CN116629109A - Low-scattering carrier optimization design method - Google Patents
Low-scattering carrier optimization design method Download PDFInfo
- Publication number
- CN116629109A CN116629109A CN202310550321.8A CN202310550321A CN116629109A CN 116629109 A CN116629109 A CN 116629109A CN 202310550321 A CN202310550321 A CN 202310550321A CN 116629109 A CN116629109 A CN 116629109A
- Authority
- CN
- China
- Prior art keywords
- low
- scattering
- geometric
- scattering carrier
- carrier
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000013461 design Methods 0.000 title claims abstract description 30
- 238000005457 optimization Methods 0.000 title claims abstract description 18
- 230000007246 mechanism Effects 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 17
- 238000004088 simulation Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 44
- 230000006870 function Effects 0.000 claims description 24
- 238000005192 partition Methods 0.000 claims description 22
- 238000000354 decomposition reaction Methods 0.000 claims description 18
- 238000005516 engineering process Methods 0.000 claims description 16
- 230000010287 polarization Effects 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000010801 machine learning Methods 0.000 claims description 10
- 230000007704 transition Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 230000006835 compression Effects 0.000 claims description 7
- 238000007906 compression Methods 0.000 claims description 7
- 238000000638 solvent extraction Methods 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 239000007787 solid Substances 0.000 claims description 3
- 230000007480 spreading Effects 0.000 claims description 3
- 238000003892 spreading Methods 0.000 claims description 3
- 239000000969 carrier Substances 0.000 abstract description 3
- 238000011176 pooling Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000005284 excitation Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000031018 biological processes and functions Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000012938 design process Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000003032 molecular docking Methods 0.000 description 1
- 230000005405 multipole Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005381 potential energy Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 210000000857 visual cortex Anatomy 0.000 description 1
- 238000013316 zoning Methods 0.000 description 1
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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computer Hardware Design (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Mathematical Optimization (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Medical Informatics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a low-scattering carrier optimization design method, and belongs to the field of low-scattering carriers. The method comprises the following steps: constructing an original model of the low-scattering carrier; carrying out regional parameterized geometric treatment to obtain a geometric model of the low-scattering carrier; electromagnetic modeling is carried out on the geometric model; implementing parametric modeling constraint condition limitation; optimizing a low-scattering carrier by using a convolutional neural network; and obtaining the quantized output of the simulation calculation result. The method can deeply reflect the basic characteristics of the low-scattering carrier and the physical mechanism behind the low-scattering carrier, and can give consideration to full-wave high-precision calculation efficiency in the aspects of single-round large bandwidth, wide-angle domain, multi-polarization and the like in geometric modeling and electromagnetic modeling.
Description
Technical Field
The invention relates to the field of low-scattering carriers, in particular to a low-scattering carrier optimal design method combining regional decomposition-direct solution electromagnetic modeling and convolutional neural network.
Background
Currently, on-board, carrier-based, and even on-board platforms have created stealth requirements for various sensor systems and various components. Due to various realistic conditions, the stealth characteristics of the system cannot be analyzed or verified based on the final platform at the design stage, but cannot be directly evaluated only for the subsystem or component. This is because the hard boundaries of the sub-system or component, including edges, and the internal structures that would otherwise be shielded, tend to be extremely strong, and contribute significantly to the scattering of the entire platform in a truly installed state. Therefore, a new carrier platform, i.e. "low scattering carrier", that can be substituted for the equivalent evaluation must be sought to aid in design and analysis. The subsystem or the component to be evaluated is arranged on the low scattering carrier, and the details in the integration process are further specially processed, so that scattering interference of the edge and the inner cavity structure can be eliminated, and the result is closer to the actual contribution value.
It should be noted that, although there is a common design principle based on the high-frequency technology, the design of the low-scattering carrier belongs to the inverse problem, and the solution is not unique. Meanwhile, because the related radar has wide frequency band, wide angular range and multiple polarization conditions, the cost of electromagnetic simulation and optimization iteration is extremely high, and the improvement of the efficiency of the design process is urgently needed.
Chinese patent application No. 201610825510.1 discloses a low scattering carrier for RCS testing, where the main given method of docking application in the test. The chinese patent with application No. 201811173412.X discloses a low scattering carrier with both forward and lateral designs, which gives a modeling design under specific application scenarios instead of the design and optimization method itself.
Disclosure of Invention
In view of the above, the invention provides a low scattering carrier optimization design method by combining the regional decomposition-direct solution electromagnetic modeling and convolutional neural network technology. The method can deeply reflect the basic characteristics of the low-scattering carrier and the physical mechanism behind the low-scattering carrier, and can give consideration to full-wave high-precision calculation efficiency in the aspects of single-round large bandwidth, wide-angle domain, multi-polarization and the like in geometric modeling and electromagnetic modeling.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
the low scattering carrier optimizing design method includes the following steps:
(1) According to the local skin of the subsystem or the component to be tested, the extension is performed in combination with the examination angle domain, so that a rough geometric framework of the low-scattering carrier is formed;
(2) Partitioning the general geometric framework of the low scattering carrier, and performing parameterized geometric expression, wherein the partition geometric framework is at least divided into a front region, a middle region and a rear region, so as to obtain a partitioned geometric model of the low scattering carrier, the front portion generally comprises a peak, the front portion and the middle portion can also comprise edges, the rear portion mostly adopts smooth ending, and the other regions are continuous smooth regions;
(3) Constructing parameter constraint conditions of the partition geometric model in the subsequent change process, wherein the parameter constraint conditions comprise second-order continuity constraint of a smooth transition zone of the curved surface, a curvature/curvature radius change range of a peak and an edge, a normal jump range of the curved surface at an edge stitching joint and an overall size range;
(4) Electromagnetic modeling is carried out on the partition geometric model by adopting an area decomposition method IE-DDM based on an integral equation; specifically, firstly, adopting a high-order non-conformal curved surface grid, and adopting a complete free sectioning and differentiating mode for different partitions of the partition geometric model to obtain the best fitting capacity and the minimum grid number under the condition of containing a sharp peak, a flat area and an edge; further, defining a high order basis function for current spreading on the non-conformal curved surface mesh;
(5) The method comprises the steps of optimizing a low scattering carrier by using a convolutional neural network in machine learning, carrying out mathematical expression on a high-frequency scattering mechanism and a resonance region scattering mechanism, combining the changed key local or whole geometric characteristics, carrying out auxiliary analysis in the machine learning, constructing network logic with integrated technical index requirements, geometric boundary constraint and physical mechanism limitation, and carrying out iteration by means of a forward and backward transmission method, wherein the core of the iteration is the electromagnetic simulation based on the region decomposition technology and the direct solution technology;
(6) For the optimized geometric model, the radar scattering sectional areas under different irradiation conditions are quantitatively output, namely, the average value of RCS in different frequency bands covers typical examination frequency points, all examination directions and two polarization conditions, namely horizontal polarization conditions and vertical polarization conditions, in each frequency band;
the final geometric model is the required low scattering carrier.
Further, the specific mode of the step (1) is as follows:
partially cutting the skin;
setting key points and constructing a line model;
generating a curved surface based on the line;
and (5) primarily closing the curved surface to form a solid.
Further, the specific mode of the step (2) is as follows:
determining partition key boundary points according to the boundary shape, and parameterizing space coordinates;
carrying out parameterized expression of curve boundaries or key envelope scanning lines between boundary points in the region;
the parametric representation is also performed on curved surfaces bordered by curves.
The low scattering carrier optimizing design method includes the following steps:
(1) According to the local skin where the subsystem or the component to be tested is positioned, the extension is carried out by combining the checking angular domain, and an original model of the low scattering carrier is constructed;
(2) Carrying out regional parameterized geometric treatment on the low-scattering carrier by adopting a regional decomposition method, and dividing the low-scattering carrier into at least three regions of front, middle and rear to obtain a geometric model of the low-scattering carrier;
(3) Electromagnetic modeling is carried out on the geometric model by adopting a region decomposition method based on an integral equation;
(4) Adopting non-conformal grids and high-order basis functions to adapt to different geometric characteristics in different partitions, wherein the geometric characteristics comprise a peak area, a flat area and a prismatic area;
(5) The direct solving method is adopted, and the low-rank compression method of the impedance matrix is combined with the order adjustment of the basis function, so that the method is suitable for the multi-excitation characteristics of the low-scattering carrier in electromagnetic simulation processes of different frequency bands, different angles and different polarizations;
(6) Implementing parametric modeling constraint condition limitations, including second-order continuity constraint of a smooth transition zone of a curved surface, curvature/curvature radius variation ranges at the points and edges, normal jump ranges of the curved surface at the connection position and overall size ranges;
(7) Optimizing a low scattering carrier by using a convolutional neural network, mathematically expressing a high-frequency scattering mechanism and a resonance region scattering mechanism, constructing network logic, and iterating by means of forward and backward transmission;
(8) Obtaining quantitative output of simulation calculation results, wherein the quantitative output comprises average values of different frequency bands, and the average values cover typical frequency points in a band, all examination directions and two polarization modes of HH and VV;
and (5) completing the optimal design of the low-scattering carrier.
Compared with the prior art, the invention has the following beneficial effects:
1. the full-wave high-precision computing efficiency in the aspects of single-wheel large bandwidth, wide-angle domain, multi-polarization and the like can be considered in geometric modeling and electromagnetic modeling, and the comprehensive efficiency is far superior to that of the traditional moment method MOM and the common multipole method MLFMA.
2. The invention combines the advanced capability in the current artificial intelligence machine learning, and can remarkably compress the period in multi-round optimization.
Drawings
FIG. 1 is a flow chart of a low scattering carrier optimization design method in an embodiment of the invention;
FIG. 2 is a schematic diagram of a typical partitioning of a metal structured low scattering carrier, comprising front, middle and rear three regions, according to an embodiment of the present invention;
FIG. 3 is a schematic illustration of a portion of the electromagnetic mechanism of a low scattering carrier in an embodiment of the present invention;
FIG. 4 is a schematic diagram of a non-conformal curved higher order grid cell;
FIG. 5 is a schematic illustration of an application of region splitting in an embodiment of the present invention, showing a region splitting of overlap at a boundary;
FIG. 6 is a schematic diagram of the distribution of the compatibility matrix block and the non-compatibility matrix block directly solved in the embodiment of the invention;
FIG. 7 is a flowchart of an exemplary optimization based on convolutional neural network in an embodiment of the present invention.
Detailed Description
The conception, technical advantages, and technical effects of the present invention will be clearly and completely described below with reference to the accompanying drawings to fully understand the objects, features, and effects of the present invention. It should be noted that the specific embodiments described herein are only for explaining the present invention, and do not limit the present invention.
As shown in fig. 1-7, a low scattering carrier optimization design method includes the following steps:
(1) According to the local skin of the subsystem or the component to be tested, the extension is performed in combination with the examination angle domain, so that a rough geometric framework of the low-scattering carrier is formed;
(2) Partitioning the general geometric framework of the low scattering carrier, and performing parameterized geometric expression, wherein the partition geometric framework is at least divided into a front region, a middle region and a rear region, so as to obtain a partitioned geometric model of the low scattering carrier, the front portion generally comprises a peak, the front portion and the middle portion can also comprise edges, the rear portion mostly adopts smooth ending, and the other regions are continuous smooth regions;
(3) Constructing parameter constraint conditions of the partition geometric model in the subsequent change process, wherein the parameter constraint conditions comprise second-order continuity constraint of a smooth transition zone of the curved surface, a curvature/curvature radius change range of a peak and an edge, a normal jump range of the curved surface at an edge stitching joint and an overall size range;
(4) Electromagnetic modeling is carried out on the partition geometric model by adopting an area decomposition method IE-DDM based on an integral equation; specifically, firstly, adopting a high-order non-conformal curved surface grid, and adopting a complete free sectioning and differentiating mode for different partitions of the partition geometric model to obtain the best fitting capacity and the minimum grid number under the condition of containing a sharp peak, a flat area and an edge; further, defining a high order basis function for current spreading on the non-conformal curved surface mesh;
(5) The method comprises the steps of optimizing a low scattering carrier by using a convolutional neural network in machine learning, carrying out mathematical expression on a high-frequency scattering mechanism (including specular reflection, edge diffraction, tip scattering and the like) and a resonance region scattering mechanism, combining the changed key local (specular, edge and tip) or whole (low-frequency resonance) geometric characteristics, carrying out auxiliary analysis in the machine learning, constructing network logic with integrated technical index requirements, geometric boundary constraint and physical mechanism limitation, and carrying out iteration by using a forward and backward transmission method, wherein the core of iteration is the electromagnetic simulation based on the region decomposition technology and the direct solution technology;
(6) For the optimized geometric model, the radar scattering sectional areas under different irradiation conditions are quantitatively output, and the radar scattering sectional areas are mainly the average value of RCS in different frequency bands, wherein the average value covers typical assessment frequency points, all assessment directions (such as independent forward, lateral or backward areas) and two polarization conditions of horizontal (HH) and vertical (VV) in each frequency band such as P, L, S, C, X, ku;
(7) The final geometric model, which is the required low scattering carrier, is also the optimal model under the constraint of geometric boundaries.
The specific mode of the step (1) is as follows:
partially cutting the skin;
setting key points and constructing a line model;
generating a curved surface based on the line;
and (5) primarily closing the curved surface to form a solid.
The specific mode of the step (2) is as follows:
determining partition key boundary points according to the boundary shape, and parameterizing space coordinates;
between boundary points in the region, parametric expression of curve boundaries or key envelope scanning lines is carried out, and polynomials can be adopted;
the curve surface taking the curve as the boundary is also parameterized and expressed, and a rational B spline form can be adopted.
Wherein, FIG. 1 shows a flow chart of the low scattering carrier optimization design method of the invention; FIG. 2 shows a typical partitioning of a low scattering carrier of the metallic structure of the present invention, comprising front, middle and rear three regions; FIG. 3 illustrates a part of the electromagnetic mechanism of a low scattering carrier, namely deflection in the specular strong scattering direction, wherein the rear part and the middle part also contain the points of reducing projection area, reducing discontinuity, reducing travelling wave scattering and the like; FIG. 4 is a non-conformal curved surface high-order grid unit, comprising a curved triangle unit and a curved quadrilateral unit, according to which the low-density subdivision can be performed, and the total unknown quantity number is reduced by combining the application of a high-order basis function on the premise of reflecting the shape of a carrier with high precision; FIG. 5 is a schematic illustration of an area resolution application, showing an area resolution of overlap at a boundary; FIG. 6 is a schematic diagram of the distribution of a compatibility matrix block and a non-compatibility matrix block in direct solution, wherein the compatibility matrix block accords with low-rank characteristics and can be compressed with high efficiency by adopting means of adaptive cross approximation and the like; FIG. 7 is a typical optimization application flow based on convolutional neural networks.
The method for realizing the optimal design of the low-scattering carrier by combining regional decomposition-direct solution electromagnetic modeling and convolution neural network comprises the following steps:
(1) According to the local skin where the subsystem or the component to be tested is positioned, combining technical requirements including examination angle domain, rationalizing and extending, and constructing a preliminary rough model or an original model;
(2) Carrying out zonal parameterization on the low scattering carrier by adopting a zone decomposition technology, and specifically dividing the low scattering carrier into at least three typical zones, namely a front zone, a middle zone and a rear zone;
(3) On the basis of the characteristics, the constructed geometric model is subjected to electromagnetic modeling by adopting a region decomposition (DDM) technology based on an Integral Equation (IE) framework, and particularly, an overlapped (end face) transmission condition or a non-overlapped (gird) transmission condition can be adopted;
(4) Based on the characteristics, non-conformal grids and high-order basis functions are further adopted to adapt to different geometric characteristics in different partitions, wherein the geometric characteristics comprise a peak area, a flat area and a prismatic area;
(5) Based on the characteristics, an efficient direct solving technology is adopted, including but not limited to an H matrix technology, and a basis function order adjustment and impedance matrix low-rank compression technology are combined so as to adapt to the multi-excitation characteristics of a low scattering carrier in electromagnetic simulation processes of different frequency bands, different angles and different polarization;
(6) Based on the characteristics, parametric modeling constraint condition limitation is implemented, wherein the constraint condition comprises second-order continuity constraint of a smooth transition zone of a curved surface, the constraint comprises curvature/curvature radius change ranges at the points and edges, the normal jump range of the curved surface at the joint and the overall size range;
(7) On the basis of the characteristics, a Convolutional Neural Network (CNN) in machine learning is combined to optimize a low-scattering carrier, a part of high-frequency scattering mechanism and a resonance region scattering mechanism are expressed mathematically, network logic is constructed, and process iteration is carried out by means of forward and backward transmission;
(8) Based on the steps, quantitative output based on simulation calculation results is given, wherein the quantitative output comprises average values of different frequency bands, and the average values cover typical frequency points in the band, all examination directions (azimuth and pitching), and two polarization modes of HH and VV;
(9) The optimization method can be applied to metal type, medium type and metal medium composite type low-scattering carriers.
The basic principle of the method is as follows:
(1) Zoning design of carrier
The original or coarse low scattering carrier of the preliminary construction can generally be further divided into front, middle and rear three regions.
The front part of the low scattering carrier is mainly characterized by a sharp-top mode, the normal directions (scattering peak directions) of two sides of the sharp-top inclination angle are required to deviate from the index checking range as far as possible in consideration of main scattering contribution, the higher-order scattering sources corresponding to the area mainly are mainly diffraction energy, when the front part is subjected to curved surface construction, the surface smoothness of a curved surface is ensured as far as possible, discontinuous points are reduced, and a creeping wave path towards a checking area is removed to the maximum extent.
In the middle design, a plurality of surfaces are closed by using a smooth curved surface, a smooth curved surface structure is selected, discontinuous boundary diffraction sources of the surfaces can be removed, the smooth curved surface meets the second-order continuity as far as possible, and a non-uniform rational B-spline curved surface widely adopted in the international industry can be selected.
In the rear design, in order to reduce the projection of the area to the incidence direction of far-field plane waves in the examination area, the transition from a curved surface with a large curvature radius to a curved surface with a small curvature radius is selected as far as possible for ending treatment, and the method has the advantages that the inherent forward projection scattering area is not increased basically. As with the front and middle designs, the curved surface of the region should also meet the second order continuity as much as possible to eliminate discontinuous boundary diffraction and traveling wave reflection.
(2) Mesh subdivision and basis function
To implement mesh subdivision, non-conformal curved triangles and curved quadrilateral meshes may be employed and higher order basis functions defined. Taking a parallelogram grid as an example, the adaptive higher-order basis functions can be expressed as follows in the local parameter coordinate system:
the edge basis functions and the face element basis functions are included.
(3) Regarding region decomposition
In particular, an overlapping type or non-overlapping type region decomposition algorithm can be adopted.
For the overlap-type region decomposition algorithm, for example, the closed surface S of the original low scattering carrier can be decomposed into three open sub-surfaces, the remaining two sub-surfaces being considered as a new open surface when one of the sub-surfaces is considered. The buffer area introduced by each subdomain is only one layer of saw-tooth triangle units. Further decomposing each sub-domain into two overlapping open surfaces, in each sub-domain, the basis function conditions defined at the overlapping surfaces ensure normal continuity of the current across the artificial boundary.
In the described sub-domain model, the overlapping currents do not participate in the scattering contribution at the same time, in general only one set of scattering contributions of the overlapping currents need to be considered in each sub-domain model. At the same time, different overlapping current participation contributions need to be selected in different sub-domain models. From a mathematical perspective, such a process can avoid the singular matrix problem caused by the same set of overlapping currents participating in scatter calculations in both sub-domain models.
In addition, if non-overlapping transmission boundaries are employed for the low scattering carrier in structural decomposition, a new internal penalty term can be introduced to ensure normal continuity of the current across the boundary. In the internal penalty process, an expression of the error charge and the test charge is determined by combining a current continuity equation, and the potential generated by the error charge is tested according to a dual pairing test method so as to achieve the aim of minimizing the potential energy of the error charge. The advantage of this approach is that: (1) An additional boundary subdomain or a buffer zone is not required to be introduced to construct a preprocessing matrix, so that the burden of preprocessing work is reduced; (2) The selection of the stability parameters can be avoided without introducing a stability term, and the stability parameter has better usability and robustness.
(4) With respect to direct solution
The direct solving method can construct and obtain the inverse matrix of the impedance matrix, and for any right-end term vector (namely corresponding to the excitation of radar waves), the coefficient matrix can be obtained only by carrying out back-substitution, so that the method has higher calculation efficiency.
The traditional direct solving method has higher solving complexity and is difficult to apply to practical engineering problems. The quick method for directly solving is based on the fact that sub-matrix information formed by coupling between a remote basic function set and a test function set is redundant, and the low-rank compression method is combined for compression, so that the impedance matrix is sparsely represented, the memory requirement of the system matrix is reduced, and the quick solving of the system matrix is realized.
Specifically, the invention adopts an H-matrix method, which adopts a tree data structure to represent a system matrix as a multi-layer structure for sparsely approximating a dense matrix. The matrix formed by coupling the cores of the H-matrix with two groups of basis functions that are far apart from each other is redundant, i.e. the matrix can be approximated by a low rank. The system matrix is divided into a multi-layer structure to reasonably divide the system matrix, and different processing methods are adopted for different partitioned matrixes so as to accelerate the solution of the whole system matrix. The compatibility condition is adopted to determine whether the two groups of basis functions are well separated in geometric space, namely whether the matrix block formed by coupling meets the low-rank characteristic. Matrix blocks meeting the compatibility condition are called compatibility blocks, directly serve as leaf nodes of the structural tree, and do not need to be subdivided. Matrix blocks that do not meet the compatibility condition are referred to as non-compatibility matrix blocks. If the two base function sets corresponding to the non-compatibility matrix blocks are not in the finest layer, the next layer is judged continuously. If two basic function sets corresponding to the non-compatible matrix block are both at the finest layer, the matrix block is the non-compatible matrix block. The compatibility matrix block corresponds to a low rank matrix and may be populated using a low rank algorithm (e.g., adaptive cross approximation, ACA) compression. The non-compatible matrix blocks are then filled directly.
Direct inversion of H-matrix and LU decomposition of H-matrix can be constructed by using basic arithmetic operations between H-matrix, which are computationally more efficient than direct inversion, and operations in which are recursively implemented.
By the method, high-efficiency full-wave simulation calculation of the low-scattering carrier under the condition of multiple scattering excitation can be realized.
(5) CNN with respect to convolutional neural networks
Convolutional neural networks originate from biological processes in which the pattern of linkages between neurons is very similar to that of animal visual cortex tissue, a type of artificial intelligence machine learning. In view of the complexity of the low scattering carrier design problem of the present invention, a three-dimensional convolutional neural network will be employed. The optimization model is built on the basis of a feedforward neural network model and comprises a convolution layer, a pooling layer and a full-connection layer.
In the convolution layer, the dot product result between a region of the input model and the weight matrix (i.e., the filter) is calculated, the filter is slid over the entire physical model, and the same dot product calculation operation is repeated. The pooling layer is mainly used for reducing the space dimension. In the fully connected layer, the output of the last convolutional layer is flattened and each node of the current layer is connected to a node of the next layer. The general operation is as follows:
1) Convolution operation: carrying out convolution operation on the characteristic graph of the previous layer and a learnable convolution kernel, outputting a convolution result after an activation function to form neurons of the layer, thus forming the characteristic graph of the layer, connecting the input of each neuron with a local receptive field of the previous layer, and extracting the local characteristic;
2) Pooling operation: the input model geometry and electromagnetic information are partitioned into non-overlapping regions, and the network spatial resolution is reduced for each region by pooling (downsampling) operations, which can be performed with an average value, i.e., an average value within the region is calculated.
3) Full join operation: after the input model is subjected to multiple convolution kernel pooling operation, multiple sets of result information are output, and multiple sets of information are sequentially combined into one set of information through full-connection operation.
4) Identification operation: based on the operation, a layer of network is added for regression calculation according to the problem demand.
The following are more specific examples:
the low scattering carrier shown in fig. 2, which serves a certain antenna aperture test, is designed to constrain the two-dimensional dimensions to approximately 2.10m x 1.00m. When designing, dividing the front, middle and rear three typical areas; electromagnetic modeling is carried out on the constructed geometric model by adopting a region decomposition technology based on an integral equation frame, and an overlapping transmission condition is adopted; when the grid is split, a curved quadrilateral grid is adopted and matched with a high-order non-conformal basis function so as to adapt to different geometric characteristics in different partitions, wherein the geometric characteristics comprise a peak area and a flat area; then, an efficient direct solving technology based on an H matrix is adopted, and a self-adaptive cross approximation ACA impedance matrix low-rank compression technology is combined to adapt to multiple excitation characteristics in electromagnetic simulation processes of different frequency bands, different angles and different polarizations; the method is characterized in that carrier design optimization is carried out based on a convolutional neural network in machine learning, parametric modeling constraint condition limitation is applied, the constraint comprises second-order continuity constraint, overall size range, actual technical index requirement and the like of a curved surface smooth transition region, partial high-frequency scattering mechanism and resonant region scattering mechanism are expressed mathematically, and an optimization direction and path are determined by means of advanced artificial intelligence.
Through practice, RCS statistical average values of the obtained low-scattering carrier under the conditions of typical angular regions, different pitch angles and different polarization of 6 microwave and millimeter wave typical frequency bands are shown in tables 1-6, and the RCS statistical average values can be found to reach a good level, so that the RCS statistical average values can meet the actual evaluation requirements.
Table 1: freq=f1
Table 2: freq=f2
Table 3: freq=f3
Table 4: freq=f4
Table 5: freq=f5
Table 6: freq=f6
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and it should be noted that it is possible for those skilled in the art to make several improvements and modifications without departing from the technical principle of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention.
Claims (4)
1. The low scattering carrier optimization design method is characterized by comprising the following steps of:
(1) According to the local skin of the subsystem or the component to be tested, the extension is performed in combination with the examination angle domain, so that a rough geometric framework of the low-scattering carrier is formed;
(2) Partitioning the general geometric framework of the low scattering carrier, and performing parameterized geometric expression, wherein the partition geometric framework is at least divided into a front region, a middle region and a rear region, so as to obtain a partitioned geometric model of the low scattering carrier, the front portion generally comprises a peak, the front portion and the middle portion can also comprise edges, the rear portion mostly adopts smooth ending, and the other regions are continuous smooth regions;
(3) Constructing parameter constraint conditions of the partition geometric model in the subsequent change process, wherein the parameter constraint conditions comprise second-order continuity constraint of a smooth transition zone of the curved surface, a curvature/curvature radius change range of a peak and an edge, a normal jump range of the curved surface at an edge stitching joint and an overall size range;
(4) Electromagnetic modeling is carried out on the partition geometric model by adopting an area decomposition method IE-DDM based on an integral equation; specifically, firstly, adopting a high-order non-conformal curved surface grid, and adopting a complete free sectioning and differentiating mode for different partitions of the partition geometric model to obtain the best fitting capacity and the minimum grid number under the condition of containing a sharp peak, a flat area and an edge; further, defining a high order basis function for current spreading on the non-conformal curved surface mesh;
(5) The method comprises the steps of optimizing a low scattering carrier by using a convolutional neural network in machine learning, carrying out mathematical expression on a high-frequency scattering mechanism and a resonance region scattering mechanism, combining the changed key local or whole geometric characteristics, carrying out auxiliary analysis in the machine learning, constructing network logic with integrated technical index requirements, geometric boundary constraint and physical mechanism limitation, and carrying out iteration by means of a forward and backward transmission method, wherein the core of the iteration is the electromagnetic simulation based on the region decomposition technology and the direct solution technology;
(6) For the optimized geometric model, the radar scattering sectional areas under different irradiation conditions are quantitatively output, namely, the average value of RCS in different frequency bands covers typical examination frequency points, all examination directions and two polarization conditions, namely horizontal polarization conditions and vertical polarization conditions, in each frequency band;
the final geometric model is the required low scattering carrier.
2. The method for optimizing design of a low scattering carrier according to claim 1, wherein the specific mode of step (1) is as follows:
partially cutting the skin;
setting key points and constructing a line model;
generating a curved surface based on the line;
and (5) primarily closing the curved surface to form a solid.
3. The method for optimizing design of a low scattering carrier according to claim 1, wherein the specific mode of the step (2) is as follows:
determining partition key boundary points according to the boundary shape, and parameterizing space coordinates;
carrying out parameterized expression of curve boundaries or key envelope scanning lines between boundary points in the region;
the parametric representation is also performed on curved surfaces bordered by curves.
4. The low scattering carrier optimization design method is characterized by comprising the following steps of:
(1) According to the local skin where the subsystem or the component to be tested is positioned, the extension is carried out by combining the checking angular domain, and an original model of the low scattering carrier is constructed;
(2) Carrying out regional parameterized geometric treatment on the low-scattering carrier by adopting a regional decomposition method, and dividing the low-scattering carrier into at least three regions of front, middle and rear to obtain a geometric model of the low-scattering carrier;
(3) Electromagnetic modeling is carried out on the geometric model by adopting a region decomposition method based on an integral equation;
(4) Adopting non-conformal grids and high-order basis functions to adapt to different geometric characteristics in different partitions, wherein the geometric characteristics comprise a peak area, a flat area and a prismatic area;
(5) The direct solving method is adopted, and the low-rank compression method of the impedance matrix is combined with the order adjustment of the basis function, so that the method is suitable for the multi-excitation characteristics of the low-scattering carrier in electromagnetic simulation processes of different frequency bands, different angles and different polarizations;
(6) Implementing parametric modeling constraint conditions including second order continuity constraint of smooth transition zone of curved surface, curvature/curvature radius variation range of sharp top and edge, normal jump range of curved surface at connection position and
an overall size range;
(7) Optimizing a low scattering carrier by using a convolutional neural network, mathematically expressing a high-frequency scattering mechanism and a resonance region scattering mechanism, constructing network logic, and iterating by means of forward and backward transmission;
(8) Obtaining quantitative output of simulation calculation results, wherein the quantitative output comprises average values of different frequency bands, and the average values cover typical frequency points in a band, all examination directions and two polarization modes of HH and VV;
and (5) completing the optimal design of the low-scattering carrier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310550321.8A CN116629109B (en) | 2023-05-16 | 2023-05-16 | Low-scattering carrier optimization design method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310550321.8A CN116629109B (en) | 2023-05-16 | 2023-05-16 | Low-scattering carrier optimization design method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116629109A true CN116629109A (en) | 2023-08-22 |
| CN116629109B CN116629109B (en) | 2024-05-24 |
Family
ID=87641049
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310550321.8A Active CN116629109B (en) | 2023-05-16 | 2023-05-16 | Low-scattering carrier optimization design method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116629109B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2521608C1 (en) * | 2013-05-13 | 2014-07-10 | Открытое акционерное общество "Конструкторское бюро по радиоконтролю систем управления, навигации и связи" (ОАО "КБ "Связь") | Method for invisible detection of mobile objects |
| US20170018118A1 (en) * | 2015-07-17 | 2017-01-19 | Board Of Trustees Of Michigan State University | Electromagnetic Scattering Analysis Module Using Subdivision Based Isogeometric Analysis |
| CN106428625A (en) * | 2016-09-14 | 2017-02-22 | 北京环境特性研究所 | Low-scattering carrier used for RCS test |
| CN109212504A (en) * | 2018-10-09 | 2019-01-15 | 北京环境特性研究所 | To scattering carrier low with what is laterally designed before taking into account |
| CN111504952A (en) * | 2020-04-15 | 2020-08-07 | 成都飞机工业（集团）有限责任公司 | Low-scattering carrier with both horizontal polarization and vertical polarization and testing method thereof |
| CN113075634A (en) * | 2021-03-29 | 2021-07-06 | 北京环境特性研究所 | RCS (Radar Cross section) evaluation carrier for vertical polarized waves |
| WO2021212680A1 (en) * | 2020-04-23 | 2021-10-28 | 南京英尼格玛工业自动化技术有限公司 | Dynamic grid division-based method and system for analyzing load of non-standard part |
| CN114239239A (en) * | 2021-11-29 | 2022-03-25 | 电子科技大学 | Direct sparse solving method for rapid simulation of electromagnetic characteristics of bullet and eye meeting target |
| CN115099161A (en) * | 2022-07-22 | 2022-09-23 | 南京理工大学 | Low-scattering carrier shape construction method |
-
2023
- 2023-05-16 CN CN202310550321.8A patent/CN116629109B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2521608C1 (en) * | 2013-05-13 | 2014-07-10 | Открытое акционерное общество "Конструкторское бюро по радиоконтролю систем управления, навигации и связи" (ОАО "КБ "Связь") | Method for invisible detection of mobile objects |
| US20170018118A1 (en) * | 2015-07-17 | 2017-01-19 | Board Of Trustees Of Michigan State University | Electromagnetic Scattering Analysis Module Using Subdivision Based Isogeometric Analysis |
| CN106428625A (en) * | 2016-09-14 | 2017-02-22 | 北京环境特性研究所 | Low-scattering carrier used for RCS test |
| CN109212504A (en) * | 2018-10-09 | 2019-01-15 | 北京环境特性研究所 | To scattering carrier low with what is laterally designed before taking into account |
| CN111504952A (en) * | 2020-04-15 | 2020-08-07 | 成都飞机工业（集团）有限责任公司 | Low-scattering carrier with both horizontal polarization and vertical polarization and testing method thereof |
| WO2021212680A1 (en) * | 2020-04-23 | 2021-10-28 | 南京英尼格玛工业自动化技术有限公司 | Dynamic grid division-based method and system for analyzing load of non-standard part |
| CN113075634A (en) * | 2021-03-29 | 2021-07-06 | 北京环境特性研究所 | RCS (Radar Cross section) evaluation carrier for vertical polarized waves |
| CN114239239A (en) * | 2021-11-29 | 2022-03-25 | 电子科技大学 | Direct sparse solving method for rapid simulation of electromagnetic characteristics of bullet and eye meeting target |
| CN115099161A (en) * | 2022-07-22 | 2022-09-23 | 南京理工大学 | Low-scattering carrier shape construction method |
Non-Patent Citations (5)
| Title |
|---|
| HANFEI ZHANG: "An Optimized Method of Low RCS Carrier Design", 《2021 INTERNATIONAL APPLIED COMPUTATIONAL ELECTROMAGNETICS SOCIETY (ACES-CHINA) SYMPOSIUM》, 11 March 2022 (2022-03-11) * |
| 张扬;艾俊强;王健;张维仁;: "隐身飞机机身侧棱电磁散射特点分析", 航空工程进展, no. 01, 28 February 2017 (2017-02-28) * |
| 徐伊达等: "飞行器部件低散射载体的分析与设计", 《电讯科技》, 23 February 2021 (2021-02-23) * |
| 梁爽;刘庆杰;聂暾;郭文;张斌;: "RCS测试中的低散射载体设计及仿真分析", 计算机测量与控制, no. 08, 25 August 2017 (2017-08-25) * |
| 王军, 聂在平, 胡俊, 弓晓东: "复杂目标电磁散射特性分析中的几何建模", 电波科学学报, no. 01, 25 February 2003 (2003-02-25) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116629109B (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fezoui et al. | Convergence and stability of a discontinuous Galerkin time-domain method for the 3D heterogeneous Maxwell equations on unstructured meshes | |
| EP2378444B1 (en) | Method, device and computer program product for determining an electromagnetic near field of a field excitation source for an electrical system | |
| Taubin | A signal processing approach to fair surface design | |
| De Adana | Practical applications of asymptotic techniques in electromagnetics | |
| CN109766588B (en) | Parallelization time domain hybrid electromagnetic method based on iteration method | |
| Pérez et al. | Analysis of antennas on board arbitrary structures modeled by NURBS surfaces | |
| King et al. | Optimal bit allocation in compressed 3D models | |
| CN110322547A (en) | A kind of adaptive Tetrahedron subdivision method of reservoir | |
| CN116629109B (en) | Low-scattering carrier optimization design method | |
| CN113985408B (en) | Inverse synthetic aperture radar imaging method combining gate unit and transfer learning | |
| CN114755652A (en) | Method for acquiring electrically large-size target broadband RCS (radar cross section) based on ACA (advanced communication architecture) and CAT (CAT) | |
| CN112327374B (en) | DGTD forward modeling method for GPU ground penetrating radar complex medium | |
| Delgado et al. | Combination of ray-tracing and the method of moments for electromagnetic radiation analysis using reduced meshes | |
| CN113255194A (en) | Autonomous intelligent straight hexahedron subdivision method based on parametric curved surface | |
| CN109614732B (en) | Electromagnetic compatibility modeling method and device for object | |
| CN111597744A (en) | Rapid frequency sweep simulation method based on region decomposition | |
| CN109241556B (en) | Random asperity object modeling method and storage medium | |
| CN114239439B (en) | Automatic filter design method based on tribal algorithm | |
| US8031957B1 (en) | Rewritable lossy compression of graphical data | |
| Di Giampaolo et al. | Astigmatic beam tracing for GTD/UTD methods in 3-D complex environments | |
| Yoo | Filling holes in large polygon models using an implicit surface scheme and the domain decomposition method | |
| CN110196961B (en) | Aircraft radar echo prediction method of non-cooperative uncertain shape | |
| Belokrys-Fedotov et al. | Adaptive Mesh Refinement for Electromagnetic Simulation | |
| Fuchs | Almost regular triangulations of trimmend nurbs-solids | |
| Clees et al. | RBF-metamodel driven multi-objective optimization and its applications |
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 |