地理信息可视化新技术综述与分析

随着技术的发展,新的地理信息可视化技术不断涌现。本文对目前地理信息可视化技术研究中较热点的四种地理信息可视化新技术进行了介绍,并分析了各技术的不足,希望能对相关的研究有所裨益。     

基于LSTM神经网络的青藏高原月降水量预测

青藏高原的降水量预测不仅为该地区水资源合理规划利用提供依据,同时对中国及周边国家气候变化研究有着重要的意义。论文利用1990—2016年青藏高原降水量数据,采用长短期记忆神经网络（LSTM）对青藏高原月降水量进行预测,主要包括：① 使用青藏高原86个测站1990—2013年的月降水资料,预测各个测站2014—2016年的月降水量,并与传统的RNN、NAR、SSA和ARIMA预测模型相比,平均决定系数R²分别提高了0.07、0.15、0.13和0.36,均方根误差（RMSE）和平均绝对误差（MAE）表现更低;② 分析了降水量预测精度的空间分布特征,将各模型的R²在青藏高原地区内插值,分析R²的空间分布特征,发现所有模型降雨稀少的干旱地区和降雨多的湿润地区R²较低,在气候稳定、降水规律性明显的地区R²较高,且LSTM模型R²≥0.6的空间范围远大于传统模型;③ 分析了不同预测长度对各模型预测精度的影响,发现所有模型会随着预测长度增加而预测精度降低,但在不同的预测长度下LSTM预测的RMSE值都低于其他模型。     

Simulation of hydraulic fracture propagation near a natural fracture using virtual multidimensional internal bonds

A virtual multidimensional internal bond (VMIB) model developed to simulate the propagation of hydraulic fractures using the finite‐element method is formulated within the framework of the virtual internal bond theory (VIB) that considers a solid as randomized material particles in the micro scale, and derives the macro constitutive relation from the cohesive law between the material particles with an implicit fracture criterion. Hydraulic pressure is applied using a new scheme that enables simulation of hydraulically driven cracks. When the model is applied to study hydraulic fracture propagation in the presence of a natural fracture, the results show the method to be very effective. It shows that although the in situ stress ratio is the dominant factor governing the propagation direction, a natural fault can also strongly influence the hydraulic fracture behavior. This influence is conditioned by the shear stiffness of the fault and the distance to the original hydraulic fracture. The model results show that when the fault is strong in shear, its impact on hydraulic fracture trajectory is weak and the hydraulic fracture will likely penetrate the fault. For a weak fault, however, the fracture tends to be arrested at the natural fault. The distance between the fault and the hydraulic fracture is also important; the fault influence increases with decreasing distance. The VMIB does not require selection of a fracture criterion and remeshing when the fracture propagates. Therefore, it is advantageous for modeling fracture initiation and propagation in naturally fractured rock. Copyright © 2010 John Wiley & Sons, Ltd.     

A virtual experiment technique on the elementary behaviour of granular materials with discrete element method

Multi‐scale investigations aided by the discrete element method (DEM) play a vital role for current state‐of‐the‐art research on the elementary behaviour of granular materials. Similar to laboratory tests, there are three important aspects to be considered carefully, which are the proper stress/strain definition and measurement, the application of target loading paths and the designed experiment setup, to be addressed in the present paper. Considering the volume sensitive characteristics of granular materials, in the proposed technique, the deformation of the tested specimen is controlled and measured by deformation gradient tensor involving both the undeformed configuration and the current configuration. Definitions of Biot strain and Cauchy stress are adopted. The expressions of them in terms of contact forces and particle displacements, respectively, are derived. The boundary of the tested specimen consists of rigid massless planar units. It is suggested that the representative element uses a convex polyhedral (polygonal) shape to minimize possible boundary arching effects. General loading paths are described by directly specifying the changes in the stress/strain invariants or directions. Loading can be applied in the strain‐controlled mode by specifying the translations and rotations of the boundary units, or in the stress‐controlled mode by using a servo‐control mechanism, or in the combination of the two methods to realize mixed boundary conditions. Taking the simulation results as the natural consequences originated from a complex system, virtual experiments provide particle‐scale information database to conduct multi‐scale investigations for better understanding in granular material behaviours and possible development of the constitutive theories provided the qualitative similarity between the simulation results from virtual experiments and observations on real material behaviour. Copyright © 2011 John Wiley & Sons, Ltd.     

虚拟企业的组织结构设计

作为2l世纪主流企业组织形态的虚拟企业，如何组织设计虚拟企业，无论在理论研究还是在实际应用中都是一个非常重要的问题。从实际出发，介绍了虚拟企业的组织机制，完成了虚拟企业组织结构的详细设计。力求能够为虚拟企业组织的建立探索一种可行的思路。     

边界虚拟钻孔在复杂地质体3维建模中的引入与确定

在矿山复杂地质体3维建模中,由于数据缺乏致使模型难以控制地质体空间边界特征,直接影响到模型的准确性和实用价值。借助边界虚拟钻孔方法来增加复杂地质体边界数据点密度,从而有效控制3维形体的边界,在一定程度上能缓解这一问题。以地层受断层破坏为例,分别就地层与断层4种典型空间分布组合特征下边界虚拟钻孔的确定方法进行研究,提出相应解决方案。边界虚拟钻孔和普通钻孔在数据存储方式、数据结构和表达方式等方面具有一致性,可用类似的处理方式,因而可以加快3维地质建模,提高建模精度。     

一种基于流形正则化的半监督指纹定位算法

针对传统指纹定位算法采集带标签训练数据成本高的问题,本文提出了一种基于流形正则化的半监督指纹定位算法。首先以流形假设为依据,利用批量输入的带标签数据与无标签数据之间的相似度构建图拉普拉斯算子;然后与极限学习机算法相结合,通过随机特征映射建立隐含层;最后在流形正则化框架下,求解隐含层和输出层之间的权值矩阵,从而建立位置估计模型。仿真结果表明,与INN、SVR、ELM 3种算法相比,该算法的训练和测试时间相对较短,且在带标签训练数据稀疏的前提下仍能保持较高的准确率与稳定性。     

红外卫星云图和相关向量机的有眼热带气旋客观定强模型

热带气旋TC(Tropical Cyclone)是全球影响最严重的自然灾害之一。TC强度和路径的准确预报,对于减轻其带来的灾害影响至关重要。本文基于静止红外卫星云图和相关向量机RVM(Relevance Vector Machine)构建有眼TC客观定强模型。首先,利用高斯平滑对红外卫星云图进行去噪;然后,利用基于测地活动轮廓GAC(Geodesic Active Contour)模型的偏微分方程PDE(Partial Differential Equation)法对有眼TC的眼壁进行分割,提取眼壁的亮温梯度信息,计算眼壁亮温梯度的最大值及梯度数据不同概率时的均值,从而构造与TC强度密切相关的特征因子;最后,利用RVM构建单特征因子、多特征因子与近地面最大中心风速的客观定强模型,研究不同特征维度对TC客观定强误差的影响。实验结果表明,在单特征因子的模型定强中,95%概率眼壁亮温梯度均值的定强误差最小,相比利用单特征因子所构建的定强模型,多特征因子的模型定强误差更小,即多特征因子中包含更多与TC强度相关的特征信息。在多特征因子的模型定强中,二特征因子优于三特征因子模型,说明应当合理选择特征因子维数,并非越多越好。本文所用RVM模型具有良好的高维非线性处理能力,能对TC强度进行有效估计。