Research On Hydraulic Parameters Estimation And Soil Deformation Controlment In Excavation Dewatering And Recharging

The groundwater seepage is a classical problem in the field of geotechnical engineering.Over the past decades,the underground space has been largely developed,which is a challenge to the construction.When the project is involved with groundwater system.the safety and difficulties will significantly increase with the development of underground space due to the nonuniform spatial distribution of the aquifer system.When the hydraulic connection between aquifers is relatively high,dewatering can largely induce land subsidence outside the excavation.It is mainly because the lack of understanding of local hydrogeology conditions.Then,the performance of artificial recharging cannot meet the requirement.Recently,in engineering,estimating hydraulic parameters are still based on laboratory tests and graphical approaches coupled with measured drawdown data.However,the estimated hydraulic parameters cannot adequately reflect the practical soil properties with respect to groundwater seepage.In this study,the field tests,optimization algorithm,numerical modelling and machine learning approaches are adopted in estimating hydraulic parameters and investigating the mechanism of controlling of land subsidence by artificial recharge.(1)By mixing the genetic algorithm and the Levenberg-Marquardt algorithm,a hybrid algorithm called GALMA is proposed in this study.Coupled with NeumanWitherspoon model and Hantush-Jacob model,the algorithm can be used in estimating hydraulic parameters for leaky aquifer.Validated by several fields pumping tests,GALMA can show its superiority by combining with complex analytical solutions in estimating hydraulic parameters for leaky aquifer.Meanwhile,the algorithm is characterized by the high accuracy result and avoiding to be influenced by the initial guess of parameters.(2)By mixing the metaheuristic algorithm and machine learning method,an algorithm called AHBRO is proposed in this study.Coupled with numerical modelling,it can be adopted in hydraulic parameters in the multi-aquifer system.A series of pumping tests are shown and used to validate the performance of the algorithm.As the results showed,the accuracy of the estimated hydraulic parameters by AHBRO is considered higher.Compared with other algorithms,the performance of the proposed algorithm demonstrates its excellent estimation ability,with a small number of program calls or functional evaluations.(3)According to the analysis of long-term groundwater level and land subsidence data,the spatial and temporal patterns of groundwater level and the land subsidence is studied.Furthermore,the large exploitation from deep aquifer groups is the main reason that land subsidence is still continuous growing,Meanwhile,influenced by the amount of pumping and recharging,the stress history of four aquifer groups are different.(4)By applying the two algorithms mentioned above,the changes of parameters during pumping and recharging and the distribution of soil deformation were investigated in this study.The result shows that,due to the different parameters between compression and rebound,the storage coefficient is smaller during recharge.The increased groundwater level will be underestimated if the same hydraulic parameters are used in the analysis.Meanwhile,the characteristic of elastoplastic of soil can lead the difference deformation in different time sequences of recharging and pumping.With the increasing duration of pumping before recharging,the unrecoverable deformation is growing.