Damage Characterization of Thermoset Composite Using Acoustic Emission and Deep Learning
Description
The vast application of carbon fiber-reinforced polymers (CFRPs) in automotive, aerospace, and construction industries is advancing due to their excellent load-bearing capabilities, and durability. These incredible mechanical properties call for a thorough study of the flaws in CFRPs. Under quasi-static tensile loading until failure, thermoset CFRPs exhibit three primary failure modes-fiber breakage, matrix cracking, and delamination. Real-time transient waves emitted by the failure events are acquired as acoustic emission (AE) signals for analysis. Due to the numerous AE signals generated during experiments, effective classification of the failure modes through manual inspection of the waveforms has been challenging. Therefore, deep learning models were developed to study and classify the failure mechanisms efficiently and accurately. One-dimensional (1-D) and two-dimensional (2-D) convolutional neural network (CNN) models were trained using the signal amplitudes and spectrogram images, respectively, as training data. A thorough data processing was done to remove any outliers that may impair the performance of the models. The performances of the two approaches were compared using model evaluation metrics. Although the 1-D CNN model yielded an accuracy of 97%, the 2-D CNN model outperformed the 1-D CNN model with a flawless accuracy of 100%. The 2-D CNN model exhibited a perfect discriminative capability as depicted in the confusion matrix