Prediction and experimental validation approach to improve performance of novel hybrid bio-inspired 3D printed lattice structures using artificial neural networks

Abstract

Novel Cellular lattice structures with lightweight designs are gaining more interest in the automobile and aerospace sectors. Additive manufacturing technologies have focused on designing and manufacturing cellular structures in recent years, increasing the versatility of these structures because of the significant benefits like high strength-to-weight ratio. In this research, a novel hybrid type of cellular lattice structure is designed, bio-inspired from the circular patterns seen in the bamboo tree structure and the overlapping patterns found on the dermal layers of fish-like species. The unit lattice cell with varied overlapping areas with a unit cell wall thickness of 0.4 to 0.6 mm. Fusion 360 software models the lattice structures with a constant volume of 40 × 40 × 40 mm. Utilizing the stereolithography (SLA) process and a vat polymerization type three-dimensional printing equipment is used to fabricate the 3D printed specimens. A quasi-static compression test was carried out on all 3D printed specimens, and the energy absorption capacity of each structure was calculated. Machine learning technique like the Artificial neural network (ANN) with Levenberg–Marquardt Algorithm (ANN-LM) was applied to the present research to predict the energy absorption of the lattice structure with parameters such as overlapping area, wall thickness, and size of the unit cell. The k-fold cross-validation technique was applied in the training phase to get the best training results. Overall, the results obtained using the ANN tool are validated and can be a favourable tool for lattice energy prediction with available data.