Convolutional Neural Network Model to Detect Internal Markers on 3D Printed Polymer Samples Using Ultrasound Imaging

Abstract

While AI and imaging technologies are dramatically transforming the process and machine condition monitoring, product inspection remains confined to probing the geometry and surface morphology. Subsurface and bulk inspection to detect defects, embedded anticounterfeit tags remains prohibitively slow and imprecise (Anticounterfeit tags are introduced inside parts using 3D printing technologies to improve the integrity of these tags). Our explainable AI (XAI)-infused ultrasound imaging technology would achieve “quick” and “accurate” scans that will extract information (defects and codes) stored in these parts. It can could profoundly impact industrial product quality and (cyber)security assurance technologies. CT scanning for a 2 cm cube can take upto 8 hours while scanning with ultrasound devices can take 30 seconds. Ultrasound scanning technology can obtain phenomenal improvements in scanning times but often have poor resolution in comparison to CT scans. Surface marker technologies like QR codes and bar codes are widely used to identify parts, but their presence on the surface makes them vulnerable to damage. Alternate identification strategy includes RFID tags, which can be expensive and cannot be introduced on all types of manufactured parts. Our new technology uses fast scanning ultrasound imaging technique coupled with a convolutional deep learning software framework to obtain speed and accuracy in scanning the internal markers and defects present inside industrial parts. The technology contains an ultrasound reading device / transducer and a machine learning software. The ultrasound transducer scans the interior of 3D printed parts for markers. Here we specifically focus on 3D printed parts made of Veroclear polymer containing spherical markers with diameter in the order of 0.48mm. The scanning images obtained from the transducer is processed by a machine learning software to identify markers in these images. In this manner, we extract the information stored in these markers based on their physical locations within the 3D printed part.

Contact Information

Name: Sheikh Ismail

Email: smismail@tamu.edu

Phone: 979-862-3273