【作者】 肖云华；

【导师】 王清；

【作者基本信息】 吉林大学 ， 地质工程， 2009， 博士

【摘要】 论文以“岩体结构控制论”、“广义系统科学”及数学建模的思想为基本指导思想,从围岩分类角度对隧道围岩稳定性进行研究。论文通过总结研究围岩分类的发展过程,指出现有围岩分类方法中存在几点问题,并针对存在的几点问题,以浙江诸永高速公路双峰隧道为项目依托,对围岩分类方法进行了详细的研究和探讨,从而对隧道围岩稳定性进行分析评价。论文首次详细阐述了BP神经网络和支持向量机在围岩分类中的应用思路问题和模型检验失效性问题,针对相关问题提出了自己的检验思路和具体步骤;首次详细阐述了基于类比法思想的模糊综合评判法、灰色聚类法和可拓评判法在实际应用中的一些问题,详细分析了权重的意义,建立了一个比RMR法精度更高的评价模型,同时结合可靠度理论,为隧道支护设计与施工提供了更加详细的围岩信息。研究结果应用于双峰隧道的围岩稳定性评价获得了可靠的结论。研究成果具有实用和推广价值,同时对工程地质领域的其他分类问题也同样具有重要意义。更多还原

【Abstract】 During the construction and use of road tunnels, the tunnel’s safe is a primary problem. If safety event happen, not only the period delay, cost significantly improve, but also may lead to the technical staff is imperiled, these have bad effects to the society. If these safety events are not handled properly, the quality of the project will be bad and repair work will be more difficultly. In fact, the stability of tunnel suurounding rock is the fundamental problem in tunnel safety. Before excavating, the surrounding rock is in a certain state of equilibrium. And after excacating, the surrounding rock is caused stress redistribution. If the rock strength is high enough, the damage will not happen; but to the low rock strength, damage will happen and the surrounding rock will lose its stability. Because of limitations of human science and technology, it tends to occur in a variety of security events in the process of tunnel construction and use. In addition, the stress on the tunnel overestimate, or underestimate the strength of the rock, it will cause the tunnel design too conservatively, and improve project cost, resulting in unnecessary waste. Therefore, the analysis of tunnel stability is directly related to the evaluation of the success or failure of the tunnel project, and it is of great practical significance and the enormous economic and social benefits. At the same time, studies on stability of surrounding rock will not only rich in theory in tunnel engineering, but also to promote awareness of the rock, similar to other rock engineering significance. Therefore, study on the stability of surrounding rock is very important to academic value and practical significance.In underground engineering, the class of rock mass quality is the scale of the surrounding rock stability. An objective surrounding rock classification can reflect to the basic characteristics of rock mass, is a correct understanding to the suruounding rock, guide the reasonable design and is the basis for project budget. It is very important practical.Base on the scientific research idea of“Rock Mass Structure Control Theory”,“Generalized Syetem Science”and the idea of mathmematical modeling, thesis study the stability of surrounding rock of tunnel from the perspective of rock classification. By summing up the development process of the rock mass classification, thesis points the there are some questions in the existing methods to rock mass classificatio. To the questions for the existence, based on the shuangfeng tunnel in the Zhejiang Zhuyong freeway, thesis detailed study the methods of the rock mass classification and get the following conclusions:1, The development process of the rock mass classification indicates that the study on the rock mass classification is mastering its influencing factors and seting up the model which describes the relationship between the influencing factors and the level of the rock mass.2, By the field investigation and experiment, the author grasps a great deal of geological information and data and analyse influencing factors to the stability of the surrounding rock of Shuangfeng tunnel. The main factors are rock strength, rock mass discontinuity, groundwater and the situation between the strike-dip of discontinuities and the direction of tunnel axis. The quantitative description to the main factors are a uniaxial compressive strength of rock, intactness index of rock mass, the state of the discontinuity, groundwater flow and the angel between the hole axis and the main discontinuity.3, Using the numerical simulation method, thesis discusses detailedly the problem about BP neural network and support vector machine used in the surrounding rock classification. There are a multi-solution characteristic on the modeling of BP neural network and support vector machine in the surrounding rock classification. It’s very necessary to chech the model set up by BP neural network and support vector machine. The existing checking method is completely ineffective.4, For the first time, thesis point out that BP neural network and support vector machine just provide a possibility of finding the realistic model for the surrounding rock classification and that the possibility is closed related to the distribution of the learning samples.5, Thesis points out that it is not the key to just focus on improving learning efficiency and effectiveness, when using BP neural network and support vector machine to solve the surrounding rock classification. And it’s just a misunderstanding. Thesis also points out that it is not reliable to use the learning effectiveness for modeling standard or use the learning effectiveness and testing effectiveness.6, It also points out that the scope of application of the model is the range of the learning samples, rather than the actual problem space.7, Thesis puts forward a basic idea of model checking: model checking should be based on numerical simulation, closely contact the basic theory and experience, using a combination of qualitative and quantitative methods, comparing the simulated results between the existing method and the model in a large number of random samples, and determine the accuracy and reliability of the model. Thesis elaborates the specific steps through an example about Shuangfeng tunnel.8, We can get the same result from using three methods which are the fuzzy comprehensive evaluation method, gray clustering method and the extension evaluation method when each method uses different method of calculating the similarity or different weight. This is cased by the principle of the maximum membership degree. The principle can make these models, which are modeled by these methods, be step functions. So it is not extremely sensitive to the evaluation result when changing the weight or membership function within a certain range in each category. As for the points near the border it is more sensitive and prone to error evaluation.9, With the influencing factors changing, the stability of surrounding rock is variational. It is a continuous variable. The class of the surrounding rock is a measure of the stability of surrounding rock. It loses too much useful information with discrete values evaluating the surrounding rock. Thesis uses the characteristic values of class to evaluate the surrounding rock. So the class of the surrounding rock is continuous and it is more detailed description of the practical and used conveniently.10, Thesis points out the weight should be a variable and reflect influence degree of facts to the evaluation result in the current method of calculating similarity and the various factors that in some cases. Based on this, thesis presents a method which translates the weight problem into an optimization problem to calculate the weight. The method indicates that the weight is a mean with statistically significant in the scope of the samples.If the weight is variational, calculated weight is not the same within different scope of samples. If the scope of the samples is close to a point and samples are reliable enough, then the weight is infuence degree in the current method of calculating similarity and the various factors that in the point.11, Based on extension evaluation method, combined with the put forward a new method of calculating the weight, thesis set up a high-precision method than the RMR of the evaluation model. By calculating, some sections of the shuangfeng tunnel is betweenⅢandⅣ, in accordance with the standard for surrounding rock classification in highway tunne. Based on the evaluation results, the stability of these sections is evaluated, being combined with field investigation.12, Based on the random characteristics of the rock and soil, the reliability theory of geotechnical engineering is introduced into surrounding rock classification. From the classification point of view, the current reliability theory of geotechnical engineering is in fact a probability analysis theory about two classes (reliability and failure).In this paper, the reliability theory to geotechnical engineering is extended to multiclass classification problem. And it is used for surrounding rock classification.The examples shows that the evaluation result is more realistic.13, For the classification problems in other fields, studying on methods of surrounding rock classification for engineering geological and geotechnical engineering also has great significance. And the above conclusions also are suitable for them.更多还原