Dataset for "Self-Supporting Shell Lattices: Explicit Design Method and Neural Accelerated Evolutionary Optimization" (Virtual and Physical Prototyping)

Shell lattice metamaterials possess unique topological and physical characteristics that demonstrate multifunctional application potential across various important fields. However, their engineering adoption has been limited by manufacturing processes, particularly the need for supports in additive manufacturing of overhanging regions. This study presents an explicit design method for shell lattices based on B-spline curves, which enables parametric control over the surface topology and achieves self-supporting characteristics at the unit cell level. The effectiveness of the design method was validated via manufacturing using both vat photopolymerization and laser powder bed fusion processes. A pixel-based representation method was developed to characterize the topological features of variable-thickness shell lattice structures. Additionally, a neural network accelerated evolutionary optimization method was proposed, which enhances the computational efficiency without increasing the data burdens. This method was successfully applied to improve the elastic properties of self-supporting shell lattices. Numerical simulations and experimental results demonstrated that the optimized structures exhibited nearly uniform stiffness across all loading orientations, achieving increases of up to 41.79% in uniaxial elastic moduli and 79.3% in shear moduli after optimization. The proposed optimization framework effectively mitigates the data dependency inherent in traditional machine-learning aided genetic algorithms, demonstrating strong potential for complex, high-dimensional optimization tasks.