Remaining Challenges in the Application of High-Performance Computing to a Process-to- Performance Pipeline for Textile Composites

Abstract

Predicting the performance of a textile composite based on a description of the mesostructure and parameters related to the processing that the textile experiences would provide many advantages. Multiple groups have made significant progress towards robust process-to-performance predictions and developed a pipeline of computational tools to perform the analyses. As more complex components and materials are considered, high-performance computing systems have been used to meet the increasing computational costs, but most tools do not effectively leverage modern HPC systems. This paper assesses a particular process-to-performance pipeline by measuring the computational cost at every step for increasingly large problem sizes. Critical limitations of each tool are discussed, and a perspective on how the pipeline may be improved to efficiently use HPC resources to perform more complex analyses is provided.