Multi-Fidelity Reliability-Based Design Optimization of Advanced Composite Structures

This paper presents a Multi-Fidelity Reliability-Based Design Optimization (MF-RBDO) strategy for Automated Fibre Placement (AFP) tow-steered composite structures. The framework combines manufacturing-aware fibre-path design, uncertainty propagation and multi-fidelity structural theories based on the Carrera Unified Formulation (CUF). The composite optimization framework is posed as a mixedinteger problem, where the number of plies is discrete and the tow-steering parameters are continuous. Deterministic AFP constraints on curvature, gaps and overlaps are enforced before the structural analysis, while stochastic variability in material properties, load, thickness and fibre angles is propagated through a multi-fidelity Gaussian Process Regression (GPR). Low-fidelity Equivalent-Single-Layer models provide inexpensive exploration, whereas high-fidelity Layerwise descriptions enrich the surrogate in regions where through-thickness stresses and failure indices are critical. A benchmark tow-steered plate shows that the reliability-based solutions require only a modest mass increase, but reduce the predicted failure indices and make the design less sensitive to manufacturing-signature variability.