Monitoring Damage Evolution of CFRP via 1D-CNN-Based Feature Fusion of Acoustic Emission and Electrical Resistance Measurement

Identifying damage processes in carbon fiber-reinforced polymer (CFRP) is critical for inspection. Existing detection methods primarily rely on single-sensor techniques, which often fail to fully capture damage evolution. Electrical resistance measurement (ERM) and acoustic emission (AE) are two key online damage monitoring techniques. This work proposes a novel method for CFRP damage identification and classification. Simultaneous online monitoring of two specimen types under tensile loading was conducted using ERM and AE. Our findings reveal that the damage evolution of the specimens was divided into three stages; ERM provides more information in the early tension stage, while AE is more sensitive to damage in the later stages. Data fusion techniques combined with a 1-D convolutional neural network (1D-CNN) model were employed to classify the damage evolution process. Compared to single-sensor classifications, the AE-ERM dataset classified by 1D-CNN achieved improvements of 56.0% and 17.2%, and 1D-CNN outperformed both support vector machines (SVMs) and long short-term memory (LSTM) models. Digital image correlation (DIC) and microscopic images characterized damage modes such as matrix cracking, delamination, and fiber breakage. The measurements from ERM and AE were consistent and effectively validated the observed damage mechanisms. The proposed method leverages the complementary nature of multisource sensor data to provide a comprehensive description of damage evolution and enable in-depth identification of damage progression.