| With urban development and population growing,land resources are becoming increasingly tense.As a result,increasing high-rise and super-high-rise buildings are built for more efficient use of land resources.Meanwhile the foundation pits are getting larger and deeper.The surrounding environment of deep foundation pit projects in cities is usually highly complicated.Many existing buildings and structures may exist around the foundation pit,such as buildings,urban roads,viaducts,tunnels,subway stations and municipal pipelines.Due to the existence of these buildings and structures,the construction of deep foundation pits in cities faces many disadvantages and safety risks.At the same time,in recent years the rapid development of automated monitoring technology has made the monitoring information of foundation pits more comprehensive and time-sensitive.Therefore,how to make full use of the monitoring information of the environment of deep foundation pit to realize the prediction and control of the safety is of far-reaching significance.In this paper,a series of studies were conducted on intelligent identification and prediction method of the environmental monitoring information of deep foundation pit.Besides,the studies are applied to the foundation pit project of Shanghai Xujiahui Center Project as verification.The research of intelligent recognition includes missing information filling,abnormal data processing and information cluster analysis.The existing methods are collated and analyzed for processing missing data and abnormal data.The principles,advantages,disadvantages and scope of application of different processing methods are compared.Combined with some of the traditional methods,a comprehensive processing method is proposed for missing and abnormal monitoring data of the surrounding environment of deep foundation pit,which can achieve higher processing efficiency and accuracy.On this basis,the monitoring information clustering method based on Gaussian mixture clustering is proposed.The monitoring points are clustered according to the characteristics such as location,changing speed of the monitoring value and current value.Besides,the algorithm is improved through the preliminary selection of core measuring points,optimizing the results and improving the efficiency by 59.6%.After the clustering,the representative monitoring point of each cluster is selected and the probability parameters of each cluster are counted.These streamlined and filtered information represents overall dynamic information of the surrounding environment.Then,the long-term and short-term memory neural network is applied to the prediction of the surrounding environment of the foundation pit.The prediction model of long-term and short-term memory neural network adjacent to the foundation pit is established first.The data set is divided into a training set,verification set and test set.Then prediction errors of different hyperparameter combination models on the verification set are compared to select the optimal model.Based on the optimal model,the representative monitoring point and probability parameters of each cluster is predicted and accurate results are obtained.The RMSE of representative monitoring point and probability parameters prediction reach 0.3294 and 0.2987 respectively,which verifies the effectiveness of long-term and short-term memory neural network model.Besides,the update prediction is carried out based on the measured data and the RMSE of results are reduced by 65% and 75% for representative monitoring points and probability parameters.Finally,the intelligent identification and prediction method proposed in this paper is applied to the foundation pit project of Xujiahui Center project in Shanghai.Environmental factors such as tunnel displacement,displacement of the enclosing structure,surface settlement outside the pit and groundwater are considered.The results verify the applicability of this method. |
PDF Full Text Request