【作者】 曹杉杉；

【导师】 夏元友；

【作者基本信息】 武汉理工大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 本文结合国土资源部三峡库区专项基金(SXKY4-041)等项目，采用范例推理技术CBR(Case Based Reasoning)结合其他人工智能技术作为边坡稳定性评估的理论基础，针对三峡库区(Three Gorges Reservoir Area)高切坡稳定性评价(Slope Stability Evaluation)，使用Matlab与VC混合编程来开发一个面向对象的边坡智能评估系统(TGSE-CBR)。首先，综述了范例推理技术的研究现状和发展趋势，阐述了范例推理的基本工作原理及工作过程，详细给出了当前应用较广泛的范例表示方法、范例检索方法与范例组织方法，另外简单介绍了目前国内外有一定影响的基于范例推理的系统。接着，针对三峡库区高切坡的特点，采用面向对象的表示方法，构造了高切坡的案例表示方法；建立了基于径向基函数RBF(Radius Based Function)神经网络和基于贝叶斯(Bayes)网络两种边坡范例检索方法，并把粗糙集理论知识引入到范例检索中；引入模拟退火算法优化特征属性权重，应用基于域理论的自适应谐振网络学习方法进行范例学习。最后构建了基于范例推理的边坡稳定性智能评估系统的框架，对边坡系统作了详细的需求分析，论述了系统的设计思想和流程，介绍了系统各个功能模块，系统后台数据库采用关系数据库SQL SERVER2000建立了边坡范例库，检索模块代码在Matlab7．0．4中实现，并在Microsoft Visual C++6.0平台上开发出了该智能评估系统。。论文采用了面向对象方法进行边坡范例表示，运用基于径向基函数的神经网络和贝叶斯网络的两种范例检索方法，并在范例检索过程中引入粗糙集知识，同时结合边坡工程领域特点，建立了相应的边坡范例检索机制，并进行了实例应用，最后分析评估结果表明，该系统能为实际边坡工程提供有效的指导信息，具有十分重要的意义。更多还原

【Abstract】 Combined with some related subjects such as the special fund on Three Gorges reservoir area of the Minister of Land and Resource P.R.C(SXKY4-041), the thesis adopts the Case Based Reasoning technology and other artificial intelligence(AI) technology to form theoretical basis to evaluate slope stability in Three Gorges reservoir Area. It is using Matlab mixed with VC programming to develop one object-oriented (OO)intelligent system, it is called TGSE-CBR System.Firstly, this thesis includes as follows, summarizing the research present status and trend in development of the CBR, expounding the basic operating principle and working process of CBR, giving in detailed case representation and case retrieval method and case fabric method which are used widely at present, introducing simply the CBR System which is appeared both here and abroad.Secondly, this thesis includes as follows: comparing in briefly conventional slope stability evaluation system with the existing popular AI methods, giving their merits and faults, leading-in case retrieval from rough sets(RS) theory adopt object-oriented representation method according to the characters of slope field, meanwhile, introducing some basic conception of RS Theory and its the application in CBR, discussing mainly and in detailed two kinds of slope case retrieval methods based on Radius Based Function and Bayes network, and introducing how to form radial primary neural network and Bayes network, giving the example with Simulate Anneal Arithmetic to optimizing characteristic attribute weights and the example with Field Theory based Adaptive Resonance Theory to learning case.Finally, the frame of the system is constructed and developed in Microsoft Visual C++ 6.0 and Matlab 7.0.4 environments. Meanwhile, it gives the detail of the system’s requirement analysis, design idea and flow, and introduces each functional module. The case representation adopts 00 representation method that is one slope instance representing one case, and the data base is constructed in relational database SQL SERVER2000.This thesis introduces 00 method for case representation, and adopts the RBF network and the Bayes network for case index respectively, while RS theory is applied during the indexing process. According with the characters of slope field, corresponding system is constructed, by applying with the practical slope for stability evaluation, the result shows that the system has very important sense for providing effective instructs in engineering practice.The research is sponsored by the special fund on Three Gorges reservoir area of the Minister of Land and Resource P.R.C (SXKY4-041) and grants on research of major Science and Technology project of the Ministry (104135) .更多还原