PCA-RBF神经网络模型在桥梁变形预测中的应用分析  

Application and Analysis of PCA-RBF Neural Network Model in Deformation Prediction of Bridge

作　　者：戴建彪 岳东杰[1] 陈健[1] 何国庆 Dai Jianbiao;Yue Dongjie;Chen Jian;He Guoqing(School of Earth Sciences and Engineering, Hohai University)

机构地区：[1]河海大学地球科学与工程学院

出　　处：《勘察科学技术》2019年第3期39-42,48,共5页Site Investigation Science and Technology

摘　　要：在由多种因素导致的桥梁变形进行变形预测时,针对大多神经网络预测方法具有局部最优等局限性,以及高维度的影响因子之间具有很强的相关性和信息重叠性等问题,该文提出基于主成分分析(PCA)的径向基函数(RBF)神经网络来进行桥梁变形预测,并和直接采用RBF神经网络的预测结果进行了对比。实例分析证明:两种方法的训练误差都在10-15mm级,均能很好地训练样本,其中基于主成分分析的径向基函数神经网络误差更小,提高了预测精度,这对变形监测中分析主要的影响因子从而建立预测模型具有参考意义。When predicting bridge deformation caused by many factors,aiming at the limitations of most neural network prediction methods such as local optimum,and the strong correlation and information overlap between high-dimensional influencing factors,this paper proposes a radial basis function(RBF) neural network based on principal component analysis(PCA) to predict bridge deformation,and compares the prediction results with RBF neural network directly. The example shows that the training errors of the two methods are all at 10-15 mm level,and both of them all can train samples well. Among them,the error of RBF neural network based on principal component analysis is smaller,the predicted accuracy is improved. It has reference significance for analysing main influencing factors in deformation monitoring and establishing prediction model.

关 键 词：主成分分析 RBF神经网络 桥梁变形 变形预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] U445.7[自动化与计算机技术—控制科学与工程]