Powerful computers and low-cost digital sensor technology have created the hardware basis for collecting data on the state of a production system on a large scale. But to use this data in real time for better control and optimization, complex digital representations of the underlying physical systems are needed. This leads to the concept of the digital twin.
In the context of Industrie 4.0, the desired implementation of a digital twin is a simulation model of a real system that integrates the best available multi-physics and multi-scale models and is capable of real-time state updates based on sensor measurements. In this form, a digital twin goes far beyond a simple structured database solution. Instead, multiple heterogeneous data sources must be combined and synchronized; simulation models must be able to be coupled based on context, as well as replaced by surrogate models depending on the situation, providing a suitable mix of accuracy and computational speed. The full-scale realization of powerful digital twins therefore represents a major methodological and software engineering challenge.
With its many years of development and implementation experience in the areas of machine learning, optimization and simulation, Fraunhofer SCAI is an ideal research partner for industry. We are accompanying the topic of digital twins in several projects:
- EVOLOPRO (project duration: 11/2019 – 12/2023): In the context of the lead project EVOLOPRO, we are conducting research on methods that allow automating the contextual generation of multifidelity surrogate models.
- digitalTPC (project duration: 02/2019 – 01/2022): The potential of digital twins is still largely untapped for cross-value chain and material-triggered process control. This also applies to plastic-based composite structures. The digitalTPC project aims to demonstrate this potential using the large-scale hybrid injection molding technology that is just becoming established on the market, in which fiber-reinforced tape laminate semi-finished products are continuously formed and back-injected. digitalTPC aims to take a comprehensive and holistic view of all sub-process steps from semi-finished product to component manufacture. Relevant material, process and component characteristics (including fiber orientation, pore content, degree of consolidation, temperature, pressure, etc.) are to be measured, recorded and virtually modeled and analyzed in a digital twin, if possible at local resolution throughout the entire real value chain. The challenge in the project is the material- and process-related intelligent acquisition of the sensor data and their linkage with the continuous simulation chain within the framework of the digital twin by SCAI.
- MADESI (project duration 10/2018 – 09/2021): In the MADESI project, we have been working on how machine learning methods can be increasingly used in systems for predictive maintenance of wind turbines with the help of the interaction of sensor data and numerical simulation.
- VMAP (project duration 08/2017 – 08/2020): In the ITEA VMAP project, 30 industry partners have been working on a new standard for the exchange of material parameters and local condition descriptions in virtual manufacturing processes. This VMAP interface standard makes it much easier to map production processes and new manufacturing methods in digital twins. For the further development and the assurance and maintenance of a uniform library standard, the VMAP Standards Community (VMAP SC) has been established. It is open to any interested party that wants to use or contribute to the standardization efforts. Specification documents and directly related software components of the VMAP Standard are available to any interested party – VMAP SC members as well as external institutions – on a royalty-free basis.