CN114372385A - Artificial intelligence-based multi-bullet fast fire planning method - Google Patents
Artificial intelligence-based multi-bullet fast fire planning method Download PDFInfo
- Publication number
- CN114372385A CN114372385A CN202110045002.2A CN202110045002A CN114372385A CN 114372385 A CN114372385 A CN 114372385A CN 202110045002 A CN202110045002 A CN 202110045002A CN 114372385 A CN114372385 A CN 114372385A
- Authority
- CN
- China
- Prior art keywords
- bullet
- target
- damage
- planning
- knowledge
- 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
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- 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
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a multi-bullet fast firepower planning method based on artificial intelligence, which comprises the following steps of S1: establishing a finite element damage simulation cloud computing platform, and computing damage effects of multiple bullet types; s2: carrying out bullet matching of multiple bullet types based on knowledge graph reasoning; s3: and based on a deep learning theory, establishing a bullet consumption planning model according to the damage effect calculation result in the step S1 and the bullet matching result in the step S2, and outputting the multiple-bullet fast fire power planning. Aiming at key technologies in the multi-bullet fast fire planning, a finite element damage simulation cloud computing platform framework is established, a bullet matching implementation process based on knowledge map reasoning is provided, a deep learning model for bullet consumption planning is established, and a system solution for solving the multi-bullet fast fire planning problem by using cloud computing and artificial intelligence technologies is provided.
Description
Technical Field
The invention relates to the technical field of multi-bomb-type fire planning, in particular to a multi-bomb-type rapid fire planning method based on artificial intelligence.
Background
In the prior combat methods, a strike method of using weapons with different damage mechanisms to strike the same target appears, and the overall optimal combat effect is achieved by comprehensively using different damage effects of various weapons. The innovative tactics fully explore the comprehensive combat effectiveness of the existing weapons. However, theoretical studies on combined shots are still in the stage of starting. The traditional optimization method is difficult to solve the problems of dimension explosion, unstructured optimization and the like, and the requirement of a modern operation mode on real-time online firepower planning is difficult to meet.
Under the multi-bullet striking mode, the warhead with different effects of penetration, killing, shock waves, combustion and the like performs combined striking on the target, and the prediction and evaluation of the composite damage effect are the premise of accurate firepower planning. At present, the calculation of the composite damage effect is mostly carried out based on finite element modeling, and in damage effect and numerical simulation of ultra-high performance cement-based composite materials under penetration and explosion (Yanhao, Nanjing university of rational Engineers, 2018.03.), numerical simulation is carried out on the penetration and explosion coupling damage effect by adopting a finite element model based on ANSYS-AUTODYN software. In the research on the composite damage of the discrete rod and the shock wave to the supersonic missile (Jiang Ying Zi, etc., bullets and arrows and guidance, 2015, 35 (3): 63-67.), the damage effect of the composite action of the discrete rod and the shock wave to the supersonic missile is researched based on LS-DYNA software. In the research on the damage effectiveness of the combined effect warhead on the tank and the armored vehicle top armor (Li Yu Pin, university of northcentral, 2019.05), the damage rule of the combined effect warhead on the tank armored vehicle top armor target under different structural parameters is researched based on LS-DYNA software.
The finite element method essentially utilizes the discretization thought, adopts the variational principle, the continuum subdivision and the piecewise interpolation, and has very complex calculation process. The method can accurately model and calculate specific targets, but is difficult to simulate large-scale system targets and traverse all target types which may be hit by battle application. The current research on the prediction and evaluation of the composite damage effect only stays at the material level and the target component level, and cannot support the prediction and evaluation of the multi-bullet combined strike damage effect at the battle-level firepower planning level.
Disclosure of Invention
Aiming at the existing problems, the invention aims to provide a multi-bullet fast fire planning method based on artificial intelligence, which carries out multi-bullet fast fire planning by using cloud computing, knowledge map reasoning and deep learning technologies and provides technical support for constructing an intelligent fire planning system.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a multi-bomb fast fire planning method based on artificial intelligence is characterized by comprising the following steps,
s1: establishing a finite element damage simulation cloud computing platform, and computing damage effects of multiple bullet types;
s2: carrying out bullet matching of multiple bullet types based on knowledge graph reasoning;
s3: and based on a deep learning theory, establishing a bullet consumption planning model according to the damage effect calculation result in the step S1 and the bullet matching result in the step S2, and outputting the multiple-bullet fast fire power planning.
Further, the mathematical description of the multiple-bomb fast fire schedule is as follows: given m types of ammunition, the ith type of ammunition has NMi(i is 1, …, m) and the total shot size is NM, and combined shots are performed on n targets. Multi-tuple MBHLijRepresenting the fire distribution scheme of the ith missile to the jth target
In the formula (d)ijRepresenting the amount of the ith bullet allocated to strike the jth target, taking a positive integer or 0, wherein i is 1, …, m; j is 1, …, n;
for given bullet type and quantity, the decision variables of firepower planning problem can use multi-element group MBHLijIs described by the set HL of (a) of (b),
in the tuple array HL, m × n tuples are provided, and each tuple
Containing 1 integer variable, and 2 xdijIndividual real number type
The overall damage effect of the target in the multi-bullet fire planning problem can be expressed as the functional of each target attribute information, each bullet attribute information and the mixed fire distribution scheme HL, and max P ═ f (MB)1，…，MBn，DZ1，…，DZmHL); where MBi represents the target parameters needed to calculate the damage effect, i is 1, …, n; DXjRepresents the weapon parameters needed to calculate the effect of the damage, j-1, …, m.
Further, the finite element damage simulation cloud computing platform in step S1 includes a fire planning user layer, a cloud platform application layer, a resource scheduling layer, a model service layer, and an infrastructure layer;
the fire planning user layer provides an operation interface for fire planning auxiliary decision-making personnel, sends damage calculation requests to the cloud computing platform through the fire planning user layer, and adjusts various parameters of weapons and targets according to needs;
the cloud platform application layer comprises model creation, damage simulation and data management, wherein finite element analysis services are dynamically combined to respond to a user request according to the damage calculation requirements of a firepower planning user layer, so that tasks are disassembled, the damage effects of various bullets are calculated, and a foundation is provided for resource scheduling;
the resource scheduling layer allocates enough virtual resources for implementing finite element-based damage calculation, so that the calculation efficiency is improved;
the model service layer comprises finite element mesh division, simulation algorithm service and dynamic model service provided by a cloud computing platform, and the damage computation based on the finite element is realized in a service mode according to a functional module mode;
the base implementation layer includes a computing host, a storage device, and a network device.
Further, the specific operation of the cloud platform application layer for calculating the damage effects of multiple bullet types comprises the following steps,
s11: dividing a hit region omega into M sub-domains delta sigma according to the shape and physical characteristics of the hit region of the warheadm(M-1, 2, …, M), i.e. a unit;
s12: numbering the units and the nodes, determining the relationship between the units and the nodes, and listing natural boundaries and corresponding boundary values;
s13: selecting a corresponding interpolation function as a unit shape function according to the number of the unit nodes in the area and the requirement of calculating an approximate solution, approximating the solving function of each area unit by using a linear combination expression of the unit shape function, substituting the approximate function into an integral equation, integrating the unit area to obtain an algebraic equation containing undetermined coefficients, namely a unit finite element equation set; the undetermined coefficient comprises a target structure parameter, a target material parameter and a target function parameter;
s14: solving by adopting a proper numerical calculation method in a finite element according to the boundary condition to obtain the coefficient value of each node, namely the value of the coefficient to be determined;
s15: warhead hit Δ σmThe damage degree D (m) of the target is determined by the damage degree and vulnerability analysis result of each target unit, and the calculation formula is as follows:
in the formula, K is the number of target units; s (m, k) is the damage degree of the kth target unit, and A (k) is the influence factor of the kth target unit;
wherein the warhead hits Δ σmThe calculation expression of the damage degree S (m, k) of the target unit is S (m, k) ═ S { Z (0, k), Y, C, O }, wherein Z ((0, k) is the damage criterion of the kth target unit in a perfect state, Y is a penetration effect parameter, C is a shock wave effect parameter, and O is a vibration effect parameter;
the calculation expression of the influence factor a (k) of the kth target unit is a (k) ═ aC(CT，MT)k，AF(FT，MT)k]In the formula, AC(CT，MT)kIs the structural influence factor of the kth target unit; a. theF(FT，MT)kIs the functional impact factor of the kth target unit; cTIs a target structure parameter; mTIs a target material parameter, FTIs a target functional parameter.
Further, the specific operation of performing the bullet matching based on knowledge-graph reasoning in step S2 includes the following steps,
s21, knowledge acquisition: extracting related knowledge of target entities, types, characteristics, damage effects and weapon application principles from expert experience, combat simulation experiments, outfield damage tests, structured, semi-structured and unstructured data of typical combat cases by using an automatic extraction technology to construct a basic knowledge base;
s22, knowledge fusion: structuring the basic knowledge base constructed in the step S21 through links of heterogeneous data integration, knowledge importance calculation and verification to establish a bullet matching rule base;
s23, knowledge storage: by adopting a data structure based on a graph, a flexible and extensible storage mode is designed, the effective storage of a knowledge base and a rule base is realized, and dynamic management is realized;
s24, query analysis: predicting the query intention of the user, classifying according to the feature words, identifying the weapon and target attributes involved in the user query, generating a standard analysis statement, and analyzing the battlefield environment and the fighting intention queried by the user;
s25, knowledge reasoning: and deducing answers inquired by the user by using an inference model and an algorithm according to a knowledge base and a rule base in the knowledge map, and giving a bullet matching scheme sequence.
Further, the specific operation of step S25 includes,
s251: establishing a dynamic key value memory network model according to a graph structure knowledge base and a rule base which are established by a knowledge graph;
s252: the external input data is divided into two parts of key and valueIs divided { (k)i，vi)}i＝1 NThe method comprises the steps of using a static key matrix key as network memory to store all target attributes, battlefield environment and target intention information, using a dynamic value matrix value to store and update weapon matching results, wherein the key is used for addressing, scoring the correlation degree of memory and questions, and the value is used for reading, weighting and summing the memorized values to obtain output, and further realizing the reasoning function.
Further, the specific operation of implementing the inference function in step S252 includes the following steps,
s2521: preprocessing of keys, matching each triplet s in each knowledge graphi，pi，oiChinese subject entity and predicate si+piCorresponds to kiObject entity oiCorresponds to viConverting the knowledge base and the rule base into a form of (k, v);
s2522: selecting a subset from (k, v), the selection conditions of the subset including: k has a common word with the input question and the common word is not a stop word; selecting the top N items according to the number sequence of the common words, constructing a corresponding memory unit vector for each input question, and inputting the memory unit vector mi＝Aφ(ki) Outputting memory cell vector ci＝Cφ(vi)；
S2523: addressing Key Addressing of the keys, calculating a probability distribution for all keys in the memory unit according to the problem, and carrying out correlation scoring on the memory unit; the probability distribution is obtained by Softmax after multiplying the memory of key and the input problem, and the size of the probability indicates the degree of correlation
S2524: reading value reading, calculating the weighted vector of value according to the probability of key
S2525: query updating QueryUpdating, adding the input vector o and the vector representation q of the query input question, passing through RiMatrix linear mapping of inputs qj+1＝Rj(qj+o)；
S2526: with qj+1Replacing the problem vector in Key Addressing as the input of the next layer, and performing iterative updating;
s2527: predicting, after iteration H, adding qH+1Inputting into classifier, predicting answers, and sorting answers to phi＝Softmax(gT j+1AΦK(kki))。
Further, the specific operation of establishing the project model of the consumption of the ammunition based on the deep learning theory in the step S3 and outputting the plan of the multiple-ammunition fast fire power includes the following steps,
s31: establishing a sample library of target attributes, weapon attributes, battlefield environment, operational intention, bullet matching results and various bullet consumptions according to expert experience, simulation, exercise tests and actual combat cases;
s32: dividing samples in a sample library into a training data set and a testing data set, performing learning training by using a deep neural network, and establishing a bomb volume planning model for multi-bomb strike according to a missile target matching result based on knowledge graph reasoning and a damage effect calculation result based on a cloud computing technology;
s33: target attributes, weapon attributes, battlefield environment and operation intentions input by the firepower planning user layer are input into the ammunition amount planning model of multi-ammunition striking established in the step S32, and a bullet matching scheme, various ammunition consumption amounts and damage effect predicted values are output.
The invention has the beneficial effects that:
1. the invention integrates finite element analysis and cloud computing, calculates the damage effect in the multi-bullet fire planning, and can overcome the difficulties of complex and huge calculation process, higher specialty, high cost of finite element software, dependence on high performance computers and the like faced by the damage effect of finite element calculation.
2. The bullet matching method based on knowledge graph reasoning can enable a computer to understand the intention of a user from the semantic perspective and output a bullet matching scheme quickly and intelligently.
3. The improved deep learning model for raising the bomb consumption is an improvement on the traditional heuristic fire planning algorithm represented by a genetic algorithm and a particle group algorithm, can meet the requirement of fast calculation of a nonlinear problem, can avoid the dependence of multi-attribute decision on the weight set by an expert, and provides a technical solution for fast and accurately developing fire planning.
Drawings
FIG. 1 is a diagram of a finite element damage simulation cloud computing platform architecture according to the present invention;
FIG. 2 is a bullet matching implementation flow based on knowledge graph reasoning in the present invention;
FIG. 3 is a knowledge graph inference model of the present invention;
FIG. 4 is a deep learning model for the project of the amount of fuel consumption in the present invention;
FIG. 5 is a graph showing the effect of target damage over time, according to one embodiment of the present invention.
Fig. 6 is a damage effect scatter diagram in a mixed fire power distribution scheme obtained by an enumeration method in the first embodiment of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
A multi-bomb fast fire planning method based on artificial intelligence comprises the following steps,
s1: establishing a finite element damage simulation cloud computing platform, and computing damage effects of multiple bullet types;
the mathematical description of a typical multi-shot fast fire schedule is: given m types of ammunition, the ith type of ammunition has NMi(i is 1, …, m) and the total shot size is NM, and combined shots are performed on n targets. Multi-tuple MBHLijRepresenting the fire distribution scheme of the ith missile to the jth target
In the formula (d)ijRepresenting the amount of the ith bullet allocated to strike the jth target, taking a positive integer or 0, wherein i is 1, …, m; j is 1, …, n;
for given bullet type and quantity, the decision variables of firepower planning problem can use multi-element group MBHLijIs described by the set HL of (a) of (b),
in the tuple array HL, m × n tuples are provided, and each tuple
Containing 1 integer variable, and 2 xdijA real-type variable; the dimension of its variables is very high. If the solution is optimized directly over the domain of all variables, it will fall into the predicament of "dimension explosion". For example, there are A, B, C kinds of shots, each of which has a shot size of 6, and 8 targets are subjected to combined fire power hitting, and the shots coexist in the target regardless of the optimal calculation of the target point of each shot
A fire distribution scheme.
The goal of fire planning problem optimization can generally be described as fully utilizing existing combat resources in pursuit of the best devastating effect on the goal. Under the multi-bullet multi-target striking mode, the aim is to achieve the optimal overall target damage effect. The objective function can be expressed as a functional of each target attribute information, each bullet attribute information, and the mixed fire allocation scheme HL, and max P ═ f (MB)1，…，MBn，DZ1，…，DZmHL); in the formula, MBiRepresenting target parameters needed for calculating the damage effect, i is 1, …, n; DXjRepresents the weapon parameters needed to calculate the effect of the damage, j-1, …, m.
Further, the cloud computing platform for finite element damage simulation comprises a fire planning user layer, a cloud platform application layer, a resource scheduling layer, a model service layer and an infrastructure layer, as shown in fig. 1.
The fire planning user layer provides an operation interface for fire planning auxiliary decision-making personnel, sends damage calculation requests to the cloud computing platform through the fire planning user layer, and adjusts various parameters of weapons and targets according to needs;
the cloud platform application layer comprises model creation, damage simulation and data management, wherein finite element analysis services are dynamically combined to respond to a user request according to the damage calculation requirements of a firepower planning user layer, so that tasks are disassembled, the damage effects of various bullets are calculated, and a foundation is provided for resource scheduling;
the resource scheduling layer allocates enough virtual resources for implementing finite element-based damage calculation, so that the calculation efficiency is improved;
the model service layer comprises finite element mesh division, simulation algorithm service and dynamic model service provided by a cloud computing platform, and the damage computation based on the finite element is realized in a service mode according to a functional module mode;
the base implementation layer includes a computing host, a storage device, and a network device.
Further, the specific operation of the cloud platform application layer for calculating the damage effects of multiple bullet types comprises the following steps,
s11: dividing a hit region omega into M sub-domains delta sigma according to the shape and physical characteristics of the hit region of the warheadm(M-1, 2, …, M), i.e. a unit;
s12: numbering the units and the nodes, determining the relationship between the units and the nodes, and listing natural boundaries and corresponding boundary values;
s13: selecting a corresponding interpolation function as a unit shape function according to the number of the unit nodes in the area and the requirement of calculating an approximate solution, approximating the solving function of each area unit by using a linear combination expression of the unit shape function, substituting the approximate function into an integral equation, integrating the unit area to obtain an algebraic equation containing undetermined coefficients, namely a unit finite element equation set; the undetermined coefficient comprises a target structure parameter, a target material parameter and a target function parameter;
s14: solving by adopting a proper numerical calculation method in a finite element according to the boundary condition to obtain the coefficient value of each node, namely the value of the coefficient to be determined;
s15: warhead hit Δ σmThe damage degree D (m) of the target is determined by the damage degree and vulnerability analysis result of each target unit, and the calculation formula is as follows:
in the formula, K is the number of target units; s (m, k) is the damage degree of the kth target unit, and A (k) is the influence factor of the kth target unit;
wherein the warhead hits Δ σmThe calculation expression of the damage degree S (m, k) of the target unit is S (m, k) ═ S { Z (0, k), Y, C, O }, where Z (0, k) is the damage criterion of the kth target unit in a good state; y is penetration effect parameters (including penetration depth, damage area, pit forming radius, deformation curvature and the like); c is shock wave effect parameters (including overpressure peak value, overpressure duration and the like); o is vibration effect parameters (including acceleration, vibration frequency and the like);
the calculation expression of the influence factor a (k) of the kth target unit is a (k) ═ aC(CT，MT)k，AF(FT，MT)k]In the formula, AC(CT，MT)kIs the structural influence factor of the kth target unit; a. theF(FT，MT)kIs the functional impact factor of the kth target unit; cTIs a target structure parameter; mTFor target material parameters (including density, modulus of elasticity, shear modulus, Poisson's ratio, tensile and compressive failure stress, etc.), FTIs a target functional parameter.
The finite element analysis and the cloud computing are fused, so that the difficulties of complex and huge computing process, higher specialty, high cost of finite element software, dependence on a high-performance computer and the like faced by the damage effect of the finite element computing can be overcome.
Further, step S2: carrying out bullet matching of multiple bullet types based on knowledge graph reasoning;
the knowledge map is generated according to the requirement of artificial intelligence on knowledge, and maps the user query request with entities, attributes and relations in a semantic knowledge base. The knowledge graph has the characteristics of logical reasoning, heterogeneous information association, knowledge visualization, interpretability and the like, and is supported by a large-scale database, so that a computer can understand the user intention from the semantic perspective and intelligently output the query result of the user
For the match of the bullet and the target in the multi-bullet fire striking planning, the knowledge field to be identified contains multidimensional knowledge such as single damage effect dimension, combined damage effect dimension, target entity dimension, target type dimension, target characteristic dimension, weapon application dimension and the like. The single damage effect dimension identification knowledge has penetration effect, killing effect, shock wave effect, combustion effect and the like. The combined damage effect includes killing blasting, penetration blasting, blasting combustion, etc. The target entity dimension has a target type, name, location, etc. The target characteristic dimension includes target material, composition, structure, function, etc. The weapon operation dimension includes penetration bomb and blasting bomb combination, blasting bomb and combustion bomb combination, armor piercing bomb and combustion bomb combination and the like. The knowledge includes applying rule classes or typical case classes to appear in a structured, semi-structured, unstructured form. Under the environment that the scale of the combat data is exponentially increased, reasoning based on the knowledge graph provides a solution for intelligent fire planning bullet matching.
As shown in fig. 2, the specific operation of the bullet matching based on knowledge-graph reasoning includes the following steps,
s21, knowledge acquisition: extracting related knowledge of target entities, types, characteristics, damage effects and weapon application principles from expert experience, combat simulation experiments, outfield damage tests, structured, semi-structured and unstructured data of typical combat cases by using an automatic extraction technology to construct a basic knowledge base;
s22, knowledge fusion: structuring the basic knowledge base constructed in the step S21 through links of heterogeneous data integration, knowledge importance calculation and verification to establish a bullet matching rule base;
s23, knowledge storage: by adopting a data structure based on a graph, a flexible and extensible storage mode is designed, the effective storage of a knowledge base and a rule base is realized, and dynamic management is realized;
s24, query analysis: predicting the query intention of the user, classifying according to the feature words, identifying the weapon and target attributes involved in the user query, generating a standard analysis statement, and analyzing the battlefield environment and the fighting intention queried by the user;
s25, knowledge reasoning: and deducing answers inquired by the user by using an inference model and an algorithm according to a knowledge base and a rule base in the knowledge graph, and giving a bullet matching scheme ranking, wherein the inference model is shown as an attached figure 3.
Further, the specific operations of knowledge inference include,
s251: establishing a dynamic key value memory network model according to a graph structure knowledge base and a rule base which are established by a knowledge graph;
s252: dividing external input data into two parts of key and value { (k)i，vi)}i＝1 NThe method comprises the steps of using a static key matrix key as network memory to store all target attributes, battlefield environment and target intention information, using a dynamic value matrix value to store and update weapon matching results, wherein the key is used for addressing, scoring the correlation degree of memory and questions, and the value is used for reading, weighting and summing the memorized values to obtain output, and further realizing the reasoning function.
The specific algorithm flow comprises the following steps of,
s2521: preprocessing of keys, matching each triplet s in each knowledge graphipi，oiChinese subject entity and predicate si+piCorresponds to kiObject entity oiCorresponds to viConverting the knowledge base and the rule base into a form of (k, v);
s2522: selecting a subset from (k, v), the selection conditions of the subset including: k has a common word with the input question and a common wordThe word is not a stop word; selecting the top N items according to the number sequence of the common words, constructing a corresponding memory unit vector for each input question, and inputting the memory unit vector mi＝Aφ(ki) Outputting memory cell vector ci＝Cφ(vi)；
S2523: addressing Key Addressing of the keys, calculating a probability distribution for all keys in the memory unit according to the problem, and carrying out correlation scoring on the memory unit; the probability distribution is obtained by Softmax after multiplying the memory of key and the input problem, and the size of the probability indicates the degree of correlation
S2524: reading value reading, calculating the weighted vector of value according to the probability of key
S2525: query updating QueryUpdating, adding the input vector o and the vector representation q of the query input question, passing through RiMatrix linear mapping of inputs qj+1＝Rj(qj+o)；
S2526: with qj+1Replacing the problem vector in Key Addressing as the input of the next layer, and performing iterative updating;
s2527: predicting, after iteration H, adding qH+1Inputting into classifier, predicting answers, and sorting answers to phi＝Softmax(qT j+1AΦK(kki))。
Further, step S3: and based on a deep learning theory, establishing a bullet consumption planning model according to the damage effect calculation result in the step S1 and the bullet matching result in the step S2, and outputting the multiple-bullet fast fire power planning.
After the bullet matching scheme is determined, the fire planning as a whole also includes the distribution of the bullet quantity of various bullets to the target, namely the project of the bullet consumption. Under the attack of multiple bombs, the problem is a multi-dimensional nonlinear search and multi-attribute decision problem, no mathematical analytic expression exists, the calculation can be carried out, and the optimal solution is difficult to find. The informationized combat puts higher requirements on the instantaneity of firepower planning, scientific auxiliary decision is implemented, and a large amount of simulation and analog data are fully mined on the basis of fully utilizing the experienced technology by means of an artificial intelligence technology. And (3) learning and training samples by using a deep learning method, and mapping a complex nonlinear relation between output and input. The deep learning-based fuel consumption planning model is shown in fig. 4, and the specific operation comprises the following steps,
s31: establishing a sample library of target attributes, weapon attributes, battlefield environment, operational intention, bullet matching results and various bullet consumptions according to expert experience, simulation, exercise tests and actual combat cases;
s32: dividing samples in a sample library into a training data set and a testing data set, performing learning training by using a deep neural network, and establishing a bomb volume planning model for multi-bomb strike according to a missile target matching result based on knowledge graph reasoning and a damage effect calculation result based on a cloud computing technology;
s33: target attributes, weapon attributes, battlefield environment and operation intentions input by the firepower planning user layer are input into the ammunition amount planning model of multi-ammunition striking established in the step S32, and a bullet matching scheme, various ammunition consumption amounts and damage effect predicted values are output.
The deep learning-based bomb consumption planning method is an improvement of the traditional heuristic firepower planning algorithm represented by a genetic algorithm and a particle swarm algorithm, can meet the requirement of fast calculation of a nonlinear problem, and can avoid the dependence of multi-attribute decision on the weight set by an expert.
The first embodiment is as follows:
assuming that 15 missile weapons with 9 different types are provided, firepower planning is carried out on 8 targets to be hit in an airport, the target types and the missile types are respectively shown in the following tables 1 and 2, and the firepower distribution optimization objective function is optimal in overall effect.
TABLE 1 optional hit target table
| Sub-target number | Name of sub-target | Sub-target number | Name of |
| 1 | Command tower | 5 | Radar station |
| 2 | Oil depot | 6 | Sliding track |
| 3 | Parking apron | 7 | Flight control room |
| 4 | |
8 | Ammunition warehouse |
TABLE 2 bullet type ammunition meter
| Bullet type number | Bullet type | Spring rate |
| I | I-type time-delay bullet | 3 |
| II | II type penetration bomb | 2 |
| III | III type explosive shell | 2 |
| IV | IV type integral bullet | 1 |
| V | V-shaped |
1 |
| Ⅵ | VI type explosive- |
1 |
| Ⅶ | VII type explosive killing bomb | 2 |
| VIII | Type VIII penetration bomb | 2 |
| IX | IX-type |
1 |
For any given fire distribution scheme, as shown in table 3 below, the overall damage effect of the target can be calculated according to the damage effect calculation model, as shown in fig. 5.
TABLE 3 Multi-bomb fire distribution scheme
The hit target numbers and the ranks of the 9 types of bullets suitable for hit are shown in the following table 4 according to the bullet matching knowledge graph reasoning model.
TABLE 4 bullet type bullet quantity and the sub-target list suitable for each bullet type
| Bullet type number | Bullet type | Sub-target number and sequence suitable for striking |
| I | I-type time- |
4，3，1，6 |
| II | II |
4，3，1，6 |
| III | III type |
2，8 |
| IV | IV type |
3，2，8，5，7，1 |
| V | V-shaped |
4，3，1，6 |
| Ⅵ | VI type explosive- |
3，2，8，5，7，1 |
| Ⅶ | VII type |
3，5，1 |
| VIII | Type |
4，3，1，6 |
| IX | IX-type |
3，2，8，5，7，1 |
According to the project model of the bullet consumption, a scheme for distributing the firepower of multiple bullet types under the condition of optimal overall damage effect of the airport is given, and as shown in the following table 5, the overall damage effect of the scheme on target striking is 0.884.
TABLE 5 optimal firepower distribution scheme of multi-bullet type
In order to verify the advancement of the model, 600 mixed fire power distribution schemes are enumerated randomly, and the scheme with the best damage effect on the target is recorded, wherein the corresponding damage effect is HMJi. Repeating the above process 500 times to obtain calculated HMJi(i-1, …, 500) is plotted as a scatter plot, as shown in fig. 6.
And (4) analyzing a calculation result:
as shown in fig. 6, the average damage effect of the mixed fire power distribution scheme obtained by the enumeration method is 0.423, and in 500 enumerations, the effect index value exceeds 0.8 for 22 times, accounts for 4.4%, exceeds 0.88 for 2 times, and accounts for 0.4%. This indicates that the reliability of the same effect as the optimized model obtained by the enumeration method is 0.4% under the similar calculation scale conditions.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are merely illustrative of the principles of the present invention, but that various changes and modifications may be made without departing from the spirit and scope of the invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (8)
1. A multi-bomb fast fire planning method based on artificial intelligence is characterized by comprising the following steps,
s1: establishing a finite element damage simulation cloud computing platform, and computing damage effects of multiple bullet types;
s2: carrying out bullet matching of multiple bullet types based on knowledge graph reasoning;
s3: and based on a deep learning theory, establishing a bullet consumption planning model according to the damage effect calculation result in the step S1 and the bullet matching result in the step S2, and outputting the multiple-bullet fast fire power planning.
2. The artificial intelligence based multiple-missile type fast-fire planning method according to claim 1, wherein the method comprises the following steps: the mathematical description of the multiple-bullet fast fire plan is as follows: given m types of ammunition, the ith type of ammunition has NMi(i is 1, …, m) and the total bullet quantity is NM, and combined striking is carried out on n targets; multi-tuple MBHLijRepresenting the fire distribution scheme of the ith missile to the jth target
In the formula (d)ijRepresenting the amount of the ith bullet allocated to strike the jth target, taking a positive integer or 0, wherein i is 1, …, m; j is 1, …, n;
for given bullet type and quantity, the decision variables of firepower planning problem can use multi-element group MBHLijIs described by the set HL of (a) of (b),
in the multi-element matrix HL, m × n multi-elements are in total, and each multi-element group
Containing 1 integer variable, and 2 xdijA number of real-number variables;
the overall damage effect of the target in the multi-bullet fire planning problem can be expressed as the functional of each target attribute information, each bullet attribute information and the mixed fire distribution scheme HL, and max P ═ f (MB)1，…，MBn，DZ1，…，DZmHL); in the formula, MBiTarget parameters needed for calculating the damage effect are represented, i is 1, …, n; DXjRepresents the weapon parameters needed to calculate the effect of the damage, j-1, …, m.
3. The artificial intelligence based multiple-missile type fast-fire planning method according to claim 2, wherein the method comprises the following steps: the finite element damage simulation cloud computing platform in the step S1 comprises a firepower planning user layer, a cloud platform application layer, a resource scheduling layer, a model service layer and an infrastructure layer;
the fire planning user layer provides an operation interface for fire planning auxiliary decision-making personnel, sends damage calculation requests to the cloud computing platform through the fire planning user layer, and adjusts various parameters of weapons and targets according to needs;
the cloud platform application layer comprises model creation, damage simulation and data management, wherein finite element analysis services are dynamically combined to respond to a user request according to the damage calculation requirements of a firepower planning user layer, so that tasks are disassembled, the damage effects of various bullets are calculated, and a foundation is provided for resource scheduling;
the resource scheduling layer allocates enough virtual resources for implementing finite element-based damage calculation, so that the calculation efficiency is improved;
the model service layer comprises finite element mesh division, simulation algorithm service and dynamic model service provided by a cloud computing platform, and the damage calculation based on the finite element is realized in a service mode according to a functional module mode;
the base implementation layer includes a computing host, a storage device, and a network device.
4. The artificial intelligence based multiple bullet fast fire planning method according to claim 3, wherein said specific operation of computing the damage effect of multiple bullets by said cloud platform application layer comprises the following steps,
s11: dividing the hit region omega into M subdomains delta sigma according to the shape and physical characteristics of the hit region of the warheadm(M-1, 2, …, M), i.e. a unit;
s12: numbering the units and the nodes, determining the relationship between the units and the nodes, and listing natural boundaries and corresponding boundary values;
s13: selecting a corresponding interpolation function as a unit shape function according to the number of the unit nodes in the area and the requirement of calculating an approximate solution, approximating the solving function of each area unit by using a linear combination expression of the unit shape function, substituting the approximate function into an integral equation, integrating the unit area to obtain an algebraic equation containing undetermined coefficients, namely a unit finite element equation set; the undetermined coefficient comprises a target structure parameter, a target material parameter and a target function parameter;
s14: solving by adopting a proper numerical calculation method in a finite element according to the boundary condition to obtain the coefficient value of each node, namely the value of the coefficient to be determined;
s15: warhead hit Δ σmThe damage degree D (m) of the target is determined by the damage degree and vulnerability analysis result of each target unit, and the calculation formula is as follows:
in the formula, K is the number of target units; s (m, k) is the damage degree of the kth target unit, and A (k) is the influence factor of the kth target unit;
wherein the warhead hits Δ σmThe calculation expression of the damage degree S (m, k) of the target unit is S (m, k) ═ S { Z (0, k), Y, C, O }, where Z (0, k) is the damage criterion of the kth target unit in a good state; y is a penetration effect parameter; c is a shock wave effect parameter; o is a vibration effect parameter;
the calculation expression of the influence factor a (k) of the kth target unit is a (k) ═ ac(CT，MT)k，AF(FT，MT)k]In the formula, AC(CT，MT)kIs the structural influence factor of the kth target unit; a. theF(FT，MT)kIs the functional impact factor of the kth target unit; cTIs a target structure parameter; mTIs a target material parameter, FTIs a target functional parameter.
5. The artificial intelligence based multiple-missile type fast-fire planning method according to claim 2, wherein the concrete operation of performing missile matching based on knowledge-graph reasoning in the step S2 comprises the following steps,
s21, knowledge acquisition: extracting related knowledge of target entities, types, characteristics, damage effects and weapon application principles from expert experience, combat simulation experiments, outfield damage tests, structured, semi-structured and unstructured data of typical combat cases by using an automatic extraction technology to construct a basic knowledge base;
s22, knowledge fusion: structuring the basic knowledge base constructed in the step S21 through links of heterogeneous data integration, knowledge importance calculation and verification to establish a bullet matching rule base;
s23, knowledge storage: by adopting a data structure based on a graph, a flexible and extensible storage mode is designed, the effective storage of a knowledge base and a rule base is realized, and dynamic management is realized;
s24, query analysis: predicting the query intention of the user, classifying according to the feature words, identifying the weapon and target attributes involved in the user query, generating a standard analysis statement, and analyzing the battlefield environment and the fighting intention queried by the user;
s25, knowledge reasoning: and deducing answers inquired by the user by using an inference model and an algorithm according to a knowledge base and a rule base in the knowledge map, and giving a bullet matching scheme sequence.
6. The artificial intelligence based multi-missile type fast fire planning method of claim 5, wherein the specific operation of the step S25 comprises,
s251: establishing a dynamic key value memory network model according to a graph structure knowledge base and a rule base which are established by a knowledge graph;
s252: dividing external input data into two parts of key and value { (k)i，vi)}i＝1 NThe method comprises the steps of using a static key matrix key as network memory to store all target attributes, battlefield environment and target intention information, using a dynamic value matrix value to store and update weapon matching results, wherein the key is used for addressing, scoring the correlation degree of memory and problems, and the value is used for reading, and performing weighted summation on the memorized values to obtain output, thereby realizing the reasoning function.
7. The artificial intelligence based multi-missile type fast fire planning method of claim 6, wherein the specific operation of implementing the reasoning function in the step S252 comprises the following steps,
s2521: preprocessing of keys, matching each triplet s in each knowledge graphi，pi，oiSubject entities and predicates ini+piCorresponds to kiObject entity oiCorresponds to viConverting the knowledge base and the rule base into a form of (k, v);
s2522: selecting a subset from (k, v), the selection conditions of the subset including: k has a common word with the input question and the common word is not a stop word; sorting according to the number of common words, selecting the top N items, constructing corresponding memory unit vector for each input question, and inputting the memory unit vector mi＝Aφ(ki) Outputting the memory cell vector ci＝Cφ(vi)；
S2523: addressing Key Addressing of the Key, calculating a probability distribution for all keys in the memory unit according to the problem, and carrying out correlation scoring on the memory unit; the probability distribution is obtained by Softmax after multiplying the memory of key and the input problem, and the size of the probability indicates the degree of correlation
S2524: reading value reading, calculating the weighted vector of value according to the probability of key
S2525: query updating QueryUpdating, adding the input vector o and the vector representation q of the query input question, passing through RiMatrix linear mapping of inputs qj+1＝Rj(qj+o)；
S2526: with qj+1Replacing the problem vector in Key Addressing as the input of the next layer, and performing iterative updating;
s2527: predicting, after iteration H, adding qH+1Inputting into classifier, predicting answers, and sorting answers to phi＝Softmax(qT j+1AΦK(kki))。
8. The artificial intelligence-based multi-bomb fast fire planning method according to claim 2, wherein in step S3, based on deep learning theory, a bomb consumption planning model is established, and the specific operation of outputting the multi-bomb fast fire planning includes the following steps,
s31: establishing a sample library of target attributes, weapon attributes, battlefield environment, fighting intentions, bullet matching results and various bullet consumptions according to expert experience, simulation, exercise tests and actual combat cases;
s32: dividing samples in a sample library into a training data set and a testing data set, performing learning training by using a deep neural network, and establishing a bomb volume planning model for multi-bomb strike according to a missile target matching result based on knowledge graph reasoning and a damage effect calculation result based on a cloud computing technology;
s33: target attributes, weapon attributes, battlefield environment and operation intentions input by the firepower planning user layer are input into the ammunition amount planning model of multi-ammunition striking established in the step S32, and a bullet matching scheme, various ammunition consumption amounts and damage effect predicted values are output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110045002.2A CN114372385A (en) | 2021-01-13 | 2021-01-13 | Artificial intelligence-based multi-bullet fast fire planning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110045002.2A CN114372385A (en) | 2021-01-13 | 2021-01-13 | Artificial intelligence-based multi-bullet fast fire planning method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114372385A true CN114372385A (en) | 2022-04-19 |
Family
ID=81138938
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110045002.2A Pending CN114372385A (en) | 2021-01-13 | 2021-01-13 | Artificial intelligence-based multi-bullet fast fire planning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114372385A (en) |
-
2021
- 2021-01-13 CN CN202110045002.2A patent/CN114372385A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Przemieniecki | Mathematical methods in defense analyses | |
| CN108615122A (en) | A kind of air-defense anti-missile system combat capability assessment method | |
| EP2144032B1 (en) | Method and apparatus for analysis of errors, accuracy, and precision of guns and direct and indirect fire control mechanisms | |
| CN109614729A (en) | A kind of equipment Efficacy assessment result rapid analysis method based on meta-model | |
| Hu et al. | Improved Ant Colony Optimization for Weapon‐Target Assignment | |
| CN113627749B (en) | Damage efficiency evaluation and fire planning method based on convolutional neural network | |
| CN104392087A (en) | Performance evaluation method of overhead weapon station | |
| Tianle et al. | Synthetic damage effect assessment through evidential reasoning approach and neural fuzzy inference: Application in ship target | |
| Wang et al. | An uncertain optimization method for overall ballistics based on stochastic programming and a neural network surrogate model | |
| Erdal et al. | Evaluation of Anti-Tank Guided Missiles: An integrated Fuzzy Entropy and Fuzzy CoCoSo multi criteria methodology using technical and simulation data | |
| CN114926026B (en) | Target distribution optimization method for multi-dimensional feature deep learning | |
| Li et al. | Wartime industrial logistics information integration: Framework and application in optimizing deployment and formation of military logistics platforms | |
| CN113177583B (en) | Aerial target clustering method | |
| CN112163719B (en) | Missile equipment system missile variety optimization design method and device based on assignment model | |
| CN114372385A (en) | Artificial intelligence-based multi-bullet fast fire planning method | |
| Dong et al. | An efficient spatial high-utility occupancy frequent item mining algorithm for mission system integration architecture design using the MBSE method | |
| Lytvyn et al. | Ontology Using for Decision Making in a Competitive Environment | |
| CN113837644A (en) | Equipment combat effectiveness and contribution rate integrated evaluation method based on grey correlation | |
| Li et al. | Position deployment optimization of maneuvering conventional missile based on improved whale optimization algorithm | |
| CN110930054A (en) | Data-driven battle system key parameter rapid optimization method | |
| Yang et al. | Effectiveness evaluation of fire control system based on improved PSO algorithm and LSSVM algorithm | |
| Zhang | Research on Weighted Model of Air Defense Weapon-Target Assignment Based on AHP and Entropy Method | |
| Li et al. | A target damage effectiveness assessment mathematical calculation method with uncertain information based on an adaptive fuzzy neural network | |
| Jiaxin et al. | AHP-Based Interactive Target Threat Level Assessment and Visualization Method | |
| Lytvyn et al. | Decision support methods in a competitive environment based on Boyd cycle by means of ontology use |
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 |