RECOGNITION OF DEFECTS IN CARBON-FIBER REINFORCED EPOXY COMPOSITE STRUCTURES USING NON-DESTRUCTIVE TESTING

摘要

The aim of this research was to enhance automated defect detection in composite structures using Non-Destructive Testing (NDT). A series of data was collected to define and test algorithms for the recognition of defects in carbon fiber reinforced epoxy composite structures. Artificial defects (delaminations and voids) of various sizes were intentionally introduced at different depths and characterized using NDTs such as phased array ultrasound (PAUT) and laser shearography. These two methods were compared in their effectiveness to detect delaminations and voids in composite structures. Results suggest that PAUT is more effective to detect delaminations whereas laser shearography seems more suitable for detection of voids. Support Vector Machine (SVM) image and data analysis were performed on the NDT data to identify characteristic features, which were used to define specific algorithms for detection of specific defects in composite structures. Using the developed algorithms, it was possible to recognize the delamination defects in composite structures with an accuracy of 90 - 95%. Defects could be nicely plotted in a 3D space enhancing their detection.