Title: Quadric-Of-Revolution (QUADOR) beams and their applications to lattice structures
School of Interactive Computing
College of Computing
Georgia Institute of Technology
Date: Monday, Nov 25, 2019
Time: 10:00 am EST
Location: TSRB 223
Prof. Jarek Rossignac (Advisor, School of Interactive Computing, Georgia Tech)
Prof. Greg Turk (School of Interactive Computing, Georgia Tech)
Prof. Concettina Guerra (School of Interactive Computing, Georgia Tech)
Dr. Suraj Musuvathy (Siemens Corporate Research)
Additive Manufacturing (AM) or 3D printing presents the potential to fabricate parts with novel mechanical properties by allowing control over the internal microstructure of these parts. Lattices used for designing these microstructures are often defined by 3D arrangements of balls and beams. Attempting to discover optimal structures with desired mechanical properties, engineers explore variations: of the overall shape of these arrangements; of the dimensions and placements of the balls; of the profile shape of the beams and of the lattice connectivity that they form.
Current Solid Modeling technology only provides limited support for such explorations for several reasons: (1) Each beam connects two balls. It is often desired to ensure that the surface of the beam connects with tangential continuity to the surfaces of the two balls that it joins. Furthermore, it is often desired to use beams with curved profiles and not just those bounded by straight circular cone sections. The round-off errors resulting from computing the intersections of such beams, and hence the precise boundary representation of the lattice prohibit doing so reliably when the beams are relatively small compared to the overall size of the lattice. (2) Microstructure lattices have extremely high complexity (possibly billions of beams). Precise models of such structures are not supported by the current generation of commercial modelers. Furthermore, it is computationally prohibitive to compute mass properties of such lattices by iterating over its balls and beams.
In this thesis, we propose to address both of these problems: (1) The challenge of designing and modeling beams with curved profiles and of computing their surface reliably and accurately. (2) The challenge of precisely modeling highly complex lattices with profiled beams and of efficiently computing their mass properties.
To address the first challenge:
- We propose a family of beams, which we call the quador beams, that are bound by a quadric-of-revolution (quador) surface abutting tangentially to the two balls. This family includes cone beams. We propose geometric constructions for quador beams, that have simple mathematical expressions and provides intuitive control of its shape.
- We propose three analytically exact representations, Constructive Solid Geometry (CSG), Constructive Solid Trimming (CST) and Boundary Representation (BRep) of a lattice with quador beams. Each of these representations is more suitable than others in efficiently performing certain geometric queries on a lattice, e.g. CSG for Point Membership Classification (PMC), CST for ray intersection and BRep for minimum distance query. We propose compact data structures to store these three lattice representations, describe efficient methods to compute them and provide associated operators to efficiently query the lattice.
To address the second challenge:
- We propose a class of lattices, which we call the Steady Lattices. A steady lattice consists of a three-directional tensor of cells of balls and beams that each connects two balls in the same cell or two different cells. We discuss rows of balls and beams in each direction. In a steady lattice, all rows in at least one direction are steady, by which we mean that each cell of balls in a row and the beams incident on its balls are related to the previous cell by the same similarity transformation. We use an extremely concise representation of the lattice: The balls and beams of a template cell, three transformations, and 3 repetition counts. Steady lattices are easy to compute and are amenable to fast and robust queries.
- We propose closed-form expressions that exploit steadiness for computing mass properties (e.g., surface area, volume, center-of-mass) of a steady lattice without iterating over all of its elements.
As for further explorations, we are developing algorithms for computing numerically exact Brep of a lattice with quador beams. For a large lattice microstructure, numerical robustness is crucial to ensure that numerical round-offs don't build up to cause topological errors (e.g. edges with zero lengths) in the lattice.