Experimental Study On The Influence Of Low Temperature Freezing On Macroscopic And Microscopic Properties Of Gravelly Soil

With the support of China’s western development strategy,the demand for quantity and quality of earth-rock dams are more obvious in high-cold areas where construction is complicated and difficult.However,great temperature difference between day and night will lead to frost heave of dam body under construction.Conventional methods such as surface removal and laying insulation foam will not only weaken the flatness and adhesion between the old and new dam bodies,but also interfere with the already short construction time and cost.Therefore,the selection of local dam construction materials should be strictly controlled,starting from the basic material composition and grading,and using the fabric state and mechanical properties of dam construction materials,the allocation method of dam construction materials that can meet the requirements of dual impermeability and frost resistance in the construction process of earth-rock dam in high cold area should be explored.In this way,the excavation materials in local material yard can be fully used to meet the design requirements,improve the winter construction efficiency of earth-rock dam in high cold area,and provide an economic and effective choice for winter construction antifreezing of projects in high cold area.In this paper,the composition design of gravelly soil was carried out by referring to the mixture with high probability of intermediate composition in the material yard.On this basis,the basic fabric state and mechanical properties of gravelly soil were preliminarily studied,and then random distribution model of gravelly soil was simulated by Monte Carlo method and COMSOL,and the permeability properties of gravelly soil were preliminarily understood from the mesoscopic direction.Then,permeability test,NMR test and unidirectional freezing test were carried out on gravelly soil,and the influence mechanism of gravel content,particle composition,water content,dry density and temperature on full range of permeability,unfrozen water content and freezing parameters was systematically studied on the basis of micro-water migration and evolution.Subsequently,according to the test results,prediction models of macro and micro performance parameters of gravelly soil under these factors were established through machine learning,which realized the goal of real-time acquisition and verification,and also provided theoretical guidance for selection of earth-rock dam material and allocation of construction materials.The main research contents were as follows:（1）With the continuous addition of sand and gravel,gravelly soil gradually develops from a suspension-dense structure to a skeleton-void structure through a skeleton-dense structure.Therefore,the shear failure mode of gravelly soil presented a trend of strain hardening→plastic failure→strain softening,and the cohesion and internal friction angle increased first,then decreased and then increased,respectively,and the law of slow and fast growth.The boundary content of these processes was 30%and 60%.On this basis,gravelly soil could have stronger shear resistance and bearing capacity by increasing the grain size,reducing the initial water content or increasing the dry density.（2）Monte Carlo method was used to establish a random circular distribution model of gravelly soil,and the seepage field of gravelly soil was simulated by changing gravel content,Cc and Cu.Based on the correspondence between grading parameters,permeability coefficient and Kozeny-Carman equation,a better permeability prediction model for gravelly soil with variable grading was obtained.Combined with simulation and test results,it was found that the water seepage path of gravelly soil would first decrease and then increase with the increase of gravel content,which indicated that 30%～40%gravel could also reduce the order of magnitude of permeability coefficient by one order of magnitude to-5.The permeability of gravelly soil could be further improved by reducing control particle size d₆₀ and dry density of gravelly soil.（3）According to water occurrence state and freezing order obtained in NMR experiment,it was concluded that representative values（supercooling temperature,steep drop,curve slope,and final stable value）of gravelly soil in the freezing characteristic curve were larger with higher moisture content and smaller control particle size d₆₀.Increased amount of mixed gravel would increase supercooling temperature and volume,but also let the curve slope and final stable value decrease.Dry density basically did not change the type and distribution of water in gravelly soil,only when temperature was lower than-7.5℃,unfrozen water content had a relatively obvious inverse relationship with dry density.Then,based on the KC equation,hydraulic radius and microscopic pore channel theory,the permeability model of negative temperature in unsaturated gravel soil was established,which combined microscopic point of view by using unfrozen water content with T2 spectrum parameters of gravelly soil at different temperatures.（4）After freeze-thaw cycle,microscopic pore structure of gravelly soil will change with the increase of times of F-T cycle,and pore structure would increase with the increase of times of F-T cycle,and the first two F-T cycles had the strongest interference on pore structure,while structure would gradually reorganize and stabilize after the fifth F-T cycles.Under the influence of this fundamental structural change,permeability coefficient and porosity of gravelly soil with a skeleton-dense structure and low seepage path,gravel content of 30%-40%were least affected by F-T cycles,and the growth rate was relatively weak.In addition,the larger d₆₀ and dry density of gravel,the stronger structural stability of gravelly soil,and relatively insensitive to temperature change,the conditions of water seepage and migration were limited,and the increase of permeability coefficient was smaller.（4）Through the open system unidirectional freezing test,it was clear that three freezing parameters of temperature distribution,frost heave deformation and water migration of gravelly soil have minimum values when gravel content was 30%～40%.At this time,when controlled particle size d₆₀ of gravel was further increased,final temperature gradient of gravelly soil was reduced under the influence of thermal conductivity,the phenomenon of water migration and accumulation was hindered,and gravelly soil was transformed into weak frost heaving soil or non-frost heaving soil.Increasing moisture content and dry density of gravelly soil could improve the suction effect of frozen front and ductability of unfrozen water film,so that water migration amount and frost heave rate would increase accordingly.In terms of temperature difference,gravelly soil with higher water content would have more temperature difference gradient due to obvious phase transition heat dissipation,but increasing of dry density would reduce temperature difference gradient by improving thermal conductivity.（5）Combined with permeability simulation and laboratory test of gravelly soil,it was concluded that gravelly soil could have the best permeability and frost resistance when gravel content was roughly between 30%and 40%,control d₆₀ was appropriately increased,C_u≥5 and Cc was slightly larger than 1,and the maximum dry density and optimal water content should also be maintained,gravelly soil can have the best anti-seepage and freezing resistance.In addition,when gravelly soil was applied to an actual project,it was necessary to set up some engineering seepage prevention measures according to site environment and conditions and use the way of "upstream plugging and downstream discharging" to reduce and prevent seepage deformation of each part of dam.（6）In order to ensure the uncertainty of initial data and weight difference,KPCA method was improved by using "standardization and information entropy" to pre-process test results.The performance parameters of gravelly soil were predicted by GA-BP model,PSO-SVR model,and BO-RFR model.Finally,with reference to different error parameters and computational cost,it was concluded that GA-BP model was more suitable for predicting the minimum internal temperature and freezing rate of gravelly soils,PSO-SVR model was more suitable for predicting cohesion and angle of internal friction,and BO-RFR model was more suitable for predicting permeability coefficient and unfrozen water content.