Robotic Automated Fiber Placement (RAFP) has revolutionized manufacturing of large aerostructures in recent years. From producing lower costing parts to reducing scrap, RAFP has also helped in manufacturing large parts with contours precisely, with repeatable production once the RAFP machine is programmed. RAFP manufacturing allows a designer to utilize curvilinear fiber paths specifically designed to optimize structural performance, for example, maximizing buckling loads, minimizing stress concentrations etc. At the same time, RAFP introduces unintended manufacturing imperfections. These, which can include overlaps and gaps, lead to thickness non-homogeneities in structural panels. In this paper, steered fiber paths for maximizing the buckling loads of RAFP panels while explicitly accounting for manufacturing imperfections is considered. A novel method to model fiber paths is introduced where each course of the lay-up can be explicitly modeled, including manufacturing artifacts of course width, admissible tow curvature, and control on unintended manufacturing deviations that can lead to gaps and overlaps. A machine learning based surrogate modeling technique in conjunction with Genetic Algorithm is used to derive the optimal solutions.

近年来，机器人自动纤维铺放（RAFP）彻底改变了大型航空结构的制造。从生产成本较低的零件到减少废品，RAFP还帮助精确制造具有轮廓的大型零件，一旦对RAFP机器进行编程，便可以重复生产。RAFP制造允许设计人员利用专门设计的曲线光纤路径来优化结构性能，例如，最大化屈曲载荷，最小化应力集中等。同时，RAFP引入了意想不到的制造缺陷。这些可能包括重叠和间隙，导致结构面板的厚度不均匀。在本文中，考虑了转向光纤路径以最大化RAFP面板的屈曲载荷，同时明确考虑制造缺陷。引入了一种对光纤路径进行建模的新颖方法，在该方法中，可以显式地对每个叠层的路线进行建模，包括路线宽度，允许的丝束曲率的制造假象，以及对可能导致间隙和重叠的意外制造偏差的控制。基于机器学习的替代建模技术与遗传算法相结合，可以得出最优解。