Probabilistic failure prediction for early-stage design of novel composite wing structures via heteroscedastic Bayesian learning
File(s)
Published version
Author(s)
Li, Runze
Miranda, Mario
Pinho, Silvestre
Type
Journal Article
Abstract
Composite wing structures increasingly rely on advanced materials and novel configurations to meet future aviation performance and sustainability targets. However, incorporating failure constraints in early-stage design remains challenging, particularly for unconventional wing architectures beyond the traditional design space. This is mainly due to the high dimensionality of composite layups, which greatly increases the need for training data and computational resources in predictive models and optimisation design. This paper proposes a probabilistic framework for rapid failure envelope prediction of composite wing structural elements under varying geometries and loading conditions. Instead of explicitly modelling high-dimensional stacking sequences, the composite layup is treated as an unobserved stochastic variable, and the failure response is formulated as a conditional probability distribution over the population of engineering-rule-compliant layups. A heteroscedastic variational Bayesian model is developed to capture both the mean response and input-dependent uncertainty induced by layup variability. The proposed approach directly predicts probabilistic best-case and worst-case failure envelopes without requiring explicit layup optimisation, enabling efficient integration of failure constraints into early-stage design. The framework is demonstrated on a Z-beam wingbox concept, with failure indices evaluated using the LaRC05 criteria and a high-fidelity dataset. Results show that the model provides accurate and efficient predictions of the probability distributions of failure indices for different wingbox geometries under varying load conditions. The predicted envelopes offer valuable insights for structural assessment and highlight the need for further layup optimisation.
Date Issued
2027-01-01
Date Acceptance
2026-09-15
Citation
Aerospace Science and Technology, 2027, 180 (Part 3)
ISSN
1270-9638
Publisher
Elsevier
Journal / Book Title
Aerospace Science and Technology
Volume
180
Issue
Part 3
Copyright Statement
© 2026 The Author(s). Published by Elsevier Masson SAS. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
License URL
Identifier
10.1016/j.ast.2026.113816
Publication Status
Published
Article Number
113816
Date Publish Online
2026-09-17