A method for optimizing a corner reflector based on a regression model

Abstract

The present invention proposes an optimization method for a triangular reflector based on a regression model, which can set different optimization objective functions and objective function values according to actual combat needs, quickly obtain optimization results, and its optimization accuracy also meets the requirements of engineering design. By establishing a regression model between the structural parameters of the triangular reflector array and the RCS characteristics, as the "solver" for the RCS characteristics in the optimization design of the triangular reflector, it solves the problem of huge computational scale existing in optimization based on electromagnetic simulation calculations, and can set different optimization objective functions and objective function values according to actual combat needs, quickly obtain optimization results, and its optimization accuracy also meets the requirements of engineering design.

Description

Technical Field

[0001] The present invention relates to the technical field of signal feature control, and particularly relates to an optimization method for a triangular reflector based on a regression model. Background Art

[0002] The triangular reflector has characteristics such as strong reflected echo, insensitivity to radar polarization, and good applicability in multiple frequency bands, and has been widely used in passive interference and calibration body equipment. The common triangular trihedral corner reflector is composed of three mutually perpendicular triangular flat plates, which can generate a large radar cross section (RCS) within an observation angle of about 40 degrees. In the military field, the forms of using triangular reflector arrays to form false targets and targets are mostly double-layer circular layouts, regular polyhedron layouts, etc. The main existing problems are as follows: First, with the high efficiency and timeliness of reconnaissance satellites and electronic reconnaissance aircraft, the actual combat effectiveness of fixed false targets has shown a "cliff-like" decline; second, the RCS fluctuation range of the triangular reflector array is relatively large, resulting in weakened interference effectiveness and positioning errors of weapon systems; third, the common triangular reflector array false targets have disadvantages such as long erection time and high personnel requirements, resulting in the inability to meet the emergency combat requirements.

[0003] In order to improve the effectiveness of the triangular reflector as a military false target, many improved designs of triangular reflectors have emerged. For example, the vertical plane of the triangular reflector is decomposed into multiple blades, and the RCS characteristics are changed by rotating the blade angles; the RCS characteristics are improved by determining a semi-elliptical depression structure, trapezoidal cutting, etc. on the reflecting surface of the triangular reflector; the RCS characteristics are improved by coating an absorbing coating on the surface of the triangular reflector. The research on the arraying method of triangular reflector arrays mostly focuses on the fields of satellite and airborne laser ranging, and there is little research on triangular reflector array false targets. In the prior art, most are based on the simulation calculation of the triangular reflector with a changed structure to obtain the corresponding RCS characteristics, and then the structure change parameters of the triangular reflector are selected. The following problems exist:

[0004] In the common triangular reflector design, the method of variable interval sampling test is mostly used for optimization, and the persuasiveness of the conclusion is not high. The contradiction between the simulation calculation amount problem in the triangular reflector array design and the computing resources and design time efficiency ratio is difficult to solve. Summary of the Invention

[0005] In view of this, the present invention proposes an optimization method for a triangular reflector based on a regression model, which can set different optimization objective functions and objective function values according to actual combat needs, quickly obtain the optimization results, and the optimization accuracy also meets the engineering design requirements.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] An optimization method for a triangular reflector based on a regression model of the present invention includes the following steps:

[0008] Step 1: Select the target RCS solution method;

[0009] Step 2: Determine the structural parameters of the triangular reflector array; according to the geometric structure and array pattern of the triangular reflector, determine the optimization variables and their ranges;

[0010] Step 3: Based on the improved grille sequence method, perform data sampling on the optimization variables determined in Step 2 in a space-filling manner to obtain sampling data;

[0011] Step 4: Use the sampling data in Step 3 to establish a triangular reflector array, adopt the target RCS solution method determined in Step 1, and obtain the basic data of the RCS characteristics of the triangular reflector array through electromagnetic simulation calculation;

[0012] Step 5: Use the sampling data in Step 3 and the basic RCS characteristic data in Step 4 as the sample data for neural network learning, and establish a regression model of the structural parameters and the target RCS characteristics; specifically, set up a neural network learning environment, combine the first set of variable sampling points in Step 3 and the basic RCS characteristic data of the triangular reflector in Step 4 as the sample data for neural network learning, and perform neural network training based on the Levenberg-Marquardt backpropagation algorithm to generate a regression model of the variables and the RCS characteristics;

[0013] Step 6: Set the optimization objective function according to the user's requirements, select the optimization algorithm according to the number, nature, and mutual relationship of the objective functions, and use the regression model in Step 5 for intelligent optimization to obtain the optimization result.

[0014] Among them, in the above Step 1, according to the geometric structure of the triangular reflector array and the operating frequency of the combat object, considering the solution accuracy and optimization time ratio, select the RCS solution method for the triangular reflector array.

[0015] Among them, in the above Step 2, first determine the geometric structure parameters of the triangular reflector, perform an isosceles trapezoid cutting operation on the reflecting surface of the triangular reflector, where the lower base of the trapezoid and the midpoint of the lower base coincide with the inclined edge of the triangular reflector and the midpoint of the inclined edge respectively, and take the upper end point of the trapezoid on the connecting line between the lower end point of the trapezoid and the intersection point of the straight edge of the triangular reflector. The length of the lower base of the trapezoid and the height of the trapezoid are set as optimization variables;

[0016] Secondly, determine the structural parameters of the triangular reflector array elements. Move the triangular reflector along the positive Y-axis direction and then rotate it along the inclined edge parallel to the XOZ plane. The moving distance and the rotation angle are set as optimization variables;

[0017] In the above Step 3, the sampling data includes generating 2 groups of variable sampling points respectively: the first group is the cutting trapezoid bottom length and the trapezoid height, and the second group is the moving distance and the rotation angle of the triangular reflector array element.

[0018] Among them, in the fourth step, the data of the first group of variable sampling points obtained in the third step is used as the modeling data of the triangular reflector. The FEKO automatic modeling and RCS solution are realized by using the LUA script, and the basic RCS characteristic data of the triangular reflector is obtained.

[0019] Among them, in the sixth step, it also includes setting the optimization objective function according to the optimization design index of the triangular reflector, selecting the multi-objective optimization method based on the genetic algorithm for variable optimization, and obtaining the Pareto optimal solution set of the triangular reflector array elements.

[0020] Among them, the user can select the geometric structure parameters of the triangular reflector as needed, and select the most suitable variable parameters in the optimization results according to the actual application requirements to establish a triangular reflector array. Among them, the optimized geometric structure parameters of the triangular reflector and the data of the second group of variable sampling points obtained in the third step are used as the modeling data of the triangular reflector array elements, and steps four to six are repeated to obtain the Pareto optimal solution set of the triangular reflector array elements.

[0021] Beneficial effects:

[0022] 1. The optimization method of the triangular reflector based on the regression model of the present invention solves the problem of huge calculation scale in optimization based on electromagnetic simulation calculation by establishing a regression model between the structural parameters of the triangular reflector array and the RCS characteristics, which is used as the "solver" of the RCS characteristics in the optimization design of the triangular reflector. Different optimization objective functions and objective function values can be set according to actual combat needs, and the optimization results can be obtained quickly, and the optimization accuracy also meets the requirements of engineering design.

[0023] 2. Based on the actual combat requirements of military applications, the present invention first analyzes the user's requirements, determines the optimization variables and conducts spatial sampling, automatically generates the basic RCS database of the triangular reflector, and uses the backpropagation algorithm to establish a regression model between the optimization variables and the target RCS characteristics, and the coefficient of determination is close to 1. In the intelligent optimization process, this model is used to solve the RCS characteristics of the triangular reflector, which solves the problem of huge calculation amount in the optimization design of simulation calculation and provides a more scientific and effective method for the optimization design of the triangular reflector.

[0024] 3. The present invention combines the dual advantages of intelligent optimization algorithms and neural network learning in the optimized design of triangular reflectors. Based on the simulation calculation of the RCS of triangular reflectors a limited number of times, a basic database of the RCS characteristics of triangular reflectors is established. A regression model between the optimization variables and the RCS characteristics of triangular reflectors is constructed through neural network learning. This regression model is used as the "solver" for the RCS characteristics of triangular reflectors in the optimized design, solving the problem of huge computational scale in the optimization of triangular reflectors based on simulation calculations, breaking through the common data mining optimization mode in the simulation optimization design of triangular reflectors, and providing a more scientific and effective method for the optimized design of triangular reflectors. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is the optimization flowchart of the triangular reflector array of the present invention.

[0026] Figure 2 It is the population evolution flowchart of the gamultiobj function of the present invention.

[0027] Figure 3 It is the schematic diagram of the geometric structure optimization of the triangular reflector of the present invention.

[0028] Figure 4 It is the performance parameter diagram of the RCS characteristic regression model of the triangular reflector of the present invention.

[0029] Figure 5 It is the Pareto front result diagram of the geometric structure optimization of the triangular reflector of the present invention.

[0030] Figure 6 It is the schematic diagram of the element structure optimization of the triangular reflector of the present invention.

[0031] Figure 7 It is the performance parameter diagram of the RCS characteristic regression model of the triangular reflector element of the present invention.

[0032] Figure 8 It is the Pareto front result diagram of the element structure parameter optimization of the triangular reflector of the present invention.

[0033] Figure 9 It is the schematic diagram of a fast deployment structure of the triangular reflector array of the present invention.

[0034] Figure 10 It is the RCS plan view of the fast deployment triangular reflector array of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] The present invention will be described in detail below with reference to the accompanying drawings and by way of examples.

[0036] The present invention proposes an optimization method for triangular reflectors based on a regression model, and the optimization process of the triangular reflector array is asFigure 1 As shown, when specifically optimizing, different functions can be selected. Figure 2 The flowchart of the population evolution of the gamultiobj function is shown.

[0037] The present invention specifically includes the following steps:

[0038] Step 1: Select the target RCS solution method. According to the geometric structure of the triangular reflector array and the operating frequency of the combat object, considering the solution accuracy and optimization time-effect ratio, select the RCS solution method for the triangular reflector array. In this embodiment, when the triangular reflector array is used as a military decoy and a target, it belongs to electrically large-sized and strong scattering targets. Considering the requirements for the solution accuracy of the target RCS and the calculation time of the basic solution data in engineering design, the RL-GO high-frequency approximation algorithm is adopted for simulation calculation.

[0039] Step 2: Determine the structural parameters of the triangular reflector array. According to the geometric structure and array mode of the triangular reflector, determine the variables to be optimized and the variable ranges.

[0040] Specifically, first determine the geometric structure parameters of the triangular reflector. As Figure 3 shown, perform an isosceles trapezoidal cutting operation on the reflecting surface of the triangular reflector. The lower base of the trapezoid and the midpoint of the lower base coincide with the inclined edge and the midpoint of the inclined edge of the triangular reflector respectively. Take the upper end point of the trapezoid on the connecting line between the lower end point of the trapezoid and the intersection point of the straight edge of the triangular reflector. The length of the lower base of the trapezoid and the height of the trapezoid are set as optimization variables.

[0041] Secondly, determine the structural parameters of the triangular reflector array elements. As Figure 6 shown, move the triangular reflector along the positive Y-axis direction and then rotate it along the inclined edge parallel to the XOZ plane. The moving distance and the rotation angle are set as optimization variables.

[0042] Step 3: Variable sampling. Based on the space-filling method, perform data sampling on the optimization variables determined in Step 2. Specifically, based on the improved grid sequence method, perform variable data sampling in a space-filling manner, and generate 2 groups of variable sampling points respectively; the first group is the length of the bottom edge of the cutting trapezoid and the height of the trapezoid, and the second group is the moving distance and the rotation angle of the triangular reflector array element.

[0043] Step 4: Solve the target characteristics. Use the sampling point data in Step 3 to establish a triangular reflector array, and adopt the target RCS solution method determined in Step 1 to obtain the basic RCS characteristic data of the triangular reflector array through electromagnetic simulation calculation. The specific method is to use the data of the first group of variable sampling points obtained in Step 3 as the modeling data of the triangular reflector, and use the LUA script to realize FEKO automatic modeling and solve the RCS to obtain the basic RCS characteristic data of the triangular reflector.

[0044] Step 5: Establish a regression model. Use the sampling data from Step 3 and the basic RCS characteristic data from Step 4 as the sample data for neural network learning to establish a regression model between the structural parameters and the target RCS characteristics. The specific method is to build a neural network learning environment, combine the first set of variable sampling points from Step 3 and the basic RCS characteristic data of the triangular reflector from Step 4 as the sample data for neural network learning, and perform neural network training based on the Levenberg-Marquardt backpropagation algorithm to generate a regression model between the variables and the RCS characteristics.

[0045] In Steps 4 and 5 of the present invention, the dual advantages of intelligent optimization algorithms and neural network learning are combined in the optimized design of the triangular reflector. Based on the RCS simulation calculations of the triangular reflector for a limited number of times, a basic database of the RCS characteristics of the triangular reflector is established. A regression model between the optimization variables and the RCS characteristics of the triangular reflector is constructed through neural network learning. This regression model is used as the "solver" for the RCS characteristics in the optimized design, solving the problem of the huge computational scale in the optimization of the triangular reflector based on simulation calculations, breaking through the common data mining and optimization mode in the simulation-based optimized design of the triangular reflector, and providing a more scientific and effective method for the optimized design of the triangular reflector.

[0046] Step 6: Optimize the structural parameters. Set the optimization objective function according to the user's requirements, select an optimization algorithm based on the number, nature, and mutual relationship of the objective functions, and use the regression model from Step 5 as the RCS characteristic "solver" to perform intelligent optimization and obtain the optimization results.

[0047] This step may further include the following steps: Set the optimization objective function according to the optimized design indicators of the triangular reflector, select a multi-objective optimization method based on the genetic algorithm for variable optimization, and obtain a set of Pareto optimal solutions.

[0048] Specifically, the user independently selects the geometric structure parameters of the triangular reflector required, and according to the actual application requirements, selects the most suitable variable parameters from the optimization results to establish a triangular reflector array; among them, using the optimized geometric structure parameters of the triangular reflector and the second set of variable sampling point data obtained in Step 3 as the modeling data for the triangular reflector array elements, repeating Steps 4 to 6, the Pareto optimal solutions of the triangular reflector array elements can be obtained.

[0049] Taking the design and optimization of a quickly deployable triangular reflector decoy as an example, the optimization process of the present invention is described, including the following steps: First, optimize the geometric structure of the triangular reflector, second, optimize the array element structure of the triangular reflector, and finally, design a quickly deployable triangular reflector decoy using the optimization results.

[0050] Among them, the specific process of optimizing the geometric structure of the triangular reflector is asFigure 3 As shown, the incident range of the electromagnetic wave has θ equal to 55 degrees, equal to 45 degrees as the center, and the coverage range is 50 degrees. The length range of the lower base of the trapezoid is set from 0.1 m to 0.8 m, and the height range of the trapezoid is set from 0.01 m to 0.2 m. 30 groups of data are sampled using the improved grid sequence method and the RCS of the triangular reflector is simulated and calculated. The RCS characteristics of the triangular reflector select the average value and the dynamic range. The hidden layer of neural network learning is set to 10 layers. Based on the Levenberg Marquardt backpropagation algorithm for training, verification and testing, the determination coefficient R of the RCS characteristic regression model is equal to 0.99955. The specific performance parameters are as Figure 4 shown, and the optimized Pareto optimal solution set is as Figure 5 shown. The optimization index is set as: the average value of RCS is greater than 26 dBsm, and the dynamic range is less than 5 dB. The 6th group of parameters in the Pareto optimal solution set is the closest to the optimization index. The corresponding trapezoid lower base length of the Pareto optimal solution is 0.2748 m, and the trapezoid height is 0.0976 m. Verified by FEKO simulation, the difference in the average value of RCS is 0.16 dB, and the difference in the dynamic range is 0.007 dB.

[0051] The specific process of optimizing the triangular reflector array element structure is as follows: as Figure 6 shown, the incident range of the electromagnetic wave has θ equal to 45 degrees, equal to 0 degrees as the center, and the coverage ranges are 50 degrees and 60 degrees respectively. The rotation angle range is set from 5 to 15 degrees, and the moving distance range is set from 0.05 m to 0.15 m. 50 groups of data are sampled using the improved grid sequence method and the RCS of the triangular reflector array element is simulated and calculated. The RCS characteristics of the triangular reflector array element select the average value and the dynamic range. The hidden layer of neural network learning is set to 10 layers. Based on the Levenberg Marquardt backpropagation algorithm for training, verification and testing, the determination coefficient R of the RCS characteristic regression model is equal to 0.99304. The specific performance parameters are as Figure 7 shown, and the optimized Pareto optimal solution set is as Figure 8 shown. The optimization index is set as: the average value of RCS is greater than 26 dBsm, and the dynamic range is less than 20 dB. The 3rd group of parameters in the Pareto optimal solution set is the closest to the optimization index. The corresponding rotation angle of the Pareto optimal solution is 9.9813 degrees, and the moving distance is 0.0756 m. Verified by FEKO simulation, the difference in the average value of RCS is 0.04 dB, and the difference in the dynamic range is 4 dB (the reason is that the RCS dynamic range value of the triangular reflector array element changes greatly, and the standard deviation is close to 5 dB. It can be improved by increasing the sampling points to expand the basic database).

[0052] The specific process of designing a quickly deployable corner reflector false target using the optimization results is as follows: Using the optimized parameters of the corner reflector's geometric structure, the length of the lower base of the trapezoid is 0.2748 m, and the height of the trapezoid is 0.0976 m. For the structural parameters of the corner reflector array element, the rotation angle is 9.9813 degrees, and the moving distance is 0.0756 m. Design a quickly deployable corner reflector array as Figure 9 shown. Through FEKO simulation calculation, its RCS planar distribution is as Figure 10 shown. When the azimuth angle and elevation angle of the incident electromagnetic wave are both from 0 degrees to 90 degrees, compared with a conventional single corner reflector, the probability that the RCS value of the quickly deployable array falls within the range of 26.48 dBsm to 29.48 dBsm (the 3 dB range of the optimization target) increases by 72.5%. Compared with the optimized single corner reflector, the improvement amplitude still reaches 25.3%. This group of arrays has the characteristics of simple structure, fast erection speed, and a significant increase in the probability that the RCS value in the main combat direction falls within the design index. It is very suitable for multi-group simultaneous deployment around key protected targets in emergency operations in groups of 3 to 4 people, which can cause effective all-round interference to precision-guided weapons.

[0053] The corner reflector array optimization method of the present invention is applicable to the design of corner reflector targets and passive false targets, and can serve the performance testing of weapon equipment, and is used for the optimization design and actual combat quick deployment of military false targets, targets, etc. Based on the actual combat requirements of military applications, the present invention first analyzes user requirements, determines optimization variables and conducts spatial sampling, automatically generates a basic database of corner reflector RCS, uses the backpropagation algorithm to establish a regression model between the optimization variables and the target RCS characteristics, and the coefficient of determination is close to 1. In the intelligent optimization process, this model is used to solve the corner reflector RCS characteristics, solving the problem of extremely large computational workload in simulation calculation optimization design, and providing a more scientific and effective method for the optimization design of corner reflectors.

[0054] In summary, the above is only a preferred embodiment of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing a triangular reflector based on a regression model, characterized in that, it includes the following steps: Step 1: Select a target RCS solution method; Step 2: Determine the structural parameters of the triangular reflector array; according to the geometric structure and array mode of the triangular reflector, determine the optimization variables and variable ranges; Step 3: Based on the improved grille sequence method, perform data sampling on the optimization variables determined in Step 2 in a space-filling manner to obtain sampling data; Step 4: Use the sampling data in Step 3 to establish a triangular reflector array, and adopt the target RCS solution method determined in Step 1 to obtain the basic RCS characteristic data of the triangular reflector array through electromagnetic simulation calculation; Step 5: Use the sampling data in Step 3 and the basic RCS characteristic data in Step 4 as the sample data for neural network learning, and establish a regression model of the structural parameters and the target RCS characteristics; the specific method is to build a neural network learning environment, merge the first set of variable sampling points in Step 3 and the basic RCS characteristic data of the triangular reflector in Step 4 as the sample data for neural network learning, and perform neural network training based on the Levenberg-Marquardt backpropagation algorithm to generate a regression model of the variables and RCS characteristics; Step 6: Set the optimization objective function according to user requirements, select an optimization algorithm according to the number, nature and mutual relationship of the objective functions, and use the regression model in Step 5 for intelligent optimization to obtain the optimization result.

2. The method according to claim 1, characterized in that, in Step 1, according to the geometric structure of the triangular reflector array and the operating frequency of the combat object, considering the solution accuracy and optimization time ratio, select the RCS solution method for the triangular reflector array.

3. The method according to claim 1, characterized in that, in Step 2, first determine the geometric structure parameters of the triangular reflector, perform an isosceles trapezoid cutting operation on the reflecting surface of the triangular reflector, the lower base and the midpoint of the lower base of the trapezoid coincide with the inclined edge and the midpoint of the inclined edge of the triangular reflector respectively, take the upper end point of the trapezoid on the connecting line between the lower end point of the trapezoid and the intersection point of the straight edge of the triangular reflector, and set the length of the lower base of the trapezoid and the height of the trapezoid as optimization variables; Secondly, determine the structural parameters of the triangular reflector array elements, move the triangular reflector along the positive Y-axis direction, and then rotate it along the inclined edge parallel to the XOZ plane, and set the moving distance and rotation angle as optimization variables; in Step 3, the sampling data includes generating 2 groups of variable sampling points respectively: the first group is the cutting trapezoid bottom length and trapezoid height, and the second group is the moving distance and rotation angle of the triangular reflector array element.

4. The method according to claim 1, characterized in that, in Step 4, use the first group of variable sampling point data obtained in Step 3 as the modeling data of the triangular reflector, and use the LUA script to realize FEKO automatic modeling and solve the RCS to obtain the basic RCS characteristic data of the triangular reflector.

5. The method according to any one of claims 1-4, characterized in that, In the sixth step, it further includes setting an optimization objective function according to the optimized design index of the triangular reflector, selecting a multi-objective optimization method based on the genetic algorithm for variable optimization, and obtaining the Pareto optimal solution set of the triangular reflector array elements.

6. The method according to claim 5, wherein, the user independently selects the required geometric structure parameters of the triangular reflector, selects the most suitable variable parameters in the optimization results according to the actual application requirements, and establishes a triangular reflector array; wherein the optimized geometric structure parameters of the triangular reflector and the data of the second set of variable sampling points obtained in the third step are used as the modeling data of the triangular reflector array elements, and steps four to six are repeated to obtain the Pareto optimal solution set of the triangular reflector array elements.

Images