CN106646593B - A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel - Google Patents
A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel Download PDFInfo
- Publication number
- CN106646593B CN106646593B CN201610768430.7A CN201610768430A CN106646593B CN 106646593 B CN106646593 B CN 106646593B CN 201610768430 A CN201610768430 A CN 201610768430A CN 106646593 B CN106646593 B CN 106646593B
- Authority
- CN
- China
- Prior art keywords
- calculating
- node
- parallel
- communication
- relief surface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000008569 process Effects 0.000 claims abstract description 34
- 238000004891 communication Methods 0.000 claims abstract description 28
- 241001269238 Data Species 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000000903 blocking effect Effects 0.000 abstract description 7
- 238000013508 migration Methods 0.000 description 6
- 230000005012 migration Effects 0.000 description 6
- 238000004088 simulation Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 238000003384 imaging method Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005755 formation reaction Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004513 sizing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/64—Geostructures, e.g. in 3D data cubes
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The present invention relates to technical field of geophysical exploration, the parallel three-dimensional relief surface Acoustic Forward Modeling method of especially a kind of cross-node, include the following steps, model meshes is divided into different regions along dimension most slow in memory, different MPI processes handle the calculating in different zones；In the time stepping method of each step, the request of non-obstruction point-to-point communication is initiated by each node first, for the wavefield data on adjacent MPI process exchange areas boundary；Each process immediately begins to the calculating for carrying out the regions PML after non-blocking communication request returns, and the three dimensions cycle being directed to realizes multithreads computing using OpenMP；The method of the present invention uses non-blocking communication and realizes data exchange, covers communication delay using calculating, reaches relatively high parallel efficiency.
Description
Technical field
The present invention relates to technical field of geophysical exploration, the parallel three-dimensional relief surface sound wave of especially a kind of cross-node
The Forward Modeling.
Background technology
Seismic prospecting is to cause the earth's crust to vibrate first with artificial method, then utilizes seismic wave from subsurface formations interface
Reflected hourage and waveform variation information, heuristically under stratigraphic structure and lithology a kind of method.Classical
Seismic prospecting theory often assumes that earth's surface is horizontal, the seismic data process based on this approximating assumption and migration imaging skill
Art has obtained extraordinary application in the area for much meeting approximate condition.With the continuous development of seismic exploration technique, oil gas
The emphasis of seismic prospecting is just turning to the region of relief surface condition and complex geological condition.These regional surface relieves are big, underground
Construction is complicated, and does not meet the approximating assumption of flatly table.Therefore, the migration and imaging techniques under the conditions of research relief surface have
Important meaning.The migration and imaging techniques of mainstream have Kirchhoff offsets, one-way wave depth migration, reverse-time migration etc. at present.Its
Whole wave field information is utilized due to direct solution round trip wave equation in middle reverse-time migration, is imaging method the most accurate,
Core algorithm is exactly seismic forward simulation algorithm.
With the development of field seismic data acquisition, the data volume of seismic data is increasing so that oil-gas exploration
The performance requirement of memory requirements and data processing application software to high-performance computing environment is growing, especially inclined with the inverse time
Move and full waveform inversion be representative the treatment technology based on wave equation application, even more to earthquake Forward Modeling and Inversion algorithm across
Nodal parallel conceptual design proposes great challenge.
Invention content
In order to overcome above-mentioned technical problem, the present invention to provide a kind of three-dimensional relief surface sound wave forward modeling mould that cross-node is parallel
Quasi- method uses the parallel organization based on shared drive in intra-node, is additionally opened caused by avoid multiple MPI processes are started
Pin, and reduce unnecessary interprocess communication expense in node.Between node, method of the invention uses non-blocking communication
It realizes data exchange, covers communication delay using calculating, reach relatively high parallel efficiency.
The technical solution adopted by the present invention to solve the technical problems is：A kind of three-dimensional relief surface sound that cross-node is parallel
Wave the Forward Modeling, includes the following steps：
(1) model meshes are divided into different regions along dimension most slow in memory, the processing of different MPI processes is not
With the calculating in region.
(2) in the time stepping method of each step, the request of non-obstruction point-to-point communication is initiated by each node first, for
The wavefield data on adjacent MPI process exchange areas boundary.
(3) each process immediately begins to the calculating for carrying out the regions PML, is directed to after non-blocking communication request returns
Three dimensions cycle multithreads computing is realized using OpenMP, which is implemented as：
1. being calculated first with order central difference scheme：
Such as 4 the central difference schemes of rank be：
2. updating θxu
3. updating ηxu
4. updating ψxu
5. calculating partial derivative
At this point, data communication and the calculating of absorbing boundary are carried out at the same time, process each in this way carries out intensive calculating task
The delay of communication is covered, to improve the parallel efficiency of multinode.
(4) after the completion of the pending datas communication such as each process, the close of space partial derivative is calculated using order central difference scheme
Like value
Then the wave field of subsequent time is updated,
If spatial position is located at relief surface or more, the wave field value at this is set to 0.
(5) wave field is modified using the result that the regions PML calculate in borderline region.All cycles that the part is related to
Also OpenMP is used to realize parallel.
(6) finally in a manner of exchanging pointer, the wave field at new and old moment is exchanged, and the hough transformation exported will be needed to arrive
In host process.
(7) then calculating of all processes into next time step.
(8) after the calculating for completing all time steps, earthquake record is exported.
The invention has the advantages that the parallel method compared to pure MPI, method of the invention is used in intra-node and is based on
The parallel organization of shared drive to avoid overhead caused by multiple MPI processes is started, and is reduced unnecessary in node
Interprocess communication expense.Between node, method of the invention uses non-blocking communication and realizes data exchange, is covered using calculating
Communication delay reaches relatively high parallel efficiency.
Description of the drawings
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 is the strong scalability test result of this method；
Fig. 2 is the weak scalability test result of this method；
Fig. 3 is parallel acceleration curve of this method under different nodes；
Fig. 4 is the weak scalability curve of this method.
Specific implementation mode
Embodiment 1
(1) parameter of forward simulation is configured first in configuration file：
(2) using Madagascar Software Create standard Ricker wavelet time point ranges as focus：
(3) velocity field of the Madagascar Software Creates for forward simulation is utilized：
(4) parameter of observation system is configured：
Then the input file of Madagascar Software Create observation systems is utilized：
(5) after getting out the forward modeling parameter in step (1), (2), (3), program Run Script is write
Script is submitted with parallel task
Wherein N is number of nodes, and n is that in this test, will be continuously increased the number of N and n into number of passes, count forward simulation
The required time observes change curve of the program runtime with interstitial content.According to this hair after operation task submission script
The forward modeling program that the method for bright proposition is realized begins to be simulated, i.e.,：
(6) model meshes are divided into different regions along dimension most slow in memory, and the processing of different MPI processes is not
With the calculating in region.
(7) in the time stepping method of each step, the request of non-obstruction point-to-point communication is initiated by each node first, for
The wavefield data on adjacent MPI process exchange areas boundary.
(8) each process immediately begins to the calculating for carrying out the regions PML, is directed to after non-blocking communication request returns
Three dimensions cycle multithreads computing is realized using OpenMP, which is implemented as：
1. being calculated first with order central difference scheme：
Such as 4 the central difference schemes of rank be：
2. updating θxu
3. updating ηxu
4. updating ψxu
5. calculating partial derivative
At this point, data communication and the calculating of absorbing boundary are carried out at the same time, process each in this way carries out intensive calculating task
The delay of communication is covered, to improve the parallel efficiency of multinode.
(9) after the completion of the pending datas communication such as each process, the close of space partial derivative is calculated using order central difference scheme
Like value
Then the wave field of subsequent time is updated
If spatial position is located at relief surface or more, the wave field value at this is set to 0.
(10) wave field is modified using the result that the regions PML calculate in borderline region.What the part was related to all follows
It is parallel that ring also uses OpenMP to realize.
(11) finally in a manner of exchanging pointer, the wave field at new and old moment, and the hough transformation that needs are exported are exchanged
Onto host process.
(12) all processes enter the calculating of next time step.
(13) after the calculating for completing all time steps, earthquake record is exported.
Embodiment 2
(1) parameter of forward simulation is configured first in configuration file：
T=1.0
Dt=0.001
F0=15
Order=8
Jdata=2
Jsnap=2
Fm=2.5*f0
Dx=0.01
Dy=0.01
Dz=0.01
Nxx=1000
Nyy=300
Nzz=1000
In this test, the scale (the number nyy for increasing the directions y sampled point) for the problem that is continuously increased increases simultaneously
Number of processes ensures that the calculating sizing grid that each course allocation arrives is consistent, then counts the time needed for forward simulation, observation
Program runtime with change curve.
(2) using Madagascar Software Create standard Ricker wavelet time point ranges as focus：
(3) velocity field of the Madagascar Software Creates for forward simulation is utilized：
(4) parameter of observation system is configured：
Then the input file of Madagascar Software Create observation systems is utilized：
(5) after getting out the forward modeling parameter in step (1), (2), (3), program Run Script is write
Script is submitted with parallel task
Wherein N is number of nodes, and n is into number of passes, and number of processes n will be according to the size of rate pattern grid scale in this test
To determine.The forward modeling program that operation task submits the method proposed according to the present invention after script to realize begins to be simulated, i.e.,：
(6) model meshes are divided into different regions along dimension most slow in memory, and the processing of different MPI processes is not
With the calculating in region.
(7) in the time stepping method of each step, the request of non-obstruction point-to-point communication is initiated by each node first, for
The wavefield data on adjacent MPI process exchange areas boundary.
(8) each process immediately begins to the calculating for carrying out the regions PML, is directed to after non-blocking communication request returns
Three dimensions cycle multithreads computing is realized using OpenMP, which is implemented as：
1. being calculated first with order central difference scheme：
Such as 4 the central difference schemes of rank be：
2. updating θxu
3. updating ηxu
4. updating ψxu
5. calculating partial derivative
At this point, data communication and the calculating of absorbing boundary are carried out at the same time, process each in this way carries out intensive calculating task
The delay of communication is covered, to improve the parallel efficiency of multinode.
(9) after the completion of the pending datas communication such as each process, the close of space partial derivative is calculated using order central difference scheme
Like value,
Then the wave field of subsequent time is updated,
If spatial position is located at relief surface or more, the wave field value at this is set to 0.
(10) wave field is modified using the result that the regions PML calculate in borderline region.What the part was related to all follows
It is parallel that ring also uses OpenMP to realize.
(11) finally in a manner of exchanging pointer, the wave field at new and old moment, and the hough transformation that needs are exported are exchanged
Onto host process.
(12) all processes enter the calculating of next time step.
(13) after the calculating for completing all time steps, earthquake record is exported.
Claims (1)
1. a kind of parallel three-dimensional relief surface Acoustic Forward Modeling method of cross-node, which is characterized in that include the following steps：
(1) model meshes are divided into different regions along dimension most slow in memory, different MPI processes handle not same district
Calculating in domain；
(2) in the time stepping method of each step, the request of non-obstruction point-to-point communication is initiated by each node first, for adjacent
MPI process exchange areas boundary wavefield data；
(3) each process immediately begins to the calculating for carrying out the regions PML, wherein relating to after the request of non-obstruction point-to-point communication returns
And three dimensions cycle using OpenMP realize multithreads computing；
(4) after the completion of the pending datas communication such as each process, the approximation of space partial derivative is calculated using order central difference scheme,
Then the wave field value at this is set to 0 by the wave field of update subsequent time if spatial position is located at relief surface or more；
(5) wave field is modified using the result that the regions PML calculate in borderline region, all cycles which is related to also are adopted
It is realized with OpenMP parallel；
(6) finally in a manner of exchanging pointer, exchange the wave field at new and old moment, and will need the hough transformation exported to lead into
Cheng Shang；
(7) all processes enter the calculating of next time step；
(8) after the calculating for completing all time steps, earthquake record is exported.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610768430.7A CN106646593B (en) | 2016-08-30 | 2016-08-30 | A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610768430.7A CN106646593B (en) | 2016-08-30 | 2016-08-30 | A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106646593A CN106646593A (en) | 2017-05-10 |
| CN106646593B true CN106646593B (en) | 2018-10-26 |
Family
ID=58853089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610768430.7A Active CN106646593B (en) | 2016-08-30 | 2016-08-30 | A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106646593B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107561585A (en) * | 2017-09-19 | 2018-01-09 | 北京大学 | A kind of multinuclear multi-node parallel 3-D seismics wave field generation method and system |
| CN111766623B (en) * | 2020-05-25 | 2023-01-31 | 国家超级计算天津中心 | Seismic wave field three-dimensional forward modeling method based on MPI process topology |
| CN113239522B (en) * | 2021-04-20 | 2022-06-28 | 四川大学 | Atmospheric pollutant diffusion simulation method based on computer cluster |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6687659B1 (en) * | 2000-03-24 | 2004-02-03 | Conocophillips Company | Method and apparatus for absorbing boundary conditions in numerical finite-difference acoustic applications |
| JP2008234335A (en) * | 2007-03-20 | 2008-10-02 | Tokyo Metropolitan Univ | Electromagnetic field analysis device, electromagnetic field analysis method, electromagnetic field analysis program and field analysis device |
| CN103278848A (en) * | 2013-04-22 | 2013-09-04 | 中山大学 | Seismic imaging forward modeling method based on massage passing interface (MPI) parallel preconditioned iteration |
| CN105277980A (en) * | 2014-06-26 | 2016-01-27 | 中石化石油工程地球物理有限公司胜利分公司 | High-precision spatial and temporal arbitrary multiple variable grid finite difference forward modeling method |
| CN105717539A (en) * | 2016-01-28 | 2016-06-29 | 中国地质大学（北京） | Multi GPU calculation based reverse time migration imaging method of 3D TTI medium |
-
2016
- 2016-08-30 CN CN201610768430.7A patent/CN106646593B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6687659B1 (en) * | 2000-03-24 | 2004-02-03 | Conocophillips Company | Method and apparatus for absorbing boundary conditions in numerical finite-difference acoustic applications |
| JP2008234335A (en) * | 2007-03-20 | 2008-10-02 | Tokyo Metropolitan Univ | Electromagnetic field analysis device, electromagnetic field analysis method, electromagnetic field analysis program and field analysis device |
| CN103278848A (en) * | 2013-04-22 | 2013-09-04 | 中山大学 | Seismic imaging forward modeling method based on massage passing interface (MPI) parallel preconditioned iteration |
| CN105277980A (en) * | 2014-06-26 | 2016-01-27 | 中石化石油工程地球物理有限公司胜利分公司 | High-precision spatial and temporal arbitrary multiple variable grid finite difference forward modeling method |
| CN105717539A (en) * | 2016-01-28 | 2016-06-29 | 中国地质大学（北京） | Multi GPU calculation based reverse time migration imaging method of 3D TTI medium |
Non-Patent Citations (1)
| Title |
|---|
| 基于GPU集群的大规模三维有限差分正演模拟并行策略;廉西猛 等;《物探与化探》;20150630;第39卷(第3期);第615-620页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106646593A (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Abdelkhalek et al. | Fast seismic modeling and reverse time migration on a GPU cluster | |
| CN104614763B (en) | Multi-wave AVO reservoir elastic parameter inversion method and system based on reflectivity method | |
| CN108181653B (en) | For VTI medium reverse-time migration method, equipment and medium | |
| CN105137486A (en) | Elastic wave reverse-time migration imaging method and apparatus in anisotropic media | |
| CN106526674A (en) | Three-dimensional full waveform inversion energy weighted gradient preprocessing method | |
| CN106646593B (en) | A kind of three-dimensional relief surface Acoustic Forward Modeling method that cross-node is parallel | |
| CN110031896A (en) | Earthquake stochastic inversion methods and device based on Multiple-Point Geostatistics prior information | |
| WO2022078514A1 (en) | Three-dimensional speed geological modeling method capable of random distribution of structures and wave speeds | |
| CN106501852A (en) | A kind of multiple dimensioned full waveform inversion method of three-dimensional acoustic wave equation arbitrarily-shaped domain and device | |
| CN109633745A (en) | A kind of drafting method and device of three-dimensional structural map | |
| Xue et al. | An efficient GPU implementation for locating micro-seismic sources using 3D elastic wave time-reversal imaging | |
| Rietmann et al. | Newmark local time stepping on high-performance computing architectures | |
| CN105005072A (en) | PML boundary three-dimensional seismic wave propagation simulation method utilizing CUDA | |
| CN106226841A (en) | A kind of fluvial facies three-dimensional sedimentary facies model Decided modelling method | |
| CN111638551A (en) | Seismic first-motion wave travel time chromatography method and device | |
| CN108508481B (en) | A kind of method, apparatus and system of longitudinal wave converted wave seismic data time match | |
| CN105807317B (en) | Anisotropy decaying surface wave analogy method based on Chebyshev's puppet spectrometry | |
| CN109709602A (en) | A kind of remote detection sonic wave shift imaging method, apparatus and system | |
| CN106662665A (en) | Re-ordered interpolation and convolution for faster staggered-grid processing | |
| CN108983290B (en) | It is a kind of three-dimensional Method in Transverse Isotropic Medium in travel when determine method and system | |
| CN111273346B (en) | Method, device, computer equipment and readable storage medium for removing deposition background | |
| Weinbub et al. | Shared-memory parallelization of the fast marching method using an overlapping domain-decomposition approach | |
| He et al. | Modeling 3-D elastic wave propagation in TI media using discontinuous Galerkin method on tetrahedral meshes | |
| CN110568497B (en) | Accurate solving method for seismic first-motion wave travel time under complex medium condition | |
| CN115880455A (en) | Three-dimensional intelligent interpolation method based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |