Optical Quality Control System Improvement for an SME. A Case Study for Test-Before-Invest Partnerhips

The manuscript presents the results of a case study, a test-before-invest experiment, to set up an AI-based optical control system in a manufacturing company in order to improve the throughput and to improve the rate of false negative signals provided by the quality control system. The manufactured parts, small fasteners (rivets) could be critical elements in many industries, the quality requirements are set by special standards. The literature study proved us, that the accuracy of the existing machine-vision algorithms, at least at experimental level is about 98%. Our state of art review led to the conclusion that measurement-error improvement can be achieved by changing the detection technology and downsizing the false negative signals. During the research a secure environment was created in which generative AI solutions were tested, mitigating many security risks, without disturbing the real shopfloor environment and without using the resources of the manufacturing company. On the studied sample on a higher investigation speed, we had no false positive signals, and the number of false negative signals were decreased from 7% to 2%. The research intends to show how collaboration between SMEs and research groups can solve real technological problems using the latest achievements in science and technology. The proposed solution overcomes some of the weaknesses of the existing technology by a single digitalization project.