Damage Mode Identification in 2D Triaxial Carbon Fiber Braided Composites via Acoustic Emission Segmentation and Variational Mode Decomposition Optimization

ABSTRACT Acoustic emission (AE) technology is commonly employed to identify damage modes in carbon fiber–reinforced composites under deformation, yet remains challenging due to signal aliasing in AE data. This study investigates AE‐based damage characterization in two‐dimensional triaxially braided composites and proposes a comprehensive signal processing framework to address overlap in both the time and frequency domains. A novel AE signal segmentation method is developed by integrating root mean square (RMS) analysis with the Akaike Information Criterion (AIC) to improve the resolution of temporally overlapping AE events. For spectral decomposition, variational mode decomposition (VMD) is employed, with parameters adaptively optimized via the neural population dynamic optimization algorithm (NPDOA). A new similarity metric continuous wavelet transform‐based image subtraction (CWTIS) is introduced and combined with permutation entropy to construct a dual‐objective fitness function for VMD parameter tuning. Noise‐dominated components are filtered based on CWTIS scores, and physically relevant intrinsic mode functions (IMFs) are reconstructed into interpretable sub‐signals. Experimental validation through uniaxial tensile tests demonstrates that the proposed framework increases the number of detected AE events by 136% and significantly improves cluster separability across damage modes. Notably, the detection of matrix cracking signals improves by over 1100%, revealing previously masked low‐amplitude damage. Overall, the proposed method enhances the fidelity, granularity, and interpretability of AE signals and provides a robust tool for damage diagnosis and structural health monitoring in advanced composite materials.