Research On Settlement Monitoring Data De-moise Processing And Prediction Method In Project

In recent years, with the accelerated pace of urbanization in our country, in order toachieve full utilization of urban space, we have put our eyes on two research directions. Onone hand, building height record has been repeatedly refreshed, on the other hand,transportation lines mileage increase every year. This makes the related projects shouldconsider not only the reasonableness of the structure, but also to control its impact on thesurrounding environment. The settlement in project is one of the impact.Engineering settlement appeares in various forms, and the factors which affect thesettlement of project still without a certain degree of uncertainty, which make the settlementproject become a very complex geotechnical problem. This paper summarizes the basicengineering content on observation, commonly uses methods of observation andobservation requirements. The Wavelet Transform used in the processing of the measureddata conducted in-depth research. On the basis of obtaining accurate data on the monitoringof deformation of the deformable body to predict trends, which is important for ensuring thesafety of the project itself and related buildings.At present, a lot of methods can be used forsettlement prediction, they can be roughly divided into three categories, namely: theoreticalanalysis based on the traditional equation, numerical analysis method based on the theory ofsoil consolidation method and system models based on mathematical analysis.Among them,mathematical modeling forecasting method based on the measured data with its powerfulprocessing capabilities and nonlinear problems has been widely used. In this article, first ofall, determines the parameters of two single prediction models. Then mixes them togethercreat a new. In order to ensure the new model is usefull in the engineering settlementpredition, the papper uses deep foundation measured data engineering and subway projectsas example. By comparing the measured and predicted values of the settlement, get someconclusions for the analysis of similar projects to provide a reference to the settlement issue.In this paper, the settlement excavation monitoring data of a deep foundation pit projectand a rail transit project has been used for this research. First of all, it has made a noiseanalysis with the former raw observational data before the fitting analysis, then establishes aprediction model by to determine the accuracy of the model and model parameters. At last,it uses the followed observation datas to make a reference with the predicte ones.Then it hasmade an analysis of the applicability of these three models, when the number of observationdata and the gap between them is different. By using three predictive models on two typesof actual settlement prediction analysis it found that GM (1,1) model is a simple theoreticalmodel, the model predicte results are stable, but when large data volatility, refer to thepredicted results of the smaller. BP neural network model to predict the short-term effect is better, but the stability of predicted results is poor. Gray neural network model has not onlyweakened the original data characteristics of randomness, but also has a powerful neuralnetwork model of fault-tolerant performance, reliable and highly nonlinear fitting accuracyinterference by a known amount of data is small, or whether it is fitting to predict, and theresults are relatively stable in the short-term settlement prediction works better. So it can beused in the settlement prediction of related projects.