A Study On The Mechanism Of Moisture Migration In Seasonal Freezing Soil Based On Back-Propagation Neural Network

Seasonal frozen soil is a kind of special soil-water system with ice crystal, and it is composed of soil grain, ice, pore water, and gas. In China, it distributes in the northward regions of north latitude 30°and its area is up to 513.7×104cm2 that occupies about 53.5 percent of Chinese territory.Seasonal frozen soil distributes widely, and it has a close connection with the production and living of human being. The problem caused by water translocation plays a very important role in the research of agriculture, water resources, environment, and engineering. The northern area is disturbed by the problem of Seasonal freeze injury. The problems of swelling, settlement, and road mudding all come from water translocation.At the present time, the research of water translocation has obtained some achievements, but the research of northern seasonal frozen soil is not adequate. And due to the limits of test condition, most of soils freezing tests still are carried through in the laboratory which is not suitable for multivariate nature boundary condition. The thesis is sponsored by the national natural science fund—micro mechanism study on water translocation of roadbed soil in northeast seasonal frozen area. On the basis of field monitoring at the orientation of Changchun-Songyuan, Changchun-Siping, and the north to Changchun-jilin, combining with laboratory tests, this paper makes the research of mechanism of water translation in seasonal frozen soil in Changchun. By using the capability of Artificial Neural Network in dealing with large scale and nonlinear system, this paper establishes the prediction model of BP neural network to the problem of water translation in seasonal frozen soil. The main jobs and results are as follows:1. This paper studies the material composition, physical and chemical properties of seasonal frozen soil in Changchun. The granularity composition has more than 70% silt and more than 20% clay. Mineralogical compositions mostly are quartz and feldspar. Structure connection is mainly bound water. All this characters make the channels in soil free to water translocation. The soluble salt content is small in soil, but there are so many Na+ which will make bound water more thicker in the pore of silty clay which has worse dispersity. This will help the outer bound water translocation.Microstructure is analyzed by scanning electronic microscope in this paper, and it is found that the seasonal soil in Changchun area has cluster structure. There is mainly small and medium-sized pore which lies in the soil in which has not bonding materials. The grains are bonded mainly by bound water. They can supply the channel of water translocation.Through the research of heat exchange coefficients and the hydraulic power parameters, the influence of temperature can be ignore to calculate the specific heat and conductivity factor. The change laws of diffusion coefficient and water transmissibility coefficient are similar; they both increase with moisture content of soil.2. Four monitoring sites are selected to monitor water contents, frost depth, and underground temperature during a frozen-thaw. The data of site can be supplied to analyze the water translocation. Through the long-term monitoring data, the lowest weather temperature happens in January, monthly average temperature is about -20℃, and the highest weather temperature happens in July, the extreme temperature is about 40℃. Because of the hysteresis phenomenon, it can be found that the formation of max frozen depth is some later than the occurrence of lowest temperature from the monitoring data. The lowest temperature of ground surface is-8.3℃in the line of Changchun-Songyuan, the depth frost line is about 1.75m. and the lowest temperature of ground surface is-10℃in the line of the north to Changchun-jilin, the depth frost line is about 1.9m. From the change curve of water content in the freezing and thawing cycle, it can be found that the phenomenon of water translocation happen in four monitoring site. The area of rapid change with water translocation centralizes in the soil layer from 0 to 1 meter.3. In This paper capillary height test is done by Kaminski capillary cube method, direct method, and frozen-thaw method. The capillary heights both are higher by three different methods. The capillary height is 1.0 to 1.6 meter by direct method and frozen-thaw method, the capillary height is 2.56 to 3.45 meter by Kaminski capillary cube method, this soil easily produces all sorts of question on water translocation. The groundwater level depth is about 2~3 m. some parts in Changchun area are affected by environment, and its groundwater level depth is less than 6m. So the seasonal freezing soil of the majority of Changchun has obvious capillary phenomenon, capillary water and adhesive water in unfrozen soil layer are the main body of water translocation.This paper introduces the concept of soil and water driving force from points of energy capacity. The soil and water driving force is defined to the driving force of water translocation. Through the research of influencing factor about seasonal frozen soil in Changchun area, it can be found that the soil particle size and the mineral constituent influence water translocation in seasonal soil obviously. The temperature gradient is the important induction factor, and the soluble salt content, microstructure, and physicochemical property is the accessory factor.4. This paper introduces the theory of Artificial Neural Network to the research in water translocation in frozen soil and establishes the prediction model. Unfrozen water content is an important index to evaluate the character of water translocation. BP neural network is used to establish the relation model of unfrozen water content and its major factors. According to the predictive result, the improved BP algorithm model can achieve a satisfied result in the prediction of unfrozen water content.According to the alimentation condition of ground water, the system can be divided into two parts to research in this paper, one part is the open system, and the other part is the close system. Through the indoor frozen test data, the water translocation case of two systems can be analyzed. And the initial water content, dry density and temperature gradient are determined to the leading factors of water translocation flowing in frozen period. The total amount of water translocation in frozen parts increases with the initial water content and dry density, and it reduces with the increasing of temperature gradient when initial water content and dry density is constant. At the same time, the predictive model of BP neural network is used to analyze the leading influencing factor of water translocation in frozen soil. And nonlinear and uncertainty between the water translocation flowing and its influence factors can be showed well. It's easy to find the method of quantitative assessment and to predict the site case. By the appraisal of experimental findings, the feasibility of improved BP neural network which is used in forecasting water translocation in seasonal frozen soil can be approved.