摘要
为有效控制和预测深基坑开挖引起的周围地表沉降变形,以保定市汽车科技产业园深基坑工程为依托,采用MIDAS GTS NX有限元软件对基坑开挖过程进行模拟,并将实际值与模拟值进行对比,验证模型的准确性。对测点DB-1和DB-3采用不同降噪尺度进行降噪处理,选出合适的降噪尺度,以降噪后的测点DB-2监测数据作为径向基神经网络输入矢量,构建基于小波变换的RBF预测模型,以滚动预测方法对基坑测点DB-2沉降进行预测。结果表明:通过对监测数据多尺度分解,分离出监测数据中的真实信号与噪声信号,并对这些高频的噪声信号进行过滤,可以有效地达到降噪的目的;测得W-RBF模型的平均绝对误差为0.279 54、均方根误差为0.324 99、平均相对误差率为8.42%、最大误差为0.186 74 mm,R^(2)为0.983 71,通过对比,均优于RBF模型;经过实际工程验证,经过小波降噪的RBF神经网络模型具有较高的精度,能够满足实际工程的需要。
In order to effectively control and predict the settlement and deformation of the surrounding surface caused by the excavation of the deep foundation pit,the deep foundation pit project of Baoding Automobile Technology Industrial Park is used as the basis.First,the measurement points DB-1 and DB-3 are subjected to noise reduction processing with different noise reduction scales,and the appropriate noise reduction scale is selected.A wavelet transform-based RBF prediction model is constructed to predict the settlement of foundation pit measuring point DB-2 by rolling prediction method.The results show that:Through the multi-scale decomposition of the monitoring data,the real signal and the noise signal in the monitoring data are separated,and these high-frequency noise signals can be filtered,which can effectively achieve the purpose of noise reduction.The average absolute error of the WRBF model is 0.27954,the root mean square error is 0.32499,the average relative error rate is 8.42%,the maximum error is 0.18674 mm,and the R2 is 0.98371.By comparison,they are all better than the RBF model.After actual engineering verification,the RBF neural network model of wavelet noise reduction has high accuracy,which can meet the needs of practical engineering.
作者
于磊
田海川
张博
张纯
董玉雄
马林杰
Yu Lei;Tian Haichuan;Zhang Bo;Zhang Chun;Dong Yuxiong;Ma Linjie(Beijing Municipal Road and Bridge Co.,Ltd.,Beijing 100000,China)
出处
《山西建筑》
2024年第5期66-70,共5页
Shanxi Architecture
基金
河北省自然科学基金资助项目(E2018201106)
河北省交通运输厅科技项目(PHP-C34200-2503631-1)
北京市政路桥科研项目(ERDC0026)。
关键词
深基坑
小波变换
神经网络
基坑监测
deep foundation pit
wavelet transform
neural network
foundation pit monitoring
