Inverse Design of Lattices

Carbon

Mentors: Adrian Lew (Stanford) and Hardik Kabaria (Carbon)

Students: Ruiqi Chen, Zhiheng Zhou (not all students are listed)

Goal:

To use optimization and inverse design methodology to find a lattice structure that, when loaded in compression, has a very flat load-displacement response.

Problem:

Carbon’s continuous liquid interface protection (CLIP) method of 3D printing has opened the door for efficient manufacturing of complex lattice structures. Owing to their high strength to weight ratio and tailorable performance, lattice structures can replace many traditional designs. Given a lattice structure and boundary conditions, it is easy to characterize its mechanical behavior. However, the reverse problem is much more challenging. In this project, we look at developing an inverse design tool that takes in a desired load displacement curve and outputs a lattice structure. The tool should be general enough to handle an arbitrary load displacement curve, but one curve of particular interest for this project is a constant plateau load over a large range of displacement.

What did the team do?

The team developed a Python-based workflow to generate, preprocess, simulate, post process, and optimize lattice structures. Lattices are represented as graph objects using the NetworkX package. Simulations were conducted in LS-Dyna using explicit dynamics. All preprocessing and postprocessing (pre/post) was done in a custom written Python package that interfaced with the ME329 computing cluster, completely removing LS-PrePost (the standard GUI-driven pre/post software) from the picture. Genetic algorithm optimization (GA) and coordinate descent optimization was performed on multiple lattice families. In particular, running GA on the Pratt truss lattice family (inspired by the Pratt truss used in bridgework) yielded a lattice configuration with a relatively flat load-displacement response.