摘要
对城市大型、超大型建筑进行持续、高精度和实时的变形监测是预防事故发生的有效手段。针对某高层建筑的变形实测数据,本文分别采用灰色模型(Grey Model, GM)、小波变换、BP神经网络和支持向量机(Support Vector Machine, SVM)模型进行分析和处理,采用残差、残差均方根和残差百分比3项指标从不同方面对预测结果进行对比分析。结果表明:灰色模型实时性最高,SVM在小样本情况下能够获得更好的变形预测性能,小波变换模型具有更优的噪声稳健性,而BP神经网络模型在样本充足时能够获得最优的预测精度。本文的对比评估结果可供读者在实际工程实践中参考应用。
Continuous,high-precision and real-time deformation monitoring of large and super large urban buildings is an effective means to prevent accidents.According to the measured deformation data of a high-rise building,grey model(GM),wavelet transform,BP neural network and support vector machine(SVM)models are used for analysis and processing.Three indicators,residual,root mean square of residual and residual percentage are used to compare and analyze the prediction results from different aspects.The results show that GM has the highest real-time performance,SVM can obtain better deformation prediction performance in the case of small samples,the wavelet transform model has better noise robustness,and the BP neural network model can obtain the optimal prediction accuracy when the samples are sufficient.The comparative evaluation results in this article can be used as reference for readers in actual engineering practice.
作者
丁凯
DING Kai(Guangdong Underground Pipe Network Engineering Survey Company,Guangzhou 510000,China)
出处
《测绘与空间地理信息》
2023年第6期184-187,共4页
Geomatics & Spatial Information Technology
关键词
建筑物形变
灰色模型
小波变换
BP神经网络
支持向量机
building deformation
grey model(GM)
wavelet transform
BP neural network
support vector machine(SVM)