基于GEP优化的神经网络在高层建筑变形预测中的应用研究

发布时间：2018-05-20 14:32

本文选题：高层建筑物 + 基因表达式编程　；参考：《江西理工大学》2015年硕士论文

【摘要】：随着社会的高速发展和人们不断地向城市聚集,城市土地越来越紧张,大量的高层建筑物也随之产生。高层建筑物的变形可能会导致灾害的发生,所以高层建筑物的安全问题显得格外重要。如果可以对高层建筑物的变形进行准确地预测,就可以有效地减少灾害带来的损失。但是,目前常见的预测方法应用于高层建筑物的变形预测都存在不足之处,导致预测精度不够。因此,如何准确地预测高层建筑物变形对减少灾害的发生有着重大的研究意义。本文首先阐述了人工神经网络和基因表达式编程(GEP)的国内外研究现状,并介绍了高层建筑物变形监测的相关理论及目前常见的预测模型;其次,根据基因表达式编程原理,利用其简单的基因编码和强大的全局搜索能力,解决了径向基函数神经网络初始权值的选取和隐层中心向量(神经元)的个数难以确定的问题,提高了网络的收敛性和精度;然后,将GEP的优点应用于径向基函数神经网络,对网络进行优化,建立了GEP优化的径向基函数神经网络变形预测模型。最后,以陕西某高层建筑物为实例,将基于径向基函数神经网络的预测模型和基于GEP优化的径向基函数神经网络的预测模型应用于该高层建筑物沉降变形预测,并对两个模型的预测结果进行对比分析。结果表明,基于GEP优化的径向基函数神经网络的预测模型比基于径向基函数神经网络的预测模型的预测值的精度提高了将近一倍,从而说明基于GEP的径向基函数神经网络的预测模型在高层建筑物变形预测领域更具有研究的意义和推广价值。
[Abstract]:With the rapid development of the society and the continuous gathering of people to the city, the urban land is becoming more and more tense and a large number of high-rise buildings are produced. The deformation of the high-rise buildings may lead to the occurrence of disasters, so the safety of the high-rise buildings is particularly important. If the deformation of the high-rise buildings can be accurately predicted, it can be accurately predicted. It can effectively reduce the loss caused by disasters. However, the common prediction methods used in the deformation prediction of high-rise buildings are inadequacies, which results in insufficient prediction accuracy. Therefore, how to accurately predict the deformation of high-rise buildings is of great significance to reduce the occurrence of disasters. Through the domestic and foreign research status of network and gene expression programming (GEP), the related theories of deformation monitoring of high-rise buildings and the common prediction models are introduced. Secondly, based on the principle of gene expression programming, the initial weights of radial basis function neural networks are solved by using their simple gene coding and powerful global search ability. The selection and the number of the central vector (neuron) of the hidden layer are difficult to determine, and the convergence and accuracy of the network are improved. Then, the advantages of GEP are applied to the radial basis function neural network, the network is optimized, and the radial basis function neural network variable shape prediction model of the GEP optimization is established. Finally, a high-rise building in Shaanxi is taken as the real one. The prediction model based on radial basis function neural network and the prediction model of radial basis function neural network based on GEP optimization are applied to the prediction of the settlement deformation of the high-rise building, and the prediction results of the two models are compared and analyzed. The results show that the prediction model based on the radial basis function neural network based on the GEP optimization is based on the basis of the prediction model. The prediction accuracy of the prediction model of radial basis function neural network is nearly doubled, thus the prediction model of radial basis function neural network based on GEP is more meaningful and popularized in the field of high building deformation prediction.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TU196.1

【相似文献】