Research On Deformation Mechanism And Prediction Of Reservoir Landslide Based On DFOS Technology

Landslide,as a widely distributed geological hazards in China,which seriously threaten people's life and property.The reservoir landslides refer to the landslides formed on the bank of reservoirs under the fluctuation of the reservoir water level,especially after the completion of the large reservoirs.As it is often accompanied by secondary disasters such as debris flows,impulse water waves,floods,etc,reservoir landslides are more widespread,diversified and intense.Landslide is a complex system composed of solid,liquid and gas,whose deformation and failure process shows multifield evolution characteristics.The monitoring of multi-field information of landslides provides basic data for the assessment of landslide risk and the deformation prediction as well as early warning.Considering many deficiencies existed in traditional monitoring technology and methods,they are difficult to meet the requirements of multi-field information acquisition,analysis and evaluation.In this paper,based on distributed fiber optic sensing(DFOS)technology,the relevant sensing cables and sensors for multi-field monitoring are developed,and the-multi field monitoring system of reservoir landslide is established.Based on the laboratory model test,the deformation response law of the slope under the action of rainfall and water level fluctuation was studied.Combined with Majiagou landslide in the Three Gorges Reservoir area,the DFOS monitoring system was established.The DFOS monitoring system includes full distributed monitoring system and quasi-distributed monitoring system.The full distributed monitoring system can determine the deformation mode and key deformation position of Majiagou landslide.The quasi-distributed monitoring system can accurately capture the multi-field real-time information at the key positions.Then,by analyzing the monitoring data in detail,the trigger factors and deformation mechanism of Majiagou landslide were analyzed.Finally,based on the machine learning algorithm,a landslide deformation prediction model considering deformation hysteresis effect is proposed,and the deformation of Majiagou landslide is accurately predicted.The research results are as follows:(1)The research status and application progress of several kinds of distributed optical fiber sensing technology are introduced in detail.The novel optical fiber sensors developed by the research group,which is suitable for monitoring the deformation,seepage and temperature field of the reservoir landslide,are introduced.The in-place inclinometer based on FBG technology is designed and developed,and its monitoring performance is tested by laboratory test,which verifies the feasibility of its application in landslide deformation monitoring.(2)The error estimation formula for DFOS-based inclinometer displacement calculation was obtained.It is found that the displacement error is proportional to the square of the length.Whereas,it varies inversely with the diameter of tube.Based on the characteristics of the monitoring data,a machine learning algorithm is proposed to correct the displacement error of the inclinometer.(3)The slope model tests under the action of rainfall and reservoir water level fluctuation are carried out by using DFOS technology.By recording the water content,pore water pressure and deformation of the slope under the action of rainfall,it is found that due to the decrease of soil strength caused by the "sudden increase" of pore water pressure,the instability of slope was induced.The distributed strain sensing(DSS)cable can effectively identify the deformation area of the slope.Moreover,the sharply increase of strain 5 minutes before the slope failure can be detected,which proves the feasibility and accuracy of the DFOS technology deployed in slope monitoring and early warning.By arranging FBG strain sensors in the horizontal and vertical directions of the slope,the deformation response characteristics and failure mechanism of the slope under the action of the reservoir water level fluctuation are explored.It is found that when the water level in the slope rises rapidly and the reservoir water level drops rapidly,the safety factor of the slope decreases quickly.The conclusions drawn through laboratory test provides guidance and basis for the field monitoring.(4)Taking Majiagou landslide as an example,the DFOS monitoring system was established.The DFOS monitoring system includes full distributed monitoring system and quasi-distributed monitoring system.The full distributed monitoring system can determine the deformation mode and key deformation position of Majiagou landslide.The quasi distributed monitoring system can accurately capture the multi-field realtime information at the key positions.(5)The deflection and internal forces of anti-slide piles were estimated through the monitored strain data and comparison between the estimated and designed internal forces are made.Before March 2015,the overall deformation of Majiagou landslide is small,signifying it is in an relative stable state,and then the piles gradually exerts its anti-sliding function.In addition,the shear force of the anti-slide pile is close to the designed value for the present.That is,the anti-slide pile is in an unstable state and needs to be strengthened.(6)The method of data mining is used to analyze the multi field monitoring data.The displacement rate of Majiagou landslide exhibits zoning.When the reservoir water level is lower than 150 m,the overall deformation rate of Majiagou landslide is at a relatively high level;When the reservoir water level is between 150 and 160 m,and the fluctuation velocity is less than 0.4m/d,the deformation rate of Majiagou landslide is at a high level;Whereas,fluctuation velocity of the reservoir water level is greater than 0.4m/d,the deformation of Majiagou landslide tends to be stable.The deformation of Majiagou landslide is in a stable state when the reservoir water level is between 160 m and 175 m.(7)The reservoir water level and its fluctuation velocity affect the stability of Majiagou landslide by changing the seepage field.When the water level of the reservoir drops rapidly or remains at the low water level,the transient seepage force impinges along the sliding direction.Due to the hydraulic gradient and the lag of the water level of the rear edge,the deformation of the landslide accelerates.When the water level of the reservoir is slowly decreasing,rising rapidly or at the high level,the landslide is in a relatively stable state.This can be interpreted by the dissipation of pore water pressure,seepage force and the effect of hydrostatic pressure.(8)To achieve accurate prediction of the Majiagou landslide displacement in the Three Gorges Reservoir(TGR),China,a hybrid machine learning prediction model considering the deformation hysteresis effect is proposed.Before making the prediction,the grey correlation analysis was adopted to confirm that the fluctuation of the reservoir water level was the main influence factor.Then,the dynamic analysis of the correlation between periodic influence factors and periodic displacement was carried out,and the deformation lag time was identified by using set pair analysis(SPA)method.Finally,the optimal influence factors were selected and the prediction model of Majiagou landslide based on SPA-PSO-SVM(Support Vector Machine Optimized by Particle Swarm Optimization)was established and the displacement was accurately predicted.