Data-Driven Calibration of Multifidelity Multiscale Fracture Models Via Latent Map Gaussian Process

Abstract Fracture modeling of metallic alloys with microscopic pores relies on multiscale damage simulations which typically ignore the manufacturing-induced spatial variabilities in porosity. This simplification is made because prohibitive computational expenses explicitly spatially varying microstructures a macroscopic part. To address this challenge and open doors for fracture-aware design materials, we propose data-driven framework that integrates mechanistic reduced-order model (ROM) calibration scheme based random processes. Our ROM drastically accelerates direct numerical (DNS) by using stabilized algorithm systematically reducing degrees freedom via clustering. Since clustering affects local strain fields hence fracture response, calibrate constructing multifidelity process latent map Gaussian processes (LMGPs). In particular, use LMGPs to parameters an as function microstructure (i.e., fidelity) level such faithfully surrogates DNS. We demonstrate application our predicting behavior component results indicate microstructural porosity can significantly affect performance macro-components must be considered process.