一种建筑沉降叠加预测方法

A method of building settlement superposition prediction based on ARMA model

针对高层建筑的沉降监测与趋势预报问题,结合时间序列分析方法,该文提出一种基于ARMA的趋势项和随机项叠加预测法,把沉降监测时间序列数据分解为趋势项与随机项,分别建立趋势回归函数模型与随机项ARMA模型,叠加进行沉降量的预报,并通过上海外滩某高层建筑的沉降监测实例,研究并比较了该方法与传统的ARIMA差分预测法对建筑沉降预报精度的影响。实验结果表明:基于ARMA的趋势项和随机项叠加预测法在沉降预报中精度优于基于ARIMA的差分预测法。该方法利用趋势回归函数的保持作用,克服了传统的时间序列ARIMA模型在长期预测中精度不高的问题,并且随着预测步长的增加,优势更加明显。

In view of the settlement monitoring and trend forecasting of high-rise buildings. using time series analysis method. this paper presented the trend components and stochastic components superposilion prediction method based on ARMA modelinwhich the settlement monitoring time series data were decomposed into trend coniponents and stochastic components. and then the trend regression fund io n model and ARMA modelwere respectively established to forecast buildingsettlement together. Through the high-rise building settlement examples of Shanghai. the influenee of this method and the traditional ARIMA difference prediction method on the prediction accuracy of building settlement was studied and compared. And then the experiment results showed that the trend components and stochastic components superposition prediction method based on ARMA model wassuperior to the difference prediction method based on ARIMA modelin settlement prediction. which used keepingin fluence of the trend regression function and overcomedthe problem that traditionaltime series ARIMA model was not accurate in long-term prediction. With the prediction step increased,this advantage was more obvious.