CN110188424B - Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation - Google Patents
Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation Download PDFInfo
- Publication number
- CN110188424B CN110188424B CN201910410152.1A CN201910410152A CN110188424B CN 110188424 B CN110188424 B CN 110188424B CN 201910410152 A CN201910410152 A CN 201910410152A CN 110188424 B CN110188424 B CN 110188424B
- Authority
- CN
- China
- Prior art keywords
- grid
- reconstruction
- flow field
- parallel
- edges
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Abstract
The invention discloses a local area grid reconstruction parallel method facing dynamic boundary flow field numerical simulation, which comprises the steps of parallelly finding out and marking units needing grid reconstruction for input deformed distributed grids; dividing the old mesh with holes by using a parallel region decomposition method, and then distributing the obtained sub-meshes to all available processors again; carrying out grid reconstruction on the distributed holes in parallel; and distributing new grids to different processors in parallel to balance the grid scales on the processors so as to obtain a new distributed grid. The method can make up the regret that the non-structural dynamic grid reconstruction method is not used in the existing large-scale dynamic boundary flow field parallel numerical simulation, can be well integrated into the existing dynamic boundary flow field simulation, realizes the complete parallelization of the dynamic boundary flow field simulation system, and overcomes the performance bottleneck in the aspects of memory consumption and computational efficiency.
Description
Technical Field
The invention relates to a local area grid reconstruction parallel method for numerical simulation of a dynamic boundary flow field, which is applied to the field of computational fluid mechanics (computational aerodynamics) research.
Background
In the field of computational fluid dynamics, many studies, particularly aerodynamic calculations, involve flow field simulations of dynamic boundaries (i.e., the shape or position of geometric boundaries may change over time). The dynamic unstructured grid approach is a common approach to solving this type of dynamic boundary problem-a uniform unstructured grid is used in each time step. In order to adapt to the change of the geometric boundary under different time steps, the grid nodes are moved along with the change of the geometric boundary, but the initial connection relation of the grid nodes is kept unchanged. However, this mesh deformation method sometimes results in poor mesh shape, and even when large displacements are involved in the simulation process, mesh cell flipping may occur. To further solve this problem, a relatively common method is to perform local area mesh reconstruction on the current poor quality mesh. The local area grid reconstruction method comprises the following steps: cutting some holes in local areas with deteriorated grid unit quality, then carrying out grid division again on the holes, carrying out interpolation calculation on the attribute values on new grid nodes through the solution on the old grid, and finally carrying out flow field calculation of the next time step by using the reconstructed grid. The local grid region reconstruction method is a general method with higher efficiency and robustness for solving the problem of dynamic boundary, and ensures the universality and good compromise between the efficiency and the grid quality.
In the existing computational fluid mechanics parallel simulation, most computational fluid mechanics solvers can achieve effective parallelization, and the local area grid reconstruction method is usually executed in series. In order to obtain results in a reasonable time, hundreds of computational cores are often used, and at this time, performing the local area grid reconstruction process in series will become a performance bottleneck of the whole simulation process due to the following reasons:
(1) running a serialized mesh reconstruction process must contain enough memory to store the entire mesh;
(2) when the flow field solving process is accelerated by hundreds of times, the demand of local area grid reconstruction on a Central Processing Unit (CPU) becomes extremely outstanding;
(3) between the start of each grid reconstruction, distributed sub-grids (the grid is decomposed into different sub-grids in the parallelization flow field solving process and different computational cores are computed) need to be merged into the same computational core, and then after the grid is reconstructed, the grid needs to be distributed to different computational cores from the computational core so as to perform the next flow field computation.
It goes without saying that intensive data transfer between different computational cores is required in the grid assembly and distribution process, and in the particular case of the model, the time consumption of these processes may be a very significant percentage of the overall simulation cycle time (the specific percentage of occupation depends on the frequency of grid reconstruction calls and the data transfer efficiency between different computational cores).
Parallelization of the local area grid reconstruction method is a key means capable of breaking through the simulation cycle performance bottleneck brought by serialized local area grid reconstruction. The research on the technology in the engineering field is less, and European aerospace service prevention group (EADS) Termel and the like propose a method for reconstructing parallel local grids of an unstructured grid[1]However, their approach requires combining a single hole, originally distributed locally on different processors, on a given processor, which in turn causes the consequences of the serial approach described above, if the holes for performing the mesh reconstruction are too large, performance bottlenecks still occur in terms of memory footprint and computation time.
[1]Tremel,U.,K.A.,Hitzel,S.,Rieger,H.,Hassan,O.,&Weatherill,N.P.(2007).Parallel remeshing of unstructured volume grids for CFD applications.International Journal for Numerical Methods in Fluids,53(8),1361-1379.doi:10.1002/fld.1195
Disclosure of Invention
The invention aims to overcome the performance bottleneck brought by serialized local grid reconstruction in the dynamic boundary problem flow field simulation, and provides a local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation. The input of the method is a distributed grid after deformation, then low-quality grid units to be deleted are marked on corresponding processors, then the perforated grid is divided again by adopting a parallelized regional decomposition method, the obtained sub-grids are distributed to all available processors, further, the holes on the sub-grids are reconstructed in parallel, and finally, the grid units on different processors are distributed again, so that the scales of the grid units on each processor are balanced. The method can make up the regret of the local area grid reconstruction method which is not parallelized in the existing large-scale calculation fluid dynamic boundary parallel simulation circulation, can be inherited into the existing dynamic boundary flow field simulation circulation, realizes the complete parallelization of the flow field simulation system, overcomes the performance bottleneck in the aspects of memory consumption and calculation efficiency, and effectively parallelizes all time-consuming steps.
The invention is realized by the following technical scheme: a local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation comprises the following steps:
(1) for an input distributed (on different processors) grid which is deformed, finding and marking the cells needing grid reconstruction in parallel;
(2) dividing the old mesh with holes by using a parallel region decomposition method, and then distributing the obtained sub-meshes to all available processors again;
(3) carrying out grid reconstruction on the distributed holes in parallel;
(4) and distributing new grids to different processors in parallel to balance the grid scales on the processors so as to obtain a new distributed grid.
Furthermore, before the grid reconstruction, the old grid with holes marked on each processor is subjected to region decomposition, and is distributed to different processors for grid reconstruction.
Further, after the mesh reconstruction is completed, the attribute values on the newly generated mesh nodes are constructed by interpolating the old mesh, and then the subsequent steps are performed.
Further, after the parallelized grid is reconstructed, before the newly generated grid is solved in the flow field, the area decomposition is further performed, and the grid is uniformly distributed to available processors.
Furthermore, when the grid region decomposition is performed on the holed grid, the grid is simplified by using a dual graph method, and then the grid region decomposition is performed according to the graph relation.
Furthermore, the interface is restrained in the process of carrying out grid reconstruction on the distributed area with the holes.
Further, carrying out regional decomposition on the reconstructed grid by using ParMETIS.
Compared with the prior art, the invention has the beneficial effects that:
(1) the method can evenly distribute the distributed perforated grids to the available processors, can ensure that the sub-grids have good boundaries among the holes, ensures that subsequent grid reconstruction performed on each processor can repair the holes but does not influence the quality of surrounding grid units, does not relate to communication among the processors in the reconstruction process, and can realize local area grid reconstruction with high parallel efficiency.
(2) The method can be used as an algorithm module to be integrated into flow field dynamic boundary problem parallel simulation circulation, complete parallelization of simulation circulation is realized, bottlenecks of memory consumption and computing efficiency performance of the existing non-complete parallel (part of main links are executed in series) simulation circulation algorithm are broken through, and all time-consuming steps can be effectively parallelized.
(3) The method can complete large-scale local area grid reconstruction with high efficiency and high robustness, and has high applicability to dynamic boundary simulation (such as solid rocket booster separation simulation) of very complicated geometric shapes in industry.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a simplified flow chart of a trellis;
FIG. 3 is a schematic flow chart of an embodiment of the present invention; wherein the unit representations of the same texture are distributed on the same processor; (a) is an initial deformed grid; the grids (holes) which do not meet the quality requirement are bounded by thick lines; (b) to reallocate the meshed grid to a grid after all available processors; (c) the result after the grid reconstruction is carried out; (d) a result of the rebalancing of the allocation of grid cells to different processors;
FIG. 4 is a cross-sectional view of a volume grid of three external stores at different times after the continuous release process of the present invention is simulated, wherein (a) is ts0s, t (b)s0.1s, t in (c)s0.2s, t is (d)s=0.3s;
FIG. 5 is a comparison of simulated trajectories of three onlays using the algorithm proposed by the present invention with the results of Prewitt et al; wherein (a) - (c) respectively present data of bottom, outside and inside plug-in objects;
FIG. 6 is a cross-sectional view of a volume mesh in a simulation cycle of SRB separation at various times using the algorithm of the present invention; wherein (a) is ts0 s; (b) is ts1.2 s; (c) to t before repartitionings0.82 s; (d) for t after re-griddings=0.82s。
Detailed Description
As shown in fig. 1, the parallel local area mesh reconstruction method for dynamic boundary flow field numerical simulation provided by the present invention includes the following steps:
(1) analyzing the quality of a unit after grid deformation for the input distributed (on different processors) grids which are deformed, and if the quality of the grids accords with the quality of the grid, carrying out flow field calculation, advancing the grids to the next time step; otherwise, the grid cell to be deleted is marked (as shown in fig. 3 (a)). The marked cells may constitute a set of holes and regions for mesh reconstruction described by hole boundaries.
(2) The perforated mesh is segmented using a parallel region decomposition method and the resulting sub-meshes are then re-distributed over all available processors (as shown in fig. 3 (b)).
The method comprises the steps of firstly identifying and deleting the unnecessary grid surfaces by analyzing a dual graph of the grid, namely simplifying the grid, and then dividing the simplified grid by using a general graph divider so as to obtain a plurality of sub-grids without small dihedral angles and poor boundary shapes in order to realize the purposes of load balance, communication minimization, avoiding introducing artificial geometric characteristics (surfaces with poor shapes or small dihedral angles) and the like.
FIG. 2 shows a flowchart of a mesh simplification method, with the main calculations being performed on the edge-dual graph (SDG) and the element-dual graph (EDG) of the input mesh. Here, the cells and edges are (d-1) and (d-2) dimensional grid entities in the d-dimensional grid, respectively. In the edge dual graph, each graph node represents one edge (side) of the dual grid, and a graph edge (edge) exists between two adjacent edges; for example, each graph node in the EDG corresponds to a cell in the dual grid, and a graph edge exists between two adjacent cells.
First, a shape analysis step is performed on the input mesh to classify mesh edges that do not meet the specified shape quality requirements as deletable (using S)1Representing the set of grid edges). At the same time, by deleting S1The dual nodes of the medium to grid edges simplify the SDG of the input grid. On the basis, the invention adopts a node deletion algorithm proposed by ZHao et al[2]For further simplifying the SDG so that the simplified grid does not contain less than a user-specified threshold (in terms of β)thIndicated) is used. After this process, the more deletable grid edges are sorted (with S)2Representing the set of grid edges). Finally, with S1And S2For input, a simplified EDG is computed by reducing the graph edges corresponding to these deletable edges, and the dual mesh of the simplified EDG is the desired simplified mesh.
[2]Zhao,D.,Chen,J.,Zheng,Y.,Huang,Z.,&Zheng,J.(2015).Fine-grained parallel algorithm for unstructured surface mesh generation.Computers and Structures,154,177-191.doi:10.1016/j.compstruc.2015.04.004
After the mesh simplification step, an EDG is obtained that is dual to the simplified mesh. It is then sent to a general purpose graphics partitioner (e.g., Metis and ParMETIS) for domain decomposition. To balance the sub-region size and minimize the size of the inter-region interface, the EDGs need to be weighted appropriately. By default, the EDG node and edge weights are initialized to 1. Thus, in the simplified EDG, the node weights refer to the number of tetrahedral cells incorporated into the graph's node pair cells, while the edge weights refer to the number of triangular faces incorporated into the graph's edge pair cells.
One problem left is how to implement the domain decomposition method described above. In the present invention, the method employed is the following method (divided into four steps):
step 1 simplifies the EDG of each sub-mesh simultaneously by marking the inter-domain edges as boundary edges.
Step 2 penalizes the dual EDG edges with large weight for those inter-domain edges that do not meet the shape requirement or edges that form small dihedral.
Step 3, the EDG is subdivided by Parmetis. It is worth mentioning that since Parmetis minimizes the edge weight of the segmentation, it can prevent the edge penalized in step 2 from becoming a boundary edge of the segmentation.
Step 4 reallocates the sub-grids to conform to the graph partitioning result.
(3) The distributed holes are subjected to mesh reconstruction in parallel (as shown in fig. 3 (c)). The volumetric mesh reconstruction here may use a boundary constrained Delaunay mesh generator (input as surface boundaries), which the present invention proposes to use as proposed by Chen et al. Then a new grid with quality assurance fills the holes dug out before, and replaces the grid cells needing to be deleted in the data structure.
(4) And (d) parallelly allocating a new grid to different processors (as shown in (d) in fig. 3), balancing the grid sizes on the processors, and obtaining a new distributed grid.
The pseudo-code of the method described in the present invention is as follows:
pseudo code of local area grid reconstruction parallel method
Example A three plug-in continuous Release
The geometry is connected to the swept-back delta wing by three hangers through the nacelle. In order to ensure the safety of the separation trajectory, an ejector is active on each of the hangers in the initial phase. These hangers were released in a sequence from bottom to outside to inside with a 0.04 second delay between each release. The injector lasted 0.045 seconds after release. In the simulation, the ejector is fixed to the ejector so that the ejector does not generate a rolling torque. See Prewitt et al, incorporated by reference[3]To obtain more details about the solution parameters of the present simulation.
[3]Prewitt,N.C.,Belk,D.M.,&Shyy,W.(2000).Parallel computing of overset grids for aerodynamic problems with moving objects.Progress in Aerospace Sciences,36(2),117-172.doi:10.1016/S0376-0421(99)00013.
This example simulates a inviscid flow field with a tetrahedral mesh consisting of 423 ten thousand cells. The flow field solving adopts a double time stepping method, and the physical time interval is set to be 0.0005 s; since the total physical time was 0.4s, a total of 800 steps were calculated in the simulation. In FIG. 4, (a) - (d) are initial volume grids (t), respectivelys0s) and volume grid at tsCutting views at 0.1s, 0.2s, 0.3s (here tsAs separation time). During the separation process, the mesh is changed by mesh deformation and local area mesh reconstruction. Under the action of different ejection forces, the external mount moves outwards, and the internal mount moves inwards. FIGS. 5 (a) - (c) compare the simulation results with those reported in Prewitt et al. The comparison result shows that the simulated motion tracks of the bottom cabin and the outboard cabin are well matched with the reference data.
To demonstrate the benefits of the parallel local grid reconstruction procedure proposed by the present invention, table 1 gives the simulated timing performance based on the serial and parallel local grid reconstruction methods proposed by the present invention, respectively. Both simulations started from the same initial configuration and were executed on the same compute node (a total of 256 computer cores were used in both simulations). The simulation with the parallel grid reconstruction algorithm took about 3 hours and 53 minutes, with 18 calls to the parallel grid reconstruction algorithm taking 10.6 minutes. In contrast, if a serial mesh reconstruction algorithm is used, the overall simulation time increases by about 29%, which is about 5 hours. The increased time consumption comes mainly from the serial mesh reconstruction step. The serial algorithm was invoked 16 times, taking approximately 1 hour and 15 minutes. The present invention proposes a parallel approach that reduces the simulation time by one more hour (1 hour 7 minutes) than the serial approach.
Table 1 simulates statistics of time (units: seconds).
EXAMPLE two solid rocket booster separation
In order to verify the robustness and effectiveness of the parallel algorithm in complex simulation, a shuttle Solid Rocket Booster (SRB) separation problem is simulated. Each booster is about 45 meters long, the outer diameter is 3.71 meters, and the mass of the space shuttle during takeoff is about 590 tons. During the launch of the aerospace vehicle, when the fuel is depleted, the SRB provides thrust and separates from the outer cabin. At present, the speed of a space shuttle is about 1.2km/s, and the height is about 45 km. The mass of the SRB was reduced to 87 tons. Table 2 lists the physical properties of SRB. Safely separated by adding ejection force in the transverse direction. Four engines are located at the front of the supercharger and four engines are located at the lower part of the supercharger. Each engine provides about 0.68 seconds of 96kN force. The performance of the parallel mesh reconstruction algorithm proposed by the present invention is mainly concerned here.
TABLE 2 attributes of SRB
The CAD model contained 591 faces and the mesh contained 1077 million tetrahedral units and 340 triangular units. FIGS. 6 (a) to (d) show the webs from 0s to 1.2s during the separation processThe grid changes. During the simulation, the parallel local mesh reconstruction algorithm was invoked 9 times. FIGS. 6 (c) and (d) show the results at tsThe grid before and after reconstruction is 0.82s lower. It can clearly be seen that the cells are stretched before the mesh is reconstructed. The cell quality is greatly improved after the grid is reconstructed.
The total time of parallel simulation by adopting the serial local grid reconstruction method is 10 hours and 12 minutes, and the time of calling 9 times of serial local grid reconstruction is 1 hour and 24 minutes. In the serial local area mesh reconstruction process, the local area mesh reconstruction takes 37 minutes. When the parallel method provided by the invention is adopted, the time spent on local area grid reconstruction is reduced to 4.1 minutes, which is 11% of the time spent on the serial algorithm. Table 3 shows the time consumption of the serial and parallel local area mesh reconstruction methods proposed by the present invention at different steps, respectively. This indicates that under unbalanced load, the serial algorithm spends more time than the parallel algorithm proposed by the present invention.
TABLE 3 local grid repartitioning time consumption under different steps
Method of producing a composite material | The number of nuclei. | Region decomposition | Grid re-generation | Flow field interpolation | Others |
Serial reconstruction | 1 | 0.0 | 1720.4 | 63.4 | 136.3 |
Parallel reconstruction | 256 | 97.1 | 87.6 | 1.2 | 52.0 |
Claims (6)
1. A local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation is characterized by being used for numerical simulation of dynamic boundary flow fields of release of externally hung objects and rocket boosters, and comprising the following steps of:
(1) analyzing the quality of the unit after grid deformation for the input deformed distributed grid, and if the quality of the grid meets the requirement of flow field calculation, advancing the grid to the next time step; otherwise, marking the grid cells to be deleted, wherein the marked cells can form a group of holes and an area for grid reconstruction described by hole boundaries, wherein distributed representations are located on different processors, and the grids are tetrahedral grids in the flow field simulation;
(2) dividing the mesh with holes by using a parallel region decomposition method, and then distributing the obtained sub-meshes to all available processors; the method for segmenting the mesh with holes by using the parallel region decomposition method comprises the following specific steps:
2.1) identifying and deleting unneeded grid faces by analyzing the dual graph of the grid, namely simplifying the grid firstly, wherein the grid simplification is performed on the edge dual graph SDG and the unit dual graph EDG of the input grid, and the method comprises the following specific steps:
first, a shape analysis step is performed on the input meshClassifying the edges of the mesh which do not meet the specified shape quality requirements as deletable meshesS 1 Representing the set of grid edges; at the same time, by deletingS 1 The dual nodes of the middle-to-grid edges simplify the SDG of the input grid;
then, using the nodes to delete the small dihedral angles in the grid which do not contain the small dihedral angles lower than the user-specified threshold, after the process, classifying more deletable grid edges for useS 2 Representing the set of grid edges;
finally, toS 1 AndS 2 calculating a simplified EDG by reducing the graph edges corresponding to the deletable edges for input, wherein the dual grid of the simplified EDG is the required simplified grid;
in the edge dual graph, each graph node represents an edge side of the dual grid, and a graph edge exists between two adjacent edges;
2.2) after the mesh simplification step, obtaining the EDG which is dual with the simplified mesh, then sending the EDG to a general graph partitioner for domain decomposition, and further obtaining a plurality of sub-meshes which have no small dihedral angles and no boundary shape difference; the method is divided into four substeps: step 1, marking the inter-domain edge as a boundary edge, and simplifying the EDG of each sub-grid; step 2, punishing dual EDG edges of the inter-domain edges or the edges forming small dihedral angles, which do not meet the shape requirement, by using large weight; step 3, adopting Parmetis to subdivide the EDG; step 4, redistributing the sub-grids to accord with the graph division result;
(3) carrying out grid reconstruction on the distributed holes in parallel;
(4) and distributing new grids to different processors in parallel to balance the grid scales on the processors so as to obtain a new distributed grid.
2. The parallel local area grid reconstruction method for the numerical simulation of the dynamic boundary flow field according to claim 1, wherein before grid reconstruction, the old holed grids marked on each processor are subjected to area decomposition and are distributed to different processors for grid reconstruction.
3. The parallel local area grid reconstruction method for the numerical simulation of the dynamic boundary flow field according to claim 1, wherein after the grid reconstruction is completed, the attribute values on newly generated grid nodes are constructed by interpolating an old grid, and then the subsequent steps are performed.
4. The parallel local area grid reconstruction method for the numerical simulation of the dynamic boundary flow field according to claim 1, wherein after the parallelized grid reconstruction, before the flow field solution is continued for the newly generated grid, the area decomposition is further performed to uniformly distribute the grid to available processors.
5. The parallel local area grid reconstruction method for the numerical simulation of the dynamic boundary flow field according to claim 1, wherein the interface is constrained during the grid reconstruction of the distributed void area.
6. The parallel local area grid reconstruction method for the numerical simulation of the dynamic boundary flow field according to claim 1, wherein the area decomposition is performed on the reconstructed grid by using a ParMETIS.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910410152.1A CN110188424B (en) | 2019-05-16 | 2019-05-16 | Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910410152.1A CN110188424B (en) | 2019-05-16 | 2019-05-16 | Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110188424A CN110188424A (en) | 2019-08-30 |
CN110188424B true CN110188424B (en) | 2021-01-15 |
Family
ID=67716443
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910410152.1A Active CN110188424B (en) | 2019-05-16 | 2019-05-16 | Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110188424B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113034694B (en) * | 2021-03-26 | 2022-01-11 | 浙江大学 | Method and device for automatically generating measurement orthogonal grid adaptive to flow field characteristics |
CN116484702B (en) * | 2023-06-26 | 2023-09-12 | 北京凌云智擎软件有限公司 | Unstructured nested grid computing method suitable for any unit type |
CN116562066B (en) * | 2023-07-12 | 2023-09-26 | 北京凌云智擎软件有限公司 | Multi-layer region decomposition parallel computing method and device for polyhedral grid |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101408900A (en) * | 2008-11-24 | 2009-04-15 | 中国科学院地理科学与资源研究所 | Distributed space data enquiring and optimizing method under gridding calculation environment |
JP2010271312A (en) * | 2009-05-21 | 2010-12-02 | General Electric Co <Ge> | Inspection system and method with multi-image phase shift analysis |
CN102521854A (en) * | 2011-12-29 | 2012-06-27 | 北京理工大学 | Parallel flow line placing method applicable to two-dimensional flow field |
CN103268634A (en) * | 2012-02-24 | 2013-08-28 | 苏州蓝海彤翔系统科技有限公司 | Quick parallel self-adaptive simplification method based on vertex clustering for out-of-core model |
CN104240300A (en) * | 2014-08-29 | 2014-12-24 | 电子科技大学 | Large-scale point cloud complex space curved face reconstruction method based on distributed parallelism |
CN104239661A (en) * | 2013-06-08 | 2014-12-24 | 中国石油化工股份有限公司 | Large-scale numerical reservoir simulation calculation method |
CN107750441A (en) * | 2015-01-26 | 2018-03-02 | 卢森堡商创研腾智权信托有限公司 | Safety actuality communication network and agreement |
CN108667702A (en) * | 2013-03-14 | 2018-10-16 | 谷歌有限责任公司 | Device, method and the associated information processing of safety for the household equipped with intelligence sensor |
CN108829614A (en) * | 2014-12-23 | 2018-11-16 | 英特尔公司 | Predictive reading in buffer storage |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8131069B2 (en) * | 2007-07-16 | 2012-03-06 | Ecole Centrale De Paris | System and method for optimizing single and dynamic markov random fields with primal dual strategies |
-
2019
- 2019-05-16 CN CN201910410152.1A patent/CN110188424B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101408900A (en) * | 2008-11-24 | 2009-04-15 | 中国科学院地理科学与资源研究所 | Distributed space data enquiring and optimizing method under gridding calculation environment |
JP2010271312A (en) * | 2009-05-21 | 2010-12-02 | General Electric Co <Ge> | Inspection system and method with multi-image phase shift analysis |
CN102521854A (en) * | 2011-12-29 | 2012-06-27 | 北京理工大学 | Parallel flow line placing method applicable to two-dimensional flow field |
CN103268634A (en) * | 2012-02-24 | 2013-08-28 | 苏州蓝海彤翔系统科技有限公司 | Quick parallel self-adaptive simplification method based on vertex clustering for out-of-core model |
CN108667702A (en) * | 2013-03-14 | 2018-10-16 | 谷歌有限责任公司 | Device, method and the associated information processing of safety for the household equipped with intelligence sensor |
CN104239661A (en) * | 2013-06-08 | 2014-12-24 | 中国石油化工股份有限公司 | Large-scale numerical reservoir simulation calculation method |
CN104240300A (en) * | 2014-08-29 | 2014-12-24 | 电子科技大学 | Large-scale point cloud complex space curved face reconstruction method based on distributed parallelism |
CN108829614A (en) * | 2014-12-23 | 2018-11-16 | 英特尔公司 | Predictive reading in buffer storage |
CN107750441A (en) * | 2015-01-26 | 2018-03-02 | 卢森堡商创研腾智权信托有限公司 | Safety actuality communication network and agreement |
Also Published As
Publication number | Publication date |
---|---|
CN110188424A (en) | 2019-08-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110188424B (en) | Local area grid reconstruction parallel method for dynamic boundary flow field numerical simulation | |
US11275353B2 (en) | Creating a voxel representation of a three dimensional (3-D) object | |
Mavriplis et al. | Large-scale parallel unstructured mesh computations for three-dimensional high-lift analysis | |
US8259104B2 (en) | Domain decomposition by the advancing-partition method for parallel unstructured grid generation | |
Pirzadeh | Progress toward a user-oriented unstructured viscous grid generator | |
US10373371B2 (en) | Generating slicing data from a tree data structure | |
Dawes et al. | A practical demonstration of scalable, parallel mesh generation | |
Chen et al. | Improved boundary constrained tetrahedral mesh generation by shell transformation | |
Chen et al. | Scalable generation of large-scale unstructured meshes by a novel domain decomposition approach | |
CN105022928A (en) | Digitized real-time determination method for center-of-gravity position of fuel system of aircraft | |
Pandya et al. | Enhancements to TetrUSS for NASA Constellation Program | |
Chang et al. | A parallel implicit hole-cutting method based on background mesh for unstructured Chimera grid | |
Martin et al. | Overset grid assembly approach for scalable computational fluid dynamics with body motions | |
McMorris et al. | Octree-advancing front method for generation of unstructured surface and volume meshes | |
CN109753682B (en) | Finite element stiffness matrix simulation method based on GPU (graphics processing Unit) end | |
Zheng et al. | Parallel and automatic isotropic tetrahedral mesh generation of misaligned assemblies | |
CN112948643B (en) | Structured grid streamline integration method based on thread parallelism | |
CN109241620B (en) | Method for improving SSORPCG (Steady State stress gradient computed tomography) parallel solution of slope stress field based on finite element area decomposition | |
Zagaris et al. | A framework for parallel unstructured grid generation for practical aerodynamic simulations | |
Pirzadeh et al. | Domain decomposition by the advancing-partition method for parallel unstructured grid generation | |
Figueroa et al. | Experiences Porting a High Order Cartesian Finite Difference Solver to GPUs Using OpenACC | |
CN116401916B (en) | Method, device, medium and equipment for generating high-quality three-dimensional grid | |
Morton et al. | An overset unstructured grid methodology applied to a C-130 with a cargo pallet and extraction parachute | |
Luo et al. | GPU accelerated cell-based adaptive mesh refinement on unstructured quadrilateral grid | |
CN105550384A (en) | Post-processing method for automatic partition of aerodynamic flow field grid by Metis |
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 |