Civil Buildings Settlement Observation And Prediction Method Applied Research

Deformation monitoring data analysis and forecast is very important for the safety of engineering construction. Firstly, after reading a lot of literature about deformation survey, it summarized the significance, purposes, contents and methods, as well as design contents. Then it described several theoretical model of deformation prediction. Finally, it analyzed the four building settlement observation cases with the forecasting model, then got MATLAB curve fitting model, neural network model, time series model and gray prediction model to predict. And the conclusions are as follows:In the curve fitting model, cubic spline interpolation model (cubic spline) has some forecasting effect. It can be combined with other forecasting models when a large amount of sedimentation or settling velocity faster. Cubic Hermite interpolation model (shape-preserving) and polynomial model (Polynomial) are unsuitable for settlement monitoring and forecasting.In the neural network model, the GR neural network forecast effect is best. BP neural network model and RBF neural network forecasting model is better, and BP neural network model prediction accuracy is slightly higher than the RBF neural network model. They can be combined with the GR prediction neural network model.If the measured data is random, using time series forecasting model, its accuracy is higher than the neural network model. The disadvantage is that the process of establishing and programming model is more complex.The gray prediction model can predict with high accuracy even though have a small amount of measured data. However, if the data is more error, the prediction accuracy will be affected, and even can’t meet the eligibility requirements.