On the potential of machine learning assisted tomography for rapid assessment of FRP materials with defects

The present contribution is concerned with the development of methods for a rapid assessment of defects in carbon fiber reinforced materials detected during a nondestructive inspection regarding their effect on the structural integrity of the components. The nondestructive inspection is performed by means of X-ray computed tomography. Subsequently, machine learning methods are employed to assess the effect of the detected defects on the strength of the material. The training data base for the machine learning scheme is determined numerically by the analysis of representative volume elements containing selected relevant defects. Their strength is characterized in terms of the Puck failure envelope. The method is demonstrated and validated against experimental data for a space grade CFRP material containing manufacturing induced defects.