Deformation Modeling And Parameter Estimation Based On Time Series InSAR Technique Over Soft Clay Highway

With the continuous development of China’s transportation field,more and more highways in China are crossing soft clay areas.Due to the characteristics of soft clay with high natural water content,high compressibility,low strength,and poor structure,soft clay highway is more prone to settlement,resulting in deformation and instability.Therefore,it is of great engineering significance to carry out long-term settlement monitoring of soft clay highways after work.However,most of the traditional time-series InSAR deformation models represented by SBAS-InSAR technology are purely empirical mathematical models,which cannot accurately describe the nonlinear deformation time series evolution law of soft clay,and most of them adopt the solution space search method when solving the unknown parameters of the model,which not only relies on the scope of the solution space.This not only relies on the range of the solution space as a-priori information but also takes a lot of time to search,which seriously affects the calculation efficiency and the accuracy of deformation monitoring results.Based on this,this thesis combines the development of the soft ground settlement law of highways,and the research is based on SBAS-InSAR technology,focusing on deformation modeling and parameter estimation,introducing models and parameters that take into account the soft ground settlement mechanism to model the evolution of soft ground deformation with time,and improving the accuracy and reliability of soft ground highway deformation estimation The main research work and results of this thesis are summarized as follow:(1)A Poisson curve and GARN algorithm-based method for estimating time-series deformation of soft clay highways are proposed.This method addresses the complex nonlinear characteristics of soft soil deformation over time,uses the Poisson curve of the road prediction model for InSAR deformation modeling to replace the traditional linear model,and proposes a model parameter estimation algorithm based on the serial genetic algorithm and regularized Newton iteration method.The algorithm validation is based on simulation experiments and real experiments are carried out to verify the reliability of the model and the parameter solution method.The results of the simulation experiments show that the GARN parameter estimation results can still maintain good stability when a higher level of noise is added,and the accuracy is improved by 46.7%compared with the genetic algorithm alone.The real data experiments selected a part of Lungui Road in Foshan City,Guangdong Province as the study area,and obtained its Poisson curve parameters and timeseries deformation fields from January 2015 to January 2017,using The accuracy of the experimental results were evaluated using the level measurement data,and the root mean square error of deformation obtained based on the Poisson curve and GARN algorithm was±2.0 mm,which is about 67%improvement in accuracy compared with the linear model;and about 60%improvement in accuracy compared with the traditional GA algorithm.(2)A time-series InSAR deformation estimation method(FIPR)based on FastICA and inverse summation method for roads in soft soil areas are proposed.This method addresses the deformation modeling link mostly by directly assuming that each component is expressed by a fixed model in the set of differential interference phase equations with equal weight accumulation,which has some artificial assumption uncertainty.Then,the deformation components are modeled and the unknown deformation parameters are solved by the inverse summation method.Finally,the obtained deformation parameters and the contribution of each component of FastICA are used to modify the original time-series InSAR deformation model and generate the total time-series sedimentation in the region.The algorithm is validated using simulated experiments and real data experiments,respectively.The results of simulated experiments show that it is more effective to separate the deformation components using spatial ICA,so the real experiments use the spatial ICA method to obtain the time-series deformation fields of Beijing Capital Airport from January 2012 to February 2015.The modeling accuracy was evaluated using the residual high-pass deformation components of the model,and the results showed that the root-mean-square error of the FIPR method was 2.6 mm,which improved 36.6%compared to the traditional linear model with equal-weight accumulation and 16.1%compared to the traditional Poisson model with equal-weight accumulation.The reliability assessment using the level deformation results shows that the root mean square error of the FIPR method is 1.0 mm,which is 69.7%better than the traditional linear model with equal weight accumulation and about 50%better than the Poisson model with equal weight accumulation.