Lamb wave-based damage detection of composite structures using deep convolutional neural network and continuous wavelet transform

Abstract

Carbon fiber reinforced plastics composites have been widely used in aerospace, automotive and other fields for its superior performance. To avoid possible risks, Lamb waves-based structural heath monitoring techniques have been applied to detect the internal damage of these composites. However, due to their intrinsic high degree of anisotropy which makes the failure mode complicated under fatigue loads, an accurate and timely method for damage detection of these composites is still challenging to realize. This paper proposes a novel method for detecting the internal delamination of the carbon fiber reinforced plastics by combining deep convolutional neural network and continuous wavelet transform. This data-driven method can effectively leverage the large amount of data without relying on complex feature extraction. After converting the time series signal into a two-dimensional time–frequency image by the continuous wavelet transform technique, the convolutional neural network model is applied to classify the images. Based on the prediction results of different paths by the model, a damage localization method is also proposed for ply delamination. An experimental study is implemented to verify the effectiveness of this method. The result shows this method is a useful technique for precise diagnosis and positioning of delamination damage in the composite structures.