Machine Learning (ML) Technology for a Long Reach, Cobot-Enabled, Multi-Probe Aircraft Nondestructive Inspection (NDI) System

Early detection of the corrosion and related defects is critical, as it reduces remediation cost, improves the operational safety, and minimizes mission downtime of Navy aircraft assets.  The deleterious consequences of fatigue and fracture in aircraft metallic structures arising from local microstructure, mechanical loading, thermal effects, especially in the corrosive maritime environment, usually lead to corrosion/defects of naval aerial platform surfaces (e.g., wings, fuselage, etc.), and aircraft landing gear.  Traditional methods for detecting defects on large aviation assets at depots are inefficient and handheld processes. The main objective of this project as proposed is to complete a feasibility study leading to the development and implementation of new artificial intelligence/machine learning (AI/ML) tools and resources in Phase I/II for use with our existing collaborative robot (cobot) NDI system with its end of arm effectors (i.e., Phased Array Ultrasonic Test (PAUT) scanner, and a multi-topology advanced Eddy Current Array (ECA) probe). The cobot-enabled scanners can rapidly and volumetrically scan these primary surfaces of the airframe.  Even without AI/ML, with just our cobot NDI system, the typical wings, fuselage, and fastener target areas can be scanned up to 30x over all currently used UT and ET methods used in the NAVAIR Fleet Readiness Centers (FRCs) today.  This cobot, multi-modal NDI approach that we developed and integrated is operationally relevant for rapid transition to Level III FRCs.  Developing and adapting a new ML tool set under N244-P01 to be an integral part of the back end NDI data flow collected from these airframe inspections of in-service aircraft will optimize efficiencies and productivity of FRC scheduled inspections.  This project will develop deep learning algorithms specific to NDI defect detection and characterization, data validation, trend analysis, decision making, and future predictive maintenance capabilities.  This project embodies automation/robotics, digitization, especially the optimization of NDI processes/methods in for better performance at the FRC or Maintenance, Repair, and Overhaul (MRO) facilities.   By applying ML algorithms and advanced image recognition tools to the NDI inspection data streams from automated PAUT/ECA scans, this new ML tool will identify minute defects that might be overlooked by the human eye or standard COTS inspection data management system software.  The ML tool can be trained to recognize patterns and anomalies, characterize and/or classify the defect for prioritized review and disposition as a repair tasking.  ML technology is a powerful tool for any quality inspections, especially of in-service aircraft.