Slip Distance Prediction Research Of Heifangtai Loess Landslide Based On Numerical Simulation

The Loess Plateau accounts for about 6.6% of the total land area of China.Due to the special nature of the loess's spontaneous structure,it will exhibit the characteristics of yield,softening,strength reduction and collapsibility due to the interference of external factors,especially water.Disasters occur frequently,and the total number of disasters accounts for about one-third of the national total.For the Heifangtai area in Gansu,due to long-term agricultural flood irrigation,the groundwater level has been rising,resulting in an average of 3-5 landslide geological disasters per year,which seriously restricts the local social,economic and cultural development.At present,there have been a lot of research results on the disaster-causing factors,catastrophic mechanism,disaster mode and evolution process of the loess landslide in Heifangtai.However,there are few reports on the slippage prediction of the loess landslide in Heifangtai.The spatial slip prediction of the loess landslide in Heifangtai can provide a scientific basis for landslide warning and prevention.This article revolves around the purpose of slip ratio prediction.Based on the topographic data of 7 typical loess landslides before and after of Heifangtai from 2015to 2017.Firstly,through the discrete element numerical simulation software,six series of landslides are subjected to a series of orthogonal tests with parameters as variables,and the parameters of each group and their corresponding simulation forms are obtained?the lateral width of the sliding source area,the longitudinal width of the sliding source area,and the wall height of the landslide,the overall drop of the landslide,the slip distance,the lateral width of the accumulation area?;then use this as a sample for neural network training to determine the mapping relationship between the simulated shape and the parameters;and then use this mapping relationship to invert the most of the 6 landslides and the simulation parameters are excellent.Taking the range of the optimal simulation parameters of these six groups as the value range of the numerical simulation parameters of the loess landslide in this region.Then,the reliability of the parameter range is verified by the forward performance of the seventh landslide.Finally,the slip value of the Chenjia 7#potential landslide is predicted by using the mean value of the value range as the parameter.Mainly get the following conclusions:?1?The constitutive model,calculation time step,grid search size and discrete element software of the discrete element numerical simulation of the loess landslide in Heifangtai were determined.Through the study of discrete element theory,realizing that the mechanical mechanisms of different contact models are different.So according to the force characteristics of the particle motion process,the constitutive model between the particles and the particles of loess landslide in Heifangtai should be JKR wet particles.The contact model between the particle and the geometry should adopt the RVD rolling contact model.The method of determining the time step should adopt the Rayleigh wave method,the search side length value of the mesh should be set as1.5d_{ma x}.In the end,By comparing of the contact model library,particle contact search algorithm and time step calculation method finally determines the simulation of the loess landslide in Hefangtai should use EDEM discrete element software.?2?A fast balanced filling method of particles in the EDEM discrete element model is achieved.Because the EDEM discrete element program can not be directly modeled or the modeling process is very cumbersome,this paper uses the software of Surfer,Polywork and other software to process the data of the Hefangtai and constructs a three-dimensional model of the landslide.Combining the discrete element PFC software and the fast-filling module developed by the API in EDEM software,the“layer-by-layer brik filling method”is used to realize the fast balanced filling of particles in the EDEM discrete element model.Then according to the previous mechanical test study,indoor angle of repose test,and GEMM material database,the particle size,Poisson's ratio,elastic modulus,density,static friction coefficient between the particle and the sliding source area geometry,and rolling friction coefficient were determined.The reasonable values of the six parameters,such as the rolling friction coefficient,the static friction coefficient,the overall collision recovery coefficient of the model and the energy density of the JKR model,provide the basis for the design of the orthogonal test.?3?The neural network method is used to invert the optimal parameters corresponding to the six landslides,and the six sets of optimal parameters are used as the parameter range of the loess landslide in Hefangtai.Through the forward performance of the 7th landslide,the accuracy of the value range was verified.Because of the high nonlinearity of the macroscopic and meso-mechanical parameters in the discrete element,so,in order to find a fast and effective method to obtain the discrete element mesoscopic parameters that can reflect the actual situation,this paper firstly hrough the sensitivity analysis of the parameters,determining the JRK model surface density energy should be 100J/m².And then the orthogonal experiment carried out according to the principle of“small variation gradient with small sensitivity and small variation gradient for parameters with small sensitivity”,the total number of six landslides is 1548 times.The simulated morphology corresponding to different parameters is obtained.Then,By using it as a neural network training sample,a mapping relationship model between simulated morphology and parameters is constructed,and then the model is used to realize the parameter inversion of the six landslides.The optimal parameters obtained from the inversion of the six landslides are used as the parameter range of the loess landslide in Heifangtai.Finally,the accuracy of the value range is verified by the forward performance of the seventh landslide,which avoids the complexity and randomness brought by manual calibration of mechanical parameters under the same geological conditions.?4?For the first time,the"limited trailing edge"finite element strength reduction method was proposed to search for the potential sliding surface of the loess landslide in Heifangtai.The"limited trailing edge"finite element strength reduction method is based on the distribution of cracks and sinkholes as the trailing edge control boundary,using the displacement contour as the basis for the potential slip surface to search for the potential slip surface;then through the pair of 2019.03 Jiaojia 6# landslide,2015.01 Jiaojia 4#landslide and 2017.10 Dangchuan 9#landslide slip surface search,proved the feasibility,applicability and accuracy of the method.Also proved that is a new idea for the simple,fast and accurate search for the potential sliding surface of the loess landslide in Heifangtai.?5?Realize the prediction and analysis of the slip distance and motion process of Chenjia 7#potential landslide.In order to determine the slippery shape of the Chenjia7#potential landslide,this paper uses the UAV image and the actual investigation on the site to determine the possible trailing boundary of the CJ7#landslide by using the distribution of cracks and sinkholes as the control boundary.Using the"limited trailing edge"strength reduction method to search for its potential slip surface,determine the potential slip surface shape of CJ7#landslide,and then use ItasCAD 3D geological modeling software to construct a 3D discrete element model of potential landslide body.And then reverse The Chenjia section of the performance has the mean value of the landslide discrete element mesoscopic parameters as the parameters.The slip distance of the CJ7#landslide along the defined trailing edge boundary is about 380m,and the maximum stacking width is about 97m.The height difference between the edge shear outlet and the leading edge of the landslide stack is about 59m,and the maximum stack thickness is about 10m.When the landslide occurs,the starting response sequence of each area is:B3>B2>B1>T2>T1.After the landslide occurs,the relative front and rear and the stacking sequence positions of each area are basically consistent with those before sliding.