The IOTWINS project will deliver large-scale industrial test-beds leveraging and combining data related to the manufacturing and facility management optimization domains, coming from diverse sources, such as data APIs, historical data, embedded sensors, and Open Data sources. The goal is to build a reference architecture for the development and deployment of distributed and edge-enabled digital twins of production plants and processes. Digital Twins collect data from manufacturing, maintenance, operations, facilities and operating environments, and use them to create a model of each specific asset, system, or process. These models are then used to detect and diagnose anomalies, to determine an optimal set of actions that maximize key performance metrics. IOTWINS proposes a hierarchical organization of digital twins modeling manufacturing production plants and facility management deployment environments at increasing accuracy levels:
- IoT twins: featuring lightweight models of specific components performing big-data stream processing and local control for quality management operations (low latency and high reliability);
- Edge twins: deployed at plant gateways and/or at emerging Multi-access Edge Computing nodes, providing higher level control knobs and orchestrating IoT sensors and actuators in a production locality, thus fostering local optimizations and interoperability;
- Cloud twins: performing time-consuming and typically off-line parallel simulation and deep-learning, feeding the edge twin with pre-elaborated predictive models to be efficiently executed in the premises of the production plant for monitoring/control/tuning purposes
IoTwins claims that IoT, edge computing, and industrial cloud technologies together are the cornerstones for the creation of distributed digital twin infrastructures that, after test-bed experimentation, refinement, and maturity improvements, can be easily adopted by SMEs: (i) industrial cloud, also based on HPC resources, enables the creation of accurate predictive models based on advanced ML for end-to-end deep networks, which require huge computing power for training; (ii) elastic cloud resource availability creates the opportunity to boost model accuracy by fitting and complementing data produced by industrial IoT sensors with data produced by large-scale parallel simulation; (iii) edge computing makes it possible to close the loop between accurate models and optimal decisions by enabling very responsive on-line local management of operational parameters in the targeted plants and filtered/fused reporting to the cloud side of only significant monitoring data (e.g., anomalies and deviations); and (iv) edge computing can leverage and accelerate the adoption of digital twin techniques by exploiting its industry-perceived advantages in terms of increased reliability/autonomy (e.g., independently of continuous connectivity to the global remote cloud) and of improved locality preservation of critical production data that can be maintained and used directly at the plant premises.
TTTech Industrial AG together with TTTech will contribute to the project by developing a digital twin layer at the edge level providing the edge computing platform (NERVE) as central technology in the large-scale testbed together with FILL GmbH. The FILL-TTT testbed is aimed at creating multiple target-oriented digital twins of machine tools to produce automotive components/parts. It will deploy simulation and ML models of machine tools, drives, and spindles for detecting the condition and behaviour of the machine tool, drives and spindle themselves, to predict manufacturing-relevant and quality-influencing parameters (load, forces, vibrations etc.) for reducing unexpected rejects, breakdowns and downtime, by optimizing load and performance indexes.
The main target of the developments planned during IoTwins are: (i) design and installation of redundant IoT modules able to log data at high sample rates, and also to log and process in real time a set of data (stream processing) for creating embedded digital twin models of machine tools; (ii) creation or forecasting of data coming from individual subsystems in a machine tool by using simulation models of physical behaviour.
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 857191.