Cluster analysis of acoustic emission data to investigate the damage evolution in modified scarf joint under bi-axial loading. Issue 11 (17th August 2020)

• Record Type: Journal Article
• Authors: Destouesse, J.; Diakhate, M.; Badulescu, C.; Thévenet, D.; Stackler, M.; Albouy, W.; Carrere, N.
• Is Part Of: Journal of adhesion. Volume 96:Issue 11(2020)
• Page Start: 969
• Page End: 987
• Publication Date: 2020-08-17
• Subjects: Modified scarf joint -- bonded joints -- acoustic emission -- non-destructive testing -- adhesion -- failure mode -- joint design; Adhesion -- Periodicals; 541.33
• DOI: 10.1080/00218464.2018.1552854
• Languages: English
• ISSNs: 0021-8464

Abstract

ABSTRACT: Non-destructive monitoring of damage evolution within material or bonding assembly becomes an essential tool to better understand its mechanical behavior, and therefore to prevent failure risks of engineering structures that involve adhesive bonding matters. This paper presents the experimental results of monotonic tests that were conducted firstly to investigate the effects of bi-axial loadings (with different shear/peel ratios) on the mechanical damage evolution of metal/metal bonded joint, and secondly to both detect and identify the acoustic emission (AE) signatures of the different failure mechanisms involved in the bonded joint damage. Results from specimens with modified scarf joint show that the loading configuration (shear/peel ratio) strongly influences the normal stiffness of the adhesively-bonded joint. For each loading configuration, repetitive tests were performed, and loading rate effects on the mechanical behavior of adhesively-bonded joint were analyzed. In addition to these results, a k-means++ algorithm was used to achieve a cluster analysis of AE data, and to allow AE events that were generated by damage evolution of the bonded joint to be identified. A particular AE signature is highlighted since it allows monitoring damage evolution of the adhesively-bonded joint. Test results also show that the highest value of acoustic energy is detected when the slope of the mechanical behavior curve (macroscopic scale) drastically changes. This finding is used to perform a real-time detection of the adhesive yield strength.