| Summary: | 碩士 === 國立臺灣大學 === 土木工程學研究所 === 105 === Cable-stayed bridge is one of the most popular bridge types chosen by architects and structure engineers in modern bridge design because of its aesthetic appeal and economy. However, the design of cable-stayed bridge is very complex due to its highly statically indeterminate. Design of such complex structure with a large number of design variables and constraints with traditional methods is inevitably time consuming and cannot guarantee the optimality of the final design.
The research mainly focuses on the optimization design of cable-stayed bridge. A flexible structure optimization software, SODIUMM, is used and expanded with two global search algorithms, Particle Swarm Optimization (PSO) and Artificial Bee Colony (ABC). By applying these two global search algorithms, some specific problems of cable-stayed bridges can be solved and provide engineers better optimization pattern than by applying local search algorithms.
Basic concept of structure optimization, main features of SODIUMM, optimization procedure and the B-spline technique are briefly introduced in sequence. Then PSO and ABC are thoroughly discussed, including algorithm description, parameter selection, fitness evaluation, algorithm process, and pseudo code. In order to test PSO and ABC about their optimize performance and stability, simulations are made by using 9 typical functions.
After adding PSO and ABC into SODIUMM, SAP2000 is used to set up initial model of single pylon cable-stayed bridge, then choose the total bending energy of structure as objective function and analyze. In this research, a third algorithm named COBYLA is used for comparison. From the optimization results it is concluded that PSO will reach convergence fast, but with a relatively poor stability. Meanwhile, ABC is an effective global optimizer, but with relatively low efficiency. However, both PSO and ABC could get results 40 percent better than that getting from COBYLA.
During an optimization design of cable-stayed bridge, engineers could choose a suitable algorithm according to the need for design time, the complexity of the model or the economic goal.
|
|---|
