M.Sc Thesis – Gil Elbaz- part 1

Genetic Algorithms for Reconstructing Complex 3D CAD Models
using Primitive Shapes and Boolean Connectors

Background and Motivation:
With the advance of additive manufacturing new possibilities of structural complexity can be produced. In order to take full advantage of these developments, new tools for computer aided mechanical design should incorporate problem-specific substructures that would replace the solid bodies or generic substructures found in todays 3D printed structures.
Mechanical bodies with inner substructures that are optimized by geometrical constraints and physical loads would open the field of mechanical design to a new more broad spectrum of light weight, strong and tough structural designs. The hardware to produce these optimized structures is available, yet the software tools at the hands of mechanical engineers is not advanced enough. This next step in CAD design has the potential to transform 3D printers from useful, for quick prototyping of simple objects, to crucial, for production of structurally optimized models. Currently in the industry there are optimization tools, yet there is no way to rebuild the optimized boundary into the CAD software in a way that allows the engineer to edit the volumetric model. We can close this gap and create a powerful tool for mechanical engineers.

An important research direction that will bring this goal to reality, is to create an algorithmic model and implementation that can automatically build geometric models by structuring boolean connectors (AND, OR, NOT..) and primitive shapes (spheres, cubes, cylinders..). This can be done by looking at the basic shapes/connector nodes as genes and by using multiple Genetic Optimization Algorithms.

This research will directly continue previous work done for my undergraduate degree final project. In the project I created a method for analysis of complex porous structures, an important first step.

This project could potentially help develope a powerful tool that will help Mechanical Engineers design light weight structures that boast high structural integrity.
• Objective 1 – 3D Remodeling Method – Develope a method for reconstruction of 3D structures using primitive shapes and connectors.
• Objective 2 – Substructure Optimization Method – Develope substructure optimization tool based on Genetic Algorithms utilizing the 3D Remodeling Method.
This two stage approach to the problem will provide a thorough algorithmic understanding and a solid base for building a powerful structural optimization tool for mechanical engineers.

 Applying Hough Transform method on an imageApplying Hough Transform method on an image:
a) Example of Silica Microspheres; b) Input Synthetic Porous Structure;
c) Result of Hough Transform Analysis of the Porous Structure


Applying genetic algorithm method on an image

Applying GA method on an image:

a) Original input Micro-CT Bone Image; b) Binary Filtered Image;
c) Genetic Algorithm Curve Fitting which is applied on a real image.