Doctor of Engineering (DEng)
In load-bearing lightweight architectures, cellular materials were frequently utilized. While octahedron, tetrahedron, and octet truss lattice truss were built for lightweight architectures with stretching and flexural dominance, it can be believed that new cells could easily be designed that might perform much better than the present ones in terms of mechanical and architectural characteristics. Machine learning-based structure scouting and design improvisation for better mechanical performance is a growing field of study. Additionally, biomimicry—the science of imitating nature’s elements—offers people a wealth of resources from which to draw motivation as they work to create a better quality of life.
Here, utilizing machine learning approaches, novel lattice truss unit cellular architectures with enhanced architectural characteristics were designed. An inverse design methodology employing generative adversarial networks is suggested to investigate and improvise the lightweight lattice truss unit cellular architectures. The proposed framework was utilized to identify various lattice truss unit cellular architectures with load carrying capacities 40–120% greater than those of octet unit cells. A further 130–160% raise in buckling load bearing capacity was made possible by substituting porous biomimicry columns for the solid trusses in the light-weight lattice truss unit cellular architectures.
This dissertation's main goal is to investigate various improvisation strategies for creating lightweight architectures, particularly when using data analysis and machine learning methods. Lightweight cellular architectures with thin-walls and lattice truss unit cellular architectures with improved shape memory capabilities were created using the knowledge gleaned from numerous of the research projects mentioned in the preceding paragraphs load-bearing architectures and devices, lightweight architecture with shape memory and with better strength, better stretchability, and better elastic stress recovery are widely desired. As compared to the bulk shape memory polymeric cylinders, the cellular architectures with thin walls show 200% betterer elastic stress recovery that is normalized with respect to base designs. The architectural improvisation of many other additional designs and practical implementation can be accomplished using the inverse design framework.
Challapalli, Adithya, "Light-Weight Structural Optimization Through Biomimicry, Machine Learning, and Inverse Design" (2022). LSU Doctoral Dissertations. 6030.
Dr. Guoqiang Li