Abstract
Deep learning has helped achieve breakthroughs in a variety of applications; however, the lack of data from faulty states hinders the development of effective and robust diagnostic strategies using deep learning models. This work introduces a transfer learning framework for the autonomous detection, isolation, and quantification of delamination in laminated composites based on scarce low-frequency structural vibration data. Limited response data from an electromechanically coupled simulation model and from experimental testing of laminated composite coupons were encoded into high-resolution time-frequency images using SynchroExtracting Transforms (SETs). The simulated and experimental data were processed through different layers of pretrained deep learning models based on AlexNet, GoogleNet, SqueezeNet, ResNet-18, and VGG-16 to extract low- and high-level autonomous features. The support vector machine (SVM) machine learning algorithm was employed to assess how the identified autonomous features were able to assist in the detection, isolation, and quantification of delamination in laminated composites. The results obtained using these autonomous features were also compared with those obtained using handcrafted statistical features. The obtained results are encouraging and provide a new direction that will allow us to progress in the autonomous damage assessment of laminated composites despite being limited to using raw scarce structural vibration data.
Highlights
The use of composite materials is growing continuously in a variety of industries and remains at the forefront of contemporary research [1,2,3,4,5]
Statistical features may not allow us to discriminate between the health nated composite samples in the simulations and experiments based on both handcrafted states of laminated composites whereextracted the dynamic response is dominated by the excitastatistical features and autonomously features
In the deep learning framework, pretrained models can be employed for three purposes: to make predictions on new unseen data that is similar to previous data, for feature extraction using the activations of deep layers as features, and for transfer learning based on fine tuning a network that was pretrained on data from a different but related task to work with limited new data [42,43]
Summary
The use of composite materials is growing continuously in a variety of industries (e.g., aerospace, automotive, wind energy) and remains at the forefront of contemporary research [1,2,3,4,5]. From the literature survey above, it is clear AI-based models are able to detect and classify faults in composite structures with a high level of accuracy. Large-scale training data is expensive and challenging to collect in certain scenarios such as aerospace applications To overcome this issue, transfer learning-based models have been introduced recently for intelligent fault detection. Experimental results showed that the proposed model could classify faults up to 96% classification accuracy. The results showed that the proposed model could successfully identify transferable features from the laboratory scale bearings to detect faults in locomotive bearings. The current work proposes aiding autonomous feature extraction from limited data using off-the-shelf pretrained deep learning models for the assessment of delamination in laminated composites.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
