Abstract
Qualitative analysis of surface defects on aluminium High-Pressure Die-Casting components is relevant for both quality assurance and process monitoring purposes. Besides part functionality and durability, the outward appearance of a die-cast component can be of paramount importance during incoming goods inspection by the customers in order to ensure parts' functionality. Especially when it comes to inspections of parts' surfaces, the use of Artificial Intelligence is gaining traction for identifying and classifying defects. The present paper illustrates a case study on surface defect detection of aluminium die-cast components where a commercial Deep Learning system has proved to reach a 90% effectiveness in recognizing compliant and uncompliant parts. The development of the presented experimental application is intended to pursue the objective of using automated systems for defect detection in-line, which represents an original goal of the present paper. The development of the technical system used in this application has benefitted from the knowledge of TRIZ beyond the understanding of optical principles overlooked in a first-attempt design.