Uncertainty is intrinsically tied to decision-making in design. Process-Structure-Property (PSP) relations are central to development of new and improved materials. The multitude of PSP linkages for any performance objective can be explored using the top down, inductive design exploration method (IDEM). Each PS and SP linkage has associated uncertainty, arising both from the types of models or interpretation of experimental results used to form linkages, as well as model parameters. These uncertainties can propagate and significantly affect the decision-making process in design and development of materials for specific performance targets. Uncertainty quantification (UQ) can be a highly computationally expensive undertaking in materials design and development. In this research, computationally efficient protocols are developed to effectively incorporate UQ in the IDEM. The uncertainty associated with PS linkages is assigned based on existing literature results. Gaussian process (GP) surrogate models are developed for the various SP linkages of interest as lower order approximations of computational expensive computational materials science simulations (e.g., the crystal plasticity finite element method (CPFEM)). These GP models are used to propagate uncertainty in microstructure attributes to the quantities of interest associated with properties that are then optimized in design. These surrogate models are integrated into existing python IDEM (pyDEM) protocols in the form of mapping functions. In this work, novel protocols are developed and demonstrated for uncertainty-informed design and development of Ti-6Al-4V and Al7075-T6 microstructures for targeted performance requirements involving combinations of fatigue resistance, elastic stiffness, and yield strength.
Cite this work
Researchers should cite this work as follows:
MATIN Development Team