High-altitude electromagnetic pulse waveform rapid calculation method based on bayesian optimization neural network

Abstract

The application discloses a high-altitude electromagnetic pulse waveform rapid calculation method based on a Bayesian optimization neural network, solves the technical problem that the calculation efficiency of the existing high-altitude electromagnetic pulse waveform calculation is low due to the introduction of secondary calculation time and error, and by using a Bayesian optimization multilayer feedforward neural network model, an amplitude prediction model, an alpha prediction model, a beta prediction model, a k prediction model and a t0 prediction model are respectively constructed, so that the rapid calculation of the high-altitude electromagnetic pulse waveform is realized.

Description

Technical Field

[0001] This invention relates to the calculation of high-altitude electromagnetic pulses, and more specifically to a method for fast calculation of high-altitude electromagnetic pulse waveforms based on a Bayesian optimized neural network. Background Technology

[0002] High-altitude electromagnetic pulse (HEMP) is a transient, powerful electromagnetic pulse that has significant destructive and interference effects on electronic information systems. Currently, there are several methods for calculating HEMP waveforms in E1 environments (such as the early E1 environment as defined by IEC 61000-2-9). One method involves obtaining the distribution values ​​of Emax (amplitude), tr (leading edge), and tw (half-width) of the waveform through traditional numerical calculations, and then establishing a database of correspondences between these values ​​and α (slow decay constant), β (fast decay constant), and k (amplitude coefficient). Given tr and tw, α, β, and k are obtained by reverse searching the database, thus enabling the calculation of the HEMP waveform. Another method involves extensive calculations to establish a neural network model of tr, tw, α, β, and k, followed by approximate formula derivation to determine the HEMP waveform.

[0003] Existing methods for calculating the waveform of high-altitude electromagnetic pulses at different explosion heights, gamma equivalents, and distances require first calculating tr and tw, then calculating α, β, and k using tr and tw, and finally calculating the waveform of the high-altitude electromagnetic pulse based on α, β, and k. This approach introduces secondary calculation time and errors, resulting in low computational efficiency. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problem that the calculation of high-altitude electromagnetic pulse waveforms is inefficient due to the introduction of secondary calculation time and errors, and to provide a fast calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks.

[0005] To achieve the above objectives, the technical solution provided by this invention is as follows:

[0006] A fast calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks, characterized by the following steps:

[0007] S1. Obtain the target amplitude, target slow decay constant, target fast decay constant, target amplitude coefficient, and target start time of the high-altitude electromagnetic pulse at the preset explosion height, gamma equivalent, and distance through numerical calculation;

[0008] S2. Construct a multi-layer feedforward neural network model and train it 5 times. The input for each training is the explosion height, gamma equivalent and distance.

[0009] The output of the first training is the predicted amplitude, and the training label is the target amplitude. During the training process, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, an amplitude prediction model is constructed.

[0010] The output of the second training is the predicted slow decay constant, and the training label is the target slow decay constant. During the training process, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, the α prediction model is constructed.

[0011] The output of the third training is the predicted fast decay constant, and the training label is the target fast decay constant. During the training process, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, a β prediction model is constructed.

[0012] The output of the fourth training iteration is the predicted amplitude coefficient, and the training label is the target amplitude coefficient. During the training process, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, a k-prediction model is constructed.

[0013] The output of the fifth training iteration is the predicted start time, and the training label is the target start time. During the training process, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, the t0 prediction model is constructed.

[0014] S3. Based on the amplitude prediction model, α prediction model, β prediction model, k prediction model and t0 prediction model, calculate the waveform of the high-altitude electromagnetic pulse.

[0015] Furthermore, in step S2, the process of constructing the amplitude prediction model is as follows:

[0016] a1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0017] a2. Using explosion height, gamma equivalent, and distance as training inputs, target amplitude as training labels, and predicted amplitude as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted amplitude and the target amplitude, which is used as the objective function for the number of hidden layer nodes.

[0018] a3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0019] a4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0020] a5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, an amplitude prediction model is constructed, wherein the number of hidden layer nodes in the amplitude prediction model is 19.

[0021] Furthermore, the number of hidden layer nodes in the α prediction model is 16, the number of hidden layer nodes in the β prediction model is 11, the number of hidden layer nodes in the k prediction model is 13, and the number of hidden layer nodes in the t0 prediction model is 10.

[0022] Furthermore, in step a3, the process of establishing a distribution model between the number of hidden layer nodes and the objective function value using a Gaussian process is as follows:

[0023]

[0024] In the formula, For the distribution model, For Gaussian processes, Let be the objective function. Let covariance function be used. and This indicates the number of different hidden layer nodes.

[0025] Furthermore, step S3 is as follows:

[0026] S3.1 Input the explosion height, gamma equivalent, and distance into the amplitude prediction model, α prediction model, β prediction model, k prediction model, and t0 prediction model respectively to obtain the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time.

[0027] S3.2 Calculate the waveform of the high-altitude electromagnetic pulse based on the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time.

[0028] Furthermore, in step S3.2, the formula for calculating the high-altitude electromagnetic pulse waveform is as follows:

[0029]

[0030] In the formula, This is a high-altitude electromagnetic pulse waveform. For amplitude, Here, α is the amplitude coefficient, and α is the slow decay constant. It is a fast decay constant. The start time, For predicting time.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. This invention relates to a method for rapid calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks. By utilizing a Bayesian optimized multilayer feedforward neural network model, amplitude prediction model, α prediction model, β prediction model, k prediction model, and t0 prediction model are constructed respectively, thereby realizing rapid calculation of high-altitude electromagnetic pulse waveforms. Compared with traditional numerical simulation methods or unoptimized neural network models, the calculation speed is reduced from 2-3 hours to within 1 second, solving the problem of low calculation efficiency caused by the introduction of secondary calculation time and errors in existing calculation methods.

[0033] 2. The present invention provides a method for rapid calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks. It uses the Bayesian optimization algorithm to determine the optimal number of hidden layer nodes in a multi-layer feedforward neural network model, avoiding the time-consuming and laborious manual parameter tuning process in traditional neural network training, improving prediction accuracy, and enhancing the adaptability of the network.

[0034] 3. The present invention provides a method for rapid calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks. This method integrates the complex calculation process of high-altitude electromagnetic pulse waveforms into an end-to-end digital intelligent process, reducing the calculation process, improving real-time online computing capabilities, and shortening the traditional numerical calculation time of 2-3 hours to the second level.

[0035] 4. The present invention provides a method for rapid calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks. It supports rapid comparative analysis of large-scale scenarios and can complete waveform distribution calculation and horizontal comparison of hundreds of different scenarios within minutes, significantly improving the efficiency of multi-scheme comparison and selection. It is suitable for the global assessment of the high-altitude electromagnetic pulse resistance capability of large facilities. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating an embodiment of the fast calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to the present invention.

[0037] Figure 2 This is a schematic diagram of the distribution field of an embodiment of the fast calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks of the present invention;

[0038] Figure 3 This is a high-altitude electromagnetic pulse waveform diagram, representing an embodiment of the high-altitude electromagnetic pulse waveform fast calculation method based on Bayesian optimized neural networks according to the present invention. Detailed Implementation

[0039] To make the objectives, advantages, and features of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are merely used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0040] This embodiment presents a rapid calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks. Numerical calculations were used to compute waveform parameters along the north-south axis, including amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time at different explosion heights, gamma equivalents, and distances, totaling 336 groups. The data was divided into 288 training data sets and 48 prediction data sets. By comparing support vector machines, k-nearest neighbors, generalized regression neural network models, and long short-term memory network models, a multilayer feedforward neural network model was selected as the benchmark model to approximate the physical modeling process of high-altitude electromagnetic pulses. The multilayer feedforward neural network model consists of an input layer, hidden layers, and an output layer. The selection of the number of nodes in the hidden layer affects the network structure and modeling accuracy. In this embodiment, to improve the accuracy of the predicted parameters and reduce mutual interference, five models were established for amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time, respectively. Each model predicts the corresponding waveform parameters.

[0041] This embodiment presents a fast calculation method for high-altitude electromagnetic pulse waveforms based on a Bayesian optimized neural network, such as... Figure 1 As shown, it includes the following steps:

[0042] S1. Target amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time are obtained through numerical calculations at different explosion heights, gamma equivalents, and distances within the coverage area of ​​a pre-defined distribution field for high-altitude electromagnetic pulses. The calculation formula for the radius R of the distribution field's coverage area, based on the explosion height, is as follows:

[0043]

[0044] In the formula, The explosion height; the radius R of the field distribution coverage can be determined based on the hob, such as... Figure 2 As shown, the waveform of the high-altitude electromagnetic pulse is calculated within the coverage radius R.

[0045] S2. Construct a multi-layer feedforward neural network model and train it 5 times. The input for each training is the explosion height, gamma equivalent and distance.

[0046] The output of the first training iteration is the predicted amplitude, and the training label is the target amplitude. During training, a Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multi-layer feedforward neural network model. Based on the optimal number of hidden layer nodes, an amplitude prediction model is constructed. The specific process is as follows:

[0047] a1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0048] a2. Using explosion height, gamma equivalent, and distance as training inputs, target amplitude as training labels, and predicted amplitude as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted amplitude and the target amplitude, which is used as the objective function for the number of hidden layer nodes.

[0049] a3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0050] a4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0051] a5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, an amplitude prediction model is constructed, wherein the number of hidden layer nodes in the amplitude prediction model is 19.

[0052] The output of the second training iteration is the predicted slow decay constant, and the training label is the target slow decay constant. During training, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, an α prediction model is constructed. The specific process is as follows:

[0053] b1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0054] b2. Using the explosion height, gamma equivalent, and distance as training inputs, the target slow decay constant as training labels, and the predicted slow decay constant as training outputs, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted slow decay constant and the target slow decay constant, which is used as the objective function for the number of hidden layer nodes.

[0055] b3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0056] b4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0057] b5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, an α prediction model is constructed, wherein the number of hidden layer nodes in the α prediction model is 16.

[0058] The output of the third training iteration is the predicted fast decay constant, and the training label is the target fast decay constant. During training, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, a β prediction model is constructed. The specific process is as follows:

[0059] c1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0060] c2. Using explosion height, gamma equivalent and distance as training input, target fast decay constant as training label, and predicted fast decay constant as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted fast decay constant and the target fast decay constant, which is used as the objective function for the number of hidden layer nodes.

[0061] c3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0062] c4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0063] c5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, a β prediction model is constructed, wherein the number of hidden layer nodes in the β prediction model is 11.

[0064] The output of the fourth training iteration is the predicted magnitude coefficient, and the training label is the target magnitude coefficient. During training, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multilayer feedforward neural network model. Based on the optimal number of hidden layer nodes, a k-prediction model is constructed. The specific process is as follows:

[0065] d1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0066] d2. Using explosion height, gamma equivalent, and distance as training inputs, target amplitude coefficient as training labels, and predicted amplitude coefficient as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted amplitude coefficient and the target amplitude coefficient, which is used as the objective function for the number of hidden layer nodes.

[0067] d3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0068] d4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0069] d5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, a k-prediction model is constructed, wherein the number of hidden layer nodes in the k-prediction model is 13.

[0070] The output of the fifth training iteration is the predicted start time, and the training label is the target start time. During training, the Bayesian optimization algorithm is used to determine the optimal number of hidden layer nodes in the multi-layer feedforward neural network model. Based on the optimal number of hidden layer nodes, the t0 prediction model is constructed. The specific process is as follows:

[0071] e1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes;

[0072] e2. Using explosion height, gamma equivalent, and distance as training inputs, target start time as training label, and predicted start time as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted start time and the target start time, which is used as the objective function for the number of hidden layer nodes.

[0073] e3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process;

[0074] e4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training.

[0075] e5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, a t0 prediction model is constructed, wherein the number of hidden layer nodes in the t0 prediction model is 10.

[0076] The Bayesian optimization strategy selects a new number of hidden layer nodes based on a distribution model. It uses a collection function to weigh the options between "exploring unknown regions" and "utilizing existing optimal regions" to select the next set of hidden layer nodes that are most likely to improve performance.

[0077] S3. Input the explosion height, gamma equivalent, and distance into the amplitude prediction model, α prediction model, β prediction model, k prediction model, and t0 prediction model respectively to obtain the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time.

[0078] S4. Based on the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time, calculate the high-altitude electromagnetic pulse waveform, such as... Figure 3 As shown. The formula for calculating the waveform of a high-altitude electromagnetic pulse is as follows:

[0079]

[0080] In the formula, This is a high-altitude electromagnetic pulse waveform. For amplitude, Here, α is the amplitude coefficient, and α is the slow decay constant. It is a fast decay constant. The start time, Let E(t) be the time variable, representing the change of the electric field amplitude with time t at different time intervals. Figure 3 It can be seen that within the range of variable t, the generated E(t) is...

[0081] In steps a3, b3, c3, d3, and e3, the process of establishing a distribution model between the number of hidden layer nodes and the objective function value using a Gaussian process is as follows:

[0082]

[0083] In the formula, For the distribution model, For Gaussian processes, Let be the objective function. Let covariance function be used. and Indicates the number of different hidden layer nodes. The formula is as follows:

[0084]

[0085] In the formula, For variance; The Euclidean distance between different numbers of hidden layer nodes; It is a length scale used to control the rate of change of the function.

[0086] In steps a2, b2, c2, d2, and e2, the formula for calculating the mean squared error (MSE) of the loss function is as follows:

[0087]

[0088] In the formula, Indicates the target parameter value. Indicates the predicted parameter value. This indicates the number of training samples.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A fast calculation method for high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks, characterized in that, Includes the following steps: S1. Obtain the target amplitude, target slow decay constant, target fast decay constant, target amplitude coefficient, and target start time of the high-altitude electromagnetic pulse at the preset explosion height, gamma equivalent, and distance through numerical calculation; S2. Construct a multi-layer feedforward neural network model and train it five times. The input for each training iteration is the explosion height, gamma equivalent, and distance. The outputs for the five training iterations are the predicted amplitude, predicted slow decay constant, predicted fast decay constant, predicted amplitude coefficient, and predicted start time, respectively. The training labels are the target amplitude, target slow decay constant, target fast decay constant, target amplitude coefficient, and target start time, respectively. During the five training iterations, the optimal number of hidden layer nodes in the multi-layer feedforward neural network model is determined using the Bayesian optimization algorithm. Based on the optimal number of hidden layer nodes, amplitude prediction models, α prediction models, β prediction models, k prediction models, and t0 prediction models are constructed, respectively. S3. Based on the amplitude prediction model, α prediction model, β prediction model, k prediction model and t0 prediction model, calculate the waveform of the high-altitude electromagnetic pulse.

2. The method for fast calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to claim 1, characterized in that, In step S2, the process of constructing the amplitude prediction model is as follows: a1. Construct a multi-layer feedforward neural network model and set the number of hidden layer nodes; a2. Using explosion height, gamma equivalent, and distance as training inputs, target amplitude as training labels, and predicted amplitude as training output, a multi-layer feedforward neural network model is trained under the condition of the number of hidden layer nodes. The mean square error of the loss function is calculated based on the difference between the predicted amplitude and the target amplitude, which is used as the objective function for the number of hidden layer nodes. a3. Based on the objective function for the number of hidden layer nodes, establish a distribution model between the number of hidden layer nodes and the objective function using a Gaussian process; a4. Based on the distribution model, select a new number of hidden layer nodes as the number of hidden layer nodes through the Bayesian optimization strategy, return to step a2, and continue until the preset training stopping condition is met to complete the training. a5. Based on the distribution model determined during training, the number of hidden layer nodes that minimizes the objective function value is taken as the optimal number of hidden layer nodes. Based on the optimal number of hidden layer nodes, an amplitude prediction model is constructed, wherein the number of hidden layer nodes in the amplitude prediction model is 19.

3. The method for fast calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to claim 2, characterized in that: In step S2, the number of hidden layer nodes in the α prediction model is 16, the number of hidden layer nodes in the β prediction model is 11, the number of hidden layer nodes in the k prediction model is 13, and the number of hidden layer nodes in the t0 prediction model is 10.

4. The method for fast calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to claim 3, characterized in that: In step a3, the process of establishing a distribution model between the number of hidden layer nodes and the objective function value using a Gaussian process is as follows: ； In the formula, For the distribution model, For Gaussian processes, Let be the objective function. Let covariance function be used. and This represents the number of different hidden layer nodes selected by the Bayesian optimization strategy.

5. The method for fast calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to claim 4, characterized in that, Step S3 is as follows: S3.1 Input the explosion height, gamma equivalent, and distance into the amplitude prediction model, α prediction model, β prediction model, k prediction model, and t0 prediction model respectively to obtain the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time. S3.2 Calculate the waveform of the high-altitude electromagnetic pulse based on the predicted amplitude, slow decay constant, fast decay constant, amplitude coefficient, and start time.

6. The method for fast calculation of high-altitude electromagnetic pulse waveforms based on Bayesian optimized neural networks according to claim 5, characterized in that: In step S3.2, the formula for calculating the high-altitude electromagnetic pulse waveform is as follows: ； In the formula, This is a high-altitude electromagnetic pulse waveform. For amplitude, Here, α is the amplitude coefficient, and α is the slow decay constant. It is a fast decay constant. The start time, It is a time variable.

Images