Pattern Recognition Analysis of Acoustic Emission from Unidirectional Carbon Fiber-Epoxy Composites by Using Autoregressive Modeling

摘要

Frequency contents of acoustic emission (AE) from specially fabricated carbon fiber epoxy composites are investigated using autoregressive modeling. Unidirectional composite laminates were tested in tension. To identify failure mechanisms of composites by AE, the pattern recognition technique is applied with the autoregressive coefficients of AE signals as feature vectors. From a series of experiments, five reference sets of the autoregressive coefficients were extracted. By employing these sets as references, AE signals detected in unidirectional composite laminates with artificial flaw configurations were classified. Subsequently, each AE signal was assigned to a reference set by using both a minimum distance classifier and a maximum likelihood classifier. The results show that the pattern classification scheme successfully identified over 90% of the analyzed signals without relying on the peak amplitude information of AE signals.