머신러닝 기반 음향방출 특성분석에 의한 자가천공리벳 접합부의 파손모드 분류

초록

In order to improve the strength of vehicle parts, many efforts have been made to apply aluminum-CFRP (Carbon Fibre-Reinforced Plastic) hybrid composite materials to vehicles. Self-piercing rivet (SPR) is one of methods joining composites to metals mechanically. In this study, lap shear tests were performed on SPR bonded aluminum- CFPR hybrid composite specimens to identify the failure modes of SPR joints. Also, acoustic emission (AE) tests were conducted with mechanical tests and then failure modes in the SPR joints were classified in terms of amplitude, peak frequency, and energy. Subsequently, aforementioned AE parameters were clustered based on various unsupervised learning methods, and the results were compared with the results of fractography analysis to confirm the accuracy of the learning methods. Six kinds of unsupervised clustering methods, K-means, Minibatch K-means, Density-based spatial clustering of applications with noise (DBSCAN), Gaussian Mixture Model (GMM), Spectral clustering, and Hierarchical clustering, were applied. As a result, it was found that K-means and GMM showed the highest accuracy representing 94.6% and 93.7% respectively among the methods. Accordingly, the machine learning-based failure mode classification methodology presented in this study is thought to be able to utilize not only in the failure prediction of various hybrid composite materials but also in the reliability evaluation of various vehicle parts.

키워드