• Home
  • Blog
  • Fraunhofer Institute builds database with DIMATE

Looking into the future is a core competency of the Fraunhofer Institute for Nondestructive Testing (IZFP) in Saarbrücken, Germany. With its focus on application-oriented research and relevant key technologies it is the IZFP team that decides the course of research. During a technology screening four years ago, IZFP clearly identified the ‘next big thing’: cognitive sensor systems. A sensor system requires a physical unit and the ‘cognitive’ part hinges on an outstanding database that enables the development of AI applications.

Prof. Bernd Valeske, Copyright htw saar

The IZFP team has been working on both components. From the very beginning data collection and processing as well as the use of data for process optimization have played a major role. A look at conventional NDT illustrates why the digital transformation of NDT is a precondition for cognitive sensorics: conventional NDT follows agreed standards. Tests are done in pre-defined intervals and with pre-defined methods, no matter whether the test is necessary or no. Thus in many companies NDT, quite understandably, was considered nothing but a cost generator.

The sensor system as ‘object brain’

Cognitive sensor systems can revolutionize the test process: testing no longer follows pre-defined standards but is determined case by case based on continuous measurements performed by the sensor system, be it in production, predictive maintenance or other fields of application. The sensor system, however, does not only measure – it evaluates the data with the help of artificial intelligence.  Thus, a sensor system might be able to ensure consistent quality in a production process or it might recognize, analyze and classify material deterioration and decide whether, when and where NDT has to be performed. “Modern sensor systems accompany a component or a machine during its entire life cycle: they collect data, analyze it and make autonomous decisions. But they have to learn how to do that and for this learning process a large pool of harmonized and well-prepared data is required. Moreover, we need a system that can read, forward, evaluate and archive this data. This is where the DIMATE PACS comes in. We use the software package in order to learn from our research data“, explains Professor Dr. Bernd Valeske, Deputy Director of the Institute and Head of Department, Algorithms/Signal and Data P

Dr. Ralf Tschuncky, Copyright Uwe Bellhäuser

Data format: consistent, independent, DICONDE

Data format might sound like a trivial issue but far from it. NDT works with an extremely broad range of formats as each measuring method, indeed some manufacturers of test equipment use their own data format. This makes merging and analyzing all data linked to a component well-nigh impossible. “The first step towards autonomous cognitive sensor systems is the harmonization of measurement and test data in order to be able to create a usable data pool. Only a sensor-independent format will allow us to generate added value with our research data and in the longer run with customer data”, explains Dr. Ralf Tschuncky, Lead Researcher at Fraunhofer IZFP.

The research team opted for DICONDE (Digital Imaging and Communication in Non-Destructive Evaluation) as sensor-independent generic data format. DICONDE is not only internationally known and recognized, it also offers the ability to document metadata and merge it with the test data into a single data format.

DIMATE – the DICONDE pioneer

DIMATE is one of the few test data management solutions – if not the only one – that is based on DICONDE and can thus process and archive manufacturer-independent and modality-independent test data. “We knew DIMATE from several work groups in the German Society for Non-Destructive Testing – DGZfP. The DIMATE solution ticked most of our boxes which immensely facilitated work on our data base as we were able to build upon existing server architecture, many functionalities and an intuitive user interface”, says Ralf Tschuncky.

Bernd Valeske points out two more aspects that tipped the scale towards DIMATE: an excellent digital workflow and rights management: “We have several work groups focusing on different measuring methods. Each team collects their data, describes it in DICONDE format and inputs it in the system. Thus, one object is linked with data from different teams. This is a precondition for machine learning and for the development of algorithms for AI applications. At the same time we are an accredited and certified research institution and have to comply with strict privacy laws. Thus we need a very good access and rights management system which ensures that the researchers can access the data whenever they need to in the course of their work.”

DIMATE thus kills two birds with one stone: data can be merged, analyzed, processed and archived in a meaningful manner and all applicable privacy laws are complied with. The development of AI models that are used in intelligent sensors can now proceed without a hitch!

Header: Picture Fraunhofer IZFP (Copyright: Frank Blümler)

 

Go back