【作者】 常斌；

【导师】 李宁；

【作者基本信息】 西安理工大学 ， 岩土工程， 2005， 博士

【摘要】 当前众多岩土工程智能化分析系统中普遍存在的根本问题——样本源缺乏、样本代表性不足、样本关系过于离散、专家经验奇缺等,使智能方法在岩土工程中的应用尚属于望梅止渴、画饼充饥的可怜境地。同时,当前岩土工程数值分析中也广泛存在着就事论事,重解决实际个体问题、轻共性规律探索的现象,使数值分析处于辅佐工程经验、细化设计指标的次要、从属地位。 本文针对以上不足,提出了将岩土工程系统数值仿真试验与岩土工程智能化分析方法相结合的思路,以典型数值分析与试验结果作为主样本群的智能化解决方案,推动岩土工程传统设计方法向实验、数值相结合的理性化设计转变。从冻土工程热学参数取值,到冻土通风管路基温度场时域演化过程预测,从岩体力学参数取值,到节理断层硐室围岩稳定性与支护结构强度分析,在4个具体方面对岩土工程智能分析系统的研发框架设计、主样本群的数值仿真试验设计与分析、边界样本群的经验化专家取值、各样本群结构关系的专家准则与构建策略、巨型样本群模型的建模方法、系统集成的具体难点等关键问题,进行了研究探索,独立开发了岩土工程智能分析平台下的四套完整的智能分析系统。 本文主要创新点可简要归纳为以下五个方面: (1) 提出反映冻土物质组成、粒度、级配、孔隙率对导热系数影响的综合等效热学指标。以规范数据及国内外实测值为样本源,构建智能分析模型,开发了能够预测任意场地土导热系数变化规律的“冻土导热系数取值智能分析系统”。 (2) 以基于最新的冻土水热力三场耦合模型和对流换热边界以及考虑拉帘子效果的系统数值仿真试验结果为主样本群,结合冻土工程领域专家建议,构建并行分区大规模神经网络预测模型,开发了面向工程设计人员的能够预测路基温度场连续演化过程的“冻土通风管路基优化设计智能分析系统”。 (3) 指出在岩体地质强度指标和扰动性指标人为判定过程中个人主观因素(感受性因素、情绪因素、气质因素、知识结构因素)的负面影响,引入心理学概念、原更多还原

【Abstract】 The AI method was much more studied in theoretical approach rather than in application in Geotech, due to the shortage of samples, badly distribution of samples, low level of expert experience and so on. On the other hand, the numerical method is widely used in Geotech design and construction. Few of the numerical tests were also introduced to study the theoretical approach for the solution of Geotech problems.The main idea in this paper is to build new improved AI models based on the systematic numerical test results, so that the main samples are come from the sound base with determinate relations, in addition with expert experience.Two typical engineering problems were investigated in this paper with the proposed new approach: the optimization of the ventiduct embankment in the frozen soil foundation, and the stress distribution analysis of the tunnel in the rock mass with a big joint.For the ventiduct optimization designing, two new AI models were established: (1)As a basis of design, the thermal-parameter determining AI model of the frozen soil foundation is built, based on the Chinese code and many experiments data. (2)An optimizing design platform is set up for the ventiduct embankment on frozen soil foundations, based on systematic numerical test results.For tunneling analysis, two new AI models are set up: (1)As a previous study, the determining model of the parameters for the Hoek-Brown Law was investigated with the consideration of the artificial psychology correction for the subjective disturbing error during evaluating the practical jointed rock system. (2)A stress and deformation analysis platform for tunnel with a macro-joint(fault) is then developed and the software is also finished based on systematic numerical test results.Four innovation points has been summarized as follows:1 )The thermal-parameter intelligentized determining approach is proposed by defining a new Integrate Equivalent Thermal Index reflecting the influence of substance composing, granularity, grain proportion and pore ratio to heat exchange coefficient, based on the Chinese code and many experiments data.2) Based on systematic numerical test results aiming at the optimizing of the diameter, space and position of ventiduct tube, and the properties of frozen soil, with the consideration of foundation temperature limitation requirement, a parallel subareas AI forecasting model was built.3) Anew idea is proposed for correction of the artificial disturbance in evaluating the joint rock system, based on the opinion that user’s receptivity, emotion, proclivity and knowledge structure are the artificial subjective factors, which must influence the quantity describe of the geological features. And the solving schemes and math models based on Geological Strength Index and the disturbance degree factor are then proposed to eliminate the subjective disturbing error.4) Based on systematic numerical test results considering the fault position, distance, thickness and the rock properties, overburden, tunnel size and tunnel shapes, a parallel subareas AI forecasting model is built with supplement model in series.更多还原