基于改进灰色模型的建筑物变形预测分析研究  

Prediction and Analysis of Building Deformation Based on Improved Grey Model

作　　者：张恒僖 ZHANG Hengxi(Jiangxi Provincial Architectural Design and Research Institute Group Co.,Ltd.,Nanchang 330046,China)

机构地区：[1]江西省建筑设计研究总院集团有限公司,江西南昌330046

出　　处：《测绘与空间地理信息》2025年第3期177-180,183,共5页Geomatics & Spatial Information Technology

摘　　要：传统GM(1,1)模型难以准确预测复杂施工场景下建筑物的变形趋势,本文采用方差补偿自适应Kalman滤波方法对建筑物变形监测原始数据进行滤波处理,剔除随机误差影响,利用高质量数据序列构建GM(1,1)模型,预测建筑物变形趋势;然后利用时间序列AR(p)模型对GM(1,1)模型预测结果趋势项进行拟合优化,降低预测残差,进一步提高建筑物变形趋势预测结果的准确性。以某临近地铁车站基坑施工项目的建筑物沉降监测数据为数据源,分别构建GM(1,1)模型、GM(1,1)-AR模型,Kalman滤波-GM(1,1)-AR组合模型,对比分析不同模型预测结果,Kalman滤波-GM(1,1)-AR组合模型预测结果与实测值基本一致,较于GM(1,1)模型,预测精度提升了78.95%,预测结果更为准确可靠,为地铁工程安全施工及建筑物安全分析提供了可靠的数据支撑。It is difficult for traditional GM(1,1)model to accurately predict the building deformation trend in complex construction scenarios.This paper uses variance compensation adaptive Kalman filter method to filter the original building deformation monitoring data,eliminate the influence of random errors,and build a GM(1,1)model using high-quality data series to predict the building deformation trend.Then,the time series AR(p)model is used to optimize the trend item of the GM(1,1)model prediction results,reduce the prediction residual,and further improve the accuracy of the building deformation trend prediction results.Taking the building settlement monitoring data of a foundation pit construction project near a subway station as the data source,the GM(1,1)model,GM(1,1)-AR model and Kalman filter-GM(1,1)-AR combined model are constructed respectively,and the prediction results of different models are compared and analyzed.The prediction results of Kalman filter-GM(1,1)-AR combined model are basically consistent with the measured values.Compared with GM(1,1)model,the prediction accuracy is improved by 78.95%,and the prediction results are more accurate and reliable,providing reliable data support for the safe construction of subway engineering and the safety analysis of buildings.

关 键 词：GM(1 1)模型 时间序列AR(p)模型 Kalman滤波-GM(1 1)-AR组合模型 预测精度 

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]