Session: 01-05-01: Foundations of AI and Machine Learning for Aerospace Structural Analysis

Paper Number: 182844

182844 - Reliability-Based Design Optimization of Composite Structures With Multi-Fidelity Surrogates 

The increasing complexity of Automated Fiber Placement (AFP) processes has renewed attention on manufacturing defects and their impact on the design of composite structures. Both deterministic and stochastic defects originate from AFP: deterministic defects, such as tow gaps and overlaps [1], depend on machine parameters and geometry, and can be predicted in terms of location and size; stochastic defects, such as fiber misalignment, resin-rich zones, and local variations in fiber volume fraction, are inherently random and cannot be predicted a priori [2]. These imperfections pose significant challenges for the reliable design of high-performance composite structures.

This work presents a multi-fidelity framework for Reliability-Based Design Optimization (RBDO) of composite structures that explicitly accounts for both deterministic and stochastic manufacturing defects. The framework integrates low- and high-fidelity models into a Gaussian Process Regression (GPR) framework [3]. The Carrera Unified Formulation (CUF) [4] is used to formulate Equivalent Single-Layer (ESL) and Layer-Wise (LW) structural theories, based on Taylor and Lagrange expansions, respectively. ESL models are computationally efficient but unable to fully capture through-thickness shear and out-of-plane stress effects, which are essential for reliable failure predictions. In contrast, LW models provide the required accuracy at a higher computational cost. An adaptive learning strategy identifies the most informative regions of the design space, selectively refining the surrogate with high-fidelity LW simulations only when needed. In this way, the framework learns the correction functions, thereby improving predictions from numerous low-fidelity ESL analyses using a limited number of targeted high-fidelity simulations.

The RBDO formulation minimizes structural mass while satisfying probabilistic constraints on stresses, strain energy, and failure indices. Ongoing developments extend the approach to nonlinear phenomena, such as post-buckling and progressive damage, advancing the development of defect- and damage-tolerant composite structures.

[1] A. Pagani, A. R. Sánchez-Majano, D. Zamani, M. Petrolo, E. Carrera: Fundamental frequency layer-wise optimization of tow-steered composites considering gaps and overlaps. Aerotecnica Missili e Spazio (2025).

[2] A. Pagani, M. Petrolo, A. R. Sánchez-Majano. Stochastic characterization of multiscale material uncertainties on the fibre-matrix interface stress state of composite variable stiffness plates. International Journal of Engineering Science (2023).

[3] D. Zamani, A. Pagani, M. Petrolo, E. Carrera: Multi-fidelity models for failure onset analysis of composites under uncertainties. Proceedings of the ASME 2025 Aerospace Structures, Structural Dynamics, and Materials Conference, Houston (TX, USA), May 5–7, 2025

[4] E. Carrera, M. Cinefra, M. Petrolo, E. Zappino. Finite Element Analysis of Structures through Unified Formulation. Wiley & Sons, Hoboken, New Jersey, USA. 2014.

Presenting Author: Dario Zamani Politecnico di Torino

Presenting Author Biography: Dario Zamani is a PhD student in Aerospace Engineering at Politecnico di Torino (MUL2 Lab). He works on multi-fidelity surrogate modeling with Gaussian Process regression, uncertainty quantification, reliability-based design optimization, and manufacturing-aware design of tow-steered composites with AFP defects.