无粘结预应力混凝土梁桥自振频率的RBF网络识别

为研究采用神经网络的方法识别无粘结预应力混凝土梁桥的自振频率,收集以往PC梁的动力试验数据,并在此基础上补充制作5根PC梁进行动力试验,采集相关数据。构建径向基(RBF)神经网络,采用泛化回归神经网络(GRNN)进行函数逼近,径向基函数的光滑因子取为0.15。筛选9个影响PC梁自振频率的关键参数作为神经网络的输入参数,用收集到的试验数据对神经网络进行训练,并预留出1根PC梁的试验数据对网络进行仿真。仿真结果表明,采用所研究的神经网络方法识别无粘结预应力混凝土梁桥的自振频率是可行的,这种网络具有很好的预测能力和泛化能力。

Identifying of Natural Vibration Frequencies of Unbonded PC Beam Bridges by RBF Neural Network

To study the identifying of natural vibration frequencies of unbonded prestressed concrete(PC) beam bridges by the artificial neural network method,the previous dynamic test data of PC beams were collected and in addition to the collected data,another five beams were made and the dynamic tests for the beams were carried out to supplement more relevant data.The radial-based function(RBF) neural network was built,the generalization regression neural network(GRNN) was used to approach the RBF.The smooth factor of the RBF was taken as 0.15 and nine control parameters that had influence on the natural vibration frequencies of the PC beams were sifted as the input parameters of the neural network.The colleted data were then used to train the network,the test data of a PC beam were put aside and were used to simulate the network.The results of the simulation show that the identifying of the natural vibration frequencies of the unbonded PC beam bridges by the neural network method studied herewith is feasible and the network has very good prediction and generalization ability.