Classification of Fracture Types by Pattern Recognition Analysis of AE Signals from UD-GFRP and UD-C/C Composite

摘要

Acoustic emission (AE) signals from loaded UD-GFRP and UD-C/C composites were analyzed by utilizing the cross-correlation of Fourier transformed frequency spectra (FSC) and cross-correlation of Fourier phase images of wavelet coefficients (FPC-WC), and classified into four fracture types which are simultaneously determined by the waveform simulation of out-of-plane displacement. AE signals from tensile-loaded UD-GFRP with a side slit were classified into four fracture types with classification accuracy of above 56 percent. Higher classification accuracy in UD-GFRP is found to he due to a limited source location in front of the slit. Progression of fracture types of 627 events, classified by the FSC-WC and FPS, showed the same tendency for four fracture types. However, the classification accuracy of AE signals from four-point bent C/C composite with a side slit was lower than 56 percent. Waveform simulation to the four fracture types in UD-C/C composite, in which the source locations extended to a broad area due to inter-laminar delaminations, demonstrated a poor classification due to the waveform similarity caused by a SP wave.