Multi-view ultrasonic imaging enables a region of interest within a specimen to be imaged using different ray paths and wave mode combinations. These can be utilised to maximise the viewing angle range from a single location, increasing the likelihood of an observable defect signal in one or more images. This both increases the likelihood of defect detection, but also provides multiple data points for use in characterisation. Automated methods for both problems are desirable, to both provide consistent outcomes by avoiding human operator decision making, as well as increase data throughput. 
In this work, we apply machine learning (ML) algorithms to ultrasonic images generated using the multi-view total focusing method (TFM) for both detection and characterisation. We consider a typical oblique weld-inspection setup, widely used across industry. Off-the-shelf ML models are utilised to demonstrate the ease of applying these models to a new NDE inspection scenario. The Random Forest model is utilised for detection, with a deep learning model for characterisation. The models are trained on simulated data, with experimental data used during testing. 
The defects of interest are small cracks of arbitrary angle. In addition, the training dataset also contains voids to ensure that the trained network can adequately distinguish between cracks and voids. The dataset also incorporates both inspection setup variation and misalignment errors. 
&nbsp;

Detection and Characterisation of Defects in Multi-View Ultrasonic Imaging Using Machine Learning

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.