基于专家信息融合法的岩土参数概率分布推断

有限小样本条件下推断岩土参数的概率分布类型,在岩土工程可靠性分析是一个热点问题。在现场有限个小样本和存在多个大样本先验分布的条件下,提出了专家信息融合方法,探讨如何综合利用已有的经验资料并结合工程现场的试验数据来确定岩土参数的概率分布。考虑各先验分布的影响因素,通过专家打分方法,确定各先验分布的权重,进而进行整体融合,实现统计意义上概率分布类型的优化。     

岩土强度参数正态–逆伽马分布的最大后验估计

岩土强度参数的概率分布特征参数是确定强度参数标准值和可靠度分析及风险评估的基础,目前采用现场数据进行估计的方法存在小样本信息量不足的问题,为此,基于贝叶斯统计理论提出服从正态分布的岩土强度参数的概率特征参数服从于一个二维联合先验分布,并根据贝叶斯公式推导相应的共轭后验分布函数,以及岩土强度参数概率特征参数的最大后验估计量计算公式。以重庆万州区域内若干工程的泥岩和砂岩的黏聚力、内摩擦角的历史数据为例,建立先验分布函数,先统计单个工程的现场样本均值和方差,然后将若干工程的均值和方差组成新的样本,以此样本为基础采用参数估计得到理论推导确定的先验分布中的超参数,从而确定该区域泥岩和砂岩黏聚力及内摩擦角的先验分布函数,结合该区域内一个工程算例的现场数据,得到该工程在现场样本情况下泥岩和砂岩黏聚力及内摩擦角的概率分布特征参数的后验分布函数和最大后验估计值,并计算相应的黏聚力及内摩擦角的标准值,与传统方法确定的标准值进行比较,表明提出的贝叶斯方法综合了历史数据和现场数据的信息,更为科学合理。     

某滑坡体岩土参数概率分布统计分析方法研究

为了给某坝址比选提供滑坡体的岩土参数概率分布类型,解决小样本条件下岩土参数概率分布难以确定的问题,通过勘察试验数据分析,工程类比分析等方法确定了滑坡体可靠性计算参数的分布类型及参量,以此为先验函数,并基于Bayes统计理论对滑坡体岩土参数进行优化研究,以其中的“上田”滑坡体为例对岩土参数进行优化计算,计算结果表明,优化后方差和变异系数均有下降,可以提高滑坡防治工程的可靠性,从而达到优化的目的。     

基于第二类切比雪夫多项式的岩土参数概率分布推断

基于数值分析中的逼近原理,提出了描述岩土参数概率分布的第二类切比雪夫正交多项式逼近法。直接根据试验样本值,运用第二类切比雪夫正交多项式来拟合岩土参数的概率密度函数。研究结果表明,采用该方法所得到概率密度曲线非常接近实际统计频率分布,为较大样本条件下描述岩土参数的概率分布提供了一条新途径。     

岩土参数概率分布的系统推断研究

借助数据挖掘技术,将岩土参数测试样本概率分布的选择系统化。选取了19组岩土参数测试样本,选择beta分布和正态信息扩散法作为推断岩土参数概率密度的两种方法。借助数据挖掘技术中的模糊C-均值聚类方法,根据岩土参数测试样本的数字特征值,将19组岩土参数测试样本归聚为4类,再根据每一类样本的分布规律,确定其概率密度函数的推断方法。对于波动性和离散性严重的测试样本,在采用正态信息扩散法推断其概率密度函数时,为解决窗宽选择问题,提出以均方差最小为目标,在窗宽的基础上,改变窗宽的大小,确定样本最优窗宽的方法。对于一组新的测试样本,根据数字特征值即可推断出测试样本的概率分布。     

岩土参数最优概率分布推断方法及判别准则的研究

 为了获得岩土参数概率分布的最佳推断方法,首先考虑岩土参数均为非负值的特性,提出以“3?”准则为基础并考虑样本数据偏度进行调整的积分区间确定方法;以5组典型岩土参数作为基本信息,利用经典分布拟合法、最大熵法、一般多项式逼近法、正交多项式逼近法和正态信息扩散法分别对其概率分布函数进行了推断,并根据K-S检验法进行检验。通过所得概率分布的检验值、累积概率值和函数曲线的对比,研究上述方法的优劣。结果表明：与经典分布拟合法相比,其他4种方法的检验值普遍较小,均克服了经典分布无法反应样本随机波动性的缺陷,并且满足累积概率值等于1的要求。但最大熵法的检验值存在大于经典分布检验值的情况,一般多项式和正交多项式方法的概率密度函数则在样本数据局部分布区间存在负值情况。正态信息扩散法不存在上述缺陷,而且该方法得到的检验值最小,累积概率值始终为1,并可以随着样本的波动呈现多峰状态,拟合精度最高,是一种比较理想的最佳推断方法。最后给出了岩土参数最优概率分布的判别准则。     

基于岩土参数统计分析的地层划分研究

地层划分与岩土参数统计分析是岩土工程勘察工作的重要内容,划分的精度和合理性直接影响到对场地土层的岩土工程条件的分析与评价。传统的岩土工程勘察中一般结合地质成因、沉积规律及岩土参数等方面,先进行初步划分,随后结合物理力学参数的统计情况,不断调整分层,最后划分出合理的地层层序。本文结合北京市大望京地区多个工程实例,提出在工程经验比较丰富的地区,开展区域性的岩土参数统计分析工作,初步确定了大望京地区地层层序情况,对各主要地层进行重点岩土参数的统计分析工作,以标准差及变异系数为控制指标,优化了地层层序划分,提出了各主要地层的分布标高、统计参数范围,对今后该地区岩土工程勘察工作具有借鉴意义。     

有限数据条件下空间变异岩土力学参数随机反演分析及比较

现场和室内试验数据通常非常有限,基于有限数据难以确定岩土力学参数统计特征。随机反演方法可为有限数据条件下岩土力学参数统计特征及概率分布推断提供一条有效的途径。发展3种可解释岩土力学参数空间变异性的参数随机反演方法(DREAM_((zs))、BUS和aBUS),并从随机样本产生方式、收敛判据、模型证据和后验失效概率计算等方面对这3种方法的基本原理进行比较。最后,通过2个边坡案例对这3种方法的收敛性、计算精度和效率等进行了系统比较,进而总结了针对不同岩土力学参数随机反演问题应优先推荐使用的方法。结果表明:DREAM_((zs))方法对于低维问题计算精度和效率高。BUS方法在子集模拟运算之前需要提前确定似然函数乘子取值,适合分析考虑参数空间变异性和似然函数计算量较大的高维问题。aBUS方法不依赖于似然函数乘子取值,并且计算精度较高,适合求解考虑参数空间变异性和似然函数计算量相对较小的高维问题,但是该方法需要耗费一定的计算量来定量地判断计算是否收敛。     

上海奉贤区砂质粉土压缩模量的概率分布

收集、整理了上海奉贤区10个工程场地的岩土工程勘察报告,对砂质粉土层的压缩模量做了统计,并利用2χ比较法,建立了概率分布模型,并得出Beta分布是最优概率分布。     

建筑工程中渗透系数确定方法的讨论

岩土工程中地下水的渗透系数K值是施工计算中的重要参数。在此简述岩土渗透系数的各种数学表达式,并通过室内、现场各种渗透试验来确定K值的方法。以建筑地基基础工程为例,着重讨论实际工程中的确定途径和方法问题。并依据概率统计理论,结合工程勘察实际情况,对小样本容量条件下K的取值方法进行了分析并提出建议。     

随机加权法辅助小样本岩土参数分布拟合

提出小样本条件下基于随机加权法的多项式拟合方法,对岩土参数总体分布密度函数进行推断。应用随机加权法重采样技术对小样本数据信息进行提取,估计总体各阶原点矩。采用勒让德正交多项式拟合推断分布密度函数并采用精度较高的K-S检验法进行检验,确认本方法具有较高的精度。本方法直接根据样本数据进行推断,大大降低了对已有数据及经典分布概型的依赖性,具有明确的数学物理意义,能够满足岩土工程可靠性分析的需要。     

Incorporating installation effects into the probability analysis of controlled modulus columns

This technical report presents the probabilistic analysis which integrates the Monte Carlo simulation (MCS) with random field theory to model the load–displacement behavior of Controlled Modulus Columns (CMCs) in overconsolidated Poznań clay. Presented study focuses on the practical aspects of statistical analysis of geotechnical data, numerical model development, and results evaluation. Variability and spatial distribution of geotechnical parameters are based directly on field and lab testing. The inherent variability of soil parameters obtained from geotechnical investigation at the site is similar to the values reported in worldwide datasets for clays. The extensive discussion about incorporation of installation effects into numerical modelling is made. It was found that proper incorporation of installation effects is governed by correct estimation of initial stress level and interface shear strength parameters. The Anisotropic Undrained Shear Strength (AUS) model which captures nonlinear behavior and anisotropy of soil (Krabbenhøft et al., 2019) is a good choice to model overconsolidated clay in intact and interface zones. The application of total stress approach, the AUS model, installation effects, and natural (inherent) variability of soil and interface parameters is sufficient to explain differences in CMC load – displacement behavior observed in the field.     

A risk-based model for performance-based regulation of electric distribution companies

In this study, a model is presented to obtain the parameters of penalty and reward scheme (PRS) in performance-based regulation (PBR) for each electricity distribution company (EDC) using analytical hierarchy process (AHP) and fuzzy c-means clustering (FCM). In the FCM algorithm, similar companies were categorized into clusters. By using AHP, score of effective factor in reliability index was obtained. In this model, external factors affecting EDCs performance were considered to reduce the risk of PBR implementation for companies and customers. The proposed model was applied on the EDCs in Iran. The results, including AHP score, parameters of PRS and PRS cost were calculated.     

软土地区大型干坞边坡变形规律离心模型试验及实测研究

对上海外环隧道干坞边坡的开挖方法及变形规律等进行了离心模型试验研究 ,并对边坡变形监测数据进行了整理和分析 ,本文研究可对以后类似的工程提供借鉴和指导。     

考虑隶属函数特性的边坡模糊可靠性分析

针对边坡工程岩土参数所具有的随机性、模糊性与区间性,根据隶属函数分布特征与截集技术提出岩土参数隶属函数截集取值方案,构建隶属函数不同分布形式的表征方法及其取值界限确定方法,建立基于截集技术与模糊点估计法的边坡模糊可靠性分析方法。通过两个典型算例不同工况计算结果的系统分析获得了岩土参数隶属函数分布形式、截集取值方案与取值界限对边坡模糊可靠度的影响规律。岩土参数样本数据量较少时可优先采用取值界限为±2.5σ的三角形分布与正态分布,对数正态与拟正态分布次之,而Beta(4,3)与Beta(4,4)分布将使模糊可靠度计算结果偏大。界限截集水平对计算结果影响程度显著,三角形、拟正态与Beta(4,3)分布的截集区间可取[0,1.0],正态与对数正态分布的截集区间可取[0.05,0.95],并选取隶属函数界限值与0.1间隔的截集水平值进行计算。获得了考虑隶属函数特性的边坡模糊可靠性分析方法,其将使边坡模糊可靠度计算结果更具合理性。     

Rapid susceptibility mapping of earthquake-triggered slope geohazards in Lushan County by combining remote sensing with the AHP model developed for the Wenchuan earthquake

A rapid-response mapping model can be used to study the susceptibility of areas of interest to geohazards (which are commonly regarded as among the most damaging natural hazards), assuming that the model is stable (i.e., that it is generally applicable to any such area). Applying a predefined predictive geohazard-susceptibility model to an area with geoenvironmental conditions similar to those of the area for which it was originally formulated is an effective method of testing the stability of the model. In this paper, the analytic hierarchy process (AHP)-based model developed for the Wenchuan earthquake was used to study susceptibility to earthquake-triggered slope geohazards in Lushan County. Upon integrating the results of a literature review, site investigation, and remote sensing interpretation, seven main factors that influence earthquake-triggered slope geohazards were identified, including peak ground acceleration, distance from a stream, distance from a highway, slope gradient, slope position, normalized difference vegetation index, and micro-landform. In order to reduce the subjectivity of the expert evaluation method usually applied in the AHP, these factors were ranked by relative importance using regression analysis. The weight of each factor was then calculated by the AHP. The susceptibility mapping model was obtained on the ArcGIS platform, utilizing map overlaying. Finally, the results were re-classified to obtain a map of slope geohazard susceptibility. The accuracy of the AHP model was evaluated using both qualitative and quantitative methods. In the qualitative method, the modeled distribution of susceptibility was compared with the actually distribution of geohazards in the study area (identified through remote sensing interpretation), and the areas with high and very high geohazard susceptibilities in the model were found to match well with the actual locations of slope geohazards. In the quantitative method, statistical data showed that over 66% of the geohazards were located in areas of high or very high susceptibility according to the model, while only about 16% were located in areas of very low or low susceptibility, and the density of slope geohazards was about 125 times greater in the areas with very high susceptibility than in the areas with very low susceptibility. Also, the AUC value of the ROC curve for the model suggested that it has high predictive power (a predictive accuracy of 84.8%). In conclusion, it was possible to make accurate predictions about the slope geohazards in earthquake-prone areas located in mountainous regions based on geospatial data, and a high correlation between the susceptibility map generated by the AHP-based model and the true distribution of slope geohazards was observed. Therefore, the AHP-based model used here could be applied to map the slope geohazard susceptibility in other mountainous regions which may be prone to slope geohazards during earthquakes.