Prediction method for yield load and damage degree of explosive welding composite material

Abstract

The invention relates to a method for predicting the yield load and the damage degree of an explosive welding composite material, belongs to the technical field of composite material damage prediction, and solves the problems of high cost, poor safety, low accuracy and low efficiency of explosive welding composite material damage research in the prior art. Obtaining a damage image of the to-be-detected composite material, and inputting the damage image into the first image recognition model to obtain the number and size of holes and cracks in the image; inputting the damaged image into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface; and acquiring the cross sectional area change rate of the to-be-tested composite material, and inputting the cross sectional area change rate, the number of holes, the minimum value and the maximum value of the holes, the number of cracks, the minimum value and the maximum value of the cracks, the wavelength and the amplitude into the first load damage prediction model to obtain the yield load and the damage grade of the to-be-tested composite material. The method for predicting the damage degree of the explosive welding composite material is low in cost, high in safety, accurate and rapid.

Description

Technical Field

[0001] The invention relates to the technical field of composite material damage prediction, and in particular to a method for predicting the yield load and damage degree of an explosively welded composite material. Background Art

[0002] Explosive welding technology is a process that uses the energy generated by the explosion of explosives to achieve solid-state connection of the same or different materials. It is widely used in the preparation of composite materials. In the fields of aerospace, automobile manufacturing, petrochemicals, etc., many key structural parts use explosive welding composite materials. If these materials have undetected damage, catastrophic accidents may occur during service. Studying material damage can assess the safety of the structure in advance, discover potential problems in time and take measures to repair or replace them, and avoid major accidents caused by sudden material failure. For example, if the wing structure of an aircraft uses damaged explosive welding composite materials, it may break during flight due to the inability to withstand the huge aerodynamic force, endangering flight safety. By studying and detecting material damage, the safety of the wing structure can be ensured and the smooth progress of the flight mission can be guaranteed.

[0003] In the prior art, the main methods for studying the damage of explosive welding composite materials are: (1) Experimental research methods, such as tensile tests, nanoindentation tests, etc., which analyze the damage characteristics through actual welding and testing. Although they can intuitively reflect the mechanical behavior and damage of composite materials, the experiments require a large amount of materials and equipment, and involve the use of explosives, which is costly. The stress state of the samples during the test is relatively simple, which is different from the complex stress state that composite materials may be subjected to in actual applications. It may not be able to fully reflect its mechanical properties and damage characteristics. There are safety risks in the explosive welding process, and strict safety measures are required. (2) Microstructure analysis methods, such as scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray diffraction (XRD), etc. This method can analyze the microstructure and damage characteristics of explosive welding composite materials. Although it can accurately observe the interface microstructure, such as grain refinement, adiabatic shear band, etc., to provide theoretical support for the damage mechanism, it requires high-precision microanalysis equipment, which is costly, and the amount of microanalysis data is large, requiring professional analysis technology. (3) Numerical simulation methods use finite element analysis, smoothed particle hydrodynamics (SPH) and other technologies to simulate the damage mechanism during explosive welding. Although computer simulation can reduce the number of experiments and significantly reduce R&D costs, it can simulate complex phenomena such as stress, strain, and temperature field during welding and reveal details that are difficult to observe in experiments. However, the simulation results depend on the accuracy of the model. Complex physical phenomena may be difficult to fully simulate, requiring high-performance computing resources and high technical requirements for operators. The simulation results need to be verified through experiments to ensure their reliability.

[0004] Therefore, there is a need to provide a method for predicting the damage degree of explosive welded composite materials with low cost, high safety, and accuracy and speed. Summary of the Invention

[0005] In view of the above analysis, embodiments of the present invention aim to provide a method for predicting the yield load and damage degree of explosive welded composite materials to solve the problems of high cost, poor safety, low accuracy, and low efficiency in the research on the damage of explosive welded composite materials in the prior art.

[0006] Embodiments of the present invention provide a method for predicting the yield load and damage degree of explosive welded composite materials, which is characterized by including: Obtain the damage image of the composite material to be measured, input the damage image into the first image recognition model to obtain the number and size of holes and cracks in the composite material to be measured, and count the minimum and maximum values of the holes and cracks; Input the damage image into the second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be measured; Obtain the cross-sectional area change rate of the composite material to be measured, and input the cross-sectional area change rate, the number of holes and its minimum and maximum values, the number of cracks and its minimum and maximum values, the wavelength, and the amplitude into the first load damage prediction model to obtain the yield load and damage level of the composite material to be measured.

[0007] Based on a further improvement of the above method, the damage image at least includes , , , X-ray scan images; the first image recognition model is a 3D ResNet network model, the second image recognition model is a model combined by a U-NET model and a regression head, and the first load damage prediction model is a machine learning model.

[0008] Based on a further improvement of the above method, dilated convolution is introduced into each residual block in the 3D ResNet network model, and an SE attention module is introduced after each residual block.

[0009] Based on a further improvement of the above method, the first load damage prediction model is a support vector machine SVM model, which is implemented by hard parameter sharing, and the output parameters are the yield load and damage level respectively; The SVM model is trained based on a first sample data set; wherein, each data item in the first sample data set includes: cross-sectional area change rate, number of holes and its minimum and maximum values, number of cracks and its minimum and maximum values, wavelength, amplitude, yield load, and damage level.

[0010] Based on the further improvement of the above method, the first sample data set is constructed in the following way: A variety of composite materials are prepared based on explosive welding parameters. For each composite material, the corresponding first sample data is obtained in the following manner to form a first sample data set: Three in-situ tensile specimens are cut from the composite material by machine turning; one in-situ tensile specimen is selected to perform an in-situ tensile test based on an in-situ tensile test platform to obtain a load-displacement curve corresponding to the specimen, and then obtain a yield load; one in-situ tensile specimen is selected to obtain multiple X-ray scanning images based on the in-situ tensile test platform, and the images are input into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be tested; one in-situ tensile specimen is selected to perform an in-situ tensile test based on the in-situ tensile test platform to obtain the cross-sectional area change rate, the number and size distribution of holes, and the number and size distribution of cracks of the specimen under different tensile loads, and the damage level of the specimen is marked based on the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, and the wavelength and amplitude of the specimen under each tensile load, and the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, as well as the wavelength, amplitude, yield load, and damage level of the specimen under each tensile load are taken as a first sample data.

[0011] Based on the further improvement of the above method, the composite material is prepared based on the explosive welding parameters, including: The explosion welding parameters are input into a second prediction model to determine whether the output wavelength and amplitude meet the predetermined requirements. If yes, a composite material is prepared based on the explosion welding parameters. If no, a genetic algorithm is used to optimize the input parameters, and the optimized parameters are used as the explosion welding parameters.

[0012] Based on the further improvement of the above method, the second prediction model is trained based on the second sample data set; wherein each sample data in the second sample data set includes: control parameters in each explosive welding process and wavelength and amplitude corresponding to the control parameters, and the control parameters include: collision speed, collision angle, thickness of explosive, mass ratio of explosive per unit area to mass ratio of flyer plate, spacing between flyer plate and substrate, and thickness of flyer plate; The wavelength and amplitude in each piece of sample data are obtained in the following manner: a flying plate material and a substrate material are selected based on business requirements, and the flying plate and the substrate are composited based on the control parameters to form a specific explosively welded composite plate, an observation sample is prepared based on the composite plate, and the observation sample is observed under an optical microscope to obtain the wavelength and amplitude of the wavy interface of the explosive composite interface of the composite plate.

[0013] Based on the further improvement of the above method, the composite material to be tested is obtained by using a TA1 material plate as a flying plate and a Q235 material plate as a base plate through explosive welding technology; The method of preparing an observation sample based on the composite material comprises: An observation sample section is cut from the composite material, and the heterogeneous material laminate surface is mechanically polished. After the observation sample section is immersed in the metallographic etching agent for a first preset time, the surface is rinsed with clean water and blown dry, and then a cotton ball is used to pick up a volume concentration of The nitric acid alcohol solution is wiped on the surface of the sample, and after corrosion for a second preset time, it is cleaned and dried to finally obtain the observation sample.

[0014] Based on the further improvement of the above method, the in-situ tensile test is performed based on the in-situ tensile test platform to obtain the number and size distribution of holes and the number and size distribution of cracks of the sample under different tensile loads, including: Set the initial tensile load and tensile loading rate; The in-situ tensile specimen is placed in a mechanical fixture and an initial tensile load is applied. When the tensile load increases by 1 N, the in-situ tensile test platform rotates the fixture to take sample slices of the in-situ tensile specimen at different angles. The sample slices are input into the first image recognition model to obtain the number and size of holes and cracks corresponding to the sample slices, and the minimum and maximum values ​​of holes and cracks of the sample under the current tensile load are counted.

[0015] Based on the further improvement of the above method, the predetermined requirements include: the wavelength should be between 660-690 The amplitude should be between 60-90 within the scope; The optimization of the input parameters using a genetic algorithm comprises: B1: Initialize the population, set the population size N, preset the number of iterations T, and each individual in the population includes: collision speed, collision angle, explosive thickness, the ratio of the mass of the explosive per unit area to the mass of the flyer plate, the distance between the flyer plate and the base plate, and the thickness of the flyer plate; The collision velocity, collision angle, the distance between the flyer plate and the base plate, and the thickness of the flyer plate are binary coded, and the thickness of the explosive and the ratio of the mass of the explosive per unit area to the mass of the flyer plate are Gray coded; B2: Calculate the fitness value of each individual according to the fitness function, where the fitness function is: , in, is the minimum wavelength in the predetermined requirement, is the maximum value of the wavelength in the predetermined requirement, The wavelength obtained by inputting the second prediction model into the i-th individual, is the minimum value of the amplitude in the predetermined demand, The maximum value of the amplitude in the predetermined demand, The amplitude obtained by inputting the second prediction model into the i-th individual, is a constant, i=1,2,3,..,N; B3: Selection operation, based on the tournament algorithm, select individuals with higher fitness from the current population as parents to generate the next generation; B4: Crossover operation, for binary coded parameter groups, a single-point crossover method is used, and for Gray coded parameter groups, a multi-point crossover method is used; B5: mutation operation, select the one with the highest fitness from the population The individuals are searched locally. For each selected individual, several neighborhood solutions are generated near it. A better solution than the current solution is found in the neighborhood solutions. If a better solution is found, the current solution is replaced and the individuals after the local search are put back into the population. B6: Generate a new population, merge the parent and offspring individuals, select a preset number of individuals as the new generation population according to the fitness value, and determine whether the iteration stop condition is met. If so, the best individual in the current population is used as the optimized operating parameter.

[0016] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. The present invention provides a method for predicting the yield load and damage degree of explosively welded composite materials. The damage research of explosively welded composite materials is combined with machine learning technology. By mining the potential laws in material damage data, the relationship between damage patterns and characteristics that are difficult to detect by traditional methods is discovered, and the damage of composite materials is more comprehensively understood. Compared with traditional experimental research methods, microstructure analysis methods, and numerical simulation methods, the scheme proposed by the present invention can quickly and accurately obtain damage results and the yield load corresponding to the current material state, and has lower cost and higher safety.

[0017] 2. The present invention provides a method for predicting the yield load and damage degree of explosively welded composite materials. When preparing composite materials, if the current explosive welding parameters do not meet the predetermined wavelength and amplitude requirements, a genetic algorithm is used to optimize the current explosive welding parameters to obtain the predetermined wavelength and amplitude requirements of the explosive welding parameters. When preparing composite materials, the present invention reduces the difficulty of preparation and no longer requires technical personnel to make adjustments based on experience. Only relying on the provided genetic algorithm can quickly obtain the explosive welding parameters that are compatible with the current predetermined requirements. In the mutation operation of the genetic algorithm, combined with neighborhood solution optimization, high-quality solutions can be quickly found through efficient local search, which can not only improve the search efficiency of the algorithm and the quality of the solution, but also enhance the robustness and global search capabilities of the algorithm, thereby showing better performance in solving complex optimization problems.

[0018] 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

[0019] 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; Figure 1 This is an example diagram of a method for predicting the yield load and damage degree of an explosively welded composite material in an embodiment of the present invention; Figure 2 2 is an example diagram of the damage level standard in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] 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.

[0021] Hard Parameter Sharing is a common strategy in multi-task learning. The core idea is that multiple tasks share the main parameters of the model, and only retain independent parameters for each task in the output layer. This method is widely used in deep learning and has the advantages of reducing model complexity, accelerating the training process, and reducing the risk of overfitting.

[0022] A specific embodiment of the present invention discloses a method for predicting the yield load and damage degree of explosively welded composite materials, such as Figure 1 As shown, including: S1: Acquire a damage image of the composite material to be tested, and input the damage image into a first image recognition model to obtain the number and size of holes and cracks in the composite material to be tested, and count the minimum and maximum values ​​of the holes and cracks.

[0023] For example, the composite material to be tested may be a whole steel-titanium composite plate obtained by using a TA1 material plate as a flying plate and a Q235 material plate as a base plate through explosion welding technology.

[0024] The damage image in the present invention refers to the image obtained by transmission scanning of the X-ray imaging device. The present invention does not specifically limit the number of damage images, but at least includes , , , It can be understood that the more damage images input, the more accurate the prediction result.

[0025] The first image recognition model is a 3D convolutional neural network, which can directly process three-dimensional slice data, make full use of the spatial correlation between slices in multiple directions, better identify the three-dimensional structural characteristics of holes and cracks, and improve the accuracy of prediction. Common 3D convolutional neural networks include 3D ResNet, 3D U-Net Lite, 3D DenseNet, 3D U-Net, VoxNet, etc. The present invention does not specifically limit the model used. The technician can select the network model according to the needs to realize the identification of holes and cracks. Exemplarily, the present invention adopts a 3D ResNet network model, introduces a hole convolution in each residual block, and introduces an SE attention module after each residual block to improve the prediction accuracy and training efficiency of the 3D ResNet network.

[0026] The training process of the first image recognition model is as follows: Obtain the first training sample set for training the first image recognition model. The samples in the first training sample set are each image slice obtained by the transmission scan of the X-ray imaging device, and the first training sample set needs to ensure that holes and cracks of different sizes and shapes are covered. Each slice is manually annotated, and the annotation content includes the location, quantity, and size information of the holes and cracks. You can use professional image annotation tools such as OpenCV, Scikit-image, LabelMe, CVAT, ITK-SNAP, etc. to mark the boundaries or key points of each hole and crack, and record the corresponding quantity and size data. And you can perform normalization, denoising and other operations as needed to improve the accuracy of recognition.

[0027] The labeled data set is divided into training set, validation set and test set. Based on the training set, validation set and test set, the first image recognition model that can be used to identify the number and size of holes and cracks is finally obtained. Generally speaking, the training set is used to train the model, the validation set is used to adjust the hyperparameters of the model and monitor the training process to prevent overfitting, and the test set is used to evaluate the final performance of the model. Divide according to a certain ratio, such as 7:2:1 or 8:1:1.

[0028] The present invention does not impose any specific restrictions on the network structure design, loss function, optimizer selection, learning rate adjustment and other operations of the first image recognition model, so long as the recognition of the number and size of holes and cracks can be achieved.

[0029] A plurality of damage images of the composite material to be tested are obtained, and the plurality of damage images are input into a first image recognition model to obtain the number and size of holes and cracks of the composite material to be tested, and the minimum and maximum values ​​of the holes and cracks are counted based on the current prediction results.

[0030] S2: Inputting the damage image into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be tested.

[0031] The second image recognition model is a combination of the U-NET model and the regression head, which is used to identify the wavelength and amplitude of the wavy composite interface in the damage image. The input damage image is X-ray scan image.

[0032] The training process of the second image recognition model is as follows: The second training sample set for training the second image recognition model is obtained. The samples in the second training sample set are each damaged image with a wavy composite interface, and the wavelength and amplitude information are manually annotated. The images are preprocessed, including image resizing, normalizing pixel values, grayscale, and other operations. Resizing the image can unify all images to a fixed size for easy model processing; normalizing pixel values ​​can map pixel values ​​to or interval, speeding up the convergence of the model; grayscale processing can convert color images into grayscale images, reduce data dimensions, and also highlight the texture features of the image, which is conducive to the recognition of wavy interfaces.

[0033] The labeled data set is divided into a training set, a validation set, and a test set to realize the training of the second image recognition model, and finally a network that can accurately output the wavelength and amplitude of the wavy composite interface is obtained.

[0034] The present invention does not impose any specific restrictions on the network structure design, loss function, optimizer selection, learning rate adjustment and other operations of the second image recognition model, so as to accurately output the wavelength and amplitude of the wavy composite interface.

[0035] S3: Obtain the cross-sectional area change rate of the composite material to be tested, and input the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, and the amplitude into a first load damage prediction model to obtain the yield load and damage grade of the composite material to be tested.

[0036] The first load damage prediction model is a machine learning model, such as a neural network model, a decision tree model, a support vector machine model, a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, a generative adversarial network (GAN) model, etc.

[0037] Exemplarily, the present invention adopts a support vector machine SVM model, and uses a hard parameter method to adaptively modify its structure, limiting the input parameters to the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength and amplitude, and the output parameters are the yield load and the damage level. The SVM model is trained based on the first sample data set; wherein each sample data in the first sample data set includes: the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, the amplitude, the yield load and the damage level.

[0038] Among them, the damage level is obtained based on the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, and the amplitude. Different damage level standards can be quantified for different composite materials. For example, the composite material to be tested selected by the present invention is constructed according to predetermined requirements. The damage level standard of the whole steel-titanium composite plate composed of TA1 material plate and Q235 material plate under the specified wavelength requirement and the specified amplitude requirement is as follows: Figure 2 shown.

[0039] The first sample data set is constructed in the following way: A variety of composite materials are prepared based on explosive welding parameters. For each composite material, the corresponding first sample data is obtained in the following manner to form a first sample data set: At least three in-situ tensile specimens are cut from the composite material by lathe turning; An in-situ tensile specimen is selected to conduct an in-situ tensile test based on the in-situ tensile test platform to obtain the load-displacement curve corresponding to the specimen, and then the yield load is obtained; An in-situ tensile specimen is selected to obtain a plurality of X-ray scanning images based on the in-situ tensile test platform, and the images are input into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be tested; An in-situ tensile specimen is selected to carry out an in-situ tensile test based on an in-situ tensile test platform to obtain the cross-sectional area change rate, the number and size distribution of holes, and the number and size distribution of cracks of the specimen under different tensile loads; the damage level of the specimen is marked based on the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, and the amplitude of the specimen under each tensile load; the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, the amplitude, the yield load, and the damage level of the specimen under each tensile load are taken as a first sample data.

[0040] For example, when an in-situ tensile specimen is cut from a composite material, in order to prevent special situations such as accidental damage to the material during the test, several more in-situ tensile specimens may be cut as backup, thereby ensuring that the first sample data can be generated.

[0041] The cross-sectional area change rate S refers to the ratio of the initial cross-sectional area A1 of the in-situ tensile specimen cut from the composite material to the change in the cross-sectional area A2 under different tensile loads during the in-situ tensile test, specifically: .

[0042] For example, the composite material can be a whole steel-titanium composite plate obtained by explosive welding technology, which is composed of a TA1 material plate with a length of 300 mm, a width of 300 mm, and a height of 10 mm as a flyer plate and a Q235 material plate with a length of 300 mm, a width of 300 mm, and a height of 30 mm as a base plate. The in-situ tensile specimen is a sample section with a length of 20 mm, a width of 20 mm, and a height of 10 mm, wherein the thickness of the TA1 and Q235 materials is 5 mm respectively.

[0043] The in-situ tensile test platform consists of an X-ray source, a thermal-mechanical coupling in-situ loading device, a detector, and a three-dimensional reconstruction system. The X-ray microscope host model is XPloreVista2000 4D with a spatial resolution of , the tube voltage is The thermal-mechanical coupling in-situ loading device consists of two fixture structures, one end of which is fixed and the other end of which applies the load.

[0044] The in-situ tensile test is performed based on the in-situ tensile test platform to obtain the number and size distribution of holes and the number and size distribution of cracks of the sample under different tensile loads, including: Set the initial tensile load and tensile loading rate; The in-situ tensile specimen is placed in a mechanical fixture and an initial tensile load is applied. When the tensile load increases by 1 N, the in-situ tensile test platform rotates the fixture to take sample slices of the in-situ tensile specimen at different angles. The sample slices are input into the first image recognition model to obtain the number and size of holes and cracks corresponding to the sample slices, and the minimum and maximum values ​​of holes and cracks of the sample under the current tensile load are counted.

[0045] In order to capture the expansion of cracks in the material interface in real time, the tensile loading rate is constantly controlled at , the tensile specimen is placed in a mechanical fixture and a tensile load is applied. When a predetermined load threshold is reached, the loading is stopped, and the fixture is rotated to photograph the specimen from different angles, thereby obtaining slices at different angles. It can be understood that the technician can set the tensile loading rate based on experience, and the predetermined load threshold is the yield load obtained by another in-situ tensile specimen.

[0046] Preparation of composite materials based on preset explosive welding parameters, including: The explosion welding parameters are input into a second prediction model to determine whether the output wavelength and amplitude meet the predetermined requirements. If yes, a composite material is prepared based on the explosion welding parameters. If no, a genetic algorithm is used to optimize the input parameters, and the optimized parameters are used as the explosion welding parameters.

[0047] The predetermined requirements include: the wavelength should be between 660-690 The amplitude should be between 60-90 The predetermined requirements can be set by technicians according to actual needs.

[0048] The second prediction model is trained based on the second sample data set; wherein each sample data in the second sample data set includes: control parameters in each explosive welding process and wavelengths and amplitudes corresponding to the control parameters, and the control parameters include: collision speed, collision angle, explosive thickness, mass ratio of explosive per unit area to mass ratio of flyer plate, spacing between flyer plate and substrate, and flyer plate thickness; The wavelength and amplitude in each piece of sample data are obtained in the following manner: a flying plate material and a substrate material are selected based on business requirements, and the flying plate and the substrate are composited based on the control parameters to form a specific explosively welded composite plate, an observation sample is prepared based on the composite plate, and the observation sample is observed under an optical microscope to obtain the wavelength and amplitude of the wavy interface of the explosive composite interface of the composite plate.

[0049] The step of preparing an observation sample based on the composite material comprises: An observation sample section is cut from the composite material, and the heterogeneous material laminate surface is mechanically polished. After the observation sample section is immersed in the metallographic etching agent for a first preset time, the surface is rinsed with clean water and blown dry, and then a cotton ball is used to pick up a volume concentration of The nitric acid alcohol solution is wiped on the surface of the sample, and the sample is cleaned and dried after being corroded for a second preset time, and finally the observation sample is obtained. Exemplarily, the first preset time is 15s and the second preset time is 10s.

[0050] The optimization of the input parameters using a genetic algorithm comprises: B1: Initialize the population, set the population size N, preset the number of iterations T, and each individual in the population includes: collision speed, collision angle, explosive thickness, the ratio of the mass of the explosive per unit area to the mass of the flyer plate, the distance between the flyer plate and the base plate, and the thickness of the flyer plate; The collision velocity, collision angle, the distance between the flyer plate and the base plate, and the thickness of the flyer plate are binary coded, and the thickness of the explosive and the ratio of the mass of the explosive per unit area to the mass of the flyer plate are Gray coded; B2: Calculate the fitness value of each individual according to the fitness function, where the fitness function is: , in, is the minimum wavelength in the predetermined requirement, is the maximum value of the wavelength in the predetermined requirement, The wavelength obtained by inputting the second prediction model into the i-th individual, is the minimum value of the amplitude in the predetermined demand, The maximum value of the amplitude in the predetermined demand, The amplitude obtained by inputting the second prediction model into the i-th individual, is a constant, i=1,2,3,..,N; B3: Selection operation, based on the tournament algorithm, select individuals with higher fitness from the current population as parents to generate the next generation; B4: Crossover operation, for binary coded parameter groups, a single-point crossover method is used, and for Gray coded parameter groups, a multi-point crossover method is used; B5: mutation operation, select the one with the highest fitness from the population The individuals are searched locally. For each selected individual, several neighborhood solutions are generated near it. A better solution than the current solution is found in the neighborhood solutions. If a better solution is found, the current solution is replaced and the individuals after the local search are put back into the population. Preferably, the neighborhood solution is generated in the following manner: , in, is the newly generated neighborhood solution, For the current individual, , Indicates rounding up, t is the current iteration number, and T is the preset iteration number; B6: Generate a new population, merge the parent and offspring individuals, select a preset number of individuals as the new generation population according to the fitness value, and determine whether the iteration stop condition is met. If so, the best individual in the current population is used as the optimized operating parameter.

[0051] It can be understood that the implementation of the initialization operation, crossover operation, mutation operation, etc. in the genetic algorithm is common knowledge in the art, and the present invention does not make any specific limitations here.

[0052] Compared with the prior art, the present embodiment provides a method for predicting the yield load and damage degree of explosively welded composite materials, which combines the damage research of explosively welded composite materials with machine learning technology. By mining the potential laws in the material damage data, the relationship between damage patterns and characteristics that are difficult to detect by traditional methods is discovered, thereby gaining a more comprehensive understanding of the damage of the composite materials. Compared with traditional experimental research methods, microstructure analysis methods, and numerical simulation methods, the scheme proposed in the present invention can quickly and accurately obtain damage results and the yield load corresponding to the current material state, and has lower cost and higher safety. When preparing composite materials, if the current explosion welding parameters do not meet the predetermined wavelength and amplitude requirements, a genetic algorithm is used to optimize the current explosion welding parameters to obtain explosion welding parameters with the predetermined wavelength and amplitude requirements. When preparing composite materials, the present invention reduces the difficulty of preparation and no longer requires technical personnel to make adjustments based on experience. Only relying on the provided genetic algorithm can quickly obtain explosion welding parameters that are compatible with the current predetermined requirements. In the mutation operation of the genetic algorithm, neighborhood solution optimization is combined to quickly find high-quality solutions through efficient local search, which can not only improve the search efficiency of the algorithm and the quality of the solution, but also enhance the robustness and global search capabilities of the algorithm, thereby showing better performance in solving complex optimization problems.

[0053] 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.

[0054] 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 method for predicting the yield load and damage degree of explosively welded composite materials, characterized in that: include: Acquire a damage image of the composite material to be tested, and input the damage image into a first image recognition model to obtain the number and size of holes and cracks in the composite material to be tested, and count the minimum and maximum values ​​of the holes and cracks; Inputting the damage image into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be tested; The cross-sectional area change rate of the composite material to be tested is obtained, and the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, and the amplitude are input into a first load damage prediction model to obtain the yield load and damage grade of the composite material to be tested.

2. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 1, characterized in that: The damage image at least includes , , , X-ray scanning image; the first image recognition model is a 3D ResNet network model, the second image recognition model is a model composed of a U-NET model and a regression head, and the first load damage prediction model is a machine learning model.

3. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 2, characterized in that: A hole convolution is introduced into each residual block in the 3D ResNet network model, and a SE attention module is introduced after each residual block.

4. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 2, characterized in that: The first load damage prediction model is a support vector machine (SVM) model, which is implemented by hard parameter sharing, and the output parameters are yield load and damage level respectively; The SVM model is trained based on a first sample data set; wherein each sample data in the first sample data set includes: cross-sectional area change rate, number of holes and their minimum and maximum values, number of cracks and their minimum and maximum values, wavelength, amplitude, yield load and damage level.

5. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 4, characterized in that: Construct the first sample dataset as follows: A variety of composite materials are prepared based on explosive welding parameters. For each composite material, the corresponding first sample data is obtained in the following manner to form a first sample data set: Three in-situ tensile specimens were cut from the composite material by machine turning; An in-situ tensile specimen is selected to conduct an in-situ tensile test based on the in-situ tensile test platform to obtain the load-displacement curve corresponding to the specimen, and then the yield load is obtained; An in-situ tensile specimen is selected to obtain a plurality of X-ray scanning images based on the in-situ tensile test platform, and the images are input into a second image recognition model to obtain the wavelength and amplitude of the wavy composite interface in the composite material to be tested; An in-situ tensile specimen is selected to carry out an in-situ tensile test based on an in-situ tensile test platform to obtain the cross-sectional area change rate, the number and size distribution of holes, and the number and size distribution of cracks of the specimen under different tensile loads; the damage level of the specimen is marked based on the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, and the amplitude of the specimen under each tensile load; the cross-sectional area change rate, the number of holes and their minimum and maximum values, the number of cracks and their minimum and maximum values, the wavelength, the amplitude, the yield load, and the damage level of the specimen under each tensile load are taken as a first sample data.

6. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 5, characterized in that: Composite materials are prepared based on explosive welding parameters, including: The explosion welding parameters are input into a second prediction model to determine whether the output wavelength and amplitude meet the predetermined requirements. If yes, a composite material is prepared based on the explosion welding parameters. If no, a genetic algorithm is used to optimize the input parameters, and the optimized parameters are used as the explosion welding parameters.

7. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 6, characterized in that: The second prediction model is trained based on the second sample data set; wherein each sample data in the second sample data set includes: control parameters in each explosive welding process and wavelengths and amplitudes corresponding to the control parameters, and the control parameters include: collision speed, collision angle, explosive thickness, mass ratio of explosive per unit area to mass ratio of flyer plate, spacing between flyer plate and substrate, and flyer plate thickness; The wavelength and amplitude in each piece of sample data are obtained in the following manner: a flying plate material and a substrate material are selected based on business requirements, and the flying plate and the substrate are composited based on the control parameters to form a specific explosively welded composite plate, an observation sample is prepared based on the composite plate, and the observation sample is observed under an optical microscope to obtain the wavelength and amplitude of the wavy interface of the explosive composite interface of the composite plate.

8. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 7, characterized in that: The composite material to be tested is obtained by using a TA1 material plate as a flying plate and a Q235 material plate as a base plate through an explosion welding technology; The step of preparing an observation sample based on the composite material comprises: An observation sample section is cut from the composite material, and the heterogeneous material laminate surface is mechanically polished. After the observation sample section is immersed in the metallographic etching agent for a first preset time, the surface is rinsed with clean water and blown dry, and then a cotton ball is used to pick up a volume concentration of The nitric acid alcohol solution is wiped on the surface of the sample, and after corrosion for a second preset time, it is cleaned and dried to finally obtain the observation sample.

9. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 8, characterized in that: The in-situ tensile test is performed based on the in-situ tensile test platform to obtain the number and size distribution of holes and the number and size distribution of cracks of the sample under different tensile loads, including: Set the initial tensile load and tensile loading rate; The in-situ tensile specimen is placed in a mechanical fixture and an initial tensile load is applied. When the tensile load increases by 1 N, the in-situ tensile test platform rotates the fixture to take sample slices of the in-situ tensile specimen at different angles. The sample slices are input into the first image recognition model to obtain the number and size of holes and cracks corresponding to the sample slices, and the minimum and maximum values ​​of holes and cracks of the sample under the current tensile load are counted.

10. The method for predicting the yield load and damage degree of explosively welded composite materials according to claim 9, characterized in that: The predetermined requirements include: the wavelength should be between 660-690 The amplitude should be between 60-90 within the scope; The optimization of the input parameters using a genetic algorithm comprises: B1: Initialize the population, set the population size N, preset the number of iterations T, and each individual in the population includes: collision speed, collision angle, explosive thickness, the ratio of the mass of the explosive per unit area to the mass of the flyer plate, the distance between the flyer plate and the base plate, and the thickness of the flyer plate; The collision velocity, collision angle, the distance between the flyer plate and the base plate, and the thickness of the flyer plate are binary coded, and the thickness of the explosive and the mass ratio of the explosive per unit area to the mass of the flyer plate are Gray coded; B2: Calculate the fitness value of each individual according to the fitness function, where the fitness function is: ， in, is the minimum wavelength in the predetermined requirement, is the maximum value of the wavelength in the predetermined requirement, The wavelength obtained by inputting the second prediction model into the i-th individual, is the minimum value of the amplitude in the predetermined demand, The maximum value of the amplitude in the predetermined demand, The amplitude obtained by inputting the second prediction model into the i-th individual, is a constant, i=1,2,3,..,N; B3: Selection operation, based on the tournament algorithm, select individuals with higher fitness from the current population as parents to generate the next generation; B4: Crossover operation, for binary coded parameter groups, a single-point crossover method is used, and for Gray coded parameter groups, a multi-point crossover method is used; B5: mutation operation, select the one with the highest fitness from the population The individuals are searched locally. For each selected individual, several neighborhood solutions are generated near it. A better solution than the current solution is found in the neighborhood solutions. If a better solution is found, the current solution is replaced and the individuals after the local search are put back into the population. B6: Generate a new population, merge the parent and offspring individuals, select a preset number of individuals as the new generation population according to the fitness value, and determine whether the iteration stop condition is met. If so, the best individual in the current population is used as the optimized operating parameter.

Images