Research mission and vision
The research scope of CDL-MINT is industrial engineering as well as Model-Driven Engineering (MDE).
In particular, the application of MDE concepts, techniques and methods for industrial engineering will be investigated, while at the same time further developments of MDE fundamentals to address the upcoming challenges in industrial engineering will be provided and evaluated.
Our main focus is on providing a Digital Twin Platform, which is located between design-time and runtime. The Digital Twin Platform is bridging both worlds by comprising design models as well as operation models. The prerequisite is to have models coming potentially from different tools and disciplines hosted in the platform in a well-connected manner.
Based on this foundation, we are focusing on the one hand on how to manage simulation artifacts in the repository covering both, design information, such as simulation models, as well as runtime information, such as simulation runs. Thus, the platform is covering both design models as well as simulated runtime models and has to provide a megamodel, i.e., a structure for storing simulation-related artifacts, in a systematic manner covering also their workows, as well as services for running, validating, and coordinating cooperative simulations.
On the other hand, we study how model repositories are enhanced to provide reactivity. In particular, the focus is on how to connect runtime environments, such as different Internet-of-Things (IoT) platforms, to model repositories to extract operation models and their connection to design models. Thus, specific services are needed to connect to runtime environments and to deal with model streams to efficiently react to events occurring in highly distributed systems on the model level.
Module Cooperative Simulation Megamodels
We approach identified research issues along the globally shortcomings, namely “Isolated Model Simulation” and “Design Models as Early Systems Snapshots” in the context of production engineering and simulation.
We are focusing on two main research directions:
Handling heterogeneous data
Heterogeneous data handling approches generally in terms of fundamental research as well as specifically towards production systems and simulation models.
Dealing with simulation models
In the simulation domain the main focus is on data generation and integration as well as on fundamental methods for the execution of hybrid simulations and simulation slicing.
Module Reactive Model Repositories
In Module 3, we are working on research challenges on how to connect runtime environments to temporal model repositories. For this purpose, we develop a framework to manage the full life-cycle of systems.
We present specific techniques needed to connect runtime environments and to deal with data streams to eciffiently react to events occurring in physically distributed data sources at the operational level.
We have three different contributions:
Managing Temporal Models
To provide a platform for digital twin engineering we need a representation in form of a framework and query facilities for temporal models in order to reason
about changes over time.
Evolutionary Model Mining
This umbrella term covers the evolutionary aspect of engineering artifacts, such as models, for describing their change over time. Monitoring this evolution bases on data collected over time from dynamic environments initially described by such engineering artifacts. The integration and unification of this observed data is a cross-disciplinary challenge where MDE techniques have to be combined with mining techniques such as process mining, Markov models, etc.
Model-based Digital Twin Engineering
To tackle the challenges of automating digital twins, we explore the creation, evolution, and migration of them. In this contribution, we combine the previously mentioned contributions to provide a framework for digital twins that allows querying temporal models in order to reason about changes over time.
Module Conversational, AI-Enhanced Digital Twins
In Module 4, we envision an enhanced support service for CPPS that in case of a fault allows us to loop back from operations to development. Faults and associated data are supposed to be annotated with human feedback, and learning methods are intended to draw conclusions about potential fault sources using an existing knowledge base.
The combination of both should enable us to reassign faults to their causative models and to capture the complete development cycle, which is eventually expected to shorten the response time to support requests.