Tensile Damage Identification in 3D Braided Composites Based on Multi‐Dimensional Acoustic Emission Features

ABSTRACT Three‐dimensional braided composites (3DBC) are pivotal in aerospace, yet real‐time damage monitoring remains challenging. While acoustic emission (AE) detects tensile damage sensitively, existing methods neglect multi‐domain AE features and lack generalizable models. This work overcomes these limitations via synchronized AE/CT analysis of 3DBC under graded tension, yielding 23,200 AE waveforms and 21,600 CT slices. We propose TDIA‐BPL, a hybrid algorithm integrating particle swarm optimization (PSO) with LSTM networks, to fuse time‐frequency‐cascade AE features with CT‐validated damage modes. The model achieves 86.7% accuracy (5.5% higher than LSTM alone) in classifying matrix cracking, interface debonding, fiber cracking, and fracture. Key innovations include: (1) multi‐domain AE feature fusion for precise damage evolution tracking, and (2) PSO‐optimized LSTM for enhanced generalization. This framework advances real‐time structural health monitoring in polymer composites, with direct applications in aerospace integrity management.