岩爆预测方法与理论模型研究

发布时间：2018-03-20 16:47

本文选题：地质灾害　切入点：高地应力　出处：《浙江大学》2014年博士论文　论文类型：学位论文

【摘要】：岩爆是高地应力条件下的一种典型工程地质灾害,给地下工程施工人员和设备安全造成严重威胁。随着我国水利、交通和采矿事业的快速发展,高地应力环境中的深部岩体开挖越来越多,岩爆的预防与控制问题将越来越突出,成为深部地下工程地质灾害防治领域的重要课题。岩爆预测是岩爆防控的重要内容。准确的岩爆预测有助于在设计和施工中采取相应的工程对策,减少或避免岩爆灾害带来的损失。但由于岩爆机理复杂,使得岩爆预测十分困难。目前,工程实际中一般采用简单分级的方法对岩爆进行预测,由于不能考虑各种因素的综合影响,其结果往往与实际情况出入较大。针对目前岩爆预测存在的问题,本文主要在以下几方面开展了系统的研究： (1)以苍岭隧道为例,采用传统的强度理论方法对其岩爆进行了系统的预测分析,对已有方法存在的问题进行了探讨。 (2)针对传统强度理论的缺陷,考虑岩爆的特点,采用粒子群算法对广义回归神经网络进行了优化,构建了客观的岩爆预测模型,采用该模型对苍岭隧道、锦屏二级水电站两个深埋地下工程进行了岩爆预测,阐述该方法的特点和局限性。 (3)考虑到岩爆分析数据是连续数据,而岩爆等级是离散数据的特点,结合现场调查结果和国内外工程实例,采用粗糙集理论对岩爆影响因素进行了重要性区分和客观定量评价。 (4)从多目标规划原理出发,结合粗糙集理论分析成果和理想点方法,构建了粗糙集-理想点岩爆预测模型,通过对苍岭隧道和锦屏二级水电站的岩爆预测验证了其正确性和适用性。 (5)从信息融合角度出发,结合粗糙集理论分析成果和理想点方法,构建了粗糙集-理想点岩爆预测模型,同样通过上述两个工程实例对模型进行了验证。 (6)开展粗糙集-理想点法模型、粗糙集-证据理论模型和模糊数学方法模型的对比分析,评价各种方法的优缺点和预测效果。通过上述这些内容的研究,获得了以下一些创新成果： (1)苍岭隧道岩爆预测结果显示,与普通BP神经网络和普通广义回归神经网络相比,粒子群算法-广义回归神经网络模型输出结果稳定,预测结果准确,但该模型在预测锦屏二级水电站探洞岩爆时出现错误,说明其适用性存在一定的局限性。 (2)粗糙集理论分析结果显示应力集中程度对岩爆影响最大,岩体的储能情况影响居中,岩体的脆性条件影响相对较小。 (3)苍岭隧道、锦屏二级水电站探硐的岩爆预测结果显示粗糙集-理想点法模型预测结果正确,并且其预测精度高于层次分析-理想点法模型和等权重-理想点法模型。 (4)苍岭隧道、锦屏二级水电站探硐的岩爆预测结果显示粗糙集-证据理论模型预测结果正确,并且其预测精度高于通过人为指定建立的另外两组证据理论岩爆预测模型。 (5)粗糙集-理想点岩爆预测模型、粗糙集-证据理论岩爆预测模型和模糊数学岩爆预测模型三者总体预测水平相当,但粗糙集-理想点岩爆预测模型和粗糙集-证据理论模型更能反映岩爆发展的趋势,认为两者略优于模糊数学模型。
[Abstract]:Rock burst is a kind of typical ground engineering geological disasters force conditions, causing a serious threat to the safety of construction personnel and equipment in underground engineering. With the rapid development of China's water conservancy, transportation and mining industry, high deep rock excavation force environment more and more, the prevention and control of rock burst will be more and more outstanding, become an important research topic in the field of deep underground engineering geological disaster prevention and control.
Rockburst prediction is an important part of prevention and control of rock burst. Accurate prediction of rockburst can help take relevant engineering measures in design and construction, reduce or avoid the occurrence of rock burst disaster losses. But because of the rock burst mechanism is complex, the rock burst prediction is very difficult. At present, in actual engineering, a method using simple the classification of rock burst prediction, because not considering the influence of various factors, which often results with the actual situation is quite different.
In view of the existing problems of rock burst prediction, this paper has carried out systematic research in the following aspects:
(1) taking the Cang Ling tunnel as an example, using the traditional strength theory method, the rock burst is systematically predicted and analyzed, and the problems existing in the existing methods are discussed.
(2) aiming at the defects of the traditional strength theory, considering the characteristics of rock burst, the particle swarm algorithm to optimize the generalized regression neural network, construct the prediction model of rock tunnel by blasting the objective, the model of Cangling hydropower station, Jinping two two deep underground engineering for rock burst prediction. The paper describes the characteristics and limitations of the method.
(3) considering that the data of rockburst analysis are continuous data, and the classification of rock burst is the characteristics of discrete data. Combined with field survey results and domestic and foreign engineering examples, the importance and objective quantitative evaluation of rock burst factors are made by rough set theory.
(4) based on the principle of multi-objective programming, combined with the results of rough set theory and the ideal point method, a prediction model of rock burst based on rough set and ideal point is constructed. The correctness and applicability of rockburst prediction of Cang Ling tunnel and Jinping two hydropower station is verified by its prediction.
(5) from the perspective of information fusion, combined with the analysis results and ideal points method of rough set theory, a prediction model of rock burst based on rough set and ideal point is constructed, and the above two engineering examples are used to validate the model.
(6) rough set ideal point model, rough set evidence theory model and fuzzy mathematics model are compared and analyzed to evaluate the advantages and disadvantages of various methods and prediction results.
Through the study of these contents, some of the following innovative achievements have been obtained.
(1) the prediction results show that the explosion in Cangling Tunnel Rock, compared with the common BP neural network and generalized regression neural network, generalized regression neural network model output particle swarm algorithm is stable, accurate prediction results, but the model error in the prediction of Jinping two hydropower station tunnel rock burst, that there are some limitations its applicability.
(2) the results of rough set theory show that the degree of stress concentration has the greatest impact on rock burst, and the influence of energy storage on rock mass is in the middle, and the brittle condition of rock mass is relatively small.
(3) the prediction results of rock burst in Cang Ling tunnel and Jinping two hydropower station show that the rough set ideal point method is correct, and the prediction accuracy is higher than the analytic hierarchy process ideal point method and the equal weight ideal point method.
(4) the prediction results of rock burst in Cang Ling tunnel and Jinping two level Hydropower Station show that the rough set evidence theory model is correct, and the prediction accuracy is higher than the other two sets of evidence theory based on artificial designation.
(5) rough set ideal prediction model of explosion point rock, rough set and evidence theory and fuzzy mathematics model of rockburst prediction of rock burst prediction model of the three overall prediction level, but the rough set ideal prediction model and rough set and evidence theory model can reflect the development tendency of rock burst rock burst, think two slightly better than the fuzzy mathematical model.

【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TU45;TU91

【参考文献】