Non-Destructive Testing of 3D-Printed Samples Based on Machine Learning

Mostafa El Saadouny, Jan Barowski, Ilona Rolfes

IEEE MTT-S In­ter­na­tio­nal Micro­wa­ve Work­shop Se­ries on Ad­van­ced Ma­te­ri­als and Pro­ces­ses for RF and THz Ap­p­li­ca­ti­ons (IMWS-AMP 2019), pp. 22-24, DOI: 10.1109/IMWS-AMP.2019.8880141, Bo­chum, Ger­ma­ny, July 16-18, 2019


The three dimensional printing is a very important technology that participates in many applications. In this paper, we present an approach for the Non-Destructive Testing (NDT) of the three dimensional printed objects. This methodology solves the image classification problem by using the Neural Networks as they are capable of making good decisions and classifying images by proper training. The network has been trained by a large number of images of the tested sample layers. The proposed solution has been used for testing different sets of actual data for monitoring the performance under different scenarios, and the obtained results show a high degree of accuracy regarding image classification and defect detection.

[IEEE Library]