ScholarWorks@동국대학교

초록

Composite structures have been excessively used in numerous engineering applications due to their weight-saving capabilities. Due to their orthotropic nature, they are prone to complex failure models that can be assessed through vibration-based data-driven methods. However, the vibrational signals obtained from the composite structures can be susceptible to noise which can restrict the performance of the data-driven methods. To overcome this challenge, this study proposes a vibration-based noise-resistant framework for detecting delamination in composite structures using a hybrid multi-channel convolutional autoencoder-support vector machine (MC-CAE-SVM) framework. The proposed methodology integrates advanced signal processing and deep learning. Empirical mode decomposition (EMD) is employed to decompose vibration signals into intrinsic mode functions (IMFs), with correlation analysis isolating noise-free IMFs. These IMFs are transformed into time-frequency scalograms via continuous wavelet transform (CWT) to enhance feature representation. Finally, the CAE model is used to extract robust features autonomously and classify them using a support vector machine (SVM) classifier. The proposed framework achieved an accuracy of 98.89% on unseen test data and demonstrated exceptional noise resistance under varying noise levels. This approach offers a reliable and scalable solution for detecting delamination in composite structures, even in noisy operational environments.

키워드