Carbon ceramic resistor microstructure optimization method and system based on convolutional neural network

Abstract

The invention belongs to the field of functional ceramic material structure optimization design, and particularly discloses a carbon ceramic resistor microstructure optimization method and system based on a convolutional neural network. According to the method and the device, the nonlinear mapping relation between the microstructure parameters and the macroscopic performance response is established by utilizing deep learning, so that insufficient empirical performance and trial-and-error processes in a traditional design method are avoided, and resource waste possibly caused by a large number of trial-and-errors is reduced. Meanwhile, a brand new and high-performance microstructure combination can be introduced, an innovative solution is provided for optimization design, and the performance of the carbon ceramic resistor is comprehensively improved. The application of the method not only provides a brand new data driving solution for the microstructure optimization design of the carbon ceramic resistor, but also can provide reference for the performance optimization design of other complex materials, and has wide application prospects.

Description

Technical Field

[0001] The present application belongs to the field of structural optimization design of functional ceramic materials, and more specifically, to a method and system for optimizing the microstructure of carbon ceramic resistors based on convolutional neural networks. Background Art

[0002] Carbon ceramic resistors, as key basic components for high-power energy dissipation, are widely used in high-voltage circuit breakers, flexible direct current transmission systems, new energy grid-connected systems and special electromagnetic equipment, directly affecting the safe and stable operation of the system.

[0003] Carbon ceramic resistors are made by sintering carbon black, clay and ceramic aggregates at high temperatures to form a complex microstructure system including carbon chains, ceramic matrix and pores. Studies have shown that the uniformity, continuity and interface characteristics of the microstructure have a decisive influence on its energy tolerance. Therefore, regulating carbon ceramic resistors at the microstructural level is a key way to achieve a breakthrough in the performance of domestic resistors. However, there is a strong nonlinear and multi-scale coupling relationship between microstructure and performance. Currently, there is a lack of systematic modeling theory and optimization mechanism, and microstructure design still relies on empirical trial and error. This type of method not only has a long preparation cycle and high experimental cost, but also lacks the ability to fine-tune control, making it difficult to effectively identify key microstructural features that affect performance, which seriously restricts the efficiency of material development and optimization effect. Summary of the Invention

[0004] In response to the defects of the existing technology, the purpose of this application is to provide a carbon ceramic resistor microstructure optimization method and system based on convolutional neural networks, aiming to solve the lack of experience and trial and error process in traditional design methods, and the problem of waste of resources caused by a large amount of trial and error.

[0005] To solve this problem, in the first aspect, the present application provides a carbon ceramic resistor microstructure optimization method based on a convolutional neural network, comprising: obtaining the electro-thermal-mechanical performance requirements of the carbon ceramic resistor and a trained convolutional neural network; using the trained convolutional neural network to reversely deduce the electro-thermal-mechanical performance requirements of the carbon ceramic resistor to obtain microstructure characteristics that meet the performance requirements; wherein the trained convolutional neural network is obtained by: obtaining a training sample, the training sample comprising a microstructure distribution characteristic image of the carbon ceramic resistor and a corresponding macroscopic performance label, the microstructure distribution characteristic image consisting of a binary image containing a carbon chain network, a binary image containing a ceramic phase, and a binary image containing pores, and the macroscopic performance label is a score of comprehensive macroscopic electrical, thermal, and mechanical properties; using the training sample, the initial convolutional neural network is trained, and during the training process, the loss function value generated by the training is determined based on the macroscopic performance index until the loss function value meets the preset conditions, thereby obtaining a trained convolutional neural network.

[0006] Preferably, the microstructure distribution characteristic image is obtained in the following manner: using the carbon element scanning results in the elemental analysis image of the carbon ceramic resistor sample, extracting the distribution characteristics of the carbon chain network from the microstructure image of the carbon ceramic resistor sample, generating a binary image, marking the carbon chain area as 1 and other areas as 0; setting a grayscale value threshold for the microstructure image of the carbon ceramic resistor sample, separating the ceramic phase and pores in the image according to the grayscale value of the image, extracting the ceramic phase distribution and the pore distribution respectively, and generating a binary image.

[0007] Preferably, the microstructure image of the carbon ceramic resistor sample is preprocessed as follows: filtering technology is used to remove noise in the image to ensure a clearer image; image enhancement methods are used to improve the contrast and brightness of the image to highlight the microstructure distribution characteristics, including the distribution characteristics of the carbon chain network, ceramic phase and pores.

[0008] Preferably, the allowable injection energy and resistance change rate of the carbon ceramic resistor are obtained as electrical properties through electrical performance testing, the temperature rise of the resistor under energy injection is obtained as thermal properties through thermal performance testing, and the mechanical property is the fatigue strength of the resistor under multiple impacts.

[0009] Preferably, the score of the comprehensive macroscopic electrical, thermal and mechanical properties is obtained by: standardizing the same type of macroscopic properties of all carbon ceramic resistor samples; for each carbon ceramic resistor sample, weighting the standardized macroscopic properties according to the preset weights of the three macroscopic properties to obtain a score of the comprehensive macroscopic properties.

[0010] Preferably, the convolutional neural network includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the input layer receives a three-channel image of the distribution characteristics of the carbon chain network, ceramic phase and pores; the convolution layer scans the input image and gradually extracts local features; each convolution layer is followed by a pooling layer, and the pooling layer selects maximum pooling to reduce the size of the feature map through downsampling operations; after all convolution layers and pooling layers, the extracted local features are integrated through the fully connected layer and mapped to the target performance output; the output layer generates a comprehensive performance prediction result based on the output of the fully connected layer.

[0011] It should be noted that the CNN-based training and optimization process has strong generalization capabilities and can meet the performance requirements in different application scenarios.

[0012] Preferably, the loss function of the convolutional neural network includes a data-driven loss term and a physical constraint term; the data-driven loss term is used to measure the error between the output data predicted by the optimization model and the data obtained by the experiment; the physical constraint term is used to measure the deviation between the data predicted by the optimization model and the data obtained from the physical relationship between current density and temperature gradient.

[0013] It should be noted that the introduction of a physical constraint optimization strategy guides model learning to conform to physical laws, while also improving the model's convergence and stability, ensuring the reliability of the design. Combining deep learning with physical constraints can effectively extract high-dimensional structural features with limited samples.

[0014] Preferably, after each round of training, the results of the physical model are compared with the experimental data, and the weights of the network model hyperparameters and physical constraint terms are adjusted according to the loss function value.

[0015] Preferably, the electro-thermal-mechanical performance requirements of the carbon ceramic resistor are converted into a comprehensive macroscopic electrical, thermal and mechanical performance score, which is then input into the trained convolutional neural network.

[0016] In a second aspect, the present application provides a carbon ceramic resistor microstructure optimization system based on a convolutional neural network, comprising: at least one memory for storing programs; and at least one processor for entering the program stored in the memory. When the program stored in the memory is entered, the processor is used to enter the optimization method as described in the first aspect.

[0017] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0018] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the existing technologies: This application proposes a method for optimizing the microstructure of carbon ceramic resistors based on convolutional neural networks. It uses deep learning to establish a nonlinear mapping relationship between microstructure parameters and macroscopic performance responses, avoiding the lack of experience and trial-and-error process in traditional design methods, and reducing the waste of resources that may be caused by a large amount of trial and error. At the same time, this method can introduce a new, high-performance microstructure combination, provide innovative solutions for optimization design, and comprehensively improve the performance of carbon ceramic resistors. The application of this method not only provides a new data-driven solution for the microstructure optimization design of carbon ceramic resistors, but also provides a reference for the performance optimization design of other complex materials, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of the carbon ceramic resistor microstructure optimization method based on convolutional neural network provided in an embodiment of the present application.

[0020] Figure 2 This is a schematic diagram of the CNN network model recognition input layer provided in an embodiment of the present application.

[0021] Figure 3 This is a framework diagram of a CNN model based on physical prior knowledge provided in an embodiment of the present application.

[0022] Figure 4 It is an optimization path indication diagram of the CNN model based on physical prior knowledge provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0024] The term "and / or" in this application describes an association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, or B exists alone. The symbol " / " in this application indicates that the associated objects are in an "or" relationship, for example, A / B means A or B.

[0025] In this application, the terms "first" and "second" are used to distinguish different objects, rather than to describe a specific order of objects. For example, the terms "first response message" and "second response message" are used to distinguish different response messages, rather than to describe a specific order of response messages.

[0026] The term "electrical connection" in this application can be a direct circuit connection or signal transmission through a communication protocol.

[0027] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0028] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more, for example, multiple processing units means two or more processing units, etc.; multiple elements means two or more elements, etc.

[0029] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0030] The present application provides a method for optimizing the microstructure of a carbon ceramic resistor based on a convolutional neural network, comprising: obtaining the electro-thermal-mechanical performance requirements of the carbon ceramic resistor and a trained convolutional neural network; using the trained convolutional neural network to reversely deduce the electro-thermal-mechanical performance requirements of the carbon ceramic resistor to obtain microstructure characteristics that meet the performance requirements.

[0031] The trained convolutional neural network is obtained by: obtaining training samples, wherein the training samples include a microstructure distribution characteristic image of a carbon ceramic resistor and a corresponding macroscopic performance label, wherein the microstructure distribution characteristic image is composed of a binary image containing a carbon chain network, a binary image containing a ceramic phase, and a binary image containing pores, and the macroscopic performance label is a score of comprehensive macroscopic electrical, thermal, and mechanical properties; using the training samples, the initial convolutional neural network is trained, and during the training process, the loss function value generated by the training is determined based on the macroscopic performance index until the loss function value meets the preset conditions, thereby obtaining the trained convolutional neural network.

[0032] Preferably, the microstructure distribution characteristic image is obtained in the following manner: using the carbon element scanning results in the elemental analysis image of the carbon ceramic resistor sample, extracting the distribution characteristics of the carbon chain network from the microstructure image of the carbon ceramic resistor sample, generating a binary image, marking the carbon chain area as 1 and other areas as 0; setting a grayscale value threshold for the microstructure image of the carbon ceramic resistor sample, separating the ceramic phase and pores in the image according to the grayscale value of the image, extracting the ceramic phase distribution and the pore distribution respectively, and generating a binary image.

[0033] Preferably, the microstructure image of the carbon ceramic resistor sample is preprocessed as follows: filtering technology is used to remove noise in the image to ensure a clearer image; image enhancement methods are used to improve the contrast and brightness of the image to highlight the microstructure distribution characteristics, including the distribution characteristics of the carbon chain network, ceramic phase and pores.

[0034] Preferably, the allowable injection energy and resistance change rate of the carbon ceramic resistor are obtained as electrical properties through electrical performance testing, the temperature rise of the resistor under energy injection is obtained as thermal properties through thermal performance testing, and the mechanical property is the fatigue strength of the resistor under multiple impacts.

[0035] Preferably, the score of the comprehensive macroscopic electrical, thermal and mechanical properties is obtained by: standardizing the same type of macroscopic properties of all carbon ceramic resistor samples; for each carbon ceramic resistor sample, weighting the standardized macroscopic properties according to the preset weights of the three macroscopic properties to obtain a score of the comprehensive macroscopic properties.

[0036] Preferably, the convolutional neural network includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the input layer receives a three-channel image of the distribution characteristics of the carbon chain network, ceramic phase and pores; the convolution layer scans the input image and gradually extracts local features; each convolution layer is followed by a pooling layer, and the pooling layer selects maximum pooling to reduce the size of the feature map through downsampling operations; after all convolution layers and pooling layers, the extracted local features are integrated through the fully connected layer and mapped to the target performance output; the output layer generates a comprehensive performance prediction result based on the output of the fully connected layer.

[0037] Preferably, the loss function of the convolutional neural network includes a data-driven loss term and a physical constraint term; the data-driven loss term is used to measure the error between the output data predicted by the optimization model and the data obtained by the experiment; the physical constraint term is used to measure the deviation between the data predicted by the optimization model and the data obtained from the physical relationship between current density and temperature gradient.

[0038] Preferably, after each round of training, the results of the physical model are compared with the experimental data, and the weights of the network model hyperparameters and physical constraint terms are adjusted according to the loss function value.

[0039] Preferably, the electro-thermal-mechanical performance requirements of the carbon ceramic resistor are converted into a comprehensive macroscopic electrical, thermal and mechanical performance score, which is then input into the trained convolutional neural network.

[0040] Example In this embodiment, the convolutional neural network is a convolutional neural network model CNN. Figure 1 As shown, this embodiment provides a carbon ceramic resistor microstructure optimization method based on a convolutional neural network, which specifically includes the following steps: S1. Obtain microstructure data of carbon ceramic resistors.

[0041] 100 batches of commercial carbon ceramic resistors were selected as training samples, with resistance values ​​ranging from 5 to 20 Ω and volumes ranging from 40 to 430 cm 3 The maximum allowable injection energy is 250~700J / cm 3 , and there are no obvious defects on the surface.

[0042] First, physical and chemical testing and analysis were performed on the carbon ceramic resistor training samples. These tests included: elemental analysis using energy dispersive X-ray spectroscopy (EDS) to obtain elemental content information in the training samples; component analysis using X-ray photoelectron spectroscopy (XPS) to obtain composition information of the training samples; and microstructural images of the training samples using scanning electron microscopy (SEM), combined with X-ray diffraction phase analysis (XRD) to determine the average particle size of the ceramic grains.

[0043] In addition, a series of experimental tests were conducted on the carbon ceramic resistor training samples. The test indicators included electrothermal performance indicators such as allowable injection energy, resistance change rate, temperature rise, and fatigue strength. The macro test indicators of carbon ceramic resistors are divided into electrical properties, thermal properties, and mechanical properties. The electrical properties mainly include allowable injection energy and resistance change rate, the thermal properties mainly refer to the temperature rise under energy injection, and the mechanical properties mainly refer to the fatigue strength of the resistor after multiple energy injections. Among them, the allowable injection energy refers to the measurement of the carbon ceramic resistor sample's ability to withstand high pulse energy injection. The resistance change rate refers to the change in the resistance of the carbon ceramic resistor sample under high pulse energy injection. Temperature rise refers to the temperature change of the resistance of the carbon ceramic resistor sample under different high pulse energy injections. Fatigue strength refers to the ability of the carbon ceramic resistor sample to withstand fatigue damage after multiple energy injections, and tests the mechanical properties of the resistor under repeated loads.

[0044] S2. Microstructural feature encoding.

[0045] First, the microstructure images obtained through physical and chemical analysis are subjected to a series of preprocessing operations. These preprocessing operations include, but are not limited to: using filtering techniques to remove noise from the image to ensure clearer image data; using image enhancement methods to increase image contrast and brightness, highlighting the distribution characteristics of the microstructure, including the distribution characteristics of the carbon chain network, ceramic skeleton, and pores, making the distribution characteristics easier for neural network recognition.

[0046] Image information coding technology is used to extract the distribution characteristics of the pre-processed carbon ceramic resistor microstructure image. Figure 2 As shown, by performing image analysis on the surface scanning results of the carbon element in the EDS, the distribution characteristics of the carbon chain network in the SEM image are extracted, and a binary image is generated, with the carbon chain area marked as 1 and the other areas as 0. A grayscale value threshold is set for the microstructure image of the SEM, and the ceramic phase and pores in the image are separated according to the grayscale value of the image. The ceramic phase distribution and the pore distribution are extracted respectively, and a binary image is generated. The three-layer binary image containing the carbon chain, ceramics and pores is used as the input of the convolutional neural network model. Optionally, the extracted microstructure distribution characteristics are converted into a tensor matrix data format suitable for the neural network and used as input to the convolutional neural network for training.

[0047] The macroscopic performance of carbon ceramic resistor samples was tested through a series of experiments. Since the units of different performances are different, the data must be standardized first. Specifically, the macroscopic performance indicators of the same type are normalized so that all macroscopic performance data of all samples are converted to between 0 and 1. According to the requirements of the actual application conditions of carbon ceramic resistors, weights are assigned to the three performance indicators to establish a comprehensive macroscopic performance evaluation system. In this embodiment, the electrical performance of the carbon ceramic resistor accounts for 70%, the thermal performance accounts for 20%, and the mechanical performance accounts for 10%. Based on this, the macroscopic performance indicators are weighted and summed, and each carbon ceramic resistor sample will have a corresponding comprehensive macroscopic performance score as a label for its microstructure image, which together constitute the training sample set of the model.

[0048] In this embodiment, the number of training samples for the convolutional neural network is 100, and the generated training samples are divided into a training set and a test set in a ratio of 8:2.

[0049] S3. Construct a CNN model based on physical prior knowledge.

[0050] Define the network architecture of the CNN model, including input layer, convolution layer, pooling layer, fully connected layer and output layer. Figure 3 As shown, the input layer receives preprocessed microstructure image data, specifically a three-channel image containing the distribution characteristics of carbon chains, ceramic phases, and pores. Each channel represents an important phase in the carbon ceramic resistor microstructure and serves as input to the CNN model, ensuring that the network can learn the multidimensional characteristics of the microstructure. This embodiment uses four convolutional layers to extract local features from the microstructure image. The convolutional layers use a 2×2 convolution kernel, scan the input image, and gradually extract local features. Each convolutional layer is followed by a pooling layer to reduce the spatial dimension of the feature map to avoid overfitting. The pooling layer uses max pooling, which downsampling to reduce the size of the feature map and improve generalization. After all convolutional and pooling layers, a fully connected layer integrates the extracted local features and maps them to the target performance output. The fully connected layer converts the low-level features extracted by the convolutional and pooling layers into high-level features, achieving image-to-performance mapping. The final output layer generates a comprehensive performance prediction based on the output of the fully connected layer. The overall network structure consists of 10 layers and is suitable for a CNN model with a three-channel input image.

[0051] Preferably, a physical constraint term is added to the loss function of the network model to construct a CNN model based on physical priors. The loss function in the CNN model includes a data-driven loss term and a physical constraint term. The data-driven loss term is used to measure the error between the model predicted value Y1 and the actual experimental data Y2. In this embodiment, it is represented by the mean square error (MSE). The physical constraint term is based on the physical law of the resistor thermal conduction model and is used to constrain the relationship between the current density and temperature gradient of the resistor. The learning process of the network is adjusted by calculating the deviation between the predicted current density and temperature gradient and the physical model. The specific physical expression is as follows:

[0052] Where q is the heat density per unit volume (W / m 3 ), J is the current density (A / m 2 ), ρ is the resistivity (Ω·m), k is the thermal conductivity coefficient of the resistor (W / (m·K)), and dT / dx is the temperature gradient (K / m).

[0053] The total loss function combines the data-driven loss term and the physical constraint term, and the formula is as follows:

[0054] in, is the data-driven loss term, is a physical constraint, is the weight of the data-driven loss term, is the weight of the physical constraint term.

[0055] In the case of a small sample data set, insufficient data volume can cause the CNN network model to easily fall into the risk of local optimal solutions or overfitting. The physical constraint term includes a resistance heat conduction model, which provides additional physical prior knowledge, and can help the network model follow the physical laws during training and avoid blind optimization. Specifically, the network is guided by the physical loss term in the early stage, using a lower learning rate to ensure that the features learned by the model are consistent with the physical laws; in the later stage, the learning rate is gradually increased to help the model converge quickly. Figure 4 Compared with traditional purely data-driven methods, physical constraints reduce the network’s degrees of freedom and optimization space, enabling the network to quickly locate solutions that conform to physical laws and improve the model’s generalization ability.

[0056] During model training, the training set is used for training, and the test set is used to evaluate the training results. After each round of training, the results of the physical model are compared with the experimental data, and the weights of the network model hyperparameters and physical constraints are adjusted according to the loss function. Ultimately, a CNN model based on physical priors is obtained, which can efficiently optimize the microstructure design of carbon ceramic resistors under small sample conditions and meet the constraints of physical laws.

[0057] S4. Optimal microstructure configuration design.

[0058] Based on the requirements of carbon ceramic resistors in different application scenarios, for example, high-voltage circuit breaker closing resistors are required to have high allowable injection energy and excellent heat dissipation capabilities, their macro-performance is evaluated through a comprehensive macro-performance evaluation system and corresponding performance labels are obtained.

[0059] Based on the optimized physical prior CNN model, the microstructure characteristic parameters that meet the target performance requirements are reversely derived, thereby obtaining the optimal microstructure design. Utilizing this optimization model, carbon ceramic resistor microstructure design solutions suitable for different application scenarios can be screened, providing theoretical support and design basis for the performance optimization of carbon ceramic resistors.

[0060] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0061] Based on the methods in the above embodiments, embodiments of the present application provide an electronic device that may include: a processor, a communications interface, a memory, and a communication bus, wherein the processor, the communications interface, and the memory communicate with each other via the communication bus. The processor may invoke logic instructions in the memory to execute the methods in the above embodiments.

[0062] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0063] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0064] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0065] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0066] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC.

[0067] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions. When loaded and executed on a computer, the computer program instructions fully or partially produce the processes or functions described in the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)).

[0068] It will be understood that the various numerical numbers involved in the embodiments of the present application are merely distinctions for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0069] It is easy for those skilled in the art to understand that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A carbon ceramic resistor microstructure optimization method based on convolutional neural network, characterized in that: include: Obtain the electro-thermal-mechanical performance requirements of carbon ceramic resistors and the trained convolutional neural network; Using the trained convolutional neural network, the electro-thermal-mechanical performance requirements of carbon ceramic resistors were reversely deduced to obtain the microstructural characteristics that meet the performance requirements. Among them, the trained convolutional neural network is obtained by the following method: Obtaining training samples, the training samples comprising microstructural distribution characteristic images of carbon ceramic resistors and corresponding macroscopic performance labels, the microstructural distribution characteristic images comprising a binary image containing a carbon chain network, a binary image containing a ceramic phase, and a binary image containing pores, and the macroscopic performance labels being scores that comprehensively represent macroscopic electrical, thermal, and mechanical properties; The initial convolutional neural network is trained using the training samples. During the training process, the loss function value generated by the training is determined based on the macro performance index until the loss function value meets the preset conditions, thereby obtaining a trained convolutional neural network.

2. The optimization method according to claim 1, wherein: The microstructure distribution characteristic image is obtained by the following method: Using the carbon element scanning results in the elemental analysis image of the carbon ceramic resistor sample, the distribution characteristics of the carbon chain network are extracted from the microstructure image of the carbon ceramic resistor sample, and a binary image is generated, in which the carbon chain area is marked as 1 and the other areas are marked as 0; A grayscale threshold is set for the microstructure image of the carbon ceramic resistor sample, and the ceramic phase and pores in the image are separated according to the grayscale value of the image. The ceramic phase distribution and pore distribution are extracted respectively, and a binary image is generated.

3. The optimization method according to claim 2, wherein: The microstructure images of the carbon ceramic resistor samples were preprocessed as follows: filtering technology was used to remove noise in the image to ensure a clearer image; image enhancement methods were used to improve the contrast and brightness of the image and highlight the microstructural distribution characteristics, including the distribution characteristics of the carbon chain network, ceramic phase, and pores.

4. The optimization method according to claim 1, wherein: The electrical performance test is used to obtain the allowable injection energy and resistance change rate of the carbon ceramic resistor as electrical properties. The thermal performance test is used to obtain the temperature rise of the resistor under energy injection as thermal properties. The mechanical property is the fatigue strength of the resistor under multiple impacts.

5. The optimization method according to claim 1, wherein: The comprehensive macroscopic electrical, thermal and mechanical performance scores are obtained as follows: The macroscopic properties of the same kind of all carbon ceramic resistor samples were standardized; For each carbon ceramic resistor sample, the standardized macro-performance is weighted according to the preset weights of the three macro-performances to obtain a comprehensive macro-performance score.

6. The optimization method according to claim 1, wherein: The convolutional neural network includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; The input layer receives a three-channel image of the distribution characteristics of the carbon chain network, the ceramic phase and the pores; The convolutional layer scans the input image and gradually extracts local features; Each convolutional layer is followed by a pooling layer. The pooling layer selects maximum pooling and reduces the size of the feature map through downsampling operation; After all convolutional and pooling layers, the extracted local features are integrated through the fully connected layer and mapped to the target performance output; The output layer generates a comprehensive performance prediction result based on the output of the fully connected layer.

7. The optimization method according to claim 1, wherein: The loss function of the convolutional neural network includes a data-driven loss term and a physical constraint term; The data-driven loss term is used to measure the error between the output data predicted by the optimization model and the data obtained by the experiment; The physical constraint term is used to measure the deviation between the data predicted by the optimization model and the data obtained from the physical relationship between current density and temperature gradient.

8. The optimization method according to claim 7, wherein: After each round of training, the results of the physical model are compared with the experimental data, and the weights of the network model hyperparameters and physical constraint items are adjusted according to the loss function value.

9. The optimization method according to claim 1, wherein: The electrical-thermal-mechanical performance requirements of carbon ceramic resistors are converted into comprehensive macroscopic electrical, thermal and mechanical performance scores, which are then input into the trained convolutional neural network.

10. A carbon ceramic resistor microstructure optimization system based on convolutional neural network, characterized in that: include: at least one memory for storing a program; At least one processor is configured to enter the program stored in the memory, and when the program stored in the memory is entered, the processor is configured to enter the optimization method according to any one of claims 1 to 9.

Images