| Gushing water of tunnel is a common geological phenomena in the tunnel project. The gushing water have an important influence on stability of tunnel project. Different types of rock mass engineering geological conditions and hydro-geological conditions may have totally different results in water gushing. The studies of gushing water in tunnel at home and abroad have accumulated a wealth of experience and substantial results. However, there is still not a mature enough and stable way in prediction of gushing water. Baihe Tunnel is a long tunnel of the Zhuyong highway in Taizhou. By numerous structural geology in history, the tunnel area where geological conditions in larger, the tunnel construction have arisen in the course of collapses, block-fall, deformation, water and mud gushing and other issues. For needs of geological forecasting work in Baihe Tunnel, this paper focused in-depth study on the problem of gushing water in Baihe Tunnel.The author in the field work, take the geological survey and analysis as the major means. Uninterrupted description of the geological analysis and monitoring of the tunnel face were carried out under excavation. At the same time, the external topography geological condition in zone of the tunnel is investigated in detail. This paper takes geological structure cybernetics and rock structure cybernetics as the major guiding ideology. By fully grasping the characteristics of the formation lithology and geological structure of Baihe tunnel, groundwater controlling rock stratum of the export side of the Baihe tunnel was determined, and the location of gushing water of the tunnel was inferred. Against the characteristics of fissures bedrock, on the basis of a large number of joint surveys of discontinuities, a series of indoor use analytical tools was used to deal with it. Respectively, a trace map and structure pole map were drawn. Then, analysis of structural homogeneity of rock mass was made, the attitude distribution of discontinuities in the fractured rock mass and bulk density of discontinuities were solved. After that, methods based on the Monte Carlo principle were used to generate the fissures, and the statistical discontinuities three-dimensional network models were constructed.Three-dimensional network model with considerable confidence can be used as a platform for follow-up study. Parameters reflecting the rock characteristics were distilled from the three-dimensional network model in this paper. Based on graph theory of the correlation matrix, the distribution of length of fracture string in the model was coded to solve it. At the same time, the generation of the shortest path between the given fractures is solved. Visualization technology of three-dimensional network is used for the display of three-dimensional network model.In this paper, domestic and international conventional tunnel inflow forecasting methods were compared against the actual situation of the tunnel.An appropriate or more appropriate method for this tunnel was selected for the gushing water calculation of its stable inflow. According to the needs of working in situ, the author used rectangular and triangular weir as a top-water tunnel gushing measurement tools. By contrasting the gushing water worked out by conventional methods and reality inflow, the conventional methods result in quite a difference results with each other. It means that they are not effective and stable means for prediction and assessment of gushing water in tunnel.This paper takes theory of "Generalized System Science and Engineering Geology," as a guide. It's mentioned that gushing water in tunnel is a comprehensive geological processes related to many factors. And it is a highly nonlinear problems and uncertainties. It is not an effectively way to reveal the fundamental reason of water gushing in tunnel using traditional methods. In a wide range of comparison and analysis of domestic and foreign counterparts, on the basis of owned materials of this tunnel and the geological conditions, this paper use a neural network combined with three-dimensional network in inflow forecast of gushing water. That is, three-dimensional network as an objective portrait of fractured rock mass. In this paper, parameters such as the greatest length of fracture string, the average length of fracture string, the number of the most isolated fracture, were extracted from the three-dimensional network as key parameters of gushing water tunnel. In a detailed analysis on the basis of other factors, a reliable indicator system was formed, and the establishments of neural networks training samples were formed. A neural network toolbox of matlab7.1 was used for the tunnel inflow forecasting. This paper also uses genetic algorithms to optimize the value and bias of neural network, making neural networks in the training areas for convergence faster and the overall situation has improved. In this paper, predicted results response the gushing water after the tunnel excavation, the research results can provide scientific basis for the reference and anti-drainage tunnel engineering design.By in-depth systematically researching of gushing water of Baihe tunnel Yongjia exit fractured bedrock in this paper, some main results and conclusions were made as follows:1 On the basis of wide range collecting of Baihe tunnel project regional geological meteorology and hydrology materials, physical geography and geology conditions of Baihe tunnel was summarized. In some different ways, the engineering geological environmental conditions are investigated and analyzed in details. Also the influencing factor which may influence the characteristics of gushing water in fractured bedrock mass tunnel was discussed. The classification of rock mass along the tunnel line was analyzed too. For the purpose of measuring the total gushing water of the tunnel, a triangular-notch weir and rectangular weir were set up. By measuring the flow in some part in the tunnel, the totally geology conditions and gushing water characteristics are obtained.2 By real-time discontinuities survey in situ, a process called one footage one observation was carried out for tunnel face excavation. After extensive obtainment of fractures in situ and rock mass geology characteristics, indoor work was taken. A 2d trace map and 3d joint pole map of discontinuities was drawn. The discontinuities revealed after excavation whose structural homogeneity of rock mass was analyzed. The advantages grouping of structural homogeneity of rock mass was divided. Then, by Monte Carlo simulation method, according to relevant theories and steps of 3D network modeling, the model of tunnel's discontinuities of 3D network was made. In addition, according to the recycling process of footage in tunnel excavation, a new statistical method called volume method was used for in situ rock mass fracture investigation. The volume method is a more reliable means for bulk density of fractures.3 The analysis of fractures characteristics in 3D space was carried out based on 3D network model, and the specificities of discontinuities implied in 3D network model was solved. The relationship of discontinuities was indicated by graph topological theory and showed by associated matrix. A algorithm was put forward to judged the relationship between different discontinuities, and the generation of discontinuities associated matrix was made by coding. Based on the profound understanding of associated matrix, the global searching and generation of the 3D network relevant discontinuities was achieved. Some useful and important parameters, such as percentage of longest fracture series, average length of fracture series and percentage of isolated fracture which are used to describe the seepage characteristics of rock mass, are obtained from the analysis of relevant discontinuities properties. By DST(deep search traveling) of graph, the shortest path between specified discontinuities is solved.4 The author firmly grasp the geological conditions as the main line, based on the ideology of system and control, from the point of view of geological structure and rock mass structure control, water diversion and prevention of layered tuff to bedrock fissured water are explained and forecasted, and the probability of water gushing of tunnel in forward excavation if inferred by geology analysis. In analysis of geological environment, with the view of geological structure and regional structure, combining some geological underground stress of adjacent tunnel and regional focal mechanical solution, and by contrasting the different between high and low stress character, a conclusion that most part of the tunnel may be a region with low horizontal ground stress was made. This conclusion is very important for further researching, and it is proved by engineering practice.5 Conventional method of calculation and assessment of ground water resources and tunnel water gushing of the domestic and international are analyzed. Application of the methods and their existing problems are compared and analyzed. Some suitable methods are chose for calculation of tunnel's steady inflow after comparative analysis. Different ways caused different result, and the results of a certain gap were mainly due to parameters and conditions arising from the application.6 Based on the factors affecting gushing water, especially factors extracted from three-dimensional model, the neural network forecasting model was established. Using Matlab toolbox for neural networks, by contrasting at home and abroad BP neural network was selected for inflow forecast. The training samples were selected form the tunnel space that has been excavated for some time or with steady gushing water. By comparison of real total gushing water and prediction result, it is found that the prediction result is quite good.7 Using genetic algorithms to optimize the neural network model, the connection weight and bias is the optimizing object. Using the Matlab genetic algorithm toolbox as the optimizing tools, the shortage such as slow network convergence and perplex of local minimum of BP network can be overcome. A better result of gushing water prediction can be obtained using the weigh matrix optimized by genetic algorithm.
|
PDF Full Text Request