Deep Feature Extraction Based on AE Signals for the Characterization of Loaded Carbon Fiber Epoxy and Glass Fiber Epoxy Composites

Characterization of acoustic emission (AE) signals in loaded materials can reveal structural damage and consequently provide early warnings about product failures. Therefore, extraction the most informative features from AE is an important part characterization process. This study considers obtained bending experiments for carbon fiber epoxy (CFE) glass (GFE) composites. The research focused on recognition material structure (CFE or GFE) based analysis signals. We propose deep using a convolutional autoencoder (CAE). are compared with extracted standard features. Then, different feature sets analyzed through decision trees discriminant analysis, combined selection, to estimate predictive potential various sets. Results show that application increases accuracy. By only AE-based features, classification accuracy around 80% obtained, adding improves above 90%. Consequently, encouraged CFE