Physics-informed neural network-based petroleum reservoir simulation with sparse data using domain decomposition

Recent advances in deep learning have expanded new possibilities for fluid flow simulation in petroleum reservoirs.However,the predominant approach in existing research is to train neural networks using high-fidelity numerical simulation data.This presents a significant challenge because the sole source of authentic wellbore production data for training is sparse.In response to this challenge,this work introduces a novel architecture called physics-informed neural network based on domain decomposition(PINN-DD),aiming to effectively utilize the sparse production data of wells for reservoir simulation with large-scale systems.To harness the capabilities of physics-informed neural networks(PINNs)in handling small-scale spatial-temporal domain while addressing the challenges of large-scale systems with sparse labeled data,the computational domain is divided into two distinct sub-domains:the well-containing and the well-free sub-domain.Moreover,the two sub-domains and the interface are rigorously constrained by the governing equations,data matching,and boundary conditions.The accuracy of the proposed method is evaluated on two problems,and its performance is compared against state-of-the-art PINNs through numerical analysis as a benchmark.The results demonstrate the superiority of PINN-DD in handling large-scale reservoir simulation with limited data and show its potential to outperform conventional PINNs in such scenarios.

Domain Decomposition of an Optimal Control Problem for Semi-Linear Elliptic Equations on Metric Graphs with Application to Gas Networks

We consider optimal control problems for the flow of gas in a pipe network. The equations of motions are taken to be represented by a semi-linear model derived from the fully nonlinear isothermal Euler gas equations. We formulate an optimal control problem on a given network and introduce a time discretization thereof. We then study the well-posedness of the corresponding time-discrete optimal control problem. In order to further reduce the complexity, we consider an instantaneous control strategy. The main part of the paper is concerned with a non-overlapping domain decomposition of the semi-linear elliptic optimal control problem on the graph into local problems on a small part of the network, ultimately on a single edge.

High efficient parallel numerical surface wave model based on an irregular quasi-rectangular domain decomposition scheme

To achieve high parallel efficiency for the global MASNUM surface wave model, the algorithm of an irregular quasirectangular domain decomposition and related serializing of calculating points and data exchanging schemes are developed and conducted, based on the environment of Message Passing Interface(MPI). The new parallel version of the surface wave model is tested for parallel computing on the platform of the Sunway BlueLight supercomputer in the National Supercomputing Center in Jinan. The testing involves four horizontal resolutions, which are 1°×1°,(1/2)°×(1/2)°,(1/4)°×(1/4)°, and(1/8)°×(1/8)°. These tests are performed without data Input/Output(IO) and the maximum amount of processors used in these tests reaches to 131072. The testing results show that the computing speeds of the model with different resolutions are all increased with the increasing of numbers of processors. When the number of processors is four times that of the base processor number, the parallel efficiencies of all resolutions are greater than 80%. When the number of processors is eight times that of the base processor number, the parallel efficiency of tests with resolutions of 1°×1°,(1/2)°×(1/2)° and(1/4)°×(1/4)° is greater than 80%, and it is 62% for the test with a resolution of(1/8)°×(1/8)° using 131072 processors, which is the nearly all processors of Sunway BlueLight. When the processor's number is 24 times that of the base processor number, the parallel efficiencies for tests with resolutions of 1°×1°,(1/2)°×(1/2)°, and(1/4)°×(1/4)° are 72%, 62%, and 38%, respectively. The speedup and parallel efficiency indicate that the irregular quasi-rectangular domain decomposition and serialization schemes lead to high parallel efficiency and good scalability for a global numerical wave model.

Extension of Near-Wall Domain Decomposition to Modeling Flows with Laminar-Turbulent Transition

The near-wall domain decomposition method(NDD)has proved to be very efficient for modeling near-wall fully turbulent flows.In this paper the NDD is extended to non-equilibrium regimeswith laminar-turbulent transition(LTT)for the first time.The LTT is identified with the use of the e^(N)-method which is applied to both incompressible and compressible flows.TheNDD ismodified to take into account LTT in an efficientway.In addition,implementation of the intermittency expands the capabilities of NDD to model non-equilibrium turbulent flows with transition.Performance of the modified NDD approach is demonstrated on various test problems of subsonic and supersonic flows past a flat plate,a supersonic flow over a compression corner and a planar shock wave impinging on a turbulent boundary layer.The results of modeling with and without decomposition are compared in terms of wall friction and show good agreement with each other while NDD significantly reducing computational resources needed.It turns out that the NDD can reduce the computational time as much as three times while retaining practically the same accuracy of prediction.

Parallel Computing of a Climate Model on the Dawn 1000 by Domain Decomposition Method

In this paper the parallel computing of a grid-point nine-level atmospheric general circulation model on the Dawn 1000 is introduced. The model was developed by the Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (CAS). The Dawn 1000 is a MIMD massive parallel computer made by National Research Center for Intelligent Computer (NCIC), CAS. A two-dimensional domain decomposition method is adopted to perform the parallel computing. The potential ways to increase the speed-up ratio and exploit more resources of future massively parallel supercomputation are also discussed.

PARTITION PROPERLY OF DOMAIN DECOMPOSITION WITHOUT ELLIPTICITY

Partition property plays a central role in domain decomposition methods. Existing theory essentially assumes certain ellipticity. We prove the partition property for prod lems without ellipticity which are of practical importance. Example applications include implicit schemes applied to degenerate parabolic partial differential equations arising from superconductors, superfluids and liquid crystals. With this partition property, Schwarz algorithms can be applied to general non-elliptic problems with an h-independent optimal convergence rate. Application to the time-dependent Ginzburg-Landau model of superconductivity is illustrated and numerical results are presented.

Slope displacement prediction based on multisource domain transfer learning for insufficient sample data

Accurate displacement prediction is critical for the early warning of landslides.The complexity of the coupling relationship between multiple influencing factors and displacement makes the accurate prediction of displacement difficult.Moreover,in engineering practice,insufficient monitoring data limit the performance of prediction models.To alleviate this problem,a displacement prediction method based on multisource domain transfer learning,which helps accurately predict data in the target domain through the knowledge of one or more source domains,is proposed.First,an optimized variational mode decomposition model based on the minimum sample entropy is used to decompose the cumulative displacement into the trend,periodic,and stochastic components.The trend component is predicted by an autoregressive model,and the periodic component is predicted by the long short-term memory.For the stochastic component,because it is affected by uncertainties,it is predicted by a combination of a Wasserstein generative adversarial network and multisource domain transfer learning for improved prediction accuracy.Considering a real mine slope as a case study,the proposed prediction method was validated.Therefore,this study provides new insights that can be applied to scenarios lacking sample data.

Multi-Domain Parallel Computing for Strength Analysis of Whole Aircraft Model

In the Windows XP 64 bit operating system environment, several common PC were used to build a cluster system, establishing the distributed memory parallel (DMP) computing system. A finite element model of whole aircraft with about 260 million degrees of freedom (DOF) was developed using three-node and four-node thin shell element and two-node beam element. With the large commercial finite element software MSC.MARC and employing two kinds of domain decomposition method (DDM) respectively, realized the parallel solving for the static strength analysis of the whole aircraft model, which offered a high cost-effective solution for solving large-scale and complex finite element models.

基于HMM的Domain-Flux恶意域名检测及分析

目前,僵尸网络广泛采用域名生成算法(Domain Generation Algorithm,DGA)生成大量随机域名躲避检测,这种躲避检测的方法已经成为破坏网络安全的主要威胁。因此,研究DGA域名识别方法对于检测恶意程序、打击僵尸网络、保障信息安全具有重要的现实意义。文章设计了基于ELK大数据平台的DGA域名检测分析框架,在充分研究黑名单等现有DGA域名识别方法的基础上,收集域名解析(Domain Name Server,DNS)业务系统的请求查询日志,以DGA域名为识别对象,基于隐式马尔可夫模型(Hidden Markov Model,HMM)对恶意域名进行聚类分析,从而实现对DGA域名的判定,进一步为僵尸网络等网络攻击行为的取证、溯源提供思路。实验结果表明,文章采用的轻量级检测分类器对正常域名和恶意域名的区分效果较好。

Identifying the validity domain of machine learning models in building energy systems

The building sector significantly contributes to climate change.To improve its carbon footprint,applications like model predictive control and predictive maintenance rely on system models.However,the high modeling effort hinders practical application.Machine learning models can significantly reduce this modeling effort.To ensure a machine learning model’s reliability in all operating states,it is essential to know its validity domain.Operating states outside the validity domain might lead to extrapolation,resulting in unpredictable behavior.This paper addresses the challenge of identifying extrapolation in data-driven building energy system models and aims to raise knowledge about it.For that,a novel approach is proposed that calibrates novelty detection algorithms towards the machine learning model.Suitable novelty detection algorithms are identified through a literature review and a benchmark test with 15 candidates.A subset of five algorithms is then evaluated on building energy systems.First,on two-dimensional data,displaying the results with a novel visualization scheme.Then on more complex multi-dimensional use cases.The methodology performs well,and the validity domain could be approximated.The visualization allows for a profound analysis and an improved understanding of the fundamental effects behind a machine learning model’s validity domain and the extrapolation regimes.

基于时频域分析的车载毫米波雷达干扰抑制方法

文中针对车载调频连续波雷达之间相互干扰导致虚警和漏警的问题,提出一种在时频域基于改进经验模式分解和自回归模型相结合的干扰抑制方法。该方法首先使用经验模式分解区分出拍频信号中干扰分量主导的低阶本征模态函数,将其转换到短时傅里叶变换域后通过全局阈值方法进行干扰分量定位;其次,在时频域根据定位信息将拍频信号包含干扰的数据置零;最后,使用自回归模型对拍频信号中缺失的有用信号进行估计并插值。通过仿真和实测结果显示,该方法在精确地去除干扰分量的同时可以减少有用信号的功率损失,干扰抑制后的信号与参考信号的相关系数达到0.9697。与现有干扰抑制技术相比文中方法也体现出更优的干扰抑制性能。

基于惠更斯等效原理的高速高密度PCB分级建模方法

当带宽达到20 GHz以上时(波长从15 mm到无穷小),高速高密度电子系统的内部电路电磁环境变得十分复杂,越来越难以建模和分析预测,进一步当带宽达到40 GHz以上时,电路电磁环境问题将变得更加尖锐.为在设计阶段能够对电磁作用过程、作用效应进行预测评估及控制,需要精确的建模方法和针对大尺度问题的快速计算技术,尤其是超大带宽超高速混合电路和集成电路.本文提出了基于混合电路环境电磁计算基础理论的跨尺度处理技术,通过惠更斯等效原理和电磁奇异性几何结构的电磁收敛降速机理的利用,解决了多维交叉多尺度电路电磁环境场路融合的高效率高精度建模技术挑战;利用惠更斯等效原理和基尔霍夫积分方程,在区域边界面上定义了惠更斯端口,提出了对于任意复杂印制电路板(Printed Circuit Board,PCB)同时垂直方向和水平方向区域分解的一般方法,实现了任意PCB结构的分级分类处理和模块化封装,提高了高速高密度电子系统分析的灵活度;提出了基于几何结构电磁学多重本地本征展开的技术途径,发展了基于模式和区域分解分割的快速并行处理技术,通过电磁锐变区域本征描述及场分布的本征基函数表征,实现了高精度和高计算速度兼得,减少了复杂电子系统的计算时间和设计时间.统计数据表明,本文提出的方法在0~40 GHz大带宽频率范围内频偏误差为3.7%、幅度偏差为±3 dB.本文提出的PCB分级建模分析方法可以应用于高端电子通信系统设计,提升我国宽带高速数模系统的高效电路设计和环境电磁调控能力,缩短产品研发周期.

分位数视角下地缘政治风险与中国大宗商品市场溢出研究

近年来,地缘政治事件频发,地缘政治风险与大宗商品市场损失、收益间风险溢出问题不容忽视。本文基于中国大宗商品期货数据,利用分位数溢出网络模型及其频域分解技术,研究了相较于常规状态,极端右尾地缘政治风险与大宗商品市场损失、收益间的风险溢出效应。研究表明:常规状态下,地缘政治风险与大宗商品期货形成的交叉溢出网络具有稀疏性,以自反馈溢出为主导,而极端状态下,转为紧密网络,交叉溢出显著上升,自反馈溢出显著下降,形成风险从内部向外部传播的扩散效应。多数时候地缘政治风险在常规状态下表现为风险接受者,右尾状态下表现为强烈的风险发出者,尤其是在2020年和2022年两个关键时期,地缘政治风险对我国大宗商品市场产生了全面的强烈冲击。频域溢出层面,常规状态下,地缘政治风险以短期净溢出为主,长期溢出不显著;极端状态下,短期、长期溢出效应均显著。具体行业而言,原油处于交叉溢出的关键位置,且在极端时期对外溢出显著,豆粕易从常规时期的风险发出转变为极端时期的风险接受。本文的研究结论对于理解地缘政治风险与大宗商品市场的复杂关系、实现大宗商品市场风险管理具有重要意义。

An application of space-filling curves to improve results of turbulent aerodynamics modeling with convolutional neural networks

When carrying out calculations for turbulent flow simulation,one inevitably has to face the choice between accuracy and speed of calculations.In order to simultaneously obtain both a computationally efficient and more accurate model,a surrogate model can be built on the basis of some fast special model and knowledge of previous calculations obtained by more accurate base models from various test bases or some results of serial calculations.The objective of this work is to construct a surrogate model which allows to improve the accuracy of turbulent calculations obtained by a special model on unstructured meshes.For this purpose,we use 1D Convolutional Neural Network(CNN)of the encoder-decoder architecture and reduce the problem to a single dimension by applying space-filling curves.Such an approach would have the benefit of being applicable to solutions obtained on unstructured meshes.In this work,a non-local approach is applied where entire flow fields obtained by the special and base models are used as input and ground truth output respectively.Spalart-Allmaras(SA)model and Near-wall Domain Decomposition(NDD)method for SA are taken as the base and special models respectively.The efficiency and accuracy of the obtained surrogate model are demonstrated in a case of supersonic flow over a compression corner with different values for angleαand Reynolds number Re.We conducted an investigation into interpolation and extrapolation by Re and also into interpolation byα.

Exponentially Convergent Multiscale Finite Element Method

We provide a concise review of the exponentially convergent multiscale finite element method(ExpMsFEM)for efficient model reduction of PDEs in heterogeneous media without scale separation and in high-frequency wave propagation.The ExpMsFEM is built on the non-overlapped domain decomposition in the classical MsFEM while enriching the approximation space systematically to achieve a nearly exponential convergence rate regarding the number of basis functions.Unlike most generalizations of the MsFEM in the literature,the ExpMsFEM does not rely on any partition of unity functions.In general,it is necessary to use function representations dependent on the right-hand side to break the algebraic Kolmogorov n-width barrier to achieve exponential convergence.Indeed,there are online and offline parts in the function representation provided by the ExpMsFEM.The online part depends on the right-hand side locally and can be computed in parallel efficiently.The offline part contains basis functions that are used in the Galerkin method to assemble the stiffness matrix;they are all independent of the right-hand side,so the stiffness matrix can be used repeatedly in multi-query scenarios.

PFNN-2:A Domain Decomposed Penalty-Free Neural Network Method for Solving Partial Differential Equations

A new penalty-free neural network method,PFNN-2,is presented for solving partial differential equations,which is a subsequent improvement of our previously proposed PFNN method[1].PFNN-2 inherits all advantages of PFNN in handling the smoothness constraints and essential boundary conditions of self-adjoint problems with complex geometries,and extends the application to a broader range of non-self-adjoint time-dependent differential equations.In addition,PFNN-2 introduces an overlapping domain decomposition strategy to substantially improve the training efficiency without sacrificing accuracy.Experiments results on a series of partial differential equations are reported,which demonstrate that PFNN-2 can outperform state-of-the-art neural network methods in various aspects such as numerical accuracy,convergence speed,and parallel scalability.

采用变分模态分解与领域自适应的表面肌电信号手势识别

针对传统机器学习在表面肌电信号手势识别领域的适应性和准确性不足,以及新用户因个体生理和行为差异在已有模型上表现不佳的问题,提出一种利用卷积神经网络模型并有效克服肌电数据分布差异的算法,用于提升手势识别的性能。首先对肌电信号进行变分模态分解,构建易于识别的表面肌电图像,并提出了一种卷积神经网络模型进行手势识别,提升用户相关的肌电信号手势识别准确率;同时利用迁移学习中的领域自适应和模型微调技术,提升用户无关的肌电信号手势识别准确率,并将所提算法在NinaPro DB1肌电数据集中进行了3分类、4分类、5分类和12分类共4组评估验证。结果表明:在4组评估验证中,用户相关的肌电信号手势识别平均准确率分别达到了99.28%、99.30%、98.39%和93.40%,用户无关的肌电信号手势识别平均准确率分别达到了94.05%、92.60%、88.38%和70.03%,表明本文提出的算法在表面肌电信号手势识别中具有良好的效果,为实现人机交互中的普适性的肌电设备开发提供了一种可行的方案。

面向地学分析AI建模的地理信息服务层次网络模型

在人工智能生成(AIGC)和大语言模型(LLM)的大背景下,如何提升地学分析模型构建的智能化水平已成为业界广泛关注的焦点。为此,本文提出地理信息服务层次网络模型5-HiNet,基于需求描述、抽象模型、功能模块、服务接口、函数实例的5层子网络结构对海量地学分析模型进行建模,由抽象到具体逐层描述地学分析模型的实现过程,固化模型中的知识,形成面向地学分析模型的完备知识体系。5-HiNet模型能够进一步与大语言模型(LLM)进行耦合,实现地学分析模型的智能化生成。本文通过原型系统和应用案例,初步验证了5-HiNet的可行性,并为未来研究和应用提供新的方向和思路。

考虑磁-力耦合效应的混合磁滞模型研究

应力作用会影响磁性材料的磁性能,准确快速地模拟磁性材料在不同应力下的磁化特性对设计承受机械应力的装备铁心具有重要意义。该文研究拉应力作用下取向电工钢磁畴结构沿轧制方向的变化规律,将拉应力与畴壁平移运动之间的关系量化处理,由此采用基于畴壁平移运动的J-A模型对材料的磁化特性进行描述,通过提出适用于低磁通密度范围的新型无磁滞磁化公式替代传统J-A理论中的郎之万公式以考虑拉应力对取向电工钢磁化特性的影响;通过将磁畴简化为独立个体,将磁致伸缩效应在高磁通密度范围对取向电工钢磁化特性的影响转换为磁畴之间的摩擦作用,提出适用于高磁通密度范围的改进磁滞模型;应用BP神经网络寻找出最佳比例参数,将两种不同适用范围的磁滞模型构建为完整的磁-力耦合模型。通过对比实验结果与模型计算结果,验证了所提混合磁滞模型的准确性。

时域孪生网络融合Transformer的长时无人机视觉跟踪

针对无人机(UAV)执行跟踪任务时经常出现尺寸变化、低分辨率、目标遮挡等场景导致跟踪目标框漂移的问题,提出一种时域孪生网络融合Transformer的长时无人机视觉跟踪算法TTTrack。首先,使用基于孪生网络的SiamFC++(AlexNet)算法作为基线算法;其次,利用Transformer自适应地提取历史帧的时空信息并在线更新模板,从而将时空上下文信息储存为动态模板;随后,分别使用基准模板和动态模板与搜索特征图进行互相关运算,获得响应图后利用Transformer融合两个响应图,从而在连续帧之间建立时空上下文映射关系。实验结果表明,在LaSOT长序列跟踪基准上TTTrack的成功率和精确率分别为63.9%和66.6%,在UAV123跟踪基准上的成功率和精确率分别为61.4%和80.2%。与基线算法相比,该算法在完全遮挡场景下的成功率和精确率分别提升7.4和8.0个百分点。TTTrack在DTB70跟踪基准上精确率达到82.1%,并且跟踪速度为118 帧/s,满足实时性要求。测试结果验证了TTTrack具有良好的鲁棒性、实时性和抗干扰能力,能有效应对长时UAV跟踪任务。