Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Applied Artificial Intelligence in NDE

 
: Osman, Ahmad; Duan, Yuxia; Kaftandjian, Valérie

:

Meyendorf, N.:
Handbook of Nondestructive Evaluation 4.0. Online resource : "Living reference work"
Cham: Springer Nature (Springer reference)
ISBN: 978-3-030-48200-8 (Online)
35 S.
Englisch
Aufsatz in Buch
Fraunhofer IZFP ()
image processing; deep learning; Classifiers; data fusion; performance metrics

Abstract
The fourth industrial revolution is driven by the digitalization. The artificial intelligence (AI) as re-emerging technology have seen an increased application since 2012 with the breakthrough achieved through the application of convolutional neural networks to the ImageNet Large Scale Visual Recognition Challenge. The nondestructive evaluation (NDE) is expected to profit from the digitalization technologies and mainly from the AI. This has led to the notion of NDE of the future or NDE4.0, which partially stands for the application of artificial intelligence to process and interpret input inspection data. The manual interpretation of the NDE data offers no satisfactory solution for the future NDE systems and the AI is expected here to provide reliable and trusted methods for the automated interpretation. Reliability, repeatability, and transparency of the AI algorithms in defect detection and decision-making are main requirements for a broad integration of this technology in the NDE world. This chapter introduces the machine learning classical approaches incorporating human expertise in handcrafted processing and feature extraction techniques. Then, the chapter explains data-driven approaches based on deep learning network which are similar to a human brain capable of modeling the human thinking, learn to segment, and analyze NDE data. Finally, the chapter gives some concrete examples of the application of AI in the NDE4.0 context, in particular for ultrasound, X-ray, and thermography. Assessment of performances is also presented as it is a key part of an automated approach. A brief presentation of data fusion is also given, as a mean to enhance reliability of inspection.

: http://publica.fraunhofer.de/dokumente/N-640823.html