CN104504440A - Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement - Google Patents
Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement Download PDFInfo
- Publication number
- CN104504440A CN104504440A CN201410424091.1A CN201410424091A CN104504440A CN 104504440 A CN104504440 A CN 104504440A CN 201410424091 A CN201410424091 A CN 201410424091A CN 104504440 A CN104504440 A CN 104504440A
- Authority
- CN
- China
- Prior art keywords
- model
- formula
- consumption
- ammunition
- prediction
- 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
Abstract
The invention discloses a method and a system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement. The system comprises an expert knowledge input module, a prediction structure constructing module, a model base, a data acquisition module, a prediction model matching module, a differentiated structure output module, a model modification module and a consumption computation module. The method corresponds to the system, an ammunition consumption interpretation structure is constructed by an expert, a collected prediction context interpretation structure is matched with an interpretation structure corresponding to a formula, and a model is selected from the model base. A modification rule is constructed according to a differentiated structure so as to modify formula parameters or consumption, and the model is stored in the model base, so that an ammunition consumption structure is adjusted to complete predication of the ammunition consumption. According to the method and the system, an uncertain knowledge structure is enabled to be clear, so that the precision for predicting ammunition consumption is improved.
Description
Technical field
The present invention relates to prognoses system and method thereof, more particularly relevant with the prediction of ammunition consumption, particularly relate to a kind of ammunition consumption prognoses system based on ISM knowledge refining; The present invention relates to the method for above-mentioned a kind of ammunition consumption prognoses system based on ISM knowledge refining simultaneously.
Background technology
Prediction is an important basic fundamental, and carrying out ammunition consumption prediction accurately, reliably, is efficiently the key carrying out Wartime Ammunition Support work.Ammunition consumption is different from the general consumption of materials, and its prediction context is complicated, and influence factor is many, requires that accuracy is high.Current ammunition consumption to be predicted, through finding the retrieval of existing document, paper [Wu Wei, Wu Lin, fourth illumination. battlefield ammunition consumption prediction and calculation present Research is analyzed, accelerate to advance the modernization of national defense and the army and military systems engineering, pp:273-276], [Fan Shengli, Tian Weifeng, Bai Yanqi. ground main battle weapons conventional ammunition consumption forecast technique study present situation, equipment Command technical college journal, 2010, 21 (5), pp:106-110] [Fan Shengli, Bai Yanqi, Zhang Yaokun, Yao Tao, towards the ammunition consumption Intelligent Forecasting of Military Equipment Battling, 2011 (2), 25 (1), 22-27] all conclude, the prior art concluding ammunition in hand consumption forecast system and method has:
Prior art one: based on injuring theory, by Forecasting Methodology and the system of analytic formula.
Also find through retrieval, document [Yang Xuming, ground force's Damage Fire study course [M], publishing house of PLA .2007] give the ammunition consumption predictor formula of based target task amount, take into account ammunition power, killing zone, target importance, the target easily factor such as ruining property, location territory, damage fire.[Williams Freeman Jr.Study of AmmunitionConsumption (AD-A451782) [R] .Washington, D.C., 2005] give capability-based ammunition predictor formula, consider our strength, the plan factor, categories of operations.[Wang Sanxi, Yu Jie, the Xiaxin people, joint fire attack Missile requirement calculates dynamic model research, Journal of System Simulation, 2009,21 (9): 2734-2736] use lanping county method to calculate, take into account the factors such as troops' loss in time, weapon firing rate, operation duration.
Corresponding to above-mentioned Forecasting Methodology, prognoses system is in structure function, emphasize to comprise prediction module, will carry out consumption law as master data input database from historical experience, the model of setting (being often analytic formula) is used to predict, these system prediction modules formula that uses a model considers that parameter is few, just can using formula calculating by collecting a small amount of information to simple question.But though the factor relation that formula reflects is by matching, the thought process in matching is difficult to represent, so when constraint condition changes, formula just loses feasibility.A specific formula can provide very accurate result under one group of operating conditions, but provides not too accurate result under another group operating conditions.The change of several factors may cause such uncertainty in traffic, and such as, the data provided may comprise noise, and the condition of task is different.To its revise, very difficult, especially when many factors, be interweaved time, more seldom arrive valuable amendment scheme.The environmental suitability of formula method to dynamic change is poor, has specific application background to limit, narrow application range.And ammunition consumption prediction response environment is complicated, often need to tackle some emergency case, unknown situation, once situation changes, formula model is just no longer applicable.Prognoses system structure based on formula model is simple, and prediction lacks environmental suitability.
Prior art two: based on Forecasting Methodology and the system of intelligent algorithm
Document [Qin Xiangyu, Song Yizhong, Li Wensheng, based on the low consumed ammunition stock modeling and analyzing of Markov forecast techniques, Journal of Ordnance Engineering College, 2005,17 (4): 61-63], [Cui Yuquan, model of battlefield ammunition consumption forecast is studied, systems engineering and electronic technology, 2007,29 (4): 585-588] Time series analysis method is adopted, according to the similarity of historical experience, the patterns of warfare, weapon type and the characteristics of operation, describe the correlativity of each stage, time interval ammunition consumption, and next stage is predicted.
Document [Liu Tao, Peng is hereditary. the application of improved BP in aircraft ammunition prediction, detection and control journal, 2009, 31 (5), 52-56], [Cui's winter, Korean-Chinese heptan, ammunition consumption based on grey radial basis function network model is predicted], [Huang great Chen, Han Jingcai, Ye Muqing etc., ammunition consumption based on neural network is predicted, Journal of Ordnance Engineering College, 2004 (4), 30-34] all neural network is introduced ammunition consumption prediction, system introduces the study module relying on historical sample data, can according to nonlinear function implicit between training sample set matching inputoutput data, there is stronger adaptive capacity to environment.
But intelligent method usually needs a large amount of training datasets.It learning ability possessed, makes the accuracy of model improve, but the knowledge obtained does not have high illustrative, and inferior position is, when many parameters need adjustment, to be difficult to the immanent structure determining to consume.When environment changes, directly can not be used in prediction next time.The demand that instant prediction, Accurate Prediction are emphasized in emergency materials prediction can not be met.
Use the ammunition consumption prognoses system of intelligent method, training data usually relies on off-line collection, and its data processing, learning process also normally off-line procedure, expend time in length.Study module complicated operation, practicality is not strong.
Generally speaking, under system combat condition, each operation key element interdependent property strengthens, and combat system internal relations is complicated, and making affects non-linearization and complicated to Wartime Ammunition Support., there is accuracy rate not high, the problem of bad adaptability in said method and take the ammunition consumption prognoses system of above Forecasting Methodology or simple comprehensive a few person, is difficult to adapt to the needs of the Wartime Ammunition Support under new historical conditions.
Summary of the invention
The present invention tries hard to overcome existing method and system, and to there is accuracy rate not high, the problem of bad adaptability, a kind of ammunition consumption prognoses system based on ISM knowledge refining is provided, also has a kind of method of the ammunition consumption prognoses system based on ISM knowledge refining.
Technical scheme: the present invention is achieved through the following technical solutions, the embodiment of the present invention provides a kind of ammunition consumption prognoses system based on ISM knowledge refining, and this prognoses system comprises following ingredient:
Expertise load module 1, for providing the relation between the influence factor of man-machine interface for expert typing ammunition consumption and influence factor.
Predict builds module 2, explains structure tree for the overall situation being formed ammunition consumption by interpretative structure-modeling (ISM, Interpretative Structural Modeling).
Model bank 3, for storing default consumption forecast model, model shows as the formula Sum fanction that a group determines input, output parameter, for the ammunition consumption prediction under different context.Model bank along with the domination of Elements Of Expense causality knowledge, and can adjust, and changes rule and formula in model bank.
Data acquisition module 4, for obtaining data needed for prediction and the contextual supplemental characteristic of phone predicts.
Forecast model matching module 5, for prediction context being organized into interpretative structural modeling tree, the model in model bank is related to parameter and be also organized into interpretative structural modeling tree, both mate, and according to the situation of coupling, choose formula model the most close for prediction.
Differentiation structure output module 6, the overall situation for not yet relating in model bank formula model explains that a part for structure tree is as differentiation structure output.
Modifying model module 7, for data analysis, obtains the modification rule of predictor formula model, and to the modification rule that entirety predicts the outcome.
Consumption computing module 8, calculates consumption for using formula model the most close and modification rule.
Further, forecast model matching module is the matching degree utilizing the similarity calculating method of tree to judge two models.
Further, Modifying model module, adopts the method such as time series analysis, neural network to obtain modification rule to data.
The method of this prognoses system comprises the following steps:
Step S1: build the overall situation of the ammunition consumption based on ISM and explain structure tree;
Step S2: the prediction context of consumption forecast computing formula is used interpretative structural modeling tree representation, and stored in model bank;
Step S3: by the phone predicts context of image data, uses interpretative structural modeling tree representation;
Step S4: will predict that contextual interpretative structural modeling is set, correspondingly with formula in model bank explains that structure tree mates;
Step S5: the model in traversal model bank, obtains the formula model of the highest matching degree.Specify the corresponding formula of the highest matching degree to be target formula, choose formula for prediction;
Step S6: the overall situation not yet related in model formation is explained that a part for structure tree is as differentiation structure output;
Step S7: using parameter in differentiation structure as factor, to data analysis, obtain the modification rule of predictor formula model, and to the modification rule that entirety predicts the outcome, and for the structure of model in model bank and amendment;
Step S8: by the optimum prediction formula after improving, prediction ammunition consumption.Consume modified result rule if existed, utilize rule to revise premeasuring;
Step S9: expert analyzes predicting the outcome, forms the expertise that some are new.
Described step S1 builds an ammunition consumption interpretative structural modeling based on ISM and sets, and its sub-step comprises:
Step S1-1: list one group of parameter [P
1, P
2... P
n], P
1essential, be set to ammunition consumption, other parameter predicts relevant variable with ammunition consumption;
Step S1-2: parameter compared one by one, for P
i, P
j, (i, j=1,2 ... n), if both have direct relation, (P is made
i, P
j)=1, if both are without direct relation, then (P
i, P
j)=0.Adjacency matrix A is set up with this;
Step S1-3: use M=(A+I)
t, obtain the reachability matrix M of factor of system.Wherein I is and the unit matrix of A with order;
Step S1-4: divide element level, reachable set, in advance collection and jointly collect.Reachable set R (P
i) be parameter P
ithe elements combination that can arrive; Leading collection A (P
i) be to arrive key element P
ielements combination; Common collection and R ⌒ A;
Step S1-5: arranging the highest element is P
1, the highest element except can arrive oneself itself except, other key element can not be arrived;
Step S1-6: after finding out highest key element collection, can scratch corresponding row and column by it from reachability matrix, then, then continues to find new highest key element (i.e. second layer key element) from remaining reachability matrix.The like go out each layer key element collection, set up escort tree-shaped hierarchical structure from one place to another.
Described step S5, by match selection model, specifically performs following sub-step:
Step S5-1: all formula in traversal model bank, obtains the interpretative structural modeling tree that each formula correspondence builds.
Step S5-2: computing formula interpretative structural modeling tree and the matching degree predicting context interpretation structure tree;
Step S5-3., according to threshold value constraint, after traversal, finds the formula model existing and meet threshold condition, and obtains the corresponding model of the highest matching degree;
Step S5-4., for the corresponding model of the highest matching degree, points out the data that need gather, if can provide data, using this model as target prediction model, otherwise just gives up this model.
Step S5-5: terminate.
Further,
Step S4, adopt the computing formula based on the similarity of tree construction to calculate matching degree, its rule is higher the closer to the node weights of root, and the node weights more in branch is lower.The similarity of a node is calculated by the number of structural father node, child node, the brotgher of node.Total similarity is calculated by each node similarity.Comprise following sub-step:
Step S4-1: first find root node.If there is minor matters point, then the Zhi Jiedian number of the Zhi Jiedian number/total of the similarity=identical of root node;
Step S4-2: recurrence finds minor matters point.Similarity Measure=1/3 of minor matters point (the subclass number/subclass number of identical parent number/parent number+identical+identical adjacent class number/adjacent class number);
Step S4-3: total similarity=each node weights × each node similarity.
Further,
Step S7, adopts a kind of modification method based on neural metwork training, and correspondingly affects ammunition consumption structure, and the concrete step that performs comprises:
Step S7-1: as forecast model does not train, then perform S7-2, otherwise redirect S7-6;
Step S7-2: the quantification of ammunition consumption factor of influence and normalized;
Step S7-3: build neural network model, set up the Three-tider architecture of input layer, hidden layer, output layer;
Step S7-4: utilize initial training to assemble for training and practise a neural network;
Step S7-5: utilize neural network to produce modification rule from data;
Step S7-6: terminate.
The present invention solves the principle that its technical matters adopts:
(1) with ISM for semantic bridge, the ammunition consumption structural model of constructed formation, has abundant semantic, promote understanding model bank formula and modification rule.
(2) based on the Model Selection of semantic matches.By the prediction context interpretation structure tree gathered, correspondingly with formula explain that structure tree mates, by semantic matching algorithm, carry out automatic Model Selection, can make a choice according to different situation, adaptability improves greatly.
(3) prediction adopts the combined prediction that analytic formula and modification rule combine.In model bank, comprise one group of analytic formula, these formula are some standardized model units, in addition, also comprise the auxiliary modification rule that some obtain based on intelligent method, these rules in order to be reflected in single Elements Of Expense impact under, to the modification rule of certain formula or result, during prediction ammunition consumption, often according to calculating context, combine multiple analytic formula and auxiliary modification rule to form forecast model chain and carry out combined prediction, improve the accuracy of prediction.
(4) the consumption s tructure adjustment of knowledge based domination process.
The structure of consumption s tructure needs expert to participate in, the part determined will be comparatively easily solidified in expertise, in order to form consumption s tructure tree and computation model, for the factor not easily determined, reasoning study is carried out by neural network, time series, and in addition analysis expert, further distinct Elements Of Expense role, revises consumption s tructure tree and model in model bank.Repeatedly carry out in this loop iteration, improve constantly precision.
This method technique effect or advantage: be to provide a kind of adaptability high, the prediction modeling method of high precision, high intelligibility and system.Advantage is that 1. adaptability is good, under multiple condition being adapted to, and the demand of ammunition consumption prediction.2. high precision, the multiple method of prediction combination, through multilayer correction, composite factor is many, and overall precision is high.3. high illustrative, rule, the model of system are set forth through semantic, more meet the cognitive experience of battlefield user of service, and model are convenient to checking.Its technique effect is actively, significantly.
Accompanying drawing explanation
A kind of ammunition consumption prognoses system based on ISM knowledge refining of Fig. 1 forms.
The time series chart of a kind of ammunition consumption prognoses system information transmission based on ISM knowledge refining of Fig. 2.
A kind of ammunition consumption Forecasting Methodology process flow diagram based on ISM knowledge refining of Fig. 3
Fig. 4 builds the matrix diagram of ammunition consumption interpretative structural modeling tree
Fig. 5 ammunition consumption interpretative structural modeling tree example
The interpretative structural modeling tree that Fig. 6 obtains respectively according to model and prediction context
Fig. 7 uses neural network to obtain the method flow diagram of modification rule
Embodiment
Described in detail below in conjunction with accompanying drawing, should be pointed out that described embodiment is only intended to the understanding of the present invention, do not played any restriction effect.
As shown in Figure 1, described ammunition consumption prognoses system comprises:
Expertise load module 1, for providing the relation between the influence factor of man-machine interface for expert typing ammunition consumption and influence factor.
Predict builds module 2, explains structure tree for the overall situation being formed ammunition consumption by interpretative structure-modeling (ISM, Interpretative Structural Modeling).
Model bank 3, for storing default consumption forecast model, model shows as the formula Sum fanction that a group determines input, output parameter, for the ammunition consumption prediction under different context.Model bank along with the domination of Elements Of Expense causality knowledge, and can adjust, and changes rule and formula in model bank.
Data acquisition module 4, for obtaining supplemental characteristic needed for prediction and the contextual supplemental characteristic of phone predicts.
Forecast model matching module 5, for prediction context parameters being organized into interpretative structural modeling tree, the model in model bank is related to parameter and be also organized into interpretative structural modeling tree, both mate, according to the situation of coupling, choose formula model the most close for prediction.
Differentiation structure output module 6, the overall situation for not yet relating in formula model explains that a part for structure tree is as differentiation structure output.
Modifying model module 7, for data analysis, obtains the modification rule of forecast model, and to the modification rule that entirety predicts the outcome.
Consumption computing module 8, calculates consumption for using formula model the most close and modification rule.
As Fig. 2, being mainly operating as of described system;
Consumption law is expressed as the relation between the influence factor of expert's typing ammunition consumption and influence factor by expertise load module 1, and passes to predict and build module 2;
Data acquisition module 4 obtains the supplemental characteristic needed for predicting and the contextual supplemental characteristic of phone predicts, and contextual for prediction supplemental characteristic part is passed to predict structure module 2;
Model information is passed to predict and builds module 2 by model bank 3;
Predict builds module 2 and uses interpretative structure-modeling to build respectively the information that expertise load module 1, data acquisition module 4, model bank 3 are transmitted: the overall situation explains structure tree L, prediction context interpretation structure tree M, corresponding one group of interpretative structural modeling tree N [1...k] of formula model, and send one group of interpretative structural modeling tree N [1...k] produced by model information to model bank;
Model bank 3 is by corresponding with the model in model bank for one group of interpretative structural modeling tree N [1...k].
Forecast model matching module 5 obtains Model Structure Tree Ni from model bank, 1≤i≤k, with the interpretative structural modeling tree M obtained according to prediction context, both mate, judge whether the highest matching degree is greater than threshold value, specify the corresponding formula of the highest matching degree to be target formula, choose formula for prediction.
Differentiation structure output module 6 using in model formation still for the part of the interpretative structural modeling tree L related to exports as differentiation structure K, and pass to expertise load module 1.
Use intelligent method to obtain auxiliary modification rule in conjunction with expertise, these rules in order under being reflected in the impact of single Elements Of Expense, to modifying factor and the algorithm of certain formula or result.Stored in model bank 3, consumption computing module 8 will will be passed to the correction of result to the correction of publicity.
Consumption computing module 8, the data obtained by data acquisition module 4 substitute into the target formula that model bank 3 provides, and revise result, obtain the consumption forecast amount of ammunition.
Fig. 3 is the method for the ammunition consumption prognoses system based on ISM knowledge refining.Comprise the following steps:
Step S1: build the overall situation of the ammunition consumption based on ISM and explain structure tree L.
Step S2: use interpretative structural modeling tree N [1...k] to represent consumption forecast formula, and stored in model bank.
Step S3: by contextual for the phone predicts in image data supplemental characteristic, uses interpretative structural modeling tree M to represent.
Step S4: contextual interpretative structural modeling tree M will be predicted, correspondingly with formula in model bank explain that structure tree N mates.
Step S5: the model in traversal model bank, obtains the formula model of the highest matching degree.Specify the corresponding formula of the highest matching degree to be target formula, choose formula for prediction.
Step S6: the overall situation not yet related in formula model is explained that a part of structure tree L is as differentiation structure output.
Step S7: using parameter in differentiation structure as factor, to data analysis, obtain the modification rule of predictor formula model, and to the modification rule that entirety predicts the outcome, and for the structure of model in model bank and amendment.
Step S8: by the optimum prediction formula model after improving, prediction ammunition consumption.Consume modified result rule if existed, utilize rule to revise premeasuring.
Step S9: expert analyzes predicting the outcome, forms the expertise that some are new.
Step S1 builds an ammunition consumption interpretative structural modeling based on ISM and sets, and its sub-step comprises:
Step S1-1: list one group of parameter [P
1, P
2... P
n], P
1essential, and be set to ammunition consumption, other parameter predicts relevant parameter with ammunition consumption.
Step S1-2: parameter compared one by one, for P
i, P
j, (i, j=1,2 ... n), if both have direct relation, (P is made
i, P
j)=1, if both are without direct relation, then (P
i, P
j)=0.Adjacency matrix A is set up with this.
Step S1-3: use M=(A+I)
t, obtain the reachability matrix M of factor of system.Wherein I is and the unit matrix of A with order.
Step S1-4: divide element level, reachable set, in advance collection and jointly collect.Reachable set R (P
i) be parameter P
ithe elements combination that can arrive; Leading collection A (P
i) be to arrive key element P
ielements combination; Common collection and R ⌒ A.
Step S1-5: arranging the highest element is P
1, the highest element except can arrive oneself itself except, other key element can not be arrived.
Step S1-6: after finding out highest key element collection, can scratch corresponding row and column by it from reachability matrix, then, then continues to find new highest key element (i.e. second layer key element) from remaining reachability matrix.The like go out each layer key element collection, set up pass the tree-shaped hierarchical structure in rank.
In an embodiment of the present invention, to the demand forecast of artillery ammunition, in step S1-1, the influence factor P that expert inputs Wartime Ammunition Support demand is:
[P
1: Missile requirement, P
2: enemy's damage index, P
3: my viability, P
4: ammunition performance,
P
5: hit tactics, P
6: battlefield surroundings, P
7: attack precision, P
8: weapon firing rate,
P
9: enemy and we contrast, P
10: general assignment amount, P
11: target area, P
12: target property]
Fig. 4 is in step S1-2, and key element compares by step S1-3 one by one, considers node relationships, the adjacency matrix (left figure) of foundation and the reachability matrix (right figure) calculated.Again through step S1-4, step S1-5, the layering key element obtained after step S1-6 is:
Ground floor key element is { P
1,
Second layer factor is { P
10,
Third layer factor is { P
2, P
3, P
11}
4th layer of factor is { P
5, P
12}
Layer 5 factor is { P
4, P
6, P
7, P
8, P
9}
Fig. 5 corresponding passs the tree-shaped hierarchical structure L in rank.
The technological means that step S2, S3 obtain interpretative structural modeling tree M, N is respectively identical with S1.
In an embodiment of the present invention, to the demand forecast of artillery ammunition, there is one group of formula in model bank, the left figure of Fig. 6 is that in model bank, certain artillery weapon calculates explanation tree structure Ni corresponding to ammunition consumption formula i, 1≤i≤k.
This formulae discovery Suppressed Weapons task based access control amount method, N=∑ N
0p
ik
i, and
n is Suppressed Weapons ammunition consumption; N
0for reaching the standard ammunition consumption of required damage index to target; Pi is the task share of i-th kind of weapon; Ki is the reduction coefficient of standard ammunition to i-th kind of weapons and ammunitions.N
0according to the area m of a kth target
k, a kth target is to the reduction coefficient K of standard target
mk, damage index needed for a kth target is to the reduction coefficient K of standard damage index
hkwith the consumption of the per hectare standard ammunition of standard target accurate damage index up to standard
the implication of its interpretative structural modeling tree is, ammunition consumption P1 is determined by general assignment amount P10, general assignment amount P10 determines (the ammunition allocation proportion that P4 causes is different) by enemy's damage index P2 with P11 target area and by ammunition performance P4 again, which reflects the basic meaning of formula.
Formula used is well known in the art, but is different from and directly uses formula to carry out the conventional method calculated, and before use formula, formula is corresponded to semantic context structure.The expertise of formula reflection is behind presented in a kind of clear mode.
The right figure of Fig. 6 is the tree structure M that prediction context is corresponding.Prediction context reflects the implication representated by data that can collect.Namely P9 is compared by enemy's damage index P2, P11 target area and enemy and we troops.
Described step S5 match selection model, the following sub-step of concrete execution:
Step S5-1: all formula in traversal model bank, obtains the interpretative structural modeling tree that each formula correspondence builds.
Step S5-2: computing formula interpretative structural modeling tree and the matching degree predicting context interpretation structure tree;
Step S5-3., according to threshold value constraint, after traversal, finds the formula model existing and meet threshold condition, and obtains the corresponding model of the highest matching degree;
Step S5-4., for the corresponding model of the highest matching degree, points out the data that need gather, if can provide data, using this model as target prediction model, otherwise just gives up this model.
Step S5-5: terminate.
The matching process of some semantemes can be adopted the calculating of matching degree, such as adopt the number that vocabulary occurs.In an embodiment of the present invention, to the demand forecast of artillery ammunition, the structure of knowledge of the context shown in Fig. 6 and formula is mated, adopts set A=[P1, P2, P11, P10, P4], B=[P1, P10, P2, P11, P4, P9].The former is the common factor that in two trees, vocabulary occurs, the latter is the union that in two trees, vocabulary occurs.Then matching degree=A/B=60%.Find by contrasting other model, this numerical value is greater than threshold value, and formula is the highest model of matching degree in model bank.
But this method ignores architectural feature, as a kind of preference, consider tree structure, especially, step S4, adopts the computing formula based on the similarity of tree construction to calculate matching degree, its rule is higher the closer to the node weights of root, and the node weights more in branch is lower.The similarity of a node is calculated by the number of structural father node, child node, the brotgher of node.Total similarity is calculated by each node similarity.Comprise following sub-step:
Step S4-1: first find root node.If there is minor matters point, then the Zhi Jiedian number of the Zhi Jiedian number/total of the similarity=identical of root node;
Step S4-2: recurrence finds minor matters point.Similarity Measure=1/3 of minor matters point (the subclass number/subclass number of identical parent number/parent number+identical+identical adjacent class number/adjacent class number);
Step S4-3: total similarity=each node weights × each node similarity.
The closer to the node of root, ammunition consumption plays a part more important affecting, more can not ignore, contrary more away from the node of root, ammunition consumption plays a part not in proper sequence or order want affecting, sometimes can ignore, when node weights defines, arrange accordingly.
Semantic matching method is used exemplarily for about Fig. 6 two figure:
First more identical root node (being P1 in the drawings), P1 node has a minor matters point in left figure, and { P10} has a minor matters point { P10} in right figure.Zhi Jiedian number=1 of the Zhi Jiedian number of then similarity=identical/total.Recurrence finds minor matters point.First find P10 node in left figure, P10 also has in right figure.Similarity Measure=1/3 of P10 node (the subclass number/subclass number of identical parent number/parent number+identical+identical adjacent class number/adjacent class number)=8/9.By that analogy, total similarity=each node weights × each node similarity is calculated.
Step S6, judges differentiation structure.In left figure, P4 is the parameter that right figure does not have, and in right figure, P9 is the parameter that left figure does not have, and in S5-4, can point out and gather this parameter.If gather, then adopt this formula as target prediction model.P9 is as differentiation structure output, and P9 in formula " troops than " parameter there is no embodiment, can correction as a result.
In step S7, utilize history consumption information further, add up.Rule mainly provides the correction of modifying factor to formula and result, is made up of a lot of IF-then or similar type rule, and modification rule, according to data analysing method, as long as its model produced, can perform prediction task.The RIPPER introduced in machine learning textbook can be used, C4.5Rule etc.The explanation that these the form of the rules predict the outcome to this.In an embodiment of the present invention, adopt time series to carry out time series analysis to data, analysis obtains the impact of different phase troops comparison wear rate as correction,
Modification rule: revise according to enemy and we troops ratio in operation.Adopt IF troops ratio=a modified value=b.As table 1.
Table 1: the modifying factor of enemy and we troops comparison ammunition consumption result
In this example, when adopting formulae discovery, method predicated error is 7%.After correction result, method predicated error is about 2%.The visible precision substantially increasing ammunition consumption.
As a kind of more complicated application examples, in the differentiation structure of the corresponding interpretative structural modeling (the left figure of Fig. 6) of step S6 judgment formula with overall interpretative structural modeling (Fig. 5), visible P5 and minor matters point P4, P7, P8, P6, P9 does not embody in the left figure of Fig. 6, and [P5 hits tactics to the relevant factor of this group, P4 ammunition performance, P6 battlefield surroundings, P7 attack precision, P8 weapon firing rate, P9 relative military strength] as differentiation structure output, require that user provides corresponding data simultaneously.
Because factor is many, between above-mentioned parameters and the demand of ammunition, there are the mapping relations of the nonlinear inherence of certain complexity, these mapping relations are described by ammunition consumption interpretative structural modeling, but the quantification of structure is more difficult to be completed, consider to utilize neural network learning, carry out the correction of ammunition consumption, obtain consumption result.Neural network used herein can be the neural network of any type, as long as can perform prediction task, such as, can use the multi-layer feed-forward BP network introduced in neural network textbook.
As a kind of preference, especially.
Step S7, adopts a kind of modification method based on neural metwork training, and correspondingly affects ammunition consumption structure, and the concrete step that performs comprises:
Step S7-1: as forecast model does not train, then perform S7-2, otherwise redirect S7-6;
Step S7-2: the quantification of ammunition consumption factor of influence and normalized;
Step S7-3: build neural network model, set up the Three-tider architecture of input layer, hidden layer, output layer;
Step S7-4: utilize initial training to assemble for training and practise a neural network;
Step S7-5: utilize neural network to produce modification rule from data;
Step S7-6: terminate.
In an embodiment of the present invention, to the demand forecast of artillery ammunition, P1 in utilization variance structure is as target variable, P4, P6, P7, P8, P9 are as input variable, hit the correction factor of tactics P5, inputted by P4, P6, P7, P8, P9 of neural network, and then obtain output function f (P4, P6, P7, P8, P9) as modification rule.
S7-2 step is to the quantification of factor of influence and normalized, for the ease of modeling, be divided into 5 grades, represent the influence degree of this factor to ammunition consumption, be decided to be very large impact, considerable influence, General Influence respectively, substantially do not affect, do not affect, corresponding value gets 0.95,0.85,0.75,0.65,0.50.S7-3 step, using 4 evaluation indexes as input layer, namely input layer number is 4, rule of thumb getting node in hidden layer is 2, output layer nodes is 1, represents ammunition Combat Consumption modified value, sets up the Three-tider architecture of input layer, hidden layer, output layer, S7-4 step for learning sample data, gets 10 samples with the consumption data amount in the typical specific example of a battle.As table 2:
Table 2 neural network model learning sample data
In reality is fought, hit the Different Effects consumption of tactics, but consider when calculation task amount less.Like this by task amount computing formula N=∑ N
0p
ik
ibe modified to N=∑ N
0p
ik
i* f (P4, P6, P7, P8, P9).By data detection, it is substantially identical that the result drawn added up by the result that model draws and historical summary, and predicated error is within 5%, and precision improves greatly.
In S9 step, analysis result, by expert's further analysis optimization ammunition consumption ISM.
In model bank of the present invention, formula used is well known in the art, but be different from and directly use formula to carry out the conventional method calculated, before use formula, by expert's wisdom, use consumption forecast structure construction module, formula is related to relevant ammunition consumption influence factor and use ISM method to generate ammunition consumption structure.Relation between each ammunition consumption factor with visual structure exhibits, and is reflected the cause-effect relationship between factor by it.Consider between ammunition consumption factor to be not only isolated physical constraint, and be continuous change, interlaced restriction relation.The expertise of formula reflection behind can be presented in a kind of clear mode by building ammunition consumption structure, can the accurate context that uses of more accurate expression formula and correction formula thereof.In addition to the correction of formula or result, the precision of prediction adjusting a part can be easy to, obviously to identify that the difference degree of this predicted value and actual value is to reduce predicated error.Needing for only having a parameter after Selection parameter to change, simply can realize adjustment, and predicting the outcome and can be easy to distinguish.
The present invention is not limited only to above embodiment, everyly utilizes mentality of designing of the present invention, does the design of some simple change, all should count within protection scope of the present invention.
Claims (6)
1. based on an ammunition consumption prognoses system for ISM knowledge refining, it is characterized in that, described system comprises:
Expertise load module, for providing the relation between the influence factor of man-machine interface for expert typing ammunition consumption and influence factor.
Predict builds module, explains structure tree for the overall situation being formed ammunition consumption by interpretative structure-modeling (ISM, Interpretative Structural Modeling).
Model bank, for storing default consumption forecast model, model shows as the formula Sum fanction that a group determines input, output parameter, for the ammunition consumption prediction under different context.Along with the domination of Elements Of Expense causality knowledge, model bank can adjust, and changes rule and formula in model bank.
Data acquisition module, for obtaining supplemental characteristic needed for prediction and the contextual supplemental characteristic of phone predicts.
Forecast model matching module, for prediction context parameters being organized into interpretative structural modeling tree, the model in model bank is related to parameter and be also organized into interpretative structural modeling tree, both mate, and according to the situation of coupling, choose formula model the most close for prediction.
Differentiation structure output module, the overall situation for not yet relating in formula model explains that a part for structure tree is as differentiation structure output.
Modifying model module, for data analysis, obtains the modification rule of forecast model, and to the modification rule that entirety predicts the outcome.
Consumption computing module, calculates consumption for using formula model the most close and modification rule.
2., as claimed in claim 1 based on the ammunition consumption prognoses system of ISM knowledge refining, it is characterized in that described forecast model matching module is the matching degree utilizing the similarity calculating method of tree construction to judge two models.
3. as claimed in claim 1 based on the ammunition consumption prognoses system of ISM knowledge refining, it is characterized in that described Modifying model module, the method such as time series analysis, neural network is adopted to data, obtains modification rule.
4., based on an ammunition consumption Forecasting Methodology for ISM knowledge refining, it is characterized in that, said method comprising the steps of:
Step S1: build the overall situation of the ammunition consumption based on ISM and explain structure tree;
Step S2: the prediction context of consumption forecast computing formula is used interpretative structural modeling tree representation, and stored in model bank;
Step S3: by the phone predicts context in image data, uses interpretative structural modeling tree representation;
Step S4: will predict that contextual interpretative structural modeling is set, correspondingly with formula in model bank explains that structure tree mates;
Step S5: the model in traversal model bank, obtains the formula model of the highest matching degree.Specify the corresponding formula of the highest matching degree to be target formula, choose formula for prediction;
Step S6: using in formula model still for the overall situation that relates to explains that a part for structure tree is as differentiation structure output;
Step S7: using parameter in differentiation structure as factor, to data analysis, obtain the modification rule of predictor formula, and to the modification rule that entirety predicts the outcome, and for the structure of model in model bank and amendment;
Step S8: by the optimum prediction formula model after improving, prediction ammunition consumption.Consume modified result rule if existed, utilize rule to revise premeasuring;
Step S9: expert analyzes predicting the outcome, forms the expertise that some are new.
Described step S1 builds an ammunition consumption interpretative structural modeling based on ISM and sets, the following sub-step of concrete execution:
Step S1-1: list one group of parameter [P
1, P
2... P
n], P
1essential, be set to ammunition consumption, other parameter predicts relevant variable with ammunition consumption;
Step S1-2: parameter compared one by one, for P
i, P
j, (i, j=1,2 ... n), if both have direct relation, (P is made
i, P
j)=1, if both are without direct relation, then (P
i, P
j)=0.Adjacency matrix A is set up with this;
Step S1-3: use M=(A+I)
t, obtain the reachability matrix M of factor of system.Wherein I is and the unit matrix of A with order;
Step S1-4: divide element level, reachable set, in advance collection and jointly collect.Reachable set R (P
i) be parameter P
ithe elements combination that can arrive; Leading collection A (P
i) be to arrive key element P
ielements combination; Common collection and R ⌒ A;
Step S1-5: arranging the highest element of ground floor is P
1, the highest element except can arrive oneself itself except, other key element can not be arrived.
Step S1-6: after finding out highest key element, can scratch corresponding row and column by it from reachability matrix, then, then continues to find new highest key element (i.e. second layer key element) from remaining reachability matrix.The like go out each layer key element collection, set up escort tree-shaped hierarchical structure from one place to another.
Described step S5, by match selection model, specifically performs following sub-step:
Step S5-1: all formula in traversal model bank, obtains the interpretative structural modeling tree that each formula correspondence builds.
Step S5-2: computing formula interpretative structural modeling tree and the matching degree predicting context interpretation structure tree;
Step S5-3., according to threshold value constraint, after traversal, finds the formula model existing and meet threshold condition, and obtains the corresponding model of the highest matching degree;
Step S5-4., for the corresponding model of the highest matching degree, points out the data that need gather, if can provide data, using this model as target prediction model, otherwise just gives up this model.
Step S5-5: terminate.
5. a kind of ammunition consumption Forecasting Methodology based on ISM knowledge refining as claimed in claim 4, it is characterized in that, described step S4, the calculating formula of similarity based on tree construction is adopted to calculate matching degree, its rule is higher the closer to the node weights of root, and the node weights more in branch is lower.The similarity of a node is calculated by the number of structural father node, child node, the brotgher of node.Total similarity is calculated by each node similarity.Calculate and perform following steps:
Step S4-1: first find root node.If there is minor matters point, then the Zhi Jiedian number of the Zhi Jiedian number/total of the similarity=identical of root node;
Step S4-2: recurrence finds minor matters point.Similarity Measure=1/3 of minor matters point (the subclass number/subclass number of identical parent number/parent number+identical+identical adjacent class number/adjacent class number);
Step S4-3: total similarity=each node weights × each node similarity.
6. a kind of ammunition consumption Forecasting Methodology based on ISM knowledge refining as claimed in claim 4, it is characterized in that, described step S7, adopt a kind of modification method based on neural metwork training, the input neuron of this neural network takes from differentiation structure, and the concrete step that performs comprises:
Step S7-1: as forecast model does not train, then perform S7-2, otherwise redirect S7-6;
Step S7-2: the quantification of ammunition consumption factor of influence and normalized;
Step S7-3: build neural network model, set up the Three-tider architecture of input layer, hidden layer, output layer;
Step S7-4: utilize initial training to assemble for training and practise a neural network;
Step S7-5: utilize neural network to produce modification rule from data;
Step S7-6: terminate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410424091.1A CN104504440A (en) | 2014-08-25 | 2014-08-25 | Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410424091.1A CN104504440A (en) | 2014-08-25 | 2014-08-25 | Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104504440A true CN104504440A (en) | 2015-04-08 |
Family
ID=52945835
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410424091.1A Pending CN104504440A (en) | 2014-08-25 | 2014-08-25 | Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104504440A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105574276A (en) * | 2015-12-18 | 2016-05-11 | 华子荀 | Design method of interactive digital education television programs |
| CN107918358A (en) * | 2017-11-17 | 2018-04-17 | 山东师范大学 | Numerical control equipment failure analysis methods and device |
| CN108038544A (en) * | 2017-12-04 | 2018-05-15 | 华南师范大学 | Neutral net deep learning method and system based on big data and deep learning |
| CN110751283A (en) * | 2018-07-05 | 2020-02-04 | 第四范式（北京）技术有限公司 | Model interpretation method, device, equipment and storage medium |
| CN111145548A (en) * | 2019-12-27 | 2020-05-12 | 银江股份有限公司 | Important intersection identification and subregion division method based on data field and node compression |
| CN111898261A (en) * | 2020-07-22 | 2020-11-06 | 北京理工大学 | Ammunition reaction intensity quantitative evaluation method based on combustion network reaction evolution model |
| CN112818522A (en) * | 2021-01-18 | 2021-05-18 | 中国人民解放军91776部队 | Ammunition consumption measuring and calculating method and device |
| CN112883631A (en) * | 2021-01-13 | 2021-06-01 | 新智数字科技有限公司 | Method and device for predicting oxygen content of exhaust gas of gas turbine |
| CN113627489A (en) * | 2021-07-14 | 2021-11-09 | 青岛海尔能源动力有限公司 | Demand-based power consumption prediction method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101688765A (en) * | 2007-05-30 | 2010-03-31 | 劳尔·德尔加多·阿卡热塔 | The device that is used for controlling ammunition consumption in real time |
| JP2010281564A (en) * | 2003-02-07 | 2010-12-16 | Komatsu Ltd | Device and method of guiding ammunition |
-
2014
- 2014-08-25 CN CN201410424091.1A patent/CN104504440A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010281564A (en) * | 2003-02-07 | 2010-12-16 | Komatsu Ltd | Device and method of guiding ammunition |
| CN101688765A (en) * | 2007-05-30 | 2010-03-31 | 劳尔·德尔加多·阿卡热塔 | The device that is used for controlling ammunition consumption in real time |
Non-Patent Citations (3)
| Title |
|---|
| 郭泳亨等: "基于ISM的面向目标的弹药需求影响因素分析", 《四川兵工学报》 * |
| 陈俊等: "基于系统理论方法的炮兵弹药消耗量预测修正方法", 《指挥控制与仿真》 * |
| 黄天辰等: "基于神经网络的弹药消耗预测", 《军械工程学院学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105574276A (en) * | 2015-12-18 | 2016-05-11 | 华子荀 | Design method of interactive digital education television programs |
| CN107918358A (en) * | 2017-11-17 | 2018-04-17 | 山东师范大学 | Numerical control equipment failure analysis methods and device |
| CN107918358B (en) * | 2017-11-17 | 2021-02-23 | 山东师范大学 | Fault analysis method and device for numerical control equipment |
| CN108038544A (en) * | 2017-12-04 | 2018-05-15 | 华南师范大学 | Neutral net deep learning method and system based on big data and deep learning |
| CN110751283A (en) * | 2018-07-05 | 2020-02-04 | 第四范式（北京）技术有限公司 | Model interpretation method, device, equipment and storage medium |
| CN111145548A (en) * | 2019-12-27 | 2020-05-12 | 银江股份有限公司 | Important intersection identification and subregion division method based on data field and node compression |
| CN111145548B (en) * | 2019-12-27 | 2021-06-01 | 银江股份有限公司 | Important intersection identification and subregion division method based on data field and node compression |
| CN111898261A (en) * | 2020-07-22 | 2020-11-06 | 北京理工大学 | Ammunition reaction intensity quantitative evaluation method based on combustion network reaction evolution model |
| CN111898261B (en) * | 2020-07-22 | 2022-09-09 | 北京理工大学 | Ammunition reaction intensity quantitative evaluation method based on combustion network reaction evolution model |
| CN112883631A (en) * | 2021-01-13 | 2021-06-01 | 新智数字科技有限公司 | Method and device for predicting oxygen content of exhaust gas of gas turbine |
| CN112818522A (en) * | 2021-01-18 | 2021-05-18 | 中国人民解放军91776部队 | Ammunition consumption measuring and calculating method and device |
| CN113627489A (en) * | 2021-07-14 | 2021-11-09 | 青岛海尔能源动力有限公司 | Demand-based power consumption prediction method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104504440A (en) | Method and system for predicting ammunition consumption based on interpretative structural modeling knowledge refinement | |
| CN112329348B (en) | Intelligent decision-making method for military countermeasure game under incomplete information condition | |
| Grishko et al. | Management of structural components complex electronic systems on the basis of adaptive model | |
| CN109933669B (en) | Matching method of battlefield situation data labels | |
| CN110059932A (en) | Air-to-ground attack type unmanned plane Combat Efficiency Evaluation method based on SVM | |
| CN109670660A (en) | A kind of fleet dynamic air defense threat estimating method based on intuitionistic fuzzy TOPSIS | |
| CN114444201A (en) | Autonomous capability evaluation method of ground attack unmanned aerial vehicle based on Bayesian network | |
| Singh et al. | Election results prediction system based on fuzzy logic | |
| Reddy et al. | Neural networks for prediction of loan default using attribute relevance analysis | |
| CN116341924A (en) | Multitasking intelligent decision computing method and system based on meta learning and game theory | |
| Wang et al. | Application of BP neural network to prediction of library circulation | |
| Lytvyn et al. | Ontology Using for Decision Making in a Competitive Environment | |
| Lin et al. | Intelligent assessment method of air-defense & anti-missile command model based on genetic algorithm optimization | |
| CN110941904A (en) | Sensing equipment combination and collocation method based on different combat weather | |
| Rui et al. | Peacekeeping equipment support spare parts demand forecast | |
| CN117217392B (en) | Method and device for determining general equipment guarantee requirement | |
| Wiratama et al. | System Dynamics Modeling for Smart Defense System of the Archipelago Sea Defense in Indonesia | |
| Das et al. | Agent-based Decision Making for Integrated Air Defence Systems | |
| Jia et al. | Air Target Threat Assessment Based on PCNN Neural Network | |
| CN116151492B (en) | Auxiliary decision analysis system of combined combat system | |
| Zhang et al. | Intelligent Planning of Battlefield Resources Based on Rules and Capability | |
| Minghua et al. | Rough Set Neural Network Evaluation Method for UAV Reconnaissance Effectiveness | |
| Zhang et al. | Soldier Threat Assessment Method Based on Bayesian Network | |
| Chen et al. | A Prioritization Method for Battlefield Information Based on C-KVC | |
| Wiratama et al. | SMART DEFENSE MODELING OF THE ARCHIPELAGO SEA DEFENSE STRATEGY TO SUPPORT INDONESIAN NAVY DUTIES |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150408 |