大震前地震活动的图象识别

图象识别是近二十年来发展起来的一门学科,它已广泛应用于许多领域中。盖尔芬德(I.M.Gelfand)、普雷斯(F.Press)等人将它用于地震危险区的划分。本文将图象识别方法用于地震预测中,以识别强震发生的时间。 按一定标准将所研究的全部时间划分为危险时间段D和不危险时间段N。以问题表的形式提出大地震前中等地震活动的特性,然后分两步进行图象识别: 1.学习。对P个时间段m个问题的回答是mp的矩阵,回答以二进制(是或非)表示。通过学习,识别出一个、两个或三个问题组合的新特征,称之为D和N的性质。 2.投票。D和N性质数目的差是△,当△大于或等于某阈值时,则识别为危险段D,否则为N。 结果表明,大地震发生前的一定时期内,中等地震活动增至一定水平、相差半级的中等地震活动水平的比值较正常情况增高以及大震前中等地震活动随时间增强等性质的综合,表明未来时间段內可能发生大地震。 此外还作了控制试验,说明图象识别结果是稳定的。      

山西地区强震潜在震源区的图象识别

本文利用山西地区的地震地质、地质构造、地震活动性、地球物理场等特征资料,对有关特征进行了图象识别,采用综合决策方案以提高识别的稳定性。综合判定结果分析表明,在研究区内的180个对象中,得到54个D类对象,占全部识别总面积的30％,其中向31个对象,历史上已发生过6级以上强震或在强震附近,另有23个对象为新判定的6级以上强震潜在震源区。得到山西强震潜在震源的预测图。     

云南及其邻近地区地震活动的模式识别

本文用模式识别方法(CORA—3修改方法)研究了云南及其邻近地区大地震的发生时间与中强地震活动特性的关系。结果表明(不分子集):M≥7.0级大震前1—2年内,5—5.9级地震频度有明显增强。前1年6—6.9级地震相对减弱,前2—4年内6—6.9级地震可增至一定水平。对于CORA—3修改算法在采用“CLUSTERS”算法和不用“CLUSTERS”算法所得识别分类结果基本一致。FH(future history test)试验表明,用1930—1986年为学习时段,1969—1986年为预测时段。得到D类识别率为0.6,N类误识率为0.38。     

基于深度学习的隧道衬砌结构物探地雷达图像自动识别

传统探地雷达图像识别方法中存在识别准确率不高,复杂目标体识别难度大,识别流程较为繁琐,不能实现端到端识别等问题,不能准确的识别实测数据.本文将深度学习中Faster R-CNN、YOLOv3这两种具有代表性的目标检测算法运用到探地雷达的图像识别当中.选择隧道的衬砌结构作为探测识别研究目标,制作了包含钢拱架、钢筋网、施工缝三类结构物标注的实测数据集.从准确率、召回率、平均准确率,准确率-召回率曲线等评价指标,分析了这两种算法在实测数据集上的表现.并对照典型的识别结果,结合这两种算法的原理说明了其运用到探地雷达图像自动识别上的特点.在测试集上,Faster R-CNN在钢拱架、钢筋网、施工缝三类结构物的识别中分别取得了95.5%、90.5%、96.8%平均准确率,YOLOv3则分别取得了90.0%、16.6%、95.3%平均准确率.实验结果表明两种深度学习目标检测方法在隧道衬砌探地雷达图像的识别上取得了良好的效果,其中Faster R-CNN整体效果更好,但会将多次波误识别为有效信号,YOLOv3较少误识别多次波,但是对钢筋网识别效果不好,两种方法搭配使用会形成优势互补.     

云南省各深断裂带上6级以上地震危险区预测

本文应用图象综合识别法预测了云南省八条深断裂带上不同地段的地震危险性。我们把有可能发生6级以上地震的地段称为危险段,不可能发生6级以上地震的地段称为安全段。识别为危险段的有些是历史上曾经记载到6级以上地震的,而其中有些是云南1000多年地震历史记载以来不曾记载到6级以上地震的地区。 同时分析了危险段(D类)和安全段(N类)的某些共同特点。并对本方法进行了讨论。     

致密砂岩气藏动态分类委员会机器测井流体识别方法

致密砂岩流体识别难度大,智能算法能够较好地建立其流体识别模型.相较于单一智能算法,分类委员会机器通过联合多个专家(智能算法)有助于提升智能模型整体性能.而针对分类委员会机器中单个专家性能难以提升的问题,添加门网络构建动态分类委员会机器是一种更有效的模块化学习方式.本研究首先采用门网络将输入数据划分为多个子数据集,然后联合决策树、概率神经网络、贝叶斯分类、BP神经网络、最近邻算法分别训练子数据集得到多个子模型,最后利用组合器最优化子模型组合得到最佳的流体识别模型.针对塔里木盆地库车坳陷大北、克深、博孜地区致密砂岩地层测井数据和测试数据,采用平均影响值法优选敏感测井系列作为输入,构建了动态的测井流体识别模型,其训练、验证准确率分别为96.29%和91.39%.利用此模型以BZ9井为例进行流体类型判别,预测结果与测试结果一致.该方法将无监督与有监督学习相结合,引入门网络提高了数据集利用效率,避免了数据集分布不均衡对模型构建的影响;采用投票机制集成多种专家,建立了子模型与专家的适应关系,流体识别模型预测精度和泛化能力大大提高.     

鄂尔多斯盆地苏东南地区盒8段低阻气层识别

上古生界二叠系下石盒子组盒8段是苏东南地区主力产气层段之一,低阻气层相对较发育,其测井响应表现为低自然伽马、高声波时差、低密度和低电阻率特征.根据常规测井处理方法,对苏东南地区50口试气井106个试气层段进行了电阻率与声波时差参数交会图分析,结果发现低阻气层的电阻率数值分布范围与邻近水层的电阻率基本相当,甚至有些含气段的电阻率比水层的电阻率低,在电阻率与声波时差参数交会图中,含气层的分布区与气水同层分布区相混,无法准确区分.另外,低阻气层在深侧向电阻率和声波时差测井曲线重叠图上基本无幅度差,这就可能造成有效含气层的漏失.针对上述这种电阻率相对较低的气层,利用非电阻率测井信号,优选声波孔隙度(ФS)、密度孔隙度(ФD)、中子孔隙度(ФN)测井资料进行重叠图处理(ФS-ФN、ФD-ФN),并增加纵波等效弹性模量差比值参数来综合识别气层.在含气层段,一般具有ФSФN、ФDФN、等效弹性模量差比值大于0.1的特征.上述综合识别含气层的方法,其判别结果不受电阻率的影响,因此在低阻气层发育的部位能够有效识别气层.经对50口试气井106个试气层段试气资料进行处理,气层识别综合符合率达95.8%.     

2D MRI和3D MRI序列对直肠癌术前分期诊断的价值研究

目的:评估磁共振二维T2加权自旋回波成像序列(2D-T2WI TSE)与三维T2加权自旋回波成像序列(3D-T2WI TSE,VISTA)两种成像方法在直肠癌术前分期的准确性。方法:采用飞利浦3.0T MR机扫描,由两位放射科医师独立评估,回顾性分析从2016年1月至2016年4月间经病理确诊的33例直肠癌患者的2D-MRI序列及3D VISTA扫描序列图像。运用SPSS统计软件分别分析2D序列及3D VISTA序列对直肠癌术前T、N分期的准确性,并检验诊断结果的一致性。结果:33例病例,医师1与医师2其常规2D和3D VISTA检查序列T分期的准确性分别为70%和85%(P=0.274)及73%和82%(P=0.454),Kappa分别为0.576和0.790及0.618和0.746,3D VISTA序列评估T分期与病理结果的一致性优于2D序列;其N分期的准确性分别为58%和70%(P=0.182)及58%和67%(P=0.218),Kappa分别为0.252和0.407及0.267和0.376,N分期2D及3D VISTA分期与病理结果的一致性较差。结论:2D与3D VISTA序列对直肠癌TN分期的准确性无统计学差异,但是3D VISTA较2D序列在鉴别T2与早期T3及N分期方面有一定潜在优势。      

模式识别在强余震等早期综合判别中的应用研究

本文利用所收集到的２９个６级以上地震序列，探讨将模式识别ＣＯＲＡ－３算法用于５级以上强余震或震群或再发生更大地震的前震－主震－余震型地震序列早期综合判断的可能性及具体方法。文中参与学习和训练的２５个地震序列中，８个Ｉ类序列全部被正确识别，１７个ＩＩ类序列中有１６个被正确识别，只有１个被误判为Ｉ类，预报评分Ｖ值达０．９２。对于留作外推的４个地震序列中，２个ＩＩ类序列全部被正确识别，２个Ｉ类序列，１个     

基于深度学习的空间尘埃碰撞实时自动检测

准确快速地检测航天器上发生的尘埃碰撞事件能帮助我们更好地了解空间环境的尘埃分布以及减少航天器因尘埃碰撞受到的破坏.现有人工识别或基于尘埃碰撞引起的电势差信号波形特征的机器识别尘埃碰撞事件的方法虽然有较高精度,但效率低下,迫切需要高精度且自动化方法识别航天器收集的海量电势差信号.深度学习模型在信号分类和识别具有较强能力,本文把空间尘埃碰撞引起的电势差信号检测问题建模成信号分类问题,构建了一个卷积神经网络模型,该模型可以自动提取信号特征并根据特征对信号分类,同时为了训练模型和测试模型预测准确率,构建了一个由尘埃碰撞引起的电势差信号和非尘埃碰撞引起的电势差信号组成的数据集,模型在训练集上准确率为99.46%,在测试集上准确率达到98.68%,查全率为99.44%,查准率为97.95%,threat score为97.41%.实现了高精度且自动化的尘埃碰撞事件检测.     

利用区域地震体波走时评价华北地区三维地壳速度结构模型

本研究利用国家地震台网131个地震台站2009—2016年记录的1 749次近震的初至P波和S波走时数据,与采用快速行进方法正演计算得到的华北地区4个三维地壳速度模型对应的走时数据进行对比,通过统计分析的方法,评价这4个速度模型与真实地下结构的近似程度。结果表明:4个速度模型在大范围内存在较高的一致性,在整个研究区内（111°E—119.5°E,37°N—42°N）,Shen等的模型（简称“S模型”）相对优于Fang等的模型（简称“F模型”）和Duan等的模型（简称“D模型”）,Laske等的Crust1.0模型（简称“C模型”）相对较差。我们认为该结果与上述几个模型所使用的数据及其分辨率有关。对于研究区域内的构造单元,D模型在燕山褶皱带西南部、太行山山前构造带西北部和沧县隆起区表现较好,F模型在太行山隆起区中部、沧县隆起北部、黄骅凹陷区和燕山褶皱带表现较好,S模型在西部地块、山西凹陷区、太行山山前构造带和冀中凹陷区表现较好,C模型无明显连片表现较好区域。      

Preliminary15N studies on atmospheric nitrogenous trace gases

Preliminary nitrogen isotope data for ammonia from animal urine, fuel combustion, fertilizer use and fertilizer factories have been measured or estimated. It turns out that direct nitrogen isotope measurements of atmospheric ammonia at Jülich are in the expected range calculated from the ranges of different sources. For deposition of atmospheric ammonium in Jülich-rain a depletion in¹⁵N with respect to atmospheric ammonia has been found which is explained by isotope fractionations during rainout and washout. In correspondence with this fractionation model are nitrogen isotope data of rain-ammonium from coastal areas, which are enriched in¹⁵N due to the fact that sea water acts as a sink for atmospheric ammonia.For Jülich rain-nitrate a pronounced seasonal trend has been detected with lower¹⁵N data in spring and summer than in autumn and winter. This trend is interpreted by different nitrogen isotope data of anthropogenic and natural nitric oxides which have been measured or estimated from isotope fractionation effects during nitrification and denitrification reactions in soils. It should be possible to get better global estimations for anthropogenic and natural nitric oxides from nitrogen isotope measurements.     

川西—藏东大地电磁资料的阻抗张量畸变分解

根据Swift和Bahr的分解结果,得到准三维/二维模型的区域构造和各个畸变因子, 应用于川西—藏东MT剖面资料的分析,结果表明：该地区的构造特点为浅部比较简单,深部相对复杂;以雅安为界,东部构造比西部构造简单,而且比较稳定;构造主轴方位基本在近南北向,其偏差很少超过30°;分解参数对几条大的断裂有较好的显示;测线1横穿3个主要构造单元,并被几条大断裂切割,构造最复杂;测线2接近3D/2D模型;测线3具有较强的静位移畸变.     

3D soil structure of the Mygdonian basin for site response analysis

The 3D structure of the Mygdonian sedimentary basin (N. Greece) is investigated. The aim of this study is to propose a 3D model of this sedimentary structure that can later be used to model the seismic records currently being obtained by the permanent accelerograph network operating in the area. This model builds on previous efforts and incorporates new data. The geometry and dynamic properties of the soil layers were inverted using data from microtremor array measurements, seismic refraction profiles, boreholes, and geotechnical investigations. Phase-velocity dispersion curves of Rayleigh waves were determined at 27 sites in the basin using the spatial autocorrelation method (SPAC) introduced by Aki [1]. S-wave velocity profiles were inverted from these dispersion curves and the whole valley structure was interpolated using our new results and all previously available data. The proposed 3D model describes the geometry and shear-wave velocities of the Mygdonian and pre-Mygdonian sedimentary systems, and the top bedrock surface. Our results indicate that this 3D model correctly reflects the geometry and dynamic properties of the sedimentary layers. The case of Euroseistest, where the subsoil structure is the result of bringing together many disparate data, could be used as an example for similar alluvial basins throughout the world, where usually only scarce data is available.     

对各个时期钱德勒摆动主峰双频特征的检验

本文用ILS系统地极坐标资料（1900.0-1979.0）,对各个时期钱德勒谱峰的频率特性作了较细致的分析。结果表明:主峰双频只是一种长期的统计特性,从局部看,出现单频和双频的概率约各占50％,因此,还不能说“双频”是钱德勒摆动的基本特征。在物理上,可能用时变性二分模型来描述它更妥当。Q_C.W实测值的平均值为235,众数为125,这与由地震观测得到的结果较为一致。     

Short‐term flood inundation prediction using hydrologic‐hydraulic models forced with downscaled rainfall from global NWP

A short‐term flood inundation prediction model has been formulated based on the combination of the super‐tank model, forced with downscaled rainfall from a global numerical weather prediction model, and a one‐dimensional (1D) hydraulic model. Different statistical methods for downscaled rainfall have been explored, taking into account the availability of historical data. It has been found that the full implementation of a statistical downscaling model considering physically‐based corrections to the numerical weather prediction model output for rainfall prediction performs better compared with an altitudinal correction method. The integration of the super‐tank model into the 1D hydraulic model demonstrates a minimal requirement for the calibration of rainfall–runoff and flood propagation models. Updating the model with antecedent rainfall and regular forecast renewal has enhanced the model's capabilities as a result of the data assimilation processes of the runoff and numerical weather prediction models. The results show that the predicted water levels demonstrate acceptable agreement with those measured by stream gauges and comparable to those reproduced using the actual rainfall. Moreover, the predicted flood inundation depth and extent exhibit reasonably similar tendencies to those observed in the field. However, large uncertainties are observed in the prediction results in lower, flat portions of the river basin where the hydraulic conditions are not properly analysed by the 1D flood propagation model. Copyright © 2013 John Wiley & Sons, Ltd.     

Magnetospheric-ionospheric convection in an open model of the magnetosphere

Magnetospheric-ionospheric convection has been calculated for an open model of the magnetosphere with an ellipsoidal magnetopause in an approximation of uniform IMF. It is assumed that only 0.1 part of IMF falls in the magnetosphere as a result of the effect of IMF shielding by the magnetopause. The modeling of convection has been performed for the cases when the IMF B _z component is directed southward and the B _y component is westward or eastward. A Tsyganenko 96 model has been used as a magnetospheric model. The model calculations are compared with the data on the ion drift in the ionosphere. A certain disagreement between the experimental and calculated data has been found in the pattern of convection on the dayside of the ionosphere for the case of B _y >0, which manifested itself in the dimensions of a convection “tongue” and in the position of the convection throat on the dayside. It has been indicated that the convection pattern agrees with the results of observations if the azimuthally inhomogeneous magnetospheric conductivity is taken into account.     

Modelling the relative impact of rivers (Scheldt/Rhine/Seine) and Western Channel waters on the nutrient and diatoms/Phaeocystis distributions in Belgian waters (Southern North Sea)

The coastal areas of the Southern North Sea (SNS) experience eutrophication problems resulting from freshwater nitrogen (N) and phosphorus (P) inputs from rivers. In particular, massive blooms of Phaeocystis colonies occur in Belgian waters. In this region, water masses result from the mixing of Western Channel (WCH) waters transported through the Straits of Dover with nutrient-rich freshwater from the Scheldt, the Rhine and Meuse, the Seine, the Thames and other smaller rivers. However, the relative contribution of the WCH and each river to the inorganic nutrient pool and the impact on the phytoplankton community structure (diatoms and Phaeocystis) are not known. In order to effectively manage the eutrophication problems, it is necessary to know: (i) the relative contribution of the WCH and of each river impacting the region and (ii) the relative effect of a N and/or P nutrient reduction on the Phaeocystis blooms. To answer these questions, sensitivity tests (1% nutrient reduction) and nutrient reduction scenarios (50% nutrient reduction) have been performed with a three-dimensional (3D) coupled physical–biogeochemical model (MIRO&CO-3D).MIRO&CO-3D results from the coupling of the COHERENS 3D hydrodynamic model with the ecological model MIRO. The model has been set up for the region between 48.5°N, 4°W and 52.5°N, 4.5°E and run to simulate the annual cycle of carbon, inorganic and organic nutrients, phytoplankton (diatoms and Phaeocystis), bacteria and zooplankton (microzooplankton and copepods) in the SNS under realistic forcing (meteorology and river inputs) for the period 1991–2003. The relative contribution of the WCH waters and of the different rivers on the inorganic nutrient pool available for phytoplankton (diatoms and Phaeocystis) growth is assessed by decreasing by 1% the nutrient (dissolved inorganic nitrogen, DIN and inorganic phosphate, PO₄) inputs from the WCH and from, respectively, the Scheldt (and smaller Belgian rivers), the Rhine/Meuse and the Seine (and smaller French rivers) [sensitivity tests]. The relative role of N and P reduction on the diatoms/Phaeocystis distribution is further explored by simulations with 50% reduction of the total (inorganic and organic) N and total P river inputs [nutrient reduction scenarios]. These scenarios allow assessing the impact of the expected 50% reduction of river nutrient inputs resulting from the implementation of nutrient reduction policy.Results of the sensitivity tests suggest that the impact of a 1% reduction of river nutrient inputs on surface nutrients (DIN and PO₄) over the Belgian Exclusive Economic Zone (EEZ) area is similar for the Seine and the Scheldt, which are in turn greater than for the Rhine. However, a hypothetical 1% reduction of nutrient input from the WCH boundary would have a higher impact than for the Scheldt. The impact of nutrient reduction is higher for DIN than for PO₄ whatever the river (contrary to the WCH). DIN is more sensitive to riverine nutrient reduction because the rivers are over enriched in DIN compared to PO₄. The sensitivity tests suggest also that a PO₄ river input reduction would result in a N:P increase and a DIN river input reduction would result in a N:P decrease but that a combined (PO₄ and DIN) input reduction would reduce the N:P ratio at sea.From 50% nutrient reduction scenarios, model results suggest that a total P reduction would induce a significant decrease of diatoms and a small (coast) to negligible (offshore) decrease of Phaeocystis biomass. On the contrary, a total N reduction would induce a significant decrease of Phaeocystis biomass and a moderate increase of diatoms. When N and P river input reductions are combined, the model predicts a significant decrease of Phaeocystis biomass in Belgian waters and a significant decrease of diatom biomass in the coastal waters and a small increase offshore. A future management plan aiming at Phaeocystis reduction should thus prioritise N reduction.     

基于三维Taylor模型的福建及邻近地区地磁模型研究

基于福建及邻近地区高密度的256个地磁场矢量(D、I、H分量)实测数据,结合全球陆地1km基准海拔高度工程的网格数据以及最新的第十二代国际地磁参考场模型(IGRF12),采用三维Taylor多项式模型构建了福建地区的地磁场模型.通过比较均方偏差(RMS)、磁场实际分布以及残差等,结果显示:(1)当三维Taylor模型的截断阶数(N)为2时,可基本反映N为6时的二维Taylor模型的拟合效果;(2)三维Taylor模型计算方便,精度较高,但较易出现龙格现象,因此在确定截断阶数时既要考虑模拟精度,还需要考虑边界效应问题.     

Monitoring subsurface changes over time with cross-well electromagnetic tomographyt1

A fast imaging technique is developed to deduce the spatial conductivity distribution in the earth from low-frequency (> 1 MHz) cross-well electromagnetic measurements. A sinusoidally oscillating, vertically orientated, magnetic dipole employed as a source, and it is assumed that the scattering bodies are azimuthally symmetric about the source dipole axis. The use of this model geometry reduces the 3D vector problem to a more manageable 2D scalar form. Additional efficiency is obtained by using the Born series approximation which is derived from nonlinear integral equations that account for the scattered magnetic fields generated by inhomogeneities embedded in a layered earth. Stabilization of the inversion problem is accomplished through the use of bounding constraints and a regularization method which results in a smooth model that fits the data to the desired noise level. The applicability of cross-well electromagnetics for imaging and monitoring changes caused by subsurface processes has been tested by simulating plumes of conductive fluid with 2D models. The images that result from inverting these synthetic data indicate that the vertical resolution of the method is better than the horizontal, increasing the noise decreases the image resolution, and incorporating a priori knowledge in the form of positivity constraints improves the results. Although higher operating frequencies are usually associated with better resolution, frequencies as low as 100 Hz can produce acceptable images in simulated oilfield environments. The imaging scheme has been applied to data collected during a salt-water injection experiment at the Richmond Field Station test site in Richmond, California. Both the data and the resulting images clearly reveal the presence of the plume and indicate that it is migrating towards the north-northwest rather than spreading symmetrically about the injection well. Applying the imaging code to synthetic data generated by a 3D sheet model verifies the interpretation of these results.