Please use this identifier to cite or link to this item:
Title: A remesh-free finite-element method with large geometrical variations and its application to electric motor design
Authors: Liu, Xiaoyu
Degree: Ph.D.
Issue Date: 2018
Abstract: In this thesis, a novel fast remesh-free finite element method (FEM) for electric machines and other electromagnetic (EM) devices design optimization is studied to promote the speed of the process of design optimization. Two-dimensional (2D) engineering design optimization problems can be solved on the base of the developed software packages in C++. The contribution of this work includes an overlapping remesh-free mesh deformation technique which employs a virtual boundary technique for finite element algorithm. When geometrical deformation happens, the virtual boundary technique and the two layer mesh system makes it is unnecessary for researcher to regenerate entire mesh. In this research to accelerate the optimal design process of EM devices, the adaptive Newton-Raphson (N-R) method based on stabilized bordered block diagonal form (SBBDF)is first proposed and applied to practical nonlinear problem. Another approach is that two types of response surface models (RSM) are applied in optimization process to replace the objective function which is traditionally calculated by FEM at each iteration. For the purpose of saving the working load of designers, general patterns of PM motors are designed. Based on these patterns, novel optimization algorithms are proposed and verified by practical single objective and multi-objective problems. All the proposed algorithms are achieved by programming computer code in C++.
In this thesis, the following work has been done: (1) 2D FEM programs have been implemented based on A-φ potential formulations for solving general electromagnetic field problems. (2) 2D remesh-free mesh generation for problems with geometric deformation for both small and large geometrical deformation has been discussed. (3) N-R method for nonlinear problems has been studied. An Adaptive N-R method based on SBBDF for 2D finite element solver for electromagnetic problem involving nonlinear material is presented. Parallel computingis adopted for the adaptive N-R iteration. (4) Several optimization algorithms are studied and RSMsfor optimization are introduced for permanent magnet(PM) motor design problem. A dual RSM is proposed for the ALOPEX based optimization. The objective function is approximated by the combination of two types of RSMs for the purpose of reducing the error of the approximation. (4) A RBFEM combined with k-means clustering is presented and a 2D electromagnetic field problem is used to validate the accuracy and the effect of reducing computing time. (5) General patterns of PM components of PMSM are proposed for optimization problems, both single objective and multi-objective. RSMs are applied in single objective problems to accelerate the optimization process. Parallel calculation is applied for multi-objective problem and several objective function values are computed by FEM at the same time. The main contributions of this work can be concluded as: . A novel 2-D remesh-free FEM with large geometrical variations is developed. . An adaptive N-R method based on SBBDF is applied to 2D electromagnetic field problem. . An adaptive RBM to ensure that full FEM computation is only needed on limited sample points is developed. . Apply the RSM to the ALOPEX optimization method of PM motors . General pattern of PM motor is designed. Two RSM of design objective are applied for single objective problems. . General pattern of PMs of PMSM is designed.NSGA-II based on this general pattern is applied for multi-objective problem and the FEM in the optimization process is completed in parallel. . Engineering design optimization problems are solved by applying the proposed methods and evolutionary algorithms of optimization with different types of EM devices.
Subjects: Hong Kong Polytechnic University -- Dissertations
Engineering design
Machine design
Pages: xiv, 212 pages : color illustrations
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of Jun 4, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.