Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.485219
Title: Development of the Generalised Boundary Points Method for Shape Optimisation using Binary and Real-coding Genetic Algorithms
Author: Sia, Yaw Yoong
ISNI:       0000 0001 3408 3228
Awarding Body: The University of Leeds
Current Institution: University of Leeds
Date of Award: 2007
Availability of Full Text:
Access through EThOS:
Abstract:
Shape optimisation of continuum structures is concerned with changes to the external boundary of a structure to reach the optimum design. A structure is said to be optimal, if the objective function has the greatest degree of merit and the structure itself satisfies both the geometry and behaviour constraints at the same time. Genetic Algorithms (GA) are used as the optimisation tool in this work, because they have the distinct 'feature of multi-point local and global searches. Therefore, GA provide a better chance to obtain the global optimum solution. .' The aim of this work is to apply GA to the problem of shape optimisation. Past research has concentrated in applying the Binary-coding GA (BGA) to shape optimisation. In this method, where shape modification is achieved by allowing the Boundary Points (BP), which are attached to the design boundary, to move unit step-wise in a specific direction in 2D space. The presented work extends this method by allowing the BP to have multidirection movement. Three examples with different movement cases, including the Step-wise (SW), Step-wise Recursive (SWR), and Multi-direction (MD) are considered. The results indicated that the MD movement achieves the best optimum solution, followed by the SWR, and the SW movement method. In terms ofthe convergence rate, the SW method converges at the highest rate, followed by the SWR, and then MD method. Two adaptive methods, hybrid BP movement and step-size refinement, are introduced to improve the convergence rate and quality of the solution. The obtained results indicated that these adaptive methods produce better solutions With a higher convergence rate in comparison to individual BP movement met~ods. The adaptive methods are combined to optimise the shapes ofa spanner, torque arm, bracket, and·3D structures. This work also investigates the robustness of different .types of crossovers fOf the Realcoding GA (RGA). This investigation is conducted because the crossovers for RGA were developed to have different search attributes, which would affect the qu?lity o~ the solution. The crossovers selected for the investigation are:~Single-point (ip), Two''p~int (2p), Uniform (uni), Arithmetic (Ar), Geometric (Geo), Wright'S heuristic (WH), Blend crossover (BLX), Fuzzy Recombination (FR) and Simulated Binary crossover (SBX). In additional to the individual crossover investigation, the crossovers are combined to find the most robust operator for RGA. The combination of 'BLX, FR, and SBX is found to be the most rQbust crossover for RGA. In this work, the Fixed Grid Finite Element Analysis (FG-FEA) is used to determine the structural responses during the optImisation process. Past research has concentrated on the development and i~provement of the FG-FEA for 2D structural analysis. As 3D optimisation is undertaken in this work, it was necessary to extend the method to '3D structural analysis. The computational methodology' of 3D FG-FEA is presented in this thesis.
Supervisor: Not available Sponsor: Not available
Qualification Name: The University of Leeds, 2007 Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.485219  DOI: Not available
Share: