Cross-Section Design Optimization of Support Frame Beam on Slope Based on Finite Element Analysis and Intelligent Particle Swarm Optimization Algorithm
ZHANG Boxiang1,SU Songlin1,MA Wenjing2,SU Wenji1
1.School of Civil Engineering&Transportation , South China university of technology , Guangzhou 510641, China;2.Nanjing Tech University Pujiang Institute , Nanjing 211222, China
In this paper,the improved adaptive inertia weighted intelligent particle swarm optimization(PSO)algo⁃rithm was integrated with with the exterior penalty function to investigate the reinforced concrete support frame beam on a slope with a span length of 6 m subjected to uniform load. FLAC-3D finite element was used to simulate the stresses distribution in the beam subjected to uniform load. The results show that the cost of a double-reinforced beam is relatively lower than that of a single-reinforced beam when the mid-span bending momentreaches the thresh⁃old value of 100 kN · m. When the external force is kept unchanged, increasing the strength of the concrete in⁃creases the cost of the whole beam, while increasing the strength of the reinforcement decreases the cost. Stirrup spacing should be reduced in the arrangement while meeting the requirements. The pape also gives the correspond⁃ing explanation according to the calculation results.
其中pbest和gbest分别为鸟群当前最优位置和鸟群全局最优位置(均为三维空间位置),若pbest ( h, b, s )- x ( h, b, s )为正,说明离最佳位置有距离,应增大速度以靠近最优位置,为负则应当远离此位置,pbest( h, b, s )同理;c1与c2在本文取0.9,分别代表自身位置和全鸟群位置对于下一步位置的影响。
表1 粒子群变量标识Table 1 Variable identification of particle swarm
不妨设造价函数为Q,则问题可以转化为求minQ的值。设罚函数Pi ( x )为第i个约束条件惩罚函数,最终造价函数可以表示为:
其中σp为惩罚因子,主要目的在于放大惩罚函数的权重,本文取1000
外点罚函数计算的基本步骤为:1.给定初始点x,初始罚因子σp,误差要求为ε; 2.计算R ( x )=
,记录每一次迭代的R( x )并记录;3.当两次R( x )误差小于ε或达到设定迭代次数限制(本文设定5000次)时,得到的R ( x )的极值,记录此时的x为极小点。其中P( x )始终为大于等于0的数,当P( x )压缩至0时R( x )最小。以配筋率约束为例,
.
4 优化结果
模型利用MATLAB软件进行程序编写,其基本计算流程图如下图所示:
图7 智能粒子群优化算法流程图Fig.7 Flowchart of the intelligent PSO algorithm
4.1 粒子群优化算法有效性验证
粒子群算法本质为搜索最优空间位置(h,b, s)以降低造价,目标函数R ( x )应当收敛于一常数。
[5] Kennedy J,Eberhart R. Particle swarm optimization[C]//Pro⁃ceedings of ICNN'95-International Conference on Neural Networks. Perth,WA,Australia:IEEE,1995.