<p>Accurate and efficient defect identification and characterisation in non-destructive evaluation (NDE) are key elements for any safety-critical industry. Many NDE modalities, such as ultrasonic inspection or radiography, may encounter identification difficulties stemming from structural features that cause the appearance of artefacts in the data. In the case of ultrasonic immersion using phased-arrays and the full matrix capture (FMC) approach along with the imaging method of the multi-view total focusing method (TFM), the signal features originated from boundary reflections (i.e., from the front and back walls) produce strong artefacts that can potentially obscure defects and make the region of interest relatively small. In this context, this paper proposes a deep learning approach based on defect-free ray-tracing model data in order to identify and suppress the artefacts in the FMC dataset efficiently and accurately. The deep learning models, which are based on the fundamentals of autoencoders, are formed of two parts: (1) an encoder model that takes as input the raw FMC data and gives as output some physical parameters such as the specimen thickness or the probe location; and (2) a decoder model that takes as input the physical parameters from (1) and provides the arrival times of the boundary reflections that causes the artefacts. Then, masking windows are created around the predicted arrival times which effectively suppress the artefacts from the time-domain data. The methodology, which is illustrated using several case studies, demonstrates an accurate prediction and suppression of artefacts, rendering a larger and cleaner region of interest in the ultrasonic images while making them more easily interpretable.</p>

Artefact Identification and Suppression for Ultrasonic Nde Data Using a Model-Assisted Deep Learning Approach

Paper Proceedings: https://asmedigitalcollection.asme.org/qnde/QNDE2021/volume/85529

ASME 2021 48th Annual Review of Progress in Quantitative Nondestructive Evaluation (QNDE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.