ABSTRACT
Fatigue fracture is one of the main factors that affects structural failures in aircrafts. However, various uncertainties in the fatigue crack growth make it difficult to perform a reliable damage diagnosis and prognosis. In this chapter, by leveraging two parts of the work, a digital twin–driven framework is established to achieve this goal. First, we developed a reduced-order fatigue crack growth prediction method, utilizing the capability of the SGBEM super element–FEM coupling method for fast database generation and the machine learning method for its nonlinear fitting capability. The reduced-order model, with load inputs and fatigue growth laws, enables the fast crack growth prediction for multi-queries in the digital twin. Then, a dynamic Bayesian network–based digital twin, considering multiple uncertainties in crack growth, is constructed to conduct the damage diagnosis and prognosis for complex aircraft structures. In an example of a round-robin helicopter component, the uncertainty in the digital twin is reduced significantly, and the evolution of structural damage can be well predicted. The proposed method, by utilizing the capability of the digital twin, would be helpful to achieve condition-based maintenance.
