A virtual test method based on multi-source data fusion

Abstract

The application discloses a kind of virtual test methods based on multi-source data fusion, adopt the mode of real test data and simulation test to realize virtual test, through test data, batch simulation, machine model training, reverse prediction and virtual test etc., a kind of comprehensive, reproducible, repeated analysis, physical field reproduction, multi-source data fusion virtual test method is constructed.The present application is based on limited test data, combined with batch numerical simulation calculation, machine learning algorithm, constructs virtual test.The present application aims to solve the problem of difficult acquisition and less acquisition of offshore test data, can realize the offshore test data " from less to more", point data becomes field data, maximizes the utilization value of real test data (only a small amount).

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of ship sea test and equipment, and particularly relates to a virtual test method based on multi-source data fusion. BACKGROUND

[0002] The ship may be subjected to collision, heavy wave load or fuel gas explosion impact during driving, which seriously threatens the stability and safety of core device equipment such as propulsion system, communication equipment and data device on the ship. Through the virtual test method, important and valuable data basis can be provided for evaluating the protection capability of the ship against complex sea conditions.

[0003] At present, from the perspective of measurement data and scientific research, the amount of data that can be obtained by sea test is still limited, and mainly related video, impact environment, strain, temperature, water level and other information of limited measurement points can be obtained, and it can be predicted that in the future, there will be no great breakthrough in measurement instruments and measurement technology. Therefore, the single sea test has the disadvantages of high test cost, great test difficulty, difficult data acquisition, small amount of data acquisition, unclear physical process of sea test, etc. In addition, since the relative position of the explosion point and the target cannot be accurately obtained by sea test, the simulation test lacks accurate working condition, so it is difficult to carry out a simulation test with strong persuasion.

[0004] In order to solve the above existing problems, the present application considers the above-mentioned situation by means of the currently available technical means, adopts the mode of combining real test data and simulation test to realize virtual test, and constructs a comprehensive, reproducible, repeated analysis, physical field reproduction, multi-source data fusion virtual test method. SUMMARY

[0005] The purpose of the present application is to provide a new method of constructing virtual test based on limited test data, combining batch numerical simulation calculation and machine learning algorithm. The method can realize the change of sea test data from less to more, and maximize the research value of a small amount of real test data.

[0006] The technical solution adopted by the present application is a virtual test method based on multi-source data fusion.

[0007] A virtual test method based on multi-source data fusion, the steps of which are:

[0008] S01: Obtain effective data of measurement points by test or existing test, including acceleration measurement data, pressure measurement data, strain measurement data, break data, etc., and perform data processing to establish simplified characteristic values;

[0009] S02: According to the target ship information, a finite element model is established, and a random working condition is established. Batch calculation of random working condition, extract the effective data corresponding to the test in the result, that is, acceleration measurement data, pressure measurement data, strain measurement data, break data, etc., and take the same data processing method as the test data, obtain the simplified characteristic value, and thus establish the measurement information corresponding to various different working conditions. Normalize the original data to form a database used for machine learning.

[0010] S03: Using machine learning algorithm (support vector machine), taking result data as characteristic (i.e. independent variable X), taking working condition information as target (i.e. dependent variable Y), training the basic database generated in step S02, wherein the break data needs to be classified by using the classification algorithm of machine learning for classification model training, and other data uses the regression algorithm for regression model training, so as to establish the corresponding machine learning prediction model for all working condition information conditions (such as spatial coordinates, charge, water depth), and perform model accuracy evaluation and parameter optimization.

[0011] S04: Using the prediction model of corresponding working condition information established in step S03, taking the data obtained in step S01 as a test set, and pretesting the corresponding working condition information of the test, that is, charge, explosive burst point coordinates, etc.

[0012] S05: Comparing the working condition obtained in step S04 with all known working conditions of the test, that is, comparing the charge, burst point coordinates, etc. Among them, the test burst point spatial coordinates are analyzed through video, picture and other materials to obtain a distribution range. The predicted value of the coordinate is considered reasonable within the distribution range, and other predicted values (such as charge) are considered reasonable within the error range of 20%. For the predicted values with obvious differences, return and execute steps S03, S04, S05 until the predicted values are within the error range.

[0013] S06: Taking the working condition information predicted in step S04 as input condition, inputting into the finite element simulation calculation model established in step S02, simulating the test to obtain the simulation result of the test. Using the corresponding post-processing software, processing the simulation result to form the corresponding video animation data, physical field data and other data, so as to visualize the data, and packaging the result file to form the final virtual experiment platform.

[0014] As preferred, the measurement data processing in steps S01 and S02 establishes simplified characteristic values, and the specific method is:

[0015] S01-01: For the acceleration data of a certain measuring point, the acceleration data is converted into shock spectrum data to obtain spectrum velocity, spectrum displacement and spectrum acceleration. Specifically, the shock spectrum is obtained according to the following steps:

[0016] (1) The acceleration spectrum A(ω) is calculated according to the following formula

[0017]

[0018] In the above formula, is the time-domain acceleration curve of the test point measured by the test; τ is the time variable; t is the time point at which the acceleration curve ends; and ω is the circular frequency.

[0019] (2) The velocity spectrum V(ω) is calculated according to the following formula

[0020]

[0021] (3) The displacement spectrum D(ω) is calculated according to the following formula

[0022]

[0023] (4) The spectrum is plotted according to the following relationship in the logarithmic coordinate system to obtain the impact spectrum.

[0024] logV = logω + logD, logV = -logω + logA (5)

[0025] S01-02: For the strain data of a certain test point, the average processing is performed on the stable strain data thereof.

[0026] S01-03: For the pressure data of a certain test point, if it is an underwater test, the initial shock wave peak value, the initial shock wave pulse width, the bubble pulsation pressure peak value, the bubble pulsation pressure pulse width, and the bubble pulsation period are extracted; if it is an above-water test, the shock wave peak value, the pulse width, and the quasi-static pressure value are extracted.

[0027] S01-04: For the area data of a certain break, the equivalent break radius r is obtained e , and the maximum length l of the actual break is extracted max ; wherein the equivalent break radius r e is calculated according to the following formula:

[0028] πr e 2 = S real (6)

[0029] In the formula, S real is the actual measured or simulated break area.

[0030] As preferred, the finite element simulation calculation in the steps S02 and S06 specifically refers to the explosion fluid-structure coupling simulation calculation under specific working conditions, which is usually simulated by using the ALE or CEL method, and the simulation calculation steps mainly include:

[0031] (1) hull structure modeling, water and air domain modeling, explosive domain modeling;

[0032] (2) hull structure, water medium, air medium, explosive medium material attribute assignment;

[0033] (3) fluid domain boundary condition setting;

[0034] (4) fluid-structure coupling setting;

[0035] (5) solving time and result output setting;

[0036] (6) executing solving calculation.

[0037] As preferred, the original data normalization in the step S02 is specifically calculated according to the following formula:

[0038]

[0039] wherein X nom is the normalized data, X is the original data of the variable to be normalized, X min , X max are the minimum value and the maximum value of the variable data.

[0040] As preferred, the machine learning algorithm in the step S03 is specifically referring to, for the breach data, firstly, the support vector machine algorithm is used for classification training according to whether there is a breach, and then the specific breach data of the breach is used for regression training by the support vector machine algorithm; all the other measurement data are used for regression training by the support vector machine algorithm according to the specific data.

[0041] wherein the support vector machine classification / regression prediction is taking a certain concerned physical quantity as the output variable, i.e. the target set, and taking the explosive XYZ three-dimensional space coordinates, the explosive TNT equivalent, and the explosive underwater explosion water depth as the input variable, i.e. the feature set, through the training of the data set, the regression / regression model of the certain concerned physical quantity and the explosive XYZ three-dimensional space coordinates, the explosive TNT equivalent, and the explosive water depth is obtained, which is shown in the following form:

[0042] x i = f(X, Y, Z, M TNT , d) (8)

[0043] wherein x i is a certain concerned physical quantity; X is the x-direction coordinate of the explosive shot point; Y is the y-direction coordinate of the explosive shot point; Z is the z-direction coordinate of the explosive shot point; M TNT is the explosive TNT equivalent; d is the explosive underwater explosion water depth, which is only used for underwater explosion, and is always taken as 0 for air explosion; f represents the support vector machine classification / regression model.

[0044] As preferred, the step S03 model precision evaluation and parameter optimization, in particular, refers to taking mean square error as loss function, adopting cross validation, grid search, and combining genetic algorithm to optimize the model parameters C and G of the support vector machine, so as to improve the prediction accuracy of the model. Wherein, the mean square error MSE calculation formula is:

[0045]

[0046] Wherein, n is the sample quantity, is the predicted value, y i is the true value.

[0047] Compared with the prior art, the present application has the following advantages:

[0048] 1. The comprehensive physical field of the real test is reproduced by virtual means, so that more measurement data is extracted, and the data value of one test is improved, that is, through one test, comprehensive, reproducible and repeated analysis test data is obtained.

[0049] 2. The present application adopts machine learning algorithm to predict and test measurement data correction and optimization, and has high accuracy.

[0050] 3. The present application can provide intuitive and high credibility process reproduction based on real physical field for test, and also provides rich data materials for weapon equipment development and scientific research. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The system architecture diagram of the present application example;

[0052] Figure 2 The running flowchart of the present application example;

[0053] Figure 3 The principle logic diagram of the present application example;

[0054] Figure 4 The three-fold line impact spectrum diagram of the present application example;

[0055] Figure 5 The underwater explosion pressure curve diagram of the present application example;

[0056] Figure 6 The cabin explosion pressure curve diagram of the present application example;

[0057] Figure 7 The underwater explosion modeling diagram of the present application example. DETAILED DESCRIPTION

[0058] The application will be described in detail below with reference to the drawings and specific embodiments, which are only used to explain the application and do not limit the application.

[0059] As shown in Figure 1 , Figure 2 , the system structure of the virtual test method based on multi-source data fusion mainly includes a test part, a machine learning prediction model part and a virtual test platform part. Referring to Figure 3 there are:

[0060] S01: the test part, mainly for sorting, analyzing and simplifying the test data to form a data set with typical representatives and capable of being used for subsequent machine learning model training.

[0061] Through test or existing test, effective data of the measuring point are obtained, including acceleration measurement data, pressure measurement data, strain measurement data, breakage data, etc., and data processing is performed to establish simplified characteristic values;

[0062] S02: according to the target ship information, a finite element model is established, and random working conditions are established, batch calculation of random working conditions is performed, effective data corresponding to the test in the results, i.e. acceleration measurement data, pressure measurement data, strain measurement data, breakage data, etc. are extracted, the same data processing method as the test data is adopted, and simplified characteristic values are obtained, so as to establish measuring information corresponding to various different working conditions, normalize the original data, and form a database used for machine learning.

[0063] The measurement data processing in steps S01 and S02 establishes simplified characteristic values, and the specific method is:

[0064] S01-01: for the acceleration data of a certain measuring point, the acceleration data is converted into impact spectrum data to obtain spectrum velocity, spectrum displacement and spectrum acceleration. Specifically, the impact spectrum is obtained according to the following steps:

[0065] (1) the acceleration spectrum A(ω) is calculated according to the following formula

[0066]

[0067] In the above formula, is the acceleration time-domain curve of the measuring point measured by the test; τ is the time variable; t is the time point at which the acceleration curve ends; and ω is the circular frequency.

[0068] (2) the velocity spectrum V(ω) is calculated according to the following formula

[0069]

[0070] (3) the displacement spectrum D(ω) is calculated according to the following formula

[0071]

[0072] (4) In the logarithmic coordinate system, the spectrum is plotted according to the following formula relationship, and the impact spectrum is obtained.

[0073] logV = logω + logD, logV = -logω + logA (5)

[0074] Finally, a three-fold line impact spectrum as shown in Figure 4 is obtained.

[0075] S01-02: For the strain data of a certain measuring point, the average processing of the stabilized strain data is carried out.

[0076] S01-03: For the pressure data of a certain measuring point, if it is an underwater test, the underwater explosion pressure time domain curve is obtained, as shown in Figure 5 , the initial shock wave peak value, the initial shock wave pulse width, the bubble pulsation pressure peak value, the bubble pulsation pressure pulse width and the bubble pulsation period are extracted; if it is an overwater test, the in-cabin explosion pressure time domain curve is obtained, as shown in Figure 6 , the shock wave peak value, the pulse width and the quasi-static pressure value are extracted.

[0077] S01-04: For the area data of a certain breach, the equivalent breach radius r e is obtained max , and the maximum length l e of the real breach is extracted; wherein the equivalent breach radius r e is calculated according to the following formula:

[0078] πr 2 real = S

[0079] In the formula, S real is the actual measured or simulated breach area.

[0080] For the step S02 of "establishing a finite element model according to the target ship information", it specifically refers to the explosion fluid-structure coupling simulation modeling of a specific working condition, and a simulation model based on ALE or CEL method is usually established, and the steps mainly include:

[0081] (1) ship structure modeling, water and air domain modeling, explosive domain modeling. As shown in Figure 7 , the ship structure modeling is mainly based on the real ship scale, geometry and structure, material properties, to establish a real scale fine finite element model, the water and air domain is generally set as a hexahedral domain, which is required to be as large as possible based on covering the structure and explosive, and the explosive is set as a sphere with a certain radius according to the equivalent TNT equivalent.

[0082] (2) ship structure, water medium, air medium, explosive medium material property assignment

[0083] Wherein, the ship structure is generally made of steel, for example, Q235 steel, the CS model is used to represent the dynamic response characteristics of the ship steel under the explosion response, the Cowper-Symonds model is a classic model, and the dynamic constitutive equation is

[0084]

[0085] σ0and σ eff are static and dynamic yield stress, respectively, and σ0is 235 MPa; is the strain rate; C, P are Cowper-Symmonds strain rate parameters, which are 40.4 and 5, respectively, representing the material sensitivity characteristics; β is the hardening parameter, β = 0 represents only dynamic hardening, and β = 1 represents isotropic hardening; is the effective plastic strain;

[0086] E p is the plastic hardening modulus, which is given by the following formula:

[0087]

[0088] In the formula, E is the elastic modulus, which is 2.1 GPa; E tan is the tangent modulus, which is 1180 MPa.

[0089] The response of water medium under the action of explosion is described by the Mie-Gruneisen state equation:

[0090]

[0091] In the formula, ρ0is the initial density of water medium; μ is the compression coefficient, μ > 0 represents the compression state, and μ < 0 represents the tension state; c is the sound speed of water; S1, S2 and S3 are dimensionless parameters; γ0is the initial Grüneisen coefficient; a is the first-order volume correction coefficient; E is the internal energy per unit volume. The parameters of water are shown in Table 1.

[0092] Table 1 Parameters of water

[0093]

[0094] Air is described by a linear polynomial state equation:

[0095] p = C0 + C1μ + C2μ 2 + C3μ 3 + (C4 + C5μ + C6μ 2 )E (10)

[0096] In the formula, p is the pressure; Ci (i = 0, …, 3) are constants with pressure dimension; C i (i = 4, …, 6) are dimensionless constants; μ = ρ / ρ0-1; E is the internal energy term. The air-related parameters are shown in Table 2.

[0097] Table 2 Air-related parameters

[0098] ρ kg / m3 [C0] [C1] [C2] [C3] [C4] [C5] [C6] 1.25 0 0 0 0 0.4 0.4 0

[0099] The JWL state equation is used for the explosive,

[0100] The specific form is:

[0101]

[0102] In the formula, V is the relative specific volume of the detonation product; E is the internal energy per unit volume of the explosive, with pressure dimension; A, B, R1, R2, ω are five undetermined parameters of the JWL state equation. The related data of TNT explosive are shown in Table 3.

[0103] Table 3 Explosive materials and state equation parameters

[0104]

[0105] (3) Setting of boundary conditions of the fluid domain;

[0106] For an explosion on water, the air domain is generally set as a non-reflective boundary; for an underwater explosion, the top and bottom surfaces are generally set as non-reflective boundaries, and the four sides are set according to whether the water area is an open water area or a limited water area. If it is an open water area, the static water pressure gradient and the environmental boundary conditions are set, and if it is a limited water area, the static water pressure gradient and the reflective boundary are set.

[0107] (4) Setting of fluid-structure coupling;

[0108] The outer surface of the ship body and the water area, the air domain, and the explosive are set as a fluid-structure coupling group, considering the action force of the ship body structure and the fluid.

[0109] (5) Setting of solving time and result output;

[0110] The calculation termination time is set, and the frequency of result output is set, including animation files, explosion pressure, strain, stress, acceleration, etc.

[0111] The random working conditions in the S02 step are, for example, Figure 3As shown, the key factors such as spatial coordinates of explosives, explosive charge, etc. are randomly sampled based on the probability of attack, a series of random working conditions are established, a series of simulation calculations are carried out based on the set simulation model, that is, Monte Carlo simulation is realized, and a batch of simulation result database is generated, which is shown in Table 4, including explosion point coordinates, equivalent charge, attack speed, impact spectrum (spectrum displacement, spectrum velocity, spectrum acceleration) generated by acceleration, pressure peak, maximum strain, break size, maximum deflection, etc.

[0112] Table 4 Test set and training set data items

[0113]

[0114] In S02, the original data is normalized, specifically, the key data required to be extracted in batch simulation, including explosion point coordinates, equivalent charge, attack speed, impact spectrum (spectrum displacement, spectrum velocity, spectrum acceleration) generated by acceleration, pressure peak, maximum strain, break size, maximum deflection, etc. are normalized according to the following formula to form a standard data set for subsequent training.

[0115]

[0116] Wherein, X nom is the normalized data, X is the original data of the variable to be normalized, X min , X max are the minimum and maximum values of the variable data.

[0117] S03: Adopting machine learning algorithm (support vector machine), taking result data as features (i.e. independent variable X) and working condition information as target (i.e. dependent variable Y), the data set generated in step S02 is trained.

[0118] Firstly, support vector machine algorithm is used for classification training according to whether there is a break, then specific break data is used for regression training by support vector machine algorithm; other measurement data are used for regression training by support vector machine algorithm according to specific data.

[0119] Wherein, support vector machine classification / regression prediction is to take a certain concerned physical quantity as output variable, that is, target set, to take explosive XYZ three-dimensional spatial coordinates, explosive TNT equivalent, explosive underwater explosion water depth as input variable, that is, feature set, through training of data set, regression / regression model of a certain concerned physical quantity and explosive XYZ three-dimensional spatial coordinates, explosive TNT equivalent, explosive water depth is obtained, which is shown as follows:

[0120] x i =f(X,Y,Z,M TNT ,d) (13)

[0121] wherein x i is a certain physical quantity of interest; X is the x-coordinate of the explosive detonation point; Y is the y-coordinate of the explosive detonation point; Z is the z-coordinate of the explosive detonation point; M TNT is the TNT equivalent of the explosive; d is the water depth of the underwater explosion of the explosive, and is only used in the case of underwater explosion, and is taken as 0 in the case of air explosion; f represents the support vector machine classification / regression model.

[0122] The accuracy evaluation and parameter optimization of the model in the model generation process are specifically as follows: taking the mean square error as the loss function, as shown in Figure 3 , cross-validation, grid search, and genetic algorithm are used to optimize the model parameters C and G of the support vector machine, so as to improve the prediction accuracy of the model. The mean square error MSE calculation formula is as follows:

[0123]

[0124] wherein n is the sample number, is the predicted value, and y i is the true value.

[0125] S04: Using the prediction model corresponding to the working condition information established in step S03, the data obtained in step S01 is taken as a test set to predict the working condition information corresponding to the sea test, i.e. the charge weight and the explosive detonation point coordinates.

[0126] S05: Comparing the working condition obtained in step S04 with all known working conditions of the test, i.e. comparing the charge weight and the detonation point coordinates. The test detonation point space coordinates are analyzed through video, picture and other data to obtain a distribution range, and the predicted value of the coordinates is considered reasonable within the distribution range, and the predicted value is considered reasonable within the error range of 20%. For the predicted value with obvious difference, return and execute steps S03, S04 and S05 until the predicted value is within the error range.

[0127] S06: Taking the working condition information predicted in step S04 as an input condition, inputting into the finite element simulation calculation model established in step S02 to simulate and calculate the test to obtain the simulation result of the test. Using the corresponding post-processing software, processing the simulation result to form corresponding video animation data, physical field data and the like, as shown in Figure 2 , Figure 3 , obtaining the explosion response process, full-ship response data and physical field data, expanding the limited measurement point data into a three-dimensional full-field data, visualizing the data, and packaging the result file to form the final virtual experiment platform.

[0128] The above merely illustrates the preferred embodiments of the present application, and is not used to limit the present application, and various changes and modifications can be made by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall fall within the protection scope of the present application.

Claims

1. A virtual test method based on multi-source data fusion, characterized in that, Comprise the following steps are: S01: Obtain effective data of measuring points by test or existing test, including acceleration measurement data, pressure measurement data, strain measurement data, and data processing, and establish simplified characteristic value; S02: According to the target ship information, establish finite element model, and establish random working condition, batch calculate random working condition, extract the effective data corresponding to the test in the result, that is, acceleration measurement data, pressure measurement data, strain measurement data, and break data, and adopt the same data processing method as the test data, obtain the simplified characteristic value, so as to establish the measuring information corresponding to various different working conditions, normalize the original data, form the data used for machine learning; S03: Adopt machine learning algorithm, take result data as characteristic, that is, independent variable X, and take working condition information as target, that is, dependent variable Y, train the basic database generated in step S02, wherein the break data adopts the classifiable algorithm of machine learning for classification model training, and the acceleration measurement data, pressure measurement data and strain measurement data adopt the regressive algorithm for regressive model training, so as to respectively establish corresponding machine learning prediction model under all working condition information conditions, and perform accuracy evaluation and parameter optimization of the model; Machine learning algorithm, specifically, for break data, firstly, according to whether there is break, support vector machine algorithm is adopted for classification training, and then support vector machine algorithm is adopted for regressive training of specific break data of break; acceleration measurement data, pressure measurement data and strain measurement data all adopt support vector machine algorithm for regressive training according to specific data; Wherein, support vector machine classification / regressive prediction is to take a certain concerned physical quantity as output variable, that is, target set, take explosive XYZ three-dimensional space coordinates, explosive TNT equivalent, and explosive underwater explosion water depth as input variable, that is, characteristic set, through training of data set, support vector machine classification / regressive model of a certain concerned physical quantity and explosive XYZ three-dimensional space coordinates, explosive TNT equivalent, and explosive water depth is obtained, which is shown as follows: （8） wherein, is a certain physical quantity of interest; is the x-coordinate of the explosive shot point; is the y-coordinate of the explosive shot point; is the z-coordinate of the explosive shot point; is the TNT equivalent of the explosive; is the water depth of the explosive underwater explosion, only for underwater explosion case, for air explosion always take 0; f denotes a support vector machine classification / regression model; S04: Adopt the prediction model of corresponding working condition information established in step S03, take the data obtained in step S01 as test set, and predict the working condition information corresponding to the test, that is, charge weight and explosive explosion point coordinates; S05: Compare the working condition obtained in step S04 with all known working conditions of the test, that is, compare the charge weight and explosive explosion point coordinates; wherein the explosive explosion point coordinates of the test are analyzed through video and picture data to obtain a distribution range, and the coordinate prediction value is considered reasonable in the distribution range, and the charge weight prediction value is considered reasonable within the error range of 20%; for the prediction value with obvious difference, return and execute steps S03, S04 and S05, until the prediction value is within the error range; S06: taking the working condition information predicted in step S04 as an input condition, inputting the finite element simulation calculation model established in step S02 to perform simulation calculation on the test, obtaining a simulation result of the test; using corresponding post-processing software to process the simulation result, forming corresponding video animation data and physical field data, visualizing the data, and encapsulating the result file to form a final virtual experiment platform.

2. The virtual test method based on multi-source data fusion according to claim 1, characterized in that, The measurement data processing in steps S01 and S02 establishes a simplified characteristic value, and the specific method is as follows: S01-01: for the acceleration data of a certain measuring point, the acceleration data is converted into shock spectrum data to obtain spectrum velocity, spectrum displacement and spectrum acceleration; specifically, the shock spectrum is obtained according to the following steps: (1) The acceleration spectrum is calculated by the following formula , （1） （2） In the above formula, acceleration time-domain curve of the measuring point measured by the test; is a time variable; is the time point at which the acceleration curve ends; is a circular frequency; (2) The velocity spectrum is calculated as follows , （3） (3) The displacement spectrum is calculated as follows , （4） (4) draw the spectrum line according to the following formula in the logarithmic coordinate system to obtain the shock spectrum; （5） S01-02: for the strain data of a certain measuring point, the stable strain data is averaged; S01-03: for the pressure data of a certain measuring point, if it is an underwater test, the initial shock wave peak value, initial shock wave pulse width, bubble pulsation pressure peak value, bubble pulsation pressure pulse width and bubble pulsation cycle are extracted; if it is an above-water test, the shock wave peak value, pulse width and quasi-static pressure value are extracted; S01-04: For the area data of a certain breach, get the equivalent breach radius r e and extract the maximum length of the real breach l max where the equivalent breach radius r e is calculated as follows: （6） wherein A is the actual measured or simulated breach area.

3. The virtual test method based on multi-source data fusion according to claim 1, characterized in that, The finite element simulation calculation in steps S02 and S06 specifically refers to explosion fluid-structure coupling simulation calculation of a specific working condition, and ALE or CEL method is used for simulation calculation, and the simulation calculation steps include: (1) ship structure modeling, water and air domain modeling, explosive domain modeling; (2) ship structure, water medium, air medium, explosive medium material attribute assignment; (3) fluid domain boundary condition setting; (4) fluid-structure coupling setting; (5) solution time and result output setting; (6) execute solution calculation.

4. The virtual test method based on multi-source data fusion according to claim 1, characterized in that, The original data normalization in step S02 is calculated according to the following formula: （7） wherein, X nom is the normalized data, X is the original data of the variable to be normalized, X min , X max is the minimum and maximum value of the original data of the variable to be normalized.

5. The virtual test method based on multi-source data fusion according to claim 1, characterized in that, The accuracy evaluation and parameter optimization of the model in step S03 are as follows: taking mean square error as a loss function, using cross-validation and grid search; and combining genetic algorithm, the model parameters C and G of support vector machine are optimized to improve the prediction accuracy of the model; wherein, the mean square error MSE calculation formula is: （9） wherein n is the number of samples, is the predicted value, is the true value.

Images