Multi-stage synchronous induction coil cannonball outlet speed prediction method and system

Abstract

The invention discloses a multi-stage synchronous induction coil gun bullet outlet speed prediction method and system, and the method comprises the steps: obtaining a multi-stage synchronous induction coil gun parameter sample set, and dividing the multi-stage synchronous induction coil gun parameter sample set into a training set and a test set; establishing an RNN model, inputting the training set into the RNN model for training, testing the trained RNN model by using the test set, and evaluating the prediction capability of the RNN model according to the evaluation index; if the prediction capability of the RNN model reaches the standard, applying the tested RNN model to predict, otherwise, adopting a PSO optimization recurrent neural network model RNN to perform training by applying the training set again, selecting hyper-parameters of the RNN, and establishing a PSO-RNN model; inputting the test set to test the PSO-RNN model, evaluating the prediction capability of the PSO-RNN model according to the evaluation index until the prediction capability meets the requirement, and substituting the optimal hyper-parameter into the RNN model to obtain an optimal PSO-RNN model; the parameters of the launcher are input into the optimal PSO-RNN model, the exit speed of the projectile under the current parameters is obtained, and the prediction precision can be improved.

Description

Technical Field

[0001] The invention relates to the field of electromagnetic launch technology, and in particular to a method and system for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile. Background Art

[0002] The synchronous induction coil gun is one of the important methods of electromagnetic launch. Its working process is as follows: after the energy storage capacitor is fully charged, according to a specific trigger control strategy, the trigger discharge switch is turned on to discharge the drive coil, generate a pulse current, and generate a pulse magnetic field around the coil. According to the law of electromagnetic induction, the changing magnetic field generates induced eddy currents inside the launcher, which interact with the radial component of the coil magnetic field to generate an axial electromagnetic force, which acts on the launcher and causes it to move with variable acceleration.

[0003] The dynamic characteristics of synchronous induction coil guns are usually closely related to the timing, such as the current and magnetic field changes of each stage of the driving coil, the movement process of the projectile, the triggering time, etc., and the system parameters have strong nonlinear characteristics. At present, the prediction of the projectile exit velocity of synchronous induction coil guns is mainly carried out through two methods: numerical analysis and finite element modeling. These two methods generally have problems such as modeling involving complex multi-physical field coupling, long iterative calculation time, and low exit velocity prediction accuracy due to the nonlinearity of the system. And with the increase of the number of stages, the more input parameters involved, the faster the speed of the projectile, and the more difficult it is to accurately control the triggering time of the coil. The traditional method requires a large number of repeated iterative calculations, which has low calculation efficiency. In addition, to ensure the accuracy of the exit velocity prediction, the requirements for hardware equipment will be further increased. Therefore, a more efficient method is needed to complete the modeling and exit velocity prediction of the coil gun.

[0004] RNN neural networks are good at processing time series data and can capture the nonlinear relationship between system input and output at different times. The RNN model does not rely on mathematical relationships but expresses the transfer relationship by analyzing and fitting input and output to establish a "black box" model. Especially for multi-stage synchronous induction coil guns with many input parameters and complex coupling relationships, it can effectively improve computing efficiency, reduce requirements for hardware equipment, and achieve fast and accurate exit speed prediction.

[0005] In addition, since the coil gun system has many input and output parameters and the relationship between input and output is highly nonlinear, the RNN will have problems such as poor prediction accuracy and instability during training. It is necessary to further introduce optimization algorithms to optimize the RNN model hyperparameters (the number of neurons in each layer and the regularization coefficient) to improve the prediction accuracy of the non-parametric model of the multi-stage synchronous induction coil gun. Therefore, there is an urgent need for a method that can quickly and accurately model the multi-stage synchronous induction coil gun system and predict the exit speed. Summary of the invention

[0006] The technical problem to be solved by the present invention is to solve the problems of low prediction accuracy of projectile exit velocity of the current multi-stage synchronous induction coil gun and long modeling iteration time.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] A method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile, comprising:

[0009] Obtain a multi-stage synchronous induction coil gun parameter sample set and divide it into a training set and a test set;

[0010] Establish an RNN model, input the training set data into the RNN model for training, and then use the test set to test the trained RNN model, and evaluate the prediction ability of the RNN model based on the evaluation indicators;

[0011] If the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction. Otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again to select the hyperparameters of the RNN and establish the PSO-RNN model.

[0012] Input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation index, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model;

[0013] The launcher parameters are input into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

[0014] In one embodiment of the present invention, the parameter sample set of the multi-stage synchronous induction coil gun is obtained by finite element simulation through orthogonal experiment combined with random experiment design.

[0015] In one embodiment of the present invention, each sample in the parameter sample set of the multi-stage synchronous induction coil gun takes the number of turns of each stage of the driving coil, the projectile triggering position and the triggering time of each stage of the driving coil as input, and the projectile exit speed as output.

[0016] In one embodiment of the present invention, the prediction ability of the RNN model is evaluated according to the evaluation index, including:

[0017] S1, determine the structure of the RNN model: initialize the parameters of the RNN, including the input weight W x , hidden state weight W h , output weight W y , Input bias b x , hidden bias b h , output bias b y, initial hidden state h0, regularization parameter; determine the hidden layer activation function and the number of hidden layers;

[0018] S2, input the training set to train the RNN model;

[0019] S3, input the test set to test the RNN model, calculate the evaluation index and the error percentage between the predicted value and the simulated value of the exit speed;

[0020] S4, if the exit velocity error percentage is less than 5%, the number of turns of each level of drive coils, the projectile trigger position, and the trigger time of each level of drive coils are directly input into the RNN model that has passed the test to predict the projectile exit velocity; if the exit velocity error percentage is greater than 5%, the PSO algorithm is used to optimize the RNN model hyperparameters.

[0021] In one embodiment of the present invention, a PSO-RNN model is established, including:

[0022] S10, establishing the corresponding relationship between the PSO particle swarm algorithm and the RNN model hyperparameters;

[0023] S20, initializing the PSO algorithm, including population size, inertia weight, learning factor, maximum number of iterations, initial position and initial velocity of particles;

[0024] S30, using the root mean square error of the RNN model as the fitness function to calculate the particle fitness value;

[0025] S40, updating particle position and velocity;

[0026] S50, and applying the training set for training, inputting the optimized optimal hyperparameters into the RNN model to obtain the PSO-RNN model.

[0027] In one embodiment of the present invention, a corresponding relationship between a PSO particle swarm algorithm and a RNN model hyperparameter is established, including:

[0028] The hyperparameters of the RNN model are mapped to the positions and velocities of particles in the particle swarm algorithm, where the hyperparameters of the RNN model include: the number of neurons in each hidden layer and the regularization coefficient.

[0029] In one embodiment of the present invention, the particle fitness value is calculated using the following formula:

[0030]

[0031] In the formula, RMSE is the root mean square error, n is the number of samples, S i and P i are the simulated value and predicted value of the i-th sample respectively.

[0032] In one embodiment of the present invention, the particle position and velocity are updated by the following formula:

[0033] v id (t+1)=wv id (t)+c1r1(p id -x id )+c2r2(g d -x id );

[0034] x id (t+1)=x id (t)+v id (t+1);

[0035] In the formula, v id (t+1) and x id (t+1) are the velocity and position of the ith particle in the d-dimensional space at the (t+1)th iteration; p id is the historical optimal position of the individual particle, g d is the global optimal position; c1 and c2 are two learning factors, r1 and r2 are random numbers uniformly distributed in the interval [0,1], and w is the inertia weight.

[0036] In one embodiment of the present invention, obtaining an optimal PSO-RNN model includes:

[0037] S60, the test set is input into the trained PSO-RNN model to obtain the prediction result of the test set; the evaluation index and the error percentage are calculated according to the prediction result of the test set to determine whether the iteration reaches the termination condition;

[0038] S70, if the termination condition is not met, execute S30 to S60 to continue iteration; if the termination condition is met, the final optimal hyperparameters are obtained, and the optimal hyperparameters are brought into the RNN model to obtain the best PSO-RNN model.

[0039] The present invention also provides a multi-stage synchronous induction coil gun projectile exit velocity prediction system, which uses the multi-stage synchronous induction coil gun projectile exit velocity prediction method described above, comprising:

[0040] A sample set module is used to obtain a sample set of multi-stage synchronous induction coil gun parameters and divide it into a training set and a test set;

[0041] The RNN model module is used to establish the RNN model, input the training set data into the RNN model for training, and then use the test set to test the trained RNN model, and evaluate the prediction ability of the RNN model according to the evaluation indicators;

[0042] The initial PSO-RNN model module is used to judge. If the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction. Otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again to select the optimal hyperparameters of the RNN and establish the PSO-RNN model.

[0043] The best PSO-RNN model module is used to input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation index, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model;

[0044] The prediction module is used to input the launcher parameters into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention can solve the problems of multi-physical field coupling, long iterative calculation time, low exit velocity prediction accuracy due to nonlinearity between system parameters, etc. in the traditional synchronous induction coil gun modeling method, and realize rapid prediction of exit velocity; the use of PSO algorithm can solve the problems of poor prediction accuracy and instability caused by nonlinearity between synchronous induction coil gun system parameters in the RNN algorithm, and realize accurate prediction of exit velocity. In summary, the present invention can accurately predict the exit velocity of the projectile based on the various parameters of the coil gun, and provide accurate guidance for the parameter design of the coil gun prototype in engineering applications.

[0046] The present invention performs two-step parameter adjustment and optimization on the RNN model parameters. In the first stage, the RNN model is first trained, and the evaluation index is used to determine whether the prediction error meets the requirements. If the requirements are met, there is no need to fine-tune the parameters in the second stage. If not, the RNN model is fine-tuned in the second stage. The parameter adjustment at this time is based on the optimization of the RNN model parameters in the first stage, and the PSO algorithm is used to fine-tune the parameters of the RNN model again. Compared with the current mainstream practice of directly establishing a prediction model combining at least two algorithms and using sample set training to perform a parameter optimization method, the present invention can improve computing efficiency or reduce computing time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The present invention is a flowchart of a method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to an embodiment of the present invention.

[0048] Figure 2 This is a comparison chart of the predicted output values ​​and simulation values ​​of the traditional RNN neural network and the PSO-RNN neural network according to an embodiment of the present invention.

[0049] Figure 3Schematic diagram of the error between the predicted output value and the simulated value of the traditional RNN neural network and the PSO-RNN neural network in an embodiment of the present invention.

[0050] Figure 4 The present invention is a block diagram of a multi-stage synchronous induction coil gun projectile exit velocity prediction system. DETAILED DESCRIPTION

[0051] In order to facilitate those skilled in the art to understand the technical solution of the present invention, the technical solution of the present invention is further described in conjunction with the accompanying drawings of the specification.

[0052] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0053] See also Figure 1 As shown, the present invention provides a method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile, comprising:

[0054] S100, obtaining a multi-stage synchronous induction coil gun parameter sample set and dividing it into a training set and a test set.

[0055] In one embodiment of the present invention, the parameter sample set of the multi-stage synchronous induction coil gun is obtained by finite element simulation through orthogonal experiment combined with random experiment design.

[0056] In this embodiment, taking the three-stage synchronous induction coil gun as an example, the system parameters and projectile parameters of the multi-stage synchronous induction coil gun are shown in Tables 1 and 2. 81 groups of sample data are obtained through orthogonal experiments as shown in Table 3, and 19 groups of sample data are obtained through random experiments as shown in Table 4, where N1, N2, and N3 are the turns of the three-stage drive coil respectively; P1 is the projectile trigger position; T2 and T3 are the triggering moments of the second and third stage drive coils respectively.

[0057] Table 1 Synchronous induction coil gun system parameters

[0058]

[0059] Table 2 Projectile parameters

[0060]

[0061] Table 3 Orthogonal test sample set

[0062]

[0063] Table 4 Multi-stage synchronous induction coil gun parameter sample set

[0064]

[0065] In this embodiment, each sample in the multi-stage synchronous induction coil gun parameter sample set takes the number of turns of each stage of driving coils, the projectile triggering position and the triggering time of each stage of driving coils as input, and the projectile exit speed as output.

[0066] In this embodiment, the acquired multi-stage synchronous induction coil gun parameter sample set is preprocessed, such as normalization, and the sample set is divided into a test set and a training set at a ratio of 1:4.

[0067] S200, establishing an RNN model, inputting the data of the training set into the RNN model for training, and then applying the test set to test the trained RNN model, and evaluating the prediction ability of the RNN model according to the evaluation index.

[0068] In one embodiment of the present invention, the prediction ability of the RNN model is evaluated according to the evaluation index, including:

[0069] S1, determine the structure of the RNN model: initialize the parameters of the RNN, including the input weight W x , hidden state weight W h , output weight W y , Input bias b x , hidden bias b h , output bias b y , initial hidden state h0, regularization parameter; determine the hidden layer activation function and the number of hidden layers. The number of hidden layers is 3, and the hidden layer activation function is Tanh.

[0070] S2, input the training set and train the RNN model.

[0071] S3, input the test set to test the RNN model, calculate the evaluation index and the error percentage between the predicted value and the simulated value of the exit speed.

[0072] In this embodiment, the evaluation index is, for example, mean square percentage error, mean absolute percentage error, and root mean square error.

[0073] S4, if the exit speed error percentage is less than 5%, the number of turns of each level of drive coils, the projectile trigger position, and the trigger time of each level of drive coils are directly input into the RNN model that has passed the test to predict the projectile exit speed. If the exit speed error percentage is greater than 5%, the PSO algorithm is used to optimize the RNN model hyperparameters.

[0074] S300, if the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction, otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again for training to select the hyperparameters of the RNN and establish the PSO-RNN model.

[0075] In one embodiment of the present invention, PSO is specifically a particle swarm algorithm.

[0076] S400, input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation index, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model.

[0077] In one embodiment of the present invention, obtaining an optimal PSO-RNN model includes:

[0078] S10, establishing the corresponding relationship between the PSO particle swarm algorithm and the RNN model hyperparameters.

[0079] In this embodiment, the hyperparameters of the neural network of the RNN model, including the number of neurons in each hidden layer and the regularization coefficient, are mapped to the position and velocity of the particles in the particle swarm algorithm. The particles here are the basic individuals in the algorithm, representing a potential solution in the solution space of the optimization problem. The particles gradually approach the optimal solution by moving in the search space. The particle dimension D in the PSO algorithm is the dimension of the optimization object. In this embodiment, the optimization object includes the number of neurons in the three hidden layers and the regularization coefficient, so D is 4 in the present invention.

[0080] S20, initialize the PSO algorithm, including population size, inertia weight, learning factor, maximum number of iterations, initial position and initial velocity of particles.

[0081] S30, using the root mean square error of the RNN model as the fitness function, calculating the particle fitness value.

[0082] In this embodiment, the particle fitness value is calculated using the following formula:

[0083]

[0084] In the formula, RMSE is the root mean square error, n is the number of samples, S i and P i are the simulation value and predicted value of the i-th sample, that is, the projectile exit velocity obtained by simulation and the projectile exit velocity predicted by the non-parametric model.

[0085] S40, update particle position and velocity.

[0086] In this embodiment, the particle position and velocity are updated according to the following formula:

[0087] v id (t+1)=wv id (t)+c1r1(p id -x id )+c2r2(g d -x id );

[0088] x id (t+1)=x id (t)+v id (t+1);

[0089] In the formula, v id (t+1) and x id (t+1) are the velocity and position of the ith particle in the d-dimensional space at the (t+1)th iteration; p id is the historical optimal position of the individual particle, g d is the global optimal position; c1 and c2 are two learning factors, r1 and r2 are random numbers uniformly distributed in the interval [0,1], and w is the inertia weight. The particle speed and position must be limited to the set definition domain.

[0090] S50, applying the training set for training, inputting the optimized optimal hyperparameters into the RNN model, and obtaining the PSO-RNN model.

[0091] In one embodiment of the present invention, the above is the process of establishing a PSO-RNN model.

[0092] S60, the test set is input into the trained PSO-RNN model to obtain the prediction result of the test set; the evaluation index and the error percentage are calculated according to the prediction result of the test set to determine whether the iteration reaches the termination condition.

[0093] In this embodiment, the evaluation index and the error percentage are calculated according to the prediction results of the test set to determine whether the iteration reaches the termination condition, that is, whether the error percentage between the outlet speed prediction value and the simulation value is less than 5%.

[0094] S70, if the termination condition is not met, execute S30 to S60 to continue iteration; if the termination condition is met, the final optimal hyperparameters are obtained, and the optimal PSO-RNN model is obtained.

[0095] In this embodiment, the final optimal hyperparameters are input into the RNN model, and the model is used to predict the exit velocity of the multi-stage synchronous induction coil gun projectile.

[0096] S500, inputting the launcher parameters into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

[0097] In this embodiment, the test set is input into the traditional RNN neural network and the PSO-RNN model to obtain the prediction result comparison. Figure 2 It can be seen that the prediction model of the projectile exit velocity of the multi-stage synchronous induction coil gun based on the PSO-RNN neural network is closer to the simulation value than the prediction model of the traditional RNN neural network, and the prediction result is more accurate.

[0098] In this embodiment, Figure 3 The prediction errors of the traditional RNN neural network and the PSO-RNN neural network are compared. It can be seen that the maximum relative error of the PSO-RNN prediction model is 2.58, the average relative error is 0.95, and the maximum error percentage is 4.2%. The maximum error of the prediction model of the traditional RNN neural network is 6.5, the average relative error is 1.59, and the maximum error percentage is 11.7%. Considering the influence of the actual measurement error, the error percentage between the predicted value and the simulation value of the projectile exit velocity is acceptable within 5%. In summary, the prediction result of the projectile exit velocity of the multi-stage synchronous induction coil gun based on the PSO-RNN neural network is more accurate than that of the prediction model of the traditional RNN neural network, and has better generalization ability.

[0099] See also Figure 4 As shown, the present invention also provides a multi-stage synchronous induction coil gun projectile exit velocity prediction system, which uses the multi-stage synchronous induction coil gun projectile exit velocity prediction method described above, comprising:

[0100] The sample set module is used to obtain the parameter sample set of the multi-stage synchronous induction coil gun and divide it into a training set and a test set.

[0101] The RNN model module is used to establish the RNN model, input the training set data into the RNN model for training, and then use the test set to test the trained RNN model, and evaluate the prediction ability of the RNN model based on the evaluation indicators.

[0102] The initial PSO-RNN model module is used to judge. If the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction. Otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again to train, select the hyperparameters of the RNN, and establish the PSO-RNN model.

[0103] The best PSO-RNN model module is used to input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation indicators, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model.

[0104] The prediction module is used to input the launcher parameters into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

[0105] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting from any point of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any reference numerals in the claims should not be regarded as limiting the claims involved.

[0106] The above-described embodiments merely represent implementation methods of the invention. The protection scope of the present invention is not limited to the above-described embodiments. For those skilled in the art, several modifications and improvements may be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile, characterized in that: include: Obtain a multi-stage synchronous induction coil gun parameter sample set and divide it into a training set and a test set; Establish an RNN model, input the training set data into the RNN model for training, and then use the test set to test the trained RNN model, and evaluate the prediction ability of the RNN model based on the evaluation indicators; If the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction. Otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again to select the hyperparameters of the RNN and establish the PSO-RNN model. Input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation index, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model; The launcher parameters are input into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

2. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 1, characterized in that: The parameter sample set of the multi-stage synchronous induction coil gun is obtained by finite element simulation combined with orthogonal experiment and random experiment design.

3. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 1, characterized in that: Each sample in the parameter sample set of the multi-stage synchronous induction coil gun takes the number of turns of each stage of driving coil, the projectile triggering position and the triggering time of each stage of driving coil as input, and the projectile exit speed as output.

4. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 1, characterized in that: The prediction ability of the RNN model is evaluated based on evaluation indicators, including: S1, determine the structure of the RNN model: initialize the parameters of the RNN, including the input weight W x , hidden state weight W h , output weight W y , Input bias b x , hidden bias b h , output bias b y , initial hidden state h0, regularization parameter; determine the hidden layer activation function and the number of hidden layers; S2, input the training set to train the RNN model; S3, input the test set to test the RNN model, calculate the evaluation index and the error percentage between the predicted value and the simulated value of the exit speed; S4, if the exit velocity error percentage is less than 5%, the number of turns of each level of drive coils, the projectile trigger position, and the trigger time of each level of drive coils are directly input into the RNN model that has passed the test to predict the projectile exit velocity; if the exit velocity error percentage is greater than 5%, the PSO algorithm is used to optimize the RNN model hyperparameters.

5. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 1, characterized in that: Establish a PSO-RNN model, including: S10, establishing the corresponding relationship between the PSO particle swarm algorithm and the RNN model hyperparameters; S20, initializing the PSO algorithm, including population size, inertia weight, learning factor, maximum number of iterations, initial position and initial velocity of particles; S30, using the root mean square error of the RNN model as the fitness function to calculate the particle fitness value; S40, updating particle position and velocity; S50, and applying the training set for training, inputting the optimized optimal hyperparameters into the RNN model to obtain the PSO-RNN model.

6. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 5, characterized in that: Establish the corresponding relationship between the PSO particle swarm algorithm and the RNN model hyperparameters, including: The hyperparameters of the RNN model are mapped to the positions and velocities of particles in the particle swarm algorithm, where the hyperparameters of the RNN model include: the number of neurons in each hidden layer and the regularization coefficient.

7. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 5, characterized in that: Calculate the particle fitness value using the following formula: In the formula, RMSE is the root mean square error, n is the number of samples, S i and P i are the simulated value and predicted value of the i-th sample respectively.

8. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 5, characterized in that: Update particle position and velocity using the following formula: v id (t+1)=wv id (t)+c1r1(p id -x id )+c2r2(g d -x id ); x id (t+1)=x id (t)+v id (t+1); In the formula, v id (t+1) and x id (t+1) are the velocity and position of the ith particle in the d-dimensional space at the (t+1)th iteration; p id is the historical optimal position of the individual particle, g d is the global optimal position; c1 and c2 are two learning factors, r1 and r2 are random numbers uniformly distributed in the interval [0,1], and w is the inertia weight.

9. The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to claim 5, characterized in that: Get the best PSO-RNN model, including: S60, the test set is input into the trained PSO-RNN model to obtain the prediction result of the test set; the evaluation index and the error percentage are calculated according to the prediction result of the test set to determine whether the iteration reaches the termination condition; S70, if the termination condition is not met, execute S30 to S60 to continue iteration; if the termination condition is met, the final optimal hyperparameters are obtained, and the optimal hyperparameters are brought into the RNN model to obtain the best PSO-RNN model.

10. A multi-stage synchronous induction coil gun projectile exit velocity prediction system, characterized in that: The method for predicting the exit velocity of a multi-stage synchronous induction coil gun projectile according to any one of claims 1 to 9 comprises: A sample set module is used to obtain a sample set of multi-stage synchronous induction coil gun parameters and divide it into a training set and a test set; The RNN model module is used to establish the RNN model, input the training set data into the RNN model for training, and then use the test set to test the trained RNN model, and evaluate the prediction ability of the RNN model according to the evaluation indicators; The initial PSO-RNN model module is used to judge. If the prediction ability of the RNN model meets the standard, the tested RNN model is used for prediction. Otherwise, the PSO is used to optimize the recurrent neural network model RNN, and the training set is used again to select the hyperparameters of the RNN and establish the PSO-RNN model. The best PSO-RNN model module is used to input the test set to test the PSO-RNN model, evaluate the prediction ability of the PSO-RNN model according to the evaluation index, until the prediction ability meets the requirements, bring the optimal hyperparameters into the RNN model, and obtain the best PSO-RNN model; The prediction module is used to input the launcher parameters into the optimal PSO-RNN model to obtain the exit velocity of the projectile under the current parameters.

Images