Title: A Data-Efficient Machine Learning Framework for Uncertainty Quantification of Assembly Gap in Aerospace Composites

Authors: Kendall Johnson, Rowan Preedy, Navid Zobeiry

DOI:

Abstract: Traditional composite analysis and optimization approaches rely on high-fidelity process simulations and trial-and-error experiments which are costly and typically deterministic, ignoring uncertainties of material and process. This work introduces a data-efficient framework that couples Gaussian Process Regression (GPR) with Bayesian optimization (BO) to quantify uncertainties of process-induced deformation (PID) of composites. L-shaped composite parts with various cure cycles were fabricated and experiments were repeated multiple times to investigate uncertainty in PIDs. The results showed that the uncertainty did not follow a traditional normal distribution. In order to address this deviation from Gaussian assumption, we implemented a non-Gaussian machine learning framework by taking advantage of feature transformation prior to GPR training. This framework then analyzes sparse experimental data and proposes subsequent cure cycles that maximizes expected information gain. The resulting model can quickly quantify uncertainties in assembly gap and propose optimization strategies for cure cycles.

References: [1] G. Fernlund, C. Mobuchon, and N. Zobeiry, “2.3 Autoclave Processing,” in Comprehensive Composite Materials II, Elsevier, 2018, pp. 42–62. doi: 10.1016/B978-012-803581-8.09899-4. [2] N. Zobeiry et al., “Multiscale characterization and representation of composite materials during processing,” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 374, no. 2071, p. 20150278, Jul. 2016, doi: 10.1098/rsta.2015.0278. [3] [4] [5] [6] [7] [8] N. Zobeiry and A. Poursartip, “The origins of residual stress and its evaluation in composite materials,” in Structural Integrity and Durability of Advanced Composites, Elsevier, 2015, pp. 43–72. doi: 10.1016/B978-0-08-100137-0.00003-1. T. A. Bogetti and J. W. Gillespie, “Process-Induced Stress and Deformation in ThickSection Thermoset Composite Laminates,” J. Compos. Mater., vol. 26, no. 5, pp. 626660, Mar. 1992, doi: 10.1177/002199839202600502. C. Albert, “Spring-in and warpage of angled composite laminates,” Compos. Sci. Technol., vol. 62, no. 14, pp. 1895–1912, Nov. 2002, doi: 10.1016/S0266-3538(02)001057. K. Manohar, T. Hogan, J. Buttrick, A. G. Banerjee, J. N. Kutz, and S. L. Brunton, “Predicting shim gaps in aircraft assembly with machine learning and sparse sensing,” J. Manuf. Syst., vol. 48, pp. 87–95, Jul. 2018, doi: 10.1016/j.jmsy.2018.01.011. X. Liu, Z. Guan, X. Wang, T. Jiang, K. Geng, and Z. Li, “Study on cure-induced residual stresses and spring-in deformation of L-shaped composite laminates using a simplified constitutive model considering stress relaxation,” Compos. Struct., vol. 272, p. 114203, Sep. 2021, doi: 10.1016/j.compstruct.2021.114203. J. Magnus Svanberg and J. Anders Holmberg, “Prediction of shape distortions Part I. FEimplementation of a path dependent constitutive model,” Compos. Part A Appl. Sci. Manuf., vol. 35, no. 6, pp. 711–721, Jun. 2004, doi: 10.1016/j.compositesa.2004.02.005. [9] N. Zobeiry and A. Poursartip, “Theory-Guided Machine Learning for Process Simulation of Advanced Composites,” Mar. 2021. [10] C. Schoenholz, E. Zappino, M. Petrolo, and N. Zobeiry, “Accurate Prediction of ProcessInduced Deformations in Composites Using Layer-Wise Models and Theory-Guided Probabilistic Machine Learning,” in ASME 2024 Aerospace Structures, Structural Dynamics, and Materials Conference, American Society of Mechanical Engineers, Apr. 2024. doi: 10.1115/SSDM2024-121050. [11] C. Schoenholz and N. Zobeiry, “An Accelerated Process Optimization Method to Minimize Deformations in Composites Using Theory-guided Probabilistic Machine Learning,” Compos. Part A Appl. Sci. Manuf., vol. 176, Jan. 2024, doi: 10.1016/j.compositesa.2023.107842. [12] K. Zhou, R. Enos, D. Zhang, and J. Tang, “Uncertainty analysis of curing-induced dimensional variability of composite structures utilizing physics-guided Gaussian process meta-modeling,” Compos. Struct., vol. 280, Jan. 2022, doi: 10.1016/j.compstruct.2021.114816. [13] “3900 Prepreg System Data Sheet.” [Online]. Available: www.toraycma.com [14] N. Odagiri, H. Kishi, and M. Yamashita, “Development of TORAYCA prepreg P2302 carbon fiber reinforced plastic for aircraft primary structural materials,” Advanced Composite Materials, vol. 5, no. 3, pp. 249–254, Jan. 1996, doi: 10.1163/156855196X00301. [15] G. E. P. Box and D. R. Cox, “An Analysis of Transformations,” 1964. [Online]. Available: https://www.jstor.org/stable/2984418 [16] N. Srinivas, A. Krause, S. M. Kakade, and M. Seeger, “Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design,” Jun. 2010, doi: 10.1109/TIT.2011.2182033.

Conference: SAMPE 2026

Publication Date: 2026/04/27

SKU: 57

Pages: 13

Price: $26.00