Simulated annealing technique in discrete fracture network inversion: optimizing the optimization

Simulated annealing (SA) has the capacity to handle complex problem of fracture heterogeneity. However, its applications to characterization and modeling of an actual discrete fracture network are limited. Borrowing the context of geothermal reservoirs (where extensive discrete fractures exist), this paper attempts to solve several key practical issues that persist in current models: objective function’s (OF’s) formulation, modification scheme, and stop criteria. The improvements are examined in a case study on an actual fracture outcrop, where results are compared with a current and advanced SA work.     

An algorithm for mesh simplification applied to fracture hydrology

Numerically generated discrete fracture networks are used to simulate the flow of water through rock fractures. The fractures are modeled as random assemblages of conductive elements in a nonconductive matrix. Because large numbers of fractures often are needed to represent fractured rock adequately, minimizing the required computer time and memory is crucial. For steady-state flow, any portion of the mesh that is linked to the rest of the mesh by only one point is a dead end and does not contribute to water flow. The removal of such dead-end clusters simplifies the mesh and therefore speeds up computation, without changing its response. An algorithm for removing these dead ends is described in detail. Its effectiveness is discussed with regard to the connectivity of a network.     

Parametric study on seismic ground response by finite element modelling

In this paper the results of 2D FE analyses of the seismic ground response of a clayey deposit, performed adopting linear visco-elastic and visco-elasto-plastic constitutive models, are presented. The viscous and linear elastic parameters are selected according to a novel calibration strategy, leading to FE results comparable to those obtained by 1D equivalent-linear visco-elastic frequency-domain analyses. The influence of plasticity on the numerical results is also investigated, with particular reference to the relation between the hysteretic and viscous damping effects. Finally, different boundary conditions, spatial discretisation and time integration parameters are considered and their role on the FE results discussed.     

An automated approach for conditioning discrete fracture network modelling to in situ measurements

Creation of a geological fracture network model conditioned to in situ geometric measurements is of great importance to geoprofessionals, as fractures dominate pathways for fluid flow, a major concern for many engineering applications. This paper introduces and applies the Stochastic Nelder Mead simplex method to automatically calibrate stochastic parameters of geometric characterisations of a discrete fracture network model. This method can overcome the non-convergence of a previous exploratory approach on the classic Nelder Mead method by others, and is an effective substitution to the manual trial-and-error method and is complementary to existing conditional simulation approaches. The procedure to integrate the Stochastic Nelder Mead with a discrete fracture network is presented in detail, and a case study was conducted. Results show that the improved model can better handle the stochastic nature of the underlying system and effectively simulates the observed number and mean trace length of these fractures, although the model results underestimate its standard deviation. Simulated distributions of trace lengths and spacings are within acceptable ranges except for some small offsets, which can be adjusted during model runs.     

A new excavation analysis method for slope design using discrete fracture network based polyhedral modelling

A method to analyse potential excavation geometries, termed excavation isoplethogram, has previously been proposed for use in civil or mining engineering. The results are typically presented in the form of a contour plot identifying orientations of proposed excavation planes that may lead to large volumes (or masses) of unstable blocks and wedges. In this paper, a novel algorithm is proposed to utilise the probabilistic discrete fracture network approach to quantify the risk associated with various slope designs. The method is applied using open cut mine data in order to demonstrate its potential for decision making in mine design and operations.     

地下水污染监测网多目标优化设计模型及进化求解

采用模拟-优化方法建立了一个用于地下水污染监测网设计的多目标优化模型,该模型包括最小化监测费用、污染物质量评估误差、污染羽一阶矩评估误差和二阶矩评估误差等4个目标函数,以充分揭示减少地下水污染监测费用与提高污染监测精度之间的权衡关系.将改进小生境Pareto遗传算法与地下水流模拟程序和污染物运移模拟程序相耦合用于求解地下水污染监测网多目标设计模型.算例研究表明,采用进化算法求解监测网的多目标模型,能真实地反映各个目标函数间的权衡关系,并且不用考虑传统方法中惩罚因子的影响.与单目标优化模型相比,多目标优化模型可在较短的时间内得到优化问题的一系列Pareto权衡解,以利于相应条件下决策者选择最为经济有效的地下水污染监测方案.     

Three-dimensional seismic modelling of crustal structure in the TESZ region based on POLONAISE''97 data

This paper reports the results of 3-D tomographic modelling of crustal structure in the Trans European Suture Zone region (TESZ) of Poland, eastern Germany and Lithuania. The data are the product of a large-scale seismic experiment POLONAISE'97, which was carried out in 1997. This experiment was designed to provide some 3-D coverage. The TESZ forms the boundary between the Precambrian crustal terranes of the East European Craton (EEC) and the younger Phanerozoic terranes to the southwest. The 3-D results generally confirm the earth models derived by earlier 2-D analyses, but also add some important details as well as a 3-D perspective on the structure. The velocity model obtained shows substantial horizontal variations of crustal structure across the study area. Seismic modelling shows low (<6.1 km/s) velocities suggesting the presence of sedimentary rocks down to a depth of about 20 km in the Polish basin. The shape of the basin in the vicinity of the profile P4 shows significant asymmetry. Three-dimensional modelling also allowed tracing of horizontal irregularities of the basin shape as well as variations of the Moho depth not only along profiles, but also between them. The slice between P2 and P4 profiles shows about 10-km variations of the Moho over a 100-km interval. The crustal thickness varies from about 30 km in SW, beneath the Palaeozoic platform, to about 42 km beneath East European Craton in NE. High seismic velocities of about 6.6 km/s were found in the depth range 2–10 km, which coincides with K trzyn anorthosite massif. The results of this 3-D seismic modelling of the POLONAISE'97 data will ultimately be supplemented by inversion of seismic data from previous experiments.     

Performance studies on six heuristic global optimization methods in the location of critical slip surface

Recently, many heuristic global optimization algorithms have evolved with success for treating various types of problems. Majority of these algorithms have not been applied to slope stability problem for which the presence of soft band and convergence problem (discontinuity of the objective function) may create difficulties in the minimization process. In this paper, six heuristic optimization algorithms are applied to some simple and complicated slopes. The effectiveness and efficiency of these algorithms under different cases are evaluated, and it is found that no single method can outperform all the other methods under all cases, as different method has different behavior in different types of problems. For normal cases, the particle swarm method appears to be effective and efficient over various conditions, and this method is recommended to be used. For special cases where the objective function is highly discontinuous, the simulated annealing method appears to be a more stable solution.     

Salt movements in the Northeast German Basin and its relation to major post-Permian tectonic phases—results from 3D structural modelling, backstripping and reflection seismic data

The NW–SE-striking Northeast German Basin (NEGB) forms part of the Southern Permian Basin and contains up to 8 km of Permian to Cenozoic deposits. During its polyphase evolution, mobilization of the Zechstein salt layer resulted in a complex structural configuration with thin-skinned deformation in the basin and thick-skinned deformation at the basin margins. We investigated the role of salt as a decoupling horizon between its substratum and its cover during the Mesozoic deformation by integration of 3D structural modelling, backstripping and seismic interpretation. Our results suggest that periods of Mesozoic salt movement correlate temporally with changes of the regional stress field structures. Post-depositional salt mobilisation was weakest in the area of highest initial salt thickness and thickest overburden. This also indicates that regional tectonics is responsible for the initiation of salt movements rather than stratigraphic density inversion.Salt movement mainly took place in post-Muschelkalk times. The onset of salt diapirism with the formation of N–S-oriented rim synclines in Late Triassic was synchronous with the development of the NNE–SSW-striking Rheinsberg Trough due to regional E–W extension. In the Middle and Late Jurassic, uplift affected the northern part of the basin and may have induced south-directed gravity gliding in the salt layer. In the southern part, deposition continued in the Early Cretaceous. However, rotation of salt rim synclines axes to NW–SE as well as accelerated rim syncline subsidence near the NW–SE-striking Gardelegen Fault at the southern basin margin indicates a change from E–W extension to a tectonic regime favoring the activation of NW–SE-oriented structural elements. During the Late Cretaceous–Earliest Cenozoic, diapirism was associated with regional N–S compression and progressed further north and west. The Mesozoic interval was folded with the formation of WNW-trending salt-cored anticlines parallel to inversion structures and to differentially uplifted blocks. Late Cretaceous–Early Cenozoic compression caused partial inversion of older rim synclines and reverse reactivation of some Late Triassic to Jurassic normal faults in the salt cover. Subsequent uplift and erosion affected the pre-Cenozoic layers in the entire basin. In the Cenozoic, a last phase of salt tectonic deformation was associated with regional subsidence of the basin. Diapirism of the maturest pre-Cenozoic salt structures continued with some Cenozoic rim synclines overstepping older structures. The difference between the structural wavelength of the tighter folded Mesozoic interval and the wider Cenozoic structures indicates different tectonic regimes in Late Cretaceous and Cenozoic.We suggest that horizontal strain propagation in the brittle salt cover was accommodated by viscous flow in the decoupling salt layer and thus salt motion passively balanced Late Triassic extension as well as parts of Late Cretaceous–Early Tertiary compression.     

基坑潜在隆起破裂面的Dijkstra组合优化搜索法

运用弹塑性有限元程序计算基坑围护结构与土体任意一点在开挖前后的应力状态,根据选定的屈服准则计算其极限状态函数值,通过有限元网格节点及其坐标,采用“人机对话”的优化方法选择坑底最大可能的隆起破裂面的起点和终点,通过Dijkstra算法找出基坑土体最可能发生坑底隆起破坏的潜在破裂面的形状及位置,并求出相应的安全系数。大量数值计算和实测资料表明：潜在破裂面终点的位置对围护桩插入比不敏感,而潜在破裂面起点位置则随围护桩入土深度的增加而逐渐靠近围护桩。搜索法得到的潜在破裂面突破了以往人为给定破裂面形状和极限土压力假定的限制,工程算例证实了基于组合优化Dijkstra算法的潜在破裂面搜索法的有效性和实用性。     

江苏沿海地区地面沉降监测网络建设与优化

地面沉降监测是地质灾害防治工作的基础,是维持经济社会可持续发展不可或缺的手段之一.介绍了江苏沿海地区地面沉降多手段融合立体监测网络的发展过程及应用现状,探讨在地面沉降时空演变的当下,各种监测方法的适用性及优缺点,提出了相应的优化建议.可为地面沉降监测网络建设和发展提供借鉴与指导.     

岩体裂隙网络渗流变水温影响分析

考虑岩体裂隙渗流变水温影响,推导单裂隙变水温水流近似解析解和有限元解,在此基础上分别建立裂隙二维网络变水温渗流数值求解方程,分别对应裂隙网络变水温渗流分析的近似解析法和子结构法。分析变温水流运动规律发现：（1）单裂隙内水流水头与水力坡降成非线性关系,当水流由高温区向低温区流动时,水头分布曲线为凸曲线,此时按线性渗流简化水头整体偏小;当由低温区向高温区流动时,水头分布曲线为凹曲线,此时按线性渗流简化水头整体偏大。（2）单裂隙内,高水温处水力坡降小,低水温处水力坡降大;裂隙平均水温越高,流速越快;裂隙网络内存在与裂隙宽度相似的温度偏流效应,即交叉节点水流有偏向水流温度高的裂隙流动的趋势。在温度较高和温度梯度较大的区域,应该考虑水流温度变化对渗流场的影响。     

基于卷积神经网络的福建省区域滑坡灾害预警模型

福建省滑坡灾害频发,开展区域尺度上的滑坡灾害预警是防灾减灾的重要手段,但由于滑坡成灾机理复杂,传统的区域滑坡预警方法存在精度不足等问题。深度学习是指通过构建神经网络模型进行特征的提取、抽象、表示与学习的技术,是机器学习的一种。卷积神经网络作为一种经典的深度学习算法,具有比传统机器学习更强大的分类能力与表征能力。文章以福建省为研究区,将卷积神经网络引入滑坡灾害预警领域,构建福建省区域滑坡预警模型,过程及结果如下：（1）采用SMOTE优化算法对2010—2018年福建省滑坡灾害样本库进行优化,扩充正样本的个数,将正负样本比例从1∶3.4扩充到1∶2,样本总量达到18040个;（2）构建卷积神经网络模型结构,模型结构包括一个输入层、两个卷积层、两个最大池化层和一个全连接层以及一个输出层;（3）使用卷积神经网络对优化后的样本（2010—2018年样本的80%作为训练集）进行训练,并用贝叶斯优化算法优化模型超参数,得到福建省区域滑坡预警模型;（4）以2010—2018年样本的20%作为测试集对模型进行测试,采用混淆矩阵、ROC曲线进行模型测试,结果显示模型准确度为0.96～0.97,AUC值达到0.977,模型精度与泛化能力良好;（5）以2019年汛期滑坡灾害实况作为正样本,通过时空采样的方法采集负样本,构建2019年区域滑坡样本校验集（样本数603个）,对模型进行进一步实况校验,采用混淆矩阵、ROC曲线进行模型校验,结果显示模型准确度为0.75～0.85,AUC值为0.852。虽然仅用了2019年汛期的滑坡实况样本进行校验,但也达到较好的效果。将卷积神经网络算法应用到区域滑坡预警中,为建立区域滑坡预警模型提供了一种新的途径,初步校验表明,模型效果良好,今后将在福建省对模型进行进一步的应用与校验。     

基于生态网络构建与优化的太仓国土空间生态保护修复关键区域诊断与识别

以江苏太仓作为研究区域,基于生态系统服务评估确定生态源地,采用最小累积阻力模型（MCR）和重力模型构建生态廊道,提出生态网络优化模式和对应管控策略。经计算和分析,提取面积30 365.46 hm2的重要生态功能区作为生态源地;识别生态源地161.46 km2,其中水域生态源50个、陆域生态源地47个;提取水域生态廊道53条、陆域生态廊道56条,且水域生态廊道相较于陆域生态廊道更为完善,连通性和稳定性更高;识别需优先保护的生态夹点25个,优先修复的生态障碍点18个。研究成果可为系统性开展国土空间生态保护修复提供借鉴和指导。     

An improved BP neural network based on evaluating and forecasting model of water quality in Second Songhua River of China

Taking the Second Songhua River which lies in the northeast of China as a study area, this paper firstly designs the improved BP neural network water evaluating and forecasting model of which 13 water evaluating items are selected as nodes in input layer; 6 classes of evaluating results are selected as nodes of output layer; then, with the ＂0, 1＂ identified pattern and continually practiced comparisons, ＂13-9-5-6＂ double hidden layers with optimized training structure are confirmed; on the basis of this work, the water quality of the Second Songhua River was evaluated and forecasted at the end. The results showed that in the six classes of predefined water quality in 157 stations, none of them belongs to class Ⅰ, and classes Ⅱ, Ⅲ, Ⅳ and Ⅴ are as follows： 8.91%, 58.59%, 18.47%, 1.91% and 12.1%, respectively; the precision of these evaluating and forecasting results is 82.8%.