering best-practices. (2) The heterogeneous representations and weak integration of the engineering knowledge necessary for the development and validation of complex software-intensive automation systems. Therefore, (3) the access of project participants, who want to automate and improve project-level processes, to engineering tool data is, in general, inefficient and fragile. An example is the simulation of a complex AS based on the orchestration of heterogeneous simulation tools that were not designed to work together. A traditional integration approach, e.g., based on scripts, files, and databases, it is, in general, difficult and error prone to reuse for or adapt to a new or changed context.

Flexible AS are expected to allow effective, safe, and efficient reorganization at run time, e.g., in “Cyber-Physical Production Systems” (CPPS ) that provide desirable production capabilities but are also more complex to engineer and operate. Thesis 7 in the Industry 4.0 roadmap (Bauernhansl et al. 2014b) states that the network-ing and individualization of products and business processes creates complexity, which has to be addressed with modeling, simulation, and self-organization as a foundation to better analyze a larger solution space and find solutions faster. However, the larger solution space in making flexible AS adds considerable complexity to the engineering process, which needs improvements in the ASE process, in particular, better-integrated access to engineering data sources in distributed and parallel engineering.

Research mission and vision. To address these challenges, the mission of the CDL-Flex is the development and evaluation of concepts, methods, and tools for the integration of engineering knowledge, models, systems, and tools to enable applications for the system-wide analysis, automation, and improvement of engineering processes along the product life cycle of flexible software-intensive industrial automation systems. The approaches in the CDL-Flex provide the foundations for key aspects of integrated engineering methods and processes in ASE: tool and data integration, quality management, and advanced simulation engineering approaches.

The CDL-Flex method. To achieve the goals in the research vision, the core methods in the CDL-Flex research areas (see the research organization below) include: (a) methods for the representation of engineering knowledge with a focus on the “common concepts” used by the stakeholders in the automation system engineering (ASE) team; (b) methods for the integration of local heterogeneous engineering knowledge sources to enable their unified querying and automated transformation; and (c) methods for the representation, analysis, automation, and improvement of engineering processes based on integrated engineering knowledge.

Unfortunately, the complete engineering knowledge in several engineering domains is hard and inefficient to model and collect. However, the engineering knowledge regarding “common concepts” of used in an ASE project team are, by definition, familiar to the stakeholders, and, in general, available from existing engineering processes. Therefore, these common concepts are a sound basis to model integrated knowledge from several engineering domains, and to bridge the gap illustrated in Figure 1 for analyzing and improving selected engineer-ing processes.

The free basic research in the CDL-Flex consists of the investigation and improvement of methods and tools, which are primarily of interest for the scientific community and per se not relevant for company partners. The company partners benefit from the applied basic research results in their specific use cases. Based on specific research challenges coming from the company partners we have identified the following key use cases (UCs, see this link for a collection of UC descriptions and videos.) to derive basic research goals and to evaluate research results. Each research use case captures requirements for CDL-Flex method development and evaluation. (i) the “Semantic Dropbox” extends the functionality of the well-known Dropbox application with engineering data transformation, as foundation for traceable and quality-assured engineering tool chains; (ii) the “multi-model dashboard” enables efficient monitoring of design and project conditions in a heterogeneous and complex systems-of-systems environment; (iii) “ontology-based search across heterogeneous engineering models” bridges gaps in heterogeneous data models with semantic web technologies to enable the efficient answering of advanced queries by humans and machines over engineering knowledge; and (iv) “integrated simulation” builds on heterogeneous partial simulation processes and simulator model types to design simulation interfaces for advanced integrated simulation models.

Major complexity drivers for the research UCs are the number of engineering disciplines involved, the number of domains in AS engineering, the variety of software tools and tool types/data models, and the degree of distribution in the project organization. Evaluation measures include the effectiveness, efficiency, robustness, scalability, and usability of methods and research prototypes in laboratory and industrial contexts.

Research organization in modules and research areas. The research work in the CDL-Flex is organized in two modules and four research areas (RAs, see Figure 2). A third module consisting of a fifth RA is planned, but has not yet completed the application process for a new module.

Figure 2: Overview on the research areas and interfaces in the CDL-Flex.

Module 1 “Integration Foundations for Software Quality Improvement in Automation Systems Engineering” investigates knowledge representation and integration to provide methods and a software platform as foundation for investigating solutions for ASE process analysis and improvement. Module 1 consists of the following three RAs (see Figure 2):

The RA 1.1 “Technical Integration Foundations” (TI) investigates concepts for architecture integration in ASE to provide the Engineering Service Bus (EngSB) methods and tools (see in Figure 2 the green circles numbered 1, 2, and 3) for the integration of technologically heterogeneous software tools and engineering systems as foundation for the effective and efficient automation, analysis, and improvement of engineering processes. The RA TI captures requirements and capabilities for integrating heterogeneous systems of systems that are relevant for ASE in generalized use cases, based on external requirements and alternative solutions from the scientific communities systems interoperability and ASE. Use case solution models and processes have been based on requirements coming from researchers in the CDL-Flex research areas 1.2 SRI, 1.3 QM, 2.1 Sim and from company partners, captured in specific use cases.

The RA 1.2 “Semantic Representation and Integration of Engineering Knowledge” (SRI) investigates methods for knowledge extraction, representation and integration, as well as for engineering intelligence. It pro-vides the Engineering Knowledge Base (EKB) methods and tools (see in Figure 2 the green circles numbered 4 and 5) for the semantic knowledge representation and integration of heterogeneous engineering models. The EKB serves as a foundation for queries to engineering knowledge across engineering tools and domains. The RA SRI captures requirements and capabilities for integrating heterogeneous ASE data sources in generalized use cases based on external requirements and alternative solutions from the scientific communities semantic web and ASE. Use case solution models and processes have been based on requirements coming from researchers in the CDL-Flex research areas 1.1 TI, 1.3 QM, 2.1 Sim, and from company partners, captured in specific use cases.

The RA 1.3 “Quality Management in ASE” (QM) builds on the EngSB/EKB concepts, methods, and tools to investigate more effective and efficient approaches for engineering process analysis and improvement (EPA&I) according to the VDI 3695 guideline (see in Figure 2 the green circle numbered 6). Focus of defect detection is on classes of relevant defects that can be addressed better (e.g., earlier, more efficient) with views on integrated ASE knowledge and systems: (a) engineering application defects; (b) engineering model inconsistencies and constraint violations; and (c) analysis and improvement of real-life engineering processes. The RA QM captures requirements and capabilities for EPA&I in generalized use cases based on external requirements and alternative solutions from the scientific communities software product and process improvement and ASE. Use case solution models and processes have been based on requirements coming from company partners, captured in specific use cases and build on the 1.1 TI/1.2 SRI methods and tools.

Module 2 consists of the RA 2.1 “Advanced SCADA Algorithms for Flexible Automation Systems” (Sim, see SCADA & Simulation in Figure 2), which builds on the EngSB/EKB methods and tools provided by Module 1 to investigate the design of a Simulation Integration Framework with signal- and equation-based simulation components (see in Figure 2 the green circle numbered 7) for advanced failure detection and prediction applications for Supervisory Control and Data Acquisition (SCADA), one of the basic capabilities in ASE to monitor AS for safe and efficient production; and to investigate model-driven simulation integration based on single-input/single-output simulation components and simulation model design with the extended Bond graph approach. The RA Sim captures requirements and capabilities for integrated simulation in generalized use cases based on external requirements and alternative solutions from the scientific communities simulation engineering and ASE. Use case solution models and processes have been based on requirements coming from company partners, captured in specific use cases and build on the 1.1 TI/1.2 SRI methods and tools.

Plan for a new module “Model-Based Adaptation Engineering”. The recent Industry 4.0 trend towards more flexible cyber-physical production systems (CPPS), important kinds of flexible AS, will require advanced sup-port for the engineering process of CPPS, in particular, the engineering of safe adaptation processes in CPPS. Lead researchers from the CDL-Flex and other research groups at TU Wien are involved in a new doctorate college on CPPS , which is scheduled to start in December 2014. One of these lead researchers is Manuel Wim-mer, an expert in model engineering and the designated lead of the planned new research module.

To address the challenge of the systematic engineering of adaptation processes in CPPS engineering, we plan to add a new research module to the CDL-Flex, called “Model-Based Adaptation Engineering” (MAE). The RA 3.1 MAE shall build on the EngSB/EKB methods and tools provided by Module 1 to investigate concepts, methods, and tools for model abstraction and representation, model transformation, and model-driven design and exploration for ASE, in particular, for CPPS applications. Key challenges are the heterogeneous domain-specific representations of system models and the missing design method for an abstract safe adaptation process that leads to a valid solution in the domain-specific environment. The RA MAE will capture requirements and capabilities for model abstraction, model transformation, and model-driven design in generalized use cases based on external requirements and alternative solutions from the scientific communities (software) model engineering and ASE as well as from performing empirical studies. Use case solution models and processes will be based on requirements coming from the RA 2.1 Sim and from company partners, captured in specific use cases and built on the applicable 1.1 TI/1.2 SRI/1.3 QM methods and tools.

The first planned use case is the “safe adaptation of CPPS simulations”: model abstraction from an existing do-main-specific simulation, e.g., in Matlab Simulink, definition of adaptation rules as model transformations for the derivation of valid target simulation variants and the associated engineering steps, and model-driven design and exploration of the resulting simulation scenarios in a domain-specific simulation environment for evaluation of CPPS properties such as performance and robustness. As a result, engineers at customers of the company partner will be enabled to (a) specify variants of existing system models and (b) finding and evaluating process steps to get safely from the starting system model to the target system model.