His research team hopes to significantly reduce that downtime with an in-process AFP manufacturing inspection system (IAMIS) made with off-the-shelf laser and camera systems. Attached to the AFP placement head, the IAMIS creates a digital manufacturing twin (DMT) of the part. Using machine-learning algorithms, the system analyzes this digital data. It detects manufacturing effects that are above acceptable limits, reducing time-consuming and operator-dependent manual inspection processes that require interrupting the manufacturing process. Furthermore, it records the locations of the part’s allowable defects (those that do not require repair). Having this record can be useful when the part gets damaged during service, since it provides technicians with a better way to assess the potential damage in a given area.

The system’s AI also analyzes the digital twin to detect any gaps, overlaps and other manufacturing anomalies in the part. Based on this information, system operators could adjust manufacturing processes like laydown speed, heat input and compaction force to reduce manufacturing defects on future runs. This could improve the quality of parts by 10 to 20%, says Seneviratne.

After several successful demonstrations of the IAMIS, researchers will mount the system on production lines of several aircraft manufacturers, including those in the advanced air mobility market, within the next few months. The manufacturers will continue manual inspections at the same time to compare results with those from IAMIS.

“Advanced air mobility companies want to make thousands of airplanes a year but using the AFP machine the way we do today they’re not going to make 1,000,” says Seneviratne. Eliminating the manual inspection process and employing IAMIS – with a fully-trained ML algorithm – could decrease the time required to fabricate a part by 20% or more and reduce the cost by over 30%.

“Eventually, we want to convince the FAA and the certification authorities that the IAMIS system is either equivalent or better than manual inspection,” Seneviratne adds.

Improving the Supply Chain

Many large industrial manufacturers have implemented AI and ML-based supply chain management programs to optimize supplier selection and flag potential problems. To be effective, these systems use data that’s pulled from the organization’s enterprise resource planning (ERP) systems, supplier lists and other sources stored in the cloud.

But getting that data input can be problematic for the many composites manufacturers that have not digitalized their operations. “A lot of their data is still in Excel sheets, in registers, in notebooks somewhere,” says Rajeev Sharma, chief technology officer at Grid Dynamics.