Adaptive Multi-Fidelity Framework for Defect and Damage Tolerant Composite Structures

This work presents an adaptive multi-fidelity framework for uncertainty quantification in defect-tolerant composite structures. The approach couples the Carrera Unified Formulation (CUF) for generating consistent low- and high-fidelity structural theories with an adaptive Gaussian Process Regression (GPR) surrogate that learns their correlation. High-fidelity simulations are selectively activated in regions of high uncertainty, minimizing computational cost while preserving accuracy. The framework enables probabilistic estimation of stresses and failure indices with confidence intervals and is validated against 3D Abaqus benchmarks.