Research On Deformation Analysis And Prediction Methods Based On Multivariable Grey Model

Deformation monitoring is mainly used the measuring instrument to obtain the continuous sequence of engineering project,then through some professional treatment technology to assessing the trend of the development and the security statement.With the continuous updating of the observation methods,the deformation monitoring ways are changed from the periodic measurement mode to the modern measurement mode that contain the multi-sensor?the satellite positioning?the photogrammetry and remote sensing?the mobile communication and so on.The plane displacement monitoring system is transformed into a comprehensive monitoring system which contains the temperature,pressure,stress and displacement.This thesis mainly focuses on the Multi point deformation analysis and forecast research at poor information in the early operation stage of the deformation monitoring system.In the course of the research,the author discusses several key technologies of multi point prediction and presents three different optimization methods:(1)Multi-point grey prediction model that sensor data errors is taken into account;(2)tandem residual multi-point grey model;(3)Multi scale and points parallel deformation prediction modelThe main contributions are as follows:1.Compared and analyzed five methods of the multi-point grey model.The results show that the five models can preferably simulate and predict the settlement sequence data,but the dynamic prediction method has higher accuracy than the static prediction method.2.Through the simulation experiments demonstrate that the filtering model in the situation of small data will eliminate some information.So,the sensor observation error will cause some disturbance in the model parameter calculation.This paper using the mix least squares and multivariate total least-squares solution to inhibition the disturbance in the parameter calculation.so as to achieve unbiased solution to improve the modeling accuracy and prediction accuracy.3.This paper discusses the combination forecasting model and divides the combination forecasting model into two types:parallel combination model and tandem combination model.Based on the theory of grey model,the author designs a tandem residual multi-point grey model.Intensive exploration the common forms of deformation monitoring series and extract the deformation factor sequence and actual settlement sequences by empirical mode decomposition,then using support vector machine and multi point grey model to predict the deformation factor sequence and actual settlement sequence to obtain the final prediction sequence.