Multifidelity Active Learning for the Fast Identification of Incipient Damages in Aerospace Structures

Abstract

We propose MF-FREEDOM, a multifidelity active learning framework for efficient and accurate structural health monitoring of complex aerospace systems. The method combines (i) a two-stage informative signal compression that reduces data dimensional- ity while preserving critical features, and (ii) a multifidelity inference scheme that inte- grates high- and low-fidelity physics-based models to optimize diagnostic accuracy and computational efficiency. We apply MF-FREEDOM to detect incipient fractures in the carbon fiber skin of an aircraft wing achieving accurate damage detection performance with contained computational resources. By enhancing the efficiency and reliability of structural health monitoring, MF-FREEDOM supports the deployment of lightweight and maintainable structures playing a key role in advancing sustainable aviation.