Full-range stress-strain curve estimation of aluminum alloys using machine learning-aided ultrasound

Seong-Hyun Park; Junyeon Chung; Kiyoon Yi; Hoon Sohn; Kyung-Young Jhang

doi:10.1016/j.ultras.2023.107146

Full-range stress-strain curve estimation of aluminum alloys using machine learning-aided ultrasound

Ultrasonics. 2023 Dec:135:107146. doi: 10.1016/j.ultras.2023.107146. Epub 2023 Aug 26.

Authors

Seong-Hyun Park¹, Junyeon Chung², Kiyoon Yi², Hoon Sohn², Kyung-Young Jhang³

Affiliations

¹ Department of Mechanical Engineering, Stanford University, Stanford, CA 94305, USA.
² Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology, Daejeon 34141, South Korea.
³ School of Mechanical Engineering, Hanyang University, Seoul 04763, South Korea. Electronic address: kyjhang@hanyang.ac.kr.

PMID: 37657170
DOI: 10.1016/j.ultras.2023.107146

Abstract

Full-range stress-strain (SS) curves are crucial in understanding mechanical properties of a material such as the yield strength, ultimate tensile strength, and elongation. In this study, a full-range SS-curve was nondestructively estimated by applying machine learning to the ultrasonic amplitude-scan signal propagated through the material. The performance of the developed technique was validated using five-hundred aluminum alloy specimens with a wide spectrum of mechanical properties. The analyses of various ultrasonic properties, including nonlinearity and attenuation, with respect to the elements in the SS curves revealed how ultrasonics can be used to predict the SS curves without conventional destructive tensile testing. The proposed technique has significant potential for new applications in the fields of materials science and engineering, such as inline SS curve estimation during manufacturing.

Keywords: Machine learning; Nondestructive testing; Stress–strain curve; Ultrasonic tensile testing.