Measurement and Prediction of Aerospace Corrosion Rate Using Real-Time Sensor Measurements and Machine Learning Approaches

Corrosion represents a significant cost driver for the Department of Defense, with an estimated total annual cost of corrosion for aviation and missiles of $10.2 billion, or 24.3% of maintenance expenditures. To more effectively combat corrosion, accurate diagnostics and prognostics for cumulative corrosion damage based on corrosivity and environmental conditions are needed. Current schedule-based maintenance practices may be unnecessarily conservative, leading to reduced availability and increased demand on maintenance personnel. Schedule-based practices may also miss unexpectedly aggressive conditions that result in the need for more significant maintenance and repairs. Using on-asset environment and corrosivity monitoring systems, maintenance and sustainment of aircraft can be optimized to decrease costs and increase aircraft availability. The overall goal of this work was to verify the relationships between time-dependent environmental parameters and measurements of instantaneous corrosion rates. These relationships would form the basis for diagnostic and prognostic algorithms that could be used to establish condition-based maintenance practices. Machine learning methods were used to model environmental parameters and corrosion rate data from cyclic laboratory corrosion tests. Real-time measurements of temperature, relative humidity (RH), contaminants (salt loading), free corrosion, and galvanic corrosion were collected using simplified environmental cycles with two different salt loading conditions. The laboratory corrosion test cycles were designed to simulate atmospheric conditions at a specific base location. The resulting environmental data sets were used to build regression models to predict aluminum free corrosion and aluminum/stainless steel galvanic corrosion rates obtained from standard electrochemical sensors (ANSI/NACE TM0416-2016-SG). These machine learning techniques produced excellent (R² > 0.9) predictions for both aluminum free corrosion and galvanic corrosion using the environmental parameters. Continuous measurements of environmental conditions and corrosivity may be useful for investigating the relationships between environmental parameters, contaminant accumulation, and corrosion rate, and machine learning methods may enable diagnostic and prognostic models required for condition-based maintenance.