A digital twin solution for mixed mode I/II fatigue crack assessment under large-scale yielding

This paper introduces a digital twin solution for mixed mode I/II fatigue crack assessment under large-scale yielding. Under these conditions, fatigue load magnitude significantly influences crack trajectories, unlike paths predicted by traditional linear elasticity-based criteria. Furthermore, fatigue crack growth rates under mixed mode loading may differ substantially from those under pure mode I conditions. Therefore, the proposed digital twin addresses these challenges through its diagnosis, prognosis and updating capabilities. The diagnosis component tracks the crack path under a given fatigue load magnitude and crack tip location, while remotely estimating crack size through strain data obtained from strategically placed sensors. The prognosis component computes the cyclic J -integral as the crack driving force using the same inputs for crack path generation. The digital twin finally establishes a high-fidelity connection with its physical counterpart by employing the unscented Kalman filter to update parameters within a fatigue crack growth rate model. This study investigates the performance of the digital twin against three specimens under constant amplitude loading and three specimens under variable amplitude loading. Under constant amplitude loading, the digital twin prognosticates the remaining useful life with the updated parameters. Whereas under variable amplitude loading, the digital twin considers load sequence effects and instead performs a fatigue reliability assessment against potential failure mechanisms. The digital twin performs well across these specimens, (1) ensuring that the observed remaining useful life falls consistently within two standard deviations of the predicted mean and (2) correctly identifying the dominant failure mechanism under both step-up and step-down loading regimes.