The present technology relates to a frequency selective surface (FSS) filter design method and a storage medium for storing computer software for performing the FSS filter design method.
A frequency used for wireless communication varies depending on a wireless communication provider. Base stations require a filter to separate a wireless communication frequency band used by each wireless communication provider and avoid interference which may result from various causes. The filter functions to separate frequency bands from each other in this way.
A frequency selective surface (FSS) is a curved or flat three-dimensional (3D) surface having an artificially manufactured thickness to selectively transmit or block a frequency wave wanted by a user. Such a frequency selective property of an FSS can be obtained by arranging conductors or apertures as pixels.
The frequency response characteristics of an FSS filter vary depending on not only the geometric shape of a structure selected as unit cells but also the shape of a pixel arrangement in the unit cells and material properties of a dielectric and a conductor used as a substrate supporting the unit cells. Accordingly, various methods of obtaining frequency characteristics wanted by a user have been researched and proposed.
A genetic algorithm is an algorithm fundamentally based on the theory of biogenetics in nature, and is based on Darwin's theory of survival of the fittest. A genetic algorithm expresses possible solutions to a problem to be solved in a determined form of data structure and then gradually modifying the solutions, thereby creating better solutions. Here, the data structure representing solutions may be expressed genes, and a process of creating better solutions by modifying the genes may be expressed as evolution.
Such a genetic algorithm may include crossovers and mutations. In a crossover operation, generally, a plurality of solutions are selected, and then a crossing operation is performed between the plurality of solutions. As solutions generated in this way, new genes are constructed by receiving genetic factors from positions that do not overlap each other through a crossover operation of parent solutions. A mutation operation is an operation in which the order or values of genetic factors in a given solution are arbitrarily changed and transformed into another solution.
Conventionally, frequency selective surface (FSS) filter unit cells are completed by changing a known arrangement of unit cells. Frequency characteristics in accordance with such a change in the pixel arrangement are examined to design an FSS filter. Designing an FSS filter to have a targeted frequency response often requires a high degree of expertise. Accordingly, it takes a long time to design a filter having a desired frequency response by repeating a process of adjusting the arrangement of unit cells one by one, finding frequency characteristics, and then changing the arrangement of unit cells again for the desired frequency response. Also, filter design is so difficult that it is practically impossible to implement perfect performance. Although it is possible to propose various frequency response characteristics in theory, it is practically difficult to implement the frequency response characteristics due to the combinatorial possibilities of countless arrays that are not listable.
The present invention is directed to providing a method of designing an FSS filter to have a targeted frequency response characteristic using an efficient global optimization algorithm.
One aspect of the present invention provides a method of designing a frequency selective surface (FSS) filter, the method including calculating a candidate solution corresponding to a structure of the FSS filter and an objective-function value corresponding to a difference between a frequency response resulting from the candidate solution and a targeted frequency response, modifying the candidate solution into a trial solution in accordance with a genetic algorithm, and calculating an objective-function value with the trial solution to determine whether to effectively include the trial solution in candidate solutions.
Another aspect of the present invention provides a computer program including calculating a candidate solution corresponding to a structure of an FSS filter and an objective-function value corresponding to a difference between a frequency response resulting from the candidate solution and a targeted frequency response, modifying the candidate solution into a trial solution in accordance with a genetic algorithm, and calculating an objective-function value with the trial solution to determine whether to effectively include the trial solution in candidate solutions. The present embodiment completes the design of an FSS filter and outputs patterns of all calculated frequency selection filters and frequency response characteristics each corresponding to the frequency selection filters.
According to the present embodiment, it is possible to design a frequency selective surface (FSS) filter that has a targeted general frequency response by performing generation of combinatorial patterns, which is practically almost impossible according to the conventional art, through computation in a very efficient way.
Hereinafter, a method of designing a frequency selective surface (FSS) filter according to the present embodiment will be described with reference to the accompanying drawings. In the following description, the shape of pixels is described as a square. However, a very large number of pixels are used in a design, and thus it is possible to express a pattern having a general response characteristic.
A frequency selective property of the FSS filter 10 may vary depending on the shape of unit cells 120 filled with the metal. In the example shown in the drawings, 20 square pixels 120 may be disposed horizontally, and 20 square pixels 120 may be disposed vertically so that the FSS filter 10 is formed. However, this is only an example, and the number and shape of pixels 120 and a shape formed by the pixels 120 may vary.
An edge 130 of the FSS filter 10 may be filled with a metal. The FSS filter 10 is assumed to have finite periodicity. Accordingly, the FSS filter 10 may be filled with a metal and covered with the edge 130 to obtain finite periodicity required by the FSS filter 10. In the example shown in
An irreducible zone 110 is extracted from the surface of the FSS filter 10. The irreducible zone 110 may be a unit that may cover the entire surface of the FSS filter unit cells 10 using symmetry. An example of the irreducible zone 110 is shown in
As described above, a state of pixels in the irreducible zone 110 may be referred to as a one-dimensional (1D) sequence, and any one pixel 120 in the irreducible zone 110 may be expressed as one digit in the sequence. Such expression make it possible to process crossover and mutation operations simply.
It is required for an FSS filter to have rotational symmetry. This is because electromagnetic waves may be incident on the installed filter at various angles. To ensure the same filter performance for various angles of incidence, it is necessary to design unit cells with rotational symmetry. An irreducible zone is introduced in consideration of such a physical condition. Accordingly, when an irreducible zone is determined through a combinatorial optimization process, a filter shape having rotational symmetry may be determined. The irreducible zone expressed in the form of a 1D arrangement as described above is expanded to the entire surface of the FSS filter 10 using symmetry required by FSS filters.
Referring to
Multiple candidate solutions are prepared (S100). Random numbers may be assigned to digits included in a sequence expressing a candidate solution. As described above, the candidate solutions correspond to the irreducible zone 110, and each digit included in the candidate solutions may correspond to a pixel included in the irreducible zone 110.
A value of 0 or 1 may be assigned to each digit included in the candidate solutions in accordance with a sequence number. As an example, “0” may correspond to a pixel 120a not filled with the metal film in the irreducible zone 110, and “1” may correspond to a pixel 120b filled with the metal film in the irreducible zone 110. Also, multiple (e.g., 20) candidate solutions are prepared. A set of multiple candidate solutions is referred to as a candidate solution group.
According to an embodiment, local optimization is performed on the prepared candidate solutions.
The position and the number of digits on which local optimization is performed may be determined by a random number. Accordingly, the number of digits subjected to local optimization in any one candidate solution may differ from the number of digits subjected to local optimization in another candidate solution, and positions at which local optimization is performed may also be different.
In the embodiment illustrated in
Referring back to
When the distance between a first candidate solution and a second candidate solution is 10, 10 digits are different between the first and second candidate solutions. The distance between candidate solutions may also be the difference (similarity) between the candidate solutions.
According to an embodiment, a cutoff distance is set. For example, the cutoff distance may be set to half the average of the calculated distances between the candidate solutions. However, the set cutoff distance may be adjusted in a subsequent process (S900). For example, the cutoff distance may be adjusted to decrease by a factor of 0.97 for each attempt of crossover or mutation.
Objective-function values of the candidate solutions are calculated (S300).
In
According to an embodiment, the objective function is a numerical function designated and fixed by a user for a frequency response characteristic wanted by the user. Accordingly, a frequency response function corresponding to any objective filter, such as a bandpass filter, a band-stop filter, a multiband filter, etc., may be set as the objective function. A general objective function may be defined as a function of frequency.
In the embodiment illustrated in
Subsequently, the candidate solutions are genetically transformed using a genetic algorithm (S400). According to an embodiment, the candidate solutions may be genetically transformed by performing crossover on any two or more candidate solutions to generate a crossover trial solution or performing mutation on any one or more candidate solutions to generate a mutation trial solution (S500b).
According to an embodiment, whether to perform crossover or mutation on the candidate solutions may be determined by a random number. For example, a random number generator (not shown) may output a value between 0 and 1, and any one of crossover and mutation may be selected depending on whether the output value is greater or smaller than a threshold value of 0.5.
The operation S500a of generating a crossover trial solution will be described with reference to
As an embodiment of selecting a candidate solution, a candidate solution of which a frequency response is more similar to a target frequency response is more likely to be selected from among multiple candidate solutions. When an objective-function value is lower, a crossover or mutation operation is more likely to be performed on a corresponding candidate solution.
One of the single tournament method or the Poisson's distribution method is selected to select a candidate solution. In the single tournament method, two randomly selected different candidate solutions are selected first, and then a candidate solution having a smaller objective-function value is finally selected. In the Poisson's distribution method, when candidate solutions are sorted in increasing order of the objective-function value, a Poisson distribution function is created using an average rank and a rank deviation to finally select a candidate solution. A selected rank is probabilistically in accordance with the Poisson's distribution. The best solution has the smallest objective-function value, and the corresponding rank is the first. The positions and the number of digits on which crossover occurs are both randomly determined by the random number generator (not shown).
The operation S500b of generating a mutation trial solution will be described with reference to
The objective-function value of the trial solution is calculated (S600). As the objective-function value is calculated, similarity is determined between the frequency response characteristic of the FSS filter 10 provided by the trial solution and a target frequency response characteristic.
Whether to discard or use the trial solution is determined (S700). According to an embodiment, a process of calculating the distances between the trial solution and the candidate solutions is performed. A closest candidate solution which is closest (most similar) to the trial solution is determined from the distance calculation results between the trial solution and the candidate solutions.
When the distance between the trial solution and the closest candidate solution is shorter than the current cutoff distance value, an objective-function value of the trial solution is compared with an objective-function value of the closest candidate solution. When the objective-function value of the trial solution is larger than the objective-function value of the closest candidate solution (i.e., when the frequency response characteristic of the closest candidate solution is more similar to the target frequency response characteristic than the frequency response characteristic of the trial solution), the trial solution is discarded.
On the other hand, when the objective-function value of the trial solution is smaller than the objective-function value of the closest candidate solution (i.e., when the frequency response characteristic of the trial solution is more similar to the target frequency response characteristic than the frequency response characteristic of the closest candidate solution), the trial solution replaces the closest candidate solution, and the existing closest candidate solution is discarded.
When the distance between the trial solution and the closest candidate solution is larger than the current cutoff distance value, the objective-function value of the trial solution is compared with an objective-function value of a candidate solution having the largest objective-function value among the existing candidate solutions (i.e., a candidate solution having a frequency response characteristic that is most dissimilar to a desired frequency response characteristic among the candidate solutions).
When the objective-function value of the trial solution is smaller than the largest objective-function value of the candidate solutions (i.e., when the frequency response characteristic of the trial solution is more similar to the target frequency response characteristic than the frequency response characteristic of the compared candidate solution), the trial solution replaces the compared candidate solution, and the compared candidate solution is discarded. On the other hand, when the objective-function value of the trial solution is larger than the largest objective-function value of the candidate solutions (i.e., when the frequency response characteristic of the compared candidate solution is more similar to the target frequency response characteristic than the frequency response characteristic of the trial solution), the trial solution is discarded.
Trial solutions are included in a group of the existing candidate solutions through this process, and among the existing candidate solutions, candidate solutions resulting in a frequency response characteristic dissimilar to the target frequency response characteristic are discarded. Accordingly, the frequency response characteristic of the FSS filter 10 formed by the candidate solutions belonging to the candidate solution group gradually approaches the target frequency response characteristic.
Global optimization is performed to reduce the cutoff distance value (S800). The cutoff distance value is a criterion for determining whether to replace a candidate solution with the trial solution. In general, different forms of trial solutions may be replaced with the candidate solution group, and thus the candidate solution group may ensure diversity. Such a replacement method ensures the diversity of candidate solutions and is a computation that is not found in existing genetic algorithms.
It is determined whether to continue the above process (S900). According to an embodiment, whether to continue the above process may be determined in accordance with a change of the objective-function value. When the objective-function value is determined not to be reduced any more and thus the frequency response characteristic sufficiently approaches the frequency response characteristic wanted by the user, a plurality of unit FSS filters 10 designed as described above may be arranged in an array to constitute an FSS filter.
The device 400 on which the software for the method of designing an FSS filter is installed includes an input device 410, a computation device 420, and a storage device 430. In addition, the device 400 on which the software for the method of designing an FSS filter is installed may include an output device 440.
The input device 410 receives target frequency response data. The input device 410 may be a communication device or an interface device that receives measurement data from a network. Also, the input device 410 may be an interface device that receives measurement data through a wired network. Meanwhile, the input device 410 may receive an external control signal. For example, target frequency response data may be input by a user through the input device.
The storage device 430 may store a software model for the method of designing an FSS filter. The storage device 430 may be implemented as one of various media, such as a semiconductor storage device, a hard disk, etc., for storing data. The storage device 430 may store the software for the method of designing an FSS filter, various information and parameters used in a computation process, and the computation results.
The computation device 440 runs the software for the method of designing an FSS filter using the provided measurement data. Also, the computation device 440 may compute a frequency response of the FSS filter 10 on the basis of the computation results and derive a result value by inputting the provided target frequency response data to the software for the method of designing an FSS filter.
The computation device 440 corresponds to a device that processes data by running a certain instruction or program. The computation device 440 may be implemented as a memory (buffer) for temporarily storing an instruction or information and a processor for performing a computation process. The processor may be implemented as a central processing unit (CPU), an application processor (AP), a field programmable gate array (FPGA), etc. in accordance with the type of device.
The output device 440 may be a communication device that externally transmits necessary data. The output device 440 may externally transmit the result value derived by the trained software for the method of designing an FSS filter. In some cases, the output device may be a device that outputs a training process of the software for the method of designing an FSS filter or the result value derived by the trained software for the method of designing an FSS filter through a screen.
Also, the above-described method of designing an FSS filter may be implemented as a program (or an application) including a computer-executable algorithm. The program may be stored and provided in a non-transitory computer-readable medium.
The non-transitory computer-readable medium is not a medium that stores data for a short time period, such as a register, a cache, a memory, etc., but a medium that stores data semi-permanently and is readable by a device. Specifically, the above-described various applications or programs can be stored and provided in a non-transitory computer-readable medium such as a compact disc (CD), a digital versatile disc (DVD), a hard disk, a Blu-ray disc, a Universal Serial Bus (USB) device, a memory card, a read-only memory (ROM), etc.
A simulation example will be described below with reference to the accompanying drawings. A computer program for performing the method of designing an FSS filter according to the present embodiment was written in the language Python. Frequency response characteristics of candidate solutions are calculated by a high-frequency electromagnetic solver (HFSS) which is an electromagnetic numerical analysis program. An optimization method computer program and the HFSS are merged in the computer program language Iron Python.
An FSS filter was designed to have a pixel size of 0.1 mm2 and an overall size of 5.4 mm2 including 54×54 unit cells using the method of designing an FSS filter according to the present embodiment.
The present embodiment is useful not only in designing single-band FSS filters as illustrated in
To facilitate understanding of the present invention, the present invention has been described with reference to embodiments shown in the drawings. However, these are embodiments for implementation and only exemplary. Those of ordinary skill in the art should understand that various modifications and equivalents can be made from the embodiments. Therefore, the true technical scope of the present invention should be determined by the appended claims.
Number | Date | Country | Kind |
---|---|---|---|
10-2020-0056959 | May 2020 | KR | national |
This application is a National Stage Patent Application of PCT International Patent Application No. PCT/KR2021/005895 (filed on May 11, 2021) under 35 U.S.C. § 371, which claims priority to Korean Patent Application No. 10-2020-0056959 (filed on May 13, 2020), which are all hereby incorporated by reference in their entirety.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/KR2021/005895 | 5/11/2021 | WO |