考虑随机误差的软土基坑沉降趋势分析及预测

Analysis and prediction on settlement trend of soft soil foundation pit considering random error

为准确分析基坑沉降变性规律,基于现场监测数据,通过卡尔曼滤波对趋势项及误差项进行分解,采用M-K检验对发展趋势进行评价,利用优化广义回归神经网络和差分整合移动平均自回归模型,构建基坑沉降分项预测模型,并将预测结果与发展趋势评价结果对比分析,以实现基坑沉降变形规律综合研究。结果表明:卡尔曼滤波能有效分解基坑沉降数据趋势项与误差项,相较于传统小波分解效果更佳;基坑沉降呈持续增加趋势,但趋势性逐渐减弱;预测结果相对误差均值均不大于2%,预测模型精度较高;沉降变形会进一步增加,但增加速率明显降低,与发展趋势分析结果一致,两者相互佐证分析结果准确性。研究结果为基坑沉降变形规律分析提供新思路。

In order to realize the accurate analysis on the settlement and deformation laws of foundation pit,based on the field monitoring results of the settlement and deformation of foundation pit,the Kalman filter was used to decompose the trend and error terms,and the M-K test was used to evaluate the development trend.The optimized generalized regression neural network and differential integrated moving average autoregressive model were used to construct the sub-term prediction model of foundation pit settlement,and the prediction results were compared with the evaluation results of development trend to realize the comprehensive study on the settlement and deformation laws of foundation pit.The results showed that the Kalman filter could effectively decompose the trend and error terms of foundation pit settlement data,which had obvious advantages compared with the traditional wavelet decomposition.The foundation pit settlement had a continuous increasing trend,but the trend tended to be weakened.The average relative error of the prediction results was less than 2%,which showed that the prediction model had a high prediction accuracy.The settlement and deformation would further increase,but the increase rate reduced significantly,which was consistent with the analysis results of development trend,and the accuracy of their respective analysis results was confirmed by each other.The research results provide new ideas for the analysis on the settlement and deformation laws of foundation pit.