Bending damage characterisation of FRP-wood beams based on acoustic emission and MFCC-LightGBM prediction model

This study investigates the mechanical behavior and damage evolution of Pinus sylvestris beams reinforced with various fibre-reinforced polymer (FRP) composites (carbon, basalt, glass, and hybrids). Four-point bending tests integrated with acoustic emission (AE) monitoring were conducted. Results indicate FRP reinforcement significantly enhanced load-bearing capacity, ductility, and bending strain energy by up to 62.78%, 68.78%, and 142.6%, respectively. AE analysis revealed three distinct damage stages: initial compaction, crack propagation, and unstable fracture. Dominant AE peak frequencies shifted from 0–80 kHz to 160–240 kHz, and RA-AF analysis indicated a transition from brittle tensile fracture to mixed tensile-shear failure. This reflects a shift from localized macrocracking to distributed microcrack accumulation. The probability density distribution of AE energy exhibited critical avalanche behavior, with the index k effectively characterizing damage progression. Furthermore, Mel-Frequency Cepstral Coefficients (MFCC) extracted from AE waveforms were used to develop a LightGBM-based damage state prediction model, achieving 91.74% accuracy. This research provides a robust theoretical and technical framework for optimizing the design and health assessment of FRP-wood structures.