Physics informed machine learning: Seismic wave equation

Similar to many fields of sciences, recent deep learning advances have been applied extensively in geosciences for both small- and large-scale problems. However, the necessity of using large training data and the ‘black box’ nature of learning have limited them in practice and difficult to interpret. Furthermore, including the governing equations and physical facts in such methods is also another challenge, which entails either ignoring the physics or simplifying them using unrealistic data. To address such issues, physics informed machine learning methods have been developed which can integrate the governing physics law into the learning process. In this work, a 1-dimensional (1D) time-dependent seismic wave equation is considered and solved using two methods, namely Gaussian process (GP) and physics informed neural networks. We show that these meshless methods are trained by smaller amount of data and can predict the solution of the equation with even high accuracy. They are also capable of inverting any parameter involved in the governing equation such as wave velocity in our case. Results show that the GP can predict the solution of the seismic wave equation with a lower level of error, while our developed neural network is more accurate for velocity (P- and S-wave) and density inversion.     

基于深度学习的循环水养殖鳗鲡(Anguilla)行为状态识别的研究

对鱼类的行为进行智能监测,精准地量化与识别其健康状态,已成为研究热点。为实现养殖鳗鲡行为状态精准识别,提出一种基于DenseNet双流卷积神经网络的鳗鲡行为状态检测方法。利用混合高斯背景模型进行前景提取构建数据集,针对传统卷积神经网络对于时间动态信息提取能力有限的问题,搭建关联空间特征与时间特征的双流网络结构(Two-stream),并使用DenseNet-121网络替换原网络,对比VGGNet、ResNet等网络,通过密集连接实现特征重用,在搭建更深的网络结构基础上加强了运动特征传递并减少了参数量,更好地提取具有代表性的行为特征。传统双流网络在两端的softmax层仅作简单的决策层平均融合,无法更深程度关联时空高级特征,提出在网络卷积层提取空间特征与时间特征后,加上一层卷积层将时空特征进行卷积融合以提升模型识别精度。实验结果表明:文中提出的基于DenseNet双流卷积神经网络对6种鳗鲡行为状态检测方法准确率达到96.8%,相较于单通道的空间流与时间流网络,准确率分别提升了10.1%和9.5%;相较于以VGGNet、ResNet搭建的双流网络,准确率分别提升了12.4%和4.2%;与决...     

���ô�ֱ�����ݶ��쳣���ݺ��׵���

??????????????????????????????????????????????????????洹??????????????????????????????????????????????????????????????????????????ε?????о????о?????????Ч????????????????????????????????????????????????????????????????????????     

Modeling regional and local-scale permafrost distribution in Qinghai-Tibet Plateau using equivalent-elevation method

This study proposes an equivalent-elevation method to evaluate the integrated effects of latitude and elevation on regional and local-scale permafrost distribution in the Qinghai-Tibet Plateau,and to model the general permafrost-distribution patterns in regional and local-scale area.It is found that the Gaussian curve―an empirical model describing the relation between variations of altitudinal permafrost lower limit (PLL) and latitude in the Northern Hemisphere―could be applied in regional-and local-scale areas in the Qinghai-Tibet Plateau in a latitude-sensitive interval of 30°-50°N.The curve was then used to evaluate the latitudinal effect on permafrost distribution through transforming the latitudinal effect into a kind of altitudinal difference of PLL.This study then calculated the local equivalent-elevation value by overlaying the altitudinal difference of PLL onto real elevation at a certain location.The equivalent-elevation method was verified in an experimental subwatershed of the Qinghai-Tibet Plateau.However,feasibility of the method should be further tested in order to extend for future studies.The use of equivalent-elevation values can build a platform for comparing the regional general permafrost distribution in the plateau,and for basing further evaluations of local factors’ effects on regional permafrost distribution.     

加载速率对钢筋混凝土梁动态性能的影响

引入钢筋和混凝土的应变率效应,利用有限元分析软件ABAQUS中的显示动力分析模块ABAQUA/Explicit,对钢筋混凝土梁在不同单调加载速率下的受力性能进行了数值模拟。通过比较不同加载速率下的分析结果,探讨了地震作用下的动态荷载对钢筋混凝土粱受力性能的影响。分析结果表明,在动态荷载作用下,钢筋混凝土梁的受力性能有显著变化,其承栽力随加载速率的提高而提高,且极限强度的提高要较屈服强度明显。     

基于改进的交叉极化比率算法的南极冻融探测

针对交叉极化比率（cross-polarized gradient ratio,XPGR）算法在冰盖冻融探测中存在的问题,提出了不依赖于实地观测数据的、自动阈值分割的、改进的XPGR微波辐射计南极冰盖冻融探测方法,即运用广义高斯模型对XPGR的长时间序列数据做直方图统计,计算出干湿雪划分的最优阈值,从而得到南极冰盖冻融分布图。试验结果表明：改进前后的结果虽然相差不多,但改进后的方法由于不依赖于实地观测数据,不仅提高了冰盖冻融探测的计算效率、实用性和可操作性,而且还在一定程度上提高了冰盖冻融探测的精度。     

多波多分量高斯束叠前深度偏移

本文对基于弹性波动理论的多波多分量高斯束偏移进行了完整且详细的分析和公式推导,实现了3D空间多分量(矢量)波场的直接成像.由于当前多数基于弹性波动方程的偏移方法只是假设应力边界条件为自由地表边界条件,这种假设不符合垂直地震剖面(VSP)和海底电缆(OBC)等地震数据.为此本文详细分析了实际应用中常见的三种弹性各向同性介质模型的应力边界条件:自由空间、海底和自由地表模型.在上行传播假设情况下,获得了应力边界条件与位移边界条件的关系式.在此基础上,准确推导了3D多波多分量高斯束波场延拓和偏移成像公式,并在偏移过程中实现了完整的多波型自动分离.由于常规的互相关成像条件不适用于矢量波场成像,本文引用了散度/旋度互相关成像条件.通过约定PS转换波的正向旋转方向解决了3D空间PS成像极性翻转问题.利用2D和3D模型数据偏移成像验证了我们所提出的多波多分量高斯束偏移方法的可行性.     

Fracture detection by Gaussian beam imaging of seismic data and image spectrum analysis

Localization of fractured areas is of primary interest in the study of oil and gas geology in carbonate environments. Hydrocarbon reservoirs in these environments are embedded within an impenetrable rock matrix but possess a rich system of various microheterogeneities, i.e., cavities, cracks, and fractures. Cavities accumulate oil, but its flow is governed by a system of fractures. A distinctive feature of wave propagation in such media is the excitation of the scattered/diffracted waves by the microheterogeneities. This scattering could be a reliable attribute for characterization of the fine structure of reservoirs, but it has extremely low energy and any standard data processing renders them practically invisible in comparison with images produced by specular reflections. Therefore, any attempts to use these waves for image congestion of microheterogeneities should first have a preliminary separation of the scattering and specular reflections. In this paper, the approach to performing this separation is based on the asymmetric summation. It is implemented by double focusing of Gaussian beams. To do this, the special weights are computed by propagating Gaussian beams from the target area towards the acquisition system separately for sources and receivers. The different mutual positioning of beams in each pair introduces a variety of selective images that are destined to represent some selected singular primitives of the target objects such as fractures, cavities, and edges. In this way, one can construct various wave images of a target reservoir, particularly in scattered/diffracted waves. Additional removal of remnants of specular reflections is done by means of spectral analysis of the scattered/diffracted waves' images to recognize and cancel extended lineaments. Numerical experiments with Sigsbee 2A synthetic seismic data and some typical structures of the Yurubcheno‐Tokhomskoye oil field in East Siberia are presented and discussed.     

高斯混合模型在星载激光全波形数据反演树高中的应用

全波形激光雷达后向散射回波,通过分解返回波形获取多种地物属性信息,在森林结构参数反演方面具有显著的优势,但是,当波形变形或者存在饱和度和前向散射时,高斯分量参数确定不准确以及有效波形起始位置不准确,降低波形分解精度。本文采用高斯混合模型对波形进行拟合,利用期望最大算法估计混合模型参数,抑制高斯分量初值敏感问题,特别是在大范围树高估算且要求一定精度的时候,以确定波形分解并且反演树高。本算法基于C++编程实现,实验结果表明,高斯混合模型能较好地拟合GLAS波形数据且对树高提取精度提升明显,该方法有着很好的有效性、稳定性和精确性。