CN107656307B - Full waveform inversion calculation method and system - Google Patents
Full waveform inversion calculation method and system Download PDFInfo
- Publication number
- CN107656307B CN107656307B CN201610596765.5A CN201610596765A CN107656307B CN 107656307 B CN107656307 B CN 107656307B CN 201610596765 A CN201610596765 A CN 201610596765A CN 107656307 B CN107656307 B CN 107656307B
- Authority
- CN
- China
- Prior art keywords
- waveform inversion
- full waveform
- data
- shot point
- point data
- 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
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
- G01V1/282—Application of seismic models, synthetic seismograms
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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Disclose a kind of full waveform inversion calculation method and system.This method can include: be based on seismic data collected, obtain shot point data, and then obtain the parallel state file of the calculating state of record shot point data, pass through the full waveform inversion project of multiple calculate node Distributed Parallel Computing shot point data；The full waveform inversion project calculated multiple calculate nodes is overlapped, and obtains full waveform inversion project result；And the full waveform inversion project result renewal speed model based on shot point data, obtain full waveform inversion model, wherein, being overlapped to the full waveform inversion project that multiple calculate nodes calculate includes: to be grouped multiple calculate nodes, the full waveform inversion project for the shot point data that calculate node in every group calculates is overlapped, superposition of data is obtained；In the case where the quantity of superposition of data is equal to one, using the superposition of data of acquisition as the full waveform inversion project result of shot point data.
Description
Technical field
The present invention relates to oil-gas exploration application fields, more particularly, to a kind of Full wave shape based on parallel data superposition
Inversion Calculation method and system method and system.
Background technique
In oil-gas exploration application field, full waveform inversion method is believed using the kinematics and dynamics of pre-stack seismic wave field
Breath rebuilds earth formation, has the potentiality for disclosing construction and reservoir properties under complicated geological background.It was calculated in full waveform inversion
Cheng Zhong needs repeated multiple times iteration/loop inversion to calculate, and each Inversion Calculation includes to calculate gradient, linear search (containing repeatedly residual
The calculating of difference), renewal speed model.The processes such as gradient, linear search are calculated, need to rely on high-performance computer cluster
Parallel computation realize.Large-scale high-performance computer cluster is made of a large amount of computer nodes and shared storage device.Often
After completing gradient calculating or residual computations, calculated result is needed by being superimposed and being saved in shared storage device a node
On.There are many defects for the full waveform inversion realized with conventional MPI method, such as: do not support the dynamic of hardware device to increase
Adduction is reduced；Stability when great deal of nodes long-time numerical behavior is poor；Do not support heterogeneous device；Large-scale parallel computing inefficiency
Deng.The distributed parallel method of full waveform inversion technology can solve MPI method there are the problem of, still, in distributed computing frame
Under frame, the data of each calculate node cannot be by calling MPI application programming interfaces to realize superposition.
Inventors have found that how to solve full waveform inversion technology distributed computing realize during data file simultaneously
The problem of row superposition is a key factor for restricting full waveform inversion technical application.It is based on therefore, it is necessary to develop one kind
The full waveform inversion calculation method and system of the data investigation method of highly effective.
The information for being disclosed in background of invention part is merely intended to deepen the reason to general background technique of the invention
Solution, and it is known to those skilled in the art existing to be not construed as recognizing or imply that the information is constituted in any form
Technology.
Summary of the invention
The invention proposes a kind of full waveform inversion calculation method and systems, are superimposed by parallel data, reduce number
According to the network copy number and file stacking fold in additive process, the full waveform inversion parallel computation of efficient stable is realized.
According to an aspect of the invention, it is proposed that a kind of full waveform inversion calculation method.The method may include: it is based on
Seismic data collected obtains shot point data, and then obtains the parallel state text for the calculating state for recording the shot point data
Part passes through the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings；To the multiple calculating
The full waveform inversion project for the shot point data that node calculates is overlapped, and obtains the full waveform inversion item of the shot point data
Mesh result；And the full waveform inversion project result renewal speed model based on the shot point data, obtain full waveform inversion mould
Type, wherein being overlapped to the full waveform inversion project for the shot point data that the multiple calculate node calculates includes: by institute
Multiple calculate node groupings are stated, the full waveform inversion project for the shot point data that the calculate node in every group calculates is folded
Add, obtains superposition of data；In the case where the quantity of the superposition of data of acquisition is greater than one, by the superposition of data of acquisition
It is grouped and is overlapped respectively again；It, will be described in acquisition in the case where the quantity of the superposition of data of acquisition is equal to one
Full waveform inversion project result of the superposition of data as the shot point data.
According to another aspect of the invention, it is proposed that a kind of full waveform inversion computing system, the system may include: use
It is based on seismic data collected, obtains shot point data, and then obtain the parallel of the calculating state for recording the shot point data
Status file passes through the unit of the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings；With
It is overlapped come the full waveform inversion project of the shot point data calculated the multiple calculate node, obtains the shot point number
According to full waveform inversion project result unit；And it is updated for the full waveform inversion project result based on the shot point data
Rate pattern obtains the unit of full waveform inversion model, wherein the shot point data calculated the multiple calculate node
It includes: to be grouped the multiple calculate node that full waveform inversion project, which is overlapped, the institute calculated the calculate node in every group
The full waveform inversion project for stating shot point data is overlapped, and obtains superposition of data；It is big in the quantity of the superposition of data of acquisition
In the case where one, the superposition of data of acquisition is grouped again and is overlapped respectively；In the superposition of data of acquisition
Quantity be equal to one in the case where, using the superposition of data of acquisition as the full waveform inversion project knot of the shot point data
Fruit.
Full waveform inversion parallel calculating method according to the present invention, is superimposed by parallel data, the time required to realizing superposition
It substantially reduces, computational efficiency is obviously improved.
Methods and apparatus of the present invention has other characteristics and advantages, these characteristics and advantages are attached from what is be incorporated herein
It will be apparent in figure and subsequent specific embodiment, or will be in the attached drawing being incorporated herein and subsequent specific reality
It applies in mode and is stated in detail, the drawings and the detailed description together serve to explain specific principles of the invention.
Detailed description of the invention
Exemplary embodiment of the invention is described in more detail in conjunction with the accompanying drawings, it is of the invention above-mentioned and its
Its purpose, feature and advantage will be apparent, wherein in exemplary embodiment of the invention, identical reference label
Typically represent same parts.
Fig. 1 shows the flow chart of the step of full waveform inversion parallel calculating method according to the present invention.
Fig. 2 shows the schematic diagrames of parallel data stacking method according to embodiment of the present invention.
Fig. 3 shows a kind of schematic diagram of serial data stacking method according to prior art.
Specific embodiment
The present invention will be described in more detail below with reference to accompanying drawings.Although showing preferred implementation side of the invention in attached drawing
Formula, however, it is to be appreciated that may be realized in various forms the present invention without that should be limited by the embodiments set forth herein.Phase
Instead, these embodiments are provided so that the present invention is more thorough and complete, and can be by the scope of the present invention completely
It is communicated to those skilled in the art.
Embodiment 1
Fig. 1 shows the flow chart of the step of full waveform inversion parallel computation according to the present invention.
In this embodiment, full waveform inversion parallel calculating method according to the present invention may include: step 101, base
In seismic data collected, shot point data are obtained, and then obtain the parallel state for recording the calculating state of the shot point data
File passes through the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings；Step 102, to institute
The full waveform inversion project for stating the shot point data that multiple calculate nodes calculate is overlapped, and obtains the complete of the shot point data
Waveform inversion project result；And step 103, the full waveform inversion project result renewal speed mould based on the shot point data
Type obtains full waveform inversion model.Wherein, the full waveform inversion item of the shot point data the multiple calculate node calculated
Mesh is overlapped includes: to be grouped the multiple calculate node, the shot point data calculated the calculate node in every group
Full waveform inversion project is overlapped, and obtains superposition of data；In the case where the quantity of the superposition of data of acquisition is greater than one,
The superposition of data of acquisition is grouped again and is overlapped respectively；It is equal to one in the quantity of the superposition of data of acquisition
In the case of, using the superposition of data of acquisition as the full waveform inversion project result of the shot point data.
The embodiment is superimposed by parallel data, reduces network copy number during data investigation and file is folded
Add number, realizes the full waveform inversion parallel computation of efficient stable.
The following detailed description of the specific steps of full waveform inversion parallel calculating method according to the present invention.
In one example, seismic data collected, available shot point data are based on, and then obtains and records the big gun
The parallel state file of the calculating state of point data passes through the complete of shot point data described in multiple calculate node Distributed Parallel Computings
Waveform inversion project.
In one example, the full waveform inversion project of the shot point data may include the shot point data gradient with
Residual error.
In one example, the full waveform inversion project of the shot point data calculated the multiple calculate node can be with
It is overlapped, obtains the full waveform inversion project result of the shot point data.
In one example, the full waveform inversion project of the shot point data calculated the multiple calculate node carries out
Superposition may include: to be grouped the multiple calculate node, in every group calculate node calculate the shot point data it is complete
Waveform inversion project is overlapped, and obtains superposition of data；In the case where the superposition of data of acquisition is multiple, by acquisition
The superposition of data is grouped again and is overlapped respectively；In the case where the superposition of data of acquisition is one, will obtain
Full waveform inversion project result of the superposition of data as the shot point data.
In one example, the grouping of the multiple calculate node can be in pairs.
Fig. 2 shows the schematic diagrames of parallel data stacking method according to embodiment of the present invention.Fig. 3 is shown
The schematic diagram of serial data stacking method according to prior art.
As shown in Fig. 2, the quantity of calculate node can be 8, respectively g1, g2, g3, g4, g5, g6, g7, g8, according to
The calculate node can be grouped by method of the invention two-by-two, i.e., g1, g2 can be one group, and g3, g4 can be one group, g5,
G6 can be one group, and g7, g8 can be one group；By taking g1, g2 as an example, g1 is superimposed with the parallel data of g2, g1 can be stored in
In and be denoted as g12, i.e. superposition of data g12=g1+g2, and so on, obtain g34, g56, g78；Again two-by-two by calculate node
Grouping, i.e. g12, g34 can be one group, and g56, g78 can be one group, by taking g12, g34 as an example, by the parallel data of g12 and g34
Superposition, can be stored in g12 and be denoted as g1234, i.e. superposition of data g1234=g12+g34, and so on, obtain g5678；
G1234 is superimposed with the parallel data of g5678, can be stored in g1234 and be denoted as g12345678, i.e. superposition of data
G12345678=g1234+g5678, obtaining superposition of data is one, i.e., using the superposition of data g12345678 of acquisition as shot point
The full waveform inversion project result of shot point data as a result, and be transferred to parallel data superposition by the full waveform inversion project of data
Calculate node G.As it can be seen that according to the method for the present invention, parallel data superposition only needs 3 data investigations, 4 network transmissions.
As shown in figure 3, the quantity of calculate node can be 8, respectively g1, g2, g3, g4, g5, g6, g7, g8, according to
The method of the prior art is successively transferred to the calculate node G of parallel data superposition and is overlapped, i.e. G=g1+g2+g3+g4+
g5+g6+g7+g8.As it can be seen that method, parallel data superposition need 8 data investigations, 8 network transmissions according to prior art.
It can be seen that full waveform inversion parallel calculating method according to the present invention is than data investigation number needed for the prior art
It is substantially reduced with network transmission number, required time substantially reduces, and computational efficiency is obviously improved.
In one example, the full waveform inversion project result renewal speed model based on shot point data, in rate pattern
Update times be not up to predetermined number of times in the case where, full waveform inversion can be continued simultaneously according to updated rate pattern
Row calculates；And in the case where the update times of rate pattern reach predetermined number of times, the rate pattern that last time updates can
Using as final speed model, and then full waveform inversion model can be obtained, wherein predetermined number of times can be being manually set and
The total degree that row calculates.
Specifically, predetermined number of times can be set as 30 times, in the case where the update times of rate pattern are not up to 30 times,
Continue full waveform inversion parallel computation according to updated rate pattern；Reach 30 times in the update times of rate pattern
In the case of, the rate pattern that last time updates can be used as final speed model, and then obtain full waveform inversion model.
Using example
A concrete application example is given below in the scheme and its effect of embodiment of the present invention for ease of understanding.Ability
Field technique personnel should be understood that the example only for the purposes of understanding that the present invention, any detail are not intended in any way
The limitation present invention.
Specifically, by taking certain 3D full waveform inversion processing item as an example, which needs 1024 calculate nodes to calculate, all-wave
Shape inverting needs to carry out 30 Inversion Calculations, and each inverting averagely needs 4 parallel data superpositions, entire full waveform inversion process
120 parallel data superpositions are needed, each data volume size is 4GB.The time of each file network transmission is that (network speed was pressed in 40 seconds
100M/s is calculated), each data volume superposition time is 60 seconds.
Using parallel stacking method of the invention, each parallel data superposition needs 11 network transmissions, 10 stacked datas
Add, each parallel data superposition time is 17.3 minutes, and all data investigation holding times are 34.7 hours in entire project.
Using the serial stacking method of the prior art, each serial data superposition 1024 network transmissions of needs, 1024 times
Data investigation, each serial data are superimposed 28.4 hours of time, and data investigation holding time is 142.2 days in entire project.
It can be seen that full waveform inversion parallel calculating method according to the present invention the time required to the prior art than significantly subtracting
Few, computational efficiency is obviously improved.
It will be understood by those skilled in the art that above to the purpose of the description of embodiments of the present invention only for illustratively
The beneficial effect for illustrating embodiments of the present invention is not intended to for embodiments of the present invention to be limited to given any show
Example.
Embodiment 2
Embodiment there is provided a kind of full waveform inversion concurrent computational system, the systems for embodiment according to the present invention
It may include: to obtain shot point data, and then obtain the calculating for recording the shot point data for being based on seismic data collected
The parallel state file of state passes through the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings
Unit；The full waveform inversion project of the shot point data for calculating the multiple calculate node is overlapped, and is obtained
The unit of the full waveform inversion project result of the shot point data；And for the full waveform inversion item based on the shot point data
Mesh result renewal speed model obtains the unit of full waveform inversion model.Wherein, to described in the calculating of the multiple calculate node
It includes: to be grouped the multiple calculate node that the full waveform inversion project of shot point data, which is overlapped, to the calculating section in every group
The full waveform inversion project for the shot point data that point calculates is overlapped, and obtains superposition of data；In the superposition number of acquisition
According to quantity be greater than one in the case where, the superposition of data of acquisition is grouped again and is overlapped respectively；In the institute of acquisition
State superposition of data quantity be equal to one in the case where, the superposition of data of acquisition is anti-as the Full wave shape of the shot point data
Drill project result.
The embodiment is superimposed by parallel data, reduces network copy number during data investigation and file is folded
Add number, realizes the full waveform inversion parallel computation of efficient stable.
In one example, the full waveform inversion project of the shot point data may include the shot point data gradient with
Residual error.
In one example, the grouping of the multiple calculate node can be in pairs.
In one example, the full waveform inversion project result renewal speed model based on the shot point data obtains complete
Waveform inversion model may include: in the case where the update times of the rate pattern are not up to predetermined number of times, will be according to more
The rate pattern after new continues the full waveform inversion parallel computation；And the update times in the rate pattern
In the case where reaching predetermined number of times, the rate pattern that last time is updated obtains Full wave shape as final speed model
Inverse model, wherein the predetermined number of times is the total degree for the parallel computation being manually set.
It will be understood by those skilled in the art that above to the purpose of the description of embodiments of the present invention only for illustratively
The beneficial effect for illustrating embodiments of the present invention is not intended to for embodiments of the present invention to be limited to given any show
Example.
The embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is also not necessarily limited to disclosed each embodiment.It is right without departing from the scope and spirit of illustrated each embodiment
Many modifications and changes are obvious for those skilled in the art.The choosing of term used herein
It selects, it is intended to best explain the principle, practical application or the improvement to the technology in market of each embodiment, or make this technology
Other those of ordinary skill in field can understand each embodiment disclosed herein.
Claims (8)
1. a kind of full waveform inversion calculation method, comprising:
Based on seismic data collected, obtain shot point data, so obtain the calculating state for recording the shot point data and
Row status file passes through the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings；
The full waveform inversion project for the shot point data that the multiple calculate node calculates is overlapped, the shot point is obtained
The full waveform inversion project result of data；And
Full waveform inversion project result renewal speed model based on the shot point data obtains full waveform inversion model,
Wherein, the full waveform inversion project of the shot point data calculated the multiple calculate node, which is overlapped, includes:
The multiple calculate node is grouped, to the full waveform inversion item for the shot point data that the calculate node in every group calculates
Mesh is overlapped, and obtains superposition of data；
In the case where the superposition of data of acquisition is multiple, the superposition of data of acquisition is grouped again and is carried out respectively
Superposition；And
In the case where the superposition of data of acquisition is one, using the superposition of data of acquisition as the shot point data
Full waveform inversion project result.
2. full waveform inversion calculation method according to claim 1, wherein the full waveform inversion project of the shot point data
Gradient and residual error including the shot point data.
3. full waveform inversion calculation method according to claim 1, wherein the multiple calculate node is grouped into two-by-two
One group.
4. full waveform inversion calculation method according to claim 1, wherein the full waveform inversion based on the shot point data
Project result renewal speed model, obtaining full waveform inversion model includes:
It, will be according to the updated rate pattern in the case where the update times of the rate pattern are not up to predetermined number of times
Continue full waveform inversion parallel computation；And
In the case where the update times of the rate pattern reach predetermined number of times, using last time update rate pattern as
Final speed model, and then full waveform inversion model is obtained,
Wherein, the predetermined number of times is the total degree for the parallel computation being manually set.
5. a kind of full waveform inversion computing system, comprising:
For being based on seismic data collected, shot point data are obtained, and then obtain the calculating state for recording the shot point data
Parallel state file, pass through the list of the full waveform inversion project of shot point data described in multiple calculate node Distributed Parallel Computings
Member；
The full waveform inversion project of the shot point data for calculating the multiple calculate node is overlapped, described in acquisition
The unit of the full waveform inversion project result of shot point data；And
For the full waveform inversion project result renewal speed model based on the shot point data, full waveform inversion model is obtained
Unit,
Wherein, the full waveform inversion project of the shot point data calculated the multiple calculate node, which is overlapped, includes:
The multiple calculate node is grouped, to the full waveform inversion item for the shot point data that the calculate node in every group calculates
Mesh is overlapped, and obtains superposition of data；
In the case where the quantity of the superposition of data of acquisition is greater than one, the superposition of data of acquisition is grouped again and is divided
It is not overlapped；And
In the case where the quantity of the superposition of data of acquisition is equal to one, using the superposition of data of acquisition as the shot point
The full waveform inversion project result of data.
6. full waveform inversion computing system according to claim 5, wherein the full waveform inversion project of the shot point data
Gradient and residual error including the shot point data.
7. full waveform inversion computing system according to claim 5, wherein the multiple calculate node is grouped into two-by-two
One group.
8. full waveform inversion computing system according to claim 5, wherein the full waveform inversion based on the shot point data
Project result renewal speed model, obtaining full waveform inversion model includes:
It, will be according to the updated rate pattern in the case where the update times of the rate pattern are not up to predetermined number of times
Continue full waveform inversion parallel computation；And
In the case where the update times of the rate pattern reach predetermined number of times, using last time update rate pattern as
Final speed model, and then full waveform inversion model is obtained,
Wherein, the predetermined number of times is the total degree for the parallel computation being manually set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610596765.5A CN107656307B (en) | 2016-07-26 | 2016-07-26 | Full waveform inversion calculation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610596765.5A CN107656307B (en) | 2016-07-26 | 2016-07-26 | Full waveform inversion calculation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107656307A CN107656307A (en) | 2018-02-02 |
| CN107656307B true CN107656307B (en) | 2019-02-19 |
Family
ID=61126571
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610596765.5A Active CN107656307B (en) | 2016-07-26 | 2016-07-26 | Full waveform inversion calculation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107656307B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726354B (en) * | 2018-12-18 | 2022-12-20 | 山东交通学院 | A Parallel Cross-correlation Calculation Method for Massive Data |
| CN112578431B (en) * | 2019-09-27 | 2024-04-09 | 中国石油化工股份有限公司 | Method and system for storing full waveform inversion wave field optimization in finite state |
| CN111025388B (en) * | 2019-12-19 | 2020-10-30 | 河海大学 | A Multi-Wave Combined Prestack Waveform Inversion Method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9857488B2 (en) * | 2012-11-20 | 2018-01-02 | International Business Machines Corporation | Efficient wavefield compression in seismic imaging |
| CN103135132B (en) * | 2013-01-15 | 2015-07-01 | 中国科学院地质与地球物理研究所 | Hybrid-domain full wave form inversion method of central processing unit (CPU)/graphics processing unit (GPU) synergetic parallel computing |
| CN105738949B (en) * | 2016-03-01 | 2017-11-17 | 中国海洋石油总公司 | A kind of nine bin uniformity method for parallel processing for time-lapse seismic |
-
2016
- 2016-07-26 CN CN201610596765.5A patent/CN107656307B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107656307A (en) | 2018-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20190340510A1 (en) | Sparsifying neural network models | |
| CN107358293B (en) | Neural network training method and device | |
| Almgren et al. | Nyx: A massively parallel amr code for computational cosmology | |
| Gawedzki | Lectures on conformal field theory | |
| Zheng et al. | MIGO-NAS: Towards fast and generalizable neural architecture search | |
| Gallup et al. | Practical valence-bond calculations | |
| CN105319581B (en) | A kind of efficient time-domain full waveform inversion method | |
| Perez et al. | Ringo: Interactive graph analytics on big-memory machines | |
| CN107656307B (en) | Full waveform inversion calculation method and system | |
| Wang et al. | Scalable distributed dl training: Batching communication and computation | |
| Huang et al. | Quadratic regularization projected Barzilai–Borwein method for nonnegative matrix factorization | |
| CN106250101A (en) | Migration before stack method for parallel processing based on MapReduce and device | |
| CN112149808A (en) | Extending single-machine graph neural network training to distributed training methods, systems and media | |
| CN103473255A (en) | A data clustering method, system and data processing equipment | |
| CN103177414A (en) | Structure-based dependency graph node similarity concurrent computation method | |
| CN105403913A (en) | Pre-stack depth migration method and device | |
| CN104597499A (en) | Adjacent shot interference suppressing method and device for independent simultaneous sweeping seismic data of seismic vibrators | |
| CN106662665B (en) | The interpolation and convolution of rearrangement for the processing of faster staggered-mesh | |
| CN105929444A (en) | Cross-correlation migration and least square-based micro earthquake localization method | |
| Underwood et al. | Evostore: Towards scalable storage of evolving learning models | |
| WO2025112979A1 (en) | Parallel strategy optimal selection method, and neural network solver training method and apparatus | |
| CN109143371A (en) | A kind of noise remove method and device of seismic data | |
| Ji et al. | An apache spark implementation of block power method for computing dominant eigenvalues and eigenvectors of large-scale matrices | |
| CN108957536A (en) | Stratum inclination angle prediction method and device based on waveform matching inversion | |
| CN109214519A (en) | Data processing system, method and apparatus |
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 |