Digital twin-based probabilistic prediction of microcrack initiation and propagation in the weld zone of orthotropic steel deck

Abstract

Orthotropic steel deck (OSD) has become the primary bridge deck structure for steel bridges, but it is vulnerable to fatigue cracks in the weld zone of OSD under traffic loading. This study proposes a digital twin (DT)-based framework to give the probabilistic prediction of microcrack initiation and propagation in the weld zone of OSD. The DT-based framework couples crystal plasticity finite element model (CPFEM) simulations with fatigue tests through Dynamic Bayesian Network (DBN). The fatigue tests of the specimen (physical entity) cut from the weld zone of OSD and sliced to the centimeter scale are carried out using the standard dynamic test system together with optical microscope. The CPFEM simulations are used to develop a multiscale virtual entity to map the physical entity. The role of the DBN is to integrate the uncertainties into a hierarchical network. After completing the updating of uncertainties in the virtual entity via DBN, the microcrack propagation is predicted probabilistically by the DBN inference, and the microcrack initiation is predicted probabilistically by the inverse DBN inference. The results demonstrate the feasibility and accuracy of the DT-based framework for a full prediction of microcrack initiation and propagation in the weld zone of OSD.