A method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network

Abstract

The present invention discloses a method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network, aiming at the problem of low efficiency caused by multiple modeling and simulation required by traditional calculation methods when solving deformed targets. The method includes the following steps: generating a data set based on the Shooting and Bouncing Rays (SBR) method; constructing an artificial neural network model; learning the data set and training the artificial neural network model; using the trained artificial neural network model to predict the monostatic RCS of the deformable S-shaped cavity target. The present invention applies the trained artificial neural network model to the rapid calculation of the monostatic RCS of the deformable S-shaped cavity, thereby avoiding the modeling and solution of the traditional SBR method, greatly reducing the calculation time, and improving the calculation efficiency of the monostatic RCS of the deformable S-shaped cavity.

Description

Technical Field

[0001] The present invention relates to the field of electromagnetic computing technology, and particularly to a method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network. Background Art

[0002] RCS (Radar Cross Section) is a measure of the scattering ability of a target. In a real battlefield, the radar stealth technology of military targets (such as aircraft, ships, etc.) directly affects their combat defense and assault capabilities. For large-sized targets such as aircraft and ships, the radar stealth performance is mainly achieved by various methods to reduce the RCS, that is, to reduce the effective reflection power of the radar detection wave by itself, so as to achieve the purpose of reducing the detection range of the enemy radar.

[0003] In the field of electromagnetic scattering calculation, the acquisition of the RCS of complex targets is through theoretical analysis, formula calculation and actual measurement. However, such methods have high calculation complexity, long time period and poor timeliness, and cannot meet the requirements of simply, efficiently and accurately estimating and predicting the RCS of radar targets under real battlefields. In recent years, neural networks have been widely used in image recognition, speech recognition, recommendation systems and other aspects. By constructing a multi-layer neural network model, the characteristic information of data can be accurately and efficiently extracted.

[0004] In summary, the existing methods for calculating the monostatic RCS of a deformable S-shaped cavity still need to be further improved and accelerated. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network, so as to solve the deficiencies of high calculation complexity, long time period and poor timeliness in the prior art. The present invention calculates the monostatic RCS of a deformable S-shaped cavity with a limited number of different incident frequencies and angles, so as to rapidly estimate the electromagnetic scattering characteristics of the deformable S-shaped cavity and further improve the calculation efficiency of the RCS value of the deformable S-shaped cavity.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network, the method comprising:

[0008] Step S1, generating a data set based on the traditional SBR method, which includes: first constructing a series of structural parameters and corresponding electromagnetic parameters of a deformable S-shaped cavity; then calculating the monostatic RCS values of different S-shaped cavities based on the traditional SBR method; finally, using the structural parameters and corresponding electromagnetic parameters as the input of the data set, and using the RCS of the deformable S-shaped cavity as the label of the data set;

[0009] Step S2: Construct an artificial neural network model, where the artificial neural network model includes a plurality of fully connected layers connected in sequence;

[0010] Step S3: Divide the dataset in Step S1 into a training set and a test set according to a certain ratio, and then use the training set to train the artificial neural network model constructed in Step S2. When the loss values of the training set and the test set simultaneously reach the set threshold range, save the model parameters to obtain the trained artificial neural network model;

[0011] Step S4: Use the trained artificial neural network model obtained in Step S3 to perform the prediction of the RCS value, which includes: inputting the structural parameters and corresponding electromagnetic parameters of the S-shaped cavity to be predicted into the trained artificial neural network model, and using the output of the network model as the predicted RCS value.

[0012] Further, the structural parameters include: different radii r of the S-shaped cavity, different curvature structures α of the S-shaped cavity, and different front-to-back radius ratios ratio of the S-shaped cavity;

[0013] The electromagnetic parameters include: frequency f and plane wave polarization p.

[0014] Further, Step S1 specifically includes:

[0015] Step S101: Generate a series of radius r data of deformable S-shaped cavities, where the specific numerical range of the radius r data is: 0.11803 - 0.12803, and the step size is 0.0005; generate a series of curvature α of deformable S-shaped cavities, where the specific numerical range of the curvature α is: 1 - 1.7, and the step size is 0.01; generate a series of front-to-back radius ratios ratio of deformable S-shaped cavities, where the specific numerical range of the front-to-back radius ratio ratio is: 0.5 - 1, and the step size is 0.01; generate a series of electromagnetic parameters of deformable S-shaped cavities, which includes frequency f, the specific numerical range of the frequency f is: 8 - 12 GHz, the step size is 0.1 GHz, and the polarization direction is horizontal polarization or vertical polarization, and the incident angle is 90 degrees;

[0016] Step S102: Use the radius r, curvature α, front-to-back radius ratio ratio, frequency f, and plane wave polarization p of the deformable S-shaped cavity as the input of each sample to obtain the network input data corresponding to the unit, which includes: r, α, ratio, f, p; the network input data of each unit forms the input dataset;

[0017] Step S103: Perform row radian preprocessing on the first two columns of angles in the input data set constructed in Step S102. Multiply the angle values in the input data set by 3.14 / 180. After the radian preprocessing, the data is compressed between -3.14 and 3.14;

[0018] Step S104: Use the traditional SBR method to calculate the RCS of different structured S-shaped cavities, and use the RCS as the corresponding label of the input data set.

[0019] Further, in Step S2, the artificial neural network model includes five fully connected layers, and a LeakyReLU function is connected between each fully connected layer as the activation function;

[0020] The artificial neural network model uses the loss function nn.L1Loss() when calculating the loss value; and uses the Adam optimizer as the optimizer of the artificial neural network model

[0021] Further, the five fully connected layers are composed of 21, 121, 248, 22, and 1 neurons respectively;

[0022] And during training, the learning rate is set to 0.001, and the number of training epochs is 2000 times.

[0023] Further, Step S3 includes:

[0024] Step S301: Randomly select 20% of the samples in the data set as the test set, and then use the remaining samples as the training set and perform random shuffling;

[0025] Step S302: Use the training set to train the artificial neural network model to obtain the corresponding artificial neural network parameters.

[0026] Further, in Step S302, the process of training the artificial neural network model is specifically as follows:

[0027] First, through the data generator, several data in the training set are generated each time and put into the artificial neural network model for training. Training all the training set data once is regarded as one epoch, and the loss value of the training set is obtained;

[0028] Then, put the data of the test set into the artificial neural network model for testing and calculate the loss value of the test set through the loss function;

[0029] Next, through backpropagation, continuously adjust the network parameters to make the loss values of the training set and the test set reach the set threshold range at the same time. Finally, generate the trained artificial neural network model and save it.

[0030] Further, in the step S4, a sample is extracted from the test set, and then its structural parameters and electromagnetic parameters are determined; then the extracted sample is input into the trained artificial neural network model for prediction, and the corresponding predicted RCS value is compared with the single - station RCS value obtained by the traditional SBR method.

[0031] The beneficial effects of the present invention are as follows:

[0032] The present invention mainly predicts the single - station RCS value of the deformable S - type cavity through artificial neural network learning, so as to quickly simulate new deformable S - type cavities. By calculating the RCS values of a finite number of deformable S - type cavities and training the artificial neural network model, the single - station RCS value of the deformable S - type cavity can be predicted quickly with extremely high efficiency, saving a large amount of time. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a schematic flow chart of a method for quickly calculating the single - station RCS of a deformable S - type cavity based on an artificial neural network provided in Embodiment 1;

[0034] Figure 2 is a schematic diagram of the deformable S - type cavity provided in Embodiment 1;

[0035] Figure 3 is a comparison diagram of the single - station RCS of the plane - wave normal incidence at an incident angle of 90 degrees at different frequencies calculated by the traditional SBR method and the method of this embodiment;

[0036] Figure 4 is a schematic diagram of the artificial neural network model provided in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] Embodiment 1

[0039] Refer to Figures 1-4 , this embodiment provides a method for quickly calculating the single - station RCS of a deformable S - type cavity based on an artificial neural network. The specific process of this method is as Figure 1 shown, and specifically includes the following steps:

[0040] Step S1. Generate a dataset based on the traditional SBR method: First, construct a series of structural parameters of deformable S-shaped cavities and their corresponding electromagnetic parameters; then, calculate the monostatic RCS values of different S-shaped cavities based on the traditional SBR method; finally, use the structural parameters and their corresponding electromagnetic parameters as the input of the dataset, and use the RCS of the deformable S-shaped cavity as the label of the dataset; wherein, the structural parameters include: different radii r of the S-shaped cavity, different curvature structures α of the S-shaped cavity, and different front-to-back radius ratios ratio of the S-shaped cavity.

[0041] The electromagnetic parameters include: frequency f and plane wave polarization p.

[0042] Specifically, in this embodiment, step S1 includes:

[0043] Step S101. Generate a series of radius r data of deformable S-shaped cavities, where the specific value range of the radius r data is: 0.11803 - 0.12803, with a step size of 0.0005; generate a series of curvature α of deformable S-shaped cavities, where the specific value range of the curvature α is: 1 - 1.7, with a step size of 0.01; generate a series of front-to-back radius ratios ratio of deformable S-shaped cavities, where the specific value range of the front-to-back radius ratio ratio is: 0.5 - 1, with a step size of 0.01; generate a series of electromagnetic parameters of deformable S-shaped cavities, which include frequency f, and the specific value range of the frequency f is: 8 - 12 GHz, with a step size of 0.1 GHz, and the polarization direction is horizontal polarization or vertical polarization, and the incident angle is 90 degrees.

[0044] Step S102. Use the radius r, curvature α, front-to-back radius ratio ratio, frequency f, and plane wave polarization p of the deformable S-shaped cavity as the input of each sample to obtain the network input data (r, α, ratio, f, p) corresponding to the unit, and the network input data of each unit form the input dataset.

[0045] Step S103. Perform radian preprocessing on the first two columns of angles in the input dataset constructed in step S102. Multiply the angle values in the input dataset by 3.14 / 180. After radian preprocessing, the data is compressed between -3.14 and 3.14.

[0046] Step S104. Calculate the RCS of different structural S-shaped cavities using the traditional SBR method, and use the RCS as the corresponding label of the input dataset.

[0047] Step S2. Construct an artificial neural network model.

[0048] Specifically, in this embodiment, step S2 includes:

[0049] The above artificial neural network model includes five fully connected layers (Linear), and a LeakyReLU function is connected between each two fully connected layers as the activation function.

[0050] When calculating the loss value, the loss function nn.L1Loss() is adopted. Its function is actually to calculate the absolute value of the difference between the network output and the label, and the Adam optimizer is used as the optimizer of the artificial neural network model. The learning rate is set to 0.001, and the number of training epochs is 2000 times.

[0051] More specifically, in this embodiment, the above five fully connected layers are respectively composed of 21, 121, 248, 22, and 1 neurons.

[0052] Step S3: Divide the dataset in Step S1 into a training set and a test set according to a certain ratio, and then use the training set to train the artificial neural network model constructed in Step S2. When the loss values of the training set and the test set reach the set threshold range simultaneously, save the model parameters to obtain the trained artificial neural network model;

[0053] Specifically, in this embodiment, this Step S3 includes:

[0054] Step S301: Randomly extract 20% of the samples in the dataset as the test set, and then use the remaining samples as the training set and perform random shuffling to ensure that the trained model has a certain generalization performance;

[0055] Step S302: Use the training set to train the artificial neural network model to obtain the corresponding artificial neural network parameters.

[0056] More specifically, in this embodiment, the process of training the artificial neural network model is specifically as follows:

[0057] Through the data generator, several data in the training set are generated each time and put into the artificial neural network model for training. Training all the training set data once is regarded as one epoch to obtain the loss value of the training set, and then the data of the test set are put into the artificial neural network model for testing to obtain the loss value of the test set; through backpropagation, continuously adjust the network parameters to make the loss values of the training set and the test set reach the set threshold range simultaneously, and then generate and save the trained artificial neural network model.

[0058] Step S4: Use the trained artificial neural network model obtained in Step S3 to perform the prediction of the RCS value, which includes: inputting the structural parameters and corresponding electromagnetic parameters of the S-shaped cavity to be predicted into the trained artificial neural network model, and taking the output of this network model as the predicted RCS value.

[0059] Specifically, in this embodiment, step S4 specifically includes:

[0060] Step S401: Extract a sample from the above test set, and determine the structural parameters and electromagnetic parameters (frequency, polarization, etc.) of the S-shaped cavity therein; and determine the information of the plane wave according to the single-station RCS to be calculated.

[0061] Step S402: Then input the extracted sample into the trained artificial neural network model for prediction, and compare the obtained predicted RCS value with the single-station RCS value obtained by the traditional SBR method.

[0062] More specifically, Table 1 shows the time consumption required by the traditional SBR method and the method of this embodiment. It can be seen that the time consumption of the method of this embodiment is very small compared with the traditional method, thus proving the high efficiency of the method of this embodiment.

[0063] Table 1. Time consumption required for the calculation of the embodiments by the method of moments, the traditional SBR method, and the method of the present invention

[0064]

[0065] In summary, the present invention applies the trained artificial neural network model to the rapid calculation of the single-station RCS of the deformable S-shaped cavity, thereby avoiding the modeling and solution of the traditional SBR method, greatly reducing the calculation time, and improving the calculation efficiency of the single-station RCS of the deformable S-shaped cavity.

[0066] The parts not detailed in the present invention are all well-known technologies to those skilled in the art.

[0067] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art should be within the protection scope determined by the claims.

Claims

1. A method for quickly calculating the RCS of a single station of a deformable S-shaped cavity based on an artificial neural network, characterized in that: The method includes: Step S1: Generate a data set based on the traditional SBR method, which includes: First, construct a series of structural parameters of deformable S-shaped cavities and their corresponding electromagnetic parameters; then calculate the monostatic RCS values of different S-shaped cavities based on the traditional SBR method; finally, use the structural parameters and their corresponding electromagnetic parameters as the input of the data set, and the RCS of the deformable structure S-shaped cavity as the label of the data set. Step S2: Construct an artificial neural network model, where the artificial neural network model includes a plurality of fully connected layers connected in sequence. Step S3: Divide the data set in Step S1 into a training set and a test set according to a certain ratio, and then use the training set to train the artificial neural network model constructed in Step S2. When the loss function reaches the set threshold range at the same time, save the model parameters to obtain the trained artificial neural network model. Step S4: Use the trained artificial neural network model obtained in Step S3 to perform RCS value prediction, which includes: input the structural parameters and corresponding electromagnetic parameters of the S-shaped cavity to be predicted into the trained artificial neural network model, and use the output of the network model as the predicted RCS value. The specific steps of Step S1 include: Step S101: Generate a series of radius r data of deformable S-shaped cavities, where the specific value range of the radius r data is: 0.11803 - 0.12803, and the step size is 0.0005; generate a series of curvature α of deformable S-shaped cavities, where the specific value range of the curvature α is: 1 - 1.7, and the step size is 0.01; generate a series of front-to-back radius ratios ratio of deformable S-shaped cavities, where the specific value range of the front-to-back radius ratio ratio is: 0.5 - 1, and the step size is 0.01; generate a series of electromagnetic parameters of deformable S-shaped cavities, which include frequency f, where the specific value range of the frequency f is: 8 - 12 GHz, and the step size is 0.1 GHz, and the polarization direction is horizontal polarization or vertical polarization, and the incident angle is 90 degrees. Step S102: Use the radius r, curvature α, front-to-back radius ratio ratio, frequency f, and plane wave polarization p of the deformable S-shaped cavity as the input of each sample to obtain the corresponding network input data, which includes: r, α, ratio, f, p; the network input data of each unit forms the input data set. Step S103: Perform row radian conversion preprocessing on the first two columns of angles in the input data set constructed in Step S102. Multiply the angle values in the input data set by 3.14 / 180. After the radian conversion preprocessing, the angle data is compressed to between -3.14 and 3.

14. Step S104: Calculate the RCS of different structure S-shaped cavities using the traditional SBR method, and use the RCS as the corresponding label of the input data set.

2. The method for rapidly calculating the RCS of a single station of a deformable S-shaped cavity based on an artificial neural network according to claim 1 is characterized in that: The structural parameters include: different radii r of the S-shaped cavity, different curvature structures α of the S-shaped cavity, and different front-to-back radius ratios ratio of the S-shaped cavity. The electromagnetic parameters include: frequency f and plane wave polarization p.

3. The method for rapidly calculating the RCS of a single station of a deformable S-shaped cavity based on an artificial neural network according to claim 1, characterized in that: In the step S2, the artificial neural network model includes five fully-connected layers, and a LeakyReLU function is connected between each two fully-connected layers as the activation function; when calculating the loss value, the artificial neural network model uses the loss function nn.L1Loss(); and the Adam optimizer is used as the optimizer of the artificial neural network model.

4. A method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network according to claim 3, characterized in that, The five fully-connected layers are respectively composed of 21, 121, 248, 22, and 1 neurons; And during training, the learning rate is set to 0.001, and the number of training epochs is 2000.

5. A method for quickly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network according to claim 1, characterized in that The step S3 includes: Step S301: Randomly extract 20% of the samples in the dataset as the test set, and then use the remaining samples as the training set and perform random shuffling on them; Step S302: Use the training set to train the artificial neural network model, obtain the corresponding artificial neural network parameters and save them.

6. A method for quickly calculating the single - station RCS of a deformable S - shaped cavity based on an artificial neural network according to claim 5, characterized in that, In the step S302, the process of training the artificial neural network model is specifically as follows: First, through the data generator, several data in the training set are generated each time and put into the artificial neural network model for training. Training all the data in the training set once is regarded as one epoch, and the loss value of the training set is obtained; Then, the data in the test set are put into the artificial neural network model for testing, and the loss value of the test set is calculated through the loss function: Next, through backpropagation, the network parameters are continuously adjusted to make the loss values of the training set and the test set reach the set threshold range at the same time. Finally, the trained artificial neural network model is generated and saved.

7. A method for rapidly calculating the monostatic RCS of a deformable S-shaped cavity based on an artificial neural network according to claim 1, characterized in that, In the step S4, a sample is extracted from the test set, and then its structural parameters and electromagnetic parameters are determined; then the extracted sample is input into the trained artificial neural network model for prediction, and the corresponding predicted RCS value is compared with the single-station RCS value obtained by the traditional SBR method.

Images