Material structure generation method, device, equipment, storage medium and program product

Abstract

This application discloses a method, apparatus, device, storage medium, and program product for generating material structures, belonging to the field of computer technology. In this application, performance indicators are obtained and input into a generator model to obtain structural data of multiple energy-absorbing materials output by the generator model; the structural data of the multiple energy-absorbing materials are input into a performance prediction model to obtain structural performance data of the multiple energy-absorbing materials output by the performance prediction model; when it is determined, based on the structural data and the structural performance data, that at least one energy-absorbing material's structure satisfies specified conditions, the structure of that at least one energy-absorbing material is determined as the target structure. This application ensures that the generated energy-absorbing material structure not only has a reasonable topological configuration but also significantly reduces the deviation between the actual mechanical properties and the target performance.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, device, storage medium, and program product for generating material structures. Background Technology

[0002] With the increasing demand for both lightweighting and crash safety in automobiles, metamaterial energy-absorbing structures have become an important research direction in vehicle crashworthiness design due to their customizable mechanical properties. Metamaterials achieve superior properties not found in naturally occurring materials through artificially designed periodic microstructures, and have broad application prospects in vehicle crash protection, energy absorption, and other fields.

[0003] Existing metamaterial structure design methods primarily rely on generative adversarial networks (GANs) to directly establish a mapping relationship between performance indicators and structural topology. Conditional GANs input specified performance indicators as constraints into the generator, driving the generation of structures that meet these indicators. These methods utilize adversarial training between the discriminator and the generator to optimize the generation quality, enabling, to some extent, rapid structure generation.

[0004] However, since the generation process relies solely on the discriminator's judgment of authenticity for optimization, it lacks direct constraints on the physical properties of the generated structure. As a result, although the generated results are visually close to the real samples, there is a significant deviation between their mechanical properties and the target properties, making it difficult to guarantee the physical effectiveness and engineering practicality of the generated structure. Summary of the Invention

[0005] This application provides a method for generating material structures, which can solve the problem of poor structural performance of materials generated by generator models. The technical solution is as follows: On the one hand, a method for generating a material structure is provided, the method comprising: Obtain performance indicators, input the performance indicators into the generator model, and obtain the structural data of multiple energy-absorbing materials output by the generator model; The structural data of the plurality of energy-absorbing materials are input into the performance prediction model to obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model. When it is determined, based on the structural data and the structural performance data, that a structure with at least one energy-absorbing material satisfies the specified conditions, the structure with at least one energy-absorbing material is determined as the target structure.

[0006] In one possible implementation, determining that a structure containing at least one energy-absorbing material satisfies a specified condition based on the structural data and the structural performance data includes: For each of the plurality of energy-absorbing materials, determine the largest structural similarity index between the structural data of the energy-absorbing material and the structural data of multiple samples in the structural sample library, and determine the difference between the structural performance data and the standard structural performance data corresponding to the performance index. If the structural similarity index is greater than or equal to a first threshold and the difference is less than or equal to a second threshold, then the structure of the energy-absorbing material is determined to satisfy the specified condition.

[0007] In another possible implementation, the performance prediction model is a proxy model.

[0008] In another possible implementation, the training method for the proxy model includes: Obtain structural data samples and corresponding structural performance data samples from multiple energy-absorbing material samples, and construct a dataset based on the structural data samples and corresponding structural performance data samples; The agent model is trained based on the dataset; When the first error between the structural performance data output by the surrogate model for the structural performance data sample and the structural performance data sample decreases to below a specified error, it is determined that the surrogate model training is complete.

[0009] In another possible implementation, after the proxy model has been trained, the following is also included: The generator model is trained based on the trained proxy model.

[0010] In another possible implementation, training the generator model based on the trained proxy model includes: Obtain performance index samples and structural performance data samples corresponding to multiple energy-absorbing material samples; The performance index samples are input into the generator model to obtain the candidate structure data output by the generator model; The candidate structure data is input into the surrogate model to obtain the structural performance data of the candidate structure data; Determine the second error between the structural performance data of the candidate structural data and the structural performance data sample; The parameters of the generator model are updated based on the second error.

[0011] On the other hand, a material structure generation apparatus is provided, the apparatus comprising: The acquisition module is configured to acquire performance indicators, input the performance indicators into the generator model, and obtain the structural data of multiple energy-absorbing materials output by the generator model. The prediction module is configured to input the structural data of the plurality of energy-absorbing materials into the performance prediction model, and obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model. The determination module is configured to determine the structure with at least one energy-absorbing material as the target structure when it is determined, based on the structural data and the structural performance data, that the structure with at least one energy-absorbing material satisfies specified conditions.

[0012] In one possible implementation, the determining module is used to: For each of the plurality of energy-absorbing materials, determine the largest structural similarity index between the structural data of the energy-absorbing material and the structural data of multiple samples in the structural sample library, and determine the difference between the structural performance data and the standard structural performance data corresponding to the performance index. If the structural similarity index is greater than or equal to a first threshold and the difference is less than or equal to a second threshold, then the structure of the energy-absorbing material is determined to satisfy the specified condition.

[0013] In another possible implementation, the performance prediction model is a proxy model.

[0014] In another possible implementation, the material structure generation device also includes a training module for: Obtain structural data samples and corresponding structural performance data samples from multiple energy-absorbing material samples, and construct a dataset based on the structural data samples and corresponding structural performance data samples; The agent model is trained based on the dataset; When the first error between the structural performance data output by the surrogate model for the structural performance data sample and the structural performance data sample decreases to below a specified error, it is determined that the surrogate model training is complete.

[0015] In another possible implementation, the training module is used for: The generator model is trained based on the trained proxy model.

[0016] In another possible implementation, the training module is used for: Obtain performance index samples and structural performance data samples corresponding to multiple energy-absorbing material samples; The performance index samples are input into the generator model to obtain the candidate structure data output by the generator model; The candidate structure data is input into the surrogate model to obtain the structural performance data of the candidate structure data; Determine the second error between the structural performance data of the candidate structural data and the structural performance data sample; The parameters of the generator model are updated based on the second error.

[0017] On the other hand, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the method described in any of the above.

[0018] On the other hand, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method described in any of the preceding claims.

[0019] On the other hand, a computer program product is provided, including computer program instructions that, when run on a computer, cause the computer to perform the method described in any of the preceding claims.

[0020] The beneficial effects of the technical solution provided in this application are: in the process of generating structural data by the generator model, the physical properties of the energy-absorbing material are directly constrained by the performance prediction model, ensuring that the structure of the generated energy-absorbing material is not only topologically reasonable, but also that the deviation between the actual mechanical properties and the target properties is significantly reduced. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application; Figure 2 This is a flowchart of the material structure generation method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the material structure generation device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0024] This disclosure provides a method for generating material structures. This method can be applied to terminals. For example... Figure 1As shown, the terminal may include a processor 110, a memory 120, and a communication component 130.

[0025] Processor 110 can be a central processing unit (CPU), graphics processing unit (GPU), microcontroller unit (MCU), accelerated processing unit (APU), neural processing unit (NPU), tensor processing unit (TPU), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), digital signal processor (DSP), etc. Processor 110 can be used to acquire performance metrics, run generator models, and so on.

[0026] Memory 120 may include volatile memory and / or non-volatile memory. Volatile memory may include random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), etc. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, non-volatile random access memory (NVRAM), etc. Memory 120 can be used to store structural data, structural performance data, etc.

[0027] The communication component 130 can be a wireless communication module (WCM), a subscriber identity module (SIM), a universal subscriber identity module (USIM), an optical network unit (ONU), etc. The communication component 130 can be used for communication between the terminal and a server, or other terminals.

[0028] This application provides a method for generating material structures, such as Figure 2 As shown, in some embodiments, the method includes: S201. Obtain performance indicators and input the performance indicators into the generator model to obtain the structural data of multiple energy-absorbing materials output by the generator model.

[0029] In practical implementation, the performance indicators that the energy-absorbing material structure needs to meet can be determined according to the requirements of the actual application scenario. These performance indicators are quantitative parameters characterizing the material's energy absorption capacity, mechanical stability, etc., and can be flexibly selected according to different application scenarios. Typically, energy absorption, specific energy absorption, compressive force efficiency, and initial peak force are used as performance indicators. For example, energy absorption greater than or equal to 10 kJ, specific energy absorption greater than or equal to 20 kJ / kg, compressive force efficiency greater than or equal to 0.7, and initial peak force less than or equal to 50 kN. , , , In the formula, EA represents energy absorption. For effective compressive displacement, P(y) is the load-displacement curve function representing instantaneous compressive force, and SEA is the specific energy absorption. CFE is the average compressive load, and CFE is the compressive efficiency. Let be the initial peak force, where It is based on energy absorption efficiency What has been determined is In the formula, Let be the total energy absorbed by the material when it is compressed from its initial state to a displacement y. for Peak force within the compression range, Represents the theoretical maximum energy absorption potential, with a value range of [value missing]. Adjust the determined performance metrics to a format that meets the input requirements of the generator model, and use them as input conditions for the generator, such as standardized parameter vectors or parameter matrices.

[0030] The performance metrics are input into a pre-built and initialized generator model. The generator model can be a deep learning model or intelligent generative model that can achieve "performance metrics to structural data mapping", such as a generator in a fully connected neural network, a convolutional neural network, a generative adversarial network, or a variational autoencoder. Based on the input performance metrics, the generator outputs structural data of multiple energy-absorbing materials with different configurations. The structural data is quantitative data that can characterize the material's topological configuration and geometric parameters, such as the structure's porosity, pore size, skeleton thickness, pore shape distribution (circular, square, hexagonal), or a two-dimensional / three-dimensional structural topology matrix.

[0031] S202. Input the structural data of the plurality of energy-absorbing materials into the performance prediction model to obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model.

[0032] In practice, the structural data of multiple energy-absorbing materials output by the generator model are input into a pre-trained performance prediction model. The performance prediction model can establish a mapping relationship between structural data and structural performance, and output the structural performance data corresponding to each structural data. The structural performance data is data corresponding to the input performance index and is used to evaluate the actual performance of the generated structure, such as energy absorption efficiency.

[0033] Specifically, the structural data of the above 10 energy-absorbing materials are input into the trained performance prediction model (i.e., the surrogate model). The surrogate model is a differentiable feedforward predictive neural network. Through supervised learning, a mapping from the structural design space to the performance space is established, which can quickly output the structural performance data corresponding to each structure.

[0034] S203. When it is determined, based on the structural data and the structural performance data, that a structure with at least one energy-absorbing material satisfies the specified conditions, the structure with at least one energy-absorbing material is determined as the target structure.

[0035] In practice, based on the structural data and the structural performance data output by the performance prediction model, it is determined whether there is at least one energy-absorbing material structure that meets the specified condition. If so, the structure that meets the condition is identified as the target structure for subsequent practical applications or further optimization. For example, if 10 structures are generated, and 3 of them have reasonable porosity, are manufacturable, and have an actual energy absorption efficiency of ≥80%, these 3 structures can be identified as the target structures.

[0036] Specifically, for each of the plurality of energy-absorbing materials, the largest structural similarity index between the structural data of the energy-absorbing material and the structural data of multiple samples in the structural sample library is determined (cosine similarity method, Euclidean distance method, structural feature matching method, etc. can be used, and the maximum value among all structural similarity indices is taken), and the difference between the structural performance data (energy absorption efficiency) and the standard structural performance data corresponding to the performance index is determined. When the structural similarity index is greater than or equal to a first threshold (e.g., 0.75), and the difference is less than or equal to a second threshold (e.g., 5%), it is determined that the structure of the energy-absorbing material satisfies the specified condition.

[0037] In this embodiment, a performance prediction model is introduced to directly constrain physical performance during the generation process. This ensures that the generated energy-absorbing material structure not only has a reasonable topological configuration, but also significantly reduces the deviation between actual mechanical properties such as energy absorption efficiency and target performance indicators. This solves the problem of large deviations between mechanical performance and target performance, insufficient physical effectiveness and engineering practicality caused by the lack of direct physical performance constraints in existing generation methods that rely on generative adversarial networks. At the same time, through dual screening of structural similarity index and performance difference, it is possible to efficiently and accurately identify high-quality target structures that meet manufacturing requirements and performance needs from multiple generated structures. This improves the customization efficiency and reliability of metamaterial energy-absorbing structures in vehicle lightweighting and collision safety design, and provides a practical solution for engineering applications.

[0038] In some embodiments, the training method of the proxy model includes: Obtain structural data samples and corresponding structural performance data samples of multiple energy-absorbing material samples, and construct a dataset based on the structural data samples and corresponding structural performance data samples; train the surrogate model based on the dataset; determine that the surrogate model training is complete when the first error between the structural performance data output by the surrogate model for the structural data samples and the structural performance data samples is reduced to below a specified error.

[0039] In practice, structural data samples and corresponding structural performance data samples of multiple energy-absorbing material samples are collected. The structural data samples are quantitative data characterizing the structural configuration of the samples; the structural performance data samples are quantitative data characterizing the actual performance of the samples, with the core being the measured energy absorption efficiency of the samples. Based on these data, a dataset for training the surrogate model is constructed. The dataset can be divided into training sets, validation sets, test sets, etc., according to training requirements.

[0040] The constructed dataset is input into the surrogate model to be trained. The model parameters are updated through backpropagation and other methods to optimize the model's prediction accuracy of energy absorption efficiency. During training, appropriate optimizers, loss functions, and strategies such as learning rate scheduling and early stopping can be used to improve training efficiency and model performance. Adam optimizers or SGD optimizers can be used, with appropriate learning rates (e.g., 0.001, 0.0001) and mean absolute error as the loss function. Training is stopped when the performance on the validation set no longer improves for several consecutive training epochs (e.g., 5, 10) to avoid model overfitting.

[0041] During model training, the first error between the structural performance data output by the surrogate model for the structural data sample and the corresponding structural performance data sample is continuously monitored. The first error can be calculated using methods such as mean absolute error or mean square error. When the first error is reduced to below the preset specified error, the prediction accuracy of the surrogate model is considered to have met the requirements, and the surrogate model training is determined to be complete. For example, if the specified error is set to be less than or equal to 2%, when the model training reaches 1.8%, the requirement is met, and training is stopped.

[0042] In this embodiment, a dataset is constructed by acquiring structural data samples and corresponding structural performance data samples of multiple energy-absorbing material samples. The surrogate model is then trained under supervision based on this dataset. During the training process, the model parameters are continuously optimized until the first error between the structural performance data output by the surrogate model for the structural data samples and the structural performance data samples is reduced to below a specified error. This ensures that the surrogate model can establish a high-precision mapping relationship from the structural design space to the performance space. As a result, in the subsequent structural generation process, it provides accurate and reliable structural performance data prediction for the structural data of multiple energy-absorbing materials output by the generator model, effectively improving the robustness and engineering practicality of the entire metamaterial energy-absorbing structure generation method.

[0043] In some embodiments, after the proxy model training is completed, the method further includes: The generator model is trained based on the trained proxy model.

[0044] In specific implementation, performance index samples and structural performance data samples corresponding to multiple energy-absorbing material samples are obtained; the performance index samples are input into the generator model to obtain candidate structural data output by the generator model; the candidate structural data are input into the surrogate model to obtain structural performance data of the candidate structural data; a second error between the structural performance data of the candidate structural data and the structural performance data samples is determined; and the parameters of the generator model are updated based on the second error.

[0045] Specifically, a second error is determined between the structural performance data (predicted energy absorption efficiency) of the candidate structural data and the structural performance data samples (measured energy absorption efficiency). This error can be represented by the mean of the absolute errors. The second error equals the sum of the differences between the absolute values ​​of the predicted and measured energy absorption efficiencies of all samples, divided by the number of samples. The number of samples is the total number of energy-absorbing material samples participating in the training. The predicted performance value is the energy absorption efficiency of the i-th sample output by the surrogate model, and the measured performance value is the actual measured energy absorption efficiency of the i-th sample. For example, if there are 3 samples with predicted energy absorption efficiencies of 0.68, 0.72, and 0.69, and corresponding measured energy absorption efficiencies of 0.7, 0.7, and 0.7, then the sum of the absolute errors of all samples is (0.02 + 0.02 + 0.01) = 0.05. The second error equals 0.05 divided by 3, resulting in approximately 0.017. Simultaneously, combining the adversarial loss of the generative adversarial network (GAN), a total loss function is constructed. The total loss equals the adversarial loss of the GAN plus 0.3 multiplied by the auxiliary loss (i.e., the loss corresponding to the second error), where 0.3 is the weight of the auxiliary loss (which can be set according to specific circumstances) to balance the adversarial loss and the energy absorption efficiency error loss. The adversarial loss here is calculated based on the adversarial training results of the generator and discriminator. It is the sum of the discriminator's prediction loss for real structure data (hoping that the discriminator accurately identifies the real structure and outputs a probability close to 1) and the prediction loss for generated structure data (hoping that the discriminator cannot distinguish between generated and real structures and outputs a probability close to 0.5), plus the loss of the generator's generated structure being misjudged as a real structure by the discriminator (hoping that the structure generated by the generator can fool the discriminator and make the discriminator output a probability close to 1). Through the backpropagation of the adversarial loss, the co-optimization of the generator and discriminator is achieved.

[0046] Based on the aforementioned total loss function, the parameters of the generator model are updated using the backpropagation algorithm. The weights and biases of the generator's convolutional and transposed convolutional layers are adjusted to make the energy absorption efficiency of the generated candidate structures closer to the real performance of the samples, while ensuring that the generated structures can fool the discriminator model. Simultaneously, the parameters of the discriminator model are updated, and the weights of the discriminator's convolutional layers are optimized through backpropagation to improve its ability to distinguish between real and generated structures. The generator and discriminator models alternately update their parameters, repeating the above training steps until the total loss function decreases to a stable value, the deviation between the energy absorption efficiency of the generated structures and the target performance is less than 5%, and the discriminator's accuracy in distinguishing between real and generated structures stabilizes at around 50% (indicating that the generator-generated structures have achieved a level of realism that is indistinguishable from genuine structures). This completes the collaborative training of the generator and discriminator models.

[0047] In this embodiment, after the surrogate model is trained, the generator model is co-trained based on the trained surrogate model. The generator model is driven to generate candidate structure data by using performance index samples and structural performance data samples of multiple energy-absorbing material samples. The high-precision surrogate model is used to predict the structural performance data of the candidate structure data. The second error between the structural performance data of the candidate structure data and the structural performance data sample is calculated and used as an auxiliary loss. It is combined with the adversarial loss of the generative adversarial network to construct the total loss function. Then, based on the total loss function, the parameters of the generator model and the discriminator model are alternately updated through the backpropagation algorithm. This significantly improves the physical effectiveness and engineering practicality of the structure generated by the generator model and solves the problem of large deviation between the mechanical performance of the generated structure and the target performance in the prior art.

[0048] In some embodiments, the method further includes: When the structural similarity index is less than a first threshold (e.g., 0.75), greater than or equal to a third threshold (e.g., 0.65), and the difference is less than or equal to a fourth threshold (e.g., 3%), then the structure of the energy-absorbing material is determined to meet the specified conditions.

[0049] When the structural similarity index is greater than or equal to the fifth threshold (e.g., 0.85), and the difference is greater than the second threshold (e.g., 5%) and less than or equal to the sixth threshold (e.g., 7%), then the structure of the energy-absorbing material is determined to meet the specified conditions.

[0050] By adding conditions to the existing specified conditions, such as when the structural similarity index is less than the first threshold but greater than or equal to the third threshold and the difference is less than or equal to the fourth threshold, and when the structural similarity index is greater than or equal to the fifth threshold and the difference is greater than the second threshold but less than or equal to the sixth threshold, the flexibility and adaptability of target structure screening are significantly enhanced. This allows for a more refined trade-off between the structural similarity index and performance difference to meet the needs of different engineering scenarios. For example, when the topological configuration of the generated structure differs from that of the sample library but the energy absorption efficiency deviation is extremely small, or when the generated structure is highly similar to that of the sample library but the energy absorption efficiency deviation is slightly large, it can still be effectively identified as the target structure. This avoids the problem of potential high-quality structures being missed due to the original single fixed threshold, expands the search range of qualified structures, and thus improves the coverage and fault tolerance of the metamaterial energy-absorbing structure generation method while ensuring the physical validity and engineering practicality of the generated structure.

[0051] In some embodiments, the generative adversarial training process between the generator model and the discriminator model includes: initializing the parameters of the generator model and the discriminator model respectively to ensure that the models are in a stable initial state, with the initial parameters of the generator model being randomly set; extracting a batch of real metamaterial energy-absorbing structure data from a preset real structure sample library as real samples, while simultaneously inputting performance index samples into the generator model to be trained to generate a batch of candidate structure data as generated samples; inputting the real samples and generated samples into the discriminator model respectively, with the real sample label set to 1 (representing a real structure) and the generated sample label set to 0 (representing a generated structure), updating the discriminator model parameters through the backpropagation algorithm to optimize the discriminator model's feature extraction capability and authenticity judgment capability, enabling the discriminator model to accurately distinguish between real structures and generated structures, and using the cross-entropy loss function during training to measure the deviation between the discriminator's prediction results and the real labels; inputting the generated samples into the trained discriminator model to obtain the discriminator model's performance. The generator model results (at this point, the generator's goal is to make the discriminator misclassify the generated samples as real samples, i.e., the prediction probability is close to 1). Simultaneously, the generator's total loss function (total loss = generation adversarial loss + surrogate model performance bias loss) can be constructed by combining the performance prediction results of the generated samples output by the surrogate model. The generator parameters are then updated using the backpropagation algorithm to adjust the generation logic, improving the realism of the generated structure (deceiving the discriminator model) while ensuring the performance of the generated structure meets the target requirements (fitting the surrogate model's performance constraints). The above steps are repeated, alternating between training the discriminator and generator, gradually optimizing their parameters in each iteration until the training termination condition is met: the discriminator model's accuracy in judging real and generated samples stabilizes at around 50% (indicating that the generator model's generated structure has achieved a level of realism that is indistinguishable from the real sample), and the performance bias of the generator model's generated structure (verified by the surrogate model) is below a threshold. At this point, the generation adversarial training between the generator model and the discriminator model is complete.

[0052] In this embodiment, the problem of large deviation between the mechanical properties of the generated structure and the target performance caused by the lack of direct physical performance constraints in traditional generative adversarial network training is effectively solved. It achieves dual assurance in terms of topological realism and physical performance accuracy of the generated structure, and significantly enhances the reliability and engineering practical value of the generated metamaterial energy-absorbing structure in vehicle collision protection and energy absorption applications.

[0053] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0054] Based on the same inventive concept, and corresponding to the material structure generation method provided in the embodiments of this application, this application also provides a material structure generation apparatus.

[0055] refer to Figure 3The material structure generation apparatus includes: The acquisition module 301 is configured to acquire performance indicators, input the performance indicators into the generator model, and obtain the structural data of multiple energy-absorbing materials output by the generator model. The prediction module 302 is configured to input the structural data of the plurality of energy-absorbing materials into the performance prediction model, and obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model. The determination module 303 is configured to determine the structure with at least one energy-absorbing material as the target structure when it is determined, based on the structural data and the structural performance data, that the structure with at least one energy-absorbing material satisfies the specified conditions.

[0056] In one possible implementation, the determining module 303 is used to: For each of the plurality of energy-absorbing materials, determine the largest structural similarity index between the structural data of the energy-absorbing material and the structural data of multiple samples in the structural sample library, and determine the difference between the structural performance data and the standard structural performance data corresponding to the performance index. If the structural similarity index is greater than or equal to a first threshold and the difference is less than or equal to a second threshold, then the structure of the energy-absorbing material is determined to satisfy the specified condition.

[0057] In another possible implementation, the performance prediction model is a proxy model.

[0058] In another possible implementation, the material structure generation device further includes a training module 304, used for: Obtain structural data samples and corresponding structural performance data samples from multiple energy-absorbing material samples, and construct a dataset based on the structural data samples and corresponding structural performance data samples; The agent model is trained based on the dataset; When the first error between the structural performance data output by the surrogate model for the structural performance data sample and the structural performance data sample decreases to below a specified error, it is determined that the surrogate model training is complete.

[0059] In another possible implementation, the training module 304 is used for: The generator model is trained based on the trained proxy model.

[0060] In another possible implementation, the training module 304 is used for: Obtain performance index samples and structural performance data samples corresponding to multiple energy-absorbing material samples; The performance index samples are input into the generator model to obtain the candidate structure data output by the generator model; The candidate structure data is input into the surrogate model to obtain the structural performance data of the candidate structure data; Determine the second error between the structural performance data of the candidate structural data and the structural performance data sample; The parameters of the generator model are updated based on the second error.

[0061] It should be noted that the material structure generation apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when generating material structures. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the apparatus can be divided into different functional modules to complete all or part of the functions described above. In addition, the material structure generation apparatus and the material structure generation method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0062] Based on the same inventive concept, corresponding to the material structure generation method provided in the embodiments of this application, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the material structure generation method described in the above embodiments.

[0063] Figure 4 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0064] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0065] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0066] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0067] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0068] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0069] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0070] The electronic device described above is used to implement the corresponding material structure generation method in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0071] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to complete the material structure generation method described above. This computer-readable storage medium can be non-transitory. For example, the computer-readable storage medium can be ROM (Read-Only Memory), RAM (Random Access Memory), CD-ROM (Compact Disc Read-Only Memory), magnetic tape, floppy disk, and optical data storage devices, etc.

[0072] In an exemplary embodiment, a computer program product is also provided, including computer program instructions that, when executed on a computer, cause the computer to perform the material structure generation method described above.

[0073] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals (including but not limited to signals transmitted between user terminals and other devices, etc.) involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0074] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0075] It should be understood that "multiple" as used herein refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the step numbers described herein are merely illustrative of one possible execution order. In some other embodiments, the steps may not be executed in numerical order, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.

[0076] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for generating a material structure, characterized in that, include: Obtain performance indicators, input the performance indicators into the generator model, and obtain the structural data of multiple energy-absorbing materials output by the generator model; The structural data of the plurality of energy-absorbing materials are input into the performance prediction model to obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model. When it is determined, based on the structural data and the structural performance data, that a structure with at least one energy-absorbing material satisfies the specified conditions, the structure with at least one energy-absorbing material is determined as the target structure.

2. The material structure generation method according to claim 1, characterized in that, The determination, based on the structural data and the structural performance data, that a structure containing at least one energy-absorbing material satisfies a specified condition includes: For each of the plurality of energy-absorbing materials, determine the largest structural similarity index between the structural data of the energy-absorbing material and the structural data of multiple samples in the structural sample library, and determine the difference between the structural performance data and the standard structural performance data corresponding to the performance index. If the structural similarity index is greater than or equal to a first threshold and the difference is less than or equal to a second threshold, then the structure of the energy-absorbing material is determined to satisfy the specified condition.

3. The material structure generation method according to claim 1, characterized in that, The performance prediction model is a surrogate model.

4. The material structure generation method according to claim 3, characterized in that, The training method for the proxy model includes: Obtain structural data samples and corresponding structural performance data samples from multiple energy-absorbing material samples, and construct a dataset based on the structural data samples and corresponding structural performance data samples; The agent model is trained based on the dataset; When the first error between the structural performance data output by the surrogate model for the structural performance data sample and the structural performance data sample decreases to below a specified error, it is determined that the surrogate model training is complete.

5. The material structure generation method according to claim 4, characterized in that, After the proxy model training is completed, the following is also included: The generator model is trained based on the trained proxy model.

6. The material structure generation method according to claim 5, characterized in that, The process of training the generator model based on the trained proxy model includes: Obtain performance index samples and structural performance data samples corresponding to multiple energy-absorbing material samples; The performance index samples are input into the generator model to obtain the candidate structure data output by the generator model; The candidate structure data is input into the surrogate model to obtain the structural performance data of the candidate structure data; Determine the second error between the structural performance data of the candidate structural data and the structural performance data sample; The parameters of the generator model are updated based on the second error.

7. A material structure generation device, characterized in that, include: The acquisition module is configured to acquire performance indicators, input the performance indicators into the generator model, and obtain the structural data of multiple energy-absorbing materials output by the generator model. The prediction module is configured to input the structural data of the plurality of energy-absorbing materials into the performance prediction model, and obtain the structural performance data of the plurality of energy-absorbing materials output by the performance prediction model. The determination module is configured to determine the structure with at least one energy-absorbing material as the target structure when it is determined, based on the structural data and the structural performance data, that the structure with at least one energy-absorbing material satisfies specified conditions.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method described in any one of claims 1 to 6.

10. A computer program product comprising computer program instructions, characterized in that, When the computer program instructions are executed on a computer, the computer causes the computer to perform the method as described in any one of claims 1 to 6.

Images