A method to adapt natural structures for use in engineering and design applications.

Due to evolutionary pressures, structural systems in nature are optimized for their specific function and form. However, these specific constraints create difficulties for humans in interpreting and understanding the underlying design rules. For example, the dragonfly wing is a high-performing and lightweight structure for flight, yet limited dragonfly wing-inspired engineering applications exist due to the lack of understanding of its latent design rules. A method for uncovering these design rules would give engineers and designers access to more powerful and efficient structural designs.

The inventors developed a machine learning algorithm that generates and optimizes the structural form within a human-defined boundary using rules learned from a natural structure. Using the dragonfly wing, the machine learning algorithm can generate several airplane wing structures of varying shapes.

Using the graphics statics method, the inventors train a machine learning algorithm to extract the force diagram of a dragonfly wing from its form boundaries. A second artificial neural network then maps the topological information from the force diagram onto the geometric edges present in the real form. The combined method accurately recreates the dragonfly wing and generalizes the design to scalable boundary forms, such as full size airplane wings. 

• Maps natural structures to scalable geometric boundaries
• Optimized structures reduce material usage