Fraunhofer IZFP

For the non-contact characterization of materials with regard to mechanical-technological properties, internal structures and identity, Fraunhofer IZFP is developing techniques that synergistically combine sensor effects at the hardware and software levels. Some such hybrid sensors are manufactured at Fraunhofer IZFP or purchased sensors are hybridized on the hardware side by the institute. In order to handle the multimodal data of such hybrid sensors in research and industrial applications, Fraunhofer IZFP has developed the Modular Measurement System MMS, a software framework that modularizes the measurement and data processing flow. So far, well over 100 software modules have been developed for triggering measurements, controlling actuators, querying sensors, signal processing, feature extraction and feature processing, as well as visualization and documentation. The modules can be freely combined. The latest extension of MMS is a visualization and classification software for multimodal data spaces based on machine learning, for which the functions of the open-source library Scikit-Learn were integrated into LabVIEW and a particularly user-friendly user interface was developed. Algorithms such as Kernel-PCA and Kernel-ICA play a role in identifying and separating nonlinear relationships between target features and metrological characteristics. The modularity makes the software very versatile, for example, the algorithms originally developed for the characterization of metals have been used directly for the analysis of gas chromatographic data of the Fraunhofer IVV. Measurement and testing technology of the Fraunhofer IZFP with MMS is used more than 200 times in industrial 24/7 applications. A current high-yield application is the screening of heavy plate for hard spots with magnetic sensor technology and machine learning techniques.

www.izfp.fraunhofer.de/en