In-line Bayesian-optimization-driven shape adjustment for large composite fuselage panels considering individual variability

Aircraft manufacturers have gradually utilized composite panels to replace the aluminum alloy panels in aircraft fuselage manufacturing for structural weight reduction. However, the composite panels are prone to deform due to their large scale and have lower shape accuracy than the metal ones. Moreover, the uniqueness induced by the autoclave curing process and other random variations leads to individual variability of each composite panel, which makes it challenging to adjust its shape during the fuselage assembly. This paper proposes an in-line, Bayesian-optimization-driven shape adjustment method. In-process feedback information is adopted to update the shape adjustment action more applicable to an individual incoming composite panel, which requires less sampling data to characterize the deformation behavior of the panel. Physical experiments of shape adjustment based on a 6 m × 4 m composite panel are conducted to demonstrate the feasibility and effectiveness of the proposed methodology. The results show that the composite panel can achieve higher shape accuracy after in-line adjustment, and the dimensional deviation residual sum of squares can be further reduced by up to 78.77 % compared to that without in-process feedback information; all control points deviations remain < 0.2 mm even after fixture relocation, demonstrating robust self-adaptation to panel-to-panel variability. The proposed method therefore delivers an in-line, data-efficient and panel-adaptive solution for shape adjustment of large composite fuselage panels.