基于RBF神经网络静力有限元模型修正的双曲拱桥承载力评估

发布时间：2018-12-11 04:06

【摘要】：双曲拱桥是我国独有的、极具名族气息与特点的桥型，也是上世纪60~70年代建设最多的一种拱桥型式。由于这种结构型式所采用的积木式拼装组合构造型式及低配筋构造使其结构整体性先天不足，加之设计、施工存在先天缺陷，在自然环境及超负荷交通量情况下，在役双曲拱桥存在不同程度的损伤。为了保障交通的顺畅，了解桥梁的实际工作状态（损伤状况、实际承载力等），须对既有双曲拱桥的现实工作状况做出科学的评估。本论文基于RBF神经网络对双曲拱桥初始有限元模型进行修正，建立反映在役双曲拱桥实际状况的有限元模型，基于修正后的有限元模型进行全桥控制截面承载能力系数评估和裸拱极限承载能力评估。本文以淌沟大桥为背景，运用RBF神经网络对初始有限元模型进行修正，以修正后的有限元模型为基准进行桥梁承载力评估。论文主要工作如下： 1、对在役双曲拱桥进行外观调查，根据《公路桥梁技术状况评定标准》、《公路桥梁承载力评定规程》进行桥梁综合评定。将实际拱轴线及影响双曲拱桥承载能力的病害在有限元模型中充分考虑，从而达到模型修正的目的。进行实桥现场静载实验，提取合理实验工况及实验截面进行后面神经网络静力优化样本确定。 2、进行参数灵敏度分析，选取对结构静力特征响应量（挠度）有显著影响的设计参数作为待修正设计参数。确定待修正参数的优化空间，基于均匀设计理论合理选取神经网络训练样本进行神经网络训练。基于训练后的网络，利用RBF神经网络的泛化特性，求出设计参数的目标值即待修正参数的实际值。为了验证径向基神经网络的修正性能，，采用ANSYS自带的一阶优化算法进行有限元模型修正，进行两者的结果对比分析，验证基于径向基神经网络的可行性及实用性。 3、以修正后的桥梁有限元模型为基准从截面的真实强度、恒载与活载效应、结构损伤三方面进行桥梁承载力系数计算。考虑拱肋弹性模量折减、拱肋有效面积折减及超载计算结构控制截面的承载力系数从而综合的评价桥梁承载力。基于极限承载力方法验算双曲拱桥裸拱在各种荷载组合下的极限承载力。
[Abstract]:The hyperbolic arch bridge is a unique bridge type with unique and famous atmosphere and characteristics in our country. It is also one of the most arched bridges built in the 1960s and 1970s. Because of the structural integrity of the building block assembly and combination structure and the low reinforcement structure, and because of the inherent defects in the design, the natural environment and the overloaded traffic volume are in the condition of the natural environment and the overload of traffic. In service, the hyperbolic arch bridge has different degree of damage. In order to ensure the smooth traffic and understand the actual working state of the bridge (damage condition, actual bearing capacity, etc.), it is necessary to make a scientific evaluation of the actual working condition of the existing hyperbolic arch bridge. In this paper, the initial finite element model of hyperbolic arch bridge is modified based on RBF neural network, and a finite element model is established to reflect the actual situation of the hyperbolic arch bridge in service. Based on the modified finite element model, the load capacity coefficient of the control section of the bridge and the ultimate bearing capacity of the bare arch are evaluated. In this paper, the initial finite element model is modified by RBF neural network, and the bridge bearing capacity is evaluated based on the modified finite element model. The main work of this paper is as follows: 1. The external appearance of the double-curved arch bridge in service is investigated, and the comprehensive evaluation of the bridge bearing capacity is carried out according to the Evaluation Standard of the Technical condition of the Highway Bridge and the rules for the Evaluation of the bearing capacity of the Highway Bridge. The actual arch axis and the diseases affecting the bearing capacity of hyperbolic arch bridge are fully considered in the finite element model so as to achieve the purpose of model modification. The static load experiment of the real bridge was carried out, and the reasonable experimental conditions and the experimental section were extracted to determine the static optimization samples of the back neural network. 2. The parameter sensitivity analysis is carried out, and the design parameters which have significant influence on the static response (deflection) of the structure are selected as the design parameters to be modified. The optimization space of the parameters to be modified is determined, and the neural network training samples are reasonably selected based on uniform design theory for neural network training. Based on the trained network and the generalization characteristic of the RBF neural network, the target value of the design parameters is obtained, that is, the actual value of the parameters to be modified. In order to verify the modification performance of radial basis function neural network, the first order optimization algorithm of ANSYS is used to modify the finite element model, and the results are compared and analyzed. The feasibility and practicability of radial basis function neural network based on radial basis function neural network are verified. 3. Based on the modified finite element model of the bridge, the bearing capacity coefficient of the bridge is calculated from three aspects: the true strength of the section, the effect of dead load and live load, and the damage of the structure. Considering the reduction of elastic modulus of arch rib, the reduction of effective area of arch rib and the calculation of bearing capacity coefficient of control section of structure under overload, the bearing capacity of bridge is evaluated synthetically. The ultimate bearing capacity of bare arch of hyperbolic arch bridge under various load combinations is verified based on ultimate bearing capacity method.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U441;U448.221

【参考文献】