基于SVM-KF组合模型的建筑物变形预测方法  

Building Deformation Prediction Method Based on SVM-KF Model

作　　者：吴之林 WU Zhilin(Guangzhou Urban Planning,Survey,Design and Research Institute,Guangzhou 510095,China)

机构地区：[1]广州市城市规划勘测设计研究院,广东广州510095

出　　处：《测绘与空间地理信息》2023年第12期211-214,217,共5页Geomatics & Spatial Information Technology

摘　　要：建筑物变形过程呈现出典型的非平稳和非线性特征,传统基于单一模型的变形预测方法存在精度低、稳定性差的问题。综合考虑建筑物变形过程中的非线性和线性因素,本文提出一种基于支撑向量机(Support Vector Machine,SVM)和卡尔曼滤波(Kalman Filter,KF)相结合的变形预测方法。首先将变形时间序列分解为非线性和线性变化2部分,之后利用SVM和KF分别对非线性变化趋势和线性变化趋势进行建模预测,最后将二者的预测结果进行综合得到最终预测结果。利用建筑物变形实际数据进行试验验证,结果表明,SVM-KF组合模型相对于SVM、KF和BP神经网络等单一模型能够获得更高的预测精度和预测稳定度,具有一定的实用价值。The deformation process of buildings presents typical non-stationary and nonlinear characteristics.The traditional method of deformation prediction based on single model has low accuracy and poor stability.Considering the nonlinear and linear factors in the process of building deformation,a deformation prediction method based on the combination of support vector machine(SVM)and Kalman filter(KF)is proposed.Firstly,the deformation time series is decomposed into two parts:nonlinear and linear change.Then,SVM and KF are used to model and predict the nonlinear change trend and linear change trend respectively.Finally,the prediction results of the two are integrated to obtain the final prediction results.The experimental verification using the actual data of building deformation shows that the proposed SVM-KF model is more accurate and stable than the single model such as SVM,KF and BP neural network,and has certain practical value.

关 键 词：建筑物变形 支撑向量机(SVM) 卡尔曼滤波(KF) 变化预测 组合模型 

分 类 号：P25[天文地球—测绘科学与技术] TB22[天文地球—大地测量学与测量工程]