Smart Monitoring and Automated Real-Time Visual Inspection of Sealant Application (SMART VIStA)

As a space grade solar panel manufacture, Boeing needs to assemble solar panel by gluing multiple solar cells together. A glue deposition robot deposits the glue dots at predefined positions. In current manual operation, skilled operators monitor each dot after deposition to confirm that the dot is deposited accurately in specified position with specified shape quality – no tails or connections between dots. Early detection of gradual degradation in dot quality helps an operator to modulate several process parameters in deposition module or to take corrective actions to prevent deposition of overlapping glue dots on the solar panel. This dot-to-dot high precision inspection process is a very tedious, repetitive job and could be replaced by automated visual inspection system, as has been the case in all other industries (such as automotive) to retain US employment.

The team is developing a quality controlled visual inspection system called SMART-VIStA, which combines real-time inspection with AI learning to prevent faulty dots. After a robot deposits an adhesive dot, the first feedback loop uses one or more cameras mounted on a second robotic arm to inspect the dot from another angle. A second feedback loop employs machine learning to make up-stream adjustments and eliminate errors before the next dot is applied. Finally, a digital twin of the equipment assesses the inspection data and entire Digital Thread, and gives operators real-time insights and recommendations to improve performance and quality.

The technology developed within this project could be deployed to assist with sealant application and a range of Boeing’s own commercial and defense programs where the need to automate the application of sealant has been identified as a goal for all programs within 10 years. The solution developed will be integrated with the existing automated glue disposal system of Boeing, allowing stakeholders to improve the overall process. This technology replaces 14 human workers currently needed in this operation with an automated system defined and designed by the Boeing Company in its Solar Panel Manufacturing company. The solution not only addresses the quality control challenges for Boeing, but it also enables the integration of AI and Vision modules in several other real-time quality inspection applications throughout the US manufacturing base, such as automated welding, painting, and soldering.

The objective of the project is to develop a close-loop AI inspection system integrated with Digital Thread to provide timely, actionable insights about the quality of the current adhesive dots. The specific objectives are described below:

• Vision-based Inspection with Limited Data: Develop advanced deep neural networks to perform visual inspection of dot anomalies that is trainable on coarsely labeled limited defect data. This flexibility helps to overcome the challenges of limited defect data in manufacturing scenarios.
• Close Skills Gap: Provide the operators with real-time feedback about the current glue dot quality, type of degradation, highlighted location of degradation on Digital Twin of the solar panel, possible reasoning for the degradation, and recommendations for change in process parameters or corrective action to be taken before the next dot deposition. Thus, it helps in closing the skill gap, reducing operator variability, increasing trust on AI-system.
• Digital Thread Benefits: Create Digital Twins of solar panels to map real-time glue quality inspection results within possible defect sources, enabling downstream decisions through Digital Thread integration.