Credal identification of damage patterns in ultra-thin-ply composite bonded/bolted interference-fit joints

Abstract

Investigating the mechanical behavior and damage patterns of ultra-thin-ply composite joints is essential for ensuring their widespread application in the aerospace field. It is currently an advanced method of exploring the damage patterns of composite joint structures via a combination of acoustic emission and machine learning algorithms. However, the existing research lacks the evaluation indexes for the determination of the number of damage categories, and on the other hand, the overlapping area and noise of the monitored damage data cannot be accurately identified. In this paper, load-bearing performance of ultra-thin-ply composite hybrid bonded/bolted interference-fit joints is studied, then the SDI(Structural Damage Index) evaluation index is proposed to quantitatively analyze the number of damage categories of the complex joints, and finally the adaptive optimization algorithm based on the credal partition for the complex joints of composite is studied. The experimental results demonstrate the accuracy and generalization of the SDI formulae, and the introduction of the credal partition that submits data where overlapping regions cannot be accurately identified to ensembles of special single classes (meta-clusters) obtains deeper insights into the damage data and improves identification and isolation of noise.