JP6672938B2 - Frequency selection board design method and frequency selection board design program - Google Patents

Description

  The present invention relates to a frequency selection plate design method and a frequency selection plate design program.

2. Description of the Related Art Conventionally, a frequency selective plate (FSS: frequency selective surfaces) has been known as a device capable of controlling the selectivity (selective transmission or selective shielding) of a transmission frequency band of an electromagnetic wave propagating in space.
Various numerical calculation methods have been proposed as a design method of the frequency selection plate. For example, in the following Non-Patent Document 1, the spectral domain moment method is applied.

S. Chakravarty et al., "Application of a Microgenetic Algorithm (MGA) to the Design of Broad-Band Microwave Absorbers Using Multiple Frequency Selective Surface Screens Buried in Dielectrics", IEEE Trans. AP, Vol. 50, no. 3, March 2002

In the spectral domain moment method, each cell of the M × N cell model is usually set as a conductor cell or a non-conductor cell. Note that, since fast Fourier transform is usually used to speed up the calculation, M and N are set as 2 ^m and 2 ⁿ (m and n are arbitrary natural numbers).
In this case, in the example 16 × 16 cell model, Brute-force method for a variable space polarization symmetry ² 36 1-layer even in consideration of (70 billion combinations) (total per method), let alone Genetic Algorithms Requires a lot of evolutionary steps and time.
The present invention has been made in view of such circumstances, and an object of the present invention is to efficiently design a frequency selection plate.

In order to achieve the above object, a method for designing a frequency selection plate according to the first aspect of the present invention is a method of designing a frequency selection plate, wherein each cell region of an M × N cell model obtained by dividing the frequency selection plate into M × N cells is a conductive cell or a non-conductive cell And designating the frequency selection plate by specifying the arrangement of the conductor cells and the non-conductor cells in the M × N cell model, wherein K is an arbitrary natural number (M / N (K) × (N / K) a reduced cell model creating step of creating an arrangement geometry group of the conductor cells and the non-conductor cells in the cell model, and a predetermined frequency characteristic is obtained from the arrangement geometry group using a Brute-force method. A good solution extracting step for extracting a good good solution, and converting the unit cell length of the good solution by 1 / K to the M × N cell model, It optimizes processing to Dell, characterized in that it contains, and optimization step of extracting the arrangement geometry that conforms to the frequency characteristic.
The design method of the frequency selection plate according to the invention of claim 2, wherein, when the arrangement geometries are successively arranged, the arrangement geometry in which the same geometry as another arrangement geometry appears in the group of the arrangement geometries. The method further comprises a geometry reduction step of reducing the number of arrangement geometries included in the arrangement geometry group by contracting with another arrangement geometry.
In the designing method of a frequency selection plate according to the invention of claim 3, in the optimizing step, the arrangement geometry that matches the frequency characteristic is obtained by applying a genetic algorithm using the excellent solution M × N cell model as a gene. Extraction.
The frequency selection plate designing method according to the invention of claim 4 is characterized in that in the optimizing step, the occurrence probability of the arrangement geometry that does not match the frequency characteristic is reduced using a penalty function based on the frequency characteristic. And
In the method for designing a frequency selection plate according to a fifth aspect of the present invention, in the excellent solution extraction step, for each arrangement geometry included in the arrangement geometry group, one of the (M / K) × (N / K) cell models is used. The frequency characteristic in each arrangement period is calculated by changing the arrangement period, which is the length of a side, within a predetermined range.
According to a sixth aspect of the present invention, there is provided a frequency selection board designing program for causing a computer to execute the frequency selection board designing method according to any one of the first to fifth aspects.

According to the first aspect of the present invention, the analysis is performed by a Brute-force method using a cell model smaller than the M × N cell model to be designed, and an optimization process is performed on the superior solution. Therefore, a simple initial solution is used. The possibility that an effective design solution can be obtained in a shorter time is increased, which is advantageous in shortening the calculation time in designing the frequency selection plate.
According to the second aspect of the present invention, it is possible to reduce arrangement geometries which are substantially the same when they are arranged consecutively, which is advantageous in shortening the calculation time of the Brute-force analysis.
According to the invention of claim 3, it is possible to efficiently obtain a suitable design solution by using a genetic algorithm. Further, an effective design solution can be obtained even if the number of elements and the number of generations of the genetic algorithm are reduced, which is advantageous in shortening the calculation time when designing the frequency selection plate.
According to the invention of claim 4, when the genetic algorithm is executed, the occurrence probability of the arrangement geometry that does not match the frequency characteristic is reduced by using the penalty function based on the target frequency characteristic. This is advantageous in improving the probability of being obtained.
According to the fifth aspect of the present invention, not only the arrangement of the conductor cells and the non-conductor cells on the cell model but also the arrangement period is changed, which is advantageous in designing a frequency selection plate that is more suitable for the target frequency characteristics.
According to the invention of claim 6, it is possible to execute the above-described method for designing a frequency selection plate using a computer.

5 is a flowchart illustrating a procedure of a method of designing a frequency selection plate according to the embodiment. It is explanatory drawing which shows an example of an 8x8 cell model. It is explanatory drawing which shows an example of the same geometry used as a contraction object. It is explanatory drawing which shows an example of the excellent solution of the Brute-force method analysis. 5 is a graph showing the frequency characteristics of the excellent solution in FIG. It is explanatory drawing which shows an example of the excellent solution of a genetic algorithm operation. 7 is a graph showing the frequency characteristics of the excellent solution in FIG. It is a table | surface which shows the comparison result of calculation required time. FIG. 2 is a block diagram illustrating a configuration of a computer 20.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Each process described in the present embodiment is realized by the computer 20 shown in FIG. 9 executing a frequency selection plate design program.
As shown in FIG. 9, the computer 20 includes a CPU 2002, a ROM 2004, a RAM 2006, a hard disk device (HDD) 2008, a disk device 2010, a keyboard 2012, a mouse 2014, a display 2016, a printer 2018, an input / output interface (I / F) 2020, and the like. have.
The ROM 2004 stores a control program and the like, and the RAM 2006 provides a working area.
The hard disk device 2008 stores a dedicated program (frequency selection plate design program) for designing a frequency selection plate.
The disk device 2010 records and / or reproduces data on a recording medium such as a CD or a DVD.
The keyboard 2012 and the mouse 2014 accept an operation input by a measurer.
The display 2016 displays and outputs data, and the printer 2018 prints and outputs data. The display 2016 and the printer 2018 output data.
The input / output interface 2020 exchanges data with another computer or the like.
Note that the processes described below are not necessarily executed by one computer 20, and may be executed by a plurality of computers 20. For example, steps S10 to S20 may be performed by the first computer 20, and steps S22 to S26 may be performed by the second computer 20, respectively. Alternatively, only the analysis of the Brute-force method and the operation of the genetic algorithm, which will be described later, may be performed by a high-performance computer having a high processing capability.

FIG. 1 is a flowchart illustrating a procedure of a method for designing a frequency selection plate according to the embodiment.
The design method of the frequency selection plate according to the embodiment is such that each cell region of the M × N cell model obtained by dividing the frequency selection plate into M × N cells is designated as a conductor cell or a non-conductor cell, respectively. The frequency selection plate is designed by specifying the arrangement of the conductor cells and the non-conductor cells in.
In the present embodiment, a case will be described in which a 16 × 16 cell model (M, N = 16) obtained by dividing a unit geometry in a frequency selection plate into 16 × 16 cells is used. In addition, the laminated configuration is one layer.

Further, the frequency selection plate to be designed is required to have good predetermined frequency characteristics. For example, when the transmission band is set to α GHz ± β GHz (α and β are arbitrary positive numbers) and the shield band is required to be a band other than α GHz ± β GHz, if this specification is made more specific, for example, ( Transmission loss from α-β) GHz to (α + β) GHz is γ dB or less (γ is an arbitrary positive number), (α-δ ₁ ) GHz or less, and (α + δ ₂ ) GHz or more (δ ₁ , δ ₂ is less than β The target performance (target frequency characteristic) can be set such that the transmission loss in the band of a large positive number is equal to or more than ε dB (ε is a positive number larger than γ).

  More specifically, for example, when the required performance is a range outside the range of 9.4 ± 0.05 GHz and the shielding band is outside the range of 9.4 ± 0.05 GHz as required specifications, the target performance is set to 9.35３５. At 9.45 GHz, the transmission loss can be set to 1 dB or less (transmission band performance) and at 8 GHz or less and 11 GHz or more, the transmission loss can be set to 10 dB or more (shielding band performance).

First, a reduced cell model creation step for creating an arrangement geometry group of conductor cells and non-conductor cells in an (M / K) × (N / K) cell model where K is an arbitrary natural number is performed (step S10). .
In this embodiment, assuming that K = 2, for each of the 64 cells in the 8 × 8 cell model, it is determined whether the cell is a conductor cell or a non-conductor cell, and a placement geometry group is created.
Here, in consideration of the polarization symmetry of the frequency selection plate, it is preferable that the 8 × 8 cell model has a vertically and horizontally symmetrical geometry.
Therefore, as shown in FIG. 2, in the 8 × 8 cell model, ten cells may be determined to be either conductive cells or non-conductive cells. That is, the maximum number of arrangement geometries in the arrangement geometry group is 2 ¹⁰ = 1024.

Next, the inferior solution in the 8 × 8 cell model is deleted (step S12).
Inferior solutions are placement geometries that are clearly not expected to meet target performance. Specifically, for example, a case where all the cells in the 8 × 8 cell model are conductor cells or non-conductor cells, or a case where the ratio of the conductor cells is extremely small (for example, one conductor cell out of 64 cells). .

Next, of the placement geometry groups, when the placement geometry is sequentially arranged, the placement geometry in which the same geometry as the other placement geometry appears is reduced with the other placement geometry, and is placed in the placement geometry group. A geometry reduction step for reducing the number of included arrangement geometries is performed (step S14).
Specifically, for example, when four arrangement geometries 30 of the 8 × 8 cell model shown in FIG. 3A are arranged in a square shape as shown in FIG. 3B, another arrangement geometry 32 shown in FIG. Appears.
Therefore, the arrangement geometry 30 of FIG. 3A and the arrangement geometry 32 of FIG. 3C are substantially the same, and can be reduced to either.
After steps S12 and S14, unique arrangement geometries can be reduced to 386.

Next, a good solution extraction step of extracting a good solution having a predetermined frequency characteristic (target performance) from the arrangement geometry group using the Brute-force method is performed.
The excellent solution extraction step includes a step of performing the Brute-force method analysis (step S16) and a step of extracting a good solution from the analysis result (step S18).
In step S16, for each of the arrangement geometries included in the arrangement geometry group, the arrangement period, which is the length of one side of the (M / K) × (N / K) cell model, is changed within a predetermined range, Is calculated. That is, the calculation parameters at the time of the analysis of the Brute-force method include the calculation frequency and the arrangement period of the arrangement geometry.
The calculation frequency specifies a predetermined range centered on the center frequency of the transmission band, that is, a range of αGHz ± ζ (ζ is a positive number equal to or more than ε) in predetermined frequency steps (for example, in 1 / ηGHz steps). Specifically, for example, the band of 2.0 to 18.0 GHz is specified in 0.2 GHz steps. In this case, the frequency to be calculated is a total of 81 waves.
In the arrangement cycle, a predetermined size range is specified in predetermined size steps. Specifically, for example, a range of 3 to 50 mm is specified in 1 mm steps. In this case, the array period to be calculated has a total of 48 dimensions.
In the Brute-force method analysis, the transmission characteristics are calculated for all the patterns while changing these parameters for each arrangement geometry.
In addition, when the frequency selection plate has a multilayer structure, the lamination configuration is included in the analysis parameters.
Also, material constants (relative permittivity, dielectric loss tangent), thickness, etc. of the frequency selection plate are input as appropriate.

In the excellent solution extraction in step S18, the target performance is converted into an objective function, and each operation result is represented as a vector having the value of each objective function as a component. Then, this vector group is ranked, and a predetermined number of upper solutions are extracted as excellent solutions.
Taking the above target performance as an example, for example, the objective function is v1: the absolute value of the power transmittance (dB) of 9.4 GHz, v2: the average power transmittance (dB) of the low frequency band 2 to 8 GHz, v3: the high frequency band 11 A vector [v1; v2; v3] corresponding to each calculation result is generated with an average power transmittance (dB) of １８18 GHz. Then, from all the vectors [v1; v2; v3], a predetermined number of vectors having a high degree of matching with the target performance (that is, v1 is large and v2 and v3 are small) are extracted.

In the present embodiment, for example, the 12 solutions shown in FIG. 4 are extracted as excellent solutions.
FIG. 4 shows a geometry number for identifying the arrangement geometry in order from the top, a specific shape of the arrangement geometry, an arrangement period, and a legend name which is a serial number of 12 excellent solutions.
Of the twelve solutions, serial numbers 5 and 6 and 10 and 11 have the same arrangement geometry dimensional difference.
FIG. 5 shows the frequency characteristics of the excellent solution shown in FIG.
The vertical axis of the graph in FIG. 5 is the power transmittance (dB), and the horizontal axis is the frequency (GHz).
For comparison, FIG. 5 also shows a reference cross dipole aperture type frequency characteristic (legend name “cross”).
Excellent solutions include those that are not necessarily better than the cross dipole shape.

Next, the unit cell length of the good solution obtained by the Brute-force method analysis is multiplied by 1 / K to convert it to an M × N cell model, and an optimization process is performed on the good solution M × N cell model. Then, an optimization step of extracting an M × N cell model that matches the target frequency characteristic is performed.
In this embodiment, it is assumed that a genetic algorithm (multi-objective genetic algorithm) is applied to the above-described excellent solution M × N cell model as a gene to extract an arrangement geometry that matches the frequency characteristics.
For the optimization step, in addition to the genetic algorithm, for example, a global optimization method such as an annealing method (simulated annealing) or a PSO (particulate swarm optimization) can be used.
The optimization step includes a step of converting the (M / K) × (N / K) cell model into an M × N cell model (step S20), a step of performing a genetic algorithm operation (step S22), and extracting a good solution. A step (Step S24) and a step of determining a final design plan (Step S26) are included.

  In step S20, the unit cell length of the excellent solution, which is the (M / K) × (N / K) cell model, is multiplied by 1 / K to be converted into an M × N cell model. In the present embodiment, since K = 2, the unit cell length of the 8 × 8 cell model was halved and converted to a 16 × 16 cell model. The 16 × 16 cell model of this excellent solution becomes the initial value (seed) of the gene.

In step S22, a genetic algorithm operation is performed by specifying a GA (Genetic Algorithm) parameter including the number of elements and the number of generations. In the present embodiment, a plurality of genetic algorithm operations are performed by specifying different GA parameters.
In the genetic algorithm operation, processes such as generation of an initial element set, selection, crossover, and sudden change are performed for a specified number of generations.
In the genetic algorithm operation, a penalty function based on a target frequency characteristic may be used to reduce the occurrence probability of an arrangement geometry that does not match the frequency characteristic. The penalty function gives a penalty to a gene that does not satisfy the target frequency characteristics (the objective function value is outside the threshold).
In the present embodiment, of the genetic algorithm operations performed a plurality of times, the case where the penalty function is used and the case where the penalty function is not used are used together.

In step S24, a good solution in the genetic algorithm operation is extracted.
Specifically, in each of the genetic algorithm operations performed a plurality of times, a Pareto solution is extracted, and a duplicate solution is removed from the Pareto solution. Extract as a solution.
FIG. 6 shows an example of a good solution in the genetic algorithm operation, and FIG. 7 shows a frequency characteristic of the good solution shown in FIG. The vertical axis of the graph in FIG. 7 is the power transmittance (dB), and the horizontal axis is the frequency (GHz). In FIG. 6, legend names, which are identification numbers (No.) of the ten excellent solutions, a geometry number for identifying the arrangement geometry, a specific shape of the arrangement geometry, and an arrangement cycle are shown in order from the top. .
No. The arrangement geometry of No. 1 is a cross-dipole aperture type itself, but such a solution was extracted as a result of a genetic algorithm operation.
No. The arrangement geometries 2 to 6, 10 are a variant of the cross dipole and are characterized by being rotated 45 ° in the plane.
No. Nos. 7 and 8 are cross dipole variants with a load added to the end of the dipole. 9 can be classified as Jerusalem cloth.
From the frequency characteristics shown in FIG. 7, for example, if a solution having too sharp a peak or a solution having a too simple shape is removed, the best solution shown in FIG. 6 and no. The arrangement geometry 7 is a candidate for the design plan.

Thereafter, a final design plan is determined in step S26.
In this step, for example, regarding the arrangement geometry of the candidates extracted in step S24, the frequency characteristic is calculated while changing the arrangement cycle in smaller steps than the analysis performed by the Brute-force method, and the final design plan is determined.
At this time, if necessary, for example, fine adjustment of the material constant may be performed in accordance with the material actually used, or recalculation may be performed by applying an inclination angle in accordance with the installation location of the frequency selection plate.
More specifically, after reflecting the material constant and the inclination angle, 6 and no. With respect to the arrangement geometry of No. 7, the frequency characteristic is calculated in a range of ± predetermined thickness with respect to the thickness at the time of extraction as a good solution, with a finer step size than in the analysis by the Brute-force method.
For example, no. In the case of the arrangement geometry of 6, the arrangement cycle is changed in 0.1 mm steps in the range of 1.29 cm to 1.69 cm, which is the range of the arrangement cycle of 1.49 cm ± 2 mm in the excellent solution, and the frequency characteristic in each arrangement cycle is changed. Is calculated. In addition, No. With the arrangement geometry of 7, the arrangement cycle is changed in increments of 0.1 mm in the range of 1.18 cm to 1.58 cm, which is the range of the arrangement cycle of 1.38 cm ± 2 mm in the excellent solution, and the frequency characteristic in each arrangement cycle is changed. Is calculated.
As a result, the best arrangement period at 9.4 GHz is No. 6 for the placement geometry of 1.51 cm; 7 was 1.33 cm with a geometry of 7. In addition, No. 6 and no. When the frequency characteristics of No. 7 are compared, No. 6 showed sharper filter characteristics, but a remarkable peak was observed in the high frequency band. It was 7.
Therefore, the final design proposal is No. With an arrangement geometry of 7, the arrangement period was determined to be 1.33 cm.

FIG. 8 is a table showing the time required for calculation in the present invention and the prior art.
All values in FIG. 8 use a normal computer 20 (personal computer), and the programming language is the same (Scilab).
In the related art, a calculation amount of about 500 elements × 20 generations is required to obtain an effective design solution using a 16 × 16 cell model using a genetic algorithm. This calculation amount is a calculation amount that takes about 73 hours using a normal personal computer.
On the other hand, in the present invention, if the genetic algorithm operation of the 16 × 16 cell model is performed for 200 elements × 20 generations, almost equivalent design solutions can be obtained. This calculation amount is approximately 2 hours when a normal personal computer is used.
Further, in the present invention, it is necessary to perform the Brute-force analysis of the 8 × 8 cell model before performing the genetic algorithm operation of the 16 × 16 cell model, and the time required for this is approximately 7 hours. In addition, since the Brute-force analysis is independent for each parameter, the calculation can be easily performed by distributing to a plurality of personal computers. By doing so, the calculation time of this step can be further reduced.
Therefore, in the method for designing a frequency selection plate using the present invention, a design solution can be calculated in a total of 9 hours, which is about ８ of the conventional technology, and the calculation time required for designing the frequency selection plate is obtained. Can be greatly reduced.

As described above, the design method of the frequency selection plate according to the embodiment is shorter than performing a Brute-force analysis with a cell model smaller than the M × N cell model to be designed and using a simple initial solution. There is a high possibility that an effective design solution can be obtained in time, which is advantageous in shortening the calculation time in designing the frequency selection plate.
Further, the method of designing a frequency selection plate according to the embodiment can reduce arrangement geometries that are substantially the same when they are arranged consecutively, thereby shortening the calculation time of the Brute-force analysis. Is advantageous.
Further, the method for designing a frequency selection plate according to the embodiment can efficiently obtain a suitable design solution using a genetic algorithm. Further, an effective design solution can be obtained even if the number of elements and the number of generations of the genetic algorithm are reduced, which is advantageous in shortening the calculation time when designing the frequency selection plate.
Further, in the method for designing a frequency selection plate according to the embodiment, when the genetic algorithm is executed, a penalty function based on a target frequency characteristic is used to reduce the probability of occurrence of an arrangement geometry that does not match the frequency characteristic. This is advantageous in increasing the probability of obtaining an effective placement geometry that matches the characteristics.
Further, the method of designing a frequency selection plate according to the embodiment changes the arrangement period as well as the arrangement of the conductor cells and the non-conductor cells on the cell model, so that a frequency selection plate that is more suitable for the target frequency characteristics is designed. The above is advantageous.

  20: Computer, 30, 32: Arrangement geometry

Claims (6)

Each cell region of the M × N cell model in which the frequency selection plate is divided into M × N cells is designated as a conductor cell or a non-conductor cell, and the arrangement of the conductor cells and the non-conductor cells in the M × N cell model is specified. A design method of designing the frequency selection plate by specifying,
A reduced cell model creation step of creating an arrangement geometry group of the conductor cells and the non-conductor cells in an (M / K) × (N / K) cell model where K is an arbitrary natural number;
A good solution extraction step of extracting a good solution having a predetermined frequency characteristic from the arrangement geometry group using a Brute-force method;
The unit cell length of the excellent solution is converted to the M × N cell model by multiplying by 1 / K, the optimization process is performed on the M × N cell model of the excellent solution, and the frequency characteristic is matched. An optimization step to extract the placement geometry;
A method for designing a frequency selection plate, comprising:   Among the arrangement geometries, when the arrangement geometries are sequentially arranged, the arrangement geometry in which the same geometry as the other arrangement geometry appears is reduced with the other arrangement geometry, and is included in the arrangement geometry group. 2. The method for designing a frequency selection plate according to claim 1, further comprising a geometry reduction step for reducing the number of arrangement geometries to be set. In the optimizing step, a genetic algorithm is applied using the M × N cell model of the good solution as a gene to extract the placement geometry that matches the frequency characteristics.
3. The method for designing a frequency selection plate according to claim 1, wherein: In the optimizing step, using a penalty function based on the frequency characteristics to reduce the occurrence probability of the arrangement geometry that does not match the frequency characteristics,
4. The method for designing a frequency selection plate according to claim 1, wherein: The superior solution extraction step includes changing an arrangement cycle, which is the length of one side of the (M / K) × (N / K) cell model, within a predetermined range for each arrangement geometry included in the arrangement geometry group. And calculating the frequency characteristics in each arrangement cycle,
The method for designing a frequency selection plate according to any one of claims 1 to 4, wherein:   A frequency selection plate design program for causing a computer to execute the frequency selection plate design method according to any one of claims 1 to 5.

Images