Physics-Informed, Data-Driven Model for Atmospheric Corrosion of Carbon Steel Using Bayesian Network

Area of Science:

• Materials Science
• Aerospace Engineering
• Corrosion Science

Background:

• Atmospheric corrosion significantly impacts aircraft structural integrity over time.
• Accurate corrosion deterioration models are crucial for the aviation industry.
• Limited data for parameter estimation leads to high prediction uncertainty in existing models.

Purpose of the Study:

• To develop a more accurate and applicable corrosion deterioration model for aircraft.
• To address the challenge of limited data in practical corrosion modeling.
• To integrate physics-based simulations with real-world monitoring data.

Main Methods:

• A physics-based atmospheric corrosion model was developed using simulation methods.
• A Bayesian network (BN) was constructed using GeNie software.
• Model calibration was performed using monitoring data from aircraft parking areas, incorporating actual environmental data.

Main Results:

• The integrated physics-based model and BN approach enhances corrosion prediction accuracy.
• Calibration with real-world environmental data improves model applicability.
• The method effectively addresses challenges posed by limited data for parameter estimation.

Conclusions:

• Precise corrosion models are vital for predicting and managing atmospheric corrosion on carbon steel in aircraft.
• The developed Bayesian network approach offers improved accuracy and real-world applicability.
• Future research should explore incorporating additional data sources to further refine corrosion models.