Date of Award
8-22-2025
Date Published
September 2025
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical Engineering and Computer Science
Advisor(s)
Chilukuri Mohan
Keywords
Computer Aided Design;Evolutionary Algorithms;Isogeometric Analyses;NURBS;Self-intersection;Structural Design Optimization
Subject Categories
Computer Sciences | Physical Sciences and Mathematics
Abstract
A new end-to-end flexible framework is proposed for optimizing structural designs and geometric shapes using global and local search algorithms. Non-Uniform Rational B-Spline (NURBS)-based representations are deployed for geometric designs and Isogeometric Analyses (IGAs) are carried out for the finite element evaluations. The NURBS-based object representation and associated finite element analyses overcome the issues related to discretization errors of traditional polygon-based finite element meshes. The end-to-end framework integrates Rhino (a NURBS modeler), optimization algorithms, and Abaqus (commercial finite element software) via socket programming. Customized user element and material subroutines are developed to perform IGAs. A wireframe-based parametric modeling of the NURBS objects is used; this approach considers their geometrical features as design variables, without increasing the optimization problem size for dense meshes. Benchmark examples, convergence studies, and design optimization problems are used to illustrate the accuracy and effectiveness of the proposed framework. Several Evolutionary Algorithms (EAs) are evaluated, viz., Genetic Algorithms, Covariance Matrix Adaptation Evolution Strategies, adaptive Particle Swarm Optimization, and Differential Evolution are chosen as evolutionary methods, Sequential Least SQuares Programming is deployed for the gradient-based search. EAs demonstrate a robust performance, whereas gradient-based searches are found to get stuck in the local optima. Potential infeasibility of designs is another problem addressed here: during structural optimization, several self-intersecting designs are generated that render finite element evaluations infeasible. The method of penalizing infeasible designs during search and optimization results in poor optimal solutions. To address this bottleneck, we have proposed two novel self-intersection correction algorithms to detect and correct three-dimensional ill-posed self-intersecting designs, making them feasible for structural analysis.
Access
Open Access
Recommended Citation
Kalia, Subodh, "Structural Design using Parametric Modeling and Optimization Algorithms" (2025). Dissertations - ALL. 2195.
https://surface.syr.edu/etd/2195
