基于BP神经网络的正交异性钢桥面系多目标结构优化设计

正交异性钢桥面铺装体系结构优化就是设计出力学性能最好、重量最轻的桥面系,给出铺装结构参数的优化组合。本文建立一个钢桥面铺装体系多目标优化模型,其中目标函数是正交异性钢桥面板铺装结构众多参数的复杂函数,可以应用BP神经网络模型模拟这种函数关系。利用大量的有限元计算结果作为BP神经网络的训练样本,将训练后的BP网络用于优化求解过程中的目标函数值的仿真计算。最后给出多目标优化模型的求解步骤和算例。

Multi-Objective Optimum Design of the Paving System on the Orthotropic Steel Bridge Deck by the BP Neural Network

Structural optimum design of the paving system on the steel bridge plate is to optimize the structural parameters,by which the paving system with minimun weight and most excellent mechanical performance can be designed.A multi-objective optimum mathematical model is built,the objective functions of which are the intricate functions of the structural parameters of the paving system on the orthotropic steel bridge deck.A BP neural network can be applied to model these intricate functions.With lots of FEM calculations being the training samples,the trained network is used for simulated calculations of the objective functions during the solution of the optimum model.A numerical example is demonstrated in the end of this thesis.