动态模糊神经网络在变形预测中的应用

Application of Dynamic Fuzzy Neural Network to Deformation Prediction

为了得到更好的桥梁墩台沉降变形预测精度,减少工程监测实践的误差,分别介绍了基于扩展径向基函数神经网络(RBFNN)与动态模糊神经网络(DFNN)的学习算法和参数的确定方法。选取某一桥梁沉降监测数据分别进行基于扩展径向基函数神经网络与动态模糊神经网络的自适应学习训练,进行桥梁墩台沉降变形预测。实例分析结果表明,径向基函数神经网络预测误差达到0.15 mm,而动态模糊神经网络预测误差达到0.07 mm,显然动态模糊神经网络具有更高的预测精度,从而证实了动态模糊技术与神经网络相结合的自适应学习训练过程的优越性。

To get better prediction precision in settlement and deformation of the bridge piers and reduce errors in project monitoring practices,the learning algorithm and determination of network parameters of dynamic fuzzy neural network（DFNN） based on extended radial basis function neural networks（RBFNN） are introduced.In the selection of subsidence monitoring data from a bridge for the adaptive learning and training based on RBFNN and DFNN,the experimental results show that the prediction error of RBFNN is about 0.15 mm,while the DFNN is about 0.07 mm.The prediction precision of DFNN is better than RBFNN.Thus the advantages of dynamic fuzzy technology and neural network are confirmed in combining adaptive learning and training process.