Fraunhofer Institute for Integrated Circuits IIS, Division Engineering of Adaptive Systems EASNewsletter 01/2023
Digitization has long since arrived in the world of industrial production. Modern production technologies are equipped with integrated sensors and record numerous pieces of data. As these systems often operate in isolation from one another, however, it is hard to make in-depth analyses. To bring about a change in this situation and drive process optimization, there is a need for new approaches to the use of existing data – e.g., by artificial intelligence. We talked to Dirk Mayer, an expert in distributed data processing and control at Fraunhofer IIS/EAS, about the integration of data sources.
What advantages arise for users in industry from the integration of data sources and what approach can be used to achieve this?
The integration of data from various data sources paves the way for even deeper analyses of data collected by sensor technology. Machine learning methods, which are able to perform classifications in large, complex databases, can also be used to identify smaller abnormalities. Whereas individual measured variables can only be used to detect imminent machine failures, they can also be combined with additional data sources from the production process in order to identify variations in quality and thus to extract further added value from data analysis.
In order to establish a procedure of this kind for the purposes of process optimization, Fraunhofer IIS/EAS and Fraunhofer IWU are collaborating as part of the project How2ForProMon based on the example of a forming press.
What kind of relevant data do these sensors deliver? What kind of insights does the project provide?
In the How2ForProMon project, sensors developed by Fraunhofer IWU are integrated into the press table and provide information on its elastic deformation. In addition, vibration sensors deliver information on dynamic effects during the forming process, and – lastly – ultrasound sensor technology can be used to detect microscopic events in the interaction between die and workpiece. Other relevant data are provided by the machine control unit. Now that these data have been integrated into the IIS/EAS “DeepInsights” IIoT platform, research is now underway into how artificial intelligence can improve the detection of subtle variations in process quality or damage to the die.
What special challenges does this create when it comes to system development?
Ultrasound sensor technology produces vast quantities of data due to the high sampling rates. Just a few sensors generate several terabytes per month, which must be transferred to and managed in the cloud. Implementing a system of this kind in an environment with multiple production systems requires in situ data analysis – “on the edge” – in order to avoid the disproportionate cost of expanding your own data infrastructure.
How can a powerful digitalization solution be implemented for industrial production?
Industrial production is extremely varied, for companies seek to create unique selling points in the form of customized processes and products. A powerful digitization solution therefore relies on tailored implementation that incorporates the existing systems and supplements them with powerful edge devices or intelligent sensor technology.
At the AI Application and Test Center, Fraunhofer IIS/EAS is working to bring together services including an initial readiness check at the company, the rapid carrying out of a feasibility study to validate the benefits, and the consolidation of a company digitization solution.