Quantitative reconstruction of defects in multi-layered bonded composites using fully convolutional network-based ultrasonic inversion

Ultrasonic methods have great potential applications to detect and characterize defects in multi-layered bonded composites. However, it remains challenging quantitatively reconstruct defects, such as disbonds, that influence the integrity of adhesive bonds seriously reduce strength assemblies. In this work, an ultrasonic method based on supervised fully convolutional network (FCN) is proposed high contrast hidden training process method, FCN establishes a non-linear mapping from measured data corresponding longitudinal wave (L-wave) velocity models predicting process, obtained used directly L-wave new adhesively The presented FCN-based inversion can automatically extract useful features Although computationally expensive prediction itself online phase takes only seconds. numerical experimental results show capable accurately reconstructing which has for detection