摘要
针对5种典型单层球面网壳结构型式(联方型、施威德勒型、凯威特型、三向格子型和短程线型),本文采用基于离散变量的序列两级截面优化算法对一百个不同跨度(30m^60m)、不同矢高(1/3~1/7)的网壳进行截面优化设计,在截面优化设计结果中提取选型所需的优秀样本,一个优秀样本由3部分组成:跨度,矢跨比和该跨度、矢跨比下用钢量最少的网壳结构类型;然后利用BP人工神经网络对复杂非线性映射关系的模拟能力建立网壳跨度、矢跨比与用钢量最少的网壳型式之间的映射关系,避免了结构选型优化的多维、多重非线性模型的求解困难,通过截面优化简便地实现了网壳结构的选型优化。
From five classic types of Single-layer steel reticulated dome （Schwedler Dome, Lamella Dome, Kiewitt Dome, Three way grid Dome and Geodesic Dome）, the paper gives the section optimum design to one hundred shells of five different types and different size （rise-span ratios change from 1/7 to 1/3, spans change from 30m to 60m） based on sequential two-level algorithm. Selected good samples were chosen from those sectional optimization results. The good sample has 3 aspects： span, rise-span ratio and corresponding shell type with minimum weight in the span and rise-span ratio case. Building up the mapping relationship between span, rise-span ratios and type of shell with minimum weight by using the Nonlinear function simulation capabilities of BP artificial neural network. This way could eliminate many problems caused by solving a nonlinear mechanical and mathematical model of lectotype optimization, provides an easy and convenient way to accomplish lectotype optimization of Single-layer steel reticulated dome through section optimization.
出处
《建筑钢结构进展》
2010年第4期46-50,共5页
Progress in Steel Building Structures
关键词
BP神经网络
截面优化
选型优化
BP artificial neural network
section optimization
lectotype optimization
