基于Kalman滤波的组合预测模型在建筑物变形监测中的应用  

Application of Combined Prediction Model Based onKalman Filter in Building Deformation Monitoring

作　　者：王靖 杜国政 WANG Jing;DU Guozheng(Zhejiang Institute of Surveying and Mapping Science and Technology,Hangzhou 310030,China)

机构地区：[1]浙江省测绘科学技术研究院,浙江杭州310030

出　　处：《测绘与空间地理信息》2024年第5期202-204,207,共4页Geomatics & Spatial Information Technology

摘　　要：根据建筑物沉降监测数据的特点,结合Kalman滤波算法、BP神经网络模型以及AR自回归模型在数据降噪、数据预测中的优势,提出并构建了一种新的基于Kalman滤波的BP-AR沉降预测模型。该组合预测模型实现建筑物变形预测的主要步骤为:首先,通过Kalman滤波算法对原始观测数据进行降噪,消除随机噪声误差对观测数据的影响;其次,通过BP神经网络模型对降噪后序列进行建模与预测;最后使用AR模型对预测残差进行建模与预测。通过实际建筑物沉降监测数据对本文提出的组合预测模型进行验证,结果表明相较于BP神经网络模型与BP-AR模型,本文提出的组合预测模型的预测精度更高,有效降低了噪声影响,具有较高的优越性。According to the characteristics of building settlement monitoring data,combined with the advantages of Kalman filtering al-gorithm,BP neural network model and AR auto-regressive model in data noise reduction and data prediction,this paper proposes and constructs a new BP-AR settlement prediction model based on Kalman filtering.The main steps of building deformation prediction based on the combined prediction model are as follows:firstly,the original observation data are reduced by Kalman filtering algorithm to eliminate the influence of random noise error on the observation data;secondly,BP neural network model is used to model and pre-dict the noise reduction sequence;finally,AR model is used to model and predict the residual error.The combined prediction model proposed in this paper is verified by the actual building settlement monitoring data.The results show that compared with BP neural network model and BP-AR model,the combined prediction model proposed in this paper has higher prediction accuracy,effectively reduces the impact of noise,and has higher superiority.

关 键 词：建筑物 沉降预测 KALMAN滤波 BP神经网络模型 AR自回归模型 

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]