AI Assisted: Automated Quantification of Barely Visible Impact Damage in CFRP Laminates from SE UT Scans Verified with XCT Results

The inclusion of carbon fiber reinforced polymer (CFRP) composites in advanced designs continues to increase across industries due to their high strength-to-weight ratio, corrosion resistance, and tailorable properties. With this increased use of CFRP composite laminates in products such as airfoil skins, wind turbine blades, and car hoods, overall performance increases, but at the risk of exposure to bird strikes, accidental tool drops, hail, runway debris impact, etc. Such impact damage in CFRP composite laminates is often difficult to detect, let alone quantify, with visual inspections due to the opaque nature of the carbon fibers. Often, nondestructive testing (NDT) techniques are required to detect this barely visible impact damage (BVID). NDT techniques such as thermography, X-ray computed tomography (XCT), and ultrasonic testing (UT) are all effective methods of detecting BVID in composites. Emphasis is placed on ultrasonic testing in this study due to its portability and safety, while still maintaining accurate BVID quantification. The focus of this study is improving the analysis process of ultrasonic data, both in accuracy and automation of BVID quantification. This study utilizes a custom MATLAB script which applies waveform peak analysis in combination with artificial intelligence (AI), specifically k-means clustering (KMC), to better assess BVID in UT scan data with reduced user input. This extended abstract is part of a study that quantifies BVID in CFRP laminates which were scanned with a pulse-echo, single element ultrasonic transducer in a custom portable ultrasonic testing setup. The novel, automated method of BVID UT analysis discussed above is compared to the conventional method of UT analysis that produces less optimal BVID measurements and requires extensive operator input and knowledge. Additionally, both UT analysis techniques are compared to XCT scans to further validate the increased accuracy of the improved UT analysis technique.