基于自组织神经网络的砂土液化评价

在分析自组织特征映射(SOFM)神经网络基本学习算法的基础上,从提高算法收敛速度和性能出发,提出了一种改进算法:根据实际应用并结合专家经验确定初始连接权值;采用高斯函数作为拓扑邻域函数;将算法分为粗调整和细调整两个阶段,分别采用不同的学习率和邻域函数,然后采用改进后的SOFM算法对砂土液化进行评价。实例研究表明,应用SOFM神经网络评价砂土液化高效可行,为砂土液化评价提供了新方法。     

污染场地地下水调查布点及样品采集技术研究

地下水监测井布点和采样,是地下水污染调查的重要部分。由于污染场地的规模较小和其特殊性,很多区域地下水污染普查的方法并不完全适用于污染场地。本文结合地下水的分布及运动特征,从全面性、系统性、代表性、可行性、经济性五方面,阐述污染场地地下水调查布点应遵循的基本原则。同时,通过探讨国内外地下水采样的规范和要求,总结出适用于我国污染场地的地下水采样技术。     

地下水污染的自组织预测模型及其应用

自组织预测模型是前苏联学者Ivakhnenko于 70年代提出的一种非线性建模预测方法 ,它能有效地解决复杂非线性系统的数据处理和建模问题 .应用自组织方法进行地下水污染预测 ,以实际监测数据为基础 ,建立了地下水污染预测的自组织模型 ,其预测精度较高 ,检验误差在 5 .2 %以内     

地下水污染场地污染的控制与修复

我国存在大量的地下水污染场地,给地下水资源的使用带来了严重威胁。将地下水污染场地划分为4大类,15个亚类,为制定不同地下水污染场地的管理、控制和修复规定提供了依据;对地下水污染防治规划的内容和方法技术进行了论述。提出了建立地下水污染的预警系统,为污染的预防奠定基础;介绍了地下水污染的控制与修复技术,并对地下水污染防控和治理的基本原则进行了探讨。     

ASTER数据的自组织神经网络分类研究

传统的遥感数据分类方法大多基于统计学的参数估计，假设数据分布服从高斯正态分布。神经网络方法无需参数估计和统计假设，因而，近来越来越多地应用于遥感数据分类之中。介绍了基于聚类分析的自组织特征映射分类方法。ASTER卫星数据是新型遥感数据，包括 3个15 m分辨率波段和 3个30 m分辨率的短波红外波段。选择北京地区的ASTER数据作为方法实验数据，首先对数据进行了小波融合，然后进行了土地覆盖类型的自组织特征映射神经网络分类研究，把研究结果同最大似然判别法得到的分类结果进行了比较，分类精度比最大似然判别法总体提高了9％。     

基于自组织映射神经网络的水库洪水过程分类

针对传统洪水分类方法中洪水特征提取时存在信息损失和主观性强的问题,本文基于洪水全过程构建自组织映射神经网络(Self-Organizing Map, SOM),综合考虑代表性和拓扑性等评价指标以优选网络规模,实现洪水全过程的拓扑逻辑关系挖掘及分类。以三峡水库洪水过程为研究对象,研究结果表明：(1) 2×3维SOM覆盖率达到56.7%,与3×3维SOM相比,仅有约2%的覆盖率差距,具有良好代表性;2×3维SOM输出层仅有1处翻转,拓扑结构比3×3维SOM更优,更适合三峡水库洪水过程分类。(2) 2×3维SOM将洪水过程划分为6类,其神经元拓扑结构可有效刻画各分类的差异与联系,说明SOM可基于可视化拓扑逻辑关系实现高维洪水数据的可靠客观分类。(3)与传统方法的历史典型洪水分类结果相比,SOM能提供可靠且丰富的分类信息。     

地下水不的自组织预测模型及其应用

自组织预测模型是前苏联学者Ｉｖａｋｈｎｅｎｋｏ于７０年代提出的一种非线性建模预测方法,它能有效地解决复杂非线性系统的数据处理和建模问题,应用自组织方法进行地下水污染预测,以实际监测数据为基础,建立了地下不不污染预测的自组织模型,其预测精度较高,检验误差在５．２％以内。     

多环芳烃类污染场地地下水风险评估参数敏感性分析

以华北平原某多环芳烃类污染场地为例,选取苯并(a)芘为目标污染物,建立了污染场地地下水风险评估模型,采用局部敏感性分析方法研究EFOa(成人室外暴露频率)、BWa(成人平均体重)、DAIRa(成人每日空气呼吸量)、ATca(致癌效应平均时间)、ATnc(非致癌效应平均时间)以及土壤特征参数Pws(土壤含水率)、ρs(土...     

石油类污染场地土壤与地下水污染调查实例分析

以冀中平原某石油类污染场地为例,从污染源分布勘察、场地水文地质模型建立、土壤及地下水的现场调查入手,采用物探、坑探、钻探综合调查技术和定深取样等一些取样方法,对不同深度土壤及地下水的有机污染进行调查和样品分析。结果表明:整个场地的土壤和地下水受到不同程度的污染,30m深度内包气带和饱水带已被污染,50m深度的地下水中有有机污染物检出,石油类场地的污染特征主要表现为土壤及其地下水中含有高浓度单环芳烃和卤代烃。且单环芳烃在土壤与地下水中的浓度高于其它有机污染物。     

Utilization of self-organizing map and fuzzy clustering for site characterization using piezocone data

In this paper soft computing techniques, self-organizing maps and fuzzy clustering techniques have been proposed to isolate different layers in stratified soil based on available cone penetration test results. The results have been compared with that obtained from cone classification chart, hierarchical and K-mean clustering techniques. It was observed that variation in result with self-organizing map (SOM) and fuzzy clustering for isolating soil layers is marginal. These techniques are found to be efficient compared to hierarchical clustering technique. The results of K-mean clustering show that the identified soil strata are similar to that obtained from cone classification chart, SOM and fuzzy clustering.     

基于多元回归分析的连续压实监控技术

现实中的碾压参数难以得到严格控制以满足连续压实控制技术的应用要求,限制了该技术的适用范围和应用效果。为使碾压参数动态变化与工程实际相适应,对碾压参数进行了多元回归分析,并采用克里金插值法对连续压实控制指标的目标值/实测值进行全仓面化估计,提出了以压实合格率作为质量评价指标的连续压实监测模型。开展了现场碾压试验,建立了压实指标与碾压参数的多元回归关系,结合克里金插值法对常规质量检测指标进行了反向预测分析,验证了模型的有效性。工程实例分析表明,针对我国南方地区广泛分布的花岗岩残积土填料,遵循基于碾压参数多元回归的连续压实控制方法,深圳机场T4扩建工程2-1区的常规质量检测均满足设计要求。基于碾压参数多元回归分析的连续压实质量监测能够有效控制工程碾压质量,从而为连续压实控制技术的实际应用提供有益参考。     

Regional debris flow susceptibility analysis based on principal component analysis and self-organizing map: a case study in Southwest China

A method was developed to analyze the susceptibilities of 541 regional basins affected by debris flows at the Wudongde Dam site in southwest China. Determining susceptibility requires information on source material quantity and occurrence frequency. However, the large number of debris flows can hinder the individual field investigation in a each small basin. Factors that may trigger debris flows can be identified using remotely sensed interpretation information. Susceptibility analysis can then be conducted based on these factors. In this study, SPOT5 satellite imagery, digital elevation models (DEM), a lithology distribution map, and rainfall monitoring data were used to identify 12 debris flow trigger factors: basin relief ratio, slope gradient in the initiation zone, drainage density, downslope curvature of the main channel, vegetation coverage, main channel aspect, topographic wetness index, Melton’s ruggedness number, lithology, annual rainfall, form factor, and cross-slope curvature of the transportation zone. Principal component analysis was used to obtain the eight principal components of these factors that contribute to susceptibility results. Then, a self-organizing map method was adopted to analyze the principal components, which resulted in a debris flow susceptibility classification. Field validation of 26 debris flow basins was used to evaluate the errors of the susceptibility classification, as well as assess the causes of such errors. The study found that principle component analysis and self-organizing map methodologies are good predictors of basin susceptibility to debris flows.     

Fractal analysis based on the continuous wavelet transform and lithofacies classification from well-logs data using the self-organizing map neural network

The main goal of this paper is to show that the fractal analysis based on the continuous wavelet transform is not able to improve lithofacies classification using the self-organizing map (SOM) neural network model from well-logs data. The proposed idea consists to inject many inputs in SOM neural network machines and to choose the best map. These inputs are: data set 1: the five raw well-logs data which are: the gamma ray, density, neutron porosity, photoelectric absorption coefficient and sonic well-log; data set 2: the estimated Hölder exponents using the continuous wavelet transform of the data set 1; data set 3: data set 1 and the three radioactive elements concentrations; data set 4: the estimated Hölder exponents of the data set 1 and the Hölder exponents of the radioactive elements concentrations; data set 5: the estimated Hölder exponents of the data set 1 and the three radioactive elements concentrations logs. Application of the proposed idea at two boreholes located in the Algerian Sahara shows that the Hölder exponents estimated with the continuous wavelet transform as an input of the SOM neural network are not able to give geological details. However, the raw well-logs as an input give more details and precision especially when they are enhanced with the natural gamma ray spectrometry data.     

基于多元监测信息融合分析的突水灾害状态判识方法

突水灾变演化过程中伴随着防突结构内的岩土体的信息变化,通过对潜在突水通道围岩多物理场信息进行深度融合分析,揭示了多元信息的逻辑共生关系,建立了突水演化状态判识的理论方法。从岩体渐进破坏突水与充填结构失稳突水两种典型的突水相似模型试验入手,建立了多元监测信息与量化表征函数的关联性。基于主成分分析理论揭示了多参量间的逻辑共生关系,确定了围岩多物理场信息对突水灾害的影响权重,这为突水灾害监测设计与预警提供了理论依据。结合函数拟合曲线中表征函数变化趋势发生转变的极值点与驻点,建立了突水灾害演化状态综合判识方法,最终将两种突水类型的演化阶段划分为：平静期、发展期、突变期和灾后期,并提出了任意时刻下突水灾害发生概率的能量判别方法,该理论方法为隧道及地下工程突水灾害监测预警提供了参考。     

基于自组织特征映射网络-随机森林模型的滑坡易发性评价

为深入探讨评价单元和非滑坡样本选取对滑坡易发性预测的影响,构建了一种基于自组织特征映射网络-随机森林模型的滑坡易发性评价模型。该模型针对栅格单元和斜坡单元在滑坡易发性评价中的不足,结合栅格单元和斜坡单元的相互关系,提出了滑坡易发性指数的优化计算方法。在此基础上,基于随机森林Tree Bagger分类器构建滑坡易发性评价模型,通过对比分析自组织特征映射网络和随机方法选取非滑坡样本对评价结果的影响,探讨自组织特征映射网络、随机森林和自组织特征映射网络-随机森林三种评价模型的有效性;将评价模型应用于大余县滑坡易发性评价。结果显示,随机森林模型和自组织特征映射网络-随机森林模型的预测精度较高,分别达到91.19%和94.94%,成功率曲线的AUC值分别为0.822和0.849,表明自组织特征映射网络-随机森林模型具有更高的预测率和成功率, 自组织特征映射网络聚类的预测精度虽然有限,但作为非滑坡样本的选择方法,能够有效提高随机森林模型的评价精度。     

浅埋大跨隧道施工力学响应模拟与监测分析

以沪－蓉－西庙垭分岔隧道为工程背景,采用三维快速拉格朗日差分法(FLAC3D)对其浅埋大跨段施工开挖力学响应进行了数值模拟,分析了开挖后围岩的应力场和位移场特征,并结合现场围岩变形和支护结构性态的现场监测,研究了典型地质代表断面的施工力学状态,获得一些有益的结论,对实际工程进行直接指导并为同类工程提供参考。     

某地块污染调查及二氯甲烷在土壤和地下水中的分布特征

以某生物医药企业地块为研究对象,结合企业生产历史,识别该地块疑似污染区域以及关注的污染物,通过布设土壤及地下水现场采样点进行现场和实验室测试,研究土壤及地下水中监测因子的含量及分布特征,特别是对二氯甲烷的污染源及其在土壤和地下水中的分布特征进行分析,指出由于地层特征和多种有机污染物的影响,土壤中二氯甲烷的含量未超过国家标准,但在地下水中含量异常高;分析地层对二氯甲烷在土壤和地下水中垂向和侧向迁移的影响,为下一步污染防控提供依据。     

Setting up evaluate indicators for slope control engineering based on spatial clustering analysis

Rainfall-induced landslides (RILs) have been a source of social and economic disruption in the mountainous Baguio area in northern Philippines. Prolonged heavy rainfall usually happens during tropical cyclone and southwest monsoon activity. A pragmatic approach to RIL mitigation is to develop rainfall-based early warning. We implemented a modified regression method to derive the empirical minimum intensity (I)–duration (D) threshold I = 6.46 D ^?0.28 and a normalized ID threshold NI = 0.002 D ^?0.28 for rainfall duration ranging between 24 and 264 h. Using a separate data set to evaluate the applicability of the threshold, 93% of the landslide-triggering rainfall events fell above the derived threshold. RILs also occurred when 24-h rainfall was 0.02–28% of the mean annual precipitation or after accumulating at least 500 mm of rainfall from the onset of the rainy season. The thresholds may be further refined as more landslide data become available in the future.