Abstract

The development of practical engineering models that accurately reflect microstructurally informed phenomena is a critical challenge in materials engineering, particularly for high-performance applications like nuclear fusion. Conventional approaches typically rely on disjointed workflows requiring multiple specialized models and significant manual data transfer. This process is resource-intensive, demanding constant intervention from multidisciplinary experts at each stage of development. This seminar presents progress towards automated workflow management that addresses these inefficiencies. The proposed pipeline integrates probabilistic calibration of physics-based constitutive models, large scale generation of training datasets and utilizes the resulting datasets to train reduced order models. The result is a streamlined, end-to-end development pipeline capable of taking statistical distributions of microstructural features and desired macroscopic responses as inputs to generate robust engineering models with minimal user intervention. To demonstrate the efficacy of this approach, a case study on the generation of a microstructurally informed deformation model for pure copper will be presented. The talk will highlight how this automated framework accelerates model development, reduces human error, and provides a scalable pathway for designing advanced materials for fusion environments.