Research And Application Of Deformation Prediction Of Foundation Pit Based On Neural Network

The foundation pit is a kind of pit that is excavated at the basic design position according to the datum size and the basic elevation of the architectural design.The excavation plan shall be based on the local geological structure,the climate hydrological environment and the conditions of the nearby structures.Waterproof and drainage measures for foundation pits should also be done well.In recent years,with the accelerating process of urbanization in China,super high-rise buildings,super-large shopping malls,and subway projects have been increasing,and the scale of foundation pit projects has been continuously expanded and expanded.During the excavation of the foundation pit,the excavation unloading in the pit will cause the displacement of the structure between the inner and outer parts to be displaced,causing deformation of the external soil structure of the foundation pit and finally causing the settlement or movement of the foundation pit.At the same time,it is also affected by climate,geographical environment and various other uncertain factors.Therefore,there are many problems in the field of foundation pit monitoring.The traditional foundation pit monitoring method has a large workload and a lot of time.During the monitoring process,there are problems of inaccurate positioning and low precision of the foundation pit monitoring points.Most foundation pit monitoring only pays attention to the feedback of monitoring results after deformation,and ignores the early warning of displacement deformation.It often takes remedial support measures after the deformation displacement is found,which misses the best time for repair and repair,and also brings significant security risks.Aiming at the above problems,this paper proposes a neural network based foundation pit deformation prediction method,which is applied in a foundation pit monitoring project in S city.Firstly,the visual close-range photogrammetry technology is used,and the error compensation method is used to improve the recognition accuracy from the perspective of reducing the camera calibration error,thereby optimizing the accuracy of the position center of the pit monitoring point.Then the genetic neural network method is used to predict the deformation displacement of the foundation pit.According to the actual engineering data set,the effect of the genetic neural network method on the deformation prediction of the foundation pit is comprehensively and deeply analyzed.The deformation of the foundation pit is predicted from three angles: horizontal displacement,vertical displacement,horizontal displacement and vertical displacement.The experimental results show that the horizontal and vertical displacement prediction results are more accurate.At the same time,the displacement prediction is accurately modeled by the time dimension and the spatial dimension and its feature correlation.The quantitative analysis and research on the combination of time domain(time series only),spatial(adjacent points only)and combination of time domain features and spatial features are carried out.The experimental results show that the results of time series features and spatial neighboring point feature prediction are more accurate.Finally,the method is compared with other typical machine learning methods.It is mainly compared with support vector machine regression and random forest regression method.The experimental results show that the maximum relative error of the reverse propagation neural network prediction result after genetic algorithm optimization is 0.36%,the prediction error is basically floating within-0.05-0.05 mm,and the fitting index reaches 0.9835,which is higher than BP neural network,support vector machine regression and random forest regression by 0.0601,0.1318 and 0.1915 respectively.Close-range photogrammetry acquires monitoring data,considers the influence of different characteristics,and uses genetic algorithm to optimize neural network to predict foundation pit deformation provides a new idea for intelligent foundation pit monitoring.It improves the real-time and predictability of foundation pit monitoring and can reduce the group pit injury.The incidence of accidents provides guarantees for the development of green cities and the healthy development of cities.