SORS: scalable uncertainty quantification for progressive damage models in composite structures

We present a scalable methodology for uncertainty quantification (UQ) in high-dimensional problems involving advanced progressive damage models for composite materials. The approach combines bootstrapping and Bayesian UQ to robustly characterize the distribution of structural Quantities of Interest (QoIs), such as load-displacement curves and failure loads. Applied to a realistic 40-dimensional case study—Open Hole Tension and Compression specimens with variability in material properties, geometry, and ply orientations—the method delivers consistent and efficient predictions of QoI distributions. This approach is highly suited for industrial applications, and its computational scalability makes it ideal for leveraging supercomputing resources to handle complex models and large ensembles, ultimately accelerating the design and certification of next-generation composite structures.