环渤海虚拟台网的运行管理

针对天津市＂十一五＂期间的建设成果环渤海虚拟台网,从系统部署、运行监控、信息发布、系统维护等方面介绍了虚拟台网技术在天津地区的应用情况,并利用JAVA图像裁切技术解决了环渤海虚拟台网对天津地区速报地震的乡镇级震中位置显示问题。该台网的建成,不仅提高了天津台网对渤海海域的地震监测能力和研究水平,同时也为天津市各区县地办完善监测预报体系、提升社会显示度,提供了良好的机遇。     

高分辨率非线性储层物性参数反演方法和应用

对于陆相沉积环境下的复杂隐蔽岩性储层,由于观测信息不准确,如信息重叠、信息缺失和噪音污染,以及岩石物理关系模糊等原因,储层横向预测存在不惟一性、不稳定性和不确定性.基于线性假定的常规储层横向预测技术已不适用于复杂隐蔽岩性储层的勘探.本文采用一种非线性储层岩性物性褶积模型,建立波阻抗与孔隙度/泥质含量的函数关系;通过多级结构分解和双向边沿子波检测来刻画复杂岩石物理关系;通过Caianiello褶积神经网络实现确定性反演、统计反演和非线性理论三者有机结合;最后联合应用基于逆算子的反演方法和基于正算子的重建算法实现了综合地质、测井和地震波阻抗信息进行高分辨率储层物性参数反演.非线性储层物性参数反演采用多井约束机制和分频反演方式,在陆相和近海油气勘探资料的实际应用中,取得了明显应用效果.     

饱和砂土液化研究进展

根据国内外文献资料,从三方面总结了饱和砂土的最新进展:饱和砂土液化判别方法、砂土液化的试验研究以及液化后分析,特别是探讨了液化对上部结构的影响。最后指出了存在的问题和今后的研究方向。     

Transient deformation following the 30 January 1997 dike intrusion at Kīlauea volcano,Hawai’i

On 30 January 1997 an intrusion on Kīlauea volcano opened a new fissure within the East Rift Zone (ERZ) at Nāpau Crater, 3 km uprift from the ongoing eruptions at Pu’u ’Ō’ō. The fissure eruption lasted 22 h and opened a 5.1 km long, nearly vertical dike 1.9 m, extending from the surface to a depth of 2.4 km (Owen et al. 2000b). During the eruption, the lava pond at Pu’u ’Ō’ō drained, and eruptions ceased there. Pu’u ’Ō’ō eventually refilled in late February and eruptions resumed there on 28 March 1997. Continuous GPS data show a large transient following the 30 January 1997 dike intrusion. After lengthening 40 cm during the initial eruption, the baseline between two stations spanning the ERZ lengthened an additional 10 cm over the following 6 months. A coastal station KAEP also exhibited transient deformation, as it continued to move southward (5 cm) over the same 6-month period. The baseline between two stations spanning Kīlauea’s summit caldera contracted sharply during the eruption, but gradually recovered to slightly longer than its previous length 2 months after the intrusion. We use the extended network inversion filter (McGuire and Segall 2003) to invert continuous GPS data for volume change of a spherical pressure source under Kīlauea’s summit, opening distribution on a nearly vertical dike in the ERZ and potential slip on a decollement 9 km beneath the south flank. Following the 30 January intrusion, rift extension continued below the initial dike intrusion for the duration of the transient. Decollement slip, regardless of its assumed depth, is not required to fit the data. The modeled transient summit reinflation and rift opening patterns under Nāpau crater coincide with changes in observed behavior of Pu’u ’Ō’ō’s lava pond. Rift opening accelerated while Pu’u ’Ō’ō eruptions paused and began to decelerate after the lava pond reappeared nearly a month after the Nāpau eruption. The transient deformation is interpreted as resulting from shallow accommodation of the new dike volume.     

潜在地震灾害预测和评价神经网络系统及其若干问题

结合人工神经网络自身的特性和地震灾害预测研究的特点,本文应用神经网络模型,建立了潜在地震灾害预测和评价系统。针对网络模型参数设置、数据归一化、中间层神经元最优数目以及泛化分类评价指标等若干实际问题给出了实际可行的解决方案。通过大样本数据对网络的训练,形成了有识别和记忆功能的非线性预测和评价系统。对网络的测试和检验,论证了该系统在预测潜在地震灾害上的可行性和有效性。同时,从测试精度出发,探讨了这种预测网络存在的不足,并给出了相应的改进建议,为开展进一步的研究工作提供了参考。     

Dual models of pore spaces

We present a model for pore spaces that consists of two parts related by duality: (1) a decomposition of an open polyhedral pore space into open contractible pore bodies separated by relatively open interfaces and (2) a pore network that is homotopy equivalent to the pore space. The dual model is unique and free of parameters, but it relies on regularity conditions for the pore space. We show how to approximate any pore space by the interior of a polyhedral complex such that the regularity conditions are fulfilled. Thus, we are able to calculate the dual model from synthetic porous media and images of real porous media. The pore bodies are unions of relatively open Delaunay cells with respect to the corners of the pore boundary, and the pore network consists of certain at most two-dimensional (2D) Voronoi cells with respect to the corners of the pore boundary. The pore network describes the neighborhood relations between the pore bodies. In particular, any relatively open 2D Delaunay face f separating two pore bodies has a unique (relatively open) dual network edge. In our model, f is a pore throat only if it is hit by its dual network edge. Thus, as opposed to widespread intuition, any pore throat is convex, and adjacent pore bodies are not necessarily separated by pore throats. Due to the duality between the pore network and the decomposition of the pore space into pore bodies it is straightforward to store the geometrical properties of the pore bodies [pore throats] as attributes of the dual network vertices [edges]. Such an attributed network is used to perform 2D drainage simulations. The results agree very well with those from a pore-morphology based modeling approach performed directly on the digital image of a porous medium. Contractibility of the pore bodies and homotopy equivalence of the pore space and the pore network is proven using discrete Morse theory and the nerve theorem from combinatorial topology.     

Modelling evaporation using an artificial neural network algorithm

This paper investigates the prediction of Class A pan evaporation using the artificial neural network (ANN) technique. The ANN back propagation algorithm has been evaluated for its applicability for predicting evaporation from minimum climatic data. Four combinations of input data were considered and the resulting values of evaporation were analysed and compared with those of existing models. The results from this study suggest that the neural computing technique could be employed successfully in modelling the evaporation process from the available climatic data set. However, an analysis of the residuals from the ANN models developed revealed that the models showed significant error in predictions during the validation, implying loss of generalization properties of ANN models unless trained carefully. The study indicated that evaporation values could be reasonably estimated using temperature data only through the ANN technique. This would be of much use in instances where data availability is limited. Copyright © 2002 John Wiley & Sons, Ltd.     

基于卷积神经网络的地震波形智能识别研究

随着世界上多个国家和地区的地震预警系统投入运行,误报和漏报等问题逐渐突显,特别是将标定以及强干扰波形误识别为大震事件,快速、精确地区分地震与其他波形是一个难题。针对于此,该研究提出了基于卷积神经网络地震波形智能识别方法。首先收集并处理了2012—2017年中国境内福建以及周边邻省共683个地震和478个爆破事件,并对这些样本筛选、截取和基线校正等预处理,共得到了27 500条三通道波形。在此基础上,构建了3 s波形输入的卷积神经网络模型(SW-CNN)。结果表明：模型对地震、噪声、爆破和异常波形的识别率分别为97.9、99、99.2和99.3%。相比于人工手动分类识别,该模型更省时和更稳定,为地震预警目前所面临的问题提供了一个新的解决方法。     

基于数据分布域变换与贝叶斯神经网络的渗透率预测及不确定性估计

渗透率是储层评价和油气藏开发的关键参数.传统测井方法与常规机器学习方法估算的渗透率都是固定值.但由于测井数据本身存在噪声, 渗透率的预测结果可能受到噪声的影响出现测量性的随机误差(即任意不确定性); 同时, 当测试数据与训练数据存在差异时, 机器学习模型在预测渗透率时可能出现模型参数的不确定性(即认知不确定性).为实现渗透率的准确预测并量化两种不确定性对结果的影响, 本文提出基于数据分布域变换和贝叶斯神经网络同时实现渗透率预测及其不确定性的估计.提出方法主要包括两个部分: 一部分是不同域数据分布的相互转换, 另一部分是基于贝叶斯理论的神经网络渗透率建模预测和不确定性估计.由于贝叶斯神经网络存在数据分布的假设, 当标签的概率分布与网络的分布保持一致时, 贝叶斯神经网络可以更好的学习到数据之间的关系.因此通过寻找一个函数将一个原始域的渗透率标签转换为目标域的与渗透率有关的变量(我们称为目标域渗透率), 使得该变量符合贝叶斯神经网络的分布假设.我们使用贝叶斯神经网络预测目标域渗透率以及任意不确定性和认知不确定性.随后, 通过分布域的逆变换, 我们将目标域渗透率还原回原始域渗透率.应用本文方法到某油田的18口井的测井数据中, 使用16口井的数据进行训练, 2口井进行测试.测试井的预测渗透率与真实渗透率基本一致.同时, 任意不确定性的预测结果提供了渗透率预测值受到的测井数据噪声影响的位置.认知不确定的预测结果说明数据量少的位置具有更高的认知不确定性.我们提出的这一流程不仅显示了在储层表征方面的巨大潜力, 同时可以降低测井解释时的风险.

     

Establishment of a Non-Permanent GPS Network to Monitor the Recent NE-SW Deformation in the Granada Basin (Betic Cordillera,Southern Spain)

The Granada Basin (Central Betic Cordillera), one of the most seismically active areas of the Iberian Peninsula, is currently subjected to NW-SE compression and NE-SW extension. The present day extension is accommodated by normal faults with various orientations but particularly with a NW-SE strike. At the surface, these active NW-SE normal faults are mainly concentrated on the NE part of the Basin. In this part we have selected a 15-km long segment where several active normal faults crop out. Using the marine Tortonian rocks as a reference, we have calculated a minimum extensional rate of 0.15-0.30 mm/year. The observed block rotation, the listric geometry of faults at depth and the distribution of seismicity over the whole Basin, indicate that this rate is a minimum value. In the framework of an interdisciplinary research project a non-permanent GPS-network has been established in the central sector of Betic Cordillera to monitor the crustal deformations. The first two observation campaigns were done in 1999 and 2000.