Torpedo loading reliability evaluation method based on multi-source information grey fusion model

Abstract

The present invention relates to a torpedo loading reliability assessment method based on a multi-source information grey fusion model, and belongs to the technical field of torpedo reliability application research. The method takes a non-equidistant grey model as the core, firstly uses a grey correlation method to calculate the comprehensive development coefficient of multi-source reference information fusion, and then uses a grey estimation weighting method to fuse the grey fitting results of the loading test data and the fitting results of the reference information again. The calculation results show that the simulation prediction accuracy of the fused model is greatly improved compared with the non-equidistant grey model before fusion.

Description

Technical Field

[0001] The invention belongs to the technical field of torpedo reliability application research, and in particular relates to a torpedo loading reliability evaluation method based on a multi-source information grey fusion model. Background Art

[0002] Reliability assessment refers to the process of estimating the performance indicators of product reliability based on the reliability structure, life model and test information of the product using statistical methods and means. At present, there are great difficulties in the reliability assessment of large and complex systems. Torpedoes, as a large underwater weapon that needs to be stored for a long time and used once, often have to go through a long loading process before performing combat readiness tasks. In addition, torpedoes are not repairable when loaded on ships, and the loading environment is worse than that of cave storage. Therefore, it can be said that the reliability of torpedo loading is an important indicator for measuring the combat quality level of torpedoes, and is an important basis for formulating combat readiness plans and maintenance support strategies for torpedoes. Torpedo loading reliability refers to the ability of a torpedo to maintain the specified function under the specified ship loading conditions and within the specified loading time, which is usually measured by probability.

[0003] Due to the high cost and long time of torpedo loading test, the loading test data are usually incomplete and uncertain small sample data. In order to deal with the problem of small number of test samples, some scholars use other reliability information to evaluate the reliability of small sample systems based on successful examples of weapon system identification tests. At present, the information fusion method is widely used in the reliability evaluation of small sample weapon systems. At present, the Bayes method is usually used in the field of information fusion for torpedo loading reliability evaluation, but this method has high requirements for prior information. For torpedo systems, the acquisition of test information that has reference value for loading reliability is often incomplete or uncertain.

[0004] Grey system theory shows unique advantages in this kind of "poor information and uncertainty" problem - it does not need to know the distribution information of the original data and is not subject to the statistical constraints of the original data. Grey system theory attempts to correctly grasp the development law of the system through the generation and development of some known information, and then quantitatively predicts the future state of the system. Therefore, the grey model is very suitable for the characteristics of the small amount of torpedo loading reliability test data, and can be used to expand the source of torpedo loading reliability test data, so as to develop an evaluation method suitable for torpedo loading. Summary of the invention

[0005] Technical issues to be solved

[0006] Torpedo loading reliability is an important part of torpedo reliability assessment, but the torpedo loading test is costly, difficult, and has a long test cycle, making it difficult to conduct reliability assessment. In order to solve the problem of insufficient test data in the existing torpedo loading reliability assessment, the present invention provides a torpedo loading reliability assessment method based on a multi-source information grey fusion model.

[0007] Technical Solution

[0008] A method for evaluating the reliability of torpedo loading based on a multi-source information grey fusion model is characterized by the following steps:

[0009] Step 1: Filter out the available reference test data series through grey correlation;

[0010] Step 2: Perform dimensionless preprocessing on all test data;

[0011] Step 3: Establish the non-equidistant grey GM (1,1) model of the target test data and each reference test data;

[0012] Step 4: Establish a fusion model of the reference test data sequence;

[0013] Step 5: Fuse the model after fusion of the reference test data sequence with the fitting model of the target test data;

[0014] Step 6: Evaluate and test the final fused model.

[0015] A further technical solution of the present invention is as follows: In step 1, available reference test data sequences are screened out by using grey correlation degree, and the grey correlation degree of the reference test data sequence needs to satisfy:

[0016]

[0017] where ξ is called the resolution coefficient, γ(X0,X i ) are called X0 and X i Grey relational degree; for target sequence X0 and reference sequence X i , the lengths of the two sequences may not be equal. In this case, we can delete the excess data of the longer sequence or use the GM(1,1) model of the grey system for prediction to make up for the insufficient data of the shorter sequence. δ is the correlation threshold, which is set according to the actual situation. When calculating the correlation between other candidate sequences and the target sequence, if the correlation is lower than the threshold, the sequence is discarded, and all candidate sequences are calculated in turn until all available reference sequences are screened out.

[0018] A further technical solution of the present invention is as follows: in step 3, a non-equidistant grey GM (1,1) model of the target test data and each reference test data is established. The specific implementation process is to first perform non-equidistant processing on each test data sequence, and then establish respective grey G (1,1) models for the non-equidistant data.

[0019] A further technical solution of the present invention: in step 4, establishing a fusion model of the reference test data sequence comprises the following steps:

[0020] Step 4.1: Extract the development coefficient of the grey model established for each reference sequence - a ri ;

[0021] Step 4.2: Use the grey correlation degree to assign weights to the development coefficients of each reference sequence and calculate the integrated development coefficient - a after fusion r , the calculation method is as follows:

[0022]

[0023] where ω i (i=1,2,…,n) represents the weight of the reference sequence on the predicted trend of the modeling sequence, and γ(X0,X i ) is the target sequence X0 and the reference sequence X calculated in step 1 i Grey relational degree of

[0024] Step 4.3: Combine the comprehensive development coefficient-a r The grey model expression after the reference sequence fusion is obtained by the sum mapping method. The specific calculation method is:

[0025]

[0026] The condition for the formula to be valid is a m =a r ,

[0027] A further technical solution of the present invention is as follows: in step 5, the model after the fusion of the reference test data sequence is fused with the target test data fitting model, comprising the following steps:

[0028] Step 5.1: Use the non-equidistant grey model established by the target sequence X0 to find the fitting data corresponding to itself

[0029] Step 5.2: Use the reference sequence fusion model obtained in 4.3 to obtain the fitting data corresponding to the target sequence

[0030] Step 5.3: Calculate the sequence X0, and Grey point estimation value, the specific calculation process of grey point estimation is as follows;

[0031]

[0032] Among them, dr(x j ,x i ) is the data sample x i With the evaluated parameter x j The gray distance measure of It can be regarded as a point estimate of the grey parameter estimation, ω i Represents the sample point x i In gray, the estimated value The weight of

[0033] Step 5.4: Establish the gray estimation weighting method and Assign weights to determine the final fusion model. The calculation method is as follows:

[0034]

[0035] in, is the gray estimation of the target sequence, is the gray estimate of the fitting sequence of the target sequence, is the grey estimate of the fused fitting sequence, α is the overlap coefficient and α∈[0,1]; when α=1, the model degenerates into the traditional GM(1,1) model, and when α=0, it becomes the prediction of the change value of its mapping sequence.

[0036] A further technical solution of the present invention is as follows: in step 6, the final fused model is evaluated and tested, and the calculation method is:

[0037]

[0038] For the relative error test, is the average relative error; When , the model is of first-level accuracy. When , the model belongs to the second level of accuracy; when When , the model belongs to the third level of accuracy; When , the model belongs to the fourth level of accuracy; when The model accuracy is very poor and is not recommended in actual engineering.

[0039] Beneficial Effects

[0040] The invention proposes a method for evaluating the reliability of torpedo loading based on a multi-source information grey fusion model. Aiming at the characteristics of the small number of torpedo loading reliability test data, the invention expands the source of torpedo loading reliability test data and constructs a multi-source information grey fusion model. Since the test time corresponding to various reference information may be different and non-equidistant, the invention takes the non-equidistant grey model as the core, constructs the model through two fusions, and the model can correct the results according to other relevant information, so that the simulation and prediction accuracy is greatly improved.

[0041] Compared with the prior art, the present invention has the following beneficial effects: 1) It does not place too many requirements on the distribution information of reference information, thereby enriching the source of reference information; 2) Experimental results show that the model can better adapt to the medium- and long-term prediction of data, and its accuracy is greatly improved compared with the model without information fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings are only for the purpose of illustrating specific embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0043] Figure 1 It is a flowchart for implementing the technology of the present invention;

[0044] Figure 2 Screening models for reference sequences;

[0045] Figure 3 is the reference sequence fusion model. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] The invention proposes a method for evaluating the reliability of torpedo loading based on a multi-source information grey fusion model, comprising the following steps:

[0048] Step 1: Use grey correlation to select available reference test data series. Figure 2 For example, the grey relational degree should satisfy:

[0049]

[0050] Where ξ is called the resolution coefficient, usually ξ=0.5,γ(X0,X i ) are called X0 and Xi The grey relational degree of the target sequence X0 and the reference sequence X i , the lengths of the two sequences may not be equal. In this case, measures such as deleting the excess data of the longer sequence or using the GM (1,1) model of the grey system for prediction to fill in the insufficient data of the shorter sequence can be taken. δ is the correlation threshold, which can be set according to the actual situation. When calculating the correlation between other candidate sequences and the target sequence, if the correlation is lower than the threshold, the sequence is discarded, and all candidate sequences are calculated in turn until all available reference sequences are screened out.

[0051] Step 2: Perform dimensionless preprocessing on the test data to eliminate the influence of different dimensions between the data. The specific calculation method is as follows:

[0052] x i (k) = x i (k) / x i (1),(k=1,2,…n) (3)

[0053] Step 3: Establish the non-equidistant grey GM (1,1) model of the target test data and each reference test data. The specific calculation modeling process is as follows;

[0054] Suppose there is a non-uniformly spaced sequence Y (0) (t i )=(y (0) (t1),y (0) (t2),…,y (0) (t n )), then the corresponding time series is T (0) (i)=(t1,t2,…,t n ). The average time interval is:

[0055]

[0056] Each actual observation period t i The unit time difference coefficient with the average time interval Δt0 is:

[0057]

[0058] The difference between the actual observation periods is:

[0059] Δy (0) (t i )=μ(t i )[y (0) (t i )-y (0) (t i-1 )] (6)

[0060] The calculation formula for the equally spaced sequence is:

[0061] x (0) (i) = y (0) (t i )-Δy (0) (t i ) (7)

[0062] The non-uniformly spaced sequence can be transformed into an equally spaced sequence X by using formulas (4)-(7) (0) =(x (0) (1),x (0) (2),…,x (0) (n)). The first-order accumulation of non-negative equally spaced sequences generates the sequence X (1) =(x (1) (1),x (1) (2),…,x (1) (n)), where

[0063]

[0064] For the sequence X (1) Its adjacent generated sequence is Z (1) =(z (1) (2),z (1) (3),…,z (1) (n)), which is calculated as follows:

[0065] z (1) (k) = 0.5 (x (1) (k)+x (1) (k-1)),(k=2,3,…,n) (9)

[0066] The equation x (0) (k)+az (1) (k) = b is the grey GM (1,1) model, and its parameter a = [a, b] T The least squares method can be used for estimation, and we can get:

[0067]

[0068]

[0069] Restore the time response series to a function related to time t in a non-equally spaced series:

[0070]

[0071] Step 4: Establish a fusion model of the reference test data sequence to Figure 3 For example, the specific steps include:

[0072] Step 4.1: Extract the development coefficient of the grey model established for each reference sequence - a ri ;

[0073] Step 4.2: Use the grey correlation degree to assign weights to the development coefficients of each reference sequence and calculate the integrated development coefficient - a after fusion r , the calculation method is as follows:

[0074] For reference sequences Establish the GM (1,1) model, integrate the information of multiple development coefficients, and obtain a comprehensive development coefficient -a r , then:

[0075]

[0076] where ω i (i=1,2,…,n) represents the weight of the reference sequence on the predicted trend of the modeling sequence, and γ(X0,X i ) is the target sequence X0 and the reference sequence X calculated in step 1 i The grey correlation degree of .

[0077] Step 4.3: Combine the comprehensive development coefficient-a r The grey model expression after the reference sequence fusion is obtained by the sum mapping method. The specific modeling process is as follows:

[0078] Determine the comprehensive development coefficient -a through step 4.2 r , and its corresponding set of fusion sequences is But because Since the sequence is unknown, it cannot be used directly to model. Create a sequence X (0) The mapping sequence Its mapping development coefficient can be considered as a m =a r , so the time response sequence of the mapping sequence is generally solved as follows:

[0079]

[0080] Because the sequence Is To X (0) The mapping of the mapping sequence and the modeling sequence should have roughly the same changing trend, so it can be assumed that Substituting the assumptions, we get

[0081]

[0082] The sequence reduction value is:

[0083]

[0084] In step 5, the model after the reference test data sequence is fused is fused with the target test data, which includes the following steps:

[0085] Step 5.1: Use the non-equidistant grey model established by the target sequence X0 to find the fitting data corresponding to itself

[0086] Step 5.2: Use the reference sequence fusion model obtained in 4.3 to obtain the fitting data corresponding to the target sequence

[0087] Step 5.3: Calculate the sequence X0, and Grey point estimation value, the specific calculation process of grey point estimation is as follows;

[0088] By knowing the data sample X=(x1,x2,…,x n ) for sample parameters x j For evaluation, data sample x i With the evaluated parameter x j The calculation of the gray distance measure is as follows:

[0089]

[0090] For a given test data space X i =(x i (1),x i (2),…,x i (n i ), due to the limitation of sample quantity, it can be considered that each sample point can reflect the attribute of the true value. The grey estimation value can be calculated according to the following formula:

[0091]

[0092] Among them, ω i Represents the sample point x i In gray, the estimated value The weight of

[0093] At this time, the gray estimate generated by gray It can be regarded as a point estimate of the grey parameter estimation. i The calculation of can be determined based on the gray distance measure between each sample and the entire sample space. The calculation steps are as follows.

[0094] (1) Calculate the gray distance measure between each sample point and other sample points, and then calculate the gray distance measure between it and the entire sample space by summing and averaging:

[0095]

[0096] (2) The weight can be calculated by normalizing the gray distance measure:

[0097]

[0098] Step 5.4: Establish the gray estimation weighting method and Assign weights to determine the final fusion model. The calculation method is as follows:

[0099]

[0100] in, is the gray estimation of the target sequence, is the gray estimate of the fitting sequence of the target sequence, is the grey estimation of the fused fitting sequence. α is the overlap coefficient, and α∈[0,1]. When α=1, the model degenerates into the traditional GM(1,1) model, and when α=0, it becomes the prediction of the change value of its mapping sequence. Formula (24) is the final expression of the multi-source information grey fusion model.

[0101] Step 6: Evaluate and test the final fused model. The calculation formula is as follows;

[0102]

[0103] For the relative error test, is the average relative error. When , the model is of first-level accuracy. When , the model belongs to the second level of accuracy; when When , the model belongs to the third level of accuracy; When , the model belongs to the fourth level of accuracy; when The model accuracy is very poor and is not recommended in actual engineering.

[0104] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the reliability of torpedo loading based on a multi-source information grey fusion model, characterized in that Here are the steps: Step 1: Filter out the available reference test data series through grey correlation; Step 2: Perform dimensionless preprocessing on all test data; Step 3: Establish the non-equidistant grey GM (1,1) model of the target test data and each reference test data; Step 4: Establish a fusion model of the reference test data sequence; include: Step 4.1: Extract the development coefficient of the grey model established for each reference sequence - a ri ; Step 4.2: Use the grey correlation degree to assign weights to the development coefficients of each reference sequence and calculate the integrated development coefficient - a after fusion r , the calculation method is as follows: where ω i (i=1,2,n, represents the weight of the reference sequence on the prediction trend of the modeling sequence, and γ(X0,X i ) is the target sequence X0 and the reference sequence X calculated in step 1 i Grey relational degree of Step 4.3: Combine the comprehensive development coefficient-a r The grey model expression after the reference sequence fusion is obtained by the sum mapping method. The specific calculation method is: The condition for the formula to be valid is a m =a r , Step 5: Fusing the model after fusion of the reference test data sequence with the target test data fitting model; including: Step 5.1: Use the non-equidistant grey model established by the target sequence X0 to find the fitting data corresponding to itself Step 5.2: Use the reference sequence fusion model obtained in 4.3 to obtain the fitting data corresponding to the target sequence Step 5.3: Calculate the sequence X0, and Grey point estimation value, the specific calculation process of grey point estimation is as follows; Among them, dr(x j ,x i ) is the data sample x i With the evaluated parameter x j The gray distance measure of It can be regarded as a point estimate of the grey parameter estimation, ω i Represents the sample point x i In gray, the estimated value The weight of Step 5.4: Establish the gray estimation weighting method and Assign weights to determine the final fusion model. The calculation method is as follows: in, is the gray estimation of the target sequence, is the gray estimate of the fitting sequence of the target sequence, is the gray estimation of the fused fitting sequence, α is the overlap coefficient and α∈[0,1]; when α=1, the model degenerates into the traditional GM(1,1) model, and when α=0, it becomes the prediction of the change value of its mapping sequence; Step 6: Evaluate and test the final fused model.

2. The torpedo loading reliability assessment method based on multi-source information grey fusion model according to claim 1 is characterized by: In step 1, available reference test data sequences are screened out by using the grey correlation degree. The grey correlation degree of the reference test data sequence needs to satisfy: where ξ is called the resolution coefficient, γ(X0,X i ) are called X0 and X i Grey relational degree; for target sequence X0 and reference sequence X i , the lengths of the two sequences may not be equal. In this case, we can delete the excess data of the longer sequence or use the GM (1,1) model of the grey system for prediction to make up for the insufficient data of the shorter sequence. δ is the correlation threshold, which is set according to the actual situation. When calculating the correlation between other candidate sequences and the target sequence, if the correlation is lower than the threshold, the sequence will be discarded, and all candidate sequences will be calculated in turn until all available reference sequences are screened out.

3. The torpedo loading reliability assessment method based on multi-source information grey fusion model according to claim 1 is characterized in that: In step 3, a non-equidistant grey GM (1,1) model is established for the target test data and each reference test data. The specific implementation process is to first perform non-equidistant spacing on each test data sequence, and then establish respective grey G (1,1) models for the non-equidistant data.

4. The torpedo loading reliability assessment method based on multi-source information grey fusion model according to claim 1 is characterized in that: In step 6, the final fused model is evaluated and tested, and the calculation method is: For the relative error test, is the average relative error; When Δ>0.20, the model has very poor accuracy and is not recommended for use in actual projects.

Images