Acoustic emission-based interpretable unsupervised clustering for damage pattern recognition in steel-concrete hybrid structures

A steel-concrete hybrid structure with characteristics of multi-materials and multi-interfaces, resulting in a complex damage pattern. The identification of damage patterns of steel-concrete hybrid structure is a challenging task due to the lack of sufficient prior knowledge. Here, an interpretable unsupervised clustering method based on the AE technique is proposed to effectively identify the damage patterns of steel-concrete hybrid structures under axial compression. A two-stage clustering parameter selection method was proposed based on Pearson correlation coefficient method and Principal Component Analysis, followed by the application of the K-means+ + clustering algorithm to categorize the AE data. Decision tree algorithm was used for interpretation and validation after clustering. The correlation between AE classifications and structural damage patterns was established by analyzing the features and time evolution laws of the AE data. The study reveals that five types of damage patterns can be distinguished based on AE features. The clustering results effectively highlight similarities and distinctions in damage evolution among different structures. This research extended damage pattern identification using AE data beyond homogeneous structures to complex steel-concrete hybrid structures.