Packaging and Layout Optimization Using the Dynamic Vector Fields Method

Abstract

Ever-increasing environmental regulations and competition within industry to create the best performing, most cost-effective product has driven the development of computational design, analysis and optimization tools. Many tools exist which allow engineers to optimize specific aspects of their designs, but there are few optimization tools that can inform system level design. The complexity of system level design is a product of the many interactions, functional requirements and the layout of constituent parts and assemblies. System layout and packaging can have significant effect on the overall system performance, affecting metrics such as cost, and overall efficiency. Due to the complexity and scope of system design, there is significant benefit that could be realized from an optimization tool informing the packaging and layout of components and assemblies within a system. This thesis presents a novel method for solving packing, packaging and layout optimization problems. This method, the Dynamic Vector Fields method, solves packing and packaging optimization problems by dynamically simulating the translation and rotation of components and assemblies over time. The final layout of objects/components is determined by dynamic vector fields, which impart acceleration onto objects. These dynamic vector fields represent the problem’s objectives, constraints, and different physical interactions between objects and their environment. Using the Dynamic Vector Fields method presents several advantages; multi-objective problems can be solved by using dynamic vector fields representing different objectives, various physical effects can be represented using dynamic vector fields, relaxation of object collisions and overlap can help reduce initial condition dependence, and the scalability of the method allows it to be applied in various contexts. The development of the theory, mathematics and algorithms behind the Dynamic Vector Fields method is detailed in this thesis. Capabilities to pack non-convex geometry, and pack objects into non-convex design domains are developed, allowing practical problems to be solved. Various test cases are presented, demonstrating the capabilities and efficacy of the method in solving packing type problems.