The United States Air Force (USAF) spends roughly $5 billion annually on the prevention, inspection, and repair of corrosion on aircraft and ground support equipment. These corrosion maintenance (Mx) actions are expensive, laborious, environmentally detrimental, and hazardous to personnel health. Thus, the USAF is engaged in frequent corrosion preventative asset washes and inspections, which have been shown to be effective, but adversely impact cost and aircraft availability metrics. Therefore, the USAF Corrosion Prevention and Control Office (CPCO) initiated this work to develop a model based corrosion monitoring system capable of predicting corrosion Mx needs for the elimination unnecessary scheduled Mx.

AFCPCO leveraged a body of work sponsored by the Department of Defense that was centered upon the development and deployment of low-powered corrosivity monitoring systems. The Luna Sensor Suite for Aircraft Corrosion Monitoring (LS2A) commercial corrosion monitoring systems developed by Luna Labs were chosen as the measurement medium for the acquisition of dynamic environmental and corrosion relevant measurements for the development of a Cumulative Exposure Algorithm (CEA) to predict corrosion damage of an asset based upon it’s environmental exposure. The output of the CEA was to be used to inform a Condition Based Maintenance Plus (CBM+) algorithm to guide corrosion Mx action based upon need. To provide training data for the CEA, test articles were fabricated using common aerospace materials and coatings. These test articles were placed alongside LS2A sensors at 10 global test locations on outdoor exposure racks and subject to varied wash intervals prescribed in TO 1-1-691. Test articles were collected at specified intervals and the level of corrosion damage measured. These measurements provided ‘ground truth’ mass loss data to correlate to LS2A measurements for development of the CEA.

Though a CEA was developed, correlations to mass loss could not be made. This limited the CEA’s ability to inform a potential CBM+ algorithm. Due to the fact it was not possible to develop a CBM+ algorithm, it was not possible to recommend any changes to corrosion Mx. Consequently, concrete cost adjustment could not be performed.

Though various machine-learned models were developed to inform corrosion Mx scheduling, limitations in the study design and LS2A measurement capabilities repressed model accuracy and global performance below necessary confidence requirements.