Adaptive-Weighted Coupling-Aware Gaussian Processes for AFP

2026-99-1277

9/4/2026

Authors
Abstract
Content
Composite materials have gained widespread application in the aerospace field due to their advantages, such as high specific strength, high specific modulus, and corrosion resistance. Automated placement technology, as an emerging automated manufacturing method, is gradually replacing traditional manual placement processes and demonstrating significant advantages in composite manufacturing. Currently, the automated placement process for composite materials faces challenges such as insufficient experimental samples and strong coupling relationships between process parameters, leading to low fitting accuracy in process parameter optimization models. To address this, this paper proposes a placement process parameter optimization method based on model weight adaptive allocation. This method integrates three key technologies: a coupling-aware Gaussian process based on combined kernel functions, a weight allocation ensemble model based on leave-one-out cross-validation, and a multi-criteria adaptive sampling mechanism. Experimental validation demonstrates that the integrated model achieves a coefficient of determination R^2 = 0.82, which represents a superior fit compared to the R^2 = 0.65 achieved by a single-kernel Gaussian model and the 0.76 obtained from a single sampling. Furthermore, both the Root Mean Square Error (RMSE=0.92) and Mean Absolute Error (MAE=0.70) are lower than those of traditional baseline models. This framework provides an effective solution for optimizing parameters in the automated placement process for composite materials.
Meta TagsDetails
DOI
https://doi.org/10.4271/2026-99-1277
Citation
Zuo, R., Du, T., and Lv, R., "Adaptive-Weighted Coupling-Aware Gaussian Processes for AFP," 2025 6th International Conference on Applied Mechanics and Mechanical Engineering (ICAMME 2025), Beijing, China, December 12, 2025, https://doi.org/10.4271/2026-99-1277.
Additional Details
Publisher
Published
Yesterday
Product Code
2026-99-1277
Content Type
Technical Paper
Language
English