Dragonfly-like wing structure enabled by a novel skeleton-reinforced neural style transfer assisted topology optimization and additive manufacturing

Natural flyers, such as dragonflies, serve as excellent models for obtaining wing structures with superior performance due to their excellent mechanical characteristics, motivating the design of bionic structures with similar features. Therefore, this paper proposed a novel skeleton-reinforced neural style transfer assisted topology optimization (SNST-TO) method that integrates density-based topology optimization with a convolutional neural network to impose stylistic feature constraints, while rigorously controlling the minimum length scale using a structural skeleton. The core of this method is the incorporation of geometric skeleton information into the topology optimization process, which prevents unmanufacturable structural features by relying on geometric knowledge rather than solely on pixel similarity. The influence of the key parameters in the algorithm were deeply studied through a series of numerical examples, and the effectiveness and the robustness of the SNST-TO method were completely proved. Furthermore, the dragonfly-like wing structures were designed using the proposed SNST-TO method and commercial software ABAQUS under uniform boundary conditions for clear comparison. Especially, these designs were fabricated using fused deposition modeling additive manufacturing technology and tested through compression experiments in both spanwise and chordwise directions. Results show that the bionic dragonfly wing structure designed using the proposed algorithm outperforms the ABAQUS-optimized structure in mechanical performance, with enhanced spanwise and chordwise load capacities. The findings show that the SNST-TO method facilitates the design of lightweight, load-bearing dragonfly-like wing structures, with potential applications in creating biomimetic structures for other organisms.