基于动应变数据的桥梁移动荷载识别研究

发布时间：2018-01-05 07:01

本文关键词：基于动应变数据的桥梁移动荷载识别研究　出处：《武汉理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：监测桥上移动车辆荷载，明确其车辆参数对桥梁结构可靠度设计以及运营维护管理等方面都具有重要意义。目前的移动荷载识别理论应用于实际桥梁的技术尚未成熟，现有的车辆称重有成本高等局限性，因此探究出一种简单、快速、有效的移动荷载识别方法具有非常重要的意义。本文利用大型有限元程序ANSYS的APDL语言建立桥梁结构参数化初始有限元模型，结合实桥静载试验数据对初始模型进行有限元静力修正。运用ANSYS对修正后的有限元模型进行移动荷载的动力瞬态分析，对不同车重、不同车速、不同横向位置进行计算分析，提取应变传感器测点的计算结果，采用分阶段识别方法首先识别出车道位置和车速，最后利用计算结果建立车重与车道位置、车速及各测点动应变峰值对应关系的神经网络训练样本，采用BP神经网络的算法利用各测点的动应变峰值数据和已识别出的参数进行移动荷载车重识别，并分别建立了一个模拟模型和实桥对这种方法进行探究。具体研究工作和主要成果如下：（1）利用现场桥梁检测和荷载试验静载数据对有限元模型进行修正，选取各构件抗弯刚度为修正参数，，建立合适目标函数，比较研究不同的优化算法，运用一阶寻优法得到满意的修正结果。（2）模拟车辆荷载在桥梁上移动过程并考虑到动力作用影响，基于修正后的有限元模型，利用ANSYS的APDL语言编制程序模拟移动荷载并对其做动力瞬态分析。（3）探究出利用不同车道对应固定的唯一的各T梁底部动应变峰值大小顺序的规律来进行车道识别的方法；探究出通过找出不同测试截面的动应变峰值或峰值区域中心对应的时间点的差值与其对应的桥梁实际的距离的比值的平均值来进行车速识别的方法，并验证了该方法的有效性。（4）探究出利用不同移动荷载参数作用下各测点的动应变响应值，建立神经网络训练样本，利用训练后的BP神经网络识别车重的方法。（5）选取了广西金鲤水泥有限公司专用码头作为工程实例，对其进行验证行试验，试验结果表明：本文的基于动应变数据的桥梁移动荷载识别方法是有效的。
[Abstract]:Monitor the load of moving vehicles on the bridge. It is very important to clarify the vehicle parameters for the reliability design of bridge structure and the operation and maintenance management. The current technology of mobile load identification theory applied to practical bridge is not mature. The existing vehicle weighing has the limitation of high cost, so it is of great significance to explore a simple, fast and effective method for identification of moving load. In this paper, the parametric initial finite element model of bridge structure is established by using the APDL language of the large finite element program ANSYS. Combined with the static load test data of the real bridge, the initial model is modified by finite element static force. The dynamic transient analysis of the modified finite element model is carried out by using ANSYS, and different vehicle weights and different speeds are obtained. Different lateral positions are calculated and analyzed, the results of strain sensor measurement points are extracted, and the lane position and speed are identified by using the method of phased identification. Finally, the vehicle weight and lane position are established by using the calculation results. The training samples of the corresponding relationship between the velocity and the peak dynamic strain of each measuring point are trained by neural network. The BP neural network algorithm is used to identify the moving load vehicle weight using the dynamic strain peak data of each measuring point and the identified parameters. A simulation model and a real bridge are established to explore this method. The specific research work and main results are as follows: 1) the finite element model is modified by static load data of in-situ bridge detection and load test. The bending stiffness of each component is selected as the correction parameter, and the appropriate objective function is established, and different optimization algorithms are compared and studied. The first order optimization method is used to obtain satisfactory correction results. 2) simulating the moving process of vehicle load on the bridge and considering the influence of dynamic action, based on the modified finite element model. The moving load is simulated by APDL language of ANSYS and the dynamic transient analysis is done. 3) exploring the method of lane identification by using the law of the order of peak value of dynamic strain at the bottom of each T beam corresponding to fixed different lanes. This paper explores the method of speed identification by finding out the average value of the difference between the difference between the peak value of dynamic strain or the center of the peak region of different test sections and the ratio of the actual distance of the bridge corresponding to the difference between the difference of the time point and the actual distance of the bridge. The effectiveness of the method is verified. (4) the method of using the dynamic strain response value of each measuring point under the action of different moving load parameters to establish the training sample of neural network and identify the vehicle weight by BP neural network after training is explored. The special wharf of Guangxi Jinli cement Co., Ltd is selected as an engineering example, and the test is carried out. The experimental results show that the method of bridge moving load identification based on dynamic strain data in this paper is effective.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U446

【参考文献】