Automatic FPI

Prosjektet har nå ved sin slutt bygget en celle for automatisk penetrant inspeksjon av metalliske komponenter. Arbeidet ble rammet hardt av pandemien, og sluttføring av arbeidet ble vanskelig. Læringen av AI er ikke fullført og dermed har ikke teknologien fått vist seg levedyktig ennå. Men prosjektet har fullført alt som leder frem til valideringen. Det ser nå ut til at IoT har grobunn for et nytt produkt, som nå er i langtidstesting i produksjonen hos GKN. Det er i tillegg en vilje blant partnerne om å se på mulighetene til å starte opp en ny bedrift som skal utvikle hoved teknologien videre. Teknologien er patentsøkt. Dessverre stopper prosjektet for GKN tidlig, men det er ikke rom for videre drift i en kriserammet bedrift.

Partners IoT and Masmec are looking into the possibilities to bring the main result, automatic PT process, to commercialisation. In addition IoT are also looking at bringing a spin off to the market as well, the GUI now named Fama3D. The latter can become a product on the market quite quickly, where GAN is a test subject for implementation in the production environment. Several new positions will be generated if a new company is established in order to finalize the AutoFPI technology. The technology will with time also revolutionize the PT proceess. Fama3D will most likely end up as a product for IoT labs, and increase their turnover. Depending on the success rate, it may also cause new job positions at IoT. The Fama3D is a product that will increase the data capture and quality from production enabling analysis in a new way for a manufacturer. The existing functionality can be increased from only PT to other processes as well.

Fluorescent Penetrant Inspection (FPI) is a nondestructive test method which is very effective in detecting porosity, cracks, fractures, laps, seams and other flaws that are open to the surface of the test piece and may be caused by fatigue, impact, quenching, machining, grinding, forging, bursts, shrinkage or overload. FPI is currently used at GKN Aerospace Norway (GAN) on lots of machined parts, in particular for stator vanes, and involves a several stages process where the part is cleaned and then dipped in a fluorescent penetrant bath. Then, the penetrant is removed and a developer (dry powder) is sprayed for pulling the trapped penetrant material out of surface defects and spread it out on the surface of the part so it can be seen by an inspector. After the inspection, the part is cleaned. Some of these operations have been partially automated during the years at GAN, and today the NDI workshop is equipped with a crane for loading parts and dipping them into the penetrant tubs; however, the last part of the process is still purely visual and relies on the skills of the operator. Due to the very high number of vanes produced at GAN every week (some thousands), the human factor becomes significative in this process, since inspectors cannot ensure long-term consistency in their evaluations due to fatigue, needing breaks and rest periods to ensure high quality output which are required when inspecting jet engines parts. Besides that, inconsistency is also observed between the outputs of different inspectors. These factors highly contribute to increase the cost of the process and to undermine its reliability. The project proposes to develop and test a vision-based system for automating the FPI process, including artificial intelligence based algorithms developed for training the automatic inspection system and confirm its evaluation by comparing them with the operators' output