Efficient analysis of high-dimensional ultrasound data in non-destructive testing
Joint BMBF research project
ZeMat is a joined BMBF research project established in July 2013 for efficient analysis of high-dimensional ultrasound data in non-destructive testing.

Goal of this project is the developement of a new generation of ultrasonic testing methods that are able to reconstruct the size and type of inclusions. These methods are of particular importance in non-destructive testing of savety related components. Due to physical reasons and a limited testing time the given data is highly incomplete and noise contaminated. This leads to mathematically challenging problems where conventional methods known from inverse scattering theory cannot directly be applied. The ZeMat research project was established to elaborate new approaches for a significant improvement of the existing reconstruction methods. Four subprojects together with the industrial partner Salzgitter Mannesmann Forschung are working on different fields including:

  • Applied model reduction allowing fast and robust reconstruction techniques
  • Adapted denoising algorithms based on an ultrasonic data analysis
  • Optimized probe control for time efficient testing without loss of information
  • Real time reconstruction algorithms based on sparsity assumptions (also for the 3D case)
  • Efficient and accurate 3D simulation of complex defect geometries based on scattering models
  • Defect classification based on non-linear machine learning for high dimensional data
  • Model validation using real and simulated data as well as results from established methods like GPSS or EFIT.