Optimization method of target damage probability

Abstract

The invention relates to a target damage probability optimization method, belongs to the technical field of damage probability prediction optimization, and solves the problems that a damage probability calculation method in the prior art is complex in working condition parameter adjustment and cannot quickly and accurately meet the requirements of combat missions. Weapon parameters and target parameters are obtained, and a target damage probability prediction model is constructed based on the weapon parameters and the target parameters; obtaining to-be-tested working condition parameters, and inputting the to-be-tested working condition parameters into the target damage probability prediction model to obtain damage probabilities of the working condition parameters to a target at different positions; and judging whether the damage probability of the to-be-tested working condition parameter to the specified position meets the requirement or not based on the combat mission requirement, if so, taking the to-be-tested working condition parameter as a combat parameter, and if not, optimizing the to-be-tested working condition parameter by adopting a particle swarm algorithm, and taking the optimized parameter as the combat parameter. According to the optimization method, the working condition parameters can be rapidly and accurately corrected so as to meet the requirements of combat missions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of damage probability prediction optimization, in particular to a target damage probability optimization method. BACKGROUND

[0002] Kill Matrix is a tool for evaluating the damage effectiveness of weapon systems on specific targets, which realizes target damage probability evaluation through quantitative analysis of the damage effect that weapons may cause after hitting targets. Each cell in the Kill Matrix represents a part of the target, and the damage probability of the cell can be calculated by experience formula method, Monte Carlo simulation method, Bayesian network method, and equipment damage evaluation method based on fragment effect.

[0003] However, the above traditional calculation methods have the following disadvantages: for the experience formula method, it depends on the accuracy and relevance of historical data, and may not be applicable to new weapon systems or new targets, and is difficult to adapt to the rapid changes of the battlefield environment; for the Monte Carlo simulation method, the calculation amount is large, the time is long, a large number of simulations are needed to obtain relatively accurate results, and the requirement for computing resources is high; for the Bayesian network method, professional knowledge is needed to construct the Bayesian network, and the selection of network structure and parameters is sensitive, and the calculation process is relatively complex; for the equipment damage evaluation method based on fragment effect, it does not consider the details such as the angle of attack, the angle of attack, and the angular velocity of the fragments, which may affect the accuracy of the evaluation. In addition, for the above traditional calculation methods, only the damage probability of each cell (i.e. different positions) corresponding to a certain input working condition parameter (e.g. weapon parameter) can be obtained, if the current combat plan requires to cause a preset degree of damage to the specified position, and the current working condition parameter cannot meet the requirement, the working condition parameter needs to be adjusted and calculated again until the requirement is met, but this correction method can only rely on manual experience, and each correction needs to be recalculated, and multiple adjustments are needed to meet the requirements, which also has the problems of large calculation amount and low calculation efficiency.

[0004] Therefore, it is necessary to provide a damage probability correction method which can quickly and accurately correct the working condition parameters to meet the requirements of combat tasks. SUMMARY

[0005] In view of the above analysis, the embodiments of the present application aim to provide a target damage probability optimization method to solve the problem that the working condition parameter adjustment of the existing damage probability calculation method is complex and cannot quickly and accurately meet the requirements of combat tasks.

[0006] The embodiments of the present application provide a target damage probability optimization method, which comprises:

[0007] obtaining weapon parameters and target parameters, and constructing a target damage probability prediction model based on the weapon parameters and the target parameters;

[0008] obtaining a to-be-tested working condition parameter, inputting the to-be-tested working condition parameter into the target damage probability prediction model, and obtaining a damage probability of the working condition parameter at different positions on the target;

[0009] judging whether the damage probability of the to-be-tested working condition parameter at a specified position meets a requirement based on a combat task requirement, if yes, taking the to-be-tested working condition parameter as a combat parameter, and if no, optimizing the to-be-tested working condition parameter by using a particle swarm algorithm and taking the optimized parameter as the combat parameter;

[0010] In the optimization process, inputting an input parameter represented by each particle into the target damage probability prediction model to obtain a damage probability of the particle at the specified position, and obtaining a fitness function based on the damage probability of the particle at the specified position and a working condition parameter corresponding to the particle.

[0011] In further improvement of the above method, the weapon parameters include a type, performance parameters and a warhead type of the weapon, and the target parameters include a type, an environment and a position of the target.

[0012] In further improvement of the above method, the constructing of the target damage probability prediction model based on the weapon parameters and the target parameters includes: setting a calculation range of a weapon parameter in a damage matrix based on the type of the weapon and the performance parameters, the calculation range of the weapon parameter including a height range, a speed range, a pitch angle range, an azimuth angle range and a damage level range; setting a plurality of working condition parameters based on the calculation range, calculating a damage probability matrix of the warhead on the target and a shock wave damage probability matrix of the type of the weapon on the target corresponding to each working condition parameter, and calculating a joint damage probability matrix based on the damage probability matrix and the shock wave damage probability matrix; performing conversion and reorganization processing on the joint damage probability matrix corresponding to each working condition parameter to form a sample data set; and training a two-stage network model based on the sample data set to obtain the target damage probability prediction model.

[0013] In further improvement of the above method, the optimizing of the to-be-tested working condition parameter by using the particle swarm algorithm and the taking of the optimized parameter as the combat parameter include:

[0014] A1: initializing a particle swarm, selecting samples meeting a combat task requirement from the sample data set to form a first sample data set, taking a working condition parameter of each data in the first sample data set as a particle, taking a working condition parameter in the to-be-tested working condition parameter as a particle, judging whether a total number of the current particles meets a preset number, if yes, performing A3, and if no, performing A2.

[0015] A2: statistics of the first sample data set each parameter range, based on a random value generated by an algorithm in each parameter range, so as to generate a plurality of particles, so that the total number of particles to meet the requirements of the pre-set number;

[0016] A3: calculate the fitness function of each particle, to update the individual optimal and global optimal value, to determine whether to meet the end condition, if satisfied, then execute A5, if not satisfied, then execute A4;

[0017] A4: update the speed and position of each particle, return to A3;

[0018] A5: output the optimal position, the particle group parameters corresponding to the optimal position are the combat parameters.

[0019] Further improvement based on the above method, the fitness function is calculated based on the working condition parameters corresponding to the particle, comprising:

[0020]

[0021] Wherein, op' is the damage probability of the particle at the specified position, is the damage probability at the specified position required by the combat task, h', v', ω1', ω2' are the height, speed, pitch angle, azimuth angle corresponding to the particle, is the height, speed, pitch angle, azimuth angle in the working condition parameters to be tested, a, b, c, d are constants.

[0022] Further improvement based on the above method, for step A4, the speed of each particle is updated by the following way:

[0023] v(t+1) = ωv(t) + c1r1(p-x(t)) + c2r2(g-x(t)),

[0024]

[0025] The position of each particle is updated by the following way:

[0026] x(t+1) = x(t) + v(t+1),

[0027] Wherein, c1, c2 are learning factors, r1, r2 are random numbers, v(t+1) is the updated speed of the particle, v(t) is the current speed of the particle, p is the individual optimal position of the particle, g is the global optimal position of the whole group, x(t) is the current position of the particle, x(t+1) is the updated position of the particle, ω is the adaptive inertia weight, ω min is the minimum value of the inertia weight, ω maxis the maximum value of the inertia weight, d and k are adjustment coefficients, t is the current iteration number, M is the maximum iteration number, op best is the maximum value of the damage probability of the particles in the current population at the specified position, op ave is the average value of the damage probability of the particles in the current population at the specified position.

[0028] Further improvement based on the above method, the damage probability matrix of the warhead corresponding to each working condition parameter and the shock wave damage probability matrix of the weapon of this type to the target are calculated, including: calculating the fragment power field data corresponding to each working condition parameter, the fragment power field data including: fragment mass, speed, fragment area solid angle in different angular domains; calculating the damage probability of the warhead of this type to the target at different positions in the kill matrix based on the fragment power field data, and the damage probability of each position forming a damage probability matrix; calculating the shock wave damage probability of the weapon of this type to the target at the corresponding position in the kill matrix based on the shock wave radius of the weapon at different positions in the kill matrix, and the shock wave damage probability of each position forming a shock wave damage probability matrix.

[0029] Further improvement based on the above method, the damage probability of the warhead of this type to the target at different positions in the kill matrix based on the fragment power field data is calculated, including: establishing a target coordinate system with the position of the target as the origin; grouping the fragments in different angular domains according to mass to form a plurality of fragment mass groups, and the mass of each fragment in each fragment mass group is the same; for each position, the following operations are performed: placing the warhead at the corresponding position in the target coordinate system according to the weapon height defined in the current working condition parameter, and using the firing line analysis method to obtain the vulnerable area and exposure probability of the target at this position for each angular domain; calculating the damage probability of the warhead of this type to this position based on the vulnerable area, the exposure probability and each fragment mass group.

[0030] Further improvement based on the above method, the joint damage probability matrix corresponding to each working condition parameter is converted and reorganized to form a sample data set, including: for each working condition parameter, M*N training sample data can be constructed, each training sample data is [h, v, α, θ, x, y, l, p], wherein M and N are the dimensions of the joint damage probability matrix, h is the height of the warhead, v is the speed of the warhead, α is the azimuth angle of the warhead, θ is the elevation angle of the warhead, x is the horizontal coordinate of the warhead of this type in the target coordinate system, y is the vertical coordinate of the warhead of this type in the target coordinate system, l is the kill level of the warhead, and p is the joint damage probability of the warhead at (x, y).

[0031] Based on the further improvement of the above method, the two-stage network model includes: a one-stage network model, the input data of the one-stage network model is the working condition parameters [h, v, α, θ] and the position information [x i ,y j ], the output data is corresponding to each [x i ,y j ] one-stage predicted damage probability p′, i=1,2,3,...,M,j=1,2,3,...,N; matrix generation model, all [x i ,y j ] The corresponding one-stage predicted damage probabilities p′ are combined to form a one-stage damage probability prediction matrix; a two-stage network model, the input data of the two-stage network model is the one-stage damage probability prediction matrix, the output data is the two-stage damage probability prediction matrix, and the two-stage damage prediction probability at a specified position in the two-stage damage probability prediction matrix is ​​used as the final damage probability; wherein, the one-stage network model is a one-dimensional convolutional neural network, the two-stage network model is a convolutional network based on the attention mechanism, and the loss function of the two-stage network model is:

[0032]

[0033] Among them, w1 and w2 are constants, y x，y is the damage probability value of the warhead at (x, y) in the sample data set, is the predicted value of the two-stage damage probability at (x, y) in the two-stage damage probability prediction matrix.

[0034] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0035] 1. The present invention provides a method for optimizing the target damage probability. Based on a target damage probability prediction model, the kill probability of a certain operating condition parameter at different positions is obtained, and whether the damage probability of the current operating condition parameter to the target at the specified position meets the combat mission requirements is determined. If not, the particle swarm algorithm is used to find the optimal parameter corresponding to the current operating condition parameter. The current operating condition parameter can be corrected quickly and accurately, thereby providing technical guidance to designers when designing the operating condition parameters. In addition, when using the particle swarm algorithm, a method for adaptively adjusting the inertia weight is provided, which can more quickly and accurately obtain the damage probability of the specified position that meets the combat mission requirements.

[0036] 2. The application provides a target damage probability optimization method, studies the damage probability of a specific weapon to a target, expands the training sample set by constructing sample data under different working conditions, and can further improve the accuracy of the prediction model; the neural network prediction method can effectively extract the feature representation under different altitudes, azimuth angles, pitch angles and bullet speeds, solve the complex nonlinear mapping problem from input to output, obtain the damage matrix of the target within 100 meters, and finally realize the rapid prediction of the damage probability of the specific weapon to the specific target under different parameter conditions, solve the problem that the nonlinear relationship between various factors in the killing matrix cannot be accurately expressed in the prior art, and overcome the defects of large calculation amount and low accuracy of the existing traditional calculation method.

[0037] In the application, the above technical solutions can be combined with each other to realize more preferred combination solutions. Other features and advantages of the application will be described in the subsequent specification, and some advantages will become apparent from the specification or will be understood by implementing the application. The purposes and other advantages of the application can be realized and obtained from the contents specifically indicated in the specification and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0038] The accompanying drawings are included to provide a further understanding of the application and are incorporated herein and constitute a part of the application. The drawings illustrate embodiments of the application and, together with the description, serve to explain the principles of the application. In the drawings:

[0039] Figure 1 An example diagram of a target damage probability optimization method in an embodiment of the application;

[0040] Figure 2 An example diagram of a pitch angle and an azimuth angle in an embodiment of the application;

[0041] Figure 3 A side view of a fragment dispersion angle domain in an embodiment of the application;

[0042] Figure 4 An example diagram of a stereographic angle distribution of a fragment range in an embodiment of the application;

[0043] Figure 5 An example diagram of a shooting line grid of a target model in an embodiment of the application;

[0044] Figure 6 An example diagram of a relationship between a shock wave radius and a shock wave damage probability in an embodiment of the application;

[0045] Figure 7 An example diagram of a two-stage network model in an embodiment of the application. DETAILED DESCRIPTION

[0046] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0047] A specific embodiment of the present invention discloses a method for optimizing the target damage probability, such as Figure 1 As shown, including:

[0048] S1: Acquire weapon parameters and target parameters, and construct a target damage probability prediction model based on the weapon parameters and the target parameters.

[0049] The parameters of weapons and targets are set based on combat missions, wherein weapon parameters include: the type of weapon, performance parameters and warhead type, and target parameters include: the type of target, environment and position. Specifically, the types of weapons include missiles, artillery shells, explosives, etc. After determining the type of weapon, the performance parameters of the weapon can be determined, including range, accuracy, explosive power, etc., and then the warhead type of the weapon, such as high-explosive bombs, armor-piercing bombs, cluster bombs, etc., are analyzed to finally determine the mechanism of action of the weapon on the target (i.e., the damage mechanism of the target is studied based on the damage mechanism of fragmentation and shock wave damage to the target). The target in the present invention refers to the specific target to be hit, and the types of targets include: buildings, vehicles, clusters of people, etc. After determining the target, the environment of the target is studied, such as the terrain, climate, whether there are shelters, etc. of the target location, the location of the target, i.e. the specific coordinates of the target location, and the attributes of the target, such as parameters such as target size, structure, and material.

[0050] The constructing of a target damage probability prediction model based on the weapon parameters and the target parameters includes:

[0051] S11: Setting a calculation range of weapon parameters in a kill matrix based on the type of the weapon and the performance parameters, wherein the calculation range of the weapon parameters includes: an altitude range, a speed range, a pitch angle range, an azimuth angle range, and a kill level range.

[0052] Specifically, for the altitude parameter, it refers to the vertical distance between the weapon and the local horizontal ground, for example, the vertical distance between a missile in flight and the local horizontal ground. For the speed parameter, it refers to the flight speed or travel speed of the weapon, for example, the flight speed of a missile. For the pitch angle parameter, it refers to the angle between the longitudinal axis of the weapon and the horizontal ground. For the azimuth parameter, it refers to the angle between the weapon and a fixed direction (usually due north). See Figure 2 , which gives the positional relationship between the pitch angle and azimuth angle in the target coordinate system. As for the lethality parameter, it is used to evaluate the impact of the weapon system on the target. According to different classification standards, the lethality can be divided into different categories.

[0053] In the present invention, the damage levels are mainly divided into the following three categories: light damage, moderate damage, and severe damage. Light damage generally means the damage is relatively low, and after light damage is inflicted, the target can still continue to perform its basic tasks. Moderate damage means that after moderate damage is inflicted, the target's basic functions are significantly reduced, and it cannot function normally for a long period of time, requiring repair. Moderate damage can achieve the purpose of affecting the target's mission performance. Severe damage means that after severe damage is inflicted, the target's main parts are severely damaged, and it loses basic functions, making it unable to continue to perform other tasks.

[0054] The present invention is described by taking a certain explosive bomb and a certain light armored vehicle as examples. Figure 2 The geometric relationship between the weapon and target interaction in Figure 1 represents the working condition of an airburst weapon, which propagates a blast wave and many fragments from a detonation point. The ground zero point is defined as point O, which is directly below the center of mass of the warhead. It is the origin of the ground plane coordinate system. Assuming that the weapon's velocity vector coincides with the weapon axis, the weapon axis is extended to the ground plane to obtain point R. The line OR defines the x-axis (or longitudinal axis), and the y-axis or transverse axis is perpendicular to OR. The target, represented by the vulnerable center, is located at the ground plane coordinate point (x, y) relative to the ground zero point of the weapon detonation point. As can be seen from the figure, the target is located at a specific deflection angle of the weapon, and the weapon is located at a specific elevation angle of the target.

[0055] After determining the specific weapon to strike the target in step S11, the calculation range of the weapon parameters in the damage matrix can be set based on the weapon type and performance parameters. For example, the weapon speed range can be set to 100m / s to 500m / s, and the damage level range can be set to moderate damage and severe damage. For different combat missions, researchers can flexibly set the parameter range to meet the combat mission requirements.

[0056] S12: multiple operating condition parameters are set based on the calculation range, and the damage probability matrix of the warhead to the target and the shock wave destruction probability matrix of the type of weapon to the target corresponding to each operating condition parameter are calculated, and a joint damage probability matrix is ​​calculated based on the damage probability matrix and the shock wave destruction probability matrix.

[0057] After determining the valid range of each parameter in step S11, a random generation algorithm can be used to generate multiple sets of operating parameters within the valid range. It is understood that the more operating parameters generated, the more sample data ultimately used to train the two-stage network model, thereby further improving the model's prediction accuracy. Furthermore, using a random generation algorithm to generate multiple sets of data is also a common practice in the art and is not limited to this in the present invention; as long as a predetermined number of operating parameters can be generated, the results will be sufficient.

[0058] The damage probability matrix of the warhead on the target and the shock wave damage probability matrix of the weapon on the target corresponding to each working condition parameter are calculated, including:

[0059] S121: Calculate the fragment power field data corresponding to each working condition parameter, and the fragment power field data includes: fragment mass, speed, and fragment area solid angle in different angle domains.

[0060] The fragment power field data refers to various parameters and characteristics of fragments when the warhead explodes, including the number, speed, scattering angle, scattering direction angle, density distribution, and fragment perforation area of the fragments, which are crucial for evaluating the killing power and damage effect of the warhead. The present application limits the fragment power field data to the characterization data of the fragment power, such as the maximum mass, minimum mass, average mass, maximum speed, minimum speed, and average speed of the fragments.

[0061] For example, the present application divides the range of fragments generated by the warhead into 16 angle domains, which are referred to as scattering angles, ranging from [θ min , θ max ], and are divided by 10° intervals. The fragments are scattered from each thin slice on the warhead, rather than from the center of mass, as shown in Figure 3 . In three-dimensional space, the fragments in these conical regions form a distribution as shown in Figure 4 after scattering to the ground. The fragments are scattered along the circumference, ranging from 0° to 360°, Figure 4 The solid angle of the fragment range is denoted as φ, and according to the range of the angle domain, the maximum solid angle is denoted as φ max , and the minimum solid angle is denoted as φ min In each scattering angle interval, fragments of different mass and speed are randomly generated to represent the fragment power field generated by the weapon explosion.

[0062] The present application develops a visualization program for fragment power field calculation based on C++ development language. The input parameters are a certain working condition parameter, and the output parameters are the fragment mass, speed, and fragment area solid angle in different angle domains in the fragment power field. The business logic and calculation method of the visualization program can be adjusted in real time according to the combat plan, and are implemented using existing technology, which will not be described here.

[0063] S122: Calculate the damage probability of the type of warhead on the target at different positions in the damage matrix based on the fragment power field data, and the damage probability of each position constitutes a damage probability matrix, including:

[0064] A1: Establish a target coordinate system with the position of the target as the origin; group the fragments in different angle domains according to mass to form multiple fragment mass groups, and the mass of each fragment in each fragment mass group is the same.

[0065] See also Figure 2 , the target is equivalent to a point, recorded as the vulnerable center, and the coordinate system is established with it as the origin.

[0066] A2: Perform the following operations for each position:

[0067] A21: Place the warhead at the corresponding position in the target coordinate system based on the weapon height specified in the current operating parameters, and use the firing line analysis method for each angular region to obtain the vulnerable area and exposure probability of the target at that position.

[0068] For the fragments in each angular region divided in step S121, the target's vulnerable area and exposure probability data are calculated using a firing line analysis method.

[0069] The basic structure of the vulnerable area is the firing line module. The working method is as follows: for a fragment with a given flight direction, a grid of the same size is used to cover the target geometry. Each grid contains a firing line. The spacing between the grid lines can be determined by yourself, but the smaller the spacing, the longer the calculation time required. Assuming that the weapon explodes at a certain distance from the target, the flight paths of the individual fragments are parallel. Assuming that the firing line passes through the center of each grid or cell, the target geometry is transformed into Figure 5 The view shown can determine which component is inclined to which firing line. In this way, it can be determined which components (critical and non-critical) the fragments flying along the firing line encounter, and the order of encounter. Then, the cumulative calculation of the kill probability of each firing line is performed. The target vulnerable area is the sum of the vulnerable areas of all individual cells containing the target part, and is only valid for this azimuth and elevation angle. The analysis of the firing line is based on the ability of the fragments to penetrate the components. The FATEPEN model is used to calculate the firing line corresponding to the fragment (i.e., the trajectory of the fragment), and the number of cells covered by the firing line is counted, recorded as P k / sh , then the cell firing line vulnerable area can be calculated using the following equation:

[0070] A v-cell =P k / sh ×A p-cell ,

[0071] Where A p-cell Indicates the area of ​​the target in the cell. Here, the area of ​​the cell is used as A p-cell The value of , then, will contain the target of all p cells of A p-cell Adding them together, we can get the vulnerable area of ​​the target.

[0072]

[0073] In the shooting line method, exposure probability usually refers to the possibility of a target being hit under specific conditions. The calculation of this probability usually involves multiple factors such as the target's geometric shape, position, motion state, the shooter's position, and weapon performance. The general steps for calculating exposure probability include: (1) Target model establishment: First, it is necessary to establish a geometric model of the target, including information such as the target's shape, size, and position; (2) Shooting line generation: Based on the shooter's position and the weapon's performance, possible shooting lines are generated; (3) Hit judgment: Determine whether each shooting line intersects with the target, that is, whether it can hit the target; (4) Probability calculation: Calculate the exposure probability based on the ratio of the number of shooting lines that hit to the total number of shooting lines.

[0074] Based on the general process steps of exposure probability, the present invention developed a C++-based exposure probability calculation software, and the calculation formula of the exposure probability is:

[0075] P xpose =f w *g t *h e ,

[0076] Where, f w is the coverage probability of the power generated by the weapon (such as hit accuracy, etc.), g t is the geometric factors of target exposure (such as area, arrangement, etc.), h e is the environmental correction factor (such as terrain shielding, motion complexity, etc.).

[0077] A22: Calculate the damage probability of the warhead of the type at the location based on the vulnerable area, the exposure probability, and each fragment mass group, including:

[0078] B1: Calculate the damage probability of each fragment mass group at the location:

[0079]

[0080] Wherein, i is the number of the fragment mass group, A Vi is the target vulnerable area corresponding to the i-th fragment mass group, PE i is the target exposure probability corresponding to the i-th fragment mass group, K i is the number of fragments in the i-th fragment mass group, SR is the shortest distance from the warhead to the target to be destroyed ( Figure 2 ), Ω is the solid angle of the fragment area;

[0081] B2: Calculate the damage probability of this type of warhead at this location based on the damage probability of all fragment mass groups at that location:

[0082]

[0083] wherein i = 1, 2, 3,..., N, N is the number of fragment mass groups.

[0084] A3: Obtain the damage probability at each position, thereby forming a damage probability matrix.

[0085] S123: Calculate the blast damage probability of the type of weapon to the target at the corresponding position in the damage matrix based on the blast radius of the type of weapon at different positions in the damage matrix, and the blast damage probability at each position forms a blast damage probability matrix.

[0086] Calculate the blast damage probability P K-BLAST (SR) of the type of weapon at different positions in the damage matrix.

[0087]

[0088] wherein SR is the blast radius, RB1 and RB2 are constants, as shown in Figure 6 , for different SR, the blast damage probability is different. Combine the blast damage probability corresponding to each position to form a blast damage probability matrix.

[0089] wherein the joint damage probability matrix is calculated based on the damage probability matrix and the blast damage probability matrix, including:

[0090] Calculate the joint damage probability based on the damage probability at each position in the damage probability matrix and the blast damage probability at the position in the blast damage probability matrix, and the joint damage probability at all positions forms a joint damage probability matrix. Specifically, the joint damage probability at each position is:

[0091] P s = 1-[(1-P K-BLAST (SR))(1-P K-FRAG )].

[0092] Combine the joint damage probability at each position to form a joint damage probability matrix.

[0093] S13: Convert and reorganize the joint damage probability matrix corresponding to each working condition parameter to form a sample data set.

[0094] For each type of operating parameter, M*N target damage probability prediction model training sample data can be constructed, each training sample data is [h, v, a, q, x, y, 1, p], wherein M and N are the dimensions of the joint damage probability matrix, h is the height of the warhead, v is the speed of the warhead, a is the azimuth angle of the warhead, q is the pitch angle of the warhead, x is the horizontal coordinate of the warhead in the target coordinate system, y is the vertical coordinate of the warhead in the target coordinate system, 1 is the damage level of the warhead, and p is the damage probability of the warhead at (x, y).

[0095] For each type of operating parameter, the weapon can be placed in the specified position of the damage matrix to obtain the damage matrix based on the corresponding damage level of the specific weapon-target under the operating condition. Different damage levels of the damage matrix result in different target vulnerable areas and different damage probabilities. Therefore, for a target coordinate system and an M*N damage matrix, one operating parameter corresponds to M*N joint damage probabilities, that is, M*N training samples. Exemplarily, the damage matrix data has a dimension of 20*20, and the range of 100 meters before and after the target is predicted. The interval between different damage probabilities in the matrix is five meters, and finally a two-dimensional matrix of [20, 20] is formed, which can be divided into 400 data, that is, corresponding to 400 training samples.

[0096] Preferably, each piece of training sample data is preprocessed, including data cleaning for handling missing values and outliers. For missing values, common processing methods include filling missing values and deleting records containing missing values. For outliers, threshold setting can be used for screening or statistical methods can be used for processing.

[0097] Data integration refers to the process of integrating data from multiple data sources into a data set. In the data integration process, problems such as data matching, duplicate values, and inconsistent data types need to be solved.

[0098] Data transformation refers to converting data to adapt to specific analysis needs. Common transformation methods include feature scaling, discretization, and encoding. Data reduction is to reduce the size of data by reducing the complexity of data while maintaining the original data relationship. Common reduction methods include feature selection and feature construction.

[0099] Data verification and quality check, when data preprocessing, the integrity and consistency of the data need to be verified to ensure that the data meets the expected format and range. Duplicate data and redundant data are processed to improve data quality.

[0100] The data preprocessing is to further improve the data quality, and the technical personnel can adopt the corresponding preprocessing method to preprocess each piece of data according to the needs, and the specific implementation steps of the data preprocessing are not limited in the application.

[0101] Preferably, the kill level is One-Hot encoded as a neural network input parameter to eliminate the sequence relationship.

[0102] Exemplarily, 20000 pieces of kill matrix data csv files are constructed in the application, wherein the names of the csv files respectively represent id sequence number (no actual meaning), kill level, height, azimuth angle, pitch angle and speed of the projectile. Each of the csv files includes a 20*20 kill matrix, and each value in the matrix is a damage probability.

[0103] S14: training a two-stage network model based on the sample data set to obtain a target damage probability prediction model.

[0104] The training sample data set is divided into a training set, a test set and a validation set according to a certain proportion, wherein the training set is used for preliminary learning of the parameters of the convolutional neural network model, the test set is used for further optimization and adjustment of the parameters, and the validation set is used for evaluating the performance and generalization ability of the model. Exemplarily, the proportion can be 7:2:1. The training process of the two-stage network model is a conventional means in the art, and will not be described here.

[0105] The two-stage network model comprises:

[0106] A one-stage network model, wherein the input data of the one-stage network model is the working condition parameters [h, v, a, q] and the position information [x i , y j ] in the kill matrix, and the output data is the predicted damage probability p' corresponding to each [x i , y j ], i=1, 2, 3,..., M, J=1, 2, 3,..., N;

[0107] A matrix generation model, wherein the predicted damage probabilities p' corresponding to all [x i , y j ] are combined to form a one-stage predicted damage matrix;

[0108] A two-stage network model, wherein the input data of the two-stage network model is the one-stage predicted damage matrix, and the output data is a two-stage predicted damage matrix, and the two-stage predicted damage matrix is taken as a final predicted damage matrix;

[0109] The one-stage network model is a one-dimensional convolutional neural network, and the loss function of the one-stage is the mean square error (MSE) between the predicted probability of a certain position and the actual damage probability of the position (x, y position) in the real damage matrix.

[0110]

[0111] wherein w1 and w2 are constants, y x，y is the damage probability of the warhead at (x, y) in the damage matrix training sample data set, is the two-stage predicted damage probability at (x, y) in the two-stage predicted damage matrix of the two-stage network model.

[0112] Figure 7 An example diagram of the two-stage network model is given, the one-stage network model is a convolutional neural network model, which includes an input layer, a convolutional layer, a normalization layer, a pooling layer, and a full connection layer, and adopts a Relu activation function, and the data sample can be input in a batch mode. The two-stage network model is a convolutional network based on an attention mechanism, and the present application does not make specific limitations on the two-stage network model, as long as the designed network can complete the damage matrix prediction. Referring to Figure 7 For the two-stage network model, the feature matrix is subjected to MSE constraint with the real damage matrix to obtain Loss_1; the one-dimensional vector after one-dimension of the feature matrix is subjected to two full connection layers to obtain a predicted performance factor (the performance factor is a performance index used to measure the damage performance of a specific weapon on a specific target), and the predicted performance factor is subjected to cross entropy loss function (cross entropy) constraint with the actual performance factor to obtain Loss_2. Finally, the two-stage model training total loss function LOSS2 is obtained after weighted average.

[0113] Understandably, the technical scheme of the present application can finally obtain a target damage probability prediction model for different targets when applied to different combat tasks.

[0114] Further, after obtaining the two-stage network model, a pruning algorithm can be further used to optimize the model, so as to reduce the calculation amount and memory requirement, and improve the model efficiency and generalization ability. Specifically, the importance score of each hidden layer in the two-stage network model is calculated, for each hidden layer, the weight threshold of each neuron is calculated based on the importance score and a custom weight, the neurons with a weight lower than the weight threshold are removed, and the pruned network is fine-tuned to ensure that the prediction performance is not affected. Wherein, the threshold of each neuron is evaluated based on the importance score and the custom weight, including:

[0115]

[0116] wherein,

[0117]

[0118] w ij is the weight of the jth neuron in the ith hidden layer, and i is the importance score of the ith hidden layer, and s is the standard weight threshold of the ith hidden layer, and avg is the average weight of all neurons in the ith hidden layer, and i 0 < s < 1, 0 < s < 1.

[0119] S2: Obtain a to-be-tested working condition parameter, input the to-be-tested working condition parameter into the target damage probability prediction model, and obtain the damage probability of the target at different positions under the working condition parameter.

[0120] Based on the combat task, the working condition parameter is designed as a to-be-tested working condition parameter, which is input into the trained target damage probability prediction model as input data, and the damage probability of the target caused by the working condition parameter at different positions can be obtained.

[0121] S3: Determine whether the damage probability of the to-be-tested working condition parameter at the specified position meets the requirements based on the combat task requirements. If it meets the requirements, the to-be-tested working condition parameter is used as a combat parameter. If it does not meet the requirements, the particle swarm optimization algorithm is used to optimize the to-be-tested working condition parameter, and the optimized parameter is used as a combat parameter.

[0122] In the optimization process, the input parameter represented by each particle is input into the target damage probability prediction model to obtain the damage probability of the particle at the specified position, and the fitness function is obtained based on the damage probability of the particle at the specified position and the working condition parameter corresponding to the particle.

[0123] After obtaining the damage probability of the to-be-tested working condition parameter at each position, it is determined whether the working condition parameter can cause the expected damage effect at the specified position based on the combat task. If it can, the set of working condition parameters is used as a strike parameter. If it cannot, the optimized particle swarm optimization algorithm is used to adjust the to-be-tested working condition parameter so that it can cause the expected damage at the specified position.

[0124] The fitness function is calculated based on the working condition parameter corresponding to the particle, which includes:

[0125]

[0126] wherein, op' is the damage probability of the particle at the specified position, h', v', ω1', ω2' are height, speed, pitch angle, azimuth angle corresponding to the particle, h, v, ω1, ω2 are height, speed, pitch angle, azimuth angle in the to-be-tested working condition parameter,

[0127] The particle swarm optimization algorithm is used to optimize the to-be-tested working condition parameter, and the optimized parameter is used as the combat parameter.

[0128] C1: initializing the particle swarm, filtering out samples corresponding to the combat task requirement from the sample data set to form a first sample data set, taking the working condition parameter of each data in the first sample data set as a particle, taking the working condition parameter in the to-be-tested working condition parameter as a particle, judging whether the total number of the current particles meets the preset number, if yes, executing C3, if not, executing C2.

[0129] C2: statistically determining the range of each parameter of the first sample data set, and randomly taking values in the range of each parameter based on a random generation algorithm, thereby generating a plurality of particles, so that the total number of the particles meets the preset number requirement. As described above, the random generation algorithm is a conventional means in the art, and the present application does not limit it here, as long as it can generate a preset number of required sample data.

[0130] C3: calculating the fitness function of each particle to update the individual optimal and global optimal values, judging whether the end condition is met, if yes, executing C5, if not, executing C4.

[0131] C4: updating the speed and position of each particle, and returning to C3.

[0132] C5: outputting the optimal position, and the particle group parameter corresponding to the optimal position is the combat parameter.

[0133] For step C4, the speed of each particle is updated in the following way:

[0134] v(t+1) = ωv(t) + c1r1(p-x(t)) + c2r2(g-x(t)),

[0135]

[0136] The position of each particle is updated in the following way:

[0137] x(t+1) = x(t) + v(t+1),

[0138] wherein, c1, c2 are learning factors, r1, r2 are random numbers, v(t+1) is the updated velocity of the particle, v(t) is the current velocity of the particle, p is the individual optimal position of the particle, g is the global optimal position of the whole group, x(t) is the current position of the particle, x(t+1) is the updated position of the particle, ω is the adaptive inertia weight, ω min is the minimum value of the inertia weight, ω max is the maximum value of the inertia weight, d, k are adjustment coefficients, t is the current iteration number, M is the maximum iteration number, opbes t is the maximum value of the damage probability of the particle in the current group at the specified position, op ave is the average value of the damage probability of the particle in the current group at the specified position.

[0139] Further, the updating method of the weight in the particle swarm algorithm is the adaptive inertia weight, and the damage probability value predicted by the iteration number, the target damage probability prediction model is further optimized to update the adaptive weight, which can not only improve the global search ability, but also dynamically adjust the inertia weight according to the different stages of the algorithm running and the current position of the particle, so that the algorithm has strong global search ability in the early stage and focuses on local search in the later stage, which can explore the potential solution more widely in the search space, prevent premature convergence into local optimum, and strengthen the local search ability, so that the algorithm can perform more detailed local search when approaching the optimal solution, improve the precision of finding the global optimal solution, accelerate the convergence speed, avoid premature convergence, and improve the stability and reliability of the particle swarm algorithm.

[0140] It can be understood that the implementation method of the particle swarm algorithm, for example, the updating method of the individual optimal and global optimal values in the particle swarm algorithm, the adjustment of the acceleration constant, and the like are all common knowledge in the art, which are not limited herein.

[0141] Compared with the prior art, the target damage probability optimization method provided by the embodiment obtains the killing probability of a certain working condition parameter at different positions based on a target damage probability prediction model, and judges whether the damage probability of the current working condition parameter at the specified position to the target meets the combat task requirement, if not, the particle swarm algorithm is used to find the optimal parameter corresponding to the current working condition parameter, which can quickly and accurately correct the current working condition parameter, thereby providing technical guidance for designers when designing the working condition parameter, and when the particle swarm algorithm is used, an adaptive adjustment of the inertia weight is provided, which can more quickly and accurately obtain the damage probability of the specified position meeting the combat task requirement; the damage probability of a specific weapon to a target is studied, the sample data under different working conditions are constructed, the training sample set is expanded, and the accuracy of the prediction model can be further improved; the neural network prediction method is used, which can effectively extract the feature representation under different altitudes, azimuth angles, pitch angles and bullet speeds, solve the complex nonlinear mapping problem from input to output, obtain the damage matrix of the target within 100 meters, and finally realize the quick prediction of the damage probability of the specific weapon to the specific target under different parameter conditions, solve the problem that the nonlinear relationship between various factors in the killing matrix cannot be accurately expressed in the prior art, and overcome the defects of large calculation amount and low accuracy of the existing traditional calculation method.

[0142] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the program can be stored in a computer readable storage medium, wherein the computer readable storage medium is a disk, an optical disk, a read-only memory or a random access memory, etc.

[0143] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for optimizing target damage probability, characterized in that: include: Acquiring weapon parameters and target parameters, and building a target damage probability prediction model based on the weapon parameters and the target parameters; Acquiring operating condition parameters to be tested, inputting the operating condition parameters to be tested into the target damage probability prediction model, and obtaining the damage probability of the operating condition parameters to the target at different positions; Based on the combat mission requirements, determine whether the damage probability of the working condition parameters to be tested at the specified location meets the requirements. If so, use the working condition parameters to be tested as combat parameters. If not, use the particle swarm algorithm to optimize the working condition parameters to be tested, and use the optimized parameters as combat parameters. During the optimization process, the input parameters represented by each particle are input into the target damage probability prediction model to obtain the damage probability corresponding to the particle at the specified position, and the fitness function is obtained based on the damage probability of the particle at the specified position and the operating parameters corresponding to the particle.

2. The method for optimizing target damage probability according to claim 1, characterized in that: The weapon parameters include: weapon type, performance parameters and warhead type; the target parameters include: target type, environment and location.

3. The method for optimizing target damage probability according to claim 2, characterized in that: The constructing of a target damage probability prediction model based on the weapon parameters and the target parameters includes: Setting a calculation range of weapon parameters in a kill matrix based on the type of the weapon and the performance parameters, wherein the calculation range of the weapon parameters includes: an altitude range, a speed range, a pitch angle range, an azimuth angle range, and a kill level range; Setting a plurality of operating condition parameters based on the calculation range, calculating a damage probability matrix of the warhead to the target and a shock wave damage probability matrix of the type of weapon to the target corresponding to each operating condition parameter, and calculating a joint damage probability matrix based on the damage probability matrix and the shock wave damage probability matrix; The joint damage probability matrix corresponding to each operating condition parameter is transformed and reorganized to form a sample data set; A two-stage network model is trained based on the sample data set to obtain a target damage probability prediction model.

4. The method for optimizing target damage probability according to claim 3, wherein: The particle swarm algorithm is used to optimize the parameters of the working condition to be tested, and the optimized parameters are used as operational parameters, including: A1: Initialize the particle swarm, select samples that meet the combat mission requirements from the sample data set to form a first sample data set, use the operating condition parameter of each data in the first sample data set as a particle, and use the operating condition parameter in the operating condition parameter to be tested as a particle, and determine whether the total number of the current particles meets the preset number. If so, execute A3; if not, execute A2; A2: Counting the ranges of various parameters of the first sample data set, and randomly selecting values ​​within the ranges of the various parameters based on a random generation algorithm, thereby generating a plurality of particles such that the total number of particles meets a preset number requirement; A3: Calculate the fitness function of each particle to update the individual optimal and global optimal values, and determine whether the end condition is met. If so, execute A5; if not, execute A4. A4: Update the speed and position of each particle and return to A3; A5: Output the optimal position. The particle group parameters corresponding to the optimal position are the combat parameters.

5. The method for optimizing target damage probability according to claim 4, characterized in that: The fitness function is calculated based on the working condition parameters corresponding to the particle, including: Where op′ is the damage probability of the particle at the specified position, is the probability of damage to the designated location required by the combat mission, h′, v′, ω1′, ω2′ are the height, speed, pitch angle, and azimuth angle corresponding to the particle, are the altitude, speed, pitch angle, and azimuth angle of the working condition parameters to be tested, and a, b, c, and d are constants.

6. The method for optimizing target damage probability according to claim 5, characterized in that: For step A4, update the velocity of each particle as follows: v(t+1)=ωv(t)+c1r1(px(t))+c2r2(gx(t)), Update the position of each particle as follows: x(t+1)=x(t)+v(t+1), Among them, c1 and c2 are learning factors, r1 and r2 are random numbers, v(t+1) is the updated velocity of the particle, v(t) is the current velocity of the particle, p is the individual optimal position of the particle, g is the global optimal position of the entire group, x(t) is the current position of the particle, x(t+1) is the updated position of the particle, ω is the adaptive inertia weight, ω min is the minimum inertia weight, ω max is the maximum inertia weight, d and k are adjustment coefficients, t is the current number of iterations, M is the maximum number of iterations, and op best is the maximum probability of damage of particles in the current group at the specified position, op ave It is the average damage probability of particles in the current group at the specified position.

7. The method for optimizing target damage probability according to claim 6, characterized in that: The calculation of the damage probability matrix of the warhead to the target and the shock wave destruction probability matrix of the type of weapon to the target corresponding to each operating condition parameter includes: Calculate the fragment force field data corresponding to each operating condition parameter, the fragment force field data including: fragment mass, velocity, and fragment area solid angle in different angular regions; Calculating the damage probability of the warhead of this type on the target at different positions in the kill matrix based on the fragment power field data, the damage probability at each position forming a damage probability matrix; Based on the shock wave radius of this type of weapon at different positions in the kill matrix, the shock wave damage probability of this type of weapon on the target at the corresponding position in the kill matrix is ​​calculated. The shock wave damage probability of each position constitutes the shock wave damage probability matrix.

8. The method for optimizing target damage probability according to claim 7, characterized in that: The calculating, based on the fragment force field data, the probability of damage to the target by the warhead of this type at different positions in the kill matrix includes: A target coordinate system is established with the position of the target as the origin; fragments in different angular regions are grouped according to mass to form a plurality of fragment mass groups, wherein each fragment in each fragment mass group has the same mass; For each position, the following operations are performed: the warhead is placed at the corresponding position in the target coordinate system according to the weapon height defined in the current operating parameters, and the firing line analysis method is used for each angular region to obtain the vulnerable area and exposure probability of the target at that position; based on the vulnerable area, the exposure probability, and each fragment mass group, the damage probability of the warhead of that type at that position is calculated.

9. The method for optimizing target damage probability according to claim 8, characterized in that: The converting and reorganizing the joint damage probability matrix corresponding to each operating condition parameter to form a sample data set includes: For each operating condition parameter, M*N training sample data can be constructed, and each training sample data is [h, v, α, θ, x, y, l, p], where M and N are the dimensions of the joint damage probability matrix, h is the height of the warhead, v is the speed of the warhead, α is the azimuth angle of the warhead, θ is the pitch angle of the warhead, x is the horizontal coordinate of the type of warhead in the target coordinate system, y is the vertical coordinate of the type of warhead in the target coordinate system, l is the lethality level of the warhead, and p is the joint damage probability of the warhead at (x, y).

10. The method for optimizing target damage probability according to claim 9, characterized in that: The two-stage network model includes: One-stage network model, the input data of the one-stage network model is the working condition parameters [h, v, α, θ] and the position information [x i ,y j ], the output data is corresponding to each [x i ,y j ]’s one-stage predicted damage probability p′, i = 1, 2, 3, ..., M, j = 1, 2, 3, ..., N; Matrix generation model, all [x i ,y j ] The corresponding one-stage predicted damage probability p′ is combined to form a one-stage damage probability prediction matrix; A two-stage network model, wherein the input data of the two-stage network model is the first-stage damage probability prediction matrix, the output data is the second-stage damage probability prediction matrix, and the second-stage damage prediction probability at a specified position in the second-stage damage probability prediction matrix is ​​used as the final damage probability; The first-stage network model is a one-dimensional convolutional neural network, the second-stage network model is a convolutional network based on the attention mechanism, and the loss function of the second-stage network model is: Among them, w1 and w2 are constants, y x，y is the damage probability value of the warhead at (x, y) in the sample data set, is the predicted value of the two-stage damage probability at (x, y) in the two-stage damage probability prediction matrix.

Images