ML-Enabled Piezoelectric-Driven Internal Defect Assessment in Metal Structures

Abstract

With the growth of 3D printing in the production space, it is inevitable that quality assurance will be needed to keep final products within the constraints of requirements. Also, the variety of materials that can be used with 3D printing has increased over the years. Testing also must consider the process of manufacturing. This paper focuses its efforts on the finished product and not the process of manufacturing. Ultrasonic testing is a type of nondestructive testing. The experiments performed in this study aim to explore the usefulness of ultrasonic testing in materials that are 3D printed. The two materials used in this study are steel alloy metals and aluminum blocks of the same dimensions—120 mm × 40 mm × 15 mm. These materials represent common choices in additive manufacturing processes. The chosen alloys, such as Aluminum (6063T6) and grade-304 stainless steel, possess distinct properties crucial for validating the proposed testing method. Metal 3D-printed materials play a pivotal role in diverse industries, since ensuring their structural integrity is imperative for reliability and safety. Testing is crucial to identify and mitigate defects that could compromise the functionality and longevity of the final products, especially in applications with demanding performance requirements. An ultrasonic transducer is used to scan for subsurface defects within the samples and an oscilloscope is used to analyze the signals. Furthermore, several Machine Learning (ML) techniques are used to estimate the severity of the defects. The application of Machine Learning methods in the manufacturing industry has proven advantageous in terms of detecting defects due to its practicality and wide application. Due to their distinct benefits in processing image information, convolutional neural networks (CNNs) are the preferred method when working with picture data. In order to perform binary and multi-class classification, support vector machines that employ the alternative kernel function are a viable option for processing sensor signals and picture data. The study reveals that ultrasonic tests are viable for metallic materials. The primary objective of this work is to evaluate and validate the application of ultrasonic testing for the inspection of 3D-printed steel alloy metals and aluminum blocks. The novelty lies in the integration of Machine Learning techniques to estimate defect severity, offering a comprehensive and non-invasive approach to quality assessment in 3D-printed materials. The proposed method can successfully detect the presence of internal defects in objects, as well as estimate the location and severity of the defects.