Photoelastic sample crystallinity multi-modal data fusion analysis method and system

Abstract

The invention discloses a photoelastic sample crystallinity multi-modal data fusion analysis method and system, and the method comprises the steps: recognizing photoelastic particles from a two-dimensional image of a photoelastic sample, and extracting the center-of-mass coordinates and a strong chain network of each photoelastic particle; screening out the photoelastic particles of which the local area fractions and the bond orientation sequence parameters meet corresponding threshold values, and jointly defining the screened photoelastic particles and all the photoelastic particles in direct contact with the screened photoelastic particles as a structural crystal particle set; connecting the mass centers of the adjacent photoelastic particles in the strong chain network into a line segment, constructing a strong chain skeleton, and screening out a long force chain; defining the photoelastic particles covered by the long force chains belonging to the same cluster as a force chain crystal particle set; and calculating the fusion crystallinity of the photoelastic sample based on the proportions of the two sets in the total number of the photoelastic particles. According to the method, the limitation of a traditional single analysis method is overcome, and an accurate and reliable technical means is provided for quantitative analysis of the crystallinity of the photoelastic sample and the discrete element particle sample.

Description

Technical Field

[0001] This invention relates to the field of photoelastic particle testing technology, and more specifically to a method and system for multimodal data fusion analysis of the crystallinity of photoelastic samples. Background Technology

[0002] Photoelasticity experiments are an important technique for investigating the force chain network behavior of particulate materials. By visualizing the force transmission path through the stress birefringence effect in amorphous materials, they effectively reveal the evolution characteristics of the force chain network at the microscale in particulate media. Under gravity, single-graded particles easily form locally ordered crystalline arrangements, leading to abnormal stress concentration within the crystalline particles. This significantly interferes with the overall force chain pattern observed in photoelastic images, affecting the accurate characterization of the true force chain network state of particulate materials. Therefore, quantitative analysis of the crystallinity in photoelastic samples is a crucial prerequisite for studying the force chain network behavior of particulate materials.

[0003] Currently, crystallinity analysis based on photoelastic images mostly employs a single structural mode, focusing on identifying the geometrical order of individual particles, and has not developed a reasonable quantitative index for sample crystallinity. Furthermore, differences in loading methods can significantly alter the morphology of the force chain network under the same structure; geometrically ordered regions may not form a force chain network under pressure. Therefore, analyzing the crystallinity of photoelastic samples solely from structural modes is one-sided and limited.

[0004] Therefore, how to establish comprehensive and accurate quantitative indicators for crystallinity to solve the problem that existing crystallinity analysis based on photoelastic images relies only on a single structural mode and does not integrate information on particle geometric order and force chain network order, resulting in a one-sided assessment of the crystallization state of the system and inaccurate quantitative indicators, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of the above problems, the present invention proposes a method and system for multimodal data fusion analysis of crystallinity of photoelastic samples to overcome or at least partially solve the above problems.

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

[0007] In a first aspect, the present invention provides a method for multimodal data fusion analysis of crystallinity in photoelastic samples, comprising: Two-dimensional images of the photoelastic sample are acquired, photoelastic particles are identified from the two-dimensional images, and the centroid coordinates and strong chain network of each photoelastic particle are extracted. Based on the centroid coordinates, Voronoi cells are divided, and the local area fraction and bond orientation order parameters of each photoelastic particle are calculated. Photoelastic particles that satisfy the corresponding screening thresholds in terms of local area fraction and bond orientation order parameter are selected. The selected photoelastic particles and all photoelastic particles that are in direct contact with the selected photoelastic particles are collectively defined as a set of structural crystalline particles. Connect the centroids of adjacent photoelastic particles in the strong chain network as line segments to construct a strong chain skeleton; Based on the pointing order parameter between adjacent line segments in the strong chain skeleton, adjacent line segments that meet the merging conditions are merged into a force chain, and long force chains are selected based on the average number of photoelastic particles that make up each force chain. Cluster the directional vectors of the long force chains, and define the photoelastic particles covered by the long force chains belonging to the same cluster as the force chain crystallization particle set. The degree of fusion crystallinity of the photoelastic sample is calculated based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

[0008] Furthermore, the strong chain network of each photoelastic particle is extracted based on preset conditions; The preset conditions include: the contact force between photoelastic particles is higher than the average value, the angle between the main contact force direction and the line connecting the centroid is less than 45°, and the number of consecutive photoelastic particles is greater than 3.

[0009] Furthermore, the screening threshold corresponding to the local area fraction is greater than 0.8.

[0010] Furthermore, the screening threshold corresponding to the key orientation sequencing parameter is greater than 0.85.

[0011] Furthermore, the step of merging line segments into multiple force chains based on the pointing order parameter between adjacent line segments in the strong chain skeleton, and selecting long force chains based on the average number of photoelastic particles constituting each force chain; specifically including: Calculate the pointing order parameter between adjacent line segments in the strong chain skeleton; The pointing sequence parameter is judged as follows: when the pointing sequence parameter is greater than or equal to a preset threshold, the adjacent line segments are determined to belong to the same force chain, and the number of photoelastic particles that make up the force chain is recorded; when the pointing sequence parameter is less than the preset threshold, or when the force chain branches, it is determined to be the endpoint of the force chain. Calculate the average number of photoelastic particles that make up each force chain, and define the force chain with a number greater than the average number of photoelastic particles as a long force chain.

[0012] Furthermore, the clustering of the direction vectors of the long force chain specifically includes: The angles between each long force chain and the positive x-axis are weighted and averaged to obtain the corresponding long force chain direction angles. The orientation angles of all long force chains are standardized and converted into two-dimensional coordinate vectors. Using the two-dimensional coordinate vector as input, the neighborhood radius and minimum number of samples are set, and the DBSCAN clustering algorithm is used for clustering.

[0013] Furthermore, the fused crystalline particle set constituted by the aforementioned fusion crystallinity satisfies the following relationship with the structural crystalline particle set and the force chain crystalline particle set:

[0014] in, T Indicates a collection of fused crystalline particles; C Represents a collection of crystalline particles; F Represents a set of force chain crystalline particles; ∨ is an OR relation used to indicate that fused crystalline particles belong to a set of structural crystalline particles. C Or force chain crystallized particle aggregate F .

[0015] Furthermore, the degree of crystallinity of fusion is expressed as

[0016] in, Indicates the degree of crystallinity; This indicates the number of photoelastic particles in a set of structural crystalline particles; Indicates the number of photoelastic particles in the force chain crystalline particle set; Indicates the total number of photoelastic particles; This refers to photoelastic particles that represent the overlap of both the structural crystalline particle set and the force chain crystalline particle set.

[0017] Secondly, this invention provides a multimodal data fusion analysis system for the crystallinity of photoelastic samples, applying the above-mentioned method; the system includes: Image acquisition module, used to acquire two-dimensional images of the photoelastic sample; An image processing module is used to identify photoelastic particles from the two-dimensional image and extract the centroid coordinates and strong chain network of each photoelastic particle. The structural analysis module is used to divide Voronoi cells based on the centroid coordinates, calculate the local area fraction and bond orientation sequence parameter of each photoelastic particle, and screen out photoelastic particles whose local area fraction and bond orientation sequence parameter both meet the corresponding screening threshold. The screened photoelastic particles and all photoelastic particles that are in direct contact with the screened photoelastic particles are collectively defined as a set of structural crystalline particles. The force chain analysis module is used to connect the centroids of adjacent photoelastic particles in the strong force chain network into line segments to construct a strong force chain skeleton; based on the directional order parameter between adjacent line segments in the strong force chain skeleton, adjacent line segments that meet the merging conditions are merged into force chains, and long force chains are selected based on the average number of photoelastic particles that make up each force chain; the direction vectors of the long force chains are clustered, and the photoelastic particles covered by the long force chains belonging to the same cluster are defined as the force chain crystallized particle set. The fusion calculation module is used to calculate the fusion crystallinity of the photoelastic sample based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

[0018] Thirdly, the present invention provides a terminal, including: a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-mentioned method for multimodal data fusion analysis of crystallinity of photoelastic samples.

[0019] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for multimodal data fusion analysis of crystallinity of photoelastic samples, which has the following beneficial effects: This invention overcomes the limitations of traditional single analysis methods by combining the structural characteristics of photoelastic samples with force chain network modal information, providing a precise and reliable technical means for the quantitative analysis of crystallinity of photoelastic samples and discrete element particle samples. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the process for the multimodal data fusion analysis of crystallinity of photoelastic samples provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the outline of a discrete element particle sample after identification, provided in Embodiment 1 of the present invention. Figure 3 This is a schematic diagram of the Voronoi cell boundary division of the discrete element particle sample provided in Embodiment 1 of the present invention. Figure 4 This is a schematic diagram of the calculation of local area fraction and bond orientation sequence parameter provided in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of discrete element particles for screening composite structure threshold requirements provided in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the force chain skeleton of the discrete element particle sample provided in Embodiment 1 of the present invention; Figure 7 This is a schematic diagram of the long force chain of the discrete element particle sample provided in Embodiment 1 of the present invention; Figure 8 This is a schematic diagram of the force chain network crystalline particles of the discrete element particle sample provided in Embodiment 1 of the present invention; Figure 9 This is a schematic diagram of images taken during an additional photoelastic experiment provided in Embodiment 2 of the present invention; Figure 10 This is a schematic diagram of the two-dimensional profile of the photoelastic test particle sample provided in Embodiment 2 of the present invention; Figure 11 This is a schematic diagram of the Voronoi cell boundary of the photoelastic sample provided in Embodiment 2 of the present invention; Figure 12 This is a schematic diagram of the modal crystalline particles of the photoelastic sample structure provided in Embodiment 2 of the present invention; Figure 13 This is a schematic diagram of the force chain skeleton of the photoelastic sample provided in Embodiment 2 of the present invention; Figure 14 This is a schematic diagram of the long force chain after cluster analysis of the photoelastic sample provided in Embodiment 2 of the present invention; Figure 15 This is a schematic diagram of the modal crystallization particles of the force chain network in the photoelastic sample provided in Embodiment 2 of the present invention. Detailed Implementation

[0022] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention discloses a multimodal data fusion analysis method for the crystallinity of photoelastic samples, which can provide multimodal data fusion analysis for the quantification of crystallinity of discrete element particle samples. For example... Figure 1 As shown, the method includes the following steps: S1. Obtain a two-dimensional image of the photoelastic sample, identify the photoelastic particles from the two-dimensional image, and extract the centroid coordinates and strong chain network of each photoelastic particle. S2. Based on the centroid coordinates, perform Voronoi cell division and calculate the local area fraction and bond orientation order parameter of each photoelastic particle; S3. Select photoelastic particles whose local area fraction and bond orientation order parameters both meet the corresponding screening thresholds. Define the selected photoelastic particles and all photoelastic particles that are in direct contact with the selected photoelastic particles as a set of structural crystalline particles. S4. Connect the centroids of adjacent photoelastic particles in the strong chain network into line segments to construct the strong chain framework; S5. Based on the pointing order parameter between adjacent segments in the strong chain skeleton, adjacent segments that meet the merging conditions are merged into a force chain, and long force chains are selected based on the average number of photoelastic particles that make up each force chain. S6. Cluster the directional vectors of the long force chains and define the photoelastic particles covered by the long force chains belonging to the same cluster as the force chain crystallization particle set. S7. Calculate the degree of fusion crystallinity of the photoelastic sample based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

[0024] It should be noted that the above labels S1-S7 are only for the convenience of subsequent explanation and do not limit the implementation order of each step.

[0025] This invention provides a multimodal data fusion analysis method for the crystallinity of photoelastic samples, used to quantitatively evaluate the particle packing state of photoelastic samples. By fusing structural and force chain network ordering information, it overcomes the limitations of single-modal analysis and significantly improves the accuracy and reliability of crystallinity assessment. The term "photoelastic sample" in this invention broadly refers to systems with particle packing characteristics, including but not limited to equivalent particle samples generated by discrete element numerical simulation and physical photoelastic test samples.

[0026] Example 1: Using the "photoelastic sample" as an equivalent particle sample generated by discrete element numerical simulation, the above steps are explained in detail.

[0027] In step S1 above, a two-dimensional particle image of the sample is obtained, and the particle positions and strong chain networks are extracted; specifically including: (1) The particle outline of the particle sample was identified from the captured image using the gray-scale second-order processing method; (2) Based on the particle profile, determine the centroid coordinate information of all particles in the sample; (3) Based on the generalized definition of strong chain network, namely, the contact force between particles is higher than the mean, the angle between the main contact force direction and the line connecting the centroid is less than 45° and the number of consecutive particles is greater than 3, the strong chain network is identified and extracted; among them, the limits on the line angle and the number of consecutive particles are general parameters.

[0028] As an example, this embodiment uses the discrete element method software PFC3D to generate a particle sample model with dimensions of 18cm × 18cm for example analysis. The particles are circular in shape, and each particle has a diameter of 1.2cm. The processed particle outline is as follows. Figure 2 As shown.

[0029] In step S2 above, Voronoi cells are divided based on the centroid coordinates of the particles, and the local area fraction and bond orientation order parameters of each particle are calculated; specifically including: (1) The Voronoi cell boundary and local area fraction were determined using the Euclidean distance method. The calculation expression is:

[0030] in, S p The area of ​​the particle is obtained from particle contour recognition; S c This represents the area of ​​the corresponding Voronoi cell. (2) The bond orientation order parameter is a commonly used parameter for evaluating the hexagonal symmetry of particle arrangement. The expression for the bond orientation order parameter in a two-dimensional particle system is:

[0031] Among them, particles i The number of neighboring particles sharing the same edge in a Voronoi cell is N ( i ). θ ij Particles i With neighboring particles j The angle between the neighborhood bonds and the positive x-axis. Since the number of neighboring particles of edge particles is far less than that of internal particles, the statistical sample for calculating bond orientation ordering parameters is insufficient, leading to distortion of results such as local area fractions. Therefore, the analysis method disclosed in this invention removes edge particles that are severely truncated by boundaries, retaining only particles with complete neighborhoods for calculation.

[0032] Figure 3 A particle sample after demarcating Voronoi cell boundaries. A certain particle... i A schematic diagram illustrating the calculation of local area fraction and bond orientation ordering parameters is shown below. Figure 4 As shown, particles i Local area fraction bond orientation sequence parameter ψ 6( i The calculated results are 0.78 and 0.62, respectively.

[0033] In step S3 above, photoelastic particles whose local area fraction and bond orientation order parameters both satisfy the corresponding screening thresholds are selected. The selected photoelastic particles and all photoelastic particles in direct contact with the selected photoelastic particles are collectively defined as a set of structural crystalline particles, specifically including: The screening thresholds for local area fraction and bond orientation ordering parameters in structural modes are as follows:

[0034]

[0035] The local area fraction of 0.8 is an approximation of the volume fraction (0.785) when circular particles are packed in a square shape, and the bond orientation order parameter of 0.85 is a commonly used threshold in the field of materials science for judging the symmetry and order of particles. The number of neighboring particles of edge particles is far less than that of internal particles, resulting in insufficient statistical samples for calculating the bond orientation order parameter, and distorted characteristics of results such as the local volume fraction. Therefore, the analytical method disclosed in this invention removes edge particles that are severely truncated by boundaries, retaining only particles with complete domains for calculation. Particles whose local area fraction and bond orientation ordering parameters both meet preset thresholds are selected, and their contact particles are identified. The sum of the two quantities is then calculated. N S Defined as a collection of structurally crystalline particles; particles that meet the screening threshold requirements, such as... Figure 5 As shown.

[0036] In step S4 above, the centroids of adjacent particles in the strong chain network are connected as line segments to construct the strong chain framework; specifically, this includes: Based on the strong chain network, the centroid parameters of the particles that make up the strong chain network are recorded, and the centroids of adjacent particles in the force chain are connected as line segments to construct the strong chain skeleton.

[0037] In this embodiment, the constructed force chain skeleton is as follows: Figure 6 As shown.

[0038] In step S5 above, based on the pointing order parameter between adjacent segments in the strong chain skeleton, adjacent segments that meet the merging conditions are merged into a force chain, and long force chains are selected based on the average number of particles that make up each force chain. In force chain network modes, long force chains typically refer to load transfer paths connecting multiple orderly arranged particle contact points. By setting an average length threshold based on statistical distribution, short force chains with disordered orientations can be effectively filtered out, allowing the analysis to focus on long force chain networks. Specifically: (1) Calculate the pointing order parameter between adjacent line segments in the strong chain skeleton. S , is represented as:

[0039] in,α This represents the angle between adjacent force line segments; this formula is a general liquid crystal pointing sequence formula.

[0040] (2) For the pointer order parameter S Perform a judgment: when the pointer is an order parameter S When the value is greater than or equal to a preset threshold, adjacent line segments are determined to belong to the same force chain, and the number of particles constituting that force chain is recorded; when the pointer order parameter... S If the value is less than a preset threshold, or if a branch appears in the force chain, it is determined to be a force chain endpoint; Since 0.7 is a commonly used threshold for the consistency of liquid crystal alignment sequence parameters, the preset threshold can be set to 0.7. (3) Calculate the average number of particles constituting each force chain, and define force chains with a particle number greater than the average as long force chains. The screened long force chains are as follows: Figure 7 As shown.

[0041] In step S6 above, the direction vectors of the long force chains are clustered, and the particles covered by the long force chains belonging to the same cluster are defined as the force chain crystallization particle set. In force chain network modes, long force chains composed of crystalline particles typically extend along specific directions and are arranged in an orderly manner, forming force chain clusters with similar directional vectors. Using the DBSCAN clustering algorithm, natural clusters can be identified based on feature data density. By clustering the directional vectors of long force chains, the "force chain directional consistency" can be quantified, accurately capturing the directional cluster characteristics unique to crystalline regions and effectively screening force chain networks with ordered transmission characteristics. Specifically: (1) Take the weighted average of the angles between each long force chain and the positive x-axis to obtain the corresponding long force chain direction angle. θ rad ; (2) For all long force chain direction angles θ rad Standardize the data and convert it into a two-dimensional coordinate vector:

[0042] in, θ rad The original eigenvalues, i.e., the original long force chain direction angles. θ ; μ σ is the mean of this feature dimension; σ is the standard deviation of this feature dimension. x uniform These are the standardized feature parameter values, i.e., the two-dimensional coordinate vector.

[0043] (3) Using a two-dimensional coordinate vector as input, the neighborhood radius and minimum number of samples are set, and the DBSCAN clustering algorithm is used for clustering. Specifically: the neighborhood radius is set to 0.6, that is, force chains with a weighted direction angle difference of less than 20° are considered to be in the same cluster; the minimum number of samples for clustering is 3, that is, each cluster has at least 3 long force chains; the DBSCAN clustering algorithm is executed, and long force chains belonging to the same cluster are determined to be crystallized force chains with consistent orientation. In this embodiment, the long force chains with consistent orientation after the clustering algorithm is screened are as follows: Figure 8 As shown, the particles it covers are a collection of force chain crystallized particles.

[0044] In step S7 above, the degree of fusion crystallinity of the photoelastic sample is calculated based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of particles; specifically including: (1) Calculate the degree of crystallinity based on the proportion of the aggregate of structural crystalline particles to the total number of particles. ; indicates as:

[0045] in, Indicates the number of particles in a set of crystalline particles; Indicates the total number of particles; like Figure 5 In the structure, the number of particles in the crystalline particle aggregate. The total number of particles is 30. If the crystallinity is 52, then the structural modal crystallinity is:

[0046] (2) Calculate the crystallinity of the force chain network based on the proportion of the force chain crystalline particle set to the total number of particles. ; indicates as:

[0047] in, Indicates the number of particles in a set of force chain crystal particles; like Figure 8 In the middle, the number of particles in the force chain crystallization particle set The total number of particles is 21. If the crystallinity is 52, then the crystallinity of the force chain network is:

[0048] (3) Based on structural modal crystallinity Heli Chain Network Crystallinity Calculate the crystallinity of fusion :

[0049] in, This refers to particles where both the structural crystalline particle set and the force chain crystalline particle set overlap.

[0050] When the crystallinity of structural modes The crystallinity of the force chain network is 57.7%. 40.3% of the particles overlapped. The total number of particles is 20. When the fusion crystallinity is 52, It is 59.6%.

[0051] Example 2: Using the "photoelastic sample" as the physical photoelastic test sample, the above steps are explained in detail.

[0052] In step S1 above, a two-dimensional image of the photoelastic particles in the photoelastic sample is obtained, the photoelastic particles are identified from the two-dimensional image, and the centroid coordinates and strong chain network of each photoelastic particle are extracted; specifically including: (1) The particle outline of the photoelastic particle sample was identified from the captured image using the gray-scale second-order processing method. (2) Based on the photoelastic particle profile, determine the centroid coordinate information of all particles in the sample; (3) Based on the generalized definition of strong chain network, namely, the contact force between particles is higher than the mean, the angle between the main contact force direction and the line connecting the centroid is less than 45° and the number of consecutive particles is greater than 3, the strong chain network is identified and extracted.

[0053] As an example, this embodiment additionally conducts a set of photoelasticity tests for illustrative purposes (such as...). Figure 9 (As shown) Analysis showed that the sample size was 18cm × 18cm, the photoelastic particles were spherical, and the diameter of each photoelastic particle was 0.6cm. The total number of particles was... N T The value is 980, and the processed photoelastic particle profile is as follows: Figure 10 As shown.

[0054] In step S2 above, Voronoi cells are divided based on the centroid coordinates, and the local area fraction and bond orientation order parameters of each photoelastic particle are calculated; specifically including: (1) The Voronoi cell boundary and local area fraction were determined using the Euclidean distance method. The calculation expression is:

[0055] in, S p The area of ​​the photoelastic particle is obtained by particle contour recognition; S c This represents the area of ​​the corresponding Voronoi cell. (2) The bond orientation order parameter is a commonly used parameter for evaluating the hexagonal symmetry of particle arrangement. The expression for the bond orientation order parameter in a two-dimensional particle system is:

[0056] Among them, photoelastic particles i The number of neighboring particles sharing the same edge in a Voronoi cell is N ( i ). θ ij Particles i With neighboring particles j The angle between the neighborhood bond and the positive x-axis.

[0057] Figure 11 The particle sample after dividing the Voronoi cell boundary.

[0058] In step S3 above, photoelastic particles whose local area fraction and bond orientation order parameters both satisfy the corresponding screening thresholds are selected. The selected photoelastic particles and all photoelastic particles in direct contact with the selected photoelastic particles are collectively defined as a set of structural crystalline particles, specifically including: The screening thresholds for local area fraction and bond orientation ordering parameters in structural modes are as follows:

[0059]

[0060] The local area fraction of 0.8 is an approximation of the volume fraction (0.785) when circular particles are packed in a square shape, and the bond orientation order parameter of 0.85 is a commonly used threshold in the field of materials science for judging the symmetry and order of particles. Photoelastic particles whose local area fraction and bond orientation ordering parameters both meet preset thresholds are selected, and their contact particles are identified. The sum of the two quantities is then calculated. N S Defined as a collection of structurally crystalline particles; particles that meet the screening threshold requirements and their contact particles, such as... Figure 12 As shown.

[0061] In step S4 above, the centroids of adjacent photoelastic particles in the strong chain network are connected as line segments to construct the strong chain skeleton; based on the photoelastic particle strong chain network, the centroid parameters of the particles that make up the strong chain network are recorded, and the centroids of adjacent particles in the force chain are connected as line segments to construct the strong chain skeleton.

[0062] In this embodiment, the constructed force chain skeleton is as follows: Figure 13 As shown.

[0063] In step S5 above, based on the pointing order parameter between adjacent segments in the strong force chain skeleton, adjacent segments that meet the merging conditions are merged into force chains, and long force chains are selected based on the average number of photoelastic particles that make up each force chain; specifically including: (1) Calculate the pointing order parameter between adjacent line segments in the strong chain skeleton. S , is represented as:

[0064] in, α Indicates the angle between adjacent force line segments; (2) For the pointer order parameter S Perform a judgment: when the pointer is an order parameter S When the value is greater than or equal to a preset threshold, adjacent line segments are determined to belong to the same force chain, and the number of photoelastic particles constituting that force chain is recorded; when the pointing order parameter S If the value is less than a preset threshold, or if a branch appears in the force chain, it is determined to be a force chain endpoint; Since 0.7 is a commonly used threshold for the consistency of liquid crystal alignment sequence parameters, the preset threshold can be set to 0.7. (3) Calculate the average number of photoelastic particles constituting each force chain, and define force chains with a number greater than the average number of photoelastic particles as long force chains. The screened long force chains are as follows: Figure 14 As shown in the figure, different colors are used to distinguish each long force chain.

[0065] In step S6 above, the direction vectors of the long force chains are clustered, and the photoelastic particles covered by the long force chains belonging to the same cluster are defined as a set of force chain crystalline particles; specifically including: (1) Take the weighted average of the angles between each long force chain and the positive x-axis to obtain the corresponding long force chain direction angle. θ rad ; (2) For all long force chain direction angles θ rad Standardize the data and convert it into a two-dimensional coordinate vector:

[0066] in, θ rad The original eigenvalues, i.e., the original long force chain direction angles. θ ; μ σ is the mean of this feature dimension; σ is the standard deviation of this feature dimension. x uniform These are the standardized feature parameter values, i.e., the two-dimensional coordinate vector.

[0067] (3) Using a two-dimensional coordinate vector as input, the neighborhood radius and minimum number of samples are set, and the DBSCAN clustering algorithm is used for clustering. Specifically: the neighborhood radius is set to 0.6, that is, force chains with a weighted direction angle difference of less than 20° are considered to be in the same cluster; the minimum number of samples for clustering is 3, that is, each cluster has at least 3 long force chains; the DBSCAN clustering algorithm is executed, and long force chains belonging to the same cluster are determined to be crystallized force chains with consistent orientation. In this embodiment, the long force chains with consistent orientation after the clustering algorithm is screened are as follows: Figure 15 As shown, the particles it covers are a collection of force chain crystallized particles.

[0068] In step S7 above, the degree of fusion crystallinity of the photoelastic sample is calculated based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

[0069] (1) Calculate the degree of crystallinity based on the proportion of the aggregate of structural crystalline particles to the total number of photoelastic particles. ; indicates as:

[0070] in, This indicates the number of photoelastic particles in a set of structural crystalline particles; Indicates the total number of photoelastic particles; like Figure 12 In the structure, the number of photoelastic particles in the crystalline particle aggregate. The total number of photoelastic particles is 929. If the crystallinity is 980, then the structural modal crystallinity is:

[0071] (2) Calculate the crystallinity of the force chain network based on the proportion of the force chain crystalline particle set to the total number of photoelastic particles. ; indicates as:

[0072] in, Indicates the number of photoelastic particles in the force chain crystalline particle set; like Figure 15 In the middle, the number of photoelastic particles in the force chain crystalline particle aggregate The total number of photoelastic particles is 412. If the crystallinity is 980, then the crystallinity of the force chain network is:

[0073] (3) Based on structural modal crystallinity Heli Chain Network Crystallinity Calculate the crystallinity of fusion :

[0074] in, This refers to photoelastic particles that represent the overlap of both the structural crystalline particle set and the force chain crystalline particle set.

[0075] Due to fusion crystallinity The fused crystalline particle aggregate T , and the aggregate of structural crystalline particles C Heli Chain Crystallized Particles Collection F The relationship satisfies:

[0076] Where ∨ represents an OR relationship, used to indicate that the fused crystalline particles belong to the set of structural crystalline particles. C Or force chain crystallized particle aggregate F .

[0077] When the crystallinity of structural modes The crystallinity of the power chain network is 94.8%. The percentage was 42.0%, with overlapping photoelastic particles. The total number of photoelastic particles is 404. When the fusion crystallinity is 980, It is 95.6%.

[0078] Based on the same inventive concept, embodiments of the present invention also provide a multimodal data fusion analysis system for the crystallinity of photoelastic samples, comprising: Image acquisition module, used to acquire two-dimensional images of the photoelastic sample; The image processing module is used to identify photoelastic particles from two-dimensional images and extract the centroid coordinates and strong chain network of each photoelastic particle. The structural analysis module is used to divide Voronoi cells based on centroid coordinates, calculate the local area fraction and bond orientation sequence parameter of each photoelastic particle, and screen out photoelastic particles whose local area fraction and bond orientation sequence parameter both meet the corresponding screening threshold. The screened photoelastic particles and all photoelastic particles in direct contact with the screened photoelastic particles are collectively defined as the set of structural crystalline particles.

[0079] The force chain analysis module is used to connect the centroids of adjacent photoelastic particles in the strong force chain network into line segments to construct the strong force chain skeleton. Based on the directional order parameter between adjacent line segments in the strong force chain skeleton, adjacent line segments that meet the merging conditions are merged into force chains. Based on the average number of photoelastic particles that make up each force chain, long force chains are selected. The direction vectors of the long force chains are clustered, and the photoelastic particles covered by the long force chains belonging to the same cluster are defined as the force chain crystallized particle set. The fusion calculation module is used to calculate the fusion crystallinity of the photoelastic sample based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

[0080] Since the principle behind the problem solved by this system is similar to the aforementioned method for multimodal data fusion analysis of crystallinity in photoelastic samples, the implementation of this system can be found in the implementation of the aforementioned method, and the repetitive parts will not be repeated.

[0081] Based on the same inventive concept, embodiments of the present invention also provide a terminal, including: a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the above-mentioned method for multimodal data fusion analysis of the crystallinity of photoelastic samples.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0083] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for multimodal data fusion analysis of crystallinity in photoelastic samples, characterized in that, include: Two-dimensional images of the photoelastic sample are acquired, photoelastic particles are identified from the two-dimensional images, and the centroid coordinates and strong chain network of each photoelastic particle are extracted. Based on the centroid coordinates, Voronoi cells are divided, and the local area fraction and bond orientation order parameters of each photoelastic particle are calculated. Photoelastic particles that satisfy the corresponding screening thresholds in terms of local area fraction and bond orientation order parameter are selected. The selected photoelastic particles and all photoelastic particles that are in direct contact with the selected photoelastic particles are collectively defined as a set of structural crystalline particles. Connect the centroids of adjacent photoelastic particles in the strong chain network as line segments to construct a strong chain skeleton; Based on the pointing order parameter between adjacent line segments in the strong chain skeleton, adjacent line segments that meet the merging conditions are merged into a force chain, and long force chains are selected based on the average number of photoelastic particles that make up each force chain. Cluster the directional vectors of the long force chains, and define the photoelastic particles covered by the long force chains belonging to the same cluster as the force chain crystallization particle set. The degree of fusion crystallinity of the photoelastic sample is calculated based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

2. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The strong chain network of each photoelastic particle is extracted based on preset conditions; The preset conditions include: the contact force between photoelastic particles is higher than the average value, the angle between the main contact force direction and the line connecting the centroid is less than 45°, and the number of consecutive photoelastic particles is greater than 3.

3. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The screening threshold corresponding to the local area fraction is greater than 0.

8.

4. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The filtering threshold corresponding to the bond orientation sequence parameter is greater than 0.

85.

5. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The line segments are merged into multiple force chains based on the pointing order parameter between adjacent line segments in the strong chain skeleton, and long force chains are selected based on the average number of photoelastic particles that make up each force chain. Specifically, it includes: Calculate the pointing order parameter between adjacent line segments in the strong chain skeleton; The pointing sequence parameter is judged as follows: when the pointing sequence parameter is greater than or equal to a preset threshold, the adjacent line segments are determined to belong to the same force chain, and the number of photoelastic particles that make up the force chain is recorded; when the pointing sequence parameter is less than the preset threshold, or when the force chain branches, it is determined to be the endpoint of the force chain. Calculate the average number of photoelastic particles that make up each force chain, and define the force chain with a number greater than the average number of photoelastic particles as a long force chain.

6. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The clustering of the direction vectors of the long force chain specifically includes: The angles between each long force chain and the positive x-axis are weighted and averaged to obtain the corresponding long force chain direction angles. The orientation angles of all long force chains are standardized and converted into two-dimensional coordinate vectors. Using the two-dimensional coordinate vector as input, the neighborhood radius and minimum number of samples are set, and the DBSCAN clustering algorithm is used for clustering.

7. The method for multimodal data fusion analysis of crystallinity in photoelastic samples as described in claim 1, characterized in that, The fused crystalline particle set constituted by the aforementioned fusion crystallinity satisfies the following relationship with the structural crystalline particle set and the force chain crystalline particle set: in, T Indicates a collection of fused crystalline particles; C Represents a collection of crystalline particles; F Represents a set of force chain crystalline particles; ∨ is an OR relation used to indicate that fused crystalline particles belong to a set of structural crystalline particles. C Or force chain crystallized particle aggregate F .

8. The method for multimodal data fusion analysis of crystallinity of photoelastic samples as described in claim 1, characterized in that, The degree of crystallinity is expressed as: in, Indicates the degree of crystallinity; This indicates the number of photoelastic particles in a set of structural crystalline particles; Indicates the number of photoelastic particles in the force chain crystalline particle set; Indicates the total number of photoelastic particles; This refers to photoelastic particles that represent the overlap of both the structural crystalline particle set and the force chain crystalline particle set.

9. A multi-modal data fusion analysis system for the crystallinity of a photoelastic sample, characterized in that, The system comprises: applying the method of any one of claims 1-8; Image acquisition module, used to acquire two-dimensional images of the photoelastic sample; An image processing module is used to identify photoelastic particles from the two-dimensional image and extract the centroid coordinates and strong chain network of each photoelastic particle. The structural analysis module is used to divide Voronoi cells based on the centroid coordinates, calculate the local area fraction and bond orientation sequence parameter of each photoelastic particle, and screen out photoelastic particles whose local area fraction and bond orientation sequence parameter both meet the corresponding screening threshold. The screened photoelastic particles and all photoelastic particles that are in direct contact with the screened photoelastic particles are collectively defined as a set of structural crystalline particles. The force chain analysis module is used to connect the centroids of adjacent photoelastic particles in the strong force chain network into line segments to construct a strong force chain skeleton; based on the directional order parameter between adjacent line segments in the strong force chain skeleton, adjacent line segments that meet the merging conditions are merged into force chains, and long force chains are selected based on the average number of photoelastic particles that make up each force chain; the direction vectors of the long force chains are clustered, and the photoelastic particles covered by the long force chains belonging to the same cluster are defined as the force chain crystallized particle set. The fusion calculation module is used to calculate the fusion crystallinity of the photoelastic sample based on the proportions of the structural crystalline particle set and the force chain crystalline particle set to the total number of photoelastic particles.

10. A terminal, characterized in that, include: The processor and memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the multimodal data fusion analysis method for the crystallinity of photoelastic samples as described in any one of claims 1-8.

Images