基于BP神经网络考虑节点刚度的输电铁塔有限元建模技术

Finite element modeling technology of transmission tower considering node stiffness based on BP neural network

由于输电铁塔半刚性节点的节点刚度会受外载荷的影响而变化,所以为了修正输电铁塔有限元模型的节点刚度,在利用公式计算半刚性节点初始节点刚度的基础上,提出一种基于BP神经网络预测输电铁塔节点刚度的策略。首先采用ANSYS有限元分析软件,构造考虑节点刚度的2A-J2型输电铁塔模型。通过对铁塔结构的分析,可以将铁塔节点分为主材搭接节点K型节点。再根据不同节点处角钢尺寸与螺栓数量的不同,将节点分为7类,计算出每类节点初始刚度,并为其设置刚度系数。在大风工况下,得到输电铁塔在不同节点刚度时对应的杆件轴力,并建立数据集。通过MATLAB构建神经网络模型,在上述数据集的基础上,得到节点刚度的神经网络修正结果。并对比在神经网络隐含层不同层数、不同节点的情况下,半刚性节点节点刚度的数值模型模拟结果与神经网络预测结果,得到该神经网络性能最佳时的隐含层参数。当建立铁塔精确有限元模型时,可以通过测量部分杆件轴力,利用上述BP神经网络建立的“杆件轴力-节点刚度”映射函数进行预测,进而得到所建模型的真实节点刚度。仿真结果验证了考虑节点刚度的输电铁塔精确化建模方法的准确性与可行性。

Since the joint stiffness of semi-rigid nodes of transmission towers changes due to external loads,in order to correct the joint stiffness of the finite element model of transmission towers,proposes a strategy based on BP neural network to predict the joint stiffness of transmission towers based on BP neural network based on the calculation of the initial joint stiffness of semi-rigid joints.Firstly,ANSYS finite element analysis software was used to construct a 2A-J2 transmission tower model considering joint stiffness.Through the analysis of the tower structure,the tower nodes can be divided into main material lap nodes and K-shaped nodes.According to the difference in the size of angle steel and the number of bolts at different nodes,the joints are divided into 7 categories,the initial stiffness of each type of node is calculated,and the stiffness coefficient is set for it.Under strong wind conditions,the axial force of the member corresponding to the stiffness of the transmission tower at different nodes is obtained,and the data set is established.The neural network model is constructed by MATLAB,and the neural network correction results of node stiffness are obtained on the basis of the above data set.In addition,the numerical model simulation results of the stiffness of semi-rigid nodes and the neural network prediction results are compared with the numerical model simulation results of the node stiffness of the semi-rigid node under the condition of different layers and different nodes of the neural network,and the hidden layer parameters when the neural network has the best performance are obtained.When the accurate finite element model of the iron tower is established,the axial force of some members can be measured and predicted by using the“member axial force-node stiffness”mapping function established by the BP neural network above,and then the real joint stiffness of the built model can be obtained.The simulation results verify the accuracy and feasibility of the transmission tower precision modeling method considering the joint stiffness proposed in this paper.