Ultrasonic imaging
Ultrasound can be effectively generated by a short laser pulse due to the fast thermal expansion and contraction of a thin illuminated layer on the specimen surface. Internal features, together with geometric and material properties of the specimen, influence the propagation of the ultrasound, which is remotely detected by a laser interferometer.
We are using inventive software approaches based on machine learning algorithms to extract information-carrying parameters from the obtained signal. Varying the wave propagation properties, target parameters can be imaged in the desired field of view by a single-point ultrasound excitation and detection in the specimens of chaotic shapes.
The method is opening new possibilities for rapid, automated, and contact-free inspection of internal features such as cracks, delaminations, inclusions, and other material or geometrical imperfections.
Related publications
- Bakhtiyar Orazbayev and R. Fleury, “Far-field subwavelength imaging by deep learning”, Physical Review X 10, 031029 (2020). (web)
- Mark Buchanan, “Machine learning makes high-resolution imaging practical”, Physics 13, 124 (2020). (web)
- Anne-Muriel Brouet, “Deep learning and metamaterials make the invisible visible”, EPFL news, (2020). (web)
Industrial partners
- We are currently looking for industrial collaborations on this topic. If you would like to know whether our expertise applies to your challenging quality inspection tasks, please contact us.
Contact
- Dr Janez Rus, [email protected].