CN114925592A - Explosion-killing bomb fire power planning method based on neural network and PSO algorithm - Google Patents
Explosion-killing bomb fire power planning method based on neural network and PSO algorithm Download PDFInfo
- Publication number
- CN114925592A CN114925592A CN202210236701.XA CN202210236701A CN114925592A CN 114925592 A CN114925592 A CN 114925592A CN 202210236701 A CN202210236701 A CN 202210236701A CN 114925592 A CN114925592 A CN 114925592A
- Authority
- CN
- China
- Prior art keywords
- damage
- explosion
- bomb
- explosive
- parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an explosion-killing bomb thermal power planning method based on a neural network and a PSO algorithm, and belongs to the technical field of explosion-killing bomb opposite target damage efficiency evaluation and thermal power planning. Firstly, obtaining a damage amplitude man image of the ammunition to a target under different tail end trajectory parameters and fuze parameters by utilizing a real dynamic detonation power field of each explosive killing bomb; training a generator capable of generating a damaged spokesman image according to the type of the warhead, the parameters of the tail end trajectory and the fuze parameters by using a multilayer perceptron model; finally, the generator and the PSO algorithm are coupled to carry out firepower planning of the damage of the explosive bomb to the target; the method achieves the aim of considering both the accuracy and the rapidity of the blast fire power planning of the opposite target of the explosive-killing bomb. The method is suitable for the fields of battle command and the like, and can accurately and quickly evaluate the damage efficiency of the explosive bomb facing targets and plan fire power.
Description
Technical Field
The invention relates to an intelligent fire planning method for a bomb killer based on a neural network (a multilayer perceptron model) and a PSO (particle swarm optimization) algorithm, and belongs to the technical field of the damage efficiency evaluation and the fire planning of a target opposite to the bomb killer.
Background
In the mechanical war era, the fire intensity and the maneuvering capability of weaponry are emphasized, the task execution capability of a killer chain is focused, and the fighting capability of a single weapon platform is emphasized. In the information war era, a full-dimensional information network covering a battlefield is constructed by depending on satellites and radars, extensive combat covered by firepower is converted into efficient point damage of accurately guided ammunition, the damage power is huge, meanwhile, the casualties of civilian are reduced, and the observation and adjustment and information interaction capacity of a battlefield sensor and the accurate application of the ammunition are emphasized. If the fire planning is not carried out on the accurately guided ammunition, the 'over damage' is easily caused when multiple ammunitions hit the same position of the target, and the efficiency-cost ratio is lower; therefore, the reasonable fire planning needs to be carried out on the multiple ammunitions, the minimum ammunition consumption is realized, the maximum damage effect is achieved, and the combat effectiveness-cost ratio is improved.
The existing blast-killing target fire power distribution model of the striking surface of the blast bomb is established on an optimal aiming point, the damage range of the blast bomb is equivalent to a circular area with a blasting point as a circle center for simplifying calculation, a non-circular dynamic explosion power field under the actual near-explosion condition of the blast bomb is not considered, the use requirement of accurately guiding efficient damage of the bomb cannot be met, and the method is limited in providing help for battle command. And the calculation of the blast-killing bomb damage amplitude transformer considering the dynamic explosion power field of the real ballistic environment needs to traverse all the intersection points of thousands of fragment traces and targets, has low calculation speed and is difficult to be used for real-time firepower planning. If the firepower planning based on the dynamic and explosive power field is required, iterative calculation needs to be carried out for many times, and the calculation time is too long to be used for actual combat. Therefore, it is necessary to create a fire planning method with both accuracy and rapidity to meet the battle requirements.
Disclosure of Invention
Aiming at the problem that the calculation time of the existing dynamic explosion power field fire power planning is too long, the invention mainly aims to provide a multi-explosive-killing-bomb-striking-fire-power planning method based on a neural network and a PSO algorithm, and solves the problem that the accuracy and the rapidity are difficult to be considered in the explosive-killing-bomb opposite-target damage efficiency evaluation and the fire power planning.
The purpose of the invention is realized by the following technical scheme:
the invention discloses a blast-killing fire planning method based on a neural network and a PSO algorithm, which comprises the steps of firstly obtaining a damage spokesman image of an ammunition to a target under different terminal trajectory parameters and fuze parameters by utilizing a real dynamic blasting power field of each blast-killing bomb; training a generator capable of generating a damaged spokesman image according to the type of the warhead, the parameters of the tail end trajectory and the fuze parameters by using a multilayer perceptron model; finally, the generator and the PSO algorithm are coupled to carry out firepower planning of the damage of the explosive bomb to the target; the aim of considering both the accuracy and the rapidity of the blast bomb opposite target striking firepower planning is fulfilled.
The invention discloses an explosion-killing bomb fire power planning method based on a neural network and a PSO algorithm, which comprises the following steps:
step 1: aiming at a certain explosive bomb, a power field under the dynamic explosion condition can be obtained through experimental tests, simulation calculation of simulation software or calculation of an empirical formula.
Experimental mode: firstly, collecting the scattering angle and the scattering speed of the broken pieces through a static explosion power experiment to form a static explosion power file, and then carrying out velocity vector superposition on the broken piece static explosion power field according to the bullet meeting condition to obtain a dynamic explosion power field. And under the condition of permission, the dynamic explosion experiment can be directly carried out to obtain dynamic explosion force field data.
The simulation mode is as follows: firstly, obtaining a static explosion power field according to numerical simulation, and then carrying out velocity vector superposition on the broken piece static explosion power field according to the terminal ballistic condition to obtain an dynamic explosion power field.
Empirical formula: the ammunition structural parameters are substituted into an empirical formula to calculate a static explosion power field, and then an dynamic explosion power field is calculated.
Step 2: according to the dynamic explosive power field of the explosive-killing bomb, the parameters of the terminal trajectory and the parameters of the fuze are changed, and a damage amplitude member image is formed by traversing the intersection point of the fragment trace and the target, or a method for calculating the damage amplitude member by using the real power field is utilized to obtain enough damage amplitude member images for manufacturing a training set.
There are many factors that affect the shape of a blast-killing bomb that damages the shape of the sponders, including: the coordinates of the explosion point of the explosion-killing bomb, the azimuth angle, the falling speed, the falling angle, the explosion height, the CEP value of a random factor and the target vulnerability characteristics. The moving of the coordinates of the explosion points and the changing of the azimuth angle only translate and rotate the images of the damaged amplitude-plotter and influence the training of the image generator, so that the training set is not considered when being generated; the presence of the CEP value affects the training performance of the image generator and is therefore not considered when generating the training set. By respectively setting the falling speed, falling angle and burst height range and the variation step length of the detonation-killing bomb, images of the explosive-killing bomb damage sponders under different conditions are obtained and used as training sets.
And (3) a damage spotter image drawing mode: according to the dynamic explosion power field, obtaining moving track rays of each fragment, and calculating the intersection point of each fragment and a surface target; and judging whether the effective damage threshold is reached or not according to the kinetic energy when the fragments meet the surface target, recording the coordinates of the effective fragment falling points, and drawing the image of the damage spotter. The image of the damaged spotter can be drawn quickly by programming and calculating.
The damage width image only needs to reflect the fragment dispersion condition, and does not need to have too high definition, otherwise the training difficulty is large and the practical significance is not realized; the image needs to be large enough to encompass most of the area of the image of the marred radiator; the images need to have the same pixel value size, otherwise the training of the generator model is not facilitated.
And 3, step 3: and constructing a damaged amplitude transformer image generator based on a multilayer perceptron model, namely a fully-connected neural network.
The neural network structure comprises an input layer, an output layer and a preset number of hidden layers, wherein all layers are connected, and an activation function is added. Since the updating of the model parameters requires processing a large amount of data at the same time, the model parameters are updated by a stochastic gradient descent method.
And 4, step 4: and (4) inputting the training set obtained in the step (2) into the neural network in the step (3) for training to obtain an image generator.
And 5: and (3) carrying out fire planning on the opposite target of the bomb-killing by adopting a PSO algorithm, and considering the damage efficiency evaluation of the opposite target and the accuracy and rapidity in the fire planning.
In the steps 1 to 4, a generator for generating the damage spokesman image according to different terminal ballistic parameters and fuse parameters is used for simulating the effect of the explosion-killing bomb on the damage spokesman in a real battle scene after operations such as translation, rotation, CEP (computer aided process) processing and the like are added to the output image.
Performing mathematical modeling on the multiple explosive bombs for the damage width of the target to obtain a damage width prediction model:
s m (x m ，y m ，Dir m ，Dro m ，Fs m ，Bh m )m＝1，2，…
-180°≤Dir m ≤180°；0°≤Dro m ≤90°
wherein x and y are coordinates of a drop point, Dir is an azimuth angle, Dro is a drop angle, Fs is a drop speed, and Bh is a blast height.
s m Recording the bullet-eye intersection condition for the related parameters of a detonation-killing bullet during explosion, wherein the condition is a function of variable azimuth angle, falling speed and explosion height; s (S) m ) Provided that s is m The size of a damaged spokesman under the condition is calculated, and the conditions are summed firstly and then the spokesman calculation is carried out when multiple explosive bombs are calculated; s. the 0 Is the target area; p is the percentage of total offender area in the target. The optimal objective function is obtained as:
by a random collection of conditions
Forming a plurality of particles, optimizing collocation of operation through iteration of generations, commanding all the particles to simultaneously carry out updating of a random size step length towards a local optimal solution direction in each generation, stopping iteration updating after size convergence of a damaged member, and outputtingGenerating an optimal condition set
The bullet-and-eye intersection set is the firepower planning result.
And limiting partial parameters in Dir, Dro, Fs and Bh to optimize only the remaining parameters, and the method can be used for meeting different battle requirements on a battlefield, and specifically comprises the following steps:
and the first mode is not limited, and six variables of the meeting conditions of the explosive-killing bullets are optimized simultaneously. In the mode, the calculation amount is large, the calculation time is long, and the method is suitable for being used under the condition of complex battle;
and in a second mode, Dro, Fs and Bh are limited, only the explosive bombs x, y and Dir are optimized, and the Dro, Fs and Bh default to values under the condition that a single explosive bomb damages an amplitude-plotter to the maximum extent. Compared with the mode, the mode has the advantages that the speed is obviously improved, and the mode is suitable for conventional combat requirements;
and a third mode is that Dir, Dro, Fs and Bh are limited, only the explosive bombs x and y are optimized, Dir needs to give a certain value, and Dro, Fs and Bh default to the value of the single explosive bomb under the condition that the damage amplitude worker is maximum. The mode is used for simulating the condition that all the explosive bombs strike the target from one direction, and the calculation speed is fastest and is closest to a real battle scene.
Advantageous effects
1. The invention discloses an explosive bomb fire planning method based on a neural network and a PSO algorithm, wherein a multilayer perceptron model is introduced in explosive bomb opposite target damage amplitude worker calculation and damage efficiency evaluation, the damage efficiency evaluation time of a single explosive bomb opposite target based on a real dynamic explosion power field is shortened to one thousandth of a second, and the rapidness and the accuracy of the damage efficiency evaluation of the explosive bomb opposite target are considered.
2. The invention discloses a fire planning method for an explosion-proof bomb based on a neural network and a PSO algorithm, wherein the PSO algorithm is introduced into fire distribution planning of opposite-surface target striking of the explosion-proof bomb, and the fire planning time of the explosion-proof bomb for shortening the coverage rate of an opposite-surface target damage operator to 80% to within 3 seconds; the firepower planning time of the explosion-killing bomb covering the damage breadth factor of the large-scale target to more than 80% is shortened to be within 10 seconds, and the rapidity and the accuracy of firepower distribution planning of the explosion-killing bomb to the large-scale target are both considered.
3. The invention discloses an explosive-killing fire planning method based on a neural network and a PSO algorithm, which adopts different parameter optimization schemes, provides a plurality of calculation modes, meets different combat demands on a battlefield in different modes, and has strong selectivity.
Drawings
FIG. 1 is a diagram of a multi-layered sensor model according to the present invention.
FIG. 2 is a comparison graph of the fitting effect of the explosive-killing bomb damage spotter and the calculated effect according to the dynamic explosion power field;
the left image is a picture generated by a generator obtained by training a multilayer perceptron model, and the right image is a picture obtained by calculation according to a real dynamic and explosive power field.
FIG. 3 is a firepower planning interface programmed according to the present invention and a damage spotter interface calculated from a real life and explosive power field.
The left graph is a firepower planning operation interface, and the right graph is a damage amplitude operator size operation interface calculated according to the dynamic explosion power field.
FIG. 4 is a graph of the result of the fire planning of the present invention plotted against the kinetic energy field.
Fig. 5 is a flow chart of the fire planning of the present invention.
Detailed Description
The invention will be described in detail below with reference to the drawings and examples. While the technical problems and advantages of the present invention have been described in detail, it should be noted that the embodiments described are only intended to facilitate the understanding of the present invention, and should not be construed as limiting.
As shown in fig. 5, the present embodiment discloses a specific application of an explosive bomb damage performance evaluation and fire planning method for an explosive bomb based on a neural network and a PSO algorithm to a certain explosive bomb, which includes the following steps:
step 1: and obtaining the explosion-killing and bullet-moving power field by using a simulation mode.
Firstly, modeling explosive killing bombs by using Hypemesh, TurGrid preprocessing and other software, introducing the model into LS-DYNA software for simulation calculation to obtain a static explosion power field, and then performing velocity vector superposition on the fragment static explosion power field according to the bullet meeting conditions to obtain a dynamic explosion power field.
Step 2: changing the trajectory parameters and the fuze parameters of the tail end of the explosive bomb, calculating a damage sponders based on the dynamic explosion powerfield to obtain a plurality of images of the damage sponders of the explosive bomb, and making a training set.
The bomb is set to strike a surface target with the size of 200m x 200 m. The size can basically cover the whole shape of the explosive-killing bomb damage amplitude transformer;
setting a coordinate (0,0) point of an explosion point of the explosion killing bomb; in order to improve the accuracy of the training result, the azimuth angle is set to be 0 degree; setting the CEP value to 0 since the presence of the CEP value may cause the image to be shifted;
setting the falling speed range to be 200-900 m/s and the step length to be 20 m/s; the frying height range is 2m-9m, and the step length is 1 m; to simplify the model, the fall angle was set at 80 °. According to the calculation result of the damage spotters, the explosive bomb can be known to cause the largest damage to the spotters under the condition that the falling angle is 80 degrees.
Respectively outputting the falling speed and the explosion height as input training sets, wherein the size of the matrix is 1 multiplied by 2, and the matrix is shown in an input layer of the figure 1;
according to the dynamic explosion power field, obtaining moving track rays of all the fragments, calculating the intersection points of all the fragments and the surface targets, calculating the kinetic energy when the fragments and the surface targets are intersected, judging whether the threshold value of effective damage is reached or not, and recording the coordinates of effective fragment falling points; then 200m is produced
Dividing a 200m face target into 40000 small grids of 1m multiplied by 1m, setting a label '1' for the grid if a valid fragment is received in the grid, and otherwise setting a label '0'; the output is thus a 200 x 200 matrix, as shown in the output layer of fig. 1;
and calculating and outputting a plurality of damage spotters as an output training set.
And step 3: and constructing a multilayer perceptron model.
Because most of the cases of the target with the damaged surface of the explosive bomb can be basically covered by only a small training set; therefore, the generalization ability of the model is not the primary index for selecting the model; the actual combat environment requires that the model can accurately generate the shape of the explosive bomb damage spokesman, so the accuracy of the generator model for generating the image should be used as a primary index for selecting the model. The multi-layered perceptron model may be adequate for similar work.
According to the analysis of the input training set and the output training set in the step 2, the input is a matrix of 1 × 2, and the output is a matrix of 200 × 200, which are the input layer and the output layer of fig. 1, respectively;
aiming at the explosion killing bomb, two hidden layers which are respectively 1 multiplied by 100 and 1 multiplied by 200 are adopted; ReLU function is adopted among the input layer, the first layer hidden layer and the second layer hidden layer as an activation function, and Sigmoid function is adopted between the second layer hidden layer and the output layer as an activation function; and a random gradient descending mode is adopted for updating the neural network parameters.
And 4, step 4: and inputting the input training set and the output training set into a neural network for training to obtain an image generator.
Inputting parameters of the blast height of 9m and the falling speed of 900m/s into a trained image generator to obtain an image of the damaged spoondsman, wherein the time is one thousandth of a second, the image is shown in the left graph of figure 2, and the right graph of figure 2 is the image of the damaged spoondsman obtained according to the power field calculation. Therefore, the damage performance evaluation model can meet the requirements of accuracy and rapidity at the same time.
And 5: and designing a PSO algorithm to carry out fire planning on the targets with the determined sizes after the blast killing bomb is hit, and considering accuracy and rapidity in damage efficiency evaluation of the opposite target and fire planning.
And in the previous steps, a generator for generating a damage spokesman image according to different tail end trajectory parameters and fuze parameters of the explosive bomb is used, and translation, rotation and CEP (central processing unit) operation are added to the output image, so that the effect of the explosive bomb on damaging the spokesman in a real battle scene can be simulated.
Performing mathematical modeling on the multiple explosive bombs for the damage width of the target to obtain a damage width prediction model as follows:
sm(x m ，y m ，Dir m ，Dro m ，Fs m ，Bh m )m＝1，2，…
-180°≤Dir m ≤180°；0°≤Dro m ≤90°
s m recording the bullet-eye intersection condition for a relevant parameter when the explosive bomb explodes, wherein the condition is a function of the following variables: dir (azimuth), Dro (fall angle), Fs (fall speed), Bh (blast height); s (S) m ) Provided that s is m The size of a damage spokesman under the condition is calculated, and the conditions are summed firstly when the multiple explosive bombs are calculated, and then the spokesman calculation is carried out (the damage spokesman of the multiple explosive bombs has the overlapping condition); s. the 0 Is the target area; p is the percentage of the total lesion width in the target area. The optimized objective function is obtained as:
and defining partial parameters in Dir, Dro, Fs and Bh can optimize only the rest parameters and can be used for meeting different combat requirements on a battlefield.
The problem belongs to a continuous variable nonlinear programming problem, and a required answer can hardly be found in a specified time by adopting a traversal method. The Particle Swarm (PSO) is used as an improved version of the genetic algorithm, combines an optimization mode of simulating the vicious and vicious of nature by the genetic algorithm, improves the randomness of genetic algorithm variation, and is suitable for solving similar problems.
By a random collection of conditions
Forming a plurality of particles, optimizing collocation of operation through iteration of generations, commanding all the particles to simultaneously carry out update of a random size step length towards a local optimal solution direction by each generation, stopping iteration update after size convergence of a damaged spotter, and outputting an optimal condition set
The bullet-and-eye intersection set is the firepower planning result.
In this embodiment, 20 particles are set, each particle includes three groups of bullet intersection sets, each group of condition set includes "required bullet quantity" elements, and each element is an s m That is, the bullet-target meeting condition of the single-shot blast bomb, in this example, the mode two is selected, Dro, Fs, Bh are limited, and only the blast bombs x, y, Dir are optimized.
Adjusting the update step length parameters, wherein the update step lengths are all set to be 1.5 in the example, then circularly updating the optimal bullet-and-eye intersection condition by using a particle swarm algorithm, namely the size of a damaged member
And jumping out of the loop after the size is converged. The effect is shown in FIG. 3, the left graph is an optimization software interface, and the right graph calculates an operation interface of damage spotter software according to a real explosion power field; from the results, the optimization time is 7.08s, 5 explosive bombs cover 77.4% of the area of a target of 200m multiplied by 100m, the fire planning results are verified and plotted by using a real dynamic explosion power field, and from the results, the size of a damaged member of the 5 explosive bombs is 15255m 2 Covering an area of 76.3% of the target with only 1.1% error. From the drawing effect, the damaged members of 5 explosive bombs are uniformly distributed in each area of the surface target, and the optimization result is good.
In summary, the explosion-killing bomb thermal power planning method based on the neural network and the PSO algorithm disclosed by the invention can completely solve the problem that the prior art cannot give consideration to rapidity and accuracy.
The above detailed description is intended to illustrate the objects, aspects and advantages of the present invention, and it should be understood that the above detailed description is only exemplary of the present invention and is not intended to limit the scope of the present invention, and any modifications, equivalents, improvements and the like within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (5)
1. An explosion-killing bomb fire power planning method based on a neural network and a PSO algorithm is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
step 1: aiming at a certain explosive bomb, a power field under the dynamic explosion condition can be obtained through experimental test, simulation software simulation calculation or empirical formula calculation;
step 2: changing parameters of a tail end trajectory and fuze according to a dynamic explosion power field of the explosion-killing bomb, forming a damage spotter image by traversing a broken trace and a target intersection point, or obtaining enough damage spotter images by utilizing a method of calculating the damage spotter by utilizing a real power field for manufacturing a training set;
and 3, step 3: constructing a damage amplitude image generator based on a multilayer perceptron model, namely a fully-connected neural network;
and 4, step 4: inputting the training set obtained in the step 2 into the neural network in the step 3 for training to obtain an image generator;
and 5: and (4) carrying out fire planning on the opposite target of the bomb-killing bomb by adopting a PSO algorithm, and considering the damage efficiency evaluation of the opposite target and the accuracy and rapidity in the fire planning.
2. The method for planning the fire of the detonation gun based on the neural network and the PSO algorithm as claimed in claim 1, characterized in that: the implementation method of the step 1 is that,
experimental mode: collecting the scattering angle and scattering speed of the fragments through a static explosion power experiment to form a static explosion power file, and then performing velocity vector superposition on the static explosion power field of the fragments according to the bullet meeting condition to obtain a dynamic explosion power field; the dynamic explosion experiment can be directly carried out under the condition that the conditions allow to obtain dynamic explosion force field data;
the simulation mode is as follows: firstly, obtaining a static blasting power field according to numerical simulation, and then performing velocity vector superposition on the broken static blasting power field according to a terminal trajectory condition to obtain a dynamic blasting power field;
empirical formula: the ammunition structural parameters are substituted into an empirical formula to calculate a static explosion power field, and then an dynamic explosion power field is calculated.
3. The method for planning the fire of the detonation gun based on the neural network and the PSO algorithm as claimed in claim 1, characterized in that: the implementation method of the step 2 is that,
there are many factors that affect the shape of the blast panel from a bomb blast, including: the coordinates, azimuth angles, falling speeds, falling angles, burst heights, CEP values of random factors and target vulnerability characteristics of detonation-killing bomb explosion points; the moving of the coordinates of the explosion points and the changing of the azimuth angle only translate and rotate the images of the damaged amplitude-plotter and influence the training of the image generator, so that the training set is not considered when being generated; the existence of the CEP value can influence the training effect of the image generator, so that the CEP value is not considered when generating the training set; obtaining images of explosive-killing bomb damage sponders under different conditions as training sets by respectively setting the falling speed, falling angle and explosive height range and the variation step length of the explosive-killing bombs;
and (3) a damage spotter image drawing mode: according to the dynamic explosion power field, obtaining moving track rays of all the fragments, and calculating the intersection points of all the fragments and the surface target; judging whether an effective damage threshold is reached or not according to the kinetic energy when the fragments meet the surface target, recording the coordinates of effective fragment falling points, and drawing a damage spotter image; programming calculation can be carried out to rapidly draw the image of the damaged spotter;
the damage spotters only need to reflect the dispersion situation of the fragments without high definition, otherwise, the training difficulty is high and the practical significance is not realized; the image needs to be large enough to encompass most of the area of the image of the offending spotter; the images need to have the same pixel value size, otherwise the training of the generator model is not facilitated.
4. The method for planning the fire of the detonation gun based on the neural network and the PSO algorithm as claimed in claim 1, characterized in that: the implementation method of the step 3 is that,
the neural network structure comprises an input layer, an output layer and a preset number of hidden layers, wherein all layers are connected with one another and are added with an activation function; since the update of the model parameters requires processing a large amount of data at the same time, the model parameters are updated by a stochastic gradient descent method.
5. The method for planning the fire of the detonation gun based on the neural network and the PSO algorithm as claimed in claim 1, characterized in that: the implementation method of the step 5 is that,
in the steps 1 to 4, a generator for generating a damage spotter image according to different terminal ballistic parameters and fuze parameters is used, and the effects of the blast-killing bomb damage spotter in a real battle scene can be simulated after operations such as translation, rotation, CEP (computer aided process) processing and the like are added to the output image;
performing mathematical modeling on the multiple explosive bombs for the damage width of the target to obtain a damage width prediction model:
s m (x m ，y m ，Dir m ，Dro m ，Fs m ，Bh m )m＝1，2，…
-180°≤Dir m ≤180°；0°≤Dro m ≤90°
wherein x and y are coordinates of a drop point, Dir is an azimuth angle, Dro is a drop angle, Fs is a drop speed, and Bh is a blast height;
s m recording the bullet-eye intersection condition for the related parameters of a detonation-killing bullet during explosion, wherein the condition is a function of variable azimuth angle, falling speed and explosion height; s (S) m ) Provided that s is m The size of a damaged spokesman under the condition is calculated, and the conditions are summed firstly and then the spokesman calculation is carried out when multiple explosive bombs are calculated; s 0 Is the target area; p is the percentage of the total damage width occupying the target area; the optimal objective function is obtained as:
by a random collection of conditions
Forming a plurality of particles, optimizing the collocation of operations by iteration of generations, each generation being a command postThe particles simultaneously face the direction of the local optimal solution to update a random size step, and iteration updating is stopped after the size of the damaged member is converged to output an optimal condition set
The bullet and eye intersection set is a firepower planning result;
and limiting partial parameters in Dir, Dro, Fs and Bh to optimize only the residual parameters, wherein the parameters can be used for meeting different battle requirements on a battlefield, and the parameters are as follows:
the method comprises the following steps of firstly, optimizing six variables of meeting conditions of explosive-killing bullets without any limitation in a mode I; in the mode, the calculation amount is large, the calculation time is long, and the method is suitable for being used under the condition of complex battle;
defining Dro, Fs and Bh, optimizing the explosive bombs x, y and Dir, and acquiescently taking the value of the single explosive bomb under the maximum condition of damaging a sponderer; compared with the mode, the mode has the advantages that the speed is obviously improved, and the mode is suitable for conventional combat requirements;
defining Dir, Dro, Fs and Bh, optimizing the explosive bombs x and y only, wherein Dir needs to give a certain value, and Dro, Fs and Bh obtain the value of the single explosive bomb in the maximum condition of damaging a sponderer by default; the mode is used for simulating the condition that all the explosive bombs strike the target from one direction, and the calculation speed is fastest and is closest to a real battle scene.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210236701.XA CN114925592A (en) | 2022-03-11 | 2022-03-11 | Explosion-killing bomb fire power planning method based on neural network and PSO algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210236701.XA CN114925592A (en) | 2022-03-11 | 2022-03-11 | Explosion-killing bomb fire power planning method based on neural network and PSO algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114925592A true CN114925592A (en) | 2022-08-19 |
Family
ID=82805473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210236701.XA Pending CN114925592A (en) | 2022-03-11 | 2022-03-11 | Explosion-killing bomb fire power planning method based on neural network and PSO algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114925592A (en) |
-
2022
- 2022-03-11 CN CN202210236701.XA patent/CN114925592A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104143028B (en) | Armored equipment damage rule analysis method based on simulation experiment | |
| Przemieniecki | Mathematical methods in defense analyses | |
| CN106203870A (en) | A kind of complex analysis towards combined operation and weapon allocation method | |
| CN114722573A (en) | Method for evaluating damage of multi-type multi-explosive-bomb opposite targets | |
| CN109063331B (en) | Evaluation method for damage effect of small off-target fragment warhead on reentry warhead | |
| CN110210115A (en) | The design of operation simulating scheme and operation method emulated based on decision point and branch | |
| CN106980738A (en) | The construction method and device of ammunition power firing table | |
| CN113553777A (en) | Anti-unmanned aerial vehicle swarm air defense deployment method, device, equipment and medium | |
| CN116305988A (en) | Warhead power simulation system and method based on fuze modeling | |
| CN114925592A (en) | Explosion-killing bomb fire power planning method based on neural network and PSO algorithm | |
| Yang et al. | A Virtual Reality Approach to the Assessment of Damage Effectiveness of Naval Artillery Ammunition against Unmanned Surface Vessels | |
| Wu et al. | Dynamic multitarget assignment based on deep reinforcement learning | |
| Chen et al. | A method to determine the shell layout scheme for equipment battlefield damage tests under artillery fire | |
| Peng et al. | Terminal attitude selection method of missile attack aircraft | |
| Yan et al. | Research on intelligent minefield attack decision based on adaptive fireworks algorithm | |
| Hou et al. | Adaptive fuze-warhead coordination method based on BP artificial neural network | |
| Li et al. | Impact of tactical parameters of aircraft on jamming effectiveness of surface-source IR Decoy | |
| CN115345035B (en) | Optimization method for aiming point of body target multidirectional striking space | |
| Li et al. | Modeling and calculation method of target damage based on multi-attitude flying projectile in space intersection | |
| Hao et al. | Calculation Model and Method of Gain-Loss Game Distribution Mechanism on Projectile and Target Intersection Confront | |
| Annunziata et al. | Vulnerability-driven multi-objective topological optimization of aircraft protections | |
| Mazonka et al. | Methods and Models in Preparing Weapon-Target Interaction Data for Combat Simulations | |
| Li et al. | A target damage effectiveness assessment mathematical calculation method with uncertain information based on an adaptive fuzzy neural network | |
| Tan et al. | Influence of prefabricated fragments projectile cabin opening attitude on damage probability and attitude optimization | |
| Shi et al. | Design of Visualization Simulation Software for Fire Planning Based on Unity 3D Technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |