Frequency-based k-Nearest Neighbors algorithm for electromechanical impedance based detection of damage in adhesive joints

• Type: Article in International Scientific Journal
• Date: 2025

Authors

Tenreiro, AFG (Author) Other; António Mendes Lopes (Author) FEUP; da Silva, LFM (Author) Other

Journal

Title: Applied Soft Computing Journal

Vol. 171

ISSN: 1568-4946

Publisher: Elsevier

Indexing

ISI Web of Knowledge - 0 Citations

Scopus - 0 Citations

Other information

Authenticus ID: P-017-ZZ8

DOI: 10.1016/j.asoc.2025.112761

Abstract (EN): Structural adhesive joints are susceptible to many forms of damage that may not be detectable with currently deployed Non-Destructive Tests (NDT). The Electromechanical Impedance Spectroscopy (EMIS) based Structural Health Monitoring (SHM) methodology constitutes an alternative, where a given structure is continuously monitored for damage, thus outperforming NDT. While much has been done in the applicability of EMIS in metallic and composite structures, only preliminary research has been done on the EMIS-based integrity monitoring of adhesive joints. In this paper, two different k-Nearest Neighbor (kNN) approaches are used for both damage detection and damage quantification. Initially, peaks are extracted from the real component of the measured impedance of instrumented adhesive joint specimina, which are then inputted to either a conventional kNN or a novel parallel kNN model, where each individual model is fed with its respective peak information. For void detection, the parallel classifier approach presents a moderately better performance, with an average accuracy of 98% under optimal conditions, but, for damage quantification, a significant improvement in classification is observed. In all cases, the use of the Canberra distance allows for a significant increase in classification accuracy.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 18