Optimization of Cable Force of Extradosed Bridges
苟昌煥,吳宗達,林珮玗
Civil Engineering
College of Architecture and Design
chkou@chu.edu.tw
Abstract
In order to solve structural optimization problems in the past, it is necessary to
integrate structure analysis software and optimization software. Since structural
analysis has been the only function considered when developing most of structure
analysis software, they suffer from closeness of system. Therefore, it is not easy to
integrate them with optimization software. This study proposes an experimental
design method to solve this problem including following steps: (1) generate
experimental design, (2) implement experimental design, (3) construct a response
variable model, (4) define optimization problem, (5) solve optimization problem. The
purpose of step (1) through step (3) is to create a response variable model of
alternative structure analysis software. This model is a set of regular and simple
functions, it can easily define the optimization problem in step (4), in order to
facilitate step (5) to solve the optimization problem. The reason to employ neural
network instead of traditional regression analysis in step (3) is that the relationship
between internal forces and displaced cross section dimension of the structure are
often nonlinear. Neural network is a nonlinear system that gives itself the greatest
advantage to accurately construct a nonlinear model. This case study is based on
optimization design of cable force and tower height in an extradosed bridge,
evaluating the feasibility of above method and comparing with published references
to confirm the proposed method in this study is applicable to the optimization design
of extradosed bridge.
Keyword：extradosed bridge, cable force optimization, neural network, experimental
design method