Structural damage diagnosis and prognosis with fleet digital twin considering similarity of individual structural features

Diagnosis and prognosis of the structural health state based on online monitoring data is crucial for enabling condition-based maintenance and ensuring the safety of aeronautical structures. However, most existing studies focus on structural damage diagnosis and prognosis at the individual level, often overlooking the potential of utilizing fleet-wide data, which requires accurately measuring the similarity between structures and the correlation of damage states across individuals in the fleet. To address this, we propose a novel method for fleet-level structural damage diagnosis and prognosis that leverages the similarity of individual structural features. The method introduces a Physics-Decoded Variational Neural Network, enabling accurate extraction of structural features as well as quantifying damage. Additionally, a copula function is used to model the joint probability distribution of damage states across different structures, based on structural feature similarity metrics. This approach allows for collaborative updating of damage states across the fleet using observations from individual structures during the diagnosis process. Validation on a typical damaged aeronautical panel demonstrates that the proposed method achieves more accurate diagnosis and prognosis of individual structural damage states within a fleet, while reducing uncertainties during service compared to conventional individual-based approaches. This method shows promise for integration into a fleet-level airframe digital twin framework, advancing the implementation of condition-based maintenance across fleets.