| The limitation of earth resources,the continuous growth of population and people's increasing demand for resources lead to the obvious shortage of available space above ground.In order to alleviate the pressure of the earth,underground shopping malls and high-rise buildings gradually increase.The construction of underground engineering and high-rise buildings cannot be separated from foundation pit engineering.The safety coefficient of foundation pit construction represents the safety degree of the project itself and surrotnding things.Therefore,foundation pit deformation monitoring and data processing have gradually entered peopled vision.Deformation monitoring data can be processed in a wide range,including adjustment,de-noising,simulation,prediction and analysis of observation results.The simulation and prediction analysis of deformation monitoring data is the focus of this paper.In this paper,the engineering deformation monitoring data of Chongqing tower was taken as an example.Two prediction models,ARIMA and kalman filter,were adopted to carry out simulation prediction analysis,which was realized based on MATLAB programming.The main objects of analysis were the maximum subsidence point BM17 around the foundation pit and the sub-maximum subsidence point BM16.Due to the phenomenon of odd change in the observation data,the paper used the two times median error as the limit to judge the odd change value,which was regarded as coarse error when exceeding the limit value,and carries out experiments on it with the mean filtering and median filtering.The comparison showed that the mean filtering was more suitable for the research data in this paper.The observed values of BM17 were changed by manual intervention method to verify the adaptation of the mean square filtering.According to the experimental results,when the odd shape of V shape and inverted V shape,M shape and W shape,N shape and inverted N shape appeared,the two mean value filtering was used to process the odd value.When there was odd change of arch and inverted arch,three mean filter was used to deal with the odd change.The maximum subsidence point BM17 using ARIMA,kalman filtering,mean ARIMA and mean-kalman filter and so on four kind of simulated prediction model,the comparison shows that the mean-simulation of kalman filtering prediction effect was best,the sum of squared residuals(SSE)was 3.789,the root mean square error(RMSE)was 0.251,the mean absolute error(MAE)was 0.200,the simulation predicted value of correlation coefficient(R squared)was 0.836,two period extrapolation forecast for 20.15 and 20.18 mm,less than 30 mm,in the range of early warning.The sub-maximum sink point BM16 was used for verification,and the verification results were consistent with the previous conclusion,and the mean-kalman filter model was optimal.According to the accuracy index table,Kalman filter has better simulation prediction effect for the observation data of the foundation pit.After the two-mean processing,the mean-Kalman filter has the highest simulation prediction accuracy,which is the optimal model in this paper.Through comparison of precision indexes,it was known that kalman filter had a better simulation and prediction effect when the singular value was not processed for the observation data of the foundation pit studied in this paper.After the binaiy mean processing,the simulation and prediction accuracy of mean-kalman filter was the highest,which was the optimal model in the experiment of this paper.Therefore,in the absence of singular values,the kalman filter model had the best prediction effect.Figure[28]Table[23]Refenence[82]... |
PDF Full Text Request