Defect detection in a CFRP containing FBH defects using Unet structure trained on successive wavenumber (SWF) filtering result of Guided Wavefield

In this project, first, successive numdamage maps corresponding to different center frequency of an CFRP containing Insert dataset are obtained after applying SWF. Then with constructing a hyper image constructed from previously obtained maps and extracting multiple local patches train set is constructed. Eventually, a specteral UNET structure is trained on the previously obtained hyper windows. At the end, the trained model is tested on a FBH dataset, revealing most FBH defects.