The use of ultrasonic guided waves for non-destructive testing and structural health monitoring (SHM) of metallic structures is well understood. Composite materials such as carbon fiber reinforced plastics (CFRP), however, are still a field of research in SHM; especially, because of their inhomogeneous material characteristics and spatially varying, non-linear effects due to temperature changes. Machine Learning can be an effective tool to compensate those effects. However, for security relevant scenarios, the machine learning algorithms need to be reliable, robust to environmental changes and explainable.For this contribution, the open-source dataset “Guided wave data for varying temperature” on the Open Guided Waves platform is chosen. Moll et. al. performed multiple measurements on a CFRP plate where they attached 12 transducers, setup in a pitch-catch scenario using frequencies ranging from 40 kHz to 260 kHz.&nbsp; During the measurements, the plate is exposed to varying temperatures between 20°C to 60°C. To simulate structural damage, metal disks as a reversible defect model were attached to the specimen by tacky tape at four different locations.By using the automated statistical model building toolbox DAV³E, developed by Bastuck et.al., multiple frequencies and transducer combinations were evaluated. The toolbox tests the preprocessed data (using a combination of optimal baseline selection (OBS) and baseline signal stretch (BSS)) with five complementary methods for feature extraction with three complementary methods for feature selection. Next, a linear discriminant analysis with Mahalanobis distance as classifier follows. The toolbox automatically selects the best combination of those algorithms. The authors were able to identify 40kHz as a promising excitation frequency and achieved up to 100% accuracy by using Best Fourier Coefficients (BFC) as Feature Extractor and a Support Vector Machine as Selector. In contrast to neural networks with limited explainability, the algorithms chosen by the toolbox are physically interpretable (DAV³E selected&nbsp;plausible features/frequencies that where close to the excitation frequency) and, to this end, preferable in critical applications.

Explainable Machine Learning for Damage Detection in Carbon Fiber Composite Plates Under Varying Temperature Conditions

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.