CN106294287A - A kind of core exterior domain decomposition method based on parallel high order MoM - Google Patents
A kind of core exterior domain decomposition method based on parallel high order MoM Download PDFInfo
- Publication number
- CN106294287A CN106294287A CN201510310080.5A CN201510310080A CN106294287A CN 106294287 A CN106294287 A CN 106294287A CN 201510310080 A CN201510310080 A CN 201510310080A CN 106294287 A CN106294287 A CN 106294287A
- Authority
- CN
- China
- Prior art keywords
- sub regions
- high order
- iteration
- region
- mom
- 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.)
- Pending
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of core exterior domain decomposition method based on parallel high order MoM, be divided into n sub regions according to the actual characteristic solving target and be modeled；For discontinuous problem, then need not extend to adjacent area.For continuous problem, the most each sub regions is carried out subdivision, extraction model information.For every sub regions, solve respectively with high order MoM, introduce core outer solution technique when filling impedance matrix, use hard disk to store data；When each sub regions structure is identical, then has only to fill once, the iterative process below is reused.Initialize current value, set an iteration error scope and maximum iteration time, be iterated successively, try to achieve the electric current of each sub regions, and then obtain the field distribution in whole region.The present invention has higher accuracy, realizability and reliability.
Description
Technical field
The present invention relates to Computational electromagnetics field, a kind of core exterior domain decomposition method based on parallel high order MoM.
Background technology
Along with the development of science and technology, from weather forecast to MISSILE LAUNCHING, again to the design of various aircrafts, the solution of nearly all field question all be unable to do without numerical computations.It is concentrated mainly on three aspects for electromagnetism calculating field the most in the world to study, i.e. labyrinth, TV university problem, hybrid metal and media problem.Conventional numerical method has: moment method, Finite Element Method and time-domain finite difference etc., moment method MoM proposed by Harrington as far back as nineteen sixty-eight is widely used, compared with additive method, moment method can accurately calculate electromagnetic radiation and scattering problems.The mensuration usual employing RWG basic function of low-order moment, in order to reach certain computational accuracy, the length of each edge of each dough sheet is on the order of magnitude of λ/10, and it is very big that this results in unknown quantity number, and complexity and memory requirements during solution matrix equation reach O (N3) and O (N2), wherein N is the number of unknown quantity.Therefore, within a very long time, moment method cannot process Electrically large size object problem, 1993, the L.R.Hamilton of the U.S. etc. propose the concept of high-order basic function, on the basis of high-order basic function is built upon bilinearity dough sheet, it is possible to approach arbitrary curved surface and any CURRENT DISTRIBUTION neatly.It makes the size of solution matrix be reduced at least ten times, decreases computation complexity and memory requirements.Meanwhile, various fast algorithms and Domain Decomposition Method are emerged in large numbers in succession, and such as, fast algorithm has Fast Multiple Method, multilevel fast multipole and self-adapting integrative approach etc..Domain Decomposition Method includes Overlapping Region Decomposition and non-overlapping domain decomposition etc., scale can be arrived greatly the most treatable problem of single computer by these methods, it is divided into several minor issues to solve, allow each sub regions use different mathematical modeies and the grid of different density degree or layout, and each sub regions problem can be with estimate simultaneously.But along with being continuously increased of military requirement and Electromagnetic Engineering problem complexity, the resource making institute's Solve problems consumption is continuously increased, especially calculating the restriction in terms of time and memory source makes it cannot solve more massive electromagnetic problem, such as large-scale antenna array and airborne antenna layout etc..Therefore, how on the basis of consuming less calculating resource, more accurate simulation Electrically large size object remains a need for studying further.
Summary of the invention
It is an object of the invention to provide a kind of core exterior domain decomposition method based on parallel high order MoM, with the problem solving to propose in above-mentioned background technology.
For achieving the above object, the present invention provides following technical scheme:
A kind of core exterior domain decomposition method based on parallel high order MoM, comprises the steps:
Step 1, is divided into n sub regions according to the actual characteristic solving target and is modeled；For discontinuous problem, then need not extend to adjacent area, only the electric current of each sub regions need to be calculated respectively and then carry out suitable iteration and can obtain the field distribution of whole target；For continuous problem, then carry out next step；
Step 2, carries out subdivision, extraction model information respectively to each sub regions；
Step 3, for every sub regions, solves respectively with high order MoM, introduces the outer solution technique of core, use hard disk to store data when filling impedance matrix；When each sub regions structure is identical, then has only to fill once, the iterative process below is reused；
Step 4, initializes current value, sets an iteration error scope and maximum iteration time, be iterated successively according to above-mentioned alternative manner, try to achieve the electric current of each sub regions；
Step 5, obtains the field distribution in whole region according to electric current obtained in the previous step.
As the further scheme of the present invention: in step 3, high order MoM uses high-order basic function and bilinearity dough sheet to approach arbitrary surface, utilizes SKALAPACK data base that program is carried out parallelization simultaneously.
As the further scheme of the present invention: merely with the electric current of Non-overlapping Domain when certain zone current is iterated by step 4, eliminate the adjacent area strong interaction impact on Iterations of Multi；And the iteration expression formula when number of regions n >=3, l sub regions is hadWherein [Ii] ' for the current coefficient matrix of the Non-overlapping Domain in the i of region, [Z 'li] it is the mutual resistance matrix of Non-overlapping Domain in the i of region and region l；When | | [V]-[Z] [I '] | |/during ＜ ε, iteration terminates | | [V] | |, and wherein ε is iteration error.
Compared with prior art, the invention has the beneficial effects as follows:
1) three of the above method is effectively combined, and is applied in the calculating of extensive electromagnetism.Overcome prior art and to the calculating time and calculate the restriction in terms of resource, provide a kind of efficient method for solving of electrically large sizes problem.The method has higher accuracy and reliability.
2) will solve target and be divided into some regions, regional can use different methods to calculate according to respective feature flexibly, finally uses alternative manner to obtain the solution of whole problem.
3) for some problem, need to change certain part repeatedly to calculate, such as airborne antenna, need to analyze the electromagnetic property of dissimilar antenna installation, this is that we are not required to recalculate whole target, as long as the part changed is calculated, it is then engage in iteration and just can quickly obtain desired result.
Accompanying drawing explanation
Fig. 1 is Overlapping Region Decomposition schematic diagram；
Fig. 2 is the outer method for solving schematic diagram of core；
Fig. 3 is Region Decomposition calculation flow chart；
Fig. 4 is umbrella shape microstrip array illustraton of model；
Fig. 5 is micro-strip dipole antenna Array Model figure；
Fig. 6 is Airborne Waveguide Slot Array Antenna illustraton of model；
Fig. 7 is OOC-DDM Yu MoM result of calculation comparison diagram；
Fig. 8 is airborne antenna battle array result of calculation figure.
Detailed description of the invention
Below in conjunction with the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art are obtained under not making creative work premise, broadly fall into the scope of protection of the invention.
Present invention aims to the deficiency of above-mentioned prior art, a kind of overlapping domain decomposition method based on the outer solution technique of parallel high order MoM and core is proposed, the high order MoM proposed before and Region Decomposition technology are combined by it, and utilize Parallel implementation to realize.High order MoM uses high-order basic function and bilinearity dough sheet, it is ensured that arbitrarily metal and medium divides the seriality of electric current on interface.
For discontinuous problem, it suitable is divided into n sub regions, owing to there is not the continuity problem of electric current between each sub regions, only the electric current of each sub regions need to be calculated respectively and then carry out suitable iteration and just can obtain the field distribution of whole target.But when solving successive objective, when being iterated, owing to the interaction of adjacent area is stronger, cause progression iteration error bigger in iteration so that final iteration result is not restrained.
In order to preferably process continuity problem, use herein before and after one to overlapping decomposition method, as it is shown in figure 1, some overlapping region between the subregion that is adjacent of the n-th subregion.Illustrate as a example by n=3: assume primary iteration value [I1](0)=[I2](0)=[I3](0)=0 is iterated, and has for region 1
[I1]=[Z11]-1[V1]
Assuming 1, the basic function number of the overlapping region in 2 two regions is p, and the basic function of total area 1 is m.When utilizing the electric current in electric current iteration region 2 in region 1, in order to eliminate the adjacent area stronger interaction impact on iteration, region 2 electric current is iterated by (m-p) individual basic function electric current of a chosen area 1.I.e.
[I2]=-[Z22]-1[Z′21][I1]′+[Z22]-1[V2]
Wherein, [I1] ' for the current coefficient matrix of the Non-overlapping Domain in region 1, [Z '21] it is the mutual resistance matrix of Non-overlapping Domain in region 1 and region 2.
Method is iterated the most according to this, has for region 3
[I3]=-[Z33]-1[Z′31][I1]′-[Z33]-1[Z′32][I2]′+[Z33]-1[V3]
Utilize [I1]′[I2]′[I3] surface current tried to achieve solves whole scattered field.
Analyze according to above, can be in the hope of when number of regions n > 3 time iteration expression formula, l sub regions is had
Wherein [Ii] ' for the current coefficient matrix of the Non-overlapping Domain in the i of region, [Z 'li] it is the mutual resistance matrix of Non-overlapping Domain in the i of region and region l.
When | | [V]-[Z] [I '] | |/during ＜ ε, iteration terminates | | [V] | |, and wherein ε is iteration error.
" parallel computation " is a current hot issue, and Region Decomposition is also very suitable for parallel itself.I.e. utilize multiprocessor or multicomputer, relatively independent for process is allocated on different nodes, by the most independent operating system scheduling, enjoy independent CPU and memory source.What moment method solved is divided into two steps parallel.Initially set up a matrix equation, then solve this matrix equation.Can realize efficiently calculating to matrix fill-in and Matrix Solving process when parallel algorithm combines moment method.Thus can utilize group system, thus hardware spending and calculating time can be greatly reduced.For solving of parallel extensive dense matrix equation, it usually needs matrix in process with following this model split: the data that each process stored number is suitable, computational load to be evenly distributed on each node.The one of Parallel MoM mode effectively is exactly to be distributed in all of participation process by extensive dense matrix.In order to ensure good load balance, different matrix equation solver such as direct solutions and iterative solution method, correspond to different matrix dividing mode.
It addition, the present invention additionally uses the outer solution technique of core, when the matrix that we obtain is excessive, owing to calculator memory holds quantitative limitation, it is necessary to store data into hard disk.The great advantage solved outside core is when matrix fill-in, total data is write hard disk, and when Matrix Solving, once only need to participate in calculating by a part of element write internal memory, calculating is write back to hard disk after completing again, the most repeatedly reading and writing, until calculating whole matrix element, detailed process is as shown in Figure 2.This technology is greatly reduced the memory requirements of computer, significantly improves the analysis ability to Electrically large size object.
Embodiment 1
With reference to Fig. 3, in the embodiment of the present invention, core exterior domain decomposition method based on parallel high order MoM, comprise the steps:
Step 1, is divided into a few sub regions according to the actual features solving target and is modeled.For discontinuous problem, then need not extend to adjacent area.
Step 2, carries out quadrangle grids, extraction model information to each sub regions surface respectively.
Step 3, for every sub regions, solves respectively by the outer technology of moment method and core, when filling impedance matrix, when each sub regions structure is identical, then has only to fill once, reuse in the iterative process below.So can save the substantial amounts of matrix fill-in time.
Step 4, initializes current value, sets an iteration error scope and maximum iteration time, be iterated successively according to alternative manner above-mentioned, finally try to achieve the electric current of each sub regions.
Step 5, obtains the field distribution in whole region according to electric current obtained in the previous step.
The effectiveness of this example calculation method can be further illustrated by following simulation result:
Simulated conditions
Simulation calculation platform one: the tower work station of single node, configures 26 core INTEL XEON E5-2620 processors, 64GB DDR3 internal memory, 6T total capacity hard disk.
Simulation calculation platform two: HP BL460c high-performance blade server cluster 1 overlaps, and comprises 1 management node and 16 calculating nodes.Management node: two 4 core Intel Xeon E5310 1.6GHz CPU, 18GB internal memory, 1.8TB hard disk.Calculate node: 16 blade nodes (2011), every node configures two 6 core Intel Xeon E5650 2.66GHz CPU, 24GB internal memory, 600GB hard disk.The total check figure of CPU 192, total amount of memory 384GB, hard disk total capacity 8TB.Network connects: high speed switch.
Simulation calculation platform three: 136 cpu nodes of High Performance Cluster System, tide NX5440 each node configuration E5-2692v2 (2.20GHz) * 2,64GB internal memory, 900GB SAS*2 hard disk,
Example one: unit number is the umbrella shape micro-strip antenna array of 30 × 14, and model is as shown in Figure 4.Between each unit all discontinuously, unit interval is: x direction 0.065m, z direction 0.065m, and monopole encourages, and frequency is 2.5GHz.It is divided into 14 regions in the z-direction.Overall model uses simulation calculation platform two, and Region Decomposition uses simulation calculation platform one.
Example two: unit number is the microstrip dipole array of 1 × 96, model is as shown in Figure 5, continuous between each unit.Monopole encourages, and frequency is 3GHz.Being classified as 6 regions in the y-direction, each region extends two unit to adjacent area respectively.Overall model uses simulation calculation platform one, and Region Decomposition uses simulation calculation platform one.
Example three: Ka wave band waveguide array antenna loads aboard, calculates its radiation characteristic.Aerial array and aircraft as shown in Figure 6, a size of: 70 λ × 17.1 λ.The size of aircraft is 471 λ × 420 λ.Aircraft is divided into 6 regions, and antenna is 1 region, and aerial array distance head is 207 wavelength, and antenna height is 47 wavelength.The main beam direction of antenna points to tail.Simulation calculation platform is platform three.
2. emulation content
Under described simulated conditions, test as follows:
Above-mentioned model is divided into several regions, Surface Partition quadrangularly grid, extracts point and the information on limit.Calculate the radiation field of this target with algorithm routine recited above, and contrast with moment method overall calculation result.As shown in Figure 6, example two result is as shown in Figure 7 for example one result.As can be seen from the figure the result that two kinds of methods are obtained is coincide good.Example three is large-scale airborne antenna array, and integrative simulation cannot realize with regard to existing hardware resource, hence with the method for the present invention, is broken down into 7 regions, and the result of gained is as shown in Figure 8.Internal memory and time loss are as shown in table 1, and as can be seen from Table 1, Region Decomposition saves substantial amounts of time and memory space.
Table 1
To sum up, this simulating, verifying correctness of the present invention, realizability and reliability.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, and without departing from the spirit or essential characteristics of the present invention, it is possible to realize the present invention in other specific forms.Therefore, no matter from the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is limited by claims rather than described above, it is intended that all changes fallen in the implication of equivalency and scope of claim included in the present invention.
In addition, it is to be understood that, although this specification is been described by according to embodiment, but the most each embodiment only comprises an independent technical scheme, this narrating mode of description is only for clarity sake, description should can also be formed, through appropriately combined, other embodiments that it will be appreciated by those skilled in the art that as an entirety, the technical scheme in each embodiment by those skilled in the art.
Claims (3)
1. a core exterior domain decomposition method based on parallel high order MoM, it is characterised in that comprise the steps:
Step 1, is divided into n sub regions according to the actual characteristic solving target and is modeled；For discontinuous problem, then need not extend to adjacent area, only the electric current of each sub regions need to be calculated respectively and then carry out suitable iteration and can obtain the field distribution of whole target；For continuous problem, then carry out next step；
Step 2, carries out subdivision, extraction model information respectively to each sub regions；
Step 3, for every sub regions, solves respectively with high order MoM, introduces the outer solution technique of core, use hard disk to store data when filling impedance matrix；When each sub regions structure is identical, then has only to fill once, the iterative process below is reused；
Step 4, initializes current value, sets an iteration error scope and maximum iteration time, be iterated successively according to above-mentioned alternative manner, try to achieve the electric current of each sub regions；
Step 5, obtains the field distribution in whole region according to electric current obtained in the previous step.
Core exterior domain decomposition method based on parallel high order MoM the most according to claim 1, it is characterized in that, in step 3, high order MoM uses high-order basic function and bilinearity dough sheet to approach arbitrary surface, utilizes SKALAPACK data base that program is carried out parallelization simultaneously.
Core exterior domain decomposition method based on parallel high order MoM the most according to claim 1, it is characterized in that, merely with the electric current of Non-overlapping Domain when certain zone current is iterated by step 4, eliminate the adjacent area strong interaction impact on Iterations of Multi；And the iteration expression formula when number of regions n >=3, l sub regions is hadWherein [Ii] ' for the current coefficient matrix of the Non-overlapping Domain in the i of region, [Z 'li] it is the mutual resistance matrix of Non-overlapping Domain in the i of region and region l；When | | [V]-[Z] [I '] | |/during ＜ ε, iteration terminates | | [V] | |, and wherein ε is iteration error.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510310080.5A CN106294287A (en) | 2015-06-08 | 2015-06-08 | A kind of core exterior domain decomposition method based on parallel high order MoM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510310080.5A CN106294287A (en) | 2015-06-08 | 2015-06-08 | A kind of core exterior domain decomposition method based on parallel high order MoM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106294287A true CN106294287A (en) | 2017-01-04 |
Family
ID=57659884
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510310080.5A Pending CN106294287A (en) | 2015-06-08 | 2015-06-08 | A kind of core exterior domain decomposition method based on parallel high order MoM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106294287A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108268696A (en) * | 2017-12-15 | 2018-07-10 | 西安电子科技大学 | A kind of FSS antenna house modeling methods suitable for high order MoM |
| CN109657196A (en) * | 2018-11-27 | 2019-04-19 | 上海无线电设备研究所 | A kind of MOM-KA mixed method accelerated based on airspace close coupling path |
| CN113128090A (en) * | 2021-04-21 | 2021-07-16 | 北京航空航天大学 | Waveguide mode excitation method based on moment method, storage medium and device |
| CN115801411A (en) * | 2022-11-18 | 2023-03-14 | 国网河北省电力有限公司电力科学研究院 | High-order data feature extraction and identification method for power internet of things attack behavior |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101770542A (en) * | 2010-02-25 | 2010-07-07 | 中国科学院上海光学精密机械研究所 | Method for simulating electromagnetic wave propagation by cluster computer |
| CN103955779A (en) * | 2014-05-23 | 2014-07-30 | 武汉大学 | Wind power climbing event probability scene prediction method |
-
2015
- 2015-06-08 CN CN201510310080.5A patent/CN106294287A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101770542A (en) * | 2010-02-25 | 2010-07-07 | 中国科学院上海光学精密机械研究所 | Method for simulating electromagnetic wave propagation by cluster computer |
| CN103955779A (en) * | 2014-05-23 | 2014-07-30 | 武汉大学 | Wind power climbing event probability scene prediction method |
Non-Patent Citations (1)
| Title |
|---|
| 李艳艳 等: ""并行核外区域分解算法在电磁辐射中的应用"", 《微波学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108268696A (en) * | 2017-12-15 | 2018-07-10 | 西安电子科技大学 | A kind of FSS antenna house modeling methods suitable for high order MoM |
| CN109657196A (en) * | 2018-11-27 | 2019-04-19 | 上海无线电设备研究所 | A kind of MOM-KA mixed method accelerated based on airspace close coupling path |
| CN109657196B (en) * | 2018-11-27 | 2023-03-28 | 上海无线电设备研究所 | MOM-KA mixing method based on airspace strong coupling path acceleration |
| CN113128090A (en) * | 2021-04-21 | 2021-07-16 | 北京航空航天大学 | Waveguide mode excitation method based on moment method, storage medium and device |
| CN113128090B (en) * | 2021-04-21 | 2021-09-10 | 北京航空航天大学 | Waveguide mode excitation method based on moment method, storage medium and device |
| CN115801411A (en) * | 2022-11-18 | 2023-03-14 | 国网河北省电力有限公司电力科学研究院 | High-order data feature extraction and identification method for power internet of things attack behavior |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ergul et al. | A hierarchical partitioning strategy for an efficient parallelization of the multilevel fast multipole algorithm | |
| CN104992001B (en) | The accurate quick calculation method of extensive MIMO array Antenna Far Field radiation field | |
| CN106294287A (en) | A kind of core exterior domain decomposition method based on parallel high order MoM | |
| Plewa et al. | 3D streamers simulation in a pin to plane configuration using massively parallel computing | |
| CN102081690B (en) | MDA (Matrix Decomposition Algorithm)-combined novel SVD (Singular Value Decomposition) method for complex circuit | |
| MacKie-Mason et al. | Adaptive and parallel surface integral equation solvers for very large-scale electromagnetic modeling and simulation | |
| CN112733364B (en) | Foil cloud scattering rapid calculation method based on impedance matrix partitioning | |
| CN104200074A (en) | Multi-layer complex beam method for obtaining electromagnetic scattering characteristic of target quickly | |
| CN112329285B (en) | Method for analyzing multi-scale structure transient response by Taylor series expansion hybrid algorithm | |
| US11244094B2 (en) | Eigen augmentation methods for electromagnetic modelling and simulation | |
| CN101794355A (en) | Computer extranuclear parallel computing method for electromagnetic radiation and scattering of electric large object | |
| Cheng et al. | Reordering of hybrid unstructured grids for an implicit Navier-Stokes solver based on openMP parallelization | |
| CN101794356B (en) | Computer extranuclear computing method for electromagnetic radiation and scattering of electric large object | |
| Zhao et al. | Parallel MoM-PO method with out-of-core technique for analysis of complex arrays on electrically large platforms | |
| CN107305536A (en) | Mix the discontinuous Jia Lvejin methods of rank time domain | |
| CN105205299A (en) | Rapid dimensionality reduction analysis method for electric large object electromagnetic scattering properties | |
| Ntinas et al. | GPU and FPGA parallelization of fuzzy cellular automata for the simulation of wildfire spreading | |
| CN104954972A (en) | Distributed training method of L1 regularization kernel-based learning machine based on parallel projection method | |
| Wang et al. | A novel method for rapidly solving wideband RCS by combining UCBFM and compressive sensing | |
| Oikawa et al. | Analysis of lightning electromagnetic field on large‐scale terrain model using three‐dimensional MW‐FDTD parallel computation | |
| CN106951607B (en) | Electricity large platform modeling method for calculating isolation between antennas | |
| Weng et al. | An efficient iterative hybrid multilevel fast multipole algorithm‐physical optics method for simulating antenna mounted on the electrically large platform | |
| Miranda et al. | Hardware acceleration of an FMM-FFT solver using consumer-grade GPUs | |
| Oswald et al. | Radar cross sections of objects with simulated defects using the parallel FDTD method | |
| Simon et al. | Solving large scale em problems using FDTD analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170104 |
|
| WD01 | Invention patent application deemed withdrawn after publication |