A high-precision prediction method for target plate deformation under dual explosive source coupling loading

Abstract

The present invention relates to a high-precision prediction method for target plate deformation under dual-explosive source coupling loading, which belongs to the technical field of explosion damage effectiveness evaluation and solves the problem of lack of high-precision prediction method for target plate deformation under dual-explosive source coupling loading in the prior art. The method comprises: using the real scene and simulation scene of the target plate deformation test under dual-explosive source coupling loading to collect the target plate center deformation under typical working conditions; if the absolute value of the average value of the corresponding difference between the target plate center deformation of the real scene and the simulation scene is less than or equal to the preset deviation threshold, using the simulation scene to collect the target plate center deformation under all working conditions to form a simulation data set, and construct a sample data set accordingly; using the sample data set to train and verify the prediction model of the target plate deformation; constructing an input vector according to the new working condition, and using the prediction model to predict and output the dimensionless deformation of the target plate center corresponding to the new working condition, and obtaining the corresponding target plate center deformation after processing.

Description

Technical Field

[0001] The invention relates to the technical field of explosion damage effectiveness evaluation, and in particular to a high-precision prediction method for target plate deformation under dual explosive source coupling loading. Background Art

[0002] In the field of explosion damage effectiveness assessment, it mainly involves the assessment of explosion force field and target vulnerability, among which the target vulnerability assessment is an extremely important issue. The deformation of the target plate under different explosive loads belongs to the scope of target vulnerability assessment. At present, most studies are limited to the deformation problem caused by the explosion load of a single explosive source on the target plate, and the single explosive source generally calculates the shock wave overpressure first and then predicts the target plate deformation through the pressure value acting on the target plate. The prediction model of a single explosive source often oversimplifies the complex explosion and structural dynamics process and ignores some important physical mechanisms; and the high-quality experimental data used to verify and improve the model are relatively limited; in addition, numerical errors are inevitably introduced when solving complex nonlinear dynamic equations.

[0003] With the development of technology, unmanned swarms have developed rapidly. The strategy of using a single ammunition to attack large targets has become outdated. Most unmanned swarms carry dual-explosive source ammunition to explode and damage target plates. Therefore, it is necessary to explore a new method for predicting the deformation of target plates under dual-explosive source coupling loading to meet the needs of sophisticated damage in complex environments in the future.

[0004] The main difference between dual-explosion source coupling damage and single-explosion source damage is that the two identical shock waves generated by the dual-explosion source will collide and produce Mach reflection, making them more powerful. If the mass of an explosive is evenly divided into two explosives, and the two explosives are separated by a certain distance and then detonated simultaneously, the destructive power will be much greater than the power of a single-explosion source of the same mass, achieving the effect of 1+1>2. However, under the same explosive mass, the dual-explosion source will produce different shock wave overpressures at different dual-explosion source spacings, which is more difficult to predict and calculate.

[0005] In summary, the current single-explosion source model prediction still has the defect of insufficient prediction accuracy, and there is no mature dual-explosion source model for prediction at this stage. At the same time, compared with a single-explosion source, the single-explosion source model cannot accurately reflect the characteristics of multiple explosion sources and cannot capture complex factors such as the interaction between explosion sources. Therefore, it is urgent to study the high-precision prediction method of the target plate deformation under dual-explosion source coupling loading. Summary of the invention

[0006] In view of the above analysis, an embodiment of the present invention aims to provide a high-precision prediction method for target plate deformation under dual-explosive source coupling loading, so as to solve the problem of lack of a high-precision prediction method for target plate deformation under dual-explosive source coupling loading in the prior art.

[0007] The present invention discloses a high-precision prediction method for target plate deformation under dual-explosion source coupling loading, the method comprising:

[0008] The target plate center deformation under typical working conditions was collected by using the real scene and simulation scene of the target plate deformation test under dual explosive source coupling loading.

[0009] Determine whether the absolute value of the average value of the corresponding difference between the target plate center deformation under the real scene and the simulation scene under the typical working condition is less than or equal to the preset deviation threshold value. If so, use the simulation scene to collect the target plate center deformation under all working conditions to form a simulation data set of target plate deformation under dual explosive source coupling loading;

[0010] The target plate center deformation is normalized and dimensionless to obtain the corresponding target plate center dimensionless deformation; the characteristic variables in the working conditions are extended to construct a sample data set for characterizing the relationship between the characteristic variables of each working condition, the extended characteristic vector and the corresponding target plate center dimensionless deformation; the prediction model of the target plate deformation is trained and verified using the sample data set;

[0011] The characteristic variables and extended characteristic vectors corresponding to the new working condition are constructed as input vectors of the prediction model, and the prediction model predicts and outputs the dimensionless deformation of the target plate center corresponding to the new working condition; the dimension is added to the dimensionless deformation of the target plate center to obtain the deformation of the target plate center corresponding to the new working condition.

[0012] On the basis of the above scheme, the present invention also makes the following improvements:

[0013] Furthermore, the working conditions include the distance between the two explosion sources. , Explosion distance and target plate thickness ; Perform feature expansion on the characteristic variables in the working condition and obtain the following extended feature vector:

[0014] Dual explosion source spacing ratio , the ratio of target plate thickness to explosion distance , the area of ​​the double explosion source effect area , the square of the distance between the two explosion sources , proportional distance ;

[0015] (1)

[0016] In the formula, is the total mass of the double-source explosive.

[0017] Furthermore, the working condition also includes the explosive energy loading area; the characteristic variables in the working condition are extended to obtain the following extended characteristic vector:

[0018] (2)

[0019] In the formula, is the explosion energy density of the double explosion source; is the explosive energy loading area, The explosive heat of explosives.

[0020] Furthermore, the target plate center deformation is normalized and dimensionless to obtain the corresponding target plate center dimensionless deformation.

[0021] The ratio of the center deformation of each target plate in the simulation data set to its corresponding target plate thickness is calculated to obtain the corresponding dimensionless deformation of the center of the target plate.

[0022] Further, the sample data set is used to train and verify the prediction model of target plate deformation, and the following is performed:

[0023] The random forest regression model and the K nearest neighbor regression model are used as base models respectively; the characteristic variables and extended characteristic vectors in the sample data set are used as input variables, and the corresponding dimensionless deformation of the target plate center is used as the output variable. The two base models are trained and verified respectively, and the hyperparameters of the base models are optimized to obtain the verified base models;

[0024] The sample data sets are re-input into the two verified base models respectively to obtain the predicted values ​​of the corresponding base models; the characteristic variables and extended characteristic vectors in the sample data sets are used as input variables, and the predicted values ​​of the corresponding base models are used as output variables to obtain the predicted value data sets of the two base models; the predicted value data sets of the two base models are combined as the model data set of the prediction model for testing the deformation of the target plate; and the model data set is used to train and verify the prediction model of the deformation of the target plate.

[0025] Furthermore, the dimensionless deformation of the target plate center is increased to obtain the target plate center deformation corresponding to the new working condition, and the following is executed:

[0026] The dimensionless deformation of the target plate center corresponding to the new working condition is multiplied by the target plate thickness in the new working condition to obtain the deformation of the target plate center corresponding to the new working condition.

[0027] Furthermore, a simulation scenario of the target plate deformation test under dual explosive source coupling loading is constructed through simulation software; in the simulation scenario, the Johnson-Cook constitutive model considering the strain hardening, strain rate strengthening and temperature softening effects of metal materials is used to define the target plate material. The Johnson-Cook constitutive model is as follows:

[0028] (3)

[0029] In the formula, represents the flow stress, i.e. the stress of the material at a specific plastic strain; is the dimensionless plastic strain rate, is the strain rate, is the reference strain rate; is the dimensionless temperature, is the deformation temperature, is the melting point temperature, is room temperature; It is the temperature sensitivity index, which describes the sensitivity of material flow stress to temperature; is the yield stress at reference temperature and reference strain rate; is the strain hardening coefficient; is the strain hardening exponent; is the strain rate sensitivity coefficient; is the plastic strain.

[0030] Further, if the absolute value of the average value of the corresponding difference between the target plate center deformation under the typical working condition in the real scene and the simulation scene is greater than the preset deviation threshold, the simulation parameters of the simulation scene are adjusted, and the target plate deformation test simulation scene under the dual explosive source coupling loading is used again to collect the target plate center deformation under the typical working condition, and jump to the step of judging whether the absolute value of the average value of the corresponding difference between the target plate center deformation under the typical working condition in the real scene and the simulation scene is less than or equal to the preset deviation threshold;

[0031] The adjustment of the simulation parameters of the simulation scenario refers to adjusting the strain parameters in the Johnson-Cook constitutive model, including: strain hardening coefficient, strain hardening exponent, strain rate sensitivity coefficient, i.e., melting point temperature.

[0032] Furthermore, if the average value of the difference between the target plate center deformation under typical working conditions in the simulation scenario and the real scenario is greater than a preset deviation threshold, the strain parameter in the Johnson-Cook constitutive model is adjusted in the simulation scenario with the purpose of reducing the target plate center deformation.

[0033] Furthermore, if the average value of the difference between the target plate center deformation under typical working conditions in the real scenario and the simulation scenario is greater than a preset deviation threshold, the strain parameter in the Johnson-Cook constitutive model is adjusted in the simulation scenario with the purpose of increasing the target plate center deformation.

[0034] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0035] The high-precision prediction method for target plate deformation under dual explosive source coupling loading provided by the present invention has the following advantages:

[0036] (1) By combining the real scene with the simulation scene, the accuracy of the target plate center deformation under various working conditions collected in the simulation scene is ensured, which reduces the difficulty of obtaining the target plate center deformation under different working conditions. The deformation of target plates of different thicknesses, materials, and shapes can be predicted, which has the scalability to be modified.

[0037] (2) Based on the special problems of the target plate deformation test under dual explosive source coupling loading, the influence relationship between the deformation of the characteristic variables involved in the working condition and the deformation of the target plate center is analyzed from the physical and mathematical perspectives. The characteristic variables in the working condition are expanded in a targeted and adaptive manner. The nonlinear relationship between the characteristic variables and the deformation of the target plate center and their interaction can be captured, thereby enhancing the generalization ability of the model and reducing the risk of underfitting of the model. The comprehensiveness, completeness, adaptability and generalization performance of the sample data set are effectively guaranteed.

[0038] (3) Random forest regression (RFR) and K nearest neighbor regression (KNNR) are used as base models, combined with the 10-fold cross-validation method, so that the model is trained and validated multiple times on different training sets. By integrating the genetic algorithm and the particle swarm optimization algorithm (GA-PSO), a balance is achieved between global search and local search, avoiding falling into the local optimum, improving the hyperparameter optimization effect, and thus improving the model prediction accuracy. Considering that the prediction rate of the KNNR model is low and the residual distribution of the RFR model is not very concentrated at the zero point, the stacking model can combine the advantages of the two to form a prediction model with high accuracy and good residual concentration. LinearRegression is used as the stacking model to combine the prediction results of the optimized RFR and KNNR models, which further improves the prediction ability of the overall model and ensures the stability and accuracy of the model. That is, when predicting the deformation of the target plate under dual-explosion source coupling loading, the RFR and KNNR models were used as base models, and LinearRegression was used as a stacking model for prediction. The prediction accuracy and robustness of the stacking model will be higher than that of the base model, which can increase the accuracy and robustness of the model prediction, enabling it to be used in new unknown data sets and data sets with certain errors.

[0039] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0041] Figure 1 This is a flow chart of a method for high-precision prediction of target plate deformation under dual explosive source coupling loading provided in Example 1 of the present invention;

[0042] Figure 2 This is a flow chart of another method for high-precision prediction of target plate deformation under dual explosive source coupling loading provided in Example 1 of the present invention;

[0043] Figure 3 A schematic diagram of the positions of the deformation points to be recorded provided in Example 1 of the present invention;

[0044] Figure 4 A schematic diagram of a simulation scenario provided in Example 1 of the present invention;

[0045] Figure 5 This is a scatter plot of the data set distribution after preprocessing in Example 2 of the present invention;

[0046] Figure 6 It is a scatter plot of the predicted value and the actual value of the KNNR model in Example 2 of the present invention;

[0047] Figure 7 It is a residual distribution diagram of the predicted value and the actual value of the KNNR model in Example 2 of the present invention;

[0048] Figure 8 is a scatter plot of the predicted value and the actual value of the RFR model in Example 2 of the present invention;

[0049] Fig. 9 is a residual distribution diagram of the predicted value and the actual value of the RFR model in Example 2 of the present invention;

[0050] Fig.10 A scatter plot of the predicted value and the actual value of the stacking model in Example 2 of the present invention;

[0051] Fig.11 This is a residual distribution diagram of the predicted value and the actual value of the stacking model in Example 2 of the present invention. DETAILED DESCRIPTION

[0052] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0053] A specific embodiment 1 of the present invention discloses a high-precision prediction method for target plate deformation under dual explosive source coupling loading. The flowchart of the method is as follows: Figure 1 and Figure 2 The specific steps are described as follows.

[0054] Step S1: using the real scene and simulation scene of the target plate deformation test under dual explosive source coupling loading respectively, the target plate center deformation under typical working conditions is collected.

[0055] In actual application, considering the high cost involved in carrying out the real scene of the target plate deformation test under dual explosive source coupling loading and conducting the test, it is difficult to use the real scene to collect the target plate center deformation under all working conditions. Therefore, this embodiment constructs a simulation scene of the target plate deformation test under dual explosive source coupling loading. When the data collection results of the simulation scene and the real scene are similar, it can be considered to use the simulation scene to collect the target plate center deformation under all working conditions to replace the target plate center deformation collected under the corresponding working conditions in the real scene, and perform subsequent data processing and data prediction on this basis.

[0056] During the specific implementation process, after selecting the target plate parameters (such as the shape, size, material, etc. of the target plate), the real scene of the target plate deformation test under the dual explosive source coupling loading can be carried out. At the same time, the simulation scene of the target plate deformation test under the dual explosive source coupling loading can be constructed through simulation software (such as LS-DYNA finite element simulation software). Through repeated demonstration and analysis of the target plate deformation test process under the dual explosive source coupling loading, this embodiment selects the explosion distance, the dual explosive source spacing, and the target plate thickness as the core extended feature vectors that affect the target plate center deformation. Therefore, each working condition is composed of the explosion distance, the dual explosive source spacing and the target plate thickness. Among them, the explosion distance refers to the straight-line distance from the midpoint of the dual explosive source to the target plate, and the target plate center deformation refers to the deformation along the loading direction of the coupled shock wave generated by the dual explosive source. Exemplarily, the position of the recording point (i.e., the center of the target plate) is as follows Figure 3 shown.

[0057] In the process of constructing the simulation scene of the target plate deformation test under dual explosive source coupling loading through simulation software, in order to save calculation time, the simulation scene can be constructed through a 1 / 2 symmetric model. The schematic diagram of the simulation scene is shown in Figure 4 In the specific implementation process, the air domain and explosive fluid can be defined using the S-ALE structured transient fluid-solid interaction algorithm. Specifically, the size and mesh size of the air domain can be defined by the following keywords: At the same time, considering the calculation scale and efficiency, after comparing with the test results of the full-scale watershed, the watershed size and air domain size of the calculation model of the simulation scenario are finally determined, and the explosive size is the same as the explosive size in the real scene test, and the keyword Filling is performed in the air domain. The target plate is described by SHELL unit. Figure 4For example, set the size of the air domain to , set the grid size to 8mm, and set the target plate geometry to The load acting area is , simple support constraints are set at the 100mm boundary of each side, fixed support constraints are set at the 16 opening positions of the target plate, and the target plate thickness is In order to better match the calculation time and accuracy, the target plate grid is set to 10 mm, a symmetric boundary is set on the Oxz plane, and the remaining surfaces are set as pressure transmission boundaries.

[0058] In the simulation scenario, the target plates are all made of metal materials. The Johnson-Cook constitutive model that considers the strain hardening, strain rate strengthening and temperature softening effects of metal materials is used to define the target plate material. The Johnson-Cook constitutive model is as follows:

[0059] (1)

[0060] In the formula, represents the flow stress, i.e. the stress of the material at a specific plastic strain; is the dimensionless plastic strain rate, is the strain rate, is the reference strain rate; is the dimensionless temperature, is the deformation temperature, is the melting point temperature, is room temperature; It is the temperature sensitivity index, which describes the sensitivity of material flow stress to temperature; is the yield stress at reference temperature and reference strain rate; is the strain hardening coefficient; is the strain hardening exponent; is the strain rate sensitivity coefficient; is the plastic strain.

[0061] TNT explosive properties The model is used to describe the energy diffusion and volume change of the detonation products by using the JWL state equation. The JWL state equation is as follows:

[0062] (2)

[0063] In the formula, Indicates the pressure of detonation products; , , , , is the explosive material constant; is the relative volume of detonation products; is the initial internal energy.

[0064] Air use The constitutive model is defined by the LINRAR_POLYNOMIAL linear polynomial state equation, which is as follows:

[0065] (3)

[0066] In the formula, is the gas pressure; is the relative volume; is the material constant of air, is the initial internal energy of air.

[0067] If the average value of the difference between the target plate center deformation collected in the simulation scenario and the target plate deformation obtained in the real scenario test is greater than or equal to the deviation threshold, the strain hardening parameter ( and )、strain rate sensitivity coefficient( ) and the temperature softening effect parameter ( ), which can effectively control the maximum displacement of the target plate center in the numerical simulation, making it closer to the actual test results. In the specific implementation process, the corresponding parameters can be adjusted according to the deviation direction between the simulation results and the test results.

[0068] Step S2: Determine whether the absolute value of the average value of the corresponding difference between the target plate center deformation in the real scene and the simulation scene under typical working conditions is less than or equal to the preset deviation threshold. If so, use the simulation scene to collect the target plate center deformation under all working conditions to form a simulation data set of the target plate deformation under dual explosive source coupling loading, and execute step S3; if not, jump to step S5.

[0069] Considering the cost, only a few sets of test conditions may be conducted in real scenarios. However, the distance between the two explosion sources can be increased according to the simulation scenario. , Explosion distance and target plate thickness Different data values ​​of the three characteristic variables are used to expand the working conditions, thereby enriching the data set. Therefore, in the specific implementation process, the deviation threshold requirements can be set according to the actual accuracy requirements. It is generally believed that if the average value of the errors of all working conditions is within 15%, it can be said that the numerical simulation model and parameter settings in the simulation scenario are relatively accurate, and the working conditions can be expanded through the simulation scenario, and data under all working conditions can be collected. At this time, the simulation scenario of the target plate deformation test under dual blast source coupling loading can be used to collect the target plate center deformation under all working conditions to form a simulation data set. In this embodiment, the simulation data set is used to characterize the corresponding relationship between different working conditions and the target plate center deformation. In the specific implementation process, under the conditions of the verified finite element numerical simulation, the numerical simulation conditions (dual blast source spacing , Explosion distance and target plate thickness Three data variables) are used for numerical simulation to obtain the target plate center deformation under each working condition, and a simulation data set of the target plate deformation under dual explosive source coupling loading is constructed. For example, the simulation data set It can be expressed as:

[0070] (4)

[0071] In the formula, is the distance between the two explosion sources; For the explosion distance; is the target plate thickness; the subscript indicates the working condition, is the total number of all working conditions; Indicates working conditions The center deformation of the target plate under .

[0072] In the actual practice, although the simulation scenario can obtain the target plate center deformation under more working conditions, the simulation time of the simulation scenario is long under each working condition, and it is difficult to achieve the rapid prediction of the target plate center deformation under the newly added working conditions. Therefore, it is still necessary to explore the rapid prediction method of the target plate center deformation. The specific explanation is as follows.

[0073] Step S3: normalize and dimensionlessly process the target plate center deformation to obtain the corresponding target plate center dimensionless deformation; perform feature expansion on the characteristic variables in the working conditions to construct a sample data set for characterizing the relationship between the characteristic variables of each working condition, the expanded characteristic vectors and the corresponding target plate center dimensionless deformation; and use the sample data set to train and verify the prediction model of the target plate deformation.

[0074] The specific implementation process of step S3 is described as follows.

[0075] Step S31: Calculate the ratio of the center deformation of each target plate in the simulation data set to its corresponding target plate thickness to obtain the corresponding dimensionless center deformation of the target plate.

[0076] In this step, considering that the uneven distribution of data and different spans of data will also affect the prediction accuracy and convergence speed of the later prediction model, it is necessary to convert the original data to optimize the data quality in order to effectively improve the prediction performance of the prediction model. Since the simulation data set is obtained in a simulation scenario, there is no data anomaly such as missing data. Therefore, in this embodiment, it is only necessary to normalize the target plate center deformation, that is, the output quantity target plate center deformation in the simulation data set is converted into Divide it by the target plate thickness and make it dimensionless to obtain the dimensionless deformation of the target plate center under the action of the double explosion source. .

[0077] (5)

[0078] In the formula, is the target plate center deformation obtained in the simulation scenario; is the corresponding target plate thickness.

[0079] In this embodiment, by processing the dimensionless deformation, the target plate center deformation corresponding to different characteristic variables can be compressed within a certain range, thereby reducing the span between the target plate center deformation corresponding to different characteristic variables, which can effectively reduce the training difficulty of the prediction model and improve the prediction accuracy.

[0080] Step S32: performing feature expansion based on the characteristic variables in the working condition to obtain an extended feature vector associated with the target plate center deformation.

[0081] In this embodiment, considering that the working conditions only involve three characteristic variables (the distance between the two explosion sources , Explosion distance and target plate thickness ), while fewer feature variables will increase the risk of underfitting of the model and it is easy to lose key information, thereby limiting the decision-making ability and generalization ability of the prediction model, making it perform poorly when facing unknown data, and making the prediction model more sensitive to data noise. Therefore, it is necessary to analyze the characteristic variables in the working conditions, increase the number of characteristic variables that have a correlation with the center deformation of the target plate through feature engineering methods, and enhance the generalization ability of the prediction model. In this embodiment, by conducting an in-depth principle analysis of the target plate deformation test under dual-explosion source coupling loading, more extended feature variables with physical and mathematical significance are added, which can more comprehensively describe the experimental conditions and the response of the target plate, improve the prediction accuracy and robustness of the later prediction model, and enable it to more comprehensively describe the dynamic response of the target plate under dual-explosion source coupling loading. Specifically, this embodiment introduces the following extended feature vector:

[0082] (1) Dual explosion source spacing ratio

[0083] Under explosive load, the coupling effect between the two explosive sources has an important influence on the deformation of the target plate. and explosion distance ratio) It can help the prediction model better capture this coupling effect. When the proportional distance is large, the distance between the two explosion sources is large relative to the explosion center, and the interaction of the explosion waves may be weakened; when the proportional distance is small, the interaction of the explosion waves may be enhanced.

[0084] (2) Ratio of target plate thickness to explosion distance

[0085] The ratio of target plate thickness to explosion distance (i.e. target plate thickness Explosion distance ratio) It reflects the target plate's ability to resist deformation when subjected to shock waves. Thicker target plates have higher resistance to deformation when shock waves arrive.

[0086] (3) Area of ​​the double explosion source area

[0087] The area of ​​the double explosion source area (Double explosion source spacing and explosion distance The product of ( ) reflects the combined effect of the distance between the two explosive sources and the explosion distance on the target plate deformation, which helps to describe the intensity of the coupling between the two explosive sources and the influence of this coupling on the target plate.

[0088] (4) The square of the distance between the two explosion sources

[0089] In explosion mechanics and material mechanics, there is often a nonlinear relationship between the output (target plate center deformation) and the input variables (dual explosion source spacing, explosion distance, target plate thickness). Simple linear features may not be able to fully capture these complex relationships, and by introducing secondary features, this nonlinear relationship can be better modeled.

[0090] The distance between the two explosion sources not only affects the superposition effect of the explosion wave, but also this effect may be nonlinear. The nonlinear effect of the distance between the two explosive sources on the target plate deformation can be better described.

[0091] (5) Proportional distance

[0092] Generally, during the propagation of explosion shock waves, the intensity will decrease as the distance increases, and this attenuation is nonlinear. It is usually expressed using proportional distance.

[0093] (6)

[0094] In the formula, For the explosion distance; is the total mass of the double-source explosive. The nonlinear relationship between the intensity of explosion shock wave and distance can be described more accurately.

[0095] During the specific implementation process, if the following characteristic variables are also set in the working conditions: explosive energy loading area; the following extended characteristic variables can be considered to be added.

[0096] (6) Explosive energy density of explosives

[0097] Explosive energy density of explosives As a characteristic quantity, explosion energy density reflects the intensity of the explosion load per unit area. A higher explosion energy density means a stronger shock wave force.

[0098] (7)

[0099] In the formula, is the explosion energy density of the double explosion source, J / m 2 ; is the total mass of the double-source explosive, kg; is the explosion heat of the explosive, J / kg; is the explosive energy loading area (loaded area on the target plate), m 2 .

[0100] Step S33: constructing a sample data set for characterizing the relationship between the characteristic variables, extended characteristic vectors and their corresponding dimensionless deformation of the target plate center for each working condition.

[0101] According to the above analysis, after expanding the feature matrix, the obtained sample data set is expressed as:

[0102] (8)

[0103] Step S34: using the sample data set to train and test a prediction model for target plate deformation.

[0104] Step S341: The random forest regression (RFR) model and the K nearest neighbor regression (KNNR) model are respectively used as base models; the characteristic variables and extended characteristic vectors in the sample data set are used as input variables, and the corresponding dimensionless deformation of the target plate center is used as the output variable. The two base models are trained and verified respectively, and the hyperparameters of the base models are optimized to obtain the verified base models.

[0105] Preferably, in the specific implementation process, the constructed random forest regression model (RFR) and K nearest neighbor regression (KNNR) model are respectively used as base models, and the sample data set is divided into 10 groups of sub-sample data sets by 10-fold crossover. During the training and verification process of the base model, the base model is trained and verified 10 times on 10 different sub-sample data sets. That is: in each training process, one sub-sample data set is used as the verification set, and the remaining 9 sub-sample data sets are used as the training set. The base model is trained using the training set, and the base model is verified using the verification set to obtain a verified base model.

[0106] In order to achieve the best prediction accuracy for each base model, the hyperparameters in the base model need to be optimized during the training process. In order to solve the problem that the GA optimization algorithm and the PSO optimization algorithm have limitations when performing local and global searches and are prone to falling into local optimality, a GA-PSO hybrid method that integrates the genetic algorithm and the particle swarm optimization algorithm is used to utilize the advantages of the genetic algorithm in global search and the efficiency of the particle swarm algorithm in fast local search. After 10 fold crossovers in the above step S331, the hyperparameters of random forest regression (RFR) and K nearest neighbor regression (KNNR) are optimized. Among them, the hyperparameters optimized by KNNR are 'n_neighbors', and the hyperparameters optimized by RFR are 'n_estimators', 'max_depth', and 'random_state'. The root mean square error is used. and the coefficient of determination They are optimized together as the fitness function, and the calculation formula is as follows.

[0107] (9)

[0108] (10)

[0109] In the formula, is the predicted value of the base model; is the actual value of the base model, is the average value of the measured values ​​of the base model, Represents the logarithm of the predicted value and the actual value.

[0110] Step S342: re-input the sample data set into the two verified base models respectively to obtain the predicted values ​​of the corresponding base models; use the characteristic variables and extended characteristic vectors in the sample data set as input variables, and the predicted values ​​of the corresponding base models as output variables to obtain the predicted value data sets of the two base models; combine the predicted value data sets of the two base models as the model data set of the prediction model for testing the deformation of the target plate; and use the model data set to train and verify the prediction model of the deformation of the target plate.

[0111] That is, in step S342, the optimized hyperparameters are brought back into the base model to obtain the predicted value data set of each base model. The predicted value data sets of the random forest regression (RFR) and K nearest neighbor regression (KNNR) models are respectively expressed as and , the two prediction data sets are combined into a complete data set again (i.e., model dataset), the specific form is:

[0112] , , .

[0113] Preferably, in this embodiment, a stacked model (also called a Stacked model) constructed by LinearRegression is used as a prediction model for the deformation of the test target plate. As the model data set, the model data set is divided into a training set and a validation set in a ratio of 8:2 to train and verify the prediction model of the target plate deformation.

[0114] Step S4: constructing the characteristic variables and extended characteristic vectors corresponding to the new working condition as the input vector of the prediction model, and using the prediction model to predict and output the dimensionless deformation of the target plate center corresponding to the new working condition; adding dimensions to the dimensionless deformation of the target plate center to obtain the deformation of the target plate center corresponding to the new working condition.

[0115] It should be noted that, since the normalized dimensionless processing is performed in step S3, the output result of the prediction model here is the dimensionless deformation of the target plate center. If the target plate center deformation corresponding to the new working condition is to be obtained, the dimensionless deformation of the target plate center needs to be multiplied by the target plate thickness in the new working condition. This corresponds to the process of step S31.

[0116] Step S5: If the absolute value of the average value of the corresponding difference between the real scene and the simulation scene under the typical working condition of the target plate center deformation is greater than the preset deviation threshold, the simulation parameters of the simulation scene are adjusted, and the target plate deformation test simulation scene under the dual explosive source coupling loading is used again to collect the target plate center deformation under the typical working condition, and jump to step S2.

[0117] Step S51: If the average value of the difference between the target plate center deformation under typical working conditions in the simulation scenario and the real scenario is greater than a preset deviation threshold, in the simulation scenario, the strain parameter in the Johnson-Cook constitutive model is adjusted to reduce the target plate center deformation.

[0118] That is, when the simulation results are too large (the deformation of the target plate center is greater than the test results of the real scene), reducing the strain parameters in the Johnson-Cook constitutive model includes:

[0119] (1) Strain hardening parameters in the Johnson-Cook constitutive model and : Reduce the strain hardening coefficient or strain hardening exponent , which reduces the hardening effect of the material under large plastic strains, making the material softer and thus reducing the maximum displacement.

[0120] (2) Strain rate sensitivity coefficient : Reduce the strain rate sensitivity coefficient , which reduces the material's sensitivity to changes in strain rate, weakens the hardening effect of the material at high strain rates, and thus reduces the maximum displacement.

[0121] (3) Temperature softening effect parameters: Increasing the influence of the temperature softening effect (increasing the melting point temperature) makes the material softer under high temperature conditions, thereby reducing the maximum displacement.

[0122] Step S52: If the average value of the difference between the target plate center deformation in the real scene and the simulation scene under typical working conditions is greater than the preset deviation threshold, the strain parameter in the Johnson-Cook constitutive model is adjusted in the simulation scene with the purpose of increasing the target plate center deformation.

[0123] That is, when the simulation results are too small (the deformation of the target plate center is smaller than the test result of the real scene), reducing the strain parameters in the Johnson-Cook constitutive model includes:

[0124] (1) Strain hardening parameters and : Increase the strain hardening coefficient or strain hardening exponent , which will enhance the hardening effect of the material under large plastic strain, making the material harder and thus increasing the maximum displacement.

[0125] (2) Strain rate sensitivity coefficient :Increase the strain rate sensitivity coefficient , which will enhance the material's sensitivity to changes in strain rate, making the hardening effect of the material stronger at high strain rates, thereby increasing the maximum displacement.

[0126] (3) Temperature softening effect parameters: Reduce the influence of the temperature softening effect (reduce the melting point temperature) so that the material becomes harder under high temperature conditions, thereby increasing the maximum displacement.

[0127] The high-precision prediction method for target plate deformation under dual explosive source coupling loading provided in this embodiment has the following advantages:

[0128] (1) By combining the real scene with the simulation scene, the accuracy of the target plate center deformation under various working conditions collected in the simulation scene is ensured, which reduces the difficulty of obtaining the target plate center deformation under different working conditions. The deformation of target plates of different thicknesses, materials, and shapes can be predicted, which has the scalability to be modified.

[0129] (2) Based on the special problems of the target plate deformation test under dual explosive source coupling loading, the influence relationship between the deformation of the characteristic variables involved in the working condition and the deformation of the target plate center is analyzed from the physical and mathematical perspectives. The characteristic variables in the working condition are expanded in a targeted and adaptive manner. The nonlinear relationship between the characteristic variables and the deformation of the target plate center and their interaction can be captured, thereby enhancing the generalization ability of the model and reducing the risk of underfitting of the model. The comprehensiveness, completeness, adaptability and generalization performance of the sample data set are effectively guaranteed.

[0130] (3) Random forest regression (RFR) and K nearest neighbor regression (KNNR) are used as base models, combined with the 10-fold cross-validation method, so that the model is trained and validated multiple times on different training sets. By integrating the genetic algorithm and the particle swarm optimization algorithm (GA-PSO), a balance is achieved between global search and local search, avoiding falling into the local optimum, improving the hyperparameter optimization effect, and thus improving the model prediction accuracy. Considering that the prediction rate of the KNNR model is low and the residual distribution of the RFR model is not very concentrated at the zero point, the stacking model can combine the advantages of the two to form a prediction model with high accuracy and good residual concentration. LinearRegression is used as the stacking model to combine the prediction results of the optimized RFR and KNNR models, which further improves the prediction ability of the overall model and ensures the stability and accuracy of the model. That is, when predicting the deformation of the target plate under dual-explosion source coupling loading, the RFR and KNNR models were used as base models, and LinearRegression was used as a stacking model for prediction. The prediction accuracy and robustness of the stacking model will be higher than that of the base model, which can increase the accuracy and robustness of the model prediction, enabling it to be used in new unknown data sets and data sets with certain errors.

[0131] The specific embodiment 2 of the present invention is used to verify the feasibility and effect of the high-precision prediction method of the target plate deformation under the dual-explosion source coupling loading in the embodiment 1. In this embodiment, the target plate parameters are set as follows: The Q235 metal target plate has an explosive load loading area of According to the high-precision prediction method of target plate deformation under dual explosive source coupling loading, the required input parameters are given, as shown in Table 1 below.

[0132] Table 1 Data set extended operating conditions

[0133]

[0134] The data set obtained from the above extended operating condition table is preprocessed, and the distribution scatter plot of the preprocessed data set is as follows: Figure 5 shown.

[0135] According to the above data set, the hyper parameters of the two models are optimized by the GA-PSO hybrid optimization algorithm. The hyper parameters of the model after optimization are shown in Table 2.

[0136] Table 2 Hyperparameters of four prediction models

[0137]

[0138] This embodiment is an explosion load loading area of A case of a Q235 steel metal square plate, forming a data set by changing different thickness, explosion source spacing, explosion distance and other working condition parameters, and predicting the target plate deformation according to the stacking model. Substituting the above input parameters into the high-precision prediction method of target plate deformation under dual explosion source coupling loading in Example 1, a prediction model of target plate deformation can be obtained.

[0139] According to the base model (Random Forest Regression (RFR) and K Nearest Neighbor Regression (KNNR) model) and stacking model calculated in the above steps, the deformation of the target plate under the double explosive source coupling loading, the model error index table is shown in Table 3 below.

[0140] Table 3 Error indicators of four prediction models

[0141]

[0142] The scatter plot of the model's predicted value and actual value and the residual distribution diagram are as follows Figure 6~Figure 11 shown, specifically, Figure 6 , Figure 7 They are the scatter plot of the predicted value and the actual value of the KNNR model, and the residual distribution diagram. Figure 8 , Fig. 9 They are the scatter plot of the predicted value and the actual value of the RFR model, and the residual distribution diagram. Fig.10 , Fig.11 The scatter plots of the predicted values ​​and actual values ​​of the stacked model and the residual distribution diagram are shown in the figures above. It can be seen from the above figures that the prediction performance and residual distribution of the stacked model are better than those of the individual KNNR or RFR models. The prediction performance of the KNNR model is poor. is 3.23, It is only 0.473, showing a large prediction error and uneven residual distribution. Although the RFR model performs well, is 0.258, The prediction performance of the stacking model is 0.95, but there are still large errors and extreme values ​​in some areas. In contrast, the stacking model combines the advantages of the KNNR and RFR models and greatly improves the overall prediction performance. Reduced to 0.23, The predicted values ​​of the stacked model are closer to the actual values, the residuals are smaller and more concentrated, and the extreme values ​​are reduced, which indicates that the model has smaller prediction errors in different regions.

[0143] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0144] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A high-precision prediction method for target plate deformation under dual explosive source coupling loading, characterized in that: The method comprises: The target plate center deformation under typical working conditions was collected by using the real scene and simulation scene of the target plate deformation test under dual explosive source coupling loading. Determine whether the absolute value of the average value of the corresponding difference between the target plate center deformation under the real scene and the simulation scene under the typical working condition is less than or equal to the preset deviation threshold value. If so, use the simulation scene to collect the target plate center deformation under all working conditions to form a simulation data set of target plate deformation under dual explosive source coupling loading; The target plate center deformation is normalized and dimensionless to obtain the corresponding target plate center dimensionless deformation; the characteristic variables in the working conditions are extended to construct a sample data set for characterizing the relationship between the characteristic variables of each working condition, the extended characteristic vector and the corresponding target plate center dimensionless deformation; the prediction model of the target plate deformation is trained and verified using the sample data set; The characteristic variables and extended characteristic vectors corresponding to the new working condition are constructed as input vectors of the prediction model, and the prediction model predicts and outputs the dimensionless deformation of the target plate center corresponding to the new working condition; the dimension is added to the dimensionless deformation of the target plate center to obtain the deformation of the target plate center corresponding to the new working condition; The working conditions include double explosion source spacing , Explosion distance and target plate thickness ; Perform feature expansion on the characteristic variables in the working condition and obtain the following extended feature vector: Dual explosion source spacing ratio , the ratio of target plate thickness to explosion distance , the area of ​​the double explosion source effect area , the square of the distance between the two explosion sources , proportional distance ; (1) In the formula, is the total mass of the double-source explosive; The working condition also includes the explosive energy loading area; the characteristic variables in the working condition are extended to obtain the following extended characteristic vector: (2) In the formula, is the explosion energy density of the double explosion source; is the explosive energy loading area, The explosive heat of explosives; The simulation scenario of the target plate deformation test under dual explosive source coupling loading is constructed through simulation software; in the simulation scenario, the Johnson-Cook constitutive model considering the strain hardening, strain rate strengthening and temperature softening effects of metal materials is used to define the target plate material. The Johnson-Cook constitutive model is as follows: (3) In the formula, represents the flow stress, i.e. the stress of the material at a specific plastic strain; is the dimensionless plastic strain rate, is the strain rate, is the reference strain rate; is the dimensionless temperature, is the deformation temperature, is the melting point temperature, is room temperature; It is the temperature sensitivity index, which describes the sensitivity of material flow stress to temperature; is the yield stress at reference temperature and reference strain rate; is the strain hardening coefficient; is the strain hardening exponent; is the strain rate sensitivity coefficient; is the plastic strain.

2. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 1 is characterized in that: The target plate center deformation is normalized and dimensionless to obtain the corresponding target plate center dimensionless deformation. Execute: The ratio of the center deformation of each target plate in the simulation data set to its corresponding target plate thickness is calculated to obtain the corresponding dimensionless deformation of the center of the target plate.

3. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 2 is characterized in that: The sample data set is used to train and verify the prediction model of target plate deformation, and the following is performed: The random forest regression model and the K nearest neighbor regression model are used as base models respectively; the characteristic variables and extended characteristic vectors in the sample data set are used as input variables, and the corresponding dimensionless deformation of the target plate center is used as the output variable. The two base models are trained and verified respectively, and the hyperparameters of the base models are optimized to obtain the verified base models; The sample data sets are re-input into the two verified base models respectively to obtain the predicted values ​​of the corresponding base models; the characteristic variables and extended characteristic vectors in the sample data sets are used as input variables, and the predicted values ​​of the corresponding base models are used as output variables to obtain the predicted value data sets of the two base models; the predicted value data sets of the two base models are combined as the model data set of the prediction model for testing the deformation of the target plate; and the model data set is used to train and verify the prediction model of the deformation of the target plate.

4. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 3 is characterized in that: Add the dimension to the dimensionless deformation of the target plate center to obtain the target plate center deformation corresponding to the new working condition. Execute: The dimensionless deformation of the target plate center corresponding to the new working condition is multiplied by the target plate thickness in the new working condition to obtain the deformation of the target plate center corresponding to the new working condition.

5. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 1 is characterized in that: If the absolute value of the average value of the corresponding difference between the target plate center deformation under the typical working condition in the real scene and the simulation scene is greater than the preset deviation threshold, the simulation parameters of the simulation scene are adjusted, and the target plate center deformation under the typical working condition is collected by the simulation scene of the target plate deformation test under the dual explosive source coupling loading, and the step of jumping to judging whether the absolute value of the average value of the corresponding difference between the target plate center deformation under the typical working condition in the real scene and the simulation scene is less than or equal to the preset deviation threshold; The adjustment of the simulation parameters of the simulation scenario refers to adjusting the strain parameters in the Johnson-Cook constitutive model, including: strain hardening coefficient, strain hardening exponent, strain rate sensitivity coefficient, i.e., melting point temperature.

6. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 5 is characterized in that: If the average value of the difference between the target plate center deformation under typical working conditions in the simulation scenario and the real scenario is greater than the preset deviation threshold, the strain parameter in the Johnson-Cook constitutive model is adjusted in the simulation scenario with the purpose of reducing the target plate center deformation.

7. The high-precision prediction method for target plate deformation under dual explosive source coupling loading according to claim 5 is characterized in that: If the average value of the difference between the target plate center deformation under typical working conditions in the real scene and the simulation scene is greater than the preset deviation threshold, the strain parameter in the Johnson-Cook constitutive model is adjusted in the simulation scene with the purpose of increasing the target plate center deformation.

Images