Bayesian Data Assimilation-Integrated Phase-Field Modeling of Widmanstätten Ferrite Formation
Please login to view abstract download link
For the development of steels which have both excellent strength and ductility, it is essential to understand and predict microstructural evolution associated with the austenite-to-ferrite (γ → α) transformation. Among the various microstructures formed during γ → α transformation, lath- or plate-shaped Widmanstätten ferrite (WF) forms and significantly influences the toughness of steels. Since WF forms in an intermediate temperature range where both diffusional and displacive transformation mechanisms simultaneously operate, the mechanism of WF formation is highly complex and not yet fully understood. Although numerous experimental observations of WF formation have been reported, it remains difficult to clarify its formation mechanism based solely on the accumulation of experimental data. Therefore, numerical simulation is needed to understand the WF formation mechanism and predict the microstructural evolution during WF formation. Recently, phase-field (PF) modeling has been employed to simulate microstructural evolution during WF formation. The first PF model which can simulate WF formation has been proposed by Loginova et al.. Nevertheless, accurate microstructure prediction requires the identification of parameters related to the anisotropy of interfacial properties, including their temperature dependence. However, the parameters have not yet been identified, and no existing studies have successfully established such parameter estimation frameworks, mainly because quantitative morphological information on WF has been limited. In this study, we develop a numerical framework for estimating the parameters from the available experimental data by combining the PF simulation with Bayesian data assimilation. The application of data assimilation enables us to estimate the parameters while explicitly accounting for uncertainties in both the simulations and experimental observations. The developed framework will contribute to improving the predictive accuracy of WF formation and to establishing a more quantitative understanding of its formation mechanism.
