Search

In this paper, an improved generative adversarial network (GAN) is proposed for the crack detection problem in electromagnetic nondestructive testing (NDT). To enhance the contrast ratio of the generated image, two additional regulation terms are introduced in the loss function of the underlying GAN. By applying an appropriate threshold to the segmentation of the generated image, the real crack areas and the fake crack areas (which are affected by the noises) are accurately distinguished. Experiments are carried out to show the superiority of the improved GAN over the original one on crack detection tasks, where a real-world NDT dataset is exploited that consists of magnetic optical images obtained using the electromagnetic NDT technique. Institutional Fund Projects; Ministry of Education and King Abdulaziz University, Jeddah, Saudi Arabia; National Natural Science Foundation of China; China Postdoctoral Science Foundation; Royal Society; Alexander von Humboldt Foundation

International audience; Nondestructive testing (NDT) techniques are used to evaluate the material degradation of ferromagnetic materials, for example, in sensitive environments, such as thermal power plants, where the materials are subjected to creep damage. There is no consensus on the use of an electromagnetic NDT technique to characterize the evolution of creep damage in high-chromium ferritic steels. In this work, an overview and comparison of three different electromagnetic NDT techniques that were applied to high-chromium steels is provided to understand creep evolution in terms of microstructural changes, such as precipitation, dislocation and grain size. To quantify the empirical measurements, a modelling technique was proposed for each applied method. The model parameters were optimized for each NDT technique and tested material. Depending on the model parameters, the accuracy of the parameter determination depends strongly on the NDT technique, which indicates its correlation with the microstructural information.

International audience; Abstract—Electromagnetic non-destructive testing of damaged multi-layer fiber-reinforced laminates is of concern. Each layer involves periodically positioned circular cylindrical fibers embedded within a matrix medium. Missing, displaced, shrunk, expanded fibers as well as circular inclusions inside a fiber are modeled as properly designed equivalent sources inside the undamaged structure. Then, the location of damages is retrieved via simply searching for such equivalent sources. A sparsity-constrained solution as well as classical algorithms including truncated singular value decomposition (TSVD), multiple signal classification (MUSIC) and basic matching pursuit (BMP) are considered, and their pros and cons are illustrated by numerical results in various configurations.