Scalable Workflows for Surrogate Modeling of Simulations

Research into the biological effects of electromagnetic radiation has led to the development of high-performance modeling and simulation software. However, the high computational cost associated with these software tools makes them impractical for direct use in mission-level simulations. Fast surrogate models provide a path to reducing computational cost of accessing the results of high-fidelity numerical models. Nanohmics has developed a software framework designed to help automate the construction and evaluation of high-quality, computationally-efficient surrogate models. The framework provides capabilities for the convenient development of re-usable, scalable workflows for training and testing surrogate models of complex, physics-level simulations. A detailed software architecture has been demonstrated through software prototypes of key components and proof-of-concept sample training workflows. During the Phase II program, Nanohmics will develop a production-level implementation of an advanced version of the framework. Phase II software will include a comprehensive collection of reference material, sample workflows, and references models, as well as an informative graphical user interface. Development and analyses of surrogate models for physics-level simulations on the thermal effects of RF and optical exposures on anatomical models will demonstrate the usage, quality, and run-time performance workflows and models.