The aim of the ASSASSINN project was to develop a multifunctional cell that would optimise both the quality and time of assembly processes. In particular, assembly processes that occur during the construction of an aircraft fuselage. Technology such as artificial intelligence, augmented reality, mixed reality and collaborative robotics have been integrated into the cell.
The methodology for developing the cell consisted of defining use cases that represent assembly processes in the construction of a fuselage. The use cases have been sequentially defined and implemented.
The first case involved the assembly of intercostal parts. First, the operator must uninstall the intercostal part, then apply sealant to the part and, once the intercostal part is ready, reinstall it on the fuselage. First, a module called activity manager was developed, which controls the different components of the cell and the status of the assembly process. The following modules were then developed:
- A mixed reality application to assist the operator. The application shows the operator the assembly instructions and, using holograms, where to install the intercostal part. The operator communicates with the cell via this application.
- An automatic application of sealant. When the operator places the intercostal part on a table, the cell identifies its position using an area scanner and applies the sealant using a robotic arm.
- A quality inspection system. Using a linear scanner guided by the robotic arm, the quality of the sealant that has been applied is analysed and in case of errors, rework is carried out.
In the second use case, an augmented reality system was developed to assist the operator during cable assembly. The application provides the operator with all the information needed to perform each assembly step. In some cases, after the step has been performed, artificial intelligence algorithms are run to identify whether the step has been performed correctly. Mainly, to identify if the correct cable has been inserted in the correct hole of the connector.
The next use case involved the installation of structural parts as brackets on the fuselage. The operator is assisted with a mixed reality application similar to the one developed in the first use case. Once all the parts have been assembled, the operator using a collaborative robot and an area scanner reconstructs all the assembly area. An algorithm checks if all parts have been assembled correctly, if not, it informs the operator of the errors for repair.
Also, a feasibility evaluation to study the use of AI for automatic FOD recognition was carried. Two approaches, NERF (Neural Radiance Fields) and OLN(Object Localizer Network), have been considered to address the issue. Each has a different approach: the first one tries to find the differences between 2 images and the second one tries to learn what the foreign objects are.
After developing each use case, they have been validated at Topic Manager Plant.
Various dissemination tasks have been carried out during the project, mainly its dissemination in social networks and press. As a result of the work carried out, four articles have been published in open access journals and it is expected to publish other results the next months.
