Three-dimensional reconstruction method for porous Ni-YSZ electrode of solid oxide battery

Abstract

The invention provides a three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery. The three-dimensional reconstruction method comprises the following steps: S1, vacuum cold embedding; s2, argon ion polishing; s3, sample pretreatment and fixation; s4, FIB continuous slicing and In-lens and SE2 dual-channel image acquisition are carried out; and S5, carrying out deep learning analysis. The three-dimensional reconstruction method for the porous Ni-YSZ electrode of the solid oxide battery has the beneficial effects that accurate analysis of a multi-phase material structure is realized by integrating in-situ characterization and intelligent image analysis technologies, and the method has the advantages of high sample preparation stability, good three-dimensional reconstruction precision, high phase identification automation degree and the like. The method is especially suitable for microstructure evolution research of an electrode / electrolyte interface of a solid oxide fuel cell, and has an important value for revealing a material performance attenuation mechanism under high-temperature complex working conditions.

Description

Technical Field

[0001] The present invention relates to a battery, and particularly to a three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery. Background Art

[0002] As an efficient and reversible energy conversion device, the performance and lifespan of a solid oxide fuel cell / electrolyzer (SOFC / SOEC) critically depend on the microstructural characteristics of the electrode materials, especially the structural stability of the fuel electrode during the hydrogen production process by water electrolysis in the SOEC mode. The currently widely used Ni-YSZ cermet electrode faces severe challenges in a high-temperature steam electrolysis environment: the migration, agglomeration, and coarsening of Ni particles not only lead to a reduction in the length of the triple-phase boundaries (TPBs), but also exacerbate the thermal expansion mismatch with the YSZ electrolyte, thereby triggering a rapid decay in electrode performance. Although the FIB serial section tomography technique provides the possibility of three-dimensional microstructure reconstruction with nanoscale resolution, the application effect of this technique is severely limited by the acquisition quality of material contrast. Since the average atomic numbers of Ni and YSZ are close, traditional backscattered electron imaging is difficult to provide sufficient interphase contrast. Coupled with the intensity variations caused by differences in Ni grain orientations, image segmentation and three-dimensional reconstruction face major technical obstacles. In the prior art, although optimizing SEM imaging parameters can improve contrast to a certain extent, it still cannot meet the requirements of high-precision three-dimensional characterization of the Ni-YSZ system, which severely restricts the in-depth understanding of the evolution of the electrode microstructure and the mechanism of electrochemical performance decay, and becomes a key technical bottleneck in the development of long-life SOEC systems.

[0003] In the prior art, CN115830226A proposes a three-dimensional reconstruction method for porous media based on FIB-SEM sectioning and deep learning. Through gray-scale statistical analysis, binary section processing, and deep learning dimension expansion, high-precision three-dimensional pore structure modeling is achieved, and further combined with pore parameter calibration and regression analysis to predict thermal conductivity.

[0004] The FIB-SEM section reconstruction method proposed by CN115830226A is not restricted by the insufficient contrast of the material itself and is relatively easy to process. And it relies heavily on simulated reconstruction rather than experimental reconstruction. Therefore, the accuracy of its reconstructed structure lacks credibility.

[0005] In the prior art, CN114518378A proposes a FIB-SEM three-dimensional reconstruction method that enhances the conductivity of the sample through a metal conductive grid. The core steps include: depositing a Pt protective layer, assisting in extracting the sample with a nanomanipulator, and welding it to a metal conductive grid (copper / nickel / molybdenum grid) to form a three-sided conductive structure. This method conducts the accumulated charge through the conductive grid, significantly reducing the image drift caused by the charging effect and improving the three-dimensional reconstruction quality of non-conductive samples.

[0006] The FIB-SEM slicing and reconstruction method proposed in CN114518378A requires a nano-manipulator to assist in extracting samples. This operation is very complex and time-consuming, and the actual application efficiency is low. The method of discharging charges through a conductive grid is similar in principle to the "sputtering gold / carbon layer" or "ion beam-assisted conduction" schemes in the prior art, lacks necessity, and may introduce additional stress due to welding process problems.

[0007] In the prior art, CN114723878A proposed a three-dimensional reconstruction and correction method for FIB-SEM based on spherical particle marking. By attaching spherical particles to the material surface as a geometric calibration reference and combining epoxy resin pore filling and ion beam polishing, deformation correction of the reconstructed image is achieved. Its core innovation lies in using the known size and shape of spherical particles (of the same order of magnitude as the material microstructure) to correct image distortion caused by the ion beam inclination angle (54°) and SEM drift.

[0008] The FIB-SEM three-dimensional reconstruction and correction method proposed in CN114723878A is only applicable to the reconstruction of spherical particles, and has limited correction effect on planar or irregular structures (such as layered materials). Planar samples can correct the image sequence through known angular differences and alignment algorithms.

[0009] In summary, the conventional FIB-SEM reconstruction technology has the following core defects in the application of the Ni / YSZ system: First, due to the small difference in the average atomic numbers of Ni and YSZ, the backscattered electron imaging technology makes it impossible for the traditional gray threshold segmentation method to effectively distinguish the two phases; Second, the preferred orientation of Ni grains will cause fluctuations in the intensity of electron backscatter diffraction signals, resulting in pseudo-contrast within the same phase, seriously interfering with the reliability of the automatic segmentation algorithm; In addition, the existing anti-drift schemes (such as metal mesh welding or mechanical clamping) have extremely high operation complexity and the risk of damaging samples; At the same time, under the condition that high-contrast pictures cannot be obtained through complete experimental reconstruction, whether machine learning is used or not, a large number of slices need to be manually labeled. And due to the subjectivity of manual labeling, the cross-validation accuracy of the three-phase segmentation of pores / Ni / YSZ is often poor, and it is difficult to meet the requirements of sub-micron interface curvature analysis. These technical limitations lead to the inability to accurately correlate the microstructure evolution of the electrode with the attenuation of electrochemical performance, seriously restricting the research on the performance optimization and life prediction of SOEC electrode materials. Summary of the Invention

[0010] To solve the problems in the prior art, the present invention provides a three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery.

[0011] The present invention provides a three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery, comprising the following steps: S1, vacuum cold embedding; S2. Argon ion polishing; S3. Sample pretreatment and fixation; S4. FIB serial sectioning and In-lens and SE2 dual-channel image acquisition; S5. Deep learning analysis.

[0012] As a further improvement of the present invention, step S5 includes: training a U-net neural network model, verifying whether the accuracy reaches the expectation. If it does not reach the expectation, local manual correction is performed. If it reaches the expectation, the model is applied to all slices, and then 3D geometry is reconstructed, where the total verification accuracy > 85%, the verification accuracy of easily recognizable phases (YSZ / pore) > 90%, and the verification accuracy of difficult-to-recognize phases (Ni) > 80%.

[0013] As a further improvement of the present invention, step S1 includes: under vacuum, cold embedding the sample with epoxy resin.

[0014] As a further improvement of the present invention, step S1 includes: in a vacuum chamber (pressure ≤ 10⁻³ Pa), cold embedding the sample with Struers Epofix epoxy resin for 8 - 10 h.

[0015] As a further improvement of the present invention, step S2 includes: using an argon ion polishing machine to perform cross-section polishing and surface electrolyte thinning on the sample.

[0016] As a further improvement of the present invention, step S2 includes: after embedding, the sample is polished with 1200 - mesh sandpaper until the LSCF-CGO layer is basically removed, presenting a bright electrolyte layer; then, using a JEOL IB-19530 argon ion polishing machine (acceleration voltage 5.5 kV) to polish the cross-section and plane for 2 h and 20 min respectively to flatten the cross-section and basically remove the surface electrolyte; then, depositing a Pt conductive film protection layer about 10 nm thick on the surface of the battery slice to prevent charge accumulation.

[0017] As a further improvement of the present invention, step S3 includes the following sub-steps: S301. Depositing a Pt film on the sample surface; S302. Fixing the sample to a nail table with a pressure plate; S303. Depositing a Pt protection layer in the target area; S304. Machining grooves on both sides of the target area.

[0018] As a further improvement of the present invention, step S303 includes: sending the sample to a dual-beam electron microscope for processing and observation, depositing a Pt protective layer on the surface of the target area by using ion beam induced deposition technology, tilting the stage, adjusting the working distance to the concentric point and then performing cross-section processing, using a Ga+ ion beam, and correcting the geometric distortion of the image by using the tilt compensation and dynamic focusing functions to ensure the SEM image quality.

[0019] As a further improvement of the present invention, a 1-μm-thick Pt protective layer is deposited on the surface of the target area again by using ion beam induced deposition technology.

[0020] As a further improvement of the present invention, the stage is tilted 54° in the FIB, and cross-section processing is performed after adjusting the working distance of 5.1 mm to the concentric point.

[0021] As a further improvement of the present invention, step S304 includes: using a higher ion beam current to cut large grooves around the target area to avoid detector shadows during SEM imaging, and then reducing the ion beam current to the set value for fine processing of the final cross-section.

[0022] As a further improvement of the present invention, step S4 includes: before continuous slicing in the FIB, calling out the TEM probe to press the edge of the sample to ensure a small displacement error during FIB slicing; then performing continuous slicing with a set layer thickness, synchronously collecting In-lens and SE2 dual-channel images, collecting surface topography images through the In-lens detector, and simultaneously obtaining high-contrast phase distribution images with the SE2 detector.

[0023] As a further improvement of the present invention, step S5 includes: using the last high-contrast In-lens image in step S4 as a deep learning sample, extracting multi-phase feature data of the sample based on threshold segmentation and generating a label dataset corresponding to the SE2 image, training a deep learning model and automatically segmenting the SE2 image sequence, manually optimizing local areas and iteratively training the model, and reconstructing the 3D geometry.

[0024] As a further improvement of the present invention, step S5 includes: after continuous slicing is completed, pausing for several seconds on the last image, using the contrast change during the charging effect process, and collecting again after short-term charge accumulation to obtain a single In-lens image with high contrast although the contrast is inverted; using the collected single In-lens image as a training sample, extracting the contrast thresholds of Ni, YSZ, and pores by multi-threshold segmentation and generating a binary label, adopting a U-Net architecture with an input layer size of 96×96, setting the patch size to 128, extracting 4 patches from each image, training with the Dice coefficient + cross-entropy as the loss function, and finally importing the segmented SE2 sequence into 3D reconstruction software for voxel reconstruction.

[0025] The beneficial effects of the present invention are as follows: A three-dimensional reconstruction method for the porous Ni-YSZ electrode of a solid oxide battery is provided. By integrating in-situ characterization and intelligent image analysis techniques, accurate analysis of the structure of multiphase materials is achieved, with advantages such as high sample preparation stability, good three-dimensional reconstruction accuracy, and high degree of automation in phase identification. It is particularly suitable for the study of the microstructure evolution at the electrode / electrolyte interface of solid oxide fuel cells and has important value for revealing the mechanism of material performance degradation under high-temperature complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0027] Figure 1 is a flowchart of a three-dimensional reconstruction method for the porous Ni-YSZ electrode of a solid oxide battery according to the present invention; Figure 2 is a schematic diagram of a stage device with a pressed tablet; Figure 3 is a schematic diagram of TEM probe limit; Figure 4 is the last slice diagram under continuous section double-channel sampling; Figure 5 is the last In-lens slice diagram of resampling after stopping sectioning; Figure 6 is a diagram showing the effect (U-Net architecture) of training the corresponding SE2 pictures using the last high-contrast In-lens slice. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0029] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation on the protection scope of the present invention. In addition, terms such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more.

[0030] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "mounted", "connected", "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood through specific circumstances.

[0031] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0032] A three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery, which relates to a method for optimizing the compositional contrast and intelligent segmentation of a Ni-YSZ composite material based on the FIB serial sectioning technique, belongs to the technical field of material microstructure characterization. The purpose is to solve the technical problems such as sample drift, insufficient compositional contrast, and poor phase segmentation accuracy in the traditional three-dimensional characterization during the FIB sectioning process.

[0033] The new sample processing flow designed by the present invention successively includes cold embedding with epoxy resin under vacuum, cross-section polishing and surface electrolyte thinning using an argon ion polishing machine, Pt protective layer deposition, sample fixation with a sample stage with pressure pins (cooperating with a TEM probe to press the sample tightly to prevent drift before FIB serial sectioning), local microfabrication and FIB serial sectioning, In-lens / Se2 dual-channel acquisition (using the last high-contrast In-lens image as the deep learning sample), extracting multi-phase feature data of the sample based on threshold segmentation and generating a label data set corresponding to the SE2 image, deep learning model training and automatic segmentation of the SE2 image sequence, manual optimization of the local area and iterative training of the model. This sample processing method realizes the accurate analysis of the structure of multi-phase materials by integrating in-situ characterization and intelligent image analysis techniques, and has the advantages of high sample preparation stability, good three-dimensional reconstruction accuracy, and high degree of automation in phase recognition. It is particularly suitable for the study of the microstructure evolution of the electrode / electrolyte interface of solid oxide fuel cells, and has important value for revealing the mechanism of material performance degradation under high-temperature complex working conditions.

[0034] As Figure 1 shown, the main processes of the present invention include: vacuum cold embedding, argon ion polishing, sample pretreatment and fixation, FIB-SEM dual-channel acquisition, and deep learning analysis. The specific processes are as follows: First, in a vacuum chamber (pressure ≤ 10⁻³ Pa), the sample is cold embedded with Struers Epofix epoxy resin for 8 - 10 h. After embedding, the sample is polished with 1200-mesh sandpaper until the LSCF-CGO layer is basically removed, presenting a bright electrolyte layer. Subsequently, a JEOL IB-19530 argon ion polishing machine (acceleration voltage 5.5 kV) is used to polish the cross-section and the plane for 2 h and 20 min respectively to flatten the cross-section and basically remove the surface electrolyte. Then, a Pt conductive film protective layer with a thickness of about 10 nm is plated on the surface of the battery slice to prevent charge accumulation.

[0035] Next, as Figure 2As shown in the figure, a stage device 4 with a press plate is provided. The nailing table 2 with the press plate 3 is used to fix the sample 1 in situ by a pressing device. The stage device 4 can prevent the sample from drifting during the sectioning process. The sample is sent to a Zeiss Crossbeam 350 dual-beam electron microscope for processing and observation. The ion beam induced deposition technology is used to deposit a 1-μm-thick Pt protective layer on the surface of the target area again. In the FIB-SEM, the stage is tilted at 54°, and after adjusting the working distance of 5.1 mm to the concentric point, cross-section processing is carried out using a 30-kV Ga+ ion beam. To ensure the SEM image quality, the tilt compensation and dynamic focusing functions are used to correct the geometric distortion of the image. During the processing, a higher ion beam current (up to 65 nA) is first used to cut large grooves around the target area to avoid detector shadows during SEM imaging; then the ion beam current is reduced to 3 nA for fine processing of the final cross-section.

[0036] Before continuous FIB sectioning, as Figure 3 shown in the figure, the TEM probe is called out to press the edge of the sample to ensure a small displacement error during FIB sectioning, and the TEM probe limit is used to prevent the sample from drifting during the sectioning process; then continuous sectioning is carried out with a layer thickness of 50 nm, and In-lens and SE2 dual-channel image acquisition is carried out synchronously (acceleration voltage 1.5 kV, working distance 5.1 mm). The surface topography image is acquired through the In-lens detector, and at the same time, a high-contrast phase distribution image with a resolution of ≥2048×1536 pixels is obtained using the SE2 detector.

[0037] To solve the problem that it is difficult to achieve effective phase segmentation by the backscattered electron detector under various conditions due to the close average atomic numbers of Ni / YSZ calculated by the electron fraction weighting method and the intensity changes caused by the Ni grain orientation differences (as Figure 4 shown in the figure), after the continuous sectioning is completed, pause for a few seconds on the last image, utilize the contrast change during the charging effect process, and after the charge accumulates briefly, collect again to obtain a single In-lens image with high contrast although the contrast is inverted (as Figure 5 shown in the figure). Taking this last In-lens image as the training sample, the contrast thresholds of Ni, YSZ, and pores are extracted by multi-threshold segmentation and binary labels are generated. The U-Net architecture with an input layer size of 96×96 (including 4 layers of downsampling and upsampling) is used, the patch size is set to 128, 4 patches are extracted from each image, and training is carried out with the Dice coefficient + cross-entropy as the loss function (processing 3 images per batch, batch size = 12). Under the condition that it is difficult to obtain the label picture corresponding to the SE2 image, it is possible to achieve a verification accuracy of more than 90% for the YSZ and pore phases, a verification accuracy of more than ≥80% for the Ni phase, and a comprehensive verification accuracy of ≥85% with only 3 sample pictures and about 500 steps of iteration, as Figure 6As shown. Finally, the segmented SE2 sequence can be imported into 3D reconstruction software such as Avizo for voxel reconstruction, and parameters such as the volume fraction of each phase and the interface curvature can be quantitatively statistically analyzed to correlate with the material properties.

[0038] This solution realizes the non-destructive preparation of samples through the combination of vacuum cold embedding and argon ion polishing, uses a double fixation method of a pressing tablet nail stage + TEM probe to solve the problem of nano-section drifting, innovatively utilizes the charging effect to obtain high-contrast In-lens images as training samples, and adopts appropriate U-Net neural network parameters to achieve efficient training with small samples. It effectively overcomes the difficulty of phase segmentation caused by the close average atomic numbers of Ni / YSZ in traditional backscattered electron images, and finally forms a standardized full process of 3D reconstruction from precise sample preparation, dual-channel imaging to intelligent segmentation, providing a high-precision analysis method for the study of the interface evolution of multiphase materials.

[0039] Compared with other existing technologies, the advantages of the present invention are as follows: (1) Compared with the traditional FIB-SEM 3D reconstruction method, the present invention realizes the non-destructive preparation of samples through the combination of vacuum cold embedding and argon ion polishing, avoiding the microstructure distortion caused by mechanical damage.

[0040] (2) By using the double fixation technology of a pressing tablet nail stage + TEM probe, the problem of sample drift during the nano-level continuous sectioning process is effectively solved, the difficulty of subsequent image sequence alignment is greatly reduced, and the accuracy of 3D reconstruction is significantly improved.

[0041] (3) Innovatively utilize the charging effect to obtain high-contrast In-lens images as training samples, solve the problem of insufficient contrast caused by the close average atomic numbers of Ni / YSZ in traditional backscattered electron images, and at the same time overcome the interference of contrast changes caused by the Ni grain orientation difference on phase segmentation.

[0042] (4) Using the U-Net neural network structure, the segmentation accuracy of more than 90% for the YSZ and pore phases and more than 80% for the Ni phase can be achieved with only 3 training samples, significantly reducing the requirements of deep learning for the number of samples and manual marking, and greatly avoiding the reduction of phase segmentation accuracy caused by the randomness of manual annotation.

[0043] (5) Provide a standardized full process of 3D reconstruction from precise sample preparation, dual-channel imaging to intelligent segmentation.

[0044] (6) The In-lens imaging technology enhanced by the charging effect and the deep learning algorithm effectively solve the problem of image segmentation caused by insufficient contrast difference between phases in the Ni / YSZ system.

[0045] (7) The sample fixation technology of the cooperation between the pressing tablet nail stage and the TEM probe ensures the stability of nano-level continuous sectioning.

[0046] (8)The argon ion polishing process of specific parameters during the sample preparation process and the FIB parameters of the microfabrication process.

[0047] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A three-dimensional reconstruction method for a porous Ni-YSZ electrode of a solid oxide battery, characterized in that It includes the following steps: S1. Vacuum cold embedding; S2. Argon ion polishing; S3. Sample pretreatment and fixation; S4. FIB serial sectioning and In-lens and SE2 dual-channel image acquisition; S5. Deep learning analysis.

2. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 1, wherein: Step S1 includes: Under vacuum, use epoxy resin to cold embed the sample.

3. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 1, characterized in that: Step S2 includes: Use an argon ion polishing machine to perform cross-section polishing and surface electrolyte thinning on the sample.

4. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 1, wherein: Step S3 includes the following sub-steps: S301. Deposit a Pt film on the sample surface; S302. Fix the sample to a nail stage with a pressure plate; S303. Deposit a Pt protective layer in the target area; S304. Machine grooves on both sides of the target area.

5. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 4, characterized in that: Step S303 includes: Send the sample to a dual-beam electron microscope for processing and observation. Use the ion beam induced deposition technology to deposit a Pt protective layer on the surface of the target area. Tilt the stage, adjust the working distance to the concentric point and then perform cross-section processing. Use Ga+ ion beam. To ensure the quality of the SEM image, use the tilt compensation and dynamic focusing functions to correct the geometric distortion of the image.

6. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 5, wherein: Tilt the stage 54° in the FIB, and adjust the working distance to 5.1 mm to the concentric point and then perform cross-section processing.

7. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 4, characterized in that: Step S304 includes: Use a higher ion beam current to cut large grooves around the target area to avoid detector shadows during SEM imaging. Then reduce the ion beam current to the set value for fine processing of the final cross-section.

8. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 1, characterized in that: Step S4 includes: Before FIB serial sectioning, call out the TEM probe to press the edge of the sample to ensure a small displacement error during FIB sectioning; then perform serial sectioning at a set layer thickness, and synchronously perform In-lens and SE2 dual-channel image acquisition. Collect the surface topography image through the In-lens detector, and at the same time use the SE2 detector to obtain a high-contrast phase distribution image.

9. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 1, wherein: Step S5 includes: Use the last high-contrast In-lens image in Step S4 as the deep learning sample, extract the multi-phase feature data of the sample based on threshold segmentation and generate a label dataset corresponding to the SE2 image. Perform deep learning model training and automatic segmentation of the SE2 image sequence, manually optimize the local area and perform model iterative training, and reconstruct the 3D geometry.

10. The three-dimensional reconstruction method of the porous Ni-YSZ electrode of the solid oxide battery according to claim 9, characterized in that: Step S5 includes: After the serial sectioning is completed, pause on the last image for a few seconds, utilize the contrast change during the charging effect process, and collect again after a short charge accumulation to obtain a single high-contrast In-lens image with contrast inversion. Use the collected single In-lens image as the training sample, extract the contrast thresholds of Ni, YSZ, and pores by multi-threshold segmentation and generate a binary label. Use a U-Net architecture with an input layer size of 96×96, set the patch size to 128, extract 4 patches from each image, and train with the Dice coefficient + cross entropy as the loss function. Finally, import the segmented SE2 sequence into a 3D reconstruction software for voxel reconstruction.

Images