A method, device, medium and product for detecting silicon content in a solid-state electrode material

Abstract

This invention belongs to the field of electrochemical analysis and material composition detection technology, specifically disclosing a method, equipment, medium, and product for detecting silicon content in solid-state electrode materials. The method includes: performing non-destructive X-ray microtomography on a solid-state electrode sample to construct a three-dimensional voxel model; using a deep learning semantic segmentation algorithm to identify silicon particles, carbon coating layers, and binder regions; extracting the geometric and spatial information of the silicon particles; applying a continuous cohomology algorithm to calculate the Betti number and analyze the connectivity of the silicon particles; quantitatively assessing the effective silicon content participating in the electrochemical reaction based on this, and generating a three-dimensional visualization distribution map. Through the above technical solution, this invention achieves non-destructive, three-dimensional, and functional detection of silicon content, improving identification accuracy and structural analysis capabilities, and providing a new path for lithium-ion battery research and development and quality control.

Description

Technical Field

[0001] This invention belongs to the field of electrochemical analysis and material composition detection technology, specifically relating to a method, equipment, medium, and product for detecting silicon content in solid electrode materials. Background Technology

[0002] With the rapid development of global energy storage technology and the electric vehicle industry, silicon-based composite materials, as key anode materials for improving the energy density of lithium-ion batteries, have attracted widespread attention from academia and industry in terms of structural design and performance evaluation. In the research and development and manufacturing process of solid-state electrodes, the precise measurement of active material content and in-depth analysis of microstructure are not only core indicators for verifying the consistency of production processes, but also important prerequisites for exploring battery failure mechanisms and improving charge-discharge cycle life.

[0003] Detection techniques for the three-dimensional spatial distribution and effectiveness of silicon content in solid-state electrodes are a key technology for achieving high-performance utilization of electrode materials. This research direction aims to non-destructively detect the geometric characteristics, volume fraction, and spatial arrangement of silicon particles within the porous network of the electrode using high-resolution imaging and mathematical modeling, thereby providing microscopic data support for the macroscopic electrochemical performance of the electrode.

[0004] The inventors have discovered at least the following technical problems in the relevant technologies: Traditional chemical digestion methods rely heavily on sample pretreatment and destruction, making real-time detection impossible while maintaining the integrity of the material structure. Furthermore, because the measurement results only reflect the macroscopic total content, the non-uniformity of component distribution within the electrode is masked. Simultaneously, existing characterization methods struggle to distinguish silicon particles in different coating states, lacking the ability to quantitatively analyze the connectivity and spatial topology of active silicon networks, making it difficult to accurately define the effective silicon components participating in electrochemical reactions. Moreover, limitations in the segmentation of multiphase complex media by traditional image processing algorithms result in blurred semantic boundaries between silicon particles and components such as the carbon-based matrix and binders, making it impossible to accurately correlate microscopic damage and evolution characteristics within the material. Summary of the Invention

[0005] One object of the present invention is to provide a method, device, medium and product for detecting silicon content in solid electrode materials, thereby solving the problems in the background art.

[0006] In a first aspect, the present invention provides a method for detecting silicon content in a solid electrode material, comprising the following steps: performing non-destructive X-ray microtomography on a solid electrode sample to obtain projection data reflecting the absorption characteristics of the internal substances of the solid electrode sample, and generating internal three-dimensional density distribution data of the solid electrode sample through a back-projection reconstruction algorithm; constructing a three-dimensional voxel model based on the internal three-dimensional density distribution data; using a deep learning semantic segmentation algorithm to identify multiphase components in the three-dimensional voxel model, distinguishing silicon particle regions, carbon coating layer regions, and binder regions; extracting the spatial location, geometric morphology, and volume information of silicon particles within the silicon particle regions based on the segmentation results of the multiphase component identification; calculating the Betti number of the silicon particle regions based on the spatial location of the silicon particles, and identifying isolated silicon particles and interconnected silicon network structures; quantitatively evaluating the effective silicon content participating in the electrochemical reaction based on the identification results of the isolated silicon particles and the interconnected silicon network structures, and generating a three-dimensional visualization distribution map reflecting the distribution state of the silicon particles.

[0007] In a second aspect, the present invention also provides an electronic device comprising: one or more processors; and a memory storing computer program instructions, which, when executed, cause the processor to perform the steps of the method described above.

[0008] Thirdly, the present invention also provides a computer-readable medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the method described above.

[0009] Fourthly, the invention also provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the method described above.

[0010] Compared with the prior art, the present invention has the following beneficial effects:

[0011] 1. This invention combines X-ray microtomography with algebraic topology analysis to achieve non-destructive, three-dimensional, quantitative, and functional detection of silicon content in solid electrode materials.

[0012] 2. Compared with the limitations of traditional chemical digestion methods, which can only provide macroscopic total content, this method can accurately identify the spatial distribution, geometric shape and connectivity of silicon particles while maintaining sample integrity, thereby distinguishing between active and inactive silicon components.

[0013] 3. By combining deep learning semantic segmentation with continuous cohomology algorithms, the accuracy of target component identification and topological structure resolution in multiphase complex media are improved.

[0014] 4. The generated three-dimensional visualization distribution map not only intuitively reflects the internal microstructure characteristics of the material, but can also serve as a key input for electrode performance prediction and failure analysis.

[0015] 5. This method has good scalability and automation potential, and is suitable for laboratory research and industrial online testing scenarios, providing a new technical path for the research and development and quality control of high-energy-density lithium-ion batteries. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall technical solution architecture of a method for detecting silicon content in solid electrode materials proposed in this invention;

[0017] Figure 2 This is a schematic diagram of the core principle framework for multiphase component identification using deep learning semantic segmentation algorithms in this invention;

[0018] Figure 3 This is a schematic diagram illustrating the core principle framework of the continuous cohomology algorithm used in this invention to analyze the topological connectivity of silicon particles.

[0019] Figure 4 This is a flowchart illustrating the main stages of the process from acquiring three-dimensional density distribution data to constructing a three-dimensional voxel model in this invention.

[0020] Figure 5 This is a flowchart illustrating the logical flow of the present invention for evaluating and visualizing effective silicon content based on connectivity analysis results.

[0021] Figure 6 This is an exemplary structural diagram of an electronic device proposed in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1: Reference Figures 1 to 5The method for detecting silicon content in a solid electrode material provided by this invention specifically includes: performing non-destructive X-ray microtomography on a solid electrode sample to obtain projection data reflecting the absorption characteristics of the internal substances of the solid electrode sample, and generating internal three-dimensional density distribution data of the solid electrode sample through a back-projection reconstruction algorithm; constructing a three-dimensional voxel model based on the internal three-dimensional density distribution data; using a deep learning semantic segmentation algorithm to identify multiphase components in the three-dimensional voxel model, distinguishing silicon particle regions, carbon coating layer regions, and binder regions; extracting the spatial location, geometric shape, and volume information of silicon particles within the silicon particle regions based on the segmentation results of multiphase component identification; calculating the Betti number of the silicon particle regions based on the spatial location of the silicon particles, and identifying isolated silicon particles and interconnected silicon network structures; quantitatively evaluating the effective silicon content participating in the electrochemical reaction based on the identification results of the isolated silicon particles and the interconnected silicon network structures, and generating a three-dimensional visualization distribution map reflecting the distribution state of silicon particles.

[0024] The following sections will explain each step.

[0025] First, perform step 1: perform non-destructive X-ray microtomography on the solid electrode sample to obtain projection data reflecting the absorption characteristics of the internal material of the solid electrode sample, and generate the internal three-dimensional density distribution data of the solid electrode sample through a back projection reconstruction algorithm.

[0026] In the specific implementation process, the X-ray micro-component computed tomography (Micro-CT) equipment used includes core components such as a high-power microfocus X-ray source, a high-precision four-axis rotary stage, a scintillator-coupled high-resolution CMOS detector, and an integrated image acquisition and control system. The filament current of the X-ray source can be adjusted between 10 microamps and 500 microamps, and the accelerating voltage is set between 40 kV and 160 kV to ensure that the X-rays have sufficient penetrating power to penetrate the solid electrode sample containing high atomic number materials.

[0027] The process of fixing the solid electrode sample may include: processing the solid electrode sample into a cylinder with a diameter of 1 mm to 3 mm, and using cyanoacrylate adhesive to vertically fix the solid electrode sample on an air bearing rotary table with submicron radial runout accuracy.

[0028] During the data acquisition phase, the air-bearing rotary table can rotate 360 ​​degrees in 0.1 to 0.5 degree angular steps. At each angular position, the high-resolution CMOS detector can acquire a 16-bit depth projection image. To improve the signal-to-noise ratio, multiple exposure stacking and averaging can be performed at each angular position, for example, using an image averaging technique of 4 to 8 frames. The total number of projections can be set between 1200 and 3600. The acquired projection data is processed by 3D reconstruction software based on the Filtered Back Projection (FBP) algorithm or the Algebraic Reconstruction (ART) algorithm.

[0029] During the reconstruction process, an annular artifact removal operator and beam hardening correction parameters can be introduced to eliminate the edge enhancement effect caused by uneven X-ray energy spectrum distribution. The final output data is a series of continuous slice image sequences, which constitute the internal three-dimensional density distribution data of the solid-state electrode sample; the internal three-dimensional density distribution data is an isotropic three-dimensional density field with a spatial resolution of 200 nm to 800 nm.

[0030] Next, proceed to step 2: construct a three-dimensional voxel model based on the three-dimensional density distribution data.

[0031] Specifically, in the process of constructing the three-dimensional voxel model, the reconstructed slice image sequence can be converted into a three-dimensional numerical matrix, where each voxel represents the X-ray linear attenuation coefficient at an internal coordinate point of the solid electrode sample.

[0032] To eliminate background noise and enhance target features, the three-dimensional numerical matrix can be subjected to three-dimensional median filtering or nonlocal mean denoising. The processed grayscale values ​​are mapped to a numerical range of 0 to 65535, generating a preprocessed grayscale model. The grayscale intensity of each voxel directly reflects the electron density distribution of the material at the coordinate point.

[0033] Because silicon, carbon, binders, and pores exhibit significant differences in their X-ray absorption capabilities, in the aforementioned three-dimensional voxel model, silicon particles typically represent high-grayscale regions, carbon and binders represent medium-to-low-grayscale regions, while pores represent low-grayscale regions. The three-dimensional voxel model can be stored using a hierarchical data format (HDF5) to support efficient random access to massive amounts of data.

[0034] Then, step 3 is performed: using a deep learning semantic segmentation algorithm to identify the multiphase components of the three-dimensional voxel model, distinguishing silicon particles, carbon coating layers, and binder regions.

[0035] Optionally, in some embodiments, the step of using a deep learning semantic segmentation algorithm to identify multiphase components of the three-dimensional voxel model may include:

[0036] The deep learning semantic segmentation algorithm is based on an encoder and decoder architecture. The encoder includes a series of three-dimensional convolutional layers, residual connection structures, and max pooling layers. The three-dimensional convolutional layers perform element-wise multiplication and summation of the weight tensors with the local regions of the three-dimensional voxel model to extract multi-scale features and obtain encoded geometric features.

[0037] The decoder performs upsampling through transposed convolution and fuses the geometric features in the encoder with the semantic information in the decoder through skip connection paths to obtain a fused feature map;

[0038] The skip connection path integrates an attention mechanism module, which focuses on the boundary between the silicon particle region and the carbon coating layer region by calculating the correlation between each pixel in the fused feature map, and outputs the predicted boundary.

[0039] The loss function of the deep learning semantic segmentation algorithm adopts a weighted combination of cross-entropy loss function, similarity coefficient loss function and boundary-aware loss term. The boundary-aware loss term enhances the recognition accuracy of thin-layer structures by calculating the distance between the predicted boundary and the real boundary.

[0040] Specifically, the deep learning semantic segmentation algorithm may employ an improved U-Net 3D convolutional neural network architecture, which includes symmetrical encoder and decoder paths.

[0041] The encoder can extract multi-scale features through consecutive 3D convolutional layers and max-pooling layers, with the convolutional kernel size set to 3×3×3. To address the issue of the extremely wide size distribution of silicon particles in solid-state electrodes, a residual connection structure is integrated into each layer of the encoder to ensure that the gradients of deep networks can be effectively propagated back.

[0042] In the decoder path, upsampling can be performed using transposed convolution, and a skip connection mechanism can be used to fuse the high-resolution geometric features in the encoder with the semantic information in the decoder. Specifically, spatial attention and channel attention mechanisms can be embedded in the skip connection path. The spatial attention module can automatically focus on the boundary region between the silicon particle region and the carbon coating layer region by calculating the correlation between pixels in the fused feature map, and output the predicted boundary. The channel attention module can reweight the importance of different feature channels through global average pooling and fully connected layers.

[0043] The dataset used to train the deep learning semantic segmentation algorithm can be derived from typical silicon-carbon electrode samples that have been meticulously annotated by humans, covering different particle morphologies and loading amounts.

[0044] The loss function of the deep learning semantic segmentation algorithm can be a weighted combination of the cross-entropy loss function and the Dice similarity coefficient loss function. The Dice loss term is specifically used to alleviate class imbalance, i.e., when the volume of the silicon particle region is much smaller than the background volume. Furthermore, in some examples, a boundary-aware loss term can be introduced. By calculating the Euclidean distance between the predicted boundary and the true boundary, the recognition accuracy of thin-layer structures is enhanced, forcing the network to learn extremely small carbon coating features.

[0045] After semantic segmentation is completed, step 4 is executed: based on the segmentation results of multiphase component identification, the spatial location, geometric shape and volume information of silicon particles in the silicon particle region are extracted.

[0046] Optionally, in some embodiments, the extraction of the spatial location, geometry, and volume information of silicon particles within the silicon particle region based on the segmentation results of multiphase component identification may include:

[0047] Using a three-dimensional connected component labeling algorithm, silicon voxels within the identified silicon particle region are classified based on the connectivity principle, and a unique identifier is assigned to each independent silicon particle.

[0048] For each particle corresponding to the identification number, the centroid coordinates are calculated by arithmetically averaging the coordinates of the voxel to which it belongs, and the volume information of the silicon particle is calculated by combining the total number of silicon voxels with the physical volume of a single silicon voxel.

[0049] Based on the centroid coordinates and the silicon particle volume information, the geometry of the silicon particle is calculated; the geometry of the silicon particle includes the equivalent spherical diameter, surface area, and the ratio of surface area to volume.

[0050] Anisotropic parameters are obtained by calculating the eigenvalues ​​of the particle inertia tensor, which are used to evaluate the extension of the silicon particle in the directions of the major axis, median axis and minor axis. The extracted anisotropic parameters and the geometry of the silicon particle are stored in a database.

[0051] Specifically, the three-dimensional connected component labeling algorithm (CCL) can be used to classify the silicon voxels within the identified silicon particle region based on the 26 connectivity principle, and assign a unique identifier to each independent silicon particle.

[0052] For each silicon particle corresponding to a number, the centroid coordinates of the silicon particle are calculated. This can be done by taking the arithmetic mean of the Cartesian coordinates of the silicon voxels contained within the silicon particle. Volume information can be obtained by counting the total number of voxels occupied by the silicon particle and multiplying this by the physical volume of a single voxel.

[0053] Geometric morphology assessment may include calculating the equivalent spherical diameter (i.e., the diameter of a sphere with the same volume), surface area, and the surface area to volume ratio. To describe the degree of despherization of the particles, anisotropic parameters may also be introduced, and the extension of the silicon particles along their major, intermediate, and minor axes may be assessed by calculating the eigenvalues ​​of the particle inertia tensor. The extracted anisotropic parameters and the silicon particle geometry are stored in a structured query language database for large-scale statistical analysis.

[0054] Next, proceed to step 5: The continuous cohomology algorithm in algebraic topology can be applied to calculate the Betti number of the silicon particle region based on the spatial location of the silicon particle, and to identify isolated silicon particles and interconnected silicon network structures.

[0055] Optionally, in some embodiments, calculating the Betti number of the silicon particle region based on the spatial location of the silicon particle, and identifying isolated silicon particles and interconnected silicon network structures, may include:

[0056] The extracted spatial locations of the silicon particles are converted into centroid point cloud data as input. A sphere with a radius varying with the filtering parameters is placed at each centroid point to construct a series of nested simple complexes.

[0057] During the evolution of the simplex, as the filtering parameter increases, when the distance between any two points is less than twice the filtering parameter, a one-dimensional simplex is formed by connecting the two points.

[0058] When the pairwise distances between any three points satisfy the requirements of the filtering parameters, a two-dimensional simplex is formed.

[0059] During the simple complex evolution process, the generation and destruction of topological features are monitored in real time, and the zero-dimensional Betti number representing the number of connected components and the one-dimensional Betti number representing the number of loops are calculated.

[0060] Based on the lifecycle span of the filtering parameters during the change process, the system distinguishes between connected silicon network structures with robust topology and isolated silicon particles that do not meet the preset connectivity scale threshold.

[0061] Specifically, the spatial positions of the silicon particles obtained in step 4 can be converted into centroid point cloud data as input. The continuous cohomology algorithm constructs a series of nested simple complexes by placing a three-dimensional sphere with a radius that increases with the filtering parameter at each centroid point. The simple complexes are specifically Vietoris-Rips complexes.

[0062] As the filtering parameters gradually increase, previously isolated points begin to contact each other and form edges, subsequently forming higher-dimensional simplexes such as triangular facets and tetrahedrons. During the evolution of these simplex complexes, the generation and disappearance of topological features can be monitored in real time. Specifically, the zero-dimensional Betti number (Betti 0) represents the number of connected components at the current scale. This zero-dimensional Betti number decreases as the filtering parameters increase, reflecting the gradual aggregation of the silicon particles. The one-dimensional Betti number (Betti 1) represents the number of loops in the structure, reflecting the closed loops formed between the silicon particles. The two-dimensional Betti number (Betti 2) represents the enclosed cavities.

[0063] The persistent coherence algorithm generates persistent barcode maps or persistent graphs, where each line segment represents the lifecycle of a topological feature, with the starting point corresponding to the feature's generation scale and the ending point corresponding to the feature's extinction scale. Features with a lifecycle greater than a preset threshold are determined to be robust topological structures inherent in the solid-state electrode sample, while features with a lifecycle less than the preset threshold are determined to be noise.

[0064] By setting a connectivity scale threshold, silicon particles can be divided into two main categories: one is the connected silicon network structure located in a large-scale connected network, which constitutes long-range electron and ion transport paths; the other is the physically isolated silicon particle, which cannot effectively participate in electrochemical reactions.

[0065] Further, step 6 is performed: based on the identification results of the isolated silicon particles and the interconnected silicon network structure, the effective silicon content participating in the electrochemical reaction is quantitatively evaluated, and a three-dimensional visualization distribution map reflecting the distribution state of the silicon particles is generated.

[0066] Optionally, in some embodiments, the step of quantitatively assessing the effective silicon content participating in the electrochemical reaction based on the identification results of the isolated silicon particles and the interconnected silicon network structure may include:

[0067] Based on the identification results of the interconnected silicon network structure, the total volume of silicon particles belonging to the largest connected cluster in the interconnected silicon network structure, or the total volume of silicon particles that meet the preset topological connectivity threshold, is defined as the effective silicon volume.

[0068] The proportion of the effective silicon volume to the total volume of all silicon particles is calculated, and the performance of the electrode material is evaluated in combination with the spatial distribution uniformity index. The spatial distribution uniformity index is obtained by dividing the three-dimensional voxel model into multiple sub-regions and statistically analyzing the deviation of silicon content in each sub-region.

[0069] The results of evaluating the electrode material performance are correlated with the three-dimensional spatial location of the three-dimensional voxel model. The three-dimensional visualization distribution map is generated using a color mapping method, and silicon particles with different connectivity states are mapped to color ranges. The interconnected silicon network structure is marked with warm colors, and the isolated silicon particles are marked with cool colors. An interactive cross-sectional observation interface is provided.

[0070] Specifically, the effective silicon content is defined as the percentage of the total volume of silicon particles belonging to the largest connected cluster in the interconnected silicon network structure, or the total volume of silicon particles satisfying a preset topological connectivity threshold, relative to the total volume of all silicon particles within the detection area. The effective silicon content is directly related to the reversible capacity of the battery. Simultaneously, a three-dimensional spatial distribution uniformity index can be calculated by dividing the three-dimensional voxel model into several sub-regions and statistically analyzing the standard deviation of the effective silicon content within each sub-region.

[0071] The 3D visualization distribution map can be presented using voxel rendering or isosurface extraction techniques. In the generated interactive view, silicon particles in different connectivity states can be assigned different color mappings. For example, silicon particles within the interconnected silicon network structure are displayed in warm colors (red to orange) to identify high-risk or high-efficiency utilization areas; while isolated silicon particles are displayed in cool colors (blue to dark purple) to identify unutilized active materials.

[0072] Users can perform arbitrary sectioning operations on the 3D voxel model along the X, Y, and Z axes through a graphical interface to observe the cross-sectional morphology at any internal depth. The 3D visualization distribution map supports real-time interactive querying; after selecting any silicon particle, it displays parameters such as the particle's volume, surface area, connectivity level, and topological dimension.

[0073] Optionally, in some embodiments, the method further includes:

[0074] Scanning data of the same solid electrode sample at different electrochemical cycles were used as input. Spatial registration was performed using rigid registration based on mutual information and non-rigid registration based on adaptive grids to eliminate displacement and deformation of the solid electrode sample during assembly and disassembly, and to obtain an aligned time-series three-dimensional model.

[0075] Based on the aligned time-series 3D model, the changes in topological features at different time points are compared to quantify the degree of fragmentation, detachment behavior, and volume evolution of the silicon particles during the cycling process. Based on the quantification results, a correlation model between microstructure degradation and battery capacity decay is constructed.

[0076] In the specific implementation process, in order to monitor the dynamic evolution of the solid-state electrode sample, this embodiment may also include time-series comparative analysis of multiple scan results. The internal three-dimensional density distribution data of the same solid-state electrode sample in the initial state, after the 10th cycle, the 50th cycle, and after the 100th cycle are used as input.

[0077] Spatial registration was performed using rigid registration based on mutual information and non-rigid registration algorithms based on adaptive meshes to eliminate displacement and deformation of the solid-state electrode sample during multiple assembly and disassembly processes, resulting in an aligned temporal 3D model. Based on this aligned temporal 3D model, by comparing the changes in topological features at different time points, the degree of silicon particle fragmentation (smaller volume, sharp increase in Betti number), detachment behavior (from connected network to isolated point), and volume evolution (accumulation of electrolyte by-reaction products) during the cycling process can be quantified.

[0078] Specifically, the degree of fragmentation is quantified by observing a decrease in the volume of the silicon particles and an increase in the zero-Vibetty number; the detachment behavior is quantified by observing the transformation of the silicon particles from the interconnected silicon network structure to isolated silicon particles. Furthermore, the accumulation of electrolyte by-reaction products is quantified by analyzing the evolution of grayscale values ​​in the three-dimensional voxel model. Based on the above quantification results, a correlation model between microstructure degradation and battery capacity decay can be constructed.

[0079] Example 2: In Example 2, the method provided by the present invention is applied to a nano-silicon-carbon composite anode material with a porous structure. In the application scenario of this nano-silicon-carbon composite anode material, since the silicon particle size is typically between 50 and 200 nanometers, close to the physical limit of conventional microtomography, a nano-CT scanning technique based on a synchrotron radiation source can be used in step 1. This nano-CT scanning technique utilizes a Fresnel lens as the objective lens and can achieve an isotropic spatial resolution better than 50 nanometers.

[0080] Optionally, in some embodiments, the method further includes:

[0081] After constructing the three-dimensional voxel model, voxel clusters with gray values ​​higher than the theoretical upper limit of silicon particles are identified and determined to be impurities. When applying the continuous homology algorithm for calculation, the impurities are masked.

[0082] During the training phase of the deep learning semantic segmentation algorithm, operations such as rotation, translation, scaling, and adding noise are performed on the labeled samples to perform data augmentation.

[0083] For porous silicon materials with internal pore structures, the closed-pore regions inside the silicon framework are identified by the deep learning semantic segmentation algorithm. When quantitatively evaluating the effective silicon content participating in the electrochemical reaction, the volume of silicon voxels that are only adjacent to the closed-pore regions and not connected to the external interconnected silicon network structure are subtracted from the total silicon volume to obtain the corrected effective silicon content.

[0084] Specifically, after constructing the three-dimensional voxel model, voxel clusters with gray values ​​higher than the theoretical upper limit of silicon particles can be identified and judged as impurities. When applying the continuous homology algorithm for calculation, the impurities are masked to eliminate the interference of high atomic number impurities on topological connectivity analysis.

[0085] In step 3, to address the low image contrast at the nanoscale, the deep learning semantic segmentation algorithm can incorporate an adversarial generative network architecture. Specifically, the improved U-Net 3D convolutional neural network can be used as a generator, and a discriminator network can be constructed. The discriminator network distinguishes whether the segmentation result was generated by the generator or labeled by experts. Through game-like training between the generator and the discriminator network, the generator can be forced to produce a segmentation result mask with higher physical fidelity. During training, a gradient penalty term can be introduced to ensure training stability. Simultaneously, during the training phase of the deep learning semantic segmentation algorithm, operations such as rotation, translation, scaling, and noise addition are performed on the labeled samples to perform data augmentation.

[0086] For porous silicon materials with internal pore structures, the deep learning semantic segmentation algorithm is used to identify closed-pore regions within the silicon framework. When quantitatively evaluating the effective silicon content participating in electrochemical reactions, the volume of silicon voxels that are only adjacent to the closed-pore regions and not connected to the external interconnected silicon network structure is subtracted from the total silicon volume to obtain the corrected effective silicon content.

[0087] During step 5, the continuous cohomology algorithm can introduce a weighted complex construction strategy. Considering the thickness differences of the carbon coating layer on the silicon particle surface, each centroid point cloud data point can be assigned a weight, which is proportional to the average gray value of the local voxel. When constructing the simple complex, the connection criterion between two particles no longer depends solely on the Euclidean distance, but is based on a generalized distance function that takes into account the weights. This method can use the generalized distance function to simulate the tunneling and conduction probabilities of electrons between carbon layers of different thicknesses to define the connected silicon network structure.

[0088] During the evaluation process in step 6, fractal dimension calculation can be further incorporated. Multi-scale analysis of the identified connected silicon network structure is performed using box counting to obtain the fractal dimension of the connected silicon network structure. A higher fractal dimension indicates a more complete spatial filling and a larger contact area of ​​the connected silicon network structure.

[0089] Subsequently, the effective silicon content, the spectral characteristics of the zero-dimensional Betti number and the one-dimensional Betti number, and the fractal dimension can be input into a pre-trained support vector machine regression model. The support vector machine regression model performs nonlinear mapping calculations and outputs prediction results. The prediction results may include the predicted values ​​of the first coulombic efficiency and cycle stability of the electrode material of the batch to which the solid electrode sample belongs.

[0090] Example 3: In Example 3, the detection method provided by the present invention is integrated into an industrialized online quality control system for solid-state battery production. During the execution of step 1, the acquisition process is optimized to a fast scanning mode. By reducing the number of projections (e.g., acquiring only 600 projections) and using an iterative reconstruction algorithm based on total variation (TV) regularization, the single detection time is shortened to less than 10 minutes while ensuring image quality.

[0091] In step 5, to meet real-time requirements, a parallelized simplicial complex construction algorithm can be used. The distance matrix between point clouds is calculated concurrently using the GPU's computing cores, and the simplicial complex is simplified using discrete Morse theory to remove redundant simplicials that do not change the topological properties, thereby improving the calculation speed of the zero-dimensional Betti number and the one-dimensional Betti number.

[0092] This embodiment focuses on implementing the process feedback mechanism in step 6. The evaluation results of the effective silicon content and the identification results of the isolated silicon particles can be transmitted in real time to the upstream slurry preparation system via the industrial Ethernet protocol.

[0093] When the isolation degree of silicon particles in the silicon particle region (i.e., the ratio of the zero-dimensional Betti number to the total volume of the silicon particles) exceeds a preset alarm threshold, the control system can automatically increase the shear rate of the mixer or extend the ultrasonic dispersion time based on the silicon particle isolation degree to improve the uniformity of silicon particle distribution in the slurry. When the connectivity of the interconnected silicon network structure is insufficient (i.e., the one-dimensional Betti number is lower than a preset connectivity threshold), the control system can automatically adjust the drying temperature profile of the coating machine to slow down the solvent evaporation rate, thereby avoiding particle agglomeration caused by shrinkage stress. This closed-loop control mode improves the batch stability of solid-state electrode materials.

[0094] To further elaborate on the technical details, the following is a description of each key algorithm module:

[0095] In the deep learning semantic segmentation algorithm described in step 3, the specific operation of the convolutional layer is to perform element-wise multiplication and summation between a three-dimensional weight tensor and a local region in the three-dimensional voxel model to obtain a feature map. The pooling layer reduces the data dimensionality by selecting the maximum value within a specific window. The upsampling process restores the spatial size of the feature map to its original size through interpolation or transposed convolution.

[0096] In the attention mechanism module, the three-dimensional feature map is compressed into a one-dimensional vector by global average pooling. Each element in the one-dimensional vector represents the global information of a channel. Then, the interdependencies between channels are calculated through two consecutive fully connected layers, a set of weight coefficients are output, and the weight coefficients are multiplied by the original channel features one channel at a time.

[0097] In the continuous homology algorithm described in step 5, the construction logic of the simplex complex is as follows: For any two points in the centroid point cloud data, if the Euclidean distance between the two points is less than twice the currently set radius parameter, a line segment is connected between the two points to form a one-dimensional simplex; for any three points, if the distances between any two of the three points satisfy the set conditions, the triangular region formed by these three points is filled, forming a two-dimensional simplex. This process is repeated to construct higher-dimensional complex structures. The generation of topological features is defined as the emergence of a new homology class, while its extinction is defined as the homology class becoming the boundary of another homology class or merging.

[0098] In the color mapping process described in step 6, a mapping function is established. This function converts the volume of the connected cluster to which each silicon particle belongs into coordinates in color space. The largest connected cluster is mapped to the red end of the spectrum, while isolated silicon particles with volumes smaller than the average and not connected to other particles are mapped to the cool end of the spectrum based on their volume. Using this color mapping mechanism, engineers can readily identify electrochemical dead zones within the solid-state electrode sample.

[0099] To address the various composite materials that may exist in solid-state electrodes, such as silicon-oxygen-carbon (SiOx / C) composites, this method adds subcategories to the classification logic of the deep learning semantic segmentation algorithm in step 3. The deep learning semantic segmentation algorithm is configured for multi-class output to identify pure silicon particles and the silicon oxide phase. By calculating the topological nesting relationship between the silicon oxide phase and the pure silicon phase, the uniformity of the pre-lithiation process can be evaluated.

[0100] Furthermore, the present invention also relates to a data preprocessing mechanism. In the reconstructed internal three-dimensional density distribution data, if a voxel cluster with a gray value much higher than the theoretical upper limit of silicon particles is detected, the algorithm can automatically identify it as an impurity and perform masking processing on the impurity in subsequent topology calculations to prevent the impurity from misleading connectivity analysis.

[0101] At the software architecture level, this detection method is developed based on a distributed architecture. The front end is responsible for 3D visualization rendering and interactive control, while the back end is deployed on a high-performance computing cluster, responsible for deep learning inference and algebraic topology calculations on massive voxel data. Data is transmitted between the front end and the back end via compressed binary streams. The system supports concurrent access by multiple users and summarizes the total silicon content, effective silicon content, average particle size, and distribution curves of the zero-dimensional Betti number and the one-dimensional Betti number in the generated detection report.

[0102] During steps 1 to 6, the detection method can be extended to in-situ non-destructive monitoring at the whole battery level. During battery charging and discharging, synchrotron high-energy X-rays are used to penetrate the battery's stainless steel casing, and real-time data on the structural evolution of the electrodes inside the solid-state electrode sample is acquired. By applying steps 1 to 6, the entire process of structural compression caused by the expansion of silicon particle regions, the rupture of the carbon coating layer, and the resulting failure of the interconnected silicon network structure can be dynamically observed.

[0103] In the process of constructing the three-dimensional voxel model in step 2, the alignment and unification of the coordinate system are involved. Typically, a local coordinate system is established with the central axis of the rotary stage as the Z-axis and the detector plane as the XY plane. For the solid-state electrode sample with a large aspect ratio, a helical scanning method can be used, and the continuous three-dimensional structure over an ultra-long range can be obtained through data interpolation and reconstruction of the helical trajectory.

[0104] To enhance the robustness of the deep learning semantic segmentation algorithm described in step 3, during the training phase, operations such as rotation, translation, scaling, and adding Gaussian noise can be performed on the labeled samples to adapt to differences in image quality generated by different devices. Simultaneously, knowledge distillation technology is introduced to transfer the segmentation capabilities of the teacher network to the lightweight student network, thereby improving inference speed while maintaining accuracy.

[0105] During step 4, which involves extracting particle information, a watershed segmentation algorithm based on distance transformation can be used to address the particle adhesion problem. First, the distance from each voxel within the silicon particle region to its nearest boundary is calculated. The point with the maximum local distance is then identified as a seed point. The algorithm then expands outward from the seed point until the boundaries of different particles meet, thus distinguishing physically adjacent but logically independent silicon particles.

[0106] For the persistent barcode generated by the persistent cohomology algorithm described in step 5, the segment length is obtained by subtracting the starting point value from the ending point value. The segment length represents the spatial span of the topological feature. For the connected silicon network structure, a longer zero-dimensional barcode represents a large-sized connected cluster, while a longer one-dimensional barcode represents a large-sized inactive pore within the material. By integrating the barcode length, the persistent entropy can be obtained, which characterizes the degree of disorder in the silicon particle distribution.

[0107] For the three-dimensional visualization distribution map described in step 6, ambient light occlusion and shadow mapping techniques can be used to enhance the sense of hierarchy in the interconnected silicon network structure by simulating multiple reflections of light within the micropores. Furthermore, the three-dimensional voxel model can be exported as a standard triangular mesh format (such as STL or OBJ) for subsequent finite element analysis to simulate the mechanical stress distribution of silicon particles during lithium intercalation and expansion.

[0108] The invention also includes an environmental factor compensation mechanism. During the scanning process, temperature fluctuations are recorded in real time, and the center coordinates of each projected image are dynamically corrected according to a preset coefficient of thermal expansion to eliminate displacement deviations caused by thermal expansion and contraction.

[0109] In terms of data security and storage, the system adopts a cold and hot data classification storage strategy. The original projection data is transferred to a cold storage medium after reconstruction; the reconstructed three-dimensional voxel model and the segmented feature parameters are stored in a high-performance solid-state drive array to meet the needs of fast retrieval and secondary analysis.

[0110] In summary, this invention integrates advanced non-destructive imaging, deep learning recognition, and algebraic topology analysis techniques to construct a complete closed-loop system for detecting silicon content in solid-state electrode materials. From the physical properties of microscopic voxels to the topological functions of macroscopic networks, it achieves quantitative characterization across all scales and dimensions.

[0111] Example 4: In Example 4, the detection method provided by this invention was used to study the effect of different binder systems on the connectivity of silicon particles. Three common binders were selected for the experiment: polyvinylidene fluoride (PVDF), sodium carboxymethyl cellulose (CMC), and polyacrylic acid (PAA).

[0112] Following step 1, simultaneous data acquisition was performed on electrodes using the three binder formulations. During step 2, a standardized grayscale threshold process was used to ensure benchmark consistency in comparisons between different samples.

[0113] During step 3, the contrast of different binders in CT images varies. For example, PAA, due to its higher concentration of oxygen-containing groups, absorbs X-rays slightly better than PVDF. The deep learning semantic segmentation algorithm employs a transfer learning strategy, fine-tuning the parameters of the last convolutional layer for each binder to accommodate the subtle differences in X-ray absorption characteristics among different binders.

[0114] In the topological connectivity assessment performed in step 5, the results showed that the electrode using PAA binder exhibited the most gradual change in the one-dimensional Betti number of the silicon particle region with charge-discharge cycles, indicating that the PAA binder can maintain the stability of the interconnected silicon network structure. In contrast, the electrode using PVDF binder showed an increase in the zero-dimensional Betti number after cycling, quantitatively confirming the peeling phenomenon of the isolated silicon particles.

[0115] During step 6, the three-dimensional visualization distribution map generated by this invention can display the network framework formed by silicon particles in the PAA electrode. By extracting the decay curve of the effective silicon content with the cycle period, it was found that the rate of decrease of the effective silicon content has a linear correlation with the battery capacity decay curve, with a correlation coefficient exceeding 0.95. Using the effective silicon content index based on topological connectivity, quantitative prediction of battery life is achieved.

[0116] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this application. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but without changing the core design of the algorithm and process, are also within the scope of protection of this application.

[0117] Furthermore, some embodiments of this application also provide an electronic device. The electronic device can be various forms of digital computer, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, etc. The electronic device can also be various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices.

[0118] The electronic device includes: one or more processors; and a memory storing computer program instructions that, when executed, cause the processor to perform the steps of the methods provided in any one or more of the above embodiments. Figure 6An exemplary structural diagram of the electronic device is disclosed. The electronic device includes one or more processors 1101, a memory 1102, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations. The components, their connections and relationships, and their functions shown herein are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0119] The electronic device may further include an input device 1103 and an output device 1104. The processor 1101, memory 1102, input device 1103 and output device 1104 may be connected by a bus or other means, as shown in the figure, which is connected by a bus.

[0120] Input device 1103 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as a touch screen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 1104 may include a display device, auxiliary lighting device (e.g., LED), and haptic feedback device (e.g., vibration motor). The display device may include, but is not limited to, a liquid crystal display, a light-emitting diode display, and a plasma display. In some embodiments, the display device may be a touch screen.

[0121] To provide interaction with the user, the electronic device can be a computer. The computer has: a display device (e.g., a cathode ray tube or LCD monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback); and input from the user can be received in any form (e.g., voice input or tactile input).

[0122] In this embodiment, a computer-readable medium stores a computer program / instructions that, when executed by a processor, implement the steps of the methods provided in any one or more of the above embodiments. This computer-readable medium may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into that device. The aforementioned computer-readable medium carries one or more computer-readable instructions.

[0123] The memory 1102 can serve as a non-transitory computer-readable storage medium, used to store non-transitory software programs, non-transitory computer-executable programs, and modules. The processor 1101 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 1102, thereby implementing the program instructions / modules corresponding to the methods provided in any one or more of the embodiments described above in this application.

[0124] The memory 1102 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 1102 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 1102 may optionally include memory remotely located relative to the processor 1101, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0125] It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0126] Computer-readable media include permanent and non-permanent, removable and non-removable media, which can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory, static random access memory, dynamic random access memory, other types of random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other memory technologies, read-only optical discs, digital versatile optical discs or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0127] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including local area networks (LANs) or wide area networks (WANs), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0128] In the above embodiments, all or part of the implementation can be achieved through software, hardware, firmware, or any combination thereof. For example, it can be implemented using an application-specific integrated circuit (ASIC), a general-purpose computer, or any other similar hardware device. In some embodiments, the software program of this application can be executed by a processor to implement the above steps or functions. Similarly, the software program of this application (including related data structures) can be stored in a computer-readable recording medium, such as RAM memory, magnetic or optical drives, floppy disks, and similar devices. In addition, some steps or functions of this application can be implemented in hardware, for example, as circuitry that cooperates with a processor to perform the various steps or functions.

[0129] The computer program product provided in this application includes one or more computer programs / instructions. When executed by a processor, these computer programs / instructions generate, in whole or in part, the processes or functions described in this application. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0130] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0131] The scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to encompass all variations falling within the meaning and scope of equivalents of the claims. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device in software or hardware. Terms such as "first," "second," etc., are used only for distinguishing descriptions and do not indicate any particular order, nor should they be construed as indicating or implying relative importance.

[0132] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily made by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims, and the above embodiments should be regarded as exemplary and non-limiting.

Claims

1. A method for detecting silicon content in a solid electrode material, characterized in that, Includes the following steps: Non-destructive X-ray microtomography was performed on the solid electrode sample to obtain projection data reflecting the absorption characteristics of the internal material of the solid electrode sample, and the internal three-dimensional density distribution data of the solid electrode sample was generated by the back projection reconstruction algorithm. Based on the aforementioned internal three-dimensional density distribution data, a three-dimensional voxel model is constructed; The three-dimensional voxel model was used to identify multiphase components by using a deep learning semantic segmentation algorithm to distinguish the silicon particle region, the carbon coating layer region, and the binder region. Based on the segmentation results of multiphase component identification, the spatial location, geometry, and volume information of silicon particles within the silicon particle region are extracted. The Betti number of the silicon particle region is calculated based on the spatial location of the silicon particle to identify isolated silicon particles and interconnected silicon network structures. Based on the identification results of the isolated silicon particles and the interconnected silicon network structure, the effective silicon content participating in the electrochemical reaction is quantitatively assessed, and a three-dimensional visualization distribution map reflecting the distribution state of the silicon particles is generated.

2. The method for detecting silicon content in a solid electrode material according to claim 1, characterized in that, The step of using a deep learning semantic segmentation algorithm to identify multiphase components of the three-dimensional voxel model includes: The deep learning semantic segmentation algorithm is based on an encoder and decoder architecture. The encoder includes a series of three-dimensional convolutional layers, residual connection structures, and max pooling layers. The three-dimensional convolutional layers perform element-wise multiplication and summation of the weight tensors with the local regions of the three-dimensional voxel model to extract multi-scale features and obtain encoded geometric features. The decoder performs upsampling through transposed convolution and fuses the geometric features in the encoder with the semantic information in the decoder through skip connection paths to obtain a fused feature map; The skip connection path integrates an attention mechanism module, which focuses on the boundary between the silicon particle region and the carbon coating layer region by calculating the correlation between each pixel in the fused feature map, and outputs the predicted boundary. The loss function of the deep learning semantic segmentation algorithm adopts a weighted combination of cross-entropy loss function, similarity coefficient loss function and boundary-aware loss term. The boundary-aware loss term enhances the recognition accuracy of thin-layer structures by calculating the distance between the predicted boundary and the real boundary.

3. The method for detecting silicon content in a solid electrode material according to claim 1, characterized in that, The segmentation results based on multiphase component identification extract the spatial location, geometry, and volume information of silicon particles within the silicon particle region, including: Using a three-dimensional connected component labeling algorithm, silicon voxels within the identified silicon particle region are classified based on the connectivity principle, and a unique identifier is assigned to each independent silicon particle. For each particle corresponding to the identification number, the centroid coordinates are calculated by arithmetically averaging the coordinates of the voxel to which it belongs, and the volume information of the silicon particle is calculated by combining the total number of silicon voxels with the physical volume of a single silicon voxel. Based on the centroid coordinates and the silicon particle volume information, the geometry of the silicon particle is calculated; the geometry of the silicon particle includes the equivalent spherical diameter, surface area, and the ratio of surface area to volume. Anisotropic parameters are obtained by calculating the eigenvalues ​​of the particle inertia tensor, which are used to evaluate the extension of the silicon particle in the directions of the major axis, median axis and minor axis. The extracted anisotropic parameters and the geometry of the silicon particle are stored in a database.

4. The method for detecting silicon content in a solid electrode material according to claim 1, characterized in that, The calculation of the Betti number of the silicon particle region based on the spatial location of the silicon particle, and the identification of isolated silicon particles and interconnected silicon network structures, includes: The extracted spatial locations of the silicon particles are converted into centroid point cloud data as input. A sphere with a radius varying with the filtering parameters is placed at each centroid point to construct a series of nested simple complexes. During the evolution of the simplex, as the filtering parameter increases, when the distance between any two points is less than twice the filtering parameter, a one-dimensional simplex is formed by connecting the two points. When the pairwise distances between any three points satisfy the requirements of the filtering parameters, a two-dimensional simplex is formed. During the simple complex evolution process, the generation and destruction of topological features are monitored in real time, and the zero-dimensional Betti number representing the number of connected components and the one-dimensional Betti number representing the number of loops are calculated. Based on the lifecycle span of the filtering parameters during the change process, the system distinguishes between connected silicon network structures with robust topology and isolated silicon particles that do not meet the preset connectivity scale threshold.

5. The method for detecting silicon content in a solid electrode material according to claim 1, characterized in that, The quantitative assessment of the effective silicon content participating in the electrochemical reaction based on the identification results of the isolated silicon particles and the interconnected silicon network structure includes: Based on the identification results of the interconnected silicon network structure, the total volume of silicon particles belonging to the largest connected cluster in the interconnected silicon network structure, or the total volume of silicon particles that meet the preset topological connectivity threshold, is defined as the effective silicon volume. The proportion of the effective silicon volume to the total volume of all silicon particles is calculated, and the performance of the electrode material is evaluated in combination with the spatial distribution uniformity index. The spatial distribution uniformity index is obtained by dividing the three-dimensional voxel model into multiple sub-regions and statistically analyzing the deviation of silicon content in each sub-region. The results of evaluating the electrode material performance are correlated with the three-dimensional spatial location of the three-dimensional voxel model. The three-dimensional visualization distribution map is generated using a color mapping method, and silicon particles with different connectivity states are mapped to color ranges. The interconnected silicon network structure is marked with warm colors, and the isolated silicon particles are marked with cool colors. An interactive cross-sectional observation interface is provided.

6. The method for detecting silicon content in a solid electrode material according to claim 1, characterized in that, The method further includes: Scanning data of the same solid electrode sample at different electrochemical cycles were used as input. Spatial registration was performed using rigid registration based on mutual information and non-rigid registration based on adaptive grids to eliminate displacement and deformation of the solid electrode sample during assembly and disassembly, and to obtain an aligned time-series three-dimensional model. Based on the aligned time-series 3D model, the changes in topological features at different time points are compared to quantify the degree of fragmentation, detachment behavior, and volume evolution of the silicon particles during the cycling process. Based on the quantification results, a correlation model between microstructure degradation and battery capacity decay is constructed.

7. A method for detecting silicon content in a solid electrode material according to any one of claims 1 to 6, characterized in that, The method further includes: After constructing the three-dimensional voxel model, voxel clusters with gray values ​​higher than the theoretical upper limit of silicon particles are identified and determined to be impurities. When applying the continuous homology algorithm for calculation, the impurities are masked. During the training phase of the deep learning semantic segmentation algorithm, operations such as rotation, translation, scaling, and adding noise are performed on the labeled samples to perform data augmentation. For porous silicon materials with internal pore structures, the closed-pore regions inside the silicon framework are identified by the deep learning semantic segmentation algorithm. When quantitatively evaluating the effective silicon content participating in the electrochemical reaction, the volume of silicon voxels that are only adjacent to the closed-pore regions and not connected to the external interconnected silicon network structure are subtracted from the total silicon volume to obtain the corrected effective silicon content.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; and A memory storing computer program instructions, which, when executed, cause the processor to perform the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images