A design method for gradient buffer layer of bulletproof vest structure based on neural network

Abstract

The present invention discloses a method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network, and performs full permutation and combination of N layers of PBDMS-EVA materials of equal thickness to obtain M N buffer layer gradient combinations, select at least 10*N buffer layer gradient combinations of anti-ballistic performance numerical simulation results, and integrate the buffer layer gradient combinations and the anti-ballistic performance numerical simulation results to form a sample data set for training an artificial neural network model; the artificial neural network model is used to perform model training based on the sample data set to obtain an anti-ballistic performance prediction model, and the anti-ballistic performance prediction model is used to predict the anti-ballistic performance of each buffer layer gradient combination in turn, and based on the anti-ballistic performance prediction results, a sample data set is formed to train an artificial neural network model; N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations. The present invention uses an artificial neural network instead of the traditional method dominated by live-fire tests or numerical simulations, which can quickly output the protective performance of the buffer layer and greatly reduce the calculation cost.

Description

Technical Field

[0001] The invention relates to the technical field of bulletproof vest design, and in particular to a method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network. Background Art

[0002] With the continuous development of military technology, the protection needs of combat soldiers are also increasing. As the most commonly used human protection equipment, bulletproof vests have always played a role in protecting the lives of soldiers. Nowadays, bulletproof performance is no longer the only indicator for evaluating bulletproof vests. Ensuring the ability of soldiers to continue fighting after being attacked has become a new challenge in the design of bulletproof vests.

[0003] Traditional bulletproof vests are composed of a bulletproof layer and a buffer layer. The buffer layer is mostly made of foam material and is placed behind the bulletproof layer to reduce the blunt body damage caused by the impact of the bullet to the human body. Among them, EVA (ethylene-vinyl acetate copolymer) is widely used in the buffer layer of bulletproof vests due to its excellent buffering energy absorption performance and good resilience and flexibility. However, EVA materials will deform greatly under impact, making it difficult to provide effective protection for the human body. Studies have shown that compounding polyborosiloxane (PBDMS) with EVA materials can significantly reduce their deformation under impact effects. Therefore, using PBDMS-EVA composite foam material as a buffer layer is a new solution to improve the protective performance of bulletproof vests.

[0004] The performance of PBDMS-EVA materials is closely related to the mixing ratio between the two. Different ratios will produce different material performances. Compared with using a single material as a buffer layer, a gradient structure formed by material differences can have better impact protection performance. PBDMS-EVA materials with different ratios can form several combinations, and it is extremely difficult to study each combination one by one. Summary of the invention

[0005] The purpose of the present invention is to provide a method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network, so as to solve the technical problem in the prior art that PBDMS-EVA materials with different proportions can form several combinations, and it is extremely difficult to study each combination one by one.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0007] A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network comprises the following steps:

[0008] Step S1, setting a buffer layer composed of N layers of PBDMS-EVA materials of equal thickness and a bulletproof layer composed of X layers of fiber materials for the bulletproof vest structure, wherein the types of PBDMS-EVA materials are M types, and the material properties of each type of PBDMS-EVA materials are determined by the ratio of PBDMS and EVA;

[0009] Step S2, sequentially measuring the basic mechanical properties of different types of PBDMS-EVA materials, obtaining the performance of different types of PBDMS-EVA materials under different strain rates by changing the loading rate, and using the display power software LS-DYNA to perform numerical simulation calculation on the bulletproof performance of the bulletproof vest structure using a numerical simulation model, wherein the numerical simulation model includes a projectile, a bulletproof layer, a buffer layer and a backing material characterizing human tissue, and the material model parameters of the bulletproof layer and the buffer layer are consistent with the actual measured values, and the numerical simulation calculation results need to be verified and compared with the live ammunition test to ensure the high fidelity of the numerical simulation model, and the characterization values ​​of the bulletproof performance include the depression depth on the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer and the backing material;

[0010] Step S3: Perform full permutation and combination of N layers of PBDMS-EVA materials with equal thickness to obtain M N buffer layer gradient combinations, determine the required number of numerical simulations according to the number of buffer layers N, to ensure the sufficiency and comprehensiveness of the amount of data used for simulation, select at least 10*N buffer layer gradient combinations of anti-ballistic performance numerical simulation results, and integrate the buffer layer gradient combinations and anti-ballistic performance numerical simulation results to form a sample data set for training the artificial neural network model;

[0011] Step S4, training the model based on the sample data set by the artificial neural network model to obtain a ballistic performance prediction model that characterizes the mapping relationship between the buffer layer gradient combination and the ballistic performance, and using the ballistic performance prediction model to predict the ballistic performance of each buffer layer gradient combination in turn, and based on the ballistic performance prediction results, N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations to design the bulletproof vest structure with the best protective performance.

[0012] As a preferred solution of the present invention, the method of performing numerical simulation calculation using a numerical simulation model includes:

[0013] The bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are modeled using solid units and large deformation calculations are implemented based on the Lagrangian algorithm. The geometric dimensions, boundary conditions, and initial states of the bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are consistent with the actual working conditions, so as to be directly compared with the ballistic target experiment.

[0014] The selection of material simulation models should be determined by the material properties of each component. According to the characteristics of bullets, bulletproof materials, cushioning materials, and backing materials, the "Johnson-Cook" metal material constitutive model, the "CompositeMaterial" composite material constitutive model, the "LowDensityFoam" foam material constitutive model, and the "Cowper-Symonds" elastic-plastic material constitutive model are used respectively;

[0015] Based on the overall symmetry of the numerical simulation model and the material simulation model, numerical simulation is performed on the 1 / 4 area to ensure both calculation accuracy and calculation efficiency. After the numerical simulation is completed, the maximum deformation after stabilization is obtained through the displacement-time curve at the center point of the depression as the depression depth of the backing material, and the peak pressure and pressure distribution in the back area of ​​the buffer layer are directly obtained by setting the contact pressure sensing unit. The impact energy absorption of the bulletproof layer, the buffer layer and the backing material is directly output through the numerical simulation software to obtain the numerical simulation calculation results.

[0016] As a preferred solution of the present invention, the numerical simulation calculation results need to be verified and compared with the live ammunition test, including:

[0017] The concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material measured in the live ammunition test are compared with the concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material in the numerical simulation calculation results to obtain the simulation accuracy of the numerical simulation model;

[0018] The high fidelity of the numerical simulation model is ensured by refining the mesh model of the bullet impact area in the numerical simulation model and setting a reasonable contact stiffness to improve the simulation accuracy of the numerical simulation model.

[0019] As a preferred embodiment of the present invention, the N layers of PBDMS-EVA materials with equal thickness are fully arranged and combined to obtain M N A buffer gradient combination, including:

[0020] For N layers of PBDMS-EVA materials with equal thickness, all permutations and combinations are performed according to M categories to obtain M N The buffer layer gradient combination, the M N The buffer layer gradient combination includes: {i j |j∈[1,N],i j ∈[1,M]}, where i j It is the type of PBDMS-EVA material of the jth layer in the buffer layer gradient combination.

[0021] As a preferred solution of the present invention, the step of determining the required number of numerical simulations according to the number of buffer layers N includes:

[0022] The number of numerical simulations is set to 10 times the number of buffer layers N to ensure the sufficiency and comprehensiveness of the simulation data.

[0023] As a preferred solution of the present invention, the anti-ballistic performance numerical simulation results of selecting at least 10*N buffer layer gradient combinations include:

[0024] The fitting efficiency and fitting accuracy of artificial neural network are taken as two optimization objectives, and based on the two optimization objectives, the N The number of buffer gradient combinations selected in the buffer layer gradient combinations is converted into a multi-objective optimization problem for solution, and the multi-objective optimization algorithm is solved by using a genetic algorithm to determine the number of buffer gradient combinations selected in M N The number of buffer gradient combinations selected from the buffer layer gradient combinations is determined to be 10*N, so as to reduce the amount of numerical simulation tasks as much as possible under the condition of ensuring sufficient number of samples so as to take into account the fitting accuracy and fitting efficiency of the artificial neural network;

[0025] Using gradient change method in M N 10*N buffer layer gradient combinations are selected from the buffer layer gradient combinations to ensure the balance and representativeness of the sample selection.

[0026] As a preferred solution of the present invention, 10*N buffer layer gradient combinations are numerically simulated using a numerical simulation model to obtain the anti-ballistic performance numerical simulation result.

[0027] As a preferred solution of the present invention, the anti-ballistic performance prediction model characterizing the mapping relationship between the buffer layer gradient combination and the anti-ballistic performance is obtained by training the model based on the sample data set through the artificial neural network model, including:

[0028] The buffer layer gradient combination in the sample data set is used as an input item of the artificial neural network model, the ballistic performance numerical simulation result in the sample data set is used as an output item of the artificial neural network model, and the artificial neural network model is used to perform model training based on the input item and the output item to obtain the ballistic performance prediction model;

[0029] The input parameter of the input layer of the artificial neural network model is set to N, and each input parameter is the type of PBDMS-EVA material of each layer in the input item;

[0030] The model expression of the ballistic performance prediction model is:

[0031] [out1,out2,out3]=BP(in[ij ]);

[0032] Wherein, out1 represents the depression depth of the backing material in the anti-ballistic performance value, out2 represents the peak pressure of the back area of ​​the buffer layer in the anti-ballistic performance value, out3 represents the impact energy absorption of the bulletproof layer, buffer layer and backing material in the anti-ballistic performance value, in[i j ] is represented as a buffer layer gradient combination, and BP is represented as an artificial neural network model.

[0033] As a preferred solution of the present invention, the method of using the anti-ballistic performance prediction model to predict the anti-ballistic performance of each buffer layer gradient combination in turn includes:

[0034] M N The buffer layer gradient combinations are sequentially input into the anti-ballistic performance prediction model to obtain M N The anti-ballistic performance value of each buffer layer gradient combination is constructed by the depression depth of the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bullet-proof layer, the buffer layer and the backing material in each buffer layer gradient combination.

[0035] As a preferred embodiment of the present invention, the ballistic performance prediction result is based on M N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations, including:

[0036] The evaluation priorities are the depression depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, buffer layer and backing material, and the M is evaluated according to the evaluation priority. N Screening of buffer layer gradient combinations;

[0037] The evaluation is performed according to the priority N Screening of buffer layer gradient combinations, including:

[0038] Step 1: M N The depth of depression of the backing material in the prediction results of the anti-ballistic performance of the buffer layer gradient combination is compared, among which,

[0039] If the depression depth of the backing material of the buffer layer gradient combination is less than a preset depression depth threshold, the buffer layer gradient combination is classified into the best combination set;

[0040] If the depression depth of the backing material of the buffer layer gradient combination is greater than or equal to a preset depression depth threshold, the buffer layer gradient combination is classified into a non-optimal combination set;

[0041] Step 2: Compare the peak pressures of the back area of ​​the buffer layer in the prediction results of the buffer layer gradient combination anti-ballistic performance in the best combination set, where:

[0042] If the peak pressure of the buffer layer back region of the buffer layer gradient combination is less than a preset peak pressure threshold, retaining the buffer layer gradient combination in the best combination set;

[0043] If the peak pressure of the buffer layer back region of the buffer layer gradient combination is greater than or equal to a preset peak pressure threshold, the buffer layer gradient combination is removed from the non-optimal combination set;

[0044] Step 3: Compare the impact energy absorption of the bulletproof layer, the buffer layer and the backing material in the buffer layer gradient combination anti-ballistic performance prediction results in the best combination set, where:

[0045] If the impact energy absorption of the bulletproof layer and the buffer layer of the buffer layer gradient combination is greater than or equal to the preset absorption threshold, and the impact energy absorption of the backing material is less than the preset absorption threshold, the buffer layer gradient combination is retained in the optimal combination set;

[0046] If the impact energy absorption of the bulletproof layer or the buffer layer of the buffer layer gradient combination is less than the preset absorption threshold, or the impact energy absorption of the backing material is greater than or equal to the preset absorption threshold, the buffer layer gradient combination is removed from the non-optimal combination set;

[0047] Step 4: All buffer layer gradient combinations in the best combination set are taken as the buffer layer gradient combinations that best characterize the anti-ballistic performance.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention uses artificial neural networks to replace the traditional methods dominated by live-fire tests or numerical simulations. It only uses a certain number of numerical simulation results for training, and uses the neural network's strong nonlinear analysis ability, learning ability, and regression ability to predict the parameters of interest. It only needs to input the combination of PBDMS-EVA, and the trained neural network can quickly output the protective performance of the buffer layer, greatly reducing the calculation cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0051] Figure 1 A schematic diagram of stress-strain curves of the PBDMS-EVA material at different strain rates provided in an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of numerical simulation of the ballistic performance of a bulletproof vest provided in an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of a pressure-time curve at the center of a backing material when a bullet impacts an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of the structure of an artificial neural network provided by an embodiment of the present invention;

[0055] Figure 5 A flow chart of a method for designing a gradient buffer layer for a bulletproof vest structure provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] like Figures 1 to 5 As shown, the present invention provides a method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network, comprising the following steps:

[0058] Step S1, setting a buffer layer composed of N layers of PBDMS-EVA materials of equal thickness and a bulletproof layer composed of X layers of fiber materials for the bulletproof vest structure, wherein the types of PBDMS-EVA materials are M types, and the material properties of each type of PBDMS-EVA materials are determined by the ratio of PBDMS and EVA;

[0059] Step S2, measuring the basic mechanical properties of different types of PBDMS-EVA materials in turn, and obtaining the performance of different types of PBDMS-EVA materials at different strain rates by changing the loading rate. The stress-strain curves of typical PBDMS-EVA materials at different compression rates are as follows: Figure 1As shown, the display dynamics software LS-DYNA is used to perform numerical simulation calculations on the bulletproof performance of the bulletproof vest structure using a numerical simulation model, wherein the numerical simulation model includes a projectile, a bulletproof layer, a buffer layer, and a backing material representing human tissue, and the material model parameters of the bulletproof layer and the buffer layer are consistent with the actual measured values, and the numerical simulation calculation results need to be verified and compared with the live ammunition test to ensure the high fidelity of the numerical simulation model, and the characterization values ​​of the bulletproof performance include the depth of the depression on the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material.

[0060] Figure 2 The numerical model shown in the upper part is composed of a 54 / 51-type pistol bullet with the same shooting speed as real ammunition, 43 layers of high molecular polyethylene fiber (bulletproof layer), 5 layers of PBDMS-EVA foam with different ratios (buffer layer) and clay material (backing). The typical bullet penetration process of the bulletproof vest is as follows Figure 2 As shown in the lower part. In the process of resisting bullet penetration, the bulletproof layer and the buffer layer will be significantly deformed at the impact point and squeeze the backing material. The pressure at the center of the backing material (impact point) changes with time as shown in Figure 3 shown.

[0061] Use numerical simulation models to perform numerical simulation calculations, including:

[0062] The bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are modeled using solid units and large deformation calculations are implemented based on the Lagrangian algorithm. The geometric dimensions, boundary conditions, and initial states of the bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are consistent with the actual working conditions, so as to be directly compared with the ballistic target experiment.

[0063] The selection of material simulation models should be determined by the material properties of each component. According to the characteristics of bullets, bulletproof materials, cushioning materials, and backing materials, the "Johnson-Cook" metal material constitutive model, the "CompositeMaterial" composite material constitutive model, the "LowDensityFoam" foam material constitutive model, and the "Cowper-Symonds" elastic-plastic material constitutive model are used respectively;

[0064] Based on the overall symmetry of the numerical simulation model and the material simulation model, numerical simulation is performed on the 1 / 4 area to ensure both calculation accuracy and calculation efficiency. After the numerical simulation is completed, the maximum deformation after stabilization is obtained through the displacement-time curve at the center point of the depression as the depression depth of the backing material, and the peak pressure and pressure distribution in the back area of ​​the buffer layer are directly obtained by setting the contact pressure sensing unit. The impact energy absorption of the bulletproof layer, the buffer layer and the backing material is directly output through the numerical simulation software to obtain the numerical simulation calculation results.

[0065] The numerical simulation results need to be verified and compared with the live-fire test, including:

[0066] The concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material measured in the live ammunition test are compared with the concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material in the numerical simulation calculation results to obtain the simulation accuracy of the numerical simulation model;

[0067] The high fidelity of the numerical simulation model is ensured by refining the mesh model of the bullet impact area in the numerical simulation model and setting a reasonable contact stiffness to improve the simulation accuracy of the numerical simulation model.

[0068] The three parameters of the depression depth of the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, buffer layer, and backing material are intuitive quantifications of the dynamic mechanical properties and protective performance of the bulletproof vest, which can be obtained in the target test through direct observation and measurement. The results obtained based on the numerical simulation method need to be directly compared with the test results to verify the simulation accuracy of the numerical model.

[0069] The entire process of a bullet hitting a bulletproof vest involves the failure performance and coupling of multiple materials. It is challenging to accurately restore the actual working conditions using numerical models, especially when materials with large differences in hardness come into contact with each other, computational problems such as mesh penetration or excessive unit distortion may occur. Therefore, the accuracy of numerical simulation calculations can be effectively improved by refining the mesh model in the impact area and setting a reasonable contact stiffness.

[0070] Step S3: Perform full permutation and combination of N layers of PBDMS-EVA materials with equal thickness to obtain M N buffer layer gradient combinations, determine the required number of numerical simulations according to the number of buffer layers N, to ensure the sufficiency and comprehensiveness of the amount of data used for simulation, select at least 10*N buffer layer gradient combinations of anti-ballistic performance numerical simulation results, and integrate the buffer layer gradient combinations and anti-ballistic performance numerical simulation results to form a sample data set for training the artificial neural network model;

[0071] The N layers of PBDMS-EVA materials with equal thickness are fully arranged and combined to obtain M N A buffer gradient combination, including:

[0072] For N layers of PBDMS-EVA materials with equal thickness, all permutations and combinations are performed according to M categories to obtain M N buffer layer gradient combination, M N The buffer layer gradient combination includes: {i j |j∈[1,N],i j ∈[1,M]}, where i j It is the type of PBDMS-EVA material of the jth layer in the buffer layer gradient combination.

[0073] The number of numerical simulations required is determined according to the number of buffer layers N, including:

[0074] The number of numerical simulations is set to 10 times the number of buffer layers N to ensure the sufficiency and comprehensiveness of the simulation data.

[0075] Different types of PBDMS-EVA materials have different material properties. The dynamic response characteristics of the buffer gradient layer formed by utilizing the material differences in different combinations are also different. If each combination is exhaustively studied, the required calculation time will be extremely long, which is an unrealistic approach. Therefore, it is a more scientific path to determine a certain number of working conditions for numerical simulation and use the powerful fitting analysis capabilities of neural networks to predict the performance of other combination modes. The number of numerical simulations depends on the number of input parameters, which is the number of buffer layers N.

[0076] Select at least 10*N buffer layer gradient combinations for the anti-ballistic performance numerical simulation results, including:

[0077] The fitting efficiency and fitting accuracy of artificial neural network are taken as two optimization objectives, and based on the two optimization objectives, the N The number of buffer gradient combinations selected in the buffer layer gradient combinations is converted into a multi-objective optimization problem for solution, and the genetic algorithm is used to solve the multi-objective optimization algorithm to determine the M N The number of buffer gradient combinations selected from the buffer layer gradient combinations is determined to be 10*N, so as to reduce the amount of numerical simulation tasks as much as possible under the condition of ensuring sufficient number of samples so as to take into account the fitting accuracy and fitting efficiency of the artificial neural network;

[0078] Using gradient change method in M N 10*N buffer layer gradient combinations are selected from the buffer layer gradient combinations to ensure the balance and representativeness of the sample selection.

[0079] The prediction ability of the neural network model is affected by the number of training samples. Using a large number of data sets can effectively improve the accuracy, but it will also consume a lot of time for simulation calculations. Setting the numerical simulation results of 10*N buffer layer gradient combinations as the training set of the neural network can reduce the amount of numerical simulation tasks as much as possible while ensuring that the number of samples is sufficient.

[0080] For the determination of the 10*N buffer layer gradient combinations, a specific gradient change mode, such as monotonically increasing, monotonically decreasing, concave, convex, etc., can be preferentially selected, so as to ensure the balance and representativeness of the sample selection, that is, to avoid the 10*N buffer layer gradient combinations being concentrated in one combination range. For example, assuming that the buffer layer is composed of 5 layers of PBDMS-EVA materials with equal thickness, and the types of PBDMS-EVA materials are 6 layers, the 6 layers obtained by full arrangement 5 Buffer layer gradient combinations, gradient change is to select 10*5 buffer layer gradient combinations, to achieve uniform sample selection, 10 in 111111-222222, 10 in 222222-333333, 10 in 333333-444444, 10 in 444444-555555, 10 in 555555-666666, to avoid sample concentration in one of the combination arrangement intervals, to avoid under-fitting caused by imbalance of model samples.

[0081] The 10*N buffer layer gradient combinations are numerically simulated using a numerical simulation model to obtain the numerical simulation results of the anti-ballistic performance.

[0082] Step S4, training the model based on the sample data set by the artificial neural network model to obtain a ballistic performance prediction model that characterizes the mapping relationship between the buffer layer gradient combination and the ballistic performance, and using the ballistic performance prediction model to predict the ballistic performance of each buffer layer gradient combination in turn, and based on the ballistic performance prediction results, N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations to design the bulletproof vest structure with the best protective performance.

[0083] The anti-ballistic performance prediction model that characterizes the relationship between the buffer layer gradient combination and the anti-ballistic performance mapping is obtained by training the model based on the sample data set through the artificial neural network model, including:

[0084] The buffer layer gradient combination in the sample data set is used as the input item of the artificial neural network model, the numerical simulation result of the ballistic performance in the sample data set is used as the output item of the artificial neural network model, and the artificial neural network model is used to perform model training based on the input item and the output item to obtain the ballistic performance prediction model;

[0085] The input parameter of the input layer of the artificial neural network model is set to N, and each input parameter is the type of PBDMS-EVA material in each layer of the input item. The structure of the artificial neural network model is as follows: Figure 4 As shown;

[0086] The model expression of the ballistic performance prediction model is:

[0087] [out1,out2,out3]=BP(in[i j ]);

[0088] Wherein, out1 represents the depression depth of the backing material in the anti-ballistic performance value, out2 represents the peak pressure of the back area of ​​the buffer layer in the anti-ballistic performance value, out3 represents the impact energy absorption of the bulletproof layer, buffer layer and backing material in the anti-ballistic performance value, in[i j ] is represented as a buffer layer gradient combination, and BP is represented as an artificial neural network model.

[0089] The anti-ballistic performance prediction model is used to predict the anti-ballistic performance of each buffer layer gradient combination in turn, including:

[0090] M N The buffer layer gradient combinations are sequentially input into the anti-ballistic performance prediction model to obtain M N The anti-ballistic performance value of each buffer layer gradient combination is constructed by the depression depth of the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bullet-proof layer, the buffer layer and the backing material in each buffer layer gradient combination.

[0091] Based on the ballistic performance prediction results in M N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations, including:

[0092] The evaluation priorities are the depth of the depression of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, buffer layer and backing material, and the M is evaluated according to the evaluation priority. N Screening of buffer layer gradient combinations;

[0093] According to the evaluation priority N Screening of buffer layer gradient combinations, including:

[0094] Step 1: M N The depth of depression of the backing material in the prediction results of the anti-ballistic performance of the buffer layer gradient combination is compared, among which,

[0095] If the depression depth of the backing material of the buffer layer gradient combination is less than a preset depression depth threshold, the buffer layer gradient combination is classified into the best combination set;

[0096] If the depression depth of the backing material of the buffer layer gradient combination is greater than or equal to a preset depression depth threshold, the buffer layer gradient combination is classified into a non-optimal combination set;

[0097] Step 2: Compare the peak pressures of the back area of ​​the buffer layer in the prediction results of the buffer layer gradient combination anti-ballistic performance in the best combination set, where:

[0098] If the peak pressure of the buffer layer back region of the buffer layer gradient combination is less than a preset threshold of the peak pressure, the buffer layer gradient combination is retained in the optimal combination set;

[0099] If the peak pressure of the buffer layer back region of the buffer layer gradient combination is greater than or equal to the peak pressure preset threshold, the buffer layer gradient combination is removed from the non-optimal combination set;

[0100] Step 3: Compare the impact energy absorption of the bulletproof layer, the buffer layer and the backing material in the buffer layer gradient combination anti-ballistic performance prediction results in the best combination set, where:

[0101] If the impact energy absorption of the bulletproof layer and the buffer layer of the buffer layer gradient combination is greater than or equal to the preset absorption threshold, and the impact energy absorption of the backing material is less than the preset absorption threshold, the buffer layer gradient combination is retained in the optimal combination set;

[0102] If the impact energy absorption of the bulletproof layer or the buffer layer of the buffer layer gradient combination is less than the preset absorption threshold, or the impact energy absorption of the backing material is greater than or equal to the preset absorption threshold, the buffer layer gradient combination is removed from the non-optimal combination set;

[0103] Step 4: All buffer layer gradient combinations in the best combination set are taken as the buffer layer gradient combinations that best characterize the anti-ballistic performance.

[0104] The dent depth is the main indicator for evaluating protection. The determination of the optimal buffer layer gradient combination should first focus on the size of the dent deformation and select the combination with a smaller dent depth. For buffer layers with similar dent depths, screening is performed based on the peak pressure to obtain a gradient buffer layer that causes less pressure on the human body. Energy absorption is used as an auxiliary indicator to evaluate the degree of loss of stress waves in each layer, providing a quantitative representation of the overall response of the bulletproof vest.

[0105] Based on the trained artificial neural network model, the bulletproof and protective performance of bulletproof vests under different buffer layer gradient combinations can be predicted and evaluated quickly, accurately and comprehensively. By inputting the material combination of the buffer layer and calculating the model, the depression depth of the backing material under the impact of the bullet can be obtained first and whether the bulletproof vest under this design meets the bulletproof requirements can be judged. The model can also output the peak pressure on the backing material to evaluate whether the wearer will suffer excessive blunt injuries and lose the ability to continue fighting. In addition, the model can also calculate the energy absorbed by the buffer layer during the bullet impact and use it as a quantifiable indicator to evaluate the protective performance of the gradient buffer layer. Through the artificial neural network model established by the invention, various buffer layer combinations are globally predicted, and the optimal gradient combination form is obtained according to the size of the output indicator parameters.

[0106] The present invention uses artificial neural networks to replace the traditional methods dominated by live-fire tests or numerical simulations. It only uses a certain number of numerical simulation results for training, and uses the neural network's strong nonlinear analysis ability, learning ability, and regression ability to predict the parameters of interest. It only needs to input the combination of PBDMS-EVA, and the trained neural network can quickly output the protective performance of the buffer layer, greatly reducing the calculation cost.

[0107] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network, characterized in that: The following steps are involved: Step S1, setting a buffer layer composed of N layers of PBDMS-EVA materials of equal thickness and a bulletproof layer composed of X layers of fiber materials for the bulletproof vest structure, wherein the types of PBDMS-EVA materials are M types, and the material properties of each type of PBDMS-EVA materials are determined by the ratio of PBDMS and EVA; Step S2, sequentially measuring the basic mechanical properties of different types of PBDMS-EVA materials, obtaining the performance of different types of PBDMS-EVA materials under different strain rates by changing the loading rate, and using the display power software LS-DYNA to perform numerical simulation calculation on the bulletproof performance of the bulletproof vest structure using a numerical simulation model, wherein the numerical simulation model includes a projectile, a bulletproof layer, a buffer layer and a backing material characterizing human tissue, and the material model parameters of the bulletproof layer and the buffer layer are consistent with the actual measured values, and the numerical simulation calculation results need to be verified and compared with the live ammunition test to ensure the high fidelity of the numerical simulation model, and the characterization values ​​of the bulletproof performance include the depression depth on the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer and the backing material; Step S3: Perform full permutation and combination of N layers of PBDMS-EVA materials with equal thickness to obtain M N buffer layer gradient combinations, determine the required number of numerical simulations according to the number of buffer layers N, to ensure the sufficiency and comprehensiveness of the amount of data used for simulation, select at least 10*N buffer layer gradient combinations of anti-ballistic performance numerical simulation results, and integrate the buffer layer gradient combinations and anti-ballistic performance numerical simulation results to form a sample data set for training the artificial neural network model; Step S4, training the model based on the sample data set by the artificial neural network model to obtain a ballistic performance prediction model that characterizes the mapping relationship between the buffer layer gradient combination and the ballistic performance, and using the ballistic performance prediction model to predict the ballistic performance of each buffer layer gradient combination in turn, and based on the ballistic performance prediction results, N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations to design the bulletproof vest structure with the best protective performance.

2. The method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 1, characterized in that: The method of performing numerical simulation calculation using a numerical simulation model includes: The bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are modeled using solid units and large deformation calculations are implemented based on the Lagrangian algorithm. The geometric dimensions, boundary conditions, and initial states of the bullet, bulletproof layer, buffer layer, and backing material in the numerical simulation model are consistent with the actual working conditions, so as to be directly compared with the ballistic target experiment. The selection of material simulation models should be determined by the material properties of each component. According to the characteristics of bullets, bulletproof materials, cushioning materials, and backing materials, the "Johnson-Cook" metal material constitutive model, the "CompositeMaterial" composite material constitutive model, the "LowDensityFoam" foam material constitutive model, and the "Cowper-Symonds" elastic-plastic material constitutive model are used respectively; Based on the overall symmetry of the numerical simulation model and the material simulation model, numerical simulation is performed on the 1 / 4 area to ensure both calculation accuracy and calculation efficiency. After the numerical simulation is completed, the maximum deformation after stabilization is obtained through the displacement-time curve at the center point of the depression as the depression depth of the backing material, and the peak pressure and pressure distribution in the back area of ​​the buffer layer are directly obtained by setting the contact pressure sensing unit. The impact energy absorption of the bulletproof layer, the buffer layer and the backing material is directly output through the numerical simulation software to obtain the numerical simulation calculation results.

3. The method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 2, characterized in that: The numerical simulation results need to be verified and compared with the live-fire test, including: The concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material measured in the live ammunition test are compared with the concave depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, the buffer layer, and the backing material in the numerical simulation calculation results to obtain the simulation accuracy of the numerical simulation model; The high fidelity of the numerical simulation model is ensured by refining the mesh model of the bullet impact area in the numerical simulation model and setting a reasonable contact stiffness to improve the simulation accuracy of the numerical simulation model.

4. The method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 3, characterized in that: The method of performing full permutation and combination of N layers of PBDMS-EVA materials with equal thickness to obtain M N A buffer gradient combination, including: For N layers of PBDMS-EVA materials with equal thickness, all permutations and combinations are performed according to M categories to obtain M N The buffer layer gradient combination, the M N The buffer layer gradient combination includes: {i j |j∈[1,N],i j ∈[1,M]}, where i j It is the type of PBDMS-EVA material of the jth layer in the buffer layer gradient combination.

5. The method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 4, characterized in that: Determining the required number of numerical simulations according to the number of buffer layers N includes: The number of numerical simulations is set to 10 times the number of buffer layers N to ensure the sufficiency and comprehensiveness of the simulation data.

6. The method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 5, characterized in that: The anti-ballistic performance numerical simulation results of selecting at least 10*N buffer layer gradient combinations include: The fitting efficiency and fitting accuracy of artificial neural network are taken as two optimization objectives, and based on the two optimization objectives, the N The number of buffer gradient combinations selected in the buffer layer gradient combinations is determined by converting the number of buffer gradient combinations into a multi-objective optimization problem for solution, and a genetic algorithm is used to solve the multi-objective optimization problem to determine the number of buffer gradient combinations selected in M N The number of buffer gradient combinations selected from the buffer layer gradient combinations is determined to be 10*N, so as to reduce the amount of numerical simulation tasks as much as possible under the condition of ensuring sufficient number of samples so as to take into account the fitting accuracy and fitting efficiency of the artificial neural network; Using gradient change method in M N 10*N buffer layer gradient combinations are selected from the buffer layer gradient combinations to ensure the balance and representativeness of the sample selection.

7. A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 6, characterized in that: The 10*N buffer layer gradient combinations are numerically simulated using a numerical simulation model to obtain the anti-ballistic performance numerical simulation result.

8. A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 7, characterized in that: The anti-ballistic performance prediction model characterizing the mapping relationship between the buffer layer gradient combination and the anti-ballistic performance is obtained by performing model training based on the sample data set using the artificial neural network model, including: The buffer layer gradient combination in the sample data set is used as an input item of the artificial neural network model, the ballistic performance numerical simulation result in the sample data set is used as an output item of the artificial neural network model, and the artificial neural network model is used to perform model training based on the input item and the output item to obtain the ballistic performance prediction model; The input parameter of the input layer of the artificial neural network model is set to N, and each input parameter is the type of PBDMS-EVA material of each layer in the input item; The model expression of the ballistic performance prediction model is: [out1,out2,out3]=BP(in[i j ]); Wherein, out1 represents the depression depth of the backing material in the anti-ballistic performance value, out2 represents the peak pressure of the back area of ​​the buffer layer in the anti-ballistic performance value, out3 represents the impact energy absorption of the bulletproof layer, buffer layer and backing material in the anti-ballistic performance value, in[i j ] is represented as a buffer layer gradient combination, and BP is represented as an artificial neural network model.

9. A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 8, characterized in that: The method of using the anti-ballistic performance prediction model to predict the anti-ballistic performance of each buffer layer gradient combination in turn includes: M N The buffer layer gradient combinations are sequentially input into the anti-ballistic performance prediction model to obtain M N The anti-ballistic performance value of each buffer layer gradient combination is constructed by the depression depth of the backing material, the peak pressure in the back area of ​​the buffer layer, and the impact energy absorption of the bullet-proof layer, the buffer layer and the backing material in each buffer layer gradient combination.

10. A method for designing a gradient buffer layer of a bulletproof vest structure based on a neural network according to claim 9, characterized in that: The ballistic performance prediction results are based on the N The buffer layer gradient combination with the best anti-ballistic performance is selected from the buffer layer gradient combinations, including: The evaluation priorities are the depression depth of the backing material, the peak pressure of the back area of ​​the buffer layer, and the impact energy absorption of the bulletproof layer, buffer layer and backing material, and the M is evaluated according to the evaluation priority. N Screening of buffer layer gradient combinations; The evaluation is performed according to the priority N Screening of buffer layer gradient combinations, including: Step 1: M N The depth of depression of the backing material in the prediction results of the anti-ballistic performance of the buffer layer gradient combination is compared, among which, If the depression depth of the backing material of the buffer layer gradient combination is less than a preset depression depth threshold, the buffer layer gradient combination is classified into the best combination set; If the depression depth of the backing material of the buffer layer gradient combination is greater than or equal to a preset depression depth threshold, the buffer layer gradient combination is classified into a non-optimal combination set; Step 2: Compare the peak pressures of the back area of ​​the buffer layer in the prediction results of the buffer layer gradient combination anti-ballistic performance in the best combination set, where: If the peak pressure of the buffer layer back region of the buffer layer gradient combination is less than a preset peak pressure threshold, retaining the buffer layer gradient combination in the best combination set; If the peak pressure of the buffer layer back region of the buffer layer gradient combination is greater than or equal to a preset peak pressure threshold, the buffer layer gradient combination is removed from the non-optimal combination set; Step 3: Compare the impact energy absorption of the bulletproof layer, the buffer layer and the backing material in the buffer layer gradient combination anti-ballistic performance prediction results in the best combination set, where: If the impact energy absorption of the bulletproof layer and the buffer layer of the buffer layer gradient combination is greater than or equal to the preset absorption threshold, and the impact energy absorption of the backing material is less than the preset absorption threshold, the buffer layer gradient combination is retained in the optimal combination set; If the impact energy absorption of the bulletproof layer or the buffer layer of the buffer layer gradient combination is less than the preset absorption threshold, or the impact energy absorption of the backing material is greater than or equal to the preset absorption threshold, the buffer layer gradient combination is removed from the non-optimal combination set; Step 4: All buffer layer gradient combinations in the best combination set are taken as the buffer layer gradient combinations that best characterize the anti-ballistic performance.

Images