Date of Award

2001

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Civil and Environmental Engineering

First Advisor

Foley, Christopher M.

Second Advisor

Heirich, Stephen M.

Third Advisor

Vinnakota, Sriramulu

Abstract

The design of steel frames often involves an optimization process, in which the design evolves from an initial to a final configuration. The goal of most optimal steel design problems is to minimize the cost while satisfying performance and construction criteria. Traditionally, the design problem has been solved by trial-and-error dictated by design specifications and guided by the experience and intuition of the designer. However, researchers are continually developing analysis and optimization tools to assist engineers in the sometimes laborious design process and to foster creativity in arriving at "the optimal" design. Over the past decade, the innovation of these design tools has escalated with the advent of high-speed computer processors. The focus of this thesis is the development of an optimization algorithm. to design fully-restrained (FR) and partially-restrained (PR) steel frames. Upon review of the optimization techniques available in the literature, an evolutionary algorithm (EA) was selected to stochastically guide the algorithm through the solution space of available designs and arrive at an evolved frame. Furthermore, the EA implements an object oriented heuristic tree representation of the design variables (i.e. member sizes and connection stiffness). A method of advanced analysis is used to assess the adequacy of the steel frames in lieu of design specification and code requirements. The advanced analysis based : design uses an inelastic analysis capable of capturing the material and geometric nonlinear behavior of the frame members and incorporates a connection model to capture the nonlinear behavior associated with the connections. The EA is specific to the design of unbraced FR and PR steel frames with known geometry, material properties, and loading. Three frames selected from the literature are designed using the EA. The performance of the EA is assessed using convergence trajectories, load deformation responses of the evolved frames, and results from previous researchers. Conclusions are made as to the performance of the evolutionary algorithm and the structural behavior of the evolved frames. Suggestions for future applications of the algorithm are provided and recommendations to improve the performance of the algorithm are outlined as well.

Download

Share

COinS

Restricted Access Item

Having trouble?