Expert System for Online Defect Detection in Medical Devices Produced by Electron Beam Melting Using Layer-by-Layer...

The implementation of online, nondestructive quality control (QC) solutions is a key aspect to consider especially in medical device manufacturing, where the production process should comply to relevant quality-related standards to guarantee a safe product for the patient. For metallic implants fabricated using electron beam melting (EBM), the presence of defects (e.g., porosities, cracks, delamination, balling effect) could affect the final part quality in terms of structural integrity and mechanical properties. In this context, we propose an expert system algorithm capable of automated detection of porosities that could occur in prosthetic components, such as tibial trays, printed using a cobalt-chromium alloy. The choice of focusing only on the porosities was done as these defects are particularly critical for orthopedic prosthesis, where they can negatively impact the fatigue behavior of the printed component. The algorithm was developed ad hoc to analyze images of the manufacturing process taken from a camera embedded in the printer (Arcam Q10plus). Images can be used to identify porosities, performing a nondestructive evaluation that supports the process of part qualification. The developed algorithm automates and improves the visual inspection task conducted by human experts, including quantitative assessment on the size and location of the porosities, and reporting the presence of high porosity density areas. The defect detection performance was evaluated through the design of two tasks: The layer-wise defect detection and the large pore identification. The defect detection was performed with a sensitivity of 91% and a precision of 76%. Furthermore, a comparison with the gold-standard nondestructive evaluation technique, that is, computed tomography evaluation, allowed to validate the algorithm (percent agreement of 98%). The developed expert system allows to quickly evaluate an entire printing volume with several components representing a reliable and fast tool for defect detection and QC of EBM-printed prosthetic components.