Method for generating microstructure of porous material with effective thermal conductivity and related device

Abstract

The present application belongs to the field of porous materials, and discloses a method for generating a porous material microstructure with a set effective thermal conductivity and a related device, comprising: obtaining a random noise image and a set effective thermal conductivity; generating a porous material microstructure image with the set effective thermal conductivity according to the random noise image and the set effective thermal conductivity through a pre-trained improved conditional diffusion model; wherein the improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual up-sampling channel decoder. The improved conditional diffusion model in the present application uses a denoising neural network with a dual up-sampling channel decoder, so that the denoising neural network has flexible coding depth and can better adapt to the generation of porous material microstructures, so that the relative error between the set effective thermal conductivity coefficient of the porous material microstructure image generated by using the improved conditional diffusion model and the real effective thermal conductivity coefficient is smaller, and the generation quality is more stable.

Description

Technical Field

[0001] This invention belongs to the field of porous materials and relates to a method and related apparatus for generating the microstructure of porous materials with a set effective thermal conductivity. Background Technology

[0002] The porous structure of porous ceramic materials endows them with excellent properties such as lightweight, high strength, and tunable thermal conductivity, leading to their wide applications in sound absorption, heat insulation, filtration, catalysis, and biomedicine. Effective thermal conductivity (ETC), one of the most important thermophysical properties of porous ceramic materials, is crucial for their performance in thermal and engineering applications. By precisely controlling microscopic characteristics such as pore morphology, pore size, and pore distribution, the thermal conductivity of porous ceramic materials can be customized. This directional design can meet the highly customized performance requirements of specific application areas, such as energy and electronic devices. However, the complexity and diversity of the microstructure of porous materials make it extremely difficult to study the precise mapping relationship between microstructure and effective thermal conductivity, which greatly limits the widespread application of advanced porous ceramic materials in the field of thermal insulation.

[0003] Because convolutional neural networks can implicitly capture features beyond human prior knowledge, they can easily establish a precise mapping relationship between prior information and microstructure, while simultaneously outputting images of porous media, i.e., the structure matrix. Many researchers have used convolution-based generative adversarial networks (GANs) to reconstruct and generate microstructures of porous media. However, with the continuous improvement of generative neural network theory, previously used image generation networks based on the generative adversarial principle have many drawbacks. Training a GAN is quite difficult, and its training process is often unstable, sometimes leading to pattern collapse or pattern oscillation. This is because the optimization objectives of the generative and discriminative networks are contradictory, requiring careful adjustment of the learning rate and other hyperparameters to achieve a balance. Secondly, GANs may face model collapse during training, meaning the generative network cannot produce diverse outputs and can only generate certain patterns from the training data. In this case, the generator loses its creativity, the generated samples lack diversity, and the model's practicality is reduced.

[0004] With the continuous development of image generation models based on the diffusion principle, diffusion models can, to some extent, overcome the limitations of traditional generative adversarial networks (GANs) in generating images, which lack diversity and realism. In diffusion models, by introducing noise, the model gradually learns to recover the true data distribution from random noise. This process enables diffusion models to generate diverse and high-quality images, and the model's generalization ability is also superior. Currently, the network model architecture for diffusion models to learn to recover the true data distribution from random noise is the classic encoder-decoder architecture, namely the UNet architecture. The UNet architecture uses a simple encoder-decoder structure, meaning the decoder contains only one upsampling channel. This results in a lack of multi-level feature fusion in the decoding stage, leading to a decrease in prediction accuracy. Secondly, the encoder depth of the UNet architecture, i.e., the number of convolutional layers in the encoder, is fixed, which leads to significant differences in model performance across different datasets. Therefore, current diffusion models, when applied to the generation of porous media microstructures, produce microstructures of low quality. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and related apparatus for generating microstructures of porous materials with set effective thermal conductivity.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] In a first aspect, the present invention provides a method for generating a porous material microstructure with a set effective thermal conductivity, comprising:

[0008] Acquire random noise images and set effective thermal conductivity;

[0009] Based on random noise images and a set effective thermal conductivity, a microstructure image of a porous material with a set effective thermal conductivity is generated using a pre-trained improved conditional diffusion model.

[0010] The improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual upsampling channel decoder.

[0011] Optionally, the denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a main decoder;

[0012] The conditional coding module is used to embed the time step describing the noise diffusion level of the random noise image and the set effective thermal conductivity into the random noise image to obtain the embedded data image and output it to the encoder;

[0013] The encoder is used to compress the embedded data image and extract the data features of the embedded data image by increasing the number of channels of the feature map to obtain a high-level feature map. Based on the number of downsampling, it outputs a high-level feature map with a larger resolution to the sub-decoder and a high-level feature map with a smaller resolution to the main decoder.

[0014] The secondary decoder is used to restore the high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image, and gradually reduce the number of channels of the high-resolution feature map to obtain the decoding result of the secondary decoder, which is then sent to the main decoder.

[0015] The main decoder is used to restore the high-level feature map with a lower resolution obtained by the encoder to the same spatial resolution as the input random noise image. Then, it is fused with the decoding result of the sub-decoder to obtain a noise prediction result with the same number of channels as the input noise image.

[0016] Optionally, the pre-training of the improved conditional diffusion model includes:

[0017] Acquire microstructure images of porous materials with different pore sizes, porosities, porosity distributions, and pore morphologies;

[0018] Calculate the effective thermal conductivity of microstructure images of porous materials;

[0019] An improved conditional diffusion model is trained based on the microstructure images of porous materials and their effective thermal conductivity.

[0020] Optionally, obtaining microstructure images of porous materials with different porosities, different porosity distributions, and different pore morphologies includes:

[0021] Based on the gradient QSGS algorithm, microstructure images of porous materials with different porosities, different porosity distributions, and different pore morphologies are generated.

[0022] Optionally, the effective thermal conductivity for calculating the microstructure image of the porous material includes:

[0023] The effective thermal conductivity of porous materials is determined using the finite element method.

[0024] Optionally, the effective thermal conductivity obtained by using the finite element method to solve the microstructure image of porous materials includes:

[0025] Define the computational boundaries for the microstructure image of the porous material; where the left and right boundaries are adiabatic boundaries.

[0026] Based on the thermal conductivity of the fixed material phase and the thermal conductivity of the porous phase, the vertical heat flux of each node in the microstructure image of the porous material is calculated.

[0027] The effective thermal conductivity (ETC) of the porous material microstructure image is obtained by the following formula:

[0028]

[0029] Where q is the vertical heat flux of the node; W is the width of the microstructure image of the porous material; ΔT is the temperature difference between the upper and lower boundaries of the microstructure image of the porous material; and H is the height of the microstructure image of the porous material.

[0030] In a second aspect, the present invention provides a system for generating a porous material microstructure with a set effective thermal conductivity, comprising:

[0031] The data acquisition module is used to acquire random noise images and set the effective thermal conductivity.

[0032] The structure generation module is used to generate microstructure images of porous materials with a set effective thermal conductivity based on random noise images and a pre-trained improved conditional diffusion model.

[0033] The improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual upsampling channel decoder.

[0034] Optionally, the denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a main decoder;

[0035] The conditional coding module is used to embed the time step describing the noise diffusion level of the random noise image and the set effective thermal conductivity into the random noise image to obtain the embedded data image and output it to the encoder;

[0036] The encoder is used to compress the embedded data image and extract the data features of the embedded data image by increasing the number of channels of the feature map to obtain a high-level feature map. Based on the number of downsampling, it outputs a high-level feature map with a larger resolution to the sub-decoder and a high-level feature map with a smaller resolution to the main decoder.

[0037] The secondary decoder is used to restore the high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image, and gradually reduce the number of channels of the high-resolution feature map to obtain the decoding result of the secondary decoder, which is then sent to the main decoder.

[0038] The main decoder is used to restore the high-level feature map with a lower resolution obtained by the encoder to the same spatial resolution as the input random noise image. Then, it is fused with the decoding result of the sub-decoder to obtain a noise prediction result with the same number of channels as the input noise image.

[0039] In a third aspect, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for generating a porous material microstructure with a set effective thermal conductivity.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for generating a porous material microstructure with a set effective thermal conductivity.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] This invention provides a method for generating microstructures of porous materials with a set effective thermal conductivity. Based on a random noise image and a set effective thermal conductivity, a pre-trained improved conditional diffusion model is used to generate images of microstructures of porous materials with the set effective thermal conductivity. The improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model to a denoising neural network with a dual-upsampled-channel decoder. Since the performance of the denoising neural network in the conditional diffusion model significantly affects the quality of the generated image, and the traditional conditional diffusion model uses the classic UNet architecture with a fixed encoding depth, its performance varies greatly across different datasets. The improved conditional diffusion model in this invention, by employing a denoising neural network with a dual-upsampled-channel decoder, allows the denoising neural network to have a flexible encoding depth, which can better adapt to the generation of microstructures of porous materials. This results in a smaller relative error between the set effective thermal conductivity and the actual effective thermal conductivity of the microstructure images of porous materials generated using this improved conditional diffusion model, leading to more stable generation quality. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating a method for generating microstructures of porous materials with set effective thermal conductivity, according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the generator G structure according to an embodiment of the present invention.

[0045] Figure 3 This is a schematic diagram of the discriminator D structure according to an embodiment of the present invention.

[0046] Figure 4 This is a flowchart illustrating the training and sample generation process of the conditional diffusion model in an embodiment of the present invention.

[0047] Figure 5 This is a schematic diagram of a denoising neural network structure with a dual upsampling channel decoder according to an embodiment of the present invention.

[0048] Figure 6These are comparative images of the microstructures of different porous materials according to embodiments of the present invention.

[0049] Figure 7 This is a schematic diagram illustrating the application of boundary conditions in an embodiment of the present invention.

[0050] Figure 8 This is a comparison chart of the average porosity of the generated pore structure with a set effective thermal conductivity and the pore structure in the dataset, according to an embodiment of the present invention.

[0051] Figure 9 This is a graph showing the trend of cDCGAN model generation accuracy as a function of a set thermal conductivity coefficient, according to an embodiment of the present invention.

[0052] Figure 10 This is a graph showing the trend of the improved conditional diffusion model generation accuracy as a function of a set thermal conductivity, according to an embodiment of the present invention.

[0053] Figure 11 This is a structural block diagram of a porous material microstructure generation system with a set effective thermal conductivity, according to an embodiment of the present invention. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0055] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0056] The present invention will now be described in further detail with reference to the accompanying drawings:

[0057] See Figure 1In one embodiment of the present invention, a method for generating microstructures of porous materials with a set effective thermal conductivity is provided, which can generate high-quality gradient porous ceramic materials with a set thermal conductivity, thereby effectively promoting the directional design and customized preparation of porous materials.

[0058] Specifically, the method for generating the microstructure of porous materials with a set effective thermal conductivity includes the following steps:

[0059] S1: Obtain random noise image and set effective thermal conductivity.

[0060] S2: Based on the random noise image and the set effective thermal conductivity, the microstructure image of the porous material with the set effective thermal conductivity is generated through a pre-trained improved conditional diffusion model.

[0061] The improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual upsampling channel decoder.

[0062] Since the performance of the denoising neural network in the conditional diffusion model significantly affects the quality of the generated image, and the traditional conditional diffusion model uses the classic UNet architecture with a fixed encoding depth, its performance varies greatly across different datasets. This invention presents a method for generating porous material microstructures with a set effective thermal conductivity. It employs an improved conditional diffusion model, utilizing a denoising neural network with a dual-upsampling channel decoder. This allows the denoising neural network to have a flexible encoding depth, better adapting to the generation of porous material microstructures. Consequently, the relative error between the set effective thermal conductivity and the true effective thermal conductivity in the generated porous material microstructure image is smaller, resulting in more stable generation quality.

[0063] In one possible implementation, the denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a master decoder.

[0064] The conditional encoding module embeds a time step describing the noise diffusion level of a random noise image and a set effective thermal conductivity into the random noise image to obtain an embedded data image, which is then output to the encoder. The encoder compresses the embedded data image and extracts data features from it by increasing the number of channels in the feature map, resulting in a high-level feature map. Based on the number of downsampling operations, it outputs a high-resolution feature map to the sub-decoder and a low-resolution feature map to the main decoder. The sub-decoder restores the high-resolution feature map obtained from the encoder to the same spatial resolution as the input random noise image and gradually reduces the number of channels in the high-level feature map, sending the decoding result to the main decoder. The main decoder restores the low-resolution feature map obtained from the encoder to the same spatial resolution as the input random noise image and then fuses it with the decoding result from the sub-decoder to obtain a noise prediction result with the same number of channels as the input noise image.

[0065] Specifically, the network models currently used to generate microstructures of porous media are mostly conditional deep convolutional generative adversarial networks (cDCGANs), which generally consist of a generator G and a discriminator D.

[0066] See Figure 2 The generator G has two inputs: random noise z, used to generate arbitrary images, and label information keff, used to control the effective thermal conductivity of the generated images. Since both random noise z and label information keff are one-dimensional vectors, and the generator's output is a two-dimensional porous medium image with the same resolution as the samples in the training set, the generator G first converts the input information z and keff into 4×4 two-dimensional information with 256 channels through transposed convolution. Then, it concatenates the two, and through continuous upsampling, increases the spatial resolution of the feature map while reducing the number of channels, ultimately converting the feature map into a porous medium image with 1 channel and a resolution of 128×128, embedding the effective thermal conductivity information. See also... Figure 3The core task of the discriminator D is to distinguish the authenticity of the input data, that is, to determine whether the data comes from a real data distribution or a data distribution created by the generator. Within the cDCGGAN framework, the discriminator D not only processes image data but also considers related conditional information. Therefore, the input to the cDCGAN discriminator consists of two parts: the input image data and the label information keff of the input image data. The discriminator D concatenates the label information keff with the input image data x as the input to subsequent convolutional layers. Each subsequent convolutional operation generates a set of feature maps, which capture different aspects of the input data. As the network deepens, the feature maps gradually transform from low-level features to high-level features. Finally, the high-level features are pooled and mapped through a fully connected layer to a probability value representing the likelihood that the input data is real data.

[0067] In the training process of a Conditional Generative Adversarial Network (cDCGAN), there is an adversarial relationship between the discriminator D and the generator G. The generator G attempts to generate increasingly realistic images, while the discriminator D strives to improve its discrimination ability. However, this generation principle has certain limitations. When the generator G's goal is to generate images that fool the discriminator D, it may learn to generate only very similar images, ignoring the diversity of the data. Secondly, the generation effect of cDCGAN heavily depends on the quality of the discriminator D. When the discriminator D performs poorly, the generated data will differ significantly from the real data.

[0068] Conditional diffusion model is a generative model based on noise diffusion and denoising processes. Its training and sample generation process is as follows: Figure 4 As shown. The core idea of ​​this model is to start from a known state (usually random noise) and gradually introduce structural information into the data until samples that match the target data distribution are generated. The training process of the conditional diffusion model mainly involves recording the noise ε added to the training set images. t As labels, then the noisy data x t The time step t, which describes the degree of noise diffusion, and the image label information are concatenated and used as the input to the denoising neural network. Then, the output of the denoising network is minimized. With the recorded noise ε t MSELoss, trained on the ability of a denoising neural network to fit noise, is crucial. Therefore, the quality of images generated by the conditional diffusion model heavily depends on the architecture of the denoising neural network and its ability to fit noise.

[0069] Currently, most denoising neural network architectures using conditional diffusion models employ simple encoder-decoder structures. While these structures are simple and stable, their encoding and decoding depths are fixed, leading to significant performance variations across different datasets. Therefore, this implementation proposes a denoising neural network with a dual-upsampled-channel decoder as the denoising neural network for the conditional diffusion model.

[0070] See Figure 5 The denoising neural network with dual upsampling channel decoders mainly consists of four modules:

[0071] Conditional encoding module: By describing the input image x t The time step t of noise diffusion and the embedded effective thermal conductivity keff are related to the x-axis. t For input data x t The information is supplemented to enable the network to conditionally generate images. The encoder is used to compress the embedded data image and extract more diverse data features by increasing the number of channels in the feature map, resulting in a high-level feature map with significantly reduced spatial resolution. This high-level feature map is output to the sub-decoder and main decoder based on the number of downsampling operations. The sub-decoder restores the high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image and gradually reduces the number of channels in the high-resolution feature map, obtaining the decoding result of the sub-decoder. The main decoder restores the lower-resolution high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image, and then fuses it with the decoding result of the sub-decoder to finally obtain a noise prediction result with the same resolution and number of channels as the input noise image.

[0072] In one possible implementation, the pre-training of the improved conditional diffusion model includes: acquiring microstructure images of porous materials with different pore sizes, different porosities, different porosity distributions, and different pore morphologies; calculating the effective thermal conductivity of the microstructure images of the porous materials; and training the improved conditional diffusion model based on the microstructure images of the porous materials and their effective thermal conductivity.

[0073] Optionally, obtaining microstructure images of porous materials with different porosities, different porosity distributions, and different pore morphologies includes: generating microstructure images of porous materials with different porosities, different porosity distributions, and different pore morphologies based on the gradient QSGS algorithm.

[0074] Specifically, the diversity and realism of the training dataset have a crucial impact on improving the training and performance of the conditional diffusion model. Diversity ensures that the model can learn a wide range of features and patterns in the dataset, resulting in generated samples with rich variation and high generalization ability. Based on the gradient QSGS algorithm, this implementation generated 12,000 gradient porous ceramic structures with a resolution of 128×128 and a porosity uniformly distributed between 1% and 35%. See [link to documentation] Figure 6 Its pore morphology mainly includes layered pores, columnar pores, and other morphologies. The pore structure also has an arbitrary porosity gradient distribution, meaning that the pore distribution is not uniform in the same image, which makes the data in the training dataset closer to the real data.

[0075] In one possible implementation, the calculation of the effective thermal conductivity of the microstructure image of the porous material includes: solving for the effective thermal conductivity of the microstructure image of the porous material using the finite element method.

[0076] Optionally, the effective thermal conductivity obtained by using the finite element method to solve the microstructure image of porous materials includes:

[0077] Define the computational boundaries for the microstructure image of the porous material; where the left and right boundaries are adiabatic boundaries.

[0078] Based on the thermal conductivity of the fixed material phase and the thermal conductivity of the porous phase, the vertical heat flux of each node in the microstructure image of the porous material is calculated.

[0079] The effective thermal conductivity (ETC) of the porous material microstructure image is obtained by the following formula:

[0080]

[0081] Where q is the vertical heat flux of the node; W is the width of the microstructure image of the porous material; ΔT is the temperature difference between the upper and lower boundaries of the microstructure image of the porous material; and H is the height of the microstructure image of the porous material.

[0082] Specifically, using accurate effective thermal conductivity as a label for porous material media is significant for improving the performance of models predicting effective thermal conductivity. The effective thermal conductivity of porous material structures generated by gradient QSGS is solved using the finite element method. It is worth noting that pixel information can be directly converted into mesh information. For example, black pixels represent pore nodes, and white pixels represent matrix nodes. By applying... Figure 7 The boundary conditions shown indicate that the finite element method for solving the effective thermal conductivity of porous material structures involves the following specific steps:

[0083] (1) Set the material properties (thermal conductivity values) of the element nodes. (2) Set the boundary conditions; specifically, the temperature difference between the upper and lower boundaries is ΔT, and the left and right boundaries are adiabatic boundaries. (3) Calculate the vertical heat flux of each node. (4) Calculate the effective thermal conductivity of the structure according to the following formula:

[0084] In this embodiment, 12,000 images of porous material microstructures with a size of 128×128 pixels and different pore sizes, pore morphologies, and gradient porosity distributions were generated based on the gradient QSGS algorithm and used as a training dataset. To enable the neural network to establish an accurate mapping from microstructural features to effective thermal conductivity, the thermal conductivity of the ceramic phase and the porous phase was fixed so that the effective thermal conductivity of different structures depended only on their structural features. Then, the effective thermal conductivity was calculated using the finite element method and used as labels to train the neural network. Finally, the training dataset was divided into three parts: a training set containing 10,000 images, a validation set containing 1,000 images, and a test set containing 1,000 images.

[0085] In one possible implementation, after training on a training set containing 12,000 images of the microstructure of porous materials, the optimal weight files of the improved conditional diffusion model and the cDCGAN model of the present invention are obtained. Based on the optimal weight files of different models, the microstructure of porous materials with a set effective thermal conductivity is reconstructed.

[0086] In this embodiment, porous media pore structures with set thermal conductivity within the range of (4.5, 10.0) W / (mK) were generated, and 100 microstructure images of the porous material were generated for each set effective thermal conductivity. The performance of different generation models was discussed and analyzed by calculating the relative error R between the set effective thermal conductivity of the generated pore structures and the actual effective thermal conductivity.

[0087] The average porosity of samples with a set effective thermal conductivity in the training set was calculated and compared with the average porosity of pore structures generated by different models. The results are as follows: Figure 8As shown, the average porosity of the pore structures generated by the model of this invention is highly consistent with the average porosity of the pore structures in the dataset, and their trends and values ​​with respect to effective thermal conductivity are very close. This fully demonstrates that the model of this invention has learned the features of the sample data in the training set well and correctly established the mapping relationship between these features and effective thermal conductivity. In contrast, when the effective thermal conductivity of the cDCGAN model is set to be high, it implies a simpler pore morphology in this work, resulting in high consistency with the average porosity of the samples in the dataset. However, when the effective thermal conductivity is set to be low, the average porosity of the generated structures differs significantly from that of the dataset, indicating that the cDCGAN model has not fully learned the mapping relationship between the pore structures and their effective thermal conductivity in the dataset.

[0088] To more specifically compare the variation trend of generation accuracy of different models with the set thermal conductivity, this embodiment calculates the average value of the relative error between the true effective thermal conductivity and the set thermal conductivity for 100 generated microstructure images of porous materials under each set thermal conductivity. Specific results are as follows: Figures 9 to 10 As shown, it is clear that the improved conditional diffusion model of this invention is far superior to the cDCGAN model in terms of accuracy and stability in generating porous ceramic pore structures with a set thermal conductivity.

[0089] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0090] See Figure 11 In another embodiment of the present invention, a microstructure generation system for porous materials with a set effective thermal conductivity is provided, which can be used to implement the above-mentioned method for generating microstructures of porous materials with a set effective thermal conductivity. Specifically, the microstructure generation system for porous materials with a set effective thermal conductivity includes a data acquisition module and a structure generation module. The data acquisition module is used to acquire a random noise image and a set effective thermal conductivity; the structure generation module is used to generate a microstructure image of porous materials with a set effective thermal conductivity based on the random noise image and the set effective thermal conductivity, using a pre-trained improved conditional diffusion model; wherein the improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model to a denoising neural network with a dual upsampling channel decoder.

[0091] In one possible implementation, the denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a master decoder.

[0092] The conditional encoding module embeds a time step describing the noise diffusion level of a random noise image and a set effective thermal conductivity into the random noise image to obtain an embedded data image, which is then output to the encoder. The encoder compresses the embedded data image and extracts data features from it by increasing the number of channels in the feature map, resulting in a high-level feature map. Based on the number of downsampling operations, it outputs a high-resolution feature map to the sub-decoder and a low-resolution feature map to the main decoder. The sub-decoder restores the high-resolution feature map obtained from the encoder to the same spatial resolution as the input random noise image and gradually reduces the number of channels in the high-level feature map, sending the decoding result to the main decoder. The main decoder restores the low-resolution feature map obtained from the encoder to the same spatial resolution as the input random noise image and then fuses it with the decoding result from the sub-decoder to obtain a noise prediction result with the same number of channels as the input noise image.

[0093] All relevant content of each step involved in the aforementioned embodiments of the method for generating porous material microstructures with set effective thermal conductivity can be referenced to the functional description of the corresponding functional module of the porous material microstructure generation system with set effective thermal conductivity in the embodiments of the present invention, and will not be repeated here.

[0094] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0095] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used to operate a method for generating microstructures of porous materials with set effective thermal conductivity.

[0096] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the method for generating porous material microstructures with set effective thermal conductivity in the above embodiments.

[0097] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0098] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0099] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0100] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for generating a porous material microstructure with a predetermined effective thermal conductivity, characterized in that, include: Acquire random noise images and set effective thermal conductivity; Based on random noise images and a set effective thermal conductivity, a microstructure image of a porous material with a set effective thermal conductivity is generated using a pre-trained improved conditional diffusion model. Among them, the improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual upsampling channel decoder; The denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a main decoder; The conditional coding module is used to embed the time step describing the noise diffusion level of the random noise image and the set effective thermal conductivity into the random noise image to obtain the embedded data image and output it to the encoder; The encoder is used to compress the embedded data image and extract the data features of the embedded data image by increasing the number of channels of the feature map to obtain a high-level feature map. Based on the number of downsampling, it outputs a high-resolution high-level feature map to the sub-decoder and a high-resolution high-level feature map to the main decoder. The secondary decoder is used to restore the high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image, and gradually reduce the number of channels of the high-resolution feature map to obtain the decoding result of the secondary decoder, which is then sent to the main decoder. The main decoder is used to restore the high-level feature map with a lower resolution obtained by the encoder to the same spatial resolution as the input random noise image. Then, it is fused with the decoding result of the secondary decoder to obtain a noise prediction result with the same number of channels as the input noise image.

2. The method for generating a porous material microstructure with a set effective thermal conductivity according to claim 1, characterized in that, The pre-training of the improved conditional diffusion model includes: Acquire microstructure images of porous materials with different pore sizes, porosities, porosity distributions, and pore morphologies; Calculate the effective thermal conductivity of microstructure images of porous materials; An improved conditional diffusion model is trained based on the microstructure images of porous materials and their effective thermal conductivity.

3. The method for generating a porous material microstructure with a set effective thermal conductivity according to claim 2, characterized in that, The acquisition of microstructure images of porous materials with different pore sizes, porosities, porosity distributions, and pore morphologies includes: Based on the gradient QSGS algorithm, microstructure images of porous materials with different porosities, different porosity distributions, and different pore morphologies are generated.

4. The method for generating a porous material microstructure with a set effective thermal conductivity according to claim 3, characterized in that, The effective thermal conductivity of the microstructure image of porous materials includes: The effective thermal conductivity of porous materials is determined using the finite element method.

5. The method for generating a porous material microstructure with a set effective thermal conductivity according to claim 4, characterized in that, The effective thermal conductivity obtained by using the finite element method to solve the microstructure image of porous materials includes: Define the computational boundaries for the microstructure image of the porous material; where the left and right boundaries are adiabatic boundaries. Based on the thermal conductivity of the fixed material phase and the thermal conductivity of the porous phase, the vertical heat flux of each node in the microstructure image of the porous material is calculated. The effective thermal conductivity of the porous material microstructure image is obtained by the following formula. : in, q It is the vertical heat flux of the node; W It is the width of the microstructure image of the porous material; It is the temperature difference between the upper and lower boundaries of the microstructure image of the porous material; H It is the height of the microstructure image of the porous material.

6. A microstructure generation system for porous materials with a set effective thermal conductivity, characterized in that, include: The data acquisition module is used to acquire random noise images and set the effective thermal conductivity. The structure generation module is used to generate microstructure images of porous materials with a set effective thermal conductivity based on random noise images and a pre-trained improved conditional diffusion model. Among them, the improved conditional diffusion model is obtained by modifying the denoising neural network in the conditional diffusion model into a denoising neural network with a dual upsampling channel decoder; The denoising neural network with dual upsampling channel decoders includes: a conditional coding module, an encoder, a sub-decoder, and a main decoder; The conditional coding module is used to embed the time step describing the noise diffusion level of the random noise image and the set effective thermal conductivity into the random noise image to obtain the embedded data image and output it to the encoder; The encoder is used to compress the embedded data image and extract the data features of the embedded data image by increasing the number of channels of the feature map to obtain a high-level feature map. Based on the number of downsampling, it outputs a high-resolution high-level feature map to the sub-decoder and a high-resolution high-level feature map to the main decoder. The secondary decoder is used to restore the high-resolution feature map obtained by the encoder to the same spatial resolution as the input random noise image, and gradually reduce the number of channels of the high-resolution feature map to obtain the decoding result of the secondary decoder, which is then sent to the main decoder. The main decoder is used to restore the high-level feature map with a lower resolution obtained by the encoder to the same spatial resolution as the input random noise image. Then, it is fused with the decoding result of the secondary decoder to obtain a noise prediction result with the same number of channels as the input noise image.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating a porous material microstructure with a set effective thermal conductivity as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for generating a porous material microstructure with a set effective thermal conductivity as described in any one of claims 1 to 5.

Images