In this paper, a novel approach is introduced towards an efficient Finite-Difference Time-Domain (FDTD) algorithm by incorporating the Alternating Direction Implicit (ADI) technique to the Nonorthogonal FDTD (NFDTD) m...In this paper, a novel approach is introduced towards an efficient Finite-Difference Time-Domain (FDTD) algorithm by incorporating the Alternating Direction Implicit (ADI) technique to the Nonorthogonal FDTD (NFDTD) method. This scheme can be regarded as an extension of the conventional ADI-FDTD scheme into a generalized curvilinear coordinate system. The improvement on accuracy and the numerical efficiency of the ADI-NFDTD over the conventional nonorthogonal and the ADI-FDTD algorithms is carried out by numerical experiments. The application in the modelling of the Electromagnetic Bandgap (EBG) structure has further demonstrated the advantage of the proposed method.展开更多
In this paper, first we calculate finite-difference coefficients of implicit finite- difference methods （IFDM） for the first and second-order derivatives on normal grids and first- order derivatives on staggered gri...In this paper, first we calculate finite-difference coefficients of implicit finite- difference methods （IFDM） for the first and second-order derivatives on normal grids and first- order derivatives on staggered grids and find that small coefficients of high-order IFDMs exist. Dispersion analysis demonstrates that omitting these small coefficients can retain approximately the same order accuracy but greatly reduce computational costs. Then, we introduce a mirrorimage symmetric boundary condition to improve IFDMs accuracy and stability and adopt the hybrid absorbing boundary condition （ABC） to reduce unwanted reflections from the model boundary. Last, we give elastic wave modeling examples for homogeneous and heterogeneous models to demonstrate the advantages of the proposed scheme.展开更多
A modified alternating direction implicit algorithm is proposed to solve the full-vectorial finite-difference beam propagation method formulation based on H fields. The cross-coupling terms are neglected in the first ...A modified alternating direction implicit algorithm is proposed to solve the full-vectorial finite-difference beam propagation method formulation based on H fields. The cross-coupling terms are neglected in the first sub-step, but evaluated and doubly used in the second sub-step. The order of two sub-steps is reversed for each transverse magnetic field component so that the cross-coupling terms are always expressed in implicit form, thus the calculation is very efficient and stable. Moreover, an improved six-point finite-difference scheme with high accuracy independent of specific structures of waveguide is also constructed to approximate the cross-coupling terms along the transverse directions. The imaginary-distance procedure is used to assess the validity and utility of the present method. The field patterns and the normalized propagation constants of the fundamental mode for a buried rectangular waveguide and a rib waveguide are presented. Solutions are in excellent agreement with the benchmark results from the modal transverse resonance method.展开更多
Staggered-grid finite-difference(SGFD)schemes have been widely used in acoustic wave modeling for geophysical problems.Many improved methods are proposed to enhance the accuracy of numerical modeling.However,these met...Staggered-grid finite-difference(SGFD)schemes have been widely used in acoustic wave modeling for geophysical problems.Many improved methods are proposed to enhance the accuracy of numerical modeling.However,these methods are inevitably limited by the maximum Courant-Friedrichs-Lewy(CFL)numbers,making them unstable when modeling with large time sampling intervals or small grid spacings.To solve this problem,we extend a stable SGFD scheme by controlling SGFD dispersion relations and maximizing the maximum CFL numbers.First,to improve modeling stability,we minimize the error between the FD dispersion relation and the exact relation in the given wave-number region,and make the FD dispersion approach a given function outside the given wave-number area,thus breaking the conventional limits of the maximum CFL number.Second,to obtain high modeling accuracy,we use the SGFD scheme based on the Remez algorithm to compute the FD coefficients.In addition,the hybrid absorbing boundary condition is adopted to suppress boundary reflections and we find a suitable weighting coefficient for the proposed scheme.Theoretical derivation and numerical modeling demonstrate that the proposed scheme can maintain high accuracy in the modeling process and the value of the maximum CFL number of the proposed scheme can exceed that of the conventional SGFD scheme when adopting a small maximum effective wavenumber,indicating that the proposed scheme improves stability during the modeling.展开更多
Reverse time migration and full waveform inversion involve the crosscorrelation of two wavefields,propagated in the forward-and reverse-time directions,respectively.As a result,the forward-propagated wavefield needs t...Reverse time migration and full waveform inversion involve the crosscorrelation of two wavefields,propagated in the forward-and reverse-time directions,respectively.As a result,the forward-propagated wavefield needs to be stored,and then accessed to compute the correlation with the backward-propagated wavefield.Boundary-value methods reconstruct the source wavefield using saved boundary wavefields and can significantly reduce the storage requirements.However,the existing boundary-value methods are based on the explicit finite-difference(FD)approximations of the spatial derivatives.Implicit FD methods exhibit greater accuracy and thus allow for a smaller operator length.We develop two(an accuracy-preserving and a memory-efficient)wavefield reconstruction schemes based on an implicit staggered-grid FD(SFD)operator.The former uses boundary wavefields at M layers of grid points and the spatial derivatives of wavefields at one layer of grid points to reconstruct the source wavefield for a(2M+2)th-order implicit SFD operator.The latter applies boundary wavefields at N layers of grid points,a linear combination of wavefields at M–N layers of grid points,and the spatial derivatives of wavefields at one layer of grid points to reconstruct the source wavefield(0≤N<M).The required memory of accuracy-preserving and memory-efficient schemes is(M+1)/M and(N+2)/M times,respectively,that of the explicit reconstruction scheme.Numerical results reveal that the accuracy-preserving scheme can achieve accurate reconstruction at the cost of storage.The memory-efficient scheme with N=2 can obtain plausible reconstructed wavefields and images,and the storage amount is 4/(M+1)of the accuracy-preserving scheme.展开更多
3D eikonal equation is a partial differential equation for the calculation of first-arrival traveltimes and has been widely applied in many scopes such as ray tracing,source localization,reflection migration,seismic m...3D eikonal equation is a partial differential equation for the calculation of first-arrival traveltimes and has been widely applied in many scopes such as ray tracing,source localization,reflection migration,seismic monitoring and tomographic imaging.In recent years,many advanced methods have been developed to solve the 3D eikonal equation in heterogeneous media.However,there are still challenges for the stable and accurate calculation of first-arrival traveltimes in 3D strongly inhomogeneous media.In this paper,we propose an adaptive finite-difference(AFD)method to numerically solve the 3D eikonal equation.The novel method makes full use of the advantages of different local operators characterizing different seismic wave types to calculate factors and traveltimes,and then the most accurate factor and traveltime are adaptively selected for the convergent updating based on the Fermat principle.Combined with global fast sweeping describing seismic waves propagating along eight directions in 3D media,our novel method can achieve the robust calculation of first-arrival traveltimes with high precision at grid points either near source point or far away from source point even in a velocity model with large and sharp contrasts.Several numerical examples show the good performance of the AFD method,which will be beneficial to many scientific applications.展开更多
We propose a suite of strategies for the parallel solution of fully implicit monolithic fluid-structure interaction(FSI).The solver is based on a modeling approach that uses the velocity and pressure as the primitive ...We propose a suite of strategies for the parallel solution of fully implicit monolithic fluid-structure interaction(FSI).The solver is based on a modeling approach that uses the velocity and pressure as the primitive variables,which offers a bridge between computational fluid dynamics(CFD)and computational structural dynamics.The spatiotemporal discretization leverages the variational multiscale formulation and the generalized-αmethod as a means of providing a robust discrete scheme.In particular,the time integration scheme does not suffer from the overshoot phenomenon and optimally dissipates high-frequency spurious modes in both subproblems of FSI.Based on the chosen fully implicit scheme,we systematically develop a combined suite of nonlinear and linear solver strategies.Invoking a block factorization of the Jacobian matrix,the Newton-Raphson procedure is reduced to solving two smaller linear systems in the multi-corrector stage.The first is of the elliptic type,indicating that the algebraic multigrid method serves as a well-suited option.The second exhibits a two-by-two block structure that is analogous to the system arising in CFD.Inspired by prior studies,the additive Schwarz domain decomposition method and the block-factorization-based preconditioners are invoked to address the linear problem.Since the number of unknowns matches in both subdomains,it is straightforward to balance loads when parallelizing the algorithm for distributed-memory architectures.We use two representative FSI benchmarks to demonstrate the robustness,efficiency,and scalability of the overall FSI solver framework.In particular,it is found that the developed FSI solver is comparable to the CFD solver in several aspects,including fixed-size and isogranular scalability as well as robustness.展开更多
Realistic human reconstruction embraces an extensive range of applications as depth sensors advance.However,current stateof-the-art methods with RGB-D input still suffer from artefacts,such as noisy surfaces,non-human...Realistic human reconstruction embraces an extensive range of applications as depth sensors advance.However,current stateof-the-art methods with RGB-D input still suffer from artefacts,such as noisy surfaces,non-human shapes,and depth ambiguity,especially for the invisible parts.The authors observe the main issue is the lack of geometric semantics without using depth input priors fully.This paper focuses on improving the representation ability of implicit function,exploring an effective method to utilise depth-related semantics effectively and efficiently.The proposed geometry-enhanced implicit function enhances the geometric semantics with the extra voxel-aligned features from point clouds,promoting the completion of missing parts for unseen regions while preserving the local details on the input.For incorporating multi-scale pixel-aligned and voxelaligned features,the authors use the Squeeze-and-Excitation attention to capture and fully use channel interdependencies.For the multi-view reconstruction,the proposed depth-enhanced attention explicitly excites the network to“sense”the geometric structure for a more reasonable feature aggregation.Experiments and results show that our method outperforms current RGB and depth-based SOTA methods on the challenging data from Twindom and Thuman3.0,and achieves a detailed and completed human reconstruction,balancing performance and efficiency well.展开更多
The accuracy of numerical computation heavily relies on appropriate meshing,whichserves as the foundation for numerical computation.Although adaptive refinement methods areavailable,an adaptive numerical solution is l...The accuracy of numerical computation heavily relies on appropriate meshing,whichserves as the foundation for numerical computation.Although adaptive refinement methods areavailable,an adaptive numerical solution is likely to be ineffective if it originates from a poorly ini-tial mesh.Therefore,it is crucial to generate meshes that accurately capture the geometric features.As an indispensable input in meshing methods,the Mesh Size Function(MSF)determines the qual-ity of the generated mesh.However,the current generation of MSF involves human participation tospecify numerous parameters,leading to difficulties in practical usage.Considering the capacity ofmachine learning to reveal the latent relationships within data,this paper proposes a novel machinelearning method,Implicit Geometry Neural Network(IGNN),for automatic prediction of appro-priate MSFs based on the existing mesh data,enabling the generation of unstructured meshes thatalign precisely with geometric features.IGNN employs the generative adversarial theory to learnthe mapping between the implicit representation of the geometry(Signed Distance Function,SDF)and the corresponding MSF.Experimental results show that the proposed method is capableof automatically generating appropriate meshes and achieving comparable meshing results com-pared to traditional methods.This paper demonstrates the possibility of significantly decreasingthe workload of mesh generation using machine learning techniques,and it is expected to increasethe automation level of mesh generation.展开更多
Although conventional object detection methods achieve high accuracy through extensively annotated datasets,acquiring such large-scale labeled data remains challenging and cost-prohibitive in numerous real-world appli...Although conventional object detection methods achieve high accuracy through extensively annotated datasets,acquiring such large-scale labeled data remains challenging and cost-prohibitive in numerous real-world applications.Few-shot object detection presents a new research idea that aims to localize and classify objects in images using only limited annotated examples.However,the inherent challenge in few-shot object detection lies in the insufficient sample diversity to fully characterize the sample feature distribution,which consequently impacts model performance.Inspired by contrastive learning principles,we propose an Implicit Feature Contrastive Learning(IFCL)module to address this limitation and augment feature diversity for more robust representational learning.This module generates augmented support sample features in a mixed feature space and implicitly contrasts them with query Region of Interest(RoI)features.This approach facilitates more comprehensive learning of both intra-class feature similarity and inter-class feature diversity,thereby enhancing the model’s object classification and localization capabilities.Extensive experiments on PASCAL VOC show that our method achieves a respective improvement of 3.2%,1.8%,and 2.3%on 10-shot of three Novel Sets compared to the baseline model FPD.展开更多
Prestack reverse time migration （RTM） is an accurate imaging method ofsubsurface media. The viscoacoustic prestack RTM is of practical significance because itconsiders the viscosity of the subsurface media. One of t...Prestack reverse time migration （RTM） is an accurate imaging method ofsubsurface media. The viscoacoustic prestack RTM is of practical significance because itconsiders the viscosity of the subsurface media. One of the steps of RTM is solving thewave equation and extrapolating the wave field forward and backward; therefore, solvingaccurately and efficiently the wave equation affects the imaging results and the efficiencyof RTM. In this study, we use the optimal time-space domain dispersion high-order finite-difference （FD） method to solve the viscoacoustic wave equation. Dispersion analysis andnumerical simulations show that the optimal time-space domain FD method is more accurateand suppresses the numerical dispersion. We use hybrid absorbing boundary conditions tohandle the boundary reflection. We also use source-normalized cross-correlation imagingconditions for migration and apply Laplace filtering to remove the low-frequency noise.Numerical modeling suggests that the viscoacoustic wave equation RTM has higher imagingresolution than the acoustic wave equation RTM when the viscosity of the subsurface isconsidered. In addition, for the wave field extrapolation, we use the adaptive variable-lengthFD operator to calculate the spatial derivatives and improve the computational efficiencywithout compromising the accuracy of the numerical solution.展开更多
Generally, FD coefficients can be obtained by using Taylor series expansion （TE） or optimization methods to minimize the dispersion error. However, the TE-based FD method only achieves high modeling precision over a...Generally, FD coefficients can be obtained by using Taylor series expansion （TE） or optimization methods to minimize the dispersion error. However, the TE-based FD method only achieves high modeling precision over a limited range of wavenumbers, and produces large numerical dispersion beyond this range. The optimal FD scheme based on least squares （LS） can guarantee high precision over a larger range of wavenumbers and obtain the best optimization solution at small computational cost. We extend the LS-based optimal FD scheme from two-dimensional （2D） forward modeling to three-dimensional （3D） and develop a 3D acoustic optimal FD method with high efficiency, wide range of high accuracy and adaptability to parallel computing. Dispersion analysis and forward modeling demonstrate that the developed FD method suppresses numerical dispersion. Finally, we use the developed FD method to source wavefield extrapolation and receiver wavefield extrapolation in 3D RTM. To decrease the computation time and storage requirements, the 3D RTM is implemented by combining the efficient boundary storage with checkpointing strategies on GPU. 3D RTM imaging results suggest that the 3D optimal FD method has higher precision than conventional methods.展开更多
To improve the accuracy of the conventional finite-difference method, finitedifference numerical modeling methods of any even-order accuracy are recommended. We introduce any even-order accuracy difference schemes of ...To improve the accuracy of the conventional finite-difference method, finitedifference numerical modeling methods of any even-order accuracy are recommended. We introduce any even-order accuracy difference schemes of any-order derivatives derived from Taylor series expansion. Then, a finite-difference numerical modeling method with any evenorder accuracy is utilized to simulate seismic wave propagation in two-phase anisotropic media. Results indicate that modeling accuracy improves with the increase of difference accuracy order number. It is essential to find the optimal order number, grid size, and time step to balance modeling precision and computational complexity. Four kinds of waves, static mode in the source point, SV wave cusps, reflection and transmission waves are observed in two-phase anisotropic media through modeling.展开更多
During seismic wave propagation on a free surface, a strong material contrast boundary develops in response to interference by P- and S- waves to create a surface-wave phenomenon. To accurately determine the effects o...During seismic wave propagation on a free surface, a strong material contrast boundary develops in response to interference by P- and S- waves to create a surface-wave phenomenon. To accurately determine the effects of this interface on surface-wave propagation, the boundary conditions must be accurately modeled. In this paper, we present a numerical approach based on the dynamic poroelasticity for a space–time-domain staggered-grid finite-difference simulation in porous media that contain a free-surface boundary. We propose a generalized stess mirror formulation of the free-surface boundary for solids and fluids in porous media for the grid mesh on which lays the free-surface plane. Its analog is that used for elastic media, which is suitable for precise and stable Rayleigh-type surface-wave modeling. The results of our analysis of first kind of Rayleigh （R1） waves obtained by this model demonstrate that the discretization of the mesh in a similar way to that for elastic media can realize stable numerical solutions with acceptable precision. We present numerical examples demonstrating the efficiency and accuracy of our proposed method.展开更多
Numerical simulation in the frequency-space domain has inherent advantages, such as： it is possible to simulate wave propagation from multiple sources simultaneously; there are no cumulative errors; only the interest...Numerical simulation in the frequency-space domain has inherent advantages, such as： it is possible to simulate wave propagation from multiple sources simultaneously; there are no cumulative errors; only the interesting frequencies can be selected; and it is more suitable for wave propagation in viscoelastic media. The only obstacle to using the method is the requirement of huge computer storage. We extend the compressed format for storing the coefficient matrix. It can reduce the required computer storage dramatically. We get the optimal coefficients by least-squares method to suppress the numerical dispersion and adopt the perfectly matched layer （PML） boundary conditions to eliminate the artificial boundary reflections. Using larger grid intervals decreases computer storage requirements and provides high computational efficiency. Numerical experiments demonstrate that these means are economic and effective, providing a good basis for elastic wave imaging and inversion.展开更多
The Z-Axis tiPPer eiectromagnetic （ZTEM） technique is based on a frequency-domain airbome electromagnetic system that measures the natural magnetic field. A survey area was divided into several blocks by using the M...The Z-Axis tiPPer eiectromagnetic （ZTEM） technique is based on a frequency-domain airbome electromagnetic system that measures the natural magnetic field. A survey area was divided into several blocks by using the Maxwell＇s equations, and the magnetic components at the center of each edge of the grid cell are evaluated by applying the staggered-grid finite-difference method. The tipper and its divergence are derived to complete the 3D ZTEM forward modeling algorithm. A synthetic model is then used to compare the responses with those of 2D finite-element forward modeling to verify the accuracy of the algorithm. ZTEM offers high horizontal resolution to both simple and complex distributions of conductivity. This work is the theoretical foundation for the interpretation of ZTEM data and the study of 3D ZTEM inversion.展开更多
The staggered-grid finite-difference （SGFD） method has been widely used in seismic forward modeling. The precision of the forward modeling results directly affects the results of the subsequent seismic inversion and...The staggered-grid finite-difference （SGFD） method has been widely used in seismic forward modeling. The precision of the forward modeling results directly affects the results of the subsequent seismic inversion and migration. Numerical dispersion is one of the problems in this method. The window function method can reduce dispersion by replacing the finite-difference operators with window operators, obtained by truncating the spatial convolution series of the pseudospectral method. Although the window operators have high precision in the low-wavenumber domain, their precision decreases rapidly in the high-wavenumber domain. We develop a least squares optimization method to enhance the precision of operators obtained by the window function method. We transform the SGFD problem into a least squares problem and find the best solution iteratively. The window operator is chosen as the initial value and the optimized domain is set by the error threshold. The conjugate gradient method is also adopted to increase the stability of the solution. Approximation error analysis and numerical simulation results suggest that the proposed method increases the precision of the window function operators and decreases the numerical dispersion.展开更多
Numerical simulation in transverse isotropic media with tilted symmetry axis（TTI） using the standard staggered-grid finite-difference scheme（SSG）results in errors caused by averaging or interpolation. In order to ...Numerical simulation in transverse isotropic media with tilted symmetry axis（TTI） using the standard staggered-grid finite-difference scheme（SSG）results in errors caused by averaging or interpolation. In order to eliminate the errors, a method of rotated staggered-grid finite-difference scheme（RSG） is proposed. However, the RSG brings serious numerical dispersion. The compact staggered-grid finite-difference scheme（CSG） is an implicit difference scheme, which use fewer grid points to suppress dispersion more effectively than the SSG. This paper combines the CSG with the RSG to derive a rotated staggered-grid compact finite-difference scheme（RSGC）. The numerical experiments indicate that the RSGC has weaker numerical dispersion and better accuracy than the RSG.展开更多
To the most of velocity fields, the traveltimes of the first break that seismic waves propagate along rays can be computed on a 2-D or 3-D numerical grid by finite-difference extrapolation. Under ensuring accuracy, t...To the most of velocity fields, the traveltimes of the first break that seismic waves propagate along rays can be computed on a 2-D or 3-D numerical grid by finite-difference extrapolation. Under ensuring accuracy, to improve calculating efficiency and adaptability, the calculation method of first-arrival traveltime of finite-difference is de- rived based on any rectangular grid and a local plane wavefront approximation. In addition, head waves and scat- tering waves are properly treated and shadow and caustic zones cannot be encountered, which appear in traditional ray-tracing. The testes of two simple models and the complex Marmousi model show that the method has higher accuracy and adaptability to complex structure with strong vertical and lateral velocity variation, and Kirchhoff prestack depth migration based on this method can basically achieve the position imaging effects of wave equation prestack depth migration in major structures and targets. Because of not taking account of the later arrivals energy, the effect of its amplitude preservation is worse than that by wave equation method, but its computing efficiency is higher than that by total Green′s function method and wave equation method.展开更多
Conventional finite-difference(FD)methods cannot model acoustic wave propagation beyond Courant-Friedrichs-Lewy(CFL)numbers 0.707 and 0.577 for two-dimensional(2D)and three-dimensional(3D)equal spacing cases,respectiv...Conventional finite-difference(FD)methods cannot model acoustic wave propagation beyond Courant-Friedrichs-Lewy(CFL)numbers 0.707 and 0.577 for two-dimensional(2D)and three-dimensional(3D)equal spacing cases,respectively,thereby limiting time step selection.Based on the definition of temporal and spatial FD operators,we propose a variable-length temporal and spatial operator strategy to model wave propagation beyond those CFL numbers while preserving accuracy.First,to simulate wave propagation beyond the conventional CFL stability limit,the lengths of the temporal operators are modified to exceed the lengths of the spatial operators for high-velocity zones.Second,to preserve the modeling accuracy,the velocity-dependent lengths of the temporal and spatial operators are adaptively varied.The maximum CFL numbers for the proposed method can reach 1.25 and 1.0 in high velocity contrast 2D and 3D simulation examples,respectively.We demonstrate the effectiveness of our method by modeling wave propagation in simple and complex media.展开更多
文摘In this paper, a novel approach is introduced towards an efficient Finite-Difference Time-Domain (FDTD) algorithm by incorporating the Alternating Direction Implicit (ADI) technique to the Nonorthogonal FDTD (NFDTD) method. This scheme can be regarded as an extension of the conventional ADI-FDTD scheme into a generalized curvilinear coordinate system. The improvement on accuracy and the numerical efficiency of the ADI-NFDTD over the conventional nonorthogonal and the ADI-FDTD algorithms is carried out by numerical experiments. The application in the modelling of the Electromagnetic Bandgap (EBG) structure has further demonstrated the advantage of the proposed method.
基金supported by the National Natural Science Foundation of China(NSFC)(Grant No. 41074100)the Program for New Century Excellent Talents in University of Ministry of Education of China(Grant No. NCET-10-0812)
文摘In this paper, first we calculate finite-difference coefficients of implicit finite- difference methods （IFDM） for the first and second-order derivatives on normal grids and first- order derivatives on staggered grids and find that small coefficients of high-order IFDMs exist. Dispersion analysis demonstrates that omitting these small coefficients can retain approximately the same order accuracy but greatly reduce computational costs. Then, we introduce a mirrorimage symmetric boundary condition to improve IFDMs accuracy and stability and adopt the hybrid absorbing boundary condition （ABC） to reduce unwanted reflections from the model boundary. Last, we give elastic wave modeling examples for homogeneous and heterogeneous models to demonstrate the advantages of the proposed scheme.
文摘A modified alternating direction implicit algorithm is proposed to solve the full-vectorial finite-difference beam propagation method formulation based on H fields. The cross-coupling terms are neglected in the first sub-step, but evaluated and doubly used in the second sub-step. The order of two sub-steps is reversed for each transverse magnetic field component so that the cross-coupling terms are always expressed in implicit form, thus the calculation is very efficient and stable. Moreover, an improved six-point finite-difference scheme with high accuracy independent of specific structures of waveguide is also constructed to approximate the cross-coupling terms along the transverse directions. The imaginary-distance procedure is used to assess the validity and utility of the present method. The field patterns and the normalized propagation constants of the fundamental mode for a buried rectangular waveguide and a rib waveguide are presented. Solutions are in excellent agreement with the benchmark results from the modal transverse resonance method.
基金This research is supported by the National Natural Science Foundation of China(NSFC)under contract no.42274147.
文摘Staggered-grid finite-difference(SGFD)schemes have been widely used in acoustic wave modeling for geophysical problems.Many improved methods are proposed to enhance the accuracy of numerical modeling.However,these methods are inevitably limited by the maximum Courant-Friedrichs-Lewy(CFL)numbers,making them unstable when modeling with large time sampling intervals or small grid spacings.To solve this problem,we extend a stable SGFD scheme by controlling SGFD dispersion relations and maximizing the maximum CFL numbers.First,to improve modeling stability,we minimize the error between the FD dispersion relation and the exact relation in the given wave-number region,and make the FD dispersion approach a given function outside the given wave-number area,thus breaking the conventional limits of the maximum CFL number.Second,to obtain high modeling accuracy,we use the SGFD scheme based on the Remez algorithm to compute the FD coefficients.In addition,the hybrid absorbing boundary condition is adopted to suppress boundary reflections and we find a suitable weighting coefficient for the proposed scheme.Theoretical derivation and numerical modeling demonstrate that the proposed scheme can maintain high accuracy in the modeling process and the value of the maximum CFL number of the proposed scheme can exceed that of the conventional SGFD scheme when adopting a small maximum effective wavenumber,indicating that the proposed scheme improves stability during the modeling.
基金partially supported by National Key R&D Program of China(2021YFA0716902)the National Natural Science Foundation of China(42174156)the Fundamental Research Funds for the Central Universities,CHD(300102261107)。
文摘Reverse time migration and full waveform inversion involve the crosscorrelation of two wavefields,propagated in the forward-and reverse-time directions,respectively.As a result,the forward-propagated wavefield needs to be stored,and then accessed to compute the correlation with the backward-propagated wavefield.Boundary-value methods reconstruct the source wavefield using saved boundary wavefields and can significantly reduce the storage requirements.However,the existing boundary-value methods are based on the explicit finite-difference(FD)approximations of the spatial derivatives.Implicit FD methods exhibit greater accuracy and thus allow for a smaller operator length.We develop two(an accuracy-preserving and a memory-efficient)wavefield reconstruction schemes based on an implicit staggered-grid FD(SFD)operator.The former uses boundary wavefields at M layers of grid points and the spatial derivatives of wavefields at one layer of grid points to reconstruct the source wavefield for a(2M+2)th-order implicit SFD operator.The latter applies boundary wavefields at N layers of grid points,a linear combination of wavefields at M–N layers of grid points,and the spatial derivatives of wavefields at one layer of grid points to reconstruct the source wavefield(0≤N<M).The required memory of accuracy-preserving and memory-efficient schemes is(M+1)/M and(N+2)/M times,respectively,that of the explicit reconstruction scheme.Numerical results reveal that the accuracy-preserving scheme can achieve accurate reconstruction at the cost of storage.The memory-efficient scheme with N=2 can obtain plausible reconstructed wavefields and images,and the storage amount is 4/(M+1)of the accuracy-preserving scheme.
基金The authors thank the funds supported by the China National Nuclear Corporation under Grants Nos.WUQNYC2101 and WUHTLM2101-04National Natural Science Foundation of China(42074132,42274154).
文摘3D eikonal equation is a partial differential equation for the calculation of first-arrival traveltimes and has been widely applied in many scopes such as ray tracing,source localization,reflection migration,seismic monitoring and tomographic imaging.In recent years,many advanced methods have been developed to solve the 3D eikonal equation in heterogeneous media.However,there are still challenges for the stable and accurate calculation of first-arrival traveltimes in 3D strongly inhomogeneous media.In this paper,we propose an adaptive finite-difference(AFD)method to numerically solve the 3D eikonal equation.The novel method makes full use of the advantages of different local operators characterizing different seismic wave types to calculate factors and traveltimes,and then the most accurate factor and traveltime are adaptively selected for the convergent updating based on the Fermat principle.Combined with global fast sweeping describing seismic waves propagating along eight directions in 3D media,our novel method can achieve the robust calculation of first-arrival traveltimes with high precision at grid points either near source point or far away from source point even in a velocity model with large and sharp contrasts.Several numerical examples show the good performance of the AFD method,which will be beneficial to many scientific applications.
基金This work was supported by the National Natural Science Foundation of China(Grant No.12172160)Shenzhen Science and Technology Program(Grant No.JCYJ20220818100600002)+1 种基金South-ern University of Science and Technology(Grant No.Y01326127)the Department of Science and Technology of Guangdong Province(Grant Nos.2020B1212030001 and 2021QN020642).
文摘We propose a suite of strategies for the parallel solution of fully implicit monolithic fluid-structure interaction(FSI).The solver is based on a modeling approach that uses the velocity and pressure as the primitive variables,which offers a bridge between computational fluid dynamics(CFD)and computational structural dynamics.The spatiotemporal discretization leverages the variational multiscale formulation and the generalized-αmethod as a means of providing a robust discrete scheme.In particular,the time integration scheme does not suffer from the overshoot phenomenon and optimally dissipates high-frequency spurious modes in both subproblems of FSI.Based on the chosen fully implicit scheme,we systematically develop a combined suite of nonlinear and linear solver strategies.Invoking a block factorization of the Jacobian matrix,the Newton-Raphson procedure is reduced to solving two smaller linear systems in the multi-corrector stage.The first is of the elliptic type,indicating that the algebraic multigrid method serves as a well-suited option.The second exhibits a two-by-two block structure that is analogous to the system arising in CFD.Inspired by prior studies,the additive Schwarz domain decomposition method and the block-factorization-based preconditioners are invoked to address the linear problem.Since the number of unknowns matches in both subdomains,it is straightforward to balance loads when parallelizing the algorithm for distributed-memory architectures.We use two representative FSI benchmarks to demonstrate the robustness,efficiency,and scalability of the overall FSI solver framework.In particular,it is found that the developed FSI solver is comparable to the CFD solver in several aspects,including fixed-size and isogranular scalability as well as robustness.
基金supported by the National Key R&D Programme of China(2022YFF0902200).
文摘Realistic human reconstruction embraces an extensive range of applications as depth sensors advance.However,current stateof-the-art methods with RGB-D input still suffer from artefacts,such as noisy surfaces,non-human shapes,and depth ambiguity,especially for the invisible parts.The authors observe the main issue is the lack of geometric semantics without using depth input priors fully.This paper focuses on improving the representation ability of implicit function,exploring an effective method to utilise depth-related semantics effectively and efficiently.The proposed geometry-enhanced implicit function enhances the geometric semantics with the extra voxel-aligned features from point clouds,promoting the completion of missing parts for unseen regions while preserving the local details on the input.For incorporating multi-scale pixel-aligned and voxelaligned features,the authors use the Squeeze-and-Excitation attention to capture and fully use channel interdependencies.For the multi-view reconstruction,the proposed depth-enhanced attention explicitly excites the network to“sense”the geometric structure for a more reasonable feature aggregation.Experiments and results show that our method outperforms current RGB and depth-based SOTA methods on the challenging data from Twindom and Thuman3.0,and achieves a detailed and completed human reconstruction,balancing performance and efficiency well.
基金co-supported by the Aeronautical Science Foundation of China(Nos.2018ZA52002 and 2019ZA052011)。
文摘The accuracy of numerical computation heavily relies on appropriate meshing,whichserves as the foundation for numerical computation.Although adaptive refinement methods areavailable,an adaptive numerical solution is likely to be ineffective if it originates from a poorly ini-tial mesh.Therefore,it is crucial to generate meshes that accurately capture the geometric features.As an indispensable input in meshing methods,the Mesh Size Function(MSF)determines the qual-ity of the generated mesh.However,the current generation of MSF involves human participation tospecify numerous parameters,leading to difficulties in practical usage.Considering the capacity ofmachine learning to reveal the latent relationships within data,this paper proposes a novel machinelearning method,Implicit Geometry Neural Network(IGNN),for automatic prediction of appro-priate MSFs based on the existing mesh data,enabling the generation of unstructured meshes thatalign precisely with geometric features.IGNN employs the generative adversarial theory to learnthe mapping between the implicit representation of the geometry(Signed Distance Function,SDF)and the corresponding MSF.Experimental results show that the proposed method is capableof automatically generating appropriate meshes and achieving comparable meshing results com-pared to traditional methods.This paper demonstrates the possibility of significantly decreasingthe workload of mesh generation using machine learning techniques,and it is expected to increasethe automation level of mesh generation.
基金funded by the China Chongqing Municipal Science and Technology Bureau,grant numbers CSTB2024TIAD-CYKJCXX0009,CSTB2024NSCQ-LZX0043,CSTB2022NSCQ-MSX0288Chongqing Municipal Commission of Housing and Urban-Rural Development,grant number CKZ2024-87+3 种基金the Chongqing University of Technology Graduate Education High-Quality Development Project,grant number gzlsz202401the Chongqing University of Technology—Chongqing LINGLUE Technology Co.,Ltd.Electronic Information(Artificial Intelligence)Graduate Joint Training Basethe Postgraduate Education and Teaching Reform Research Project in Chongqing,grant number yjg213116the Chongqing University of Technology-CISDI Chongqing Information Technology Co.,Ltd.Computer Technology Graduate Joint Training Base.
文摘Although conventional object detection methods achieve high accuracy through extensively annotated datasets,acquiring such large-scale labeled data remains challenging and cost-prohibitive in numerous real-world applications.Few-shot object detection presents a new research idea that aims to localize and classify objects in images using only limited annotated examples.However,the inherent challenge in few-shot object detection lies in the insufficient sample diversity to fully characterize the sample feature distribution,which consequently impacts model performance.Inspired by contrastive learning principles,we propose an Implicit Feature Contrastive Learning(IFCL)module to address this limitation and augment feature diversity for more robust representational learning.This module generates augmented support sample features in a mixed feature space and implicitly contrasts them with query Region of Interest(RoI)features.This approach facilitates more comprehensive learning of both intra-class feature similarity and inter-class feature diversity,thereby enhancing the model’s object classification and localization capabilities.Extensive experiments on PASCAL VOC show that our method achieves a respective improvement of 3.2%,1.8%,and 2.3%on 10-shot of three Novel Sets compared to the baseline model FPD.
基金This research was supported by the National Nature Science Foundation of China （No. 41074100） and the Program for NewCentury Excellent Talents in the University of the Ministry of Education of China （No. NCET- 10-0812）.
文摘Prestack reverse time migration （RTM） is an accurate imaging method ofsubsurface media. The viscoacoustic prestack RTM is of practical significance because itconsiders the viscosity of the subsurface media. One of the steps of RTM is solving thewave equation and extrapolating the wave field forward and backward; therefore, solvingaccurately and efficiently the wave equation affects the imaging results and the efficiencyof RTM. In this study, we use the optimal time-space domain dispersion high-order finite-difference （FD） method to solve the viscoacoustic wave equation. Dispersion analysis andnumerical simulations show that the optimal time-space domain FD method is more accurateand suppresses the numerical dispersion. We use hybrid absorbing boundary conditions tohandle the boundary reflection. We also use source-normalized cross-correlation imagingconditions for migration and apply Laplace filtering to remove the low-frequency noise.Numerical modeling suggests that the viscoacoustic wave equation RTM has higher imagingresolution than the acoustic wave equation RTM when the viscosity of the subsurface isconsidered. In addition, for the wave field extrapolation, we use the adaptive variable-lengthFD operator to calculate the spatial derivatives and improve the computational efficiencywithout compromising the accuracy of the numerical solution.
基金supported by the National Natural Science Foundation of China(No.41474110)Shell Ph.D. Scholarship to support excellence in geophysical research
文摘Generally, FD coefficients can be obtained by using Taylor series expansion （TE） or optimization methods to minimize the dispersion error. However, the TE-based FD method only achieves high modeling precision over a limited range of wavenumbers, and produces large numerical dispersion beyond this range. The optimal FD scheme based on least squares （LS） can guarantee high precision over a larger range of wavenumbers and obtain the best optimization solution at small computational cost. We extend the LS-based optimal FD scheme from two-dimensional （2D） forward modeling to three-dimensional （3D） and develop a 3D acoustic optimal FD method with high efficiency, wide range of high accuracy and adaptability to parallel computing. Dispersion analysis and forward modeling demonstrate that the developed FD method suppresses numerical dispersion. Finally, we use the developed FD method to source wavefield extrapolation and receiver wavefield extrapolation in 3D RTM. To decrease the computation time and storage requirements, the 3D RTM is implemented by combining the efficient boundary storage with checkpointing strategies on GPU. 3D RTM imaging results suggest that the 3D optimal FD method has higher precision than conventional methods.
文摘To improve the accuracy of the conventional finite-difference method, finitedifference numerical modeling methods of any even-order accuracy are recommended. We introduce any even-order accuracy difference schemes of any-order derivatives derived from Taylor series expansion. Then, a finite-difference numerical modeling method with any evenorder accuracy is utilized to simulate seismic wave propagation in two-phase anisotropic media. Results indicate that modeling accuracy improves with the increase of difference accuracy order number. It is essential to find the optimal order number, grid size, and time step to balance modeling precision and computational complexity. Four kinds of waves, static mode in the source point, SV wave cusps, reflection and transmission waves are observed in two-phase anisotropic media through modeling.
基金sponsed by National Natural Science Foundation of China(NSFC,Grant No.41304077)the Natural Basic Research Program of China(the“973 Project,”Grant No.2013CB733303)Postdoctoral Science Foundation of China(Grant No.2014T70740)
文摘During seismic wave propagation on a free surface, a strong material contrast boundary develops in response to interference by P- and S- waves to create a surface-wave phenomenon. To accurately determine the effects of this interface on surface-wave propagation, the boundary conditions must be accurately modeled. In this paper, we present a numerical approach based on the dynamic poroelasticity for a space–time-domain staggered-grid finite-difference simulation in porous media that contain a free-surface boundary. We propose a generalized stess mirror formulation of the free-surface boundary for solids and fluids in porous media for the grid mesh on which lays the free-surface plane. Its analog is that used for elastic media, which is suitable for precise and stable Rayleigh-type surface-wave modeling. The results of our analysis of first kind of Rayleigh （R1） waves obtained by this model demonstrate that the discretization of the mesh in a similar way to that for elastic media can realize stable numerical solutions with acceptable precision. We present numerical examples demonstrating the efficiency and accuracy of our proposed method.
基金supported by the 863 Program (Grant no.2006AA09Z323)the 973 Program (Grant No.2006CB202402)
文摘Numerical simulation in the frequency-space domain has inherent advantages, such as： it is possible to simulate wave propagation from multiple sources simultaneously; there are no cumulative errors; only the interesting frequencies can be selected; and it is more suitable for wave propagation in viscoelastic media. The only obstacle to using the method is the requirement of huge computer storage. We extend the compressed format for storing the coefficient matrix. It can reduce the required computer storage dramatically. We get the optimal coefficients by least-squares method to suppress the numerical dispersion and adopt the perfectly matched layer （PML） boundary conditions to eliminate the artificial boundary reflections. Using larger grid intervals decreases computer storage requirements and provides high computational efficiency. Numerical experiments demonstrate that these means are economic and effective, providing a good basis for elastic wave imaging and inversion.
基金supported by the Natural Science Foundation of China(No.41374078)Geological Survey Projects of Ministry of Land and Resources of China(No.12120113086100 and 12120113101300)
文摘The Z-Axis tiPPer eiectromagnetic （ZTEM） technique is based on a frequency-domain airbome electromagnetic system that measures the natural magnetic field. A survey area was divided into several blocks by using the Maxwell＇s equations, and the magnetic components at the center of each edge of the grid cell are evaluated by applying the staggered-grid finite-difference method. The tipper and its divergence are derived to complete the 3D ZTEM forward modeling algorithm. A synthetic model is then used to compare the responses with those of 2D finite-element forward modeling to verify the accuracy of the algorithm. ZTEM offers high horizontal resolution to both simple and complex distributions of conductivity. This work is the theoretical foundation for the interpretation of ZTEM data and the study of 3D ZTEM inversion.
基金jointly supported by the NSF(No.41720104006)the Strategic Priority Research Program of the Chinese Academy of Sciences(A)(No.XDA14010303)+2 种基金the National Oil and Gas Project(Nos.2016ZX05002-005-007HZ and 2016ZX05014-001-008HZ)the Shandong Innovation Project(No.2017CXGC1602)the Qingdao Innovation Project(Nos.16-5-1-40-jch and 17CX05011)
文摘The staggered-grid finite-difference （SGFD） method has been widely used in seismic forward modeling. The precision of the forward modeling results directly affects the results of the subsequent seismic inversion and migration. Numerical dispersion is one of the problems in this method. The window function method can reduce dispersion by replacing the finite-difference operators with window operators, obtained by truncating the spatial convolution series of the pseudospectral method. Although the window operators have high precision in the low-wavenumber domain, their precision decreases rapidly in the high-wavenumber domain. We develop a least squares optimization method to enhance the precision of operators obtained by the window function method. We transform the SGFD problem into a least squares problem and find the best solution iteratively. The window operator is chosen as the initial value and the optimized domain is set by the error threshold. The conjugate gradient method is also adopted to increase the stability of the solution. Approximation error analysis and numerical simulation results suggest that the proposed method increases the precision of the window function operators and decreases the numerical dispersion.
文摘Numerical simulation in transverse isotropic media with tilted symmetry axis（TTI） using the standard staggered-grid finite-difference scheme（SSG）results in errors caused by averaging or interpolation. In order to eliminate the errors, a method of rotated staggered-grid finite-difference scheme（RSG） is proposed. However, the RSG brings serious numerical dispersion. The compact staggered-grid finite-difference scheme（CSG） is an implicit difference scheme, which use fewer grid points to suppress dispersion more effectively than the SSG. This paper combines the CSG with the RSG to derive a rotated staggered-grid compact finite-difference scheme（RSGC）. The numerical experiments indicate that the RSGC has weaker numerical dispersion and better accuracy than the RSG.
基金National Natural Science Foundation of China (49894190-024) and Geophysical Prospecting Key Laboratory Foundation of China National Petroleum Corporation.
文摘To the most of velocity fields, the traveltimes of the first break that seismic waves propagate along rays can be computed on a 2-D or 3-D numerical grid by finite-difference extrapolation. Under ensuring accuracy, to improve calculating efficiency and adaptability, the calculation method of first-arrival traveltime of finite-difference is de- rived based on any rectangular grid and a local plane wavefront approximation. In addition, head waves and scat- tering waves are properly treated and shadow and caustic zones cannot be encountered, which appear in traditional ray-tracing. The testes of two simple models and the complex Marmousi model show that the method has higher accuracy and adaptability to complex structure with strong vertical and lateral velocity variation, and Kirchhoff prestack depth migration based on this method can basically achieve the position imaging effects of wave equation prestack depth migration in major structures and targets. Because of not taking account of the later arrivals energy, the effect of its amplitude preservation is worse than that by wave equation method, but its computing efficiency is higher than that by total Green′s function method and wave equation method.
基金the National Natural Science Foundation of China(No.41874144)the Research Foundation of China University of PetroleumBeijing at Karamay(RCYJ2018A-01-001).
文摘Conventional finite-difference(FD)methods cannot model acoustic wave propagation beyond Courant-Friedrichs-Lewy(CFL)numbers 0.707 and 0.577 for two-dimensional(2D)and three-dimensional(3D)equal spacing cases,respectively,thereby limiting time step selection.Based on the definition of temporal and spatial FD operators,we propose a variable-length temporal and spatial operator strategy to model wave propagation beyond those CFL numbers while preserving accuracy.First,to simulate wave propagation beyond the conventional CFL stability limit,the lengths of the temporal operators are modified to exceed the lengths of the spatial operators for high-velocity zones.Second,to preserve the modeling accuracy,the velocity-dependent lengths of the temporal and spatial operators are adaptively varied.The maximum CFL numbers for the proposed method can reach 1.25 and 1.0 in high velocity contrast 2D and 3D simulation examples,respectively.We demonstrate the effectiveness of our method by modeling wave propagation in simple and complex media.
