Three-dimensional granular material contact force chain detection method, system, device and storage medium

Abstract

The application discloses a three-dimensional granular material contact force chain detection method, system and device and a storage medium. The method obtains granular material particle displacement data at each loading stage and contact network data of a sample before and after each loading stage; then, a graph neural network model of the granular material is established, and the graph neural network model is trained through discrete element data; then, the particle displacement data and the contact network data are input into the trained graph neural network model to predict the contact force chain of the granular material. The method can provide an effective method for obtaining a large sample three-dimensional hard granular system contact force chain in a loading test, and the application can be widely applied to the technical field of granular material mechanics research.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of granular material mechanics research, and in particular to a three-dimensional granular material contact force chain detection method, system, device and storage medium. BACKGROUND

[0002] At present, the methods for obtaining contact force chains of granular systems mainly include photoelastic experiment method, X-ray diffraction method and method combining fast photography and image correlation algorithm. However, the above three methods have different defects respectively, and there is currently still lack of an effective method for simultaneously obtaining the particle motion and contact force chain distribution of a large sample of three-dimensional hard granular system in a test. SUMMARY

[0003] The present application aims to at least partly solve one of the problems in the prior art.

[0004] To this end, an object of embodiments of the present application is to provide a three-dimensional granular material contact force chain detection method, system, device and storage medium.

[0005] In order to achieve the above technical purpose, the technical solutions adopted by the embodiments of the present application include:

[0006] On the one hand, the embodiments of the present application provide a three-dimensional granular material contact force chain detection method, comprising the following steps:

[0007] Obtaining particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage;

[0008] Establishing a graph neural network model of the granular material, and training the graph neural network model through discrete element data;

[0009] By inputting the particle displacement data and the contact network data into the trained graph neural network model, the contact force chain of the granular material is predicted.

[0010] Further, the step of obtaining particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage comprises:

[0011] Obtaining CT images of the granular material in different deformation states;

[0012] Processing the CT images to obtain particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage.

[0013] Further, the step of processing the CT images to obtain particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage comprises:

[0014] binarizing the CT image of the granular material to obtain a binary image;

[0015] processing the binary image through a watershed algorithm to obtain a binary image of a watershed line of the granular material;

[0016] performing a logical operation on the binary image and the binary image of the watershed line to obtain an inter-particle contact area of the granular material and an area of each particle;

[0017] performing a digital image logical operation according to the inter-particle contact area of the granular material and the area of each particle to obtain particle displacement data of the granular material at each loading stage and contact network data of a sample before and after each loading stage.

[0018] Further, the discrete element data is obtained through the following steps:

[0019] establishing a discrete element model of the granular material;

[0020] performing parameter calibration on the discrete element model through the particle displacement data and the contact network data;

[0021] simulating the mechanical behavior of the granular material in the loading process through the calibrated discrete element model to obtain discrete element data.

[0022] Further, the discrete element model includes a Hertz contact model and a spherical particle and cohesive contact model, the Hertz contact model is used to simulate the interaction between the granular material, and the cohesive contact model simulates the constraint effect of the flexible latex film.

[0023] Further, the prediction of the contact force chain of the granular material includes:

[0024] batch-predicting the contact force chain of the granular material at the end of different loading stages through the graph neural network model according to the particle displacement data and the contact network data.

[0025] Further, the detection method further includes:

[0026] updating the discrete element data of the granular material through a linear perceiver.

[0027] In another aspect, an embodiment of the present application provides a three-dimensional granular material contact force chain detection system, comprising:

[0028] a first module for obtaining particle displacement data of the granular material at each loading stage and contact network data of a sample before and after each loading stage;

[0029] The second module is configured to establish a graph neural network model of the granular material, and train the graph neural network model by using the discrete element data.

[0030] The third module is configured to input the particle displacement data and the contact network data into the trained graph neural network model, and predict the contact force chain of the granular material.

[0031] In another aspect, an embodiment of the present application provides a device for detecting a contact force chain of a three-dimensional granular material, comprising:

[0032] at least one processor;

[0033] at least one memory configured to store at least one program;

[0034] When the at least one program is executed by the at least one processor, the at least one processor is caused to implement the method for detecting a contact force chain of a three-dimensional granular material.

[0035] In another aspect, an embodiment of the present application provides a storage medium having processor-executable instructions stored therein, wherein the processor-executable instructions, when executed by a processor, are configured to implement the method for detecting a contact force chain of a three-dimensional granular material.

[0036] The present application discloses a method for detecting a contact force chain of a three-dimensional granular material, which has the following beneficial effects:

[0037] The present embodiment obtains particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage; then, a graph neural network model of the granular material is established, and the graph neural network model is trained by using discrete element data; then, the particle displacement data and the contact network data are input into the trained graph neural network model to predict the contact force chain of the granular material. The present embodiment simulates the macro-micro mechanical behavior of the granular material in the in-situ CT loading test by using a discrete element model to create a virtual database of discrete element data; then, a graph neural network model is used to learn the complex nonlinear relationship in the granular material system; finally, the trained graph neural network model and the virtual database of discrete element data are used to predict the contact force chain in the test, thereby providing an effective method for obtaining a large-sample three-dimensional hard granular system contact force chain in a loading test. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following introduces the drawings of the related technical solutions in the embodiments of the present application or the prior art. It should be understood that the drawings in the following introduction are merely for the convenience of clearly describing some embodiments of the technical solutions in the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0039] Figure 1 An implementation environment schematic diagram of a three-dimensional granular material contact force chain detection method provided in the embodiments of the present application;

[0040] Figure 2 A flowchart schematic diagram of a three-dimensional granular material contact force chain detection method provided in the embodiments of the present application;

[0041] FIG. 3(a) is a schematic diagram of determining graph data of particles and contact distribution of a sample before and after a loading stage provided in the embodiments of the present application;

[0042] FIG. 3(b) is a schematic diagram of a node feature and an associated edge feature in the graph provided in the embodiments of the present application;

[0043] FIG. 3(c) is a schematic diagram of determining an edge feature through relative displacement of particles provided in the embodiments of the present application;

[0044] FIG. 3(d) is a schematic diagram of a graph neural network structure provided in the embodiments of the present application;

[0045] FIG. 4(a) is a high-resolution CT image of a glass sphere sample under the action of 500kPa isotropic pressure provided in the embodiments of the present application;

[0046] FIG. 4(b) is a constant rate shear stress-strain curve of a glass sphere under the action of 500kPa confining pressure provided in the embodiments of the present application;

[0047] FIG. 5(a) is a three-dimensional space distribution diagram of vertical displacement of particles of a glass sphere sample at 0-4%, 4-8% and 8-12% provided in the embodiments of the present application;

[0048] FIG. 5(b) is a particle contact direction frequency distribution rose diagram of a glass sphere sample when the axial strain is 0%, 4%, 8% and 12% respectively provided in the embodiments of the present application;

[0049] FIG. 6(a) is a discrete element model of a glass sphere sample in an in-situ CT triaxial loading test provided in the embodiments of the present application;

[0050] FIG. 6(b) is a stress-strain curve of a sample obtained by discrete element simulation provided in the embodiments of the present application;

[0051] Fig. 6(c) is a diagram of displacement distribution of sample particles obtained by discrete element simulation according to an embodiment of the present application;

[0052] Fig. 6(d) is a diagram of frequency distribution of contact direction of particles obtained by discrete element simulation according to an embodiment of the present application;

[0053] Figure 7 Fig. 6(c) is a diagram of displacement distribution of sample particles obtained by discrete element simulation according to an embodiment of the present application;

[0054] Figure 8 Fig. 1 is a structural schematic diagram of a three-dimensional particle material contact force chain detection system according to an embodiment of the present application;

[0055] Figure 9 Fig. 1 is a structural schematic diagram of a three-dimensional particle material contact force chain detection system according to an embodiment of the present application; DETAILED DESCRIPTION

[0056] This part will describe the specific embodiments of the present application in detail, and the preferred embodiments of the present application are shown in the accompanying drawings, which serve to supplement the description in the text part of the description and enable people to intuitively and visually understand each technical feature and the overall technical scheme of the present application, but it cannot be understood as a limitation on the protection scope of the present application.

[0057] In the description of the embodiments of the present application, several meanings are one or more, and multiple meanings are two or more. Greater than, less than, more than, etc. are understood as not including the number, above, below, etc. are understood as including the number. "At least one" means one or more. "At least one of the following" and the like means any combination of these items, including single or multiple items. If there is a description of "first", "second", etc., it is only used to distinguish technical features for the purpose, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the sequence of indicated technical features.

[0058] It should be noted that the terms such as setting, installing and connecting in the embodiments of the present application should be understood in a broad sense, and the person skilled in the art can reasonably determine the specific meaning of the above terms in the embodiments of the present application in combination with the specific content of the technical scheme. For example, the term "connection" can be mechanical connection, electrical connection or can communicate with each other; it can be directly connected or indirectly connected through an intermediate medium.

[0059] In the description of the embodiments of the present application, the description of the terms "one embodiment", "another embodiment" or "certain embodiments", "in the above-described embodiments" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiments or examples are contained in at least two embodiments or examples of the present disclosure. In the present disclosure, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0060] It should be noted that the technical features involved in each of the embodiments of the present application described below can be combined with each other as long as there is no conflict.

[0061] In recent years, with the deepening understanding of the mechanical behavior of granular materials, researchers have found that the evolution law of contact force chains of granular materials under shear action is closely related to the macroscopic mechanical behavior of the granular materials. Under this background, researchers pay particular attention to the measurement of contact forces of granular materials under experimental conditions, and have proposed various testing techniques to measure the contact force chains of granular materials.

[0062] At present, the main methods for obtaining contact force chains of granular systems include photoelastic test method, X-ray diffraction method and method combining fast photography and image correlation algorithm.

[0063] Among them, the photoelastic test method determines the contact force of the particles according to the different photoelastic properties of the granular materials under different loads, and can only be used to test the contact force chains of photoelastic granular systems, and is generally used for two-dimensional granular systems (such as circular plate granular assemblies).

[0064] The X-ray diffraction method determines the contact force of the particles by measuring the lattice spacing of the loaded particles and according to the change of the lattice spacing of the particles under the action of the load, and is generally used for contact force detection of three-dimensional granular systems containing a small number of particles (such as less than 1000 particles). In addition, the photoelastic test and the X-ray diffraction method have a common limitation: they can only be used to obtain the contact force chains of the granular systems, and cannot obtain the motion information (such as particle displacement and particle rotation) of the particles during the test at the same time.

[0065] The method combining the high-speed photography with the image correlation algorithm obtains the digital images of the granular system in the loading process through the high-resolution camera, calculates the deformation of the particles in the test process by using the image correlation algorithm, and reverses the contact force of the particles by using the deformation of the particles. The method can conveniently obtain the motion field and the contact force chain of the granular system at the same time, but is only applicable to two-dimensional soft granular materials (such as a plastic circular sheet assembly). For hard granular materials (such as sand), due to the resolution limitation of the image, the image correlation algorithm cannot accurately obtain the deformation of the particles. Therefore, the method is not applicable to hard granular materials.

[0066] At present, there is still a lack of an effective method for simultaneously obtaining the particle motion and the contact force chain distribution of a large sample three-dimensional hard granular system (such as a granular system with more than 10,000 particles) in a test.

[0067] Therefore, the present application provides a three-dimensional granular material contact force chain detection method, which comprises the following steps: obtaining the particle displacement data of the granular material at each loading stage and the contact network data of the sample before and after each loading stage; then, establishing a graph neural network model of the granular material, and training the graph neural network model through discrete element data; and then inputting the particle displacement data and the contact network data into the trained graph neural network model to predict the contact force chain of the granular material. The present embodiment creates a virtual database of discrete element data by simulating the macro-micro mechanical behavior of the granular material in the in-situ CT loading test through the discrete element model; learns the complex nonlinear relationship in the granular material system through the graph neural network model; and finally uses the trained graph neural network model and the virtual database of the discrete element data to predict the contact force chain in the test, thereby providing an effective method for obtaining the contact force chain of a large sample three-dimensional hard granular system in a loading test.

[0068] Figure 1 is an implementation environment schematic diagram of a three-dimensional granular material contact force chain detection method provided by the present application. Referring to Figure 1 , the main body of the software and hardware of the implementation environment mainly includes an operation terminal 101 and a server 102, and the operation terminal 101 is in communication connection with the server 102. Among them, the three-dimensional granular material contact force chain detection method can be independently configured to execute on the operation terminal 101, or can be independently configured to execute on the server 102, or can be executed based on the interaction between the operation terminal 101 and the server 102, and the specific selection can be appropriately selected according to the actual application, and the present embodiment does not make a specific limitation. In addition, the operation terminal 101 and the server 102 can be nodes in the block chain, and the present embodiment does not make a specific limitation.

[0069] Specifically, the operation terminal 101 in the present application can include but is not limited to any one or more of a smart watch, a smart phone, a computer, a personal digital assistant (PDA), a smart voice interaction device, a smart home appliance, or a vehicle-mounted terminal. The server 102 can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms. The operation terminal 101 and the server 102 can establish a communication connection through a wireless network or a wired network using standard communication technology and / or protocols. The network can be set as the Internet or any other network, such as any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network.

[0070] Figure 2 is a flowchart of a three-dimensional granular material contact force chain detection method provided by an embodiment of the present application. The execution subject of the method can be at least one of an operation terminal or a server, Figure 2 In the present application, the three-dimensional granular material contact force chain detection method is configured to be executed by an operation terminal. Referring to Figure 2 The three-dimensional granular material contact force chain detection method includes but is not limited to steps 110 to 130.

[0071] Step 110: Obtain granular displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage.

[0072] In this step, the granular displacement data of the granular material at each loading stage and the contact network data of the sample before and after each loading stage are obtained in the in-situ CT loading test of the granular material. Specifically, in the in-situ CT loading test of the granular material, the CT images of the sample at different axial strains are obtained by a high-resolution micro-CT device, and the collected CT images are processed to obtain the granular displacement data and the contact network data.

[0073] Step 120: Establish a graph neural network model of the granular material, and train the graph neural network model through discrete element data.

[0074] In this step, to predict the contact force chain of particulate materials, a graph neural network model of the particulate materials needs to be established and trained using discrete metadata. Specifically, a discrete element model of the particulate materials is established and its parameters are calibrated. Then, samples of different particulate materials are randomly generated using the calibrated discrete element model, and simulation experiments are conducted on these samples. The particulate material data collected during the experiment is used as discrete metadata.

[0075] Specifically, taking the loading test of a glass sphere specimen with a diameter of 0.3-0.6 mm as an example, a graph neural network model of the glass sphere specimen is established, with both its input and output being graph data.

[0076] Each set of input graph data includes node and edge data (such as node eigenvectors and edge eigenvectors), as shown in Figure 3(a). This data can be determined by the particle displacement and contact evolution type of the glass sphere specimen during a loading stage. Each node in the graph corresponds to each particle of the glass sphere specimen. Referring to Figure 3(b), node attributes include the glass sphere diameter and the contact coordination number of the glass sphere at the end of the loading stage. For any two nodes in the graph, if the two glass sphere particles associated with them are in contact at the beginning or end of the loading stage, then an edge exists between them. Referring to Figure 3(c), the attributes of this edge include the components of the relative displacement of the two particles in their contact direction and contact normal direction, and the contact evolution type value during the loading stage. If the two particles are in contact at the beginning of the loading stage and not in contact at the end of the loading stage, the contact evolution type value of the corresponding edge is -1; if the two particles are in contact at both the beginning and end of the loading stage, the contact evolution type value of the corresponding edge is 0; if the two particles are not in contact at the beginning of the loading stage and are in contact at the end of the loading stage, the contact evolution type value of the corresponding edge is 1. The output graph of this neural network model has the same data structure as the input graph. This invention uses the first dimension of the output graph node as the predicted value of the normalized maximum normal contact force of the corresponding particle.

[0077] Referring to Figure 3(d), this model adopts a "encryption-processing-decryption" neural network structure, which can be used to execute the information transmission in the neural network using the following formula:

[0078] G0 = GNN enc (G inp (1)

[0079] G n =GNN core (G n-1 ;G0), for 1≤n≤N, (2)

[0080] G out =GNN dec (Gn ), (3)

[0081] wherein G inp and G out are input and output graph data of the neural network respectively; ‘;’ in the formula represents data concatenation; GNN enc , GNN core and GNN dec are encryption, processing and decryption modules of the neural network respectively; G0and G n are output graph data of the neural network encryption module and output graph data of the nth layer network of the processing module respectively; N is the number of network layers of the processing module.

[0082] For the model, the accuracy of the model prediction result can be measured by a loss function. The loss function is defined on a single training data and is used to measure the prediction error of a training data. Specifically, the loss value of a training data is determined by the label of the single training data and the prediction result of the model on the training data. In actual training, a training data set has many training data, so a cost function is generally used to measure the overall error of the training data set. The cost function is defined on the entire training data set and is used to calculate the average value of the prediction error of all training data, which can better measure the prediction effect of the model. For a general machine learning model, based on the aforementioned cost function, plus a regular term that measures the complexity of the model, it can be used as the target function of training. Based on the target function, the loss value of the entire training data set can be obtained. There are many types of commonly used loss functions, such as 0-1 loss function, square loss function, absolute loss function, logarithmic loss function, cross-entropy loss function, etc. All of them can be used as the loss function of the machine learning model, which will not be elaborated one by one. In the embodiments of the present application, any one of the loss functions can be selected to determine the loss value of the training. Based on the loss value of the training, the parameters of the model are updated using the back propagation algorithm, and after several iterations, the trained glass sphere sample graph neural network model can be obtained. The specific number of iterations can be pre-set, or the training is considered to be completed when the test set reaches the accuracy requirement.

[0083] Step 130: predicting the contact force chain of the granular material by inputting the particle displacement data and the contact network data into the trained graph neural network model.

[0084] In this step, the input of the neural network model is the graph structure data corresponding to the granular material in one loading stage, including node data and edge data. Each node in the graph corresponds to each particle of the granular system. The node attributes include the particle size of the particle and the coordination number of the particle at the end of the loading stage. For any two particles in the graph, if they are in contact at the beginning or end of the loading stage, there is an edge between them. The attributes of the edge include the components of the relative displacement of the two particles in the contact direction and the contact normal direction, and the type value of the contact evolution of the two particles in the loading stage. The output of the neural network is graph data with the same structure. The first dimension feature of the output node is taken as the contact force prediction value of the node. The neural network model is trained by discrete element data to realize the prediction of the contact force chain of the granular system by the displacement and contact network data of the particles.

[0085] Further as an optional implementation, the step of obtaining the particle displacement data of the granular material in each loading stage and the contact network data of the sample before and after each loading stage comprises:

[0086] Obtaining CT images of the granular material in different deformation states;

[0087] Processing the CT images to obtain the particle displacement data of the granular material in each loading stage and the contact network data of the sample before and after each loading stage.

[0088] Specifically, the CT images in different deformation states need to be obtained in the in-situ CT triaxial loading test. For example, taking the loading test of a glass ball sample with a diameter of 0.3-0.6 mm as an example, first, fix the micro triaxial loading instrument on the rotating table of the high-resolution micro-CT device; second, prepare the glass ball particles with a diameter of 0.3-0.6 mm into a cylindrical sample with a diameter of 8 mm and a height of 16 mm, and load it into the micro triaxial instrument; third, referring to FIG. 4(a), apply a constant confining pressure of 500 kPa to the sample by the micro triaxial instrument, and obtain the high-resolution (6.5 microns) full-field CT image of the sample by the micro-CT device; then, maintain the confining pressure of the sample unchanged, and apply an axial load to the sample at a constant strain rate of 0.1% per minute by the micro triaxial instrument, pause the axial loading when the axial strain is 4% (i.e. maintain the constant axial strain and confining pressure), and obtain the high-resolution full-field CT image of the sample by the micro-CT device; finally, repeat the above step to obtain the CT images of the sample at the axial strain of 8% and 12%, respectively; in this process, referring to FIG. 4(b), the stress-strain curve of the sample during loading can be recorded by the micro triaxial instrument.

[0089] Further as an optional implementation, the step of processing the CT images to obtain the particle displacement data of the granular material at each loading stage and the contact network data of the sample before and after each loading stage comprises:

[0090] performing binaryzation processing on the CT images of the granular material to obtain binary images;

[0091] performing processing on the binary images by a watershed algorithm to obtain a binary image of watershed lines of the granular material;

[0092] performing logical operation on the binary images and the binary image of the watershed lines to obtain inter-particle contact areas of the granular material and areas of each particle;

[0093] performing digital image logical operation according to the inter-particle contact areas of the granular material and the areas of each particle to obtain the particle displacement data of the granular material at each loading stage and the contact network data of the sample before and after each loading stage.

[0094] Specifically, still taking the loading test of the glass ball sample with a diameter of 0.3-0.6 mm as an example, the CT images of the glass ball sample obtained at different axial strains are processed to obtain the particle size, particle displacement and particle contact network data of the sample at different loading stages. First, binaryzation processing is performed on the original CT images to obtain binary images. Then, the binary images are segmented by the marked watershed algorithm to separate the areas of each particle in the CT images. Next, referring to FIG. 5(a), the pixel points of each area are calculated by the built-in function Regionprops of MATLAB to determine the particle size and center point position of each particle. Finally, the position of each particle in the CT images at different loading states is identified by a particle tracking method based on the particle volume to determine the particle displacement. Referring to FIG. 5(b), the binary image of the watershed lines of the sample can be obtained in the process of image segmentation of the CT images by the watershed algorithm. The contact areas between the particles of the sample can be obtained by performing logical AND operation on the binary image and the binary image of the original CT images. Whether there is contact between any two particles can be determined by morphological analysis, and a contact direction frequency rose diagram can be drawn.

[0095] Further as an optional implementation, the discrete element data are obtained by the following steps:

[0096] establishing a discrete element model of the granular material;

[0097] performing parameter calibration on the discrete element model by the particle displacement data and the contact network data;

[0098] Simulate the mechanical behavior of the granular material in the loading process by the calibrated discrete element model to obtain the discrete element data.

[0099] Further as an optional implementation, the discrete element model comprises a Hertz contact model and a spherical particle and cohesive contact model, the Hertz contact model is used to simulate the interaction between the granular material, and the cohesive contact model simulates the constraint effect of the flexible latex film.

[0100] Specifically, according to the initial conditions of the glass ball sample in the in-situ CT triaxial loading test, first, referring to FIG. 6(a), the discrete element model is established according to the size of the glass ball sample, the initial porosity, and the particle number and particle size distribution of the glass ball, the Hertz contact model is used to simulate the interaction between the granular material, and the flexible latex film boundary of the glass ball periphery is simulated by connecting the spherical balls with the contact cohesive model; second, referring to FIG. 6(b), the same boundary conditions as the in-situ CT triaxial loading test are applied to the discrete element model, the particle center position and contact data in the sample are recorded when the axial strain is 0%, 4%, 8% and 12%, the stress-strain curve of the sample in the loading process is recorded, and compared with the test results; referring to FIG. 6(c), the particle displacement is calculated according to the recorded particle center position coordinates of the sample at different axial strains, and compared with the test results; third, referring to FIG. 6(d), the contact direction frequency distribution of the sample is determined by the recorded particle contact data of the sample at different axial strains, and compared with the test results; finally, by modifying the parameters of the discrete element model, including the normal and tangential stiffness coefficients of the particles, and the friction coefficient of the particles, etc., the results of the discrete element simulation are close to the test results.

[0101] After the calibration of the discrete element model parameters is completed, a batch of randomly generated samples are generated, and the same boundary conditions are used for discrete element simulation, the discrete element data of the sample at different axial strains (such as 0%, 0.5%, 1%, …, 12%) are recorded, including the center point position of the particles, the particle size, the contact state of any two particles, etc. In this process, the randomly generated samples only have different initial arrangements of the particles, and other initial conditions including the number of particles, the particle size distribution, the initial porosity of the particles, the size of the sample, etc. are consistent with the discrete element model.

[0102] Further as an optional implementation, the predicting the contact force chain of the granular material comprises:

[0103] By the graph neural network model, the contact force chain of the granular material at the end of different loading stages is batch predicted according to the particle displacement data and the contact network data.

[0104] Specifically, referring to Figure 7According to the particle motion and contact network detection data of the glass ball sample in the three different loading stages of 0-4%, 4-8% and 8-12% axial strain in the in-situ CT loading test, three groups of graph data are constructed, and the trained graph neural network model is fed. The model outputs three groups of graph data, and the first dimension feature of the nodes is the predicted value of the normalized maximum normal contact force of the glass ball particle corresponding to the node at the end of the above three loading stages (i.e. 4%, 8% and 12%). According to the prediction result of each particle, the inter-particle contact force strength between the particles detected by the CT image is estimated: the contact force strength between two particles is the average value of their predicted values, and the maximum normal contact force network of the glass ball sample is determined. The contact network can reflect the transmission rule of the internal contact force of the glass ball in the triaxial loading process.

[0105] Further, as an optional implementation, the detection method further comprises:

[0106] updating the discrete element data of the granular material through the linear perceptron.

[0107] Specifically, each layer of the graph neural network is composed of a linear perceptron with 2 layers and 64 nodes to perform synchronous updating of edge attributes and node attributes.

[0108] With reference to Figure 8 The three-dimensional granular material contact force chain detection system provided by the embodiment of the application comprises:

[0109] The first module 801 is configured to acquire the particle displacement data of the granular material in each loading stage and the contact network data of the sample before and after each loading stage.

[0110] The second module 802 is configured to establish a graph neural network model of the granular material and train the graph neural network model through the discrete element data.

[0111] The third module 803 is configured to input the particle displacement data and the contact network data into the trained graph neural network model to predict the contact force chain of the granular material.

[0112] The contents in the above method embodiments are applicable to the system embodiments, the system embodiments specifically realize the same functions as the above method embodiments, and achieve the same beneficial effects as the above method embodiments.

[0113] With reference to Figure 9 The three-dimensional granular material contact force chain detection device provided by the embodiment of the application comprises:

[0114] at least one processor 901;

[0115] at least one memory 902 configured to store at least one program;

[0116] When the at least one program is executed by the at least one processor 901, the at least one processor 901 is caused to implement Figure 2 The three-dimensional granular material contact force chain detection method is shown.

[0117] The contents in the above method embodiments are all applicable to the device embodiments, the device embodiments specifically implement the functions same as the above method embodiments, and achieve the beneficial effects same as the above method embodiments.

[0118] The embodiment of the application further provides a storage medium, wherein the storage medium stores processor-executable instructions, and the processor-executable instructions are used for realizing Figure 2 The three-dimensional granular material contact force chain detection method is shown.

[0119] The above is a specific description of the preferred implementation of the application, but the application is not limited to the above-mentioned embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the application, and these equivalent modifications or replacements are all included in the scope defined by the claims of the present application.

Claims

1. A method for detecting contact force chains in three-dimensional particulate materials, characterized in that, Includes the following steps: Acquire particle displacement data of particulate material at each loading stage and contact network data of the specimen before and after each loading stage; A graph neural network model for particulate materials is established, and the graph neural network model is trained using discrete metadata. By inputting the particle displacement data and the contact network data into a trained graph neural network model, the contact force chain of the particulate material is predicted. The discrete metadata is obtained through the following steps: Establish a discrete element model for particulate materials; The discrete element model is calibrated using the particle displacement data and the contact network data. Discrete data is obtained by simulating the mechanical behavior of particulate materials during loading using a calibrated discrete element model. The discrete element model includes the Hertz contact model and the spherical particle and adhesive contact model. The Hertz contact model is used to simulate the interaction between particulate materials, and the adhesive contact model simulates the constraint effect of a flexible latex film.

2. The method for detecting contact force chains in three-dimensional particulate materials according to claim 1, characterized in that, The step of obtaining particle displacement data of the particulate material at each loading stage and contact network data of the specimen before and after each loading stage includes: Acquire CT images of the particulate material under different deformation states; The CT images are processed to obtain particle displacement data of the particulate material at each loading stage and contact network data of the specimen before and after each loading stage.

3. The method for detecting contact force chains in three-dimensional particulate materials according to claim 2, characterized in that, The step of processing the CT images to obtain particle displacement data of the particulate material at each loading stage and contact network data of the specimen before and after each loading stage includes: The CT image of the particulate material is binarized to obtain a binary image; The binary image is processed by the watershed algorithm to obtain the watershed line binary image of the particulate material; Logical operations are performed on the binary image and the watershed line binary image to obtain the interparticle contact region of the particulate material and the region of each particle. Based on the interparticle contact area of ​​the granular material and the area of ​​each particle, digital image logic operations are performed to obtain particle displacement data of the granular material at each loading stage and contact network data of the sample before and after each loading stage.

4. The method for detecting contact force chains in three-dimensional particulate materials according to claim 1, characterized in that, The prediction of the contact force chain of the particulate material includes: Using the graph neural network model, based on the particle displacement data and the contact network data, the contact force chains of the particle material at the end of different loading stages are predicted in batches.

5. The method for detecting contact force chains in three-dimensional particulate materials according to any one of claims 1-4, characterized in that, The detection method further includes: The discrete metadata of the particulate material is updated using a linear perceptron.

6. A three-dimensional particulate material contact force chain detection system, characterized in that, include: The first module is used to acquire particle displacement data of particulate material at each loading stage and contact network data of the specimen before and after each loading stage. The second module is used to establish a graph neural network model of particulate materials and train the graph neural network model using discrete metadata. The third module is used to predict the contact force chain of the particulate material by inputting the particle displacement data and the contact network data into a trained graph neural network model. The discrete metadata is obtained through the following operations: Establish a discrete element model for particulate materials; The discrete element model is calibrated using the particle displacement data and the contact network data. Discrete data is obtained by simulating the mechanical behavior of particulate materials during loading using a calibrated discrete element model. The discrete element model includes the Hertz contact model and the spherical particle and adhesive contact model. The Hertz contact model is used to simulate the interaction between particulate materials, and the adhesive contact model simulates the constraint effect of a flexible latex film.

7. A three-dimensional particulate material contact force chain detection device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the three-dimensional particulate material contact force chain detection method as described in any one of claims 1-5.

8. A computer-readable storage medium storing processor-executable instructions, characterized in that, The processor-executable instructions, when executed by the processor, are used to implement the three-dimensional particulate material contact force chain detection method as described in any one of claims 1-5.

Images