Join our welcoming International Design Community.
Complete this application form to join now.
Author: Miao, Yi
Supervisor: Fadel, Georges M.
Institution: Graduate School of Clemson University
The three dimensional packing problem, also known as configuration design, has found wide applications in the mechanical engineering field. However, due to the intractable nature of the problem, it is time prohibitive to identify the global optimum using exhaustive search methods. This work is focused on solving the three dimensional packing problem using optimization algorithms. Although there is no guarantee to find the global optimum, the algorithms can find acceptable solutions within a reasonable time. This attribute makes such an approach attractive to engineering applications.
The optimization algorithm selected in this work is the Genetic Algorithm (GA). Although the GA has been used to solve packing problems for a long time, there is no thorough study on how to apply and customize a Genetic Algorithm to deal with 3D packing problems. To achieve better performance, the “Packing GA” is developed. A test case demonstrates that the algorithm developed is far more efficient than the traditional GA in solving packing problems.
Building on the Packing GA, this work further proposes a general configuration design process, which includes four parts: Modeling, CAD Data Preparation, Optimization and Presentation. Every part in the process and several important issues related to configuration design are studied. Among these issues, the complexity of collision detection is especially analyzed through a theoretical analysis, which is further verified using several test cases.
To cater for the requirements of the vehicle configuration design problem, whose goal is to maximize vehicle performance by finding optimal positions of vehicle components, the Packing GA is extended and integrated with a Multi-Objective Genetic Algorithm so that the approximation of the Pareto front can be identified by just one run of the algorithm. The results show that the method and tools developed generate a rich set of the non-dominated solutions, from which an engineer can select the “best” alternative.