A method for calculating damage probability based on machine learning neural network

Abstract

The present invention discloses a damage probability calculation method based on a machine learning neural network, including: establishing a damage probability calculation model for a fragment warhead to a target; according to the above damage probability calculation model, a large number of random calculations are carried out based on the Monte Carlo method for all random parameters that affect the calculation results, and each parameter is randomly selected in each calculation; a large number of damage probability calculation example data obtained in the previous step are trained using the neural network in machine learning to obtain a damage probability prediction model, which can predict the damage probability according to the values ​​of each random parameter; using the obtained damage probability prediction model to use the Monte Carlo method to calculate the average single-shot damage probability of a fragment warhead to a target under specific missile-target intersection conditions. Compared with the current conventional method for calculating damage probability, the damage probability calculation method based on machine learning of the present invention not only ensures a certain accuracy of calculation precision, but also greatly reduces the calculation amount of the algorithm, greatly shortening the calculation time.

Description

Technical Field

[0001] The present invention belongs to the technical field of damage assessment, and in particular relates to a damage probability calculation method based on a machine learning neural network. Background Art

[0002] In order to accurately evaluate the damage effect of fragmentation warheads on targets and facilitate the study of ignition and warfare coordination, it is necessary to establish a corresponding damage probability calculation model based on target vulnerability analysis.

[0003] The commonly used damage probability represents the probability of a warhead damaging a target. The existing damage probability calculation method for a fragment warhead to a target is too computationally intensive. In MATLAB, it often takes hundreds or even thousands of seconds to calculate a single result for complex situations, and the amount of calculation increases exponentially with the increase in the number of fragments. When the Monte Carlo method is used to perform 1,000 random repetitions of calculations in order to obtain the average single-shot damage probability, the time required is calculated in days. Therefore, it is necessary to find a method to quickly calculate the damage probability of a fragment warhead to a target while ensuring a certain degree of accuracy in the result. Summary of the invention

[0004] The purpose of the present invention is to propose a method for calculating the probability of damage based on machine learning neural network to improve the calculation efficiency. Compared with the conventional method for calculating the probability of damage, it not only ensures a certain accuracy of the calculation precision, but also greatly reduces the amount of calculation of the algorithm and greatly shortens the calculation time.

[0005] The technical solutions to achieve the above objectives are as follows:

[0006] A method for calculating damage probability based on a machine learning neural network comprises the following steps:

[0007] 1) Establish a three-dimensional model of the target, conduct vulnerability analysis on the components of the target, and establish a model for calculating the probability of damage to the target by the fragmentation warhead;

[0008] 2) According to the damage probability calculation model, for all random parameters that affect the calculation results, repeated random calculations based on the Monte Carlo method are carried out for more than m times to obtain a large amount of damage probability calculation case data for different damage levels; the damage levels include: M-level mission damage, F-level fire control damage, and K-level catastrophic damage;

[0009] 3) Use the neural network in machine learning to train the above-mentioned large amount of damage probability calculation data (more than 10,000 groups) to obtain a damage probability prediction model for different damage levels;

[0010] 4) Using the damage probability prediction model obtained in the previous step, under the specific missile-target intersection conditions input by the superior, a part of the random parameters are fixed, and the Monte Carlo method is used to carry out more than n repeated random calculations based on the damage probability prediction model. After statistical data, the average single-shot damage probability of the warhead to the target under the specific missile-target intersection conditions is obtained.

[0011] Optionally, in step 1), the damage probability calculation model is specifically:

[0012]

[0013] Among them, N or is the number of non-redundant components, k is the number of redundant component groups, P or,i is the damage probability of each non-redundant component; M1, M2, ..., M k are the number of redundant components in each redundant component group; P and,j is the damage probability of each component in the redundant component group; the components of the target include multiple groups of redundant components and multiple non-redundant components. If all the redundant components in any one group of redundant components are damaged, the target is considered to be damaged; if any one non-redundant component is damaged, the target is considered to be damaged.

[0014] Optionally, in step 2), M-level mission damage means that the target functional components cannot fully perform their functions, and maintenance personnel need 1 to 24 hours to remove the obstacles; F-level fire control damage means that the target functional components have lost their ability to function, and it takes a special maintenance unit 1 to 7 days and nights to remove the obstacles; K-level catastrophic damage means that the target weapon system has lost its combat capability and cannot be repaired or the repair of the damage is not economically feasible.

[0015] Optionally, in step 2), m is greater than 10,000.

[0016] Optionally, the random parameters in step 2) include: terminal flight speed of the fragment warhead, entry angle, pitch angle, miss distance, and speed and scattering direction of each fragment.

[0017] Optionally, in step 3), the neural network includes: an input layer, a hidden layer, and an output layer; the number of nodes in the hidden layer is 75% of the number of nodes in the input layer; and the number of hidden layers is an integer greater than 5.

[0018] Optionally, in the step 3), the input data set of the input layer includes: a training set, a validation set and a test set; the training set is used to train the model, the validation set is used to verify and adjust parameters of the trained damage probability prediction model, and the test set is used to evaluate the performance of the damage probability prediction model.

[0019] Optionally, in step 3), the data volume ratios of the training set, the validation set and the test set are 70%, 15% and 15% respectively.

[0020] Optionally, in step 3), the algorithm used for neural network training is gradient descent method, Newton algorithm or conjugate gradient method.

[0021] Optionally, in step 4), n is greater than 500.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1) This method uses a damage probability prediction model to replace the original damage probability calculation model, which significantly reduces the amount of calculation and improves the calculation efficiency by more than 60 times;

[0024] 2) The average single-shot damage probability obtained by Monte Carlo calculation using the damage probability prediction model of this method can achieve an average accuracy of more than 85% compared with the result obtained by Monte Carlo calculation using the damage probability calculation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0026] The present invention discloses a damage probability calculation method based on a machine learning neural network, comprising: first establishing a three-dimensional model of a target, performing a vulnerability analysis on each component thereof, and then establishing a damage probability calculation model of a fragment warhead to the target. Further, for all random parameters such as projectile-target characteristics and intersection conditions that affect the calculation results, more than 10,000 random calculations are performed based on the Monte Carlo method, and each random parameter is randomly valued according to its own characteristics in each calculation. A large amount of damage probability calculation example data obtained in the previous step is trained using a machine learning neural network method to obtain a damage probability prediction model, which can predict the damage probability according to the values ​​of each parameter. The obtained damage probability prediction model is used to replace the original damage probability calculation model, and the Monte Carlo method is used again to calculate the average single-shot damage probability of the warhead to the target under specific projectile-target intersection conditions. The damage probability prediction model can be used for damage probability calculation under different projectile-target intersection conditions, and the calculation amount is very small, avoiding the huge calculation amount of the damage probability calculation model. Therefore, this method can greatly shorten the time required for calculating the average single-shot damage probability while ensuring a certain accuracy of the damage probability calculation result.

[0027] The present invention provides a damage probability calculation method based on a machine learning neural network, which uses a neural network method in machine learning to train a damage probability prediction model to calculate the damage probability of a fragmentation warhead to a target (such as an air defense missile system) under specific conditions, and includes the following steps:

[0028] S101: Establish a model for calculating the probability of damage to the target by the fragmentation warhead;

[0029] S201: Based on the damage probability calculation model, for all random parameters that affect the calculation results, such as the terminal flight speed, entry angle, pitch angle, miss distance of the target warhead, and the speed and scattering direction of each fragment, more than 10,000 repeated random calculations are carried out based on the Monte Carlo method. In each calculation, each random parameter is randomly assigned a value within a certain range according to a certain probability distribution. For example, the entry angle of the missile relative to the target is randomly assigned a value in the range of 0° to 360° according to a uniform distribution, the ballistic angle of the missile is randomly assigned a value in the range of -10° to 0° according to a uniform distribution, the yaw angle of the missile is randomly assigned a value in the range of -2° to 2° according to a uniform distribution, the terminal speed of the missile is randomly assigned a value in the range of 200m / s to 300m / s according to a uniform distribution, and the vertical miss distance of the missile is randomly assigned a value in the range of -1m to 1m according to a normal distribution, etc.;

[0030] S301: using a neural network in machine learning to train the large amount of damage probability calculation example data (more than 10,000 groups) to obtain a damage probability prediction model, which can be used to predict the damage probability according to the values ​​of various random parameters, thereby avoiding a huge amount of calculation in the damage probability calculation model;

[0031] A two-layer feedforward neural network is used for training, and "mae" is used as the network performance function, that is, the average value of the absolute error between the predicted value and the true value is used to measure the performance of the neural network; the input data set of the neural network is divided into three parts: training set, validation set and test set. The proportions of the three in the data set are generally 70%, 15% and 15% respectively, which can be adjusted according to the training results. The training set is used to train the model, the validation set is used to verify and adjust the parameters of the trained model, and the test set is used to evaluate the performance of the prediction model; when the neural network satisfies the mean square error of the predicted value and the target value is less than a decimal (such as 1e-7), the training ends and the training result is obtained.

[0032] Neural network includes: input layer, hidden layer, output layer, where the number of nodes in the input layer is generally fixed, consistent with the number of independent variables in the data set, the number of nodes in the output layer is the target number of the data set (the number of dependent variables), and the number of nodes in the hidden layer is generally 75% of the number of nodes in the input layer. A neural network can have multiple hidden layers, and the number of hidden layers is generally related to the complexity of the problem to be processed, and is generally an integer greater than 5.

[0033] S401: Using the damage probability prediction model obtained in the previous step, under specific missile-target intersection conditions, according to physical reality, a part of the random parameters are fixed, which can be one or more. The specific fixed parameters are determined by physical reality or research needs, such as entry angle and pitch angle, etc., and the Monte Carlo method is used again to carry out more than 500 repeated random calculations. After statistical data, the average single-shot damage probability of the warhead to the target under the specific missile-target intersection condition is obtained (the damage probability prediction model outputs a single-shot damage probability in a single calculation. If the Monte Carlo method is used for multiple random calculations, multiple single-shot damage probabilities can be obtained, and the average single-shot damage probability is obtained after averaging). By changing the fixed parameter type, the average single-shot damage probability of the warhead to the target under different conditions can be conveniently calculated, for example, the entry angle of the fixed missile is 270°, or the terminal velocity of the fixed missile is 200m / s, or the ballistic angle of the fixed missile is -5°, etc.

[0034] Furthermore, the specific method for establishing a damage probability calculation model for a fragment warhead to a target is as follows:

[0035] Step 1: Establish a 3D model of the target;

[0036] Establishing a three-dimensional model of the target mainly refers to equating and simplifying the main components of the target into surface units, which contain the coordinates of the quadrilateral nodes of each surface unit, as well as information such as the thickness and material of the surface unit, to form a digital model.

[0037] Step 2: Conduct vulnerability analysis on each component of the target;

[0038] The vulnerability analysis of each component of the target is mainly based on the theory described in "Target Vulnerability" (Beijing Institute of Technology Press, Li Xiangdong et al.). A vulnerability coefficient is set for each main component of the target to indicate the degree of damage to each component under a certain degree of fragment penetration or explosion shock wave. The value is equal to the ratio of the vulnerable area of ​​the component's outer surface to the presented area. The vulnerability coefficient is assigned to the face unit of each component in the three-dimensional model for the calculation of the damage probability in step 5. Thus, a three-dimensional target model with vulnerability information in the face unit is obtained, which is the basis for calculating the probability of damage to the face unit by fragments during the intersection of the projectile and the target.

[0039] Step 3: Establish a damage tree model to damage the target;

[0040] According to the Target Vulnerability, the degree of damage to a target is generally divided into three levels: M-level damage, F-level damage, and K-level damage;

[0041] M-level damage refers to mission-related damage, which means that the target functional components cannot fully function and maintenance personnel need 1 to 24 hours to eliminate the obstacles; F-level damage refers to fire control damage, which means that the target functional components lose their ability to function and require special maintenance units to spend 1 to 7 days and nights to eliminate the obstacles; K-level damage refers to annihilation damage, which means that the target weapon system loses its combat capability and cannot be repaired or it is not economically feasible to repair the damage;

[0042] Analyze the impact of the damage to each component on the function of the entire system. For example, destroying the radar antenna or shelter can lead to devastating damage to the air defense missile system. Finally, three levels of damage trees are formed, and the non-combination or non-redundant components and combination or redundant components in each level of damage tree are given;

[0043] Step 4: Establish a projectile-target intersection model between the warhead and the target;

[0044] Establish a projectile-target intersection model between the fragmentation warhead and the target, including the following contents:

[0045] 1) Establish a fragmentation field and overpressure field in the warhead coordinate system (field is a physics term that refers to the distribution of objects in space. The fragmentation field refers to the distribution of each fragment in space, including position and velocity vector information, etc. The overpressure field refers to the distribution of the explosion shock wave overpressure in space). In terms of the fragmentation field, the spatial distribution, mass distribution and velocity distribution of the fragments after the static explosion of the warhead are calculated based on the warhead charge, prefabricated fragments and shell structure, and the corresponding spatial analytical model is established to obtain a random fragmentation field in the warhead coordinate system, that is, the mass, shape coefficient and velocity vector of each fragment. In terms of the overpressure field, the peak overpressure Δp of the shock wave generated by the explosion of the fragment warhead is calculated. m And positive pressure specific impulse i:

[0046]

[0047]

[0048]

[0049]

[0050]

[0051] Where ω e is the equivalent charge; ω is the charge under shell charge; α is the charge coefficient; r0 / r f is the ratio of the initial radius of the shell to the crushing radius; γ is the polytropic index of the detonation product; Δp m It is the peak value of the shock wave generated when the charge explodes in an infinite air domain, in MPa; m eis the equivalent TNT equivalent, in kg; r is the distance from the explosion point to the detonation center, in m; C is a dimensionless constant, i is in Pa·s;

[0052] 2) Establish the intersection model between the fragmentation warhead and the target, mainly through the intersection relationship between the warhead and the target to establish the fragmentation field and overpressure field in the target coordinate system, and then determine whether each fragment in the fragmentation field intersects with each surface unit of the target, if they intersect, whether they can penetrate, and calculate whether the overpressure field can damage each specific surface unit of the target. The intersection relationship between the fragmentation field and the target is jointly determined by the terminal posture of the warhead and the terminal velocity direction of the warhead.

[0053] Step 5: Establish a damage probability calculation model for the target by the fragmentation warhead.

[0054] Calculating the probability of damage to the target by the fragmentation warhead includes the following steps:

[0055] 1) Calculate the probability of damage to target components by the fragmentation field:

[0056] Calculate the damage probability of each component of the target. Some components are considered as a whole. For such components, the sum of the number of fragments that penetrate all the face elements is used as the effective killing fragments of the entire component to calculate the damage probability of such components. Take all the fragments that penetrate the t-th component of this type of component as the effective damaging fragments of the component, and calculate the damage probability of the t-th component of this type of component according to the following formula:

[0057] p t =1-e -ξ·n

[0058] Where ζ is the vulnerability coefficient, n is the total number of fragments that penetrate the tth component;

[0059] Some components are calculated as non-redundant on each face, that is, as long as one face of the face unit that makes up this component is damaged, the entire component is considered damaged. Then the damage probability calculation formula for the sth component in this type of component is:

[0060]

[0061] Where p m is the damage probability of the mth surface unit in the sth component, and R is the total number of surface units in the sth component;

[0062] There are also types of components that have some redundant surface units, that is, if a certain percentage (e.g. 20%) of the total number of surface units are destroyed (the damage probability is 1), the component is considered to be destroyed (the damage probability is 1);

[0063] 2) Calculate the probability of damage to target components by overpressure field:

[0064] First, calculate the shock wave overpressure on a certain surface unit. If the shock wave overpressure is greater than the pressure tolerance threshold of the surface unit, the overpressure damage probability of the surface unit is considered to be 1.

[0065] The damage criterion of the shock wave is calculated according to the following formula:

[0066] (ΔP-P * )(II * )≥K

[0067] Where ΔP and I are the peak overpressure and specific impulse of the shock wave, respectively. * and I * are the critical overpressure and critical specific impulse of shock wave, respectively, and K is a constant, which depends on the vulnerability of the target. When the above formula is satisfied, it is considered that the target element has a 100% chance of being destroyed. The maximum value of the damage probability of the above fragment field and overpressure field to a certain element is taken as the final damage probability of the element, and then the damage probability of the component is further calculated according to the damage condition of the element with reference to the above section 1).

[0068] 3) Calculate the total damage probability of the system

[0069] According to the target vulnerability analysis theory, there are two damage calculation situations for the components that make up the target. One is that the target is damaged when at least one of the non-combined or non-redundant components is damaged, and the other is that the target is damaged when all the combined or redundant components are damaged. Then, based on the damage probability of each component of the target and the analysis results of the three-level damage tree, the total damage probability of the warhead to the target system can be calculated separately (divided into three levels and calculated separately), and the formula is:

[0070]

[0071] Among them, N or is the number of non-redundant components, k is the number of redundant component groups, P or,i is the damage probability of each non-redundant component, M1, M2, ..., M k are the number of redundant components in each redundant component group, P and,i is the damage probability of each component in the redundant component group.

[0072] Furthermore, the specific method of performing more than 10,000 repeated random calculations based on the Monte Carlo method for all random parameters that affect the calculation results is as follows:

[0073] Step 1: Determine the constants that need to be input when calculating the damage probability, that is, the fixed quantities that remain unchanged when performing random calculations;

[0074] The constants that need to be input when calculating the damage probability are mainly: the physical parameters of the warhead itself, such as the structure and number of prefabricated fragments, the equivalent TNT charge of the warhead, the warhead charge coefficient, etc.; the flight parameters of the warhead, such as the slip angle of the warhead, etc.; and the detection distance and detonation delay of the fuze.

[0075] Step 2: Determine the random parameters that need to be used for random calculation using the Monte Carlo method when calculating the probability of damage;

[0076] The random parameters that need to be randomly calculated using the Monte Carlo method when calculating the damage probability mainly include: the detonation position of the fragmentation warhead and the missile's entry angle, pitch angle, miss distance, terminal velocity, the angle between the terminal velocity vector and the ground, and the guidance accuracy probability error, that is, the circular probability error.

[0077] Step 3: Based on the damage probability calculation model, after determining the constants in step 1, the Monte Carlo method is used to perform more than 10,000 repeated random calculations for all random parameters in step 2, and finally a large number of damage probability calculation case data (more than 10,000 groups) can be obtained. In each calculation, each random parameter is randomly selected within a certain range according to a certain probability distribution. For example, the missile's relative angle to the target is randomly selected according to an average distribution within the range of 0° to 360°, the missile's ballistic angle is randomly selected according to a uniform distribution within the range of -10° to 0°, the missile's yaw angle is randomly selected according to a uniform distribution within the range of -2° to 2°, the missile's terminal velocity is randomly selected according to a uniform distribution between 200m / s and 300m / s, and the missile's vertical miss distance is randomly selected according to a normal distribution between -1m and 1m.

[0078] Furthermore, the specific method of using the machine learning neural network to train the large amount of damage probability calculation data obtained in the previous step to generate a damage probability prediction model is as follows:

[0079] Step 1: Determine the learning parameters when training data for a neural network using a machine learning method;

[0080] The learning parameters that need to be set when using machine learning methods to train data based on neural networks are: the maximum number of training times (parameters to ensure that training can be stopped, generally thousands to 10,000), the minimum mean square error between the predicted value and the target value, the algorithm used for training (gradient descent method, Newton algorithm and conjugate gradient method, etc.), the number of hidden layers (integers above 5), transfer functions (tansig and logsig, etc.) and the learning rate of neural network training (positive number below 1), etc. The above learning parameters are used to conduct subsequent example data training based on the neural network.

[0081] Step 2: Input more than 10,000 sets of damage probability case data obtained by random calculation in S201 into the machine learning neural network tool for training;

[0082] 1) Adjust the more than 10,000 sets of damage probability calculation example data obtained by random calculation in S201 into two parts: input and output, that is, all input parameter sets of the calculation example and the corresponding output K-level, F-level and M-level damage probability sets;

[0083] 2) Input the input data set and the output data set into various machine learning neural network tool software for training (K-level, F-level and M-level training respectively). The available tools include Matlab and Python, etc.

[0084] Step 3: Perform neural network training multiple times (dozens of times) on the damage probability calculation data, adjust the learning parameters, such as changing the number of hidden layers or the training algorithm, and find the damage probability prediction model with the highest accuracy (that is, the minimum mean square error between the predicted value and the target value of the neural network). This model can be used to predict the damage probability based on the values ​​of each random parameter, avoiding the huge amount of calculation in the damage probability calculation model.

[0085] By continuously using machine learning neural network training tools to train data and adjusting learning parameters, such as changing the number of hidden layers or training algorithms, we can eventually find a damage probability prediction model with the smallest mean square error between the predicted value and the target value. This model can be used to predict the corresponding K-level, F-level and M-level damage probabilities under different input conditions based on the values ​​of each parameter, avoiding the huge amount of calculation in the damage probability calculation model.

[0086] Furthermore, the damage probability prediction model obtained in S301 is used to adopt the Monte Carlo method to calculate the average single-shot damage probability of the warhead to the target under specific missile-target intersection conditions:

[0087] Step 1: Determine the projectile-target intersection conditions required for calculation according to user requirements;

[0088] The specific projectile-target intersection condition required for the calculation is determined by the actual needs of the user of this method. Under this condition, a part of the random parameters, such as the entry angle and pitch angle, are fixed according to physical reality or research needs, and the value selection method of other parameters remains unchanged to obtain the input data set corresponding to the projectile-target intersection condition.

[0089] Step 2: Using the obtained damage probability prediction model, the Monte Carlo method is used to perform more than 500 repeated random calculations on the input data set of step 1. After statistical data is collected, the average single-shot damage probability of the fragmentation warhead to the target under the missile-target intersection condition in step 1 is obtained. In each calculation, except for the specific missile-target intersection condition, the value selection method of other random parameters remains unchanged. In this way, the damage probability prediction model obtained by 10,000 repeated random calculations of the damage probability calculation model and then trained by the neural network replaces the original damage probability calculation model, and the calculation amount is very small, avoiding the huge calculation amount of the damage probability calculation model. Furthermore, by changing the missile-target intersection condition and its corresponding input data set (i.e., changing the fixed parameter type), the average single-shot damage probability of the warhead to the target under different conditions can be conveniently calculated.

[0090] The average single-shot damage probability obtained by Monte Carlo calculation using the damage probability prediction model of this method can achieve an average accuracy of more than 85% compared with the result obtained by directly performing Monte Carlo calculations using the damage probability calculation model. At the same time, this method significantly reduces the amount of calculation and improves the calculation efficiency by more than 60 times.

[0091] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solution of the present invention by using the above disclosed methods and technical contents without departing from the spirit and scope of the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall belong to the protection scope of the technical solution of the present invention. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0092] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A damage probability calculation method based on machine learning neural network, It is characterized in that include: Establish a three-dimensional model of the target, conduct vulnerability analysis on the components of the target, and establish a calculation model for the probability of damage to the target by the fragmentation warhead; According to the damage probability calculation model, for all random parameters that affect the calculation results, repeated random calculations based on the Monte Carlo method are carried out for more than m times to obtain a large number of damage probability calculation case data for different damage levels, where m is greater than 10000; The above-mentioned large amount of damage probability calculation example data is trained by using a neural network in machine learning to obtain a damage probability prediction model for different damage levels, wherein the neural network includes: an input layer, a hidden layer, and an output layer; the number of nodes in the hidden layer is 75% of the number of nodes in the input layer; and the number of hidden layers is an integer greater than 5; Using the damage probability prediction model obtained in the previous step, under the input specific projectile-target intersection conditions, a part of the random parameters are fixed, and the Monte Carlo method is used to carry out more than n repeated random calculations based on the damage probability prediction model. After statistical data, the average single-shot damage probability of the fragmentation warhead to the target under the specific projectile-target intersection conditions is obtained, and the n is greater than 500.

2. According to claim 1, a method for calculating damage probability based on machine learning neural network, It is characterized in that The damage probability calculation model is specifically: Among them, N or is the number of non-redundant components, k is the number of redundant component groups, P or,i is the damage probability of each non-redundant component; M 1 , M 2 , …, M k are the number of redundant components in each redundant component group; P and,j is the damage probability of each component in the redundant component group; the components of the target include multiple groups of redundant components and multiple non-redundant components. If all the redundant components in any one group of redundant components are damaged, the target is considered to be damaged; if any one non-redundant component is damaged, the target is considered to be damaged.

3. The damage probability calculation method based on machine learning neural network according to claim 1, It is characterized in that The damage levels include: M-level mission damage, F-level fire control damage and K-level catastrophic damage; M-level mission damage means that the target functional components cannot fully perform their functions and it takes 1 to 24 hours to eliminate the obstacles; F-level fire control damage means that the target functional components lose their ability to function and it takes 1 to 7 days and nights to eliminate the obstacles; K-level catastrophic damage means that the target weapon system loses its combat capability and cannot be repaired or it is not economically feasible to repair the damage.

4. The damage probability calculation method based on machine learning neural network according to claim 1, It is characterized in that The random parameters include: terminal flight speed of the fragment warhead, entry angle, pitch angle, miss distance, and speed and flying direction of each fragment.

5. The damage probability calculation method based on machine learning neural network according to claim 1, It is characterized in that The input data set of the input layer includes: a training set, a validation set and a test set; the training set is used to train the model, the validation set is used to verify and adjust parameters of the trained damage probability prediction model, and the test set is used to evaluate the performance of the damage probability prediction model.

6. The damage probability calculation method based on machine learning neural network according to claim 5, It is characterized in that The data volume ratio of the training set, validation set and test set is 70%:15%:15%.

7. The damage probability calculation method based on machine learning neural network according to claim 6, It is characterized in that The algorithm used in the neural network training is the gradient descent method, Newton algorithm or conjugate gradient method.

Images