A digital twin control system for mineral processing, sampling, sample preparation, and testing
By using a digital twin control system for mineral processing, sampling, and testing, the spectral gradient and geometric contour of the ore surface are captured in real time. A heterogeneous tetrahedral topological substrate is constructed, and the internal damage and energy state of the ore are calculated. This solves the problem of dynamically evaluating the heterogeneous characteristics of the ore in existing technologies and achieves representativeness and precise control of large-volume sampling and testing.
Patent Information
- Application Number
- CN202611468775.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies struggle to dynamically deduce physical mechanisms based on the heterogeneous characteristics and real-time stress deformation state of ore during mineral processing, crushing, and sampling. This results in an inability to accurately assess the spatial segregation distribution of refractory intergrowth particles in continuous ore flows, and an inability to achieve online dynamic feedback and control of crushing and fractionation sampling equipment. Consequently, it becomes difficult to ensure the representativeness of large-volume sampling and testing.
A digital twin control system for mineral processing, sampling, and testing is adopted. By collecting ore point cloud data and crusher operating data, the system identifies the spatial gradient of ore spectral reflectance, generates an interface metric tensor, calculates the control repulsion radius, generates a tetrahedral topological basis, calculates nodal strain energy and damage variables, sampling weight distribution, evaluates the degree of dissociation and spatial segregation parameters, and generates adaptive control commands.
It enables online real-time capture of the spectral gradient and geometric contour of the ore surface, construction of heterogeneous mineral crystal tetrahedral topological substrate, calculation of tensile micro-damage and compressive residual mechanical energy inside the ore, quantification of the wear resistance and dissociation state of micro-nodes, generation of adaptive control commands, and ensures the true representativeness and control accuracy of large-volume sampling and testing.
Smart Images

Figure CN123113594A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of mineral processing and sample preparation technology, and in particular to a digital twin control system for mineral processing sampling, sample preparation and testing. Background Technology
[0002] Mineral processing sampling and testing is a crucial step in mineral resource development and mineral processing technology control. Its purpose is to obtain accurate data such as mineral grade, degree of liberation, and phase composition by extracting representative ore samples. Accurate sampling not only requires the samples to conform to statistical distribution at the macroscopic level, but also needs to overcome the microstructural differences caused by the heterogeneity of the ore. The invention patent application with publication number CN121805524A discloses a tailings sampling, testing, and evaluation method. This method classifies and samples tailings according to the type of tailings discharge, and combines chemical phase analysis, microscopic degree of liberation analysis, and particle size analysis under laboratory conditions to determine multi-dimensional influence coefficients, and then calculates the concentrate recovery rate to achieve a static assessment of the economic value of tailings. The aforementioned existing technical solutions rely on offline static physical sampling and lagging laboratory equipment analysis in actual continuous mineral processing production and control. They lack the ability to perceive the surface texture and internal heterogeneous interfaces of ore during conveyor belt transport and entry into the crusher. They cannot extract the three-dimensional geometric and spectral characteristics of the ore in real time. When assessing the degree of liberation, the existing technology only stays at the level of statistical analysis of static samples and fails to deeply explore the physical and mechanical mechanisms such as stress, strain and energy conversion of ore under the actual working conditions of crushing equipment. They lack dynamic calculation models for tensile damage and residual mechanical energy inside the ore and cannot accurately predict the propagation of microcracks and the real-time peeling state of grain boundaries. Summary of the Invention
[0003] The technical problem solved by this invention is that existing technologies are unable to perform dynamic physical mechanism deduction based on the heterogeneous characteristics and real-time stress deformation state of ore during mineral processing, crushing and sampling. This makes it impossible to accurately assess the spatial segregation distribution of refractory intergrowth particles in continuous ore flow, and consequently, it is impossible to achieve online dynamic feedback and control of crushing equipment and fractionation sampling equipment. Ultimately, it is difficult to ensure the representativeness of large-volume sampling and testing.
[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a digital twin control system for mineral processing, sampling, sample preparation, and testing, comprising: The data acquisition module is used to acquire ore point cloud data and crusher operating data; The identification module is used to extract boundary points based on the spatial gradient of spectral reflectance in the ore point cloud data, and to calculate the heterogeneous texture vector and heterogeneous phase boundary density to generate an interface metric tensor. The topology module is used to calculate the control repulsion radius based on the interface metric tensor and the heterogeneous phase boundary density, generate discrete nodes, and perform spatial partitioning and tetrahedral subdivision on the discrete nodes to generate a tetrahedral topological basis. The calculation module is used to obtain the outer contour nodes of the tetrahedral topological base, solve the node displacement of the outer contour nodes and calculate the node strain tensor based on the crusher working condition data, and decompose the node strain tensor to obtain the node tensile strain energy and node compressive strain energy. Based on the tensile strain energy of the nodes, the partial differential equation of phase field damage is constructed and solved to obtain the set of node damage variables, the degree of dissociation is calculated, and the compressive strain energy of the nodes is extracted as the set of residual mechanical strain energy. The sampling module is used to calculate the node sampling weight distribution based on the set of node damage variables and the set of residual mechanical strain energy. The evaluation module is used to extract high-risk nodes based on the node sampling weight distribution, construct and update the spatial scatter matrix, and calculate the spatial segregation parameters. The control module is used to compare the degree of dissociation with a preset dissociation threshold to generate a first control command, and to compare the spatial segregation parameter with a preset segregation threshold to generate a second control command.
[0005] Preferably, the ore point cloud data is a fused point cloud matrix obtained by fusing and reconstructing ore surface spectral data and ore surface three-dimensional spatial coordinates.
[0006] Preferably, in the recognition module, the process of obtaining the heterogeneous texture vector and the heterogeneous phase boundary density includes: Points with spectral reflectance values within the effective range are extracted from the fused point cloud matrix as valid spectral point cloud data; The absolute average of the difference between the spectral reflectance values of each point and its neighboring points in the effective spectral point cloud data is calculated as the spectral gradient of the point. Points with spectral gradients greater than a preset spectral gradient threshold are constructed as a boundary point set matrix. The principal component analysis algorithm is called to solve for the eigenvalues and eigenvectors of the covariance matrix of the boundary point set matrix. The eigenvectors corresponding to the minimum eigenvalues are extracted and normalized to obtain the heterogeneous texture vectors. The ratio of the total number of points in the boundary point set matrix to the total number of points in the effective spectral point cloud data is calculated as the heterogeneous phase boundary density. The process of obtaining the interface metric tensor includes: An interface metric tensor is constructed based on the heterogeneous texture vector and the heterogeneous phase boundary density.
[0007] Preferably, in the topology module, the process of obtaining the control repulsion radius includes: The control repulsion radius is calculated based on the heterogeneous phase boundary density, the preset maximum grid spacing, and the preset geometric scaling constant. The process of generating the discrete nodes includes: Using the three-dimensional convex hull corresponding to the fused point cloud matrix as the spatial envelope, and within the spatial envelope, with the control repulsion radius as the minimum node spacing parameter, the Poisson disk sampling algorithm is called to generate the node coordinate matrix. The process of generating the tetrahedral topological basis includes: Using the node coordinate matrix and the interface metric tensor as input, the anisotropic Thiessen polygon algorithm is invoked. The anisotropic Thiessen polygon algorithm uses the inverse matrix of the interface metric tensor as the metric tensor to define the anisotropic distance between nodes and performs spatial partitioning on the node coordinate matrix to obtain a polyhedral region. The constrained Delaunay tetrahedral partitioning algorithm is called to partition the polyhedral region to obtain topological connection data. The node coordinate matrix and the topological connection data together constitute the tetrahedral topological basis.
[0008] Preferably, the process of obtaining the outer contour nodes in the calculation module includes: Traverse the tetrahedral cells in the topology connection data, count the number of times each triangular face appears in all tetrahedral cells, extract and deduplicate the nodes on the triangular face that appears only once to obtain the outer contour nodes. The process of obtaining the nodal displacement and nodal strain tensors includes: Divide the modulus of the extrusion force vector by the total number of the outer contour nodes to obtain the nodal force load of each outer contour node along the direction of the extrusion force vector. The units of the node coordinate matrix are uniformly converted to the International System of Units (SI). Based on the converted coordinate data and the topological connection data, the overall stiffness matrix is assembled. The element volume and strain displacement matrix are calculated using the tetrahedral shape function and its gradient. The elastic matrix is obtained by combining the linear elastic constitutive relation. Then, the element stiffness matrix of each tetrahedral element is calculated and accumulated into the overall stiffness matrix according to the node index. Solve the system of linear equations containing the overall stiffness matrix and the nodal force loads to obtain the nodal displacement column vector; For each node in the tetrahedral topological basis, the strain tensor of the tetrahedral element containing that node is extracted and averaged by element volume to obtain the nodal strain tensor of that node.
[0009] Preferably, the process of decomposing the nodal strain tensor in the calculation module includes: Based on the linear elastic constitutive relation, the nodal strain tensor is converted into elastic strain energy density. Then, using an asymmetric energy decomposition method based on the principal strain symbol, the elastic strain energy density is decomposed into nodal tensile strain energy and nodal compressive strain energy.
[0010] Preferably, the process of obtaining the set of node damage variables includes: Using the tensile strain energy of the nodes as energy input, a phase-field damage partial differential equation is constructed on each node of the tetrahedral topological substrate. The set of node damage variables for all nodes is obtained by solving the phase-field damage partial differential equation using the explicit Euler method. The degree of dissociation is the arithmetic mean of all elements in the set of node damage variables; The residual mechanical strain energy set is the nodal compressive strain energy of all nodes.
[0011] Preferably, in the sampling module, the process of calculating the node sampling weight distribution includes: For each node, the node sampling weight is calculated based on the node damage variable of that node and the corresponding compressive strain energy in the set of residual mechanical strain energy. The node sampling weights of all nodes are normalized to obtain a node sampling weight distribution where the sum of all elements is 1.
[0012] Preferably, in the evaluation module, the process of constructing and updating the spatial scatter matrix includes: The elements in the node sampling weight distribution are sorted in descending order, and the nodes with the highest sorting ratio are extracted as high-risk nodes. Extract the three-dimensional spatial coordinates of the high-risk nodes to construct a high-risk node coordinate matrix, calculate the mean vector of the high-risk node coordinate matrix, center the high-risk node coordinate matrix, and perform dimensionless processing based on the preset maximum grid spacing. Based on the temporal forgetting factor, the dimensionless high-risk node coordinate matrix and its transpose matrix, and the total number of high-risk nodes, the spatial scatter matrix is calculated and updated incrementally. The process of calculating the spatial segregation parameters includes: The inverse of the spatial scattering matrix is obtained, and the sum of the main diagonal elements of the inverse matrix is calculated as the spatial segregation parameter.
[0013] Preferably, in the control module, the process of generating the first control command includes: When the degree of dissociation is less than the preset dissociation threshold, the target discharge port gap is calculated, and based on the comparison result between the target discharge port gap and the preset minimum discharge port gap, the first control command is generated. The first control command is used to control the discharge port gap. The process of generating the second control command includes: The spatial segregation parameter is compared with a preset segregation threshold. When the spatial segregation parameter is greater than the preset segregation threshold, the time delay is calculated based on the ore transport distance and speed, and a second control command is generated. The second control command is used to control the sampling reducer.
[0014] The beneficial effects of this invention are as follows: This invention can capture the spectral gradient and geometric contour of the ore surface online in real time, and use this to construct a heterogeneous mineral crystal tetrahedral topological substrate. Then, the measured crusher extrusion pressure is converted into nodal loads, and the tensile micro-damage evolution and compressive residual mechanical energy inside the ore are calculated. This invention eliminates the dependence on pure macroscopic statistics and can directly quantify the grinding difficulty and dissociation state of each micro-node. Furthermore, the spatial scattering matrix is used to evaluate the spatial segregation degree of high-risk intergrowths in dynamic ore flow in real time, and an adaptive closed-loop control command is generated accordingly. This allows for timely adjustment of the crusher discharge port to ensure physical breakdown effect when the ore dissociation is insufficient. Attached Figure Description
[0015] Figure 1 This is a basic flowchart of a digital twin control system for mineral processing, sampling, and testing, provided as an embodiment of the present invention. Detailed Implementation
[0016] To make the above-mentioned objects, features and advantages of the present invention more readily understood, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0017] Example, refer to Figure 1 A digital twin control system for mineral processing, sampling, sample preparation, and testing is provided, comprising: The data acquisition module is used to acquire ore point cloud data and crusher operating data; The identification module is used to extract boundary points based on the spatial gradient of spectral reflectance in ore point cloud data, and to calculate heterogeneous texture vectors and heterogeneous phase boundary densities to generate interface metric tensors. The topology module is used to calculate the control repulsion radius based on the interface metric tensor and the heterogeneous phase boundary density, generate discrete nodes, and perform spatial partitioning and tetrahedral subdivision on the discrete nodes to generate a tetrahedral topological basis. The calculation module is used to obtain the outer contour nodes of the tetrahedral topological base. Based on the crusher working condition data, it solves the node displacement of the outer contour nodes and calculates the node strain tensor. It decomposes the node strain tensor to obtain the node tensile strain energy and node compressive strain energy. Based on the nodal tensile strain energy, the phase field damage partial differential equation is constructed and solved to obtain the nodal damage variable set, the degree of dissociation is calculated, and the nodal compressive strain energy is extracted as the residual mechanical strain energy set. The sampling module is used to calculate the node sampling weight distribution based on the set of node damage variables and the set of residual mechanical strain energy. The evaluation module is used to extract high-risk nodes based on the node sampling weight distribution, construct and update the spatial scatter matrix, and calculate the spatial segregation parameters. The control module is used to compare the degree of dissociation with a preset dissociation threshold to generate a first control command, and to compare the spatial segregation parameters with a preset segregation threshold to generate a second control command.
[0018] This invention integrates multidimensional ore data in real time and extrapolates stress deformation and energy conversion, enabling online dynamic quantification of the propagation of micro-cracks within the ore and the degree of segregation of particles in space. It can also adaptively adjust the physical actuator based on the dynamic dissociation state and segregation parameters, ensuring the true representativeness and control accuracy of large-flow ore sampling and testing.
[0019] The ore point cloud data is a fused point cloud matrix obtained by fusing and reconstructing spectral data and three-dimensional spatial coordinates of the ore surface; Set time window length The time window is 0.5s, which covers the physical retention process of a single batch of ore from falling off the conveyor belt to being first bitten by the crusher teeth and undergoing deformation. A multispectral camera with a sampling frequency of 100Hz is deployed directly above the end of the jaw crusher's feed belt to obtain the time window length. The spectral data of the ore surface is multi-frame multispectral image. Each frame has a dimension of H×W×16, where H is the number of pixels in the image height, W is the number of pixels in the image width, and 16 is the number of bands. Each pixel has a spectral reflectance value in each band, which is a dimensionless decimal between 0 and 1. A 3D laser profilometer was deployed 0.2m beside the multispectral camera, with a sampling frequency of 100Hz, to obtain the inter-window length. The surface coordinate data of the ore is multi-frame point cloud data, and each frame of point cloud consists of one... The matrix representation, This represents the total number of laser reflection points contained in the point cloud of this frame. It is a dynamically changing value. Each row of the matrix records the three-dimensional spatial coordinates of a laser reflection point in the three-dimensional laser profilometer coordinate system, in millimeters. Within each time window, the spectral data of the ore surface collected at the same time and the coordinate data of the ore surface are fused together; The integration process includes: Based on a multispectral camera and a 3D laser profilometer with relatively fixed installation positions, the extrinsic parameter calibration matrix is obtained. Based on the 3D spatial coordinates of the laser reflection point in the coordinate system of the 3D laser profilometer, the coordinates of the laser reflection point in the coordinate system of the multispectral camera are obtained using the extrinsic parameter calibration matrix. Using the intrinsic parameter calibration matrix of the multispectral camera, the coordinates of the laser reflection point in the coordinate system of the multispectral camera are projected onto the 2D pixel plane of the multispectral camera. For each laser reflection point in the 3D point cloud, if its projection corresponds to a multispectral pixel on a 2D pixel plane of a multispectral camera, then the spectral reflectance value extraction operation is performed, specifically including: For each multispectral pixel, the difference between the reflectance of the target mineral characteristic band and the associated gangue characteristic band is taken, and the difference is normalized to 0 to 1 according to the preset calibration curve, which is used as the spectral reflectance value of the pixel. In this embodiment, the target mineral is gold mineral phase, the associated gangue is quartz gangue phase, the characteristic band is the reflectance difference between the two bands of 980nm and 650nm, the calculated spectral reflectance value is assigned to the corresponding laser reflection point, and if the projection does not correspond to any multispectral pixel, the laser reflection point is removed. Constructing a surface-fused point cloud matrix Surface fusion point cloud matrix The dimension is , This represents the total number of laser reflection points remaining after elimination. Columns 1 to 3 represent the three-dimensional spatial coordinates of each laser reflection point. The unit is millimeters. The fourth column is the spectral reflectance value of the laser reflection point, which is dimensionless and ranges from 0 to 1. Since laser scanning can only obtain the spatial coordinates of the ore's upward-facing surface and cannot directly form a closed three-dimensional entity, this embodiment performs a symmetric envelope reconstruction operation. Specifically, it fuses the surface point cloud matrix. Least squares plane fitting is performed on all spatial coordinates to obtain the fitted plane and its normal vector. ; Based on the fitting plane normal vector unit vector in the Z-axis direction Calculate the rotation matrix The rotation matrix satisfies The rotation matrix is calculated using the Rodriguez rotation formula. ; Surface fusion point cloud matrix A rigid body transformation is performed on the three-dimensional coordinates to obtain the coordinate matrix in the local coordinate system. The specific process includes: Extracting the surface fused point cloud matrix coordinate matrix in coordinate matrix For surface fusion point cloud matrix The coordinate matrix formed by the first three columns of data; Calculate the coordinate matrix The centroid of all three-dimensional spatial coordinates in the coordinate matrix Subtract the centroid coordinates from all three-dimensional spatial coordinates, then multiply by the rotation matrix on the left. The transpose of the matrix is used to perform the rigid body transformation; After the rigid body transformation, the fitting plane passes through the origin of the local coordinate system and its normal is aligned with the local Z-axis. The origin of the local coordinate system is the centroid of the surface, which is the average position of the laser reflection points. The Z-axis of the local coordinate system is aligned with the normal vector of the fitting plane. coincide, point Consistent with this, the X and Y axes of the local coordinate system are determined by the rotation matrix. In During the rotation to the Z-axis, it is simultaneously determined that the two are mutually orthogonal and orthogonal to the Z-axis. The local coordinate system is a right-handed rectangular coordinate system, defined by the centroid and the rotation matrix. Sure; coordinate matrix in local coordinate system Each point in Construct its mirror point After the mirroring is completed, the mirrored points are inversely transformed back to the original coordinate system to obtain the spatial coordinates of the mirrored points in the coordinate system of the 3D laser profilometer. The spectral reflectance value of the mirror point is the spectral reflectance value of the corresponding surface laser reflection point. The 3D spatial coordinates of the surface laser reflection point and the mirror point are merged to obtain a merged point cloud matrix. Using this merged point cloud matrix as input, a 3D convex hull algorithm, such as the fast convex hull algorithm, is applied to output the convex hull vertex matrix. Dimension ,in This represents the number of vertices in the convex hull. Each line contains the coordinates of a vertex of the convex hull, in millimeters. Output convex hull patch index matrix Dimension ,in Let be the number of convex hull triangles, and each row represents the three vertices of that triangle. Row index in; Based on convex hull vertex matrix The minimum and maximum values of the 3D convex hull in the x, y, and z directions are obtained to obtain the bounding box. Then, grid points are uniformly generated in the bounding box along the three coordinate axes with a step size of 5 mm. The coordinates of the grid points form a candidate point matrix, with each row being the 3D coordinates of a candidate point in millimeters.
[0020] For each candidate point in the candidate point matrix, it is necessary to determine whether the point is located inside the 3D convex hull. The determination method is as follows: emit a ray in any direction from the point, and count the index matrix of the ray and the convex hull facets. The number of intersections of all triangular faces in the convex hull; if the number of intersections is odd, the point is inside the convex hull; if the number of intersections is even, the point is outside the convex hull. The system retains all candidate points located inside the convex hull as internal fill points, forming an internal fill point coordinate matrix. ; Surface laser reflection points, mirror points, and internal filling points together constitute the fused point cloud matrix of the three-dimensional entity. fusion point cloud matrix The dimension is , This represents the total number of surface points, mirror points, and internal filling points. Columns 1 to 3 contain the 3D coordinates of each point. The unit is millimeters. The fourth column is the spectral reflectance value. The values for surface laser reflection points and mirror points are 0 to 1, and the values for internal filling points are -1. Crusher operating data includes the extrusion force vector; Specifically, a piezoelectric triaxial force sensor is deployed on the main bearing base of the moving jaw of the jaw crusher, with a sampling frequency of 1000Hz, to pick up the dynamic reaction force transmitted through the bearing base when the crusher's jaw plates crush the ore.
[0021] For the triaxial force signals acquired by the piezoelectric triaxial force sensor, the system performs high-pass filtering on each axis signal with a cutoff frequency of 10Hz within the time window. Internally, the composite force amplitude is calculated from the high-pass filtered triaxial force signal. The sampling moment when the composite force amplitude reaches its maximum value is extracted, and the triaxial force components at that moment are used to form a compressive force vector. Extrusion force vector for Column vector, unit is Newton.
[0022] In the recognition module, the process of obtaining the heterogeneous texture vector and the heterogeneous phase boundary density includes: Points with spectral reflectance values within the effective range are extracted from the fused point cloud matrix and used as valid spectral point cloud data. The purpose of this step is to extract the orientation of the contact interface between the target mineral and the associated gangue by utilizing the spatial gradient change of spectral reflectance on the ore surface, and to construct a geometric constraint tensor for subsequent mesh generation. Calculate the average absolute value of the difference between the spectral reflectance values of each point and its neighboring points in the effective spectral point cloud data, and use it as the spectral gradient of the point. Construct a boundary point set matrix for points whose spectral gradient is greater than a preset spectral gradient threshold. From the fused point cloud matrix Extract the points in the fourth column whose spectral reflectance values are in the range of 0 to 1 to form an effective spectral point cloud matrix. Effective spectral point cloud matrix The dimension is , The total number of effective spectral points, for the effective spectral point cloud matrix. For each point in the spectrum, calculate the Euclidean distance between that point and the other effective spectral points, sort them in ascending order, select the top 10 effective spectral points as adjacent points, calculate the average of the absolute values of the differences between the spectral reflectance value of that point and the spectral reflectance values of its 10 adjacent points, and use this as the spectral gradient of that point. The larger the spectral gradient, the more likely there is a contact interface between different mineral phases at that point. Extracting spectral gradients greater than the spectral gradient threshold All valid spectral points constitute the boundary point set matrix. Boundary point set matrix The dimension is , This represents the total number of extracted boundary points, with each line containing the three-dimensional spatial coordinates of one boundary point. The unit is millimeters; In this embodiment, the spectral gradient threshold The value was 0.20, obtained through offline calibration experiments. The experimental procedure was as follows: for the same batch of heterogeneous gold ore samples, under the same illumination and scanning conditions, spectral gradient thresholds were set respectively. Multispectral images and 3D point clouds were acquired with thresholds of 0.10, 0.20, and 0.30. Based on manually annotated CT / lithological thin section images, the accuracy of boundary point recognition and the false recognition rate of noise points under different thresholds were statistically analyzed. The experimental results are shown in Table 1. Table 1: Spectral Gradient Thresholds Comparison table of boundary point recognition accuracy; As shown in Table 1, when the spectral gradient threshold is 0.20, the boundary point recognition accuracy reaches 96.2%, and the noise point misjudgment rate is only 2.1%, which is the best overall performance. Therefore, in this embodiment, 0.20 is chosen as the spectral gradient threshold value. The principal component analysis algorithm is called to solve for the eigenvalues and eigenvectors of the covariance matrix of the boundary point set matrix. The eigenvectors corresponding to the minimum eigenvalues are extracted and normalized to obtain the heterogeneous texture vectors. Boundary point set matrix As input, the principal component analysis algorithm is called to solve for the boundary point set matrix. The eigenvalues and eigenvectors of the covariance matrix are obtained, the eigenvectors corresponding to the smallest eigenvalues are extracted, and then normalized to obtain the heterogeneous texture vectors. Heterogeneous texture vectors for The column vector of is dimensionless and represents the direction of minimum variance of the boundary point cloud, i.e., the normal tendency of the mineral contact interface in space. The ratio of the total number of points in the boundary point set matrix to the total number of points in the effective spectral point cloud data is used as the heterogeneous phase boundary density. Calculate the total number of boundary points Total number of effective spectral points The ratio of the two values yields the boundary density of the heterogeneous phase. heterogeneous phase boundary density It is a dimensionless scalar quantity, ranging from 0 to 1, used to characterize the density of the interfaces between different minerals on the ore surface, and the heterogeneous phase boundary density. The larger the value, the more complex the internal embedding of the ore; The process of obtaining the interface metric tensor includes: An interface metric tensor is constructed based on heterogeneous texture vectors and heterogeneous phase boundary densities.
[0023] Based on heterogeneous texture vectors heterogeneous phase boundary density Calculate the interface metric tensor Interface metric tensor The mathematical expression is: ; in , for the interface metric tensor for identity matrix The boundary density of the heterogeneous phase. For heterogeneous texture vectors, This is the transpose of the heterogeneous texture vector; Interface metric tensor in It degenerates into an isotropic identity matrix at time. Time-based heterogeneous texture vector Enhanced spatial metric, and always a positive definite matrix.
[0024] This invention utilizes spectral gradients to extract the interface orientation and constructs an interface metric tensor. Through dimensionality reduction analysis, the interface orientation of heterogeneous phases is transformed into a mathematical tensor, which enables the abstract visual material differences to have clear geometric constraints, ensuring that the simulation substrate can closely match the mosaic interface structure of the real target minerals and gangue.
[0025] In the topology module, the process of obtaining the control repulsion radius includes: The control repulsion radius is calculated based on the heterogeneous phase boundary density, the preset maximum grid spacing, and the preset geometric scaling constant. This step utilizes the interface metric tensor. and heterogeneous phase boundary density This generates a static tetrahedral topological basis that does not split in subsequent mechanical derivations, thereby avoiding memory overflow caused by infinite subdivision of the finite element mesh in crack simulation. Specifically, the control repulsion radius is calculated, and its mathematical expression is as follows: ; in, To control the repulsion radius, The preset maximum grid spacing is 10.0 mm. This is a preset geometric scaling constant, with a value of 0.5, which is dimensionless. This represents the boundary density of the heterogeneous phase. The larger the radius, the better the control of the repulsion radius. The smaller the value, the denser the candidate nodes will be generated in the complex regions of the mineral contact interface to adapt to the complex embedding structure. The process of generating discrete nodes includes: Using the 3D convex hull corresponding to the fused point cloud matrix as the spatial envelope, and using the repulsion radius as the minimum node spacing parameter within the spatial envelope, the Poisson disk sampling algorithm is called to generate the node coordinate matrix. With fused point cloud matrix The three-dimensional convex hull of all points is used as the spatial envelope, that is, the convex hull vertex matrix. and convex hull patch index matrix As input, to control the repulsion radius Use the minimum node spacing parameter to generate discrete nodes using the Poisson disk sampling algorithm. Poisson disk sampling algorithm outputs node coordinate matrix Node coordinate matrix The dimension is , The total number of nodes generated; each line records the three-dimensional spatial coordinates of a node, in millimeters. The process of generating a tetrahedral topological basis includes: With the node coordinate matrix and interface metric tensor as input, the anisotropic Thiessen polygon algorithm is called. The anisotropic Thiessen polygon algorithm uses the inverse matrix of the interface metric tensor as the metric tensor to define the anisotropic distance between nodes and performs spatial partitioning on the node coordinate matrix to obtain polyhedral regions. The constrained Delaunay tetrahedral partitioning algorithm is called to partition the polyhedral region to obtain topological connection data. The node coordinate matrix and the topological connection data together constitute the tetrahedral topological basis.
[0026] With node coordinate matrix and interface metric tensor As input, the anisotropic Thiessen polygon algorithm is invoked, which uses the interface metric tensor. As a metric tensor, any two nodes are defined and The anisotropic distance between them is ,in Interface metric tensor The inverse matrix; With node coordinate matrix Each node in the array is a seed point. The space is divided according to the anisotropic distance to obtain a set of polyhedral regions. Each polyhedral region contains a seed node. The anisotropic distance from any point in the region to the seed node is less than the distance to any other seed node. The polyhedral regions are represented by a vertex coordinate list and a face index list. Extract the polyhedral regions output by the anisotropic Thiessen polygon algorithm, and perform tetrahedral subdivision of each polyhedral region using the constrained Delaunay tetrahedral subdivision algorithm. The constrained Delaunay tetrahedral subdivision algorithm outputs a topological connectivity matrix. ; Topology connection matrix The dimension is , To represent the total number of tetrahedral elements generated, each row records the coordinates of the four nodes of a tetrahedral element in the node coordinate matrix. Row index in; Node coordinate matrix and topological connectivity matrix Together they form a heterogeneous mineral crystal tetrahedral topological substrate ; This invention solves the problem of computational memory overflow caused by infinite mesh subdivision in traditional finite element method simulation of crack propagation. By automatically generating a dense mesh in the complex interface region of mineral contact and keeping it sparse in the homogeneous region, it achieves optimal allocation of computing resources on a static and non-split topological substrate, ensuring high resolution of microscopic physical deformation calculation.
[0027] In the calculation module, the process of obtaining the outer contour nodes includes: Traverse the tetrahedral elements in the topology connection data, count the number of times each triangular face appears in all tetrahedral elements, extract and remove duplicate nodes from the triangular face that appears only once, and obtain the outer contour nodes. This embodiment utilizes an unencrypted heterogeneous mineral tetrahedral topological substrate. Under these conditions, finite element static analysis is used to transform the measured extrusion pressure vector F into nodal force loads, solve the nodal displacements, and then calculate the strain, strain energy and micro-damage state inside the ore to locate intergrowth particles that are difficult to grind and have high residual energy. Specifically, the collected extrusion force vector is received. ; This embodiment uses a heterogeneous mineral tetrahedral topological substrate. Nodes located on the mesh boundary surface are designated as outer contour nodes. These outer contour nodes are used as the loading objects in the finite element analysis on a heterogeneous mineral tetrahedral topological substrate. In this process, determining the outer contour nodes includes: Traversing the topological connectivity matrix All tetrahedral elements, topological connection matrix Each line records the 4 node indices of a tetrahedral element. For each tetrahedral element, its 4 triangular faces are extracted. Each triangular face is composed of 3 node indices in the tetrahedral element. The 3 node indices of each triangular face are arranged in ascending order to generate a unique identifier for the triangular face. Count the number of times each triangular facet identifier appears in all tetrahedral elements. If a triangular facet identifier appears only once, then that triangular facet belongs to the heterogeneous mineral tetrahedral topological basis. On the mesh boundary surface, if a triangular facet appears twice or more, then the triangular facet is shared by multiple tetrahedral elements and belongs to the mesh interior facet. Extract the node indices from all the triangular faces belonging to the mesh boundary surface, remove duplicates to obtain the outer contour node index set, and each node in this set is the outer contour node. The process of obtaining nodal displacement and nodal strain tensors includes: Divide the modulus of the extrusion force vector by the total number of outer contour nodes to obtain the nodal force load of each outer contour node along the direction of the extrusion force vector. The extrusion force vector The modulus length is divided by the total number of all outer contour nodes, and the result is used as the nodal force load on each outer contour node. In this embodiment, the nodal force load is used as the force load along the extrusion force vector. A force of the same magnitude and direction is applied to all outer contour nodes, and is a compressive force vector. In this embodiment, the macroscopic extrusion force of the crusher tooth plate on the ore is equivalent to the nodal force load evenly distributed on the nodes of the outer surface of the ore. This ensures that the total resultant force is equal to the actual measured extrusion force and avoids the problem of not being able to accurately distribute the spatial force due to the unknown ore posture. The units of the node coordinate matrix are uniformly converted to the International System of Units (SI). The overall stiffness matrix is assembled based on the converted coordinate data and topological connection data. The element volume and strain displacement matrix are calculated using the tetrahedral shape function and its gradient. The elastic matrix is obtained by combining the linear elastic constitutive relation. Then, the element stiffness matrix of each tetrahedral element is calculated and accumulated into the overall stiffness matrix according to the node index. After applying nodal force loads, and before proceeding with subsequent element assembly and mechanical calculations, an SI nodal coordinate matrix is generated for finite element calculations. Specifically, the heterogeneous mineral crystal tetrahedral topological substrate The three-dimensional spatial coordinates of the mid-nodes are uniformly converted from millimeters to meters to ensure dimensional consistency in the finite element method and phase field partial differential equation calculations. Subsequent finite element method and phase field partial differential equation calculations are all based on... The spatial scatter matrix calculation still uses the original millimeter coordinates. ; Based on the SI nodal coordinate matrix and heterogeneous mineral tetrahedral topological substrate The topological connection relationship assembles the stiffness matrix of tetrahedral elements. The stiffness matrix of a tetrahedral element is determined by the linear elastic constitutive relation and the tetrahedral shape function; Specifically, for the topological connectivity matrix Each tetrahedral unit in Extract the coordinate indices of the four nodes of this unit, and then use these coordinate indices to obtain the coordinates from the SI node coordinate matrix. Extract the tetrahedral unit The coordinates of each node form a Local coordinate matrix ; Based on local coordinate matrix The volume of a tetrahedral element is calculated using the tetrahedral shape function and its gradient. and strain-displacement matrix ; The dimension is Because each node has 3 degrees of freedom, and 4 nodes have a total of 12 degrees of freedom, there are 6 corresponding strain components. ; Using the linear elastic constitutive relation, the elastic modulus Compared to Poisson Convert to elasticity matrix , The dimension is Describes the linear relationship between stress and strain; ore elastic modulus Compared to Poisson Based on the ore mechanical tests, this embodiment takes... , Node displacement column vector The dimension is The unit is meters, and each node stores data in meters. , , Displacement components in three directions; Calculate the degenerate elasticity matrix based on the damage variables of the node in the current iteration step. , ,in, For each tetrahedral element, its element degenerate elasticity matrix Take four node damage variables The arithmetic mean; Calculate the element stiffness matrix of this tetrahedral element. The mathematical expression for the element stiffness matrix is: ; in, The element stiffness matrix, This represents the volume of the tetrahedral unit, in cubic meters. The strain-displacement matrix, This is the transpose of the strain-displacement matrix. The degenerate elasticity matrix of the unit; The dimension is ; Initialize the global stiffness matrix to all zeros , dimension , Let be the total number of nodes, and each node has 3 degrees of freedom, therefore the total number of degrees of freedom is . ; For the element stiffness matrix of each tetrahedral element , and The elements are accumulated into the global stiffness matrix according to the positions of the four nodes of the element in the overall degrees of freedom. Specifically, in the corresponding rows and columns: Unit 1 The node has an index of [number] in the overall node. The corresponding three degrees of freedom are located in the row / column positions of the overall matrix. , , ; Will The Middle The corresponding nodes Sub-block, add to The Middle The corresponding nodes At the position of the sub-block.
[0028] Repeat the above operation for all elements to obtain the final global stiffness matrix. After assembly, It is A real symmetric sparse matrix whose main diagonal elements are always non-negative; Solve the system of linear equations containing the global stiffness matrix and nodal force loads to obtain the column vector of nodal displacements; Solve the system of linear equations The nodal displacement column vector is obtained. ,in This is the column vector of external loads for all nodes; the component corresponding to unloaded nodes is 0. Calculate the tensile strain energy at the nodes and substitute it into the phase field damage partial differential equation for updating. If the node damage variable in adjacent iteration steps The maximum change is greater than the convergence threshold of 10. -4 Then return to the global stiffness matrix assembly step, using the updated... Recalculate and Repeat the iteration; otherwise, output the set of node damage variables. For heterogeneous mineral tetrahedral topological substrates Each node in , From 1 to The integer, based on the nodal displacement column vector obtained from the solution. ; Nodal displacement column vector The dimension is The unit is meters. The displacement components of each node in the x, y, and z directions are stored in order of node number. Each node stores its x-direction displacement component in [the following locations]. The Line, the displacement component in the y-direction is stored in The Line, the z-direction displacement component is stored in The Okay, the external load component corresponding to the unloaded node is 0, but the node displacement still exists, which can be obtained by solving the global stiffness equation simultaneously; For each node in the tetrahedral topological basis, the strain tensor of the tetrahedral elements containing that node is extracted and averaged by element volume to obtain the nodal strain tensor of that node. compute nodes Nodal strain tensor in adjacent tetrahedral elements ; Specifically, for each tetrahedral element, the nodal displacements of that element are used. and strain-displacement matrix Calculate the constant strain tensor of the unit Element node displacement from Extracted by node index; For containing nodes The strain tensors of all tetrahedral elements are obtained by weighting the element volume and then calculating the nodal strain tensors. strain tensor ; strain tensor of node i It is The matrix is a symmetric matrix, where the main diagonal elements represent the normal strain in the x, y, and z directions, and the off-diagonal elements represent the shear strain components in the xy, yz, and zx planes. The element should be transformed to the node, which facilitates subsequent energy decomposition and damage calculation on a node-by-node basis.
[0029] This invention solves the problem by extracting the force load distributed across the outer contour nodes and assembling the overall stiffness matrix that takes into account damage degradation. This avoids the technical problem of being unable to apply spatially distributed loads due to the random and unknown orientation of the ore feed. By solving a system of linear equations containing degradation functions, the true displacement and strain tensor of each node inside the ore can be calculated without changing the basic mesh topology.
[0030] In the calculation module, the process of decomposing the nodal strain tensor includes: Based on the linear elastic constitutive relation, the nodal strain tensor is converted into elastic strain energy density. Then, using an asymmetric energy decomposition method based on the principal strain symbol, the elastic strain energy density is decomposed into nodal tensile strain energy and nodal compressive strain energy.
[0031] Since the subsequent phase-field fracture equation requires an energy term to drive the damage evolution, the essence of rock fracture is that the work done by external force is converted into strain energy. When the tensile strain energy exceeds the rock's fracturing energy, the crack will propagate. Therefore, in this embodiment, the strain is first converted into an energy scalar before the phase-field fracture equation is constructed. This embodiment converts strain into an energy scalar based on the linear elastic constitutive relation. Specifically, it calculates the nodes... elastic strain energy density, elastic strain energy density The mathematical expression is: ; in, and Let be the Lamé constant, and its conversion formula is: , , by elastic modulus Compared to Poisson The conversion yielded the result. For nodes The nodal strain tensor is the trace operator for a matrix; is the tensor double dot product; The elastic strain energy density is obtained by using an asymmetric energy decomposition method based on principal strain symbols. Decomposed into nodal tensile strain energy and nodal compressive strain energy Nodal tensile strain energy and nodal compressive strain energy The units are all joules per cubic meter, nodal tensile strain energy. and nodal compressive strain energy The mathematical expression is: ; ; in, , ; and From strain tensor Specifically, it is obtained through spectral decomposition, that is, Represented as , To adapt to changing circumstances The main strain direction; ; .
[0032] This invention ensures that only tensile strain energy that leads to microcrack propagation is used as the driving force for damage evolution by eliminating compressive strain energy that does not cause rock cracking, while retaining compressive strain energy for subsequent identification of indestructible intergrowths. This energy diversion mechanism improves the simulation realism of microfracture simulation. The process of creating a set of node damage variables includes: Because rocks are difficult to break under compression but prone to cracking under tension, the phase-field damage partial differential equation only allows tensile strain energy to drive damage, while compressive strain energy is retained and not used as an energy source for crack propagation. The tensile strain energy obtained after decomposition is... Used to drive damage variables Evolution of compressive strain energy This is then used as residual mechanical strain energy and included in the calculation of subsequent statistical sampling weights; Since rock failure is mainly characterized by tensile fracture, phase field fracture evolution is solely determined by tensile strain energy. Driven by tensile strain energy, this embodiment uses tensile strain energy. As energy input, on a heterogeneous mineral tetrahedral topological substrate Solve a phase-field damage partial differential equation at each node to obtain the damage variable for that node. Damage variables It is a dimensionless scalar, initialized to 0, and its value ranges from 0 to 1. Represents a node No fracture or dissociation occurred at that location. Represents a node The area has completely fractured and peeled off along the grain boundary; Using the tensile strain energy at the nodes as the energy input, a partial differential equation of phase field damage is constructed on each node of the tetrahedral topological substrate. The set of node damage variables for all nodes is obtained by solving the partial differential equation of phase field damage using the explicit Euler method. The mathematical expression for the partial differential equation of phase field damage is: ; in, For nodes In the Node damage variables for each iteration step Represents a node In the Damage variables for each iteration step; The time step is in seconds. Damage evolution migration rate, expressed in cubic meters per joule second; Damage degradation function right The derivative; For nodes The tensile strain energy, expressed in joules per cubic meter; Critical energy release rate, expressed in joules per square meter; This is a parameter for crack width, in meters; For nodes In the Discrete Laplace operator values at each time step; The initial conditions for the phase field damage partial differential equation are all nodes. The boundary condition is 0; zero flux boundary is used on the outer surface of the tetrahedral topological basis. In this embodiment , , , ; The critical energy release rate was determined by a three-point bending test. Specifically, a three-point bending specimen was cut from the heterogeneous gold ore in this embodiment. The specimen was placed on a three-point bending test fixture, a concentrated load was applied, and the load-displacement curve was recorded until the specimen completely fractured. The fracture toughness was calculated from the peak load and the specimen geometry according to the fracture toughness calculation formula. Then, under plane strain conditions The critical energy release rate was obtained by conversion. ,in, For elastic modulus, Poisson's ratio; Crack width parameters These are parameters controlling the crack transition zone width in the phase-field damage model. This embodiment uses a numerical experimental comparison method for calibration, utilizing experimentally measured values. , , A partial differential equation for phase-field damage was established, and a numerical simulation of a three-point bending test was performed. The results were adjusted under the same mesh size. This ensures that the crack morphology and load-displacement curves obtained from numerical simulation match the experimental observations. In this embodiment, a heterogeneous mineral tetrahedral topological substrate is used. The feature mesh size is After calibration This allows the crack transition zone width to cover to Each tetrahedral element ensures numerical stability and crack morphology accuracy. Damage evolution mobility The rate of evolution of damage variables over time is controlled. This embodiment uses a relaxation test method for calibration. A constant displacement is applied to a three-point bending specimen and held, and the relaxation curve of the load over time is recorded. The experimental relaxation curve is then compared with different... The numerical relaxation curves of the phase-field model under different values were compared, and the curve that makes the decay rates of the two consistent and has the fastest numerical iteration convergence was selected. The value, obtained through calibration in this embodiment, is... ; After each update, if Then let ;like Then let ; Discrete Laplace operator value heterogeneous mineral tetrahedral topological substrate The above calculation uses the node adjacency graph Laplace discrete scheme. The mathematical expression of the discrete Laplace operator is: ; in, For nodes The set of adjacent nodes in a tetrahedral topological basis; For nodes With nodes The total lateral surface area of the tetrahedron, in square meters; For nodes With nodes Euclidean distance, in meters; For nodes The equivalent control volume, taking the node One-quarter of the sum of the volumes of adjacent tetrahedra, in cubic meters; Damage Degradation Function The mathematical expression is: ; in, For nodes Node damage variables, To prevent tiny positive numbers with singular stiffness during complete damage, we take 10. −6 ; After the phase-field damage partial differential equation is solved iteratively and converges, the damage variables of all nodes are... Arranged into a damage column vector according to node index order. Damage column vector The dimension is Each element represents a node's damage variable; The degree of dissociation is the arithmetic mean of all elements in the set of node damage variables; For the damage column vector The degree of dissociation is obtained by taking the arithmetic mean of all elements in the equation. degree of dissociation It is a dimensionless scalar, with a value range from 0 to 1. The closer it is to 1, the more completely the current batch of ore is dissociated along the grain boundaries; The closer it is to 0, the worse the dissociation.
[0033] The residual mechanical strain energy is the sum of the nodal compressive strain energy of all nodes.
[0034] Compressive strain energy of all nodes The residual mechanical strain energy is arranged in column vector order according to node index. Residual mechanical strain energy column vector The dimension is The unit of the element is joules per cubic meter.
[0035] This invention enables dynamic and continuous tracking of the grain boundary fragmentation and exfoliation process in ore, eliminating the lag of relying on offline chemical phase or microscopic analysis. It can intuitively output quantitative indicators reflecting the dissociation effect of the entire batch of ore based on the equation iteration convergence results, and accurately mark nodes with high residual mechanical energy, providing a reliable physical state for locating and cutting off difficult-to-break intergrowth particles.
[0036] In the sampling module, the process of calculating the node sampling weight distribution includes: For each node, the node sampling weight is calculated based on the node damage variable and the corresponding compressive strain energy in the residual mechanical strain energy set of that node. Traditional sampling typically uses fixed-proportion or random sampling, with sampling probability independent of the ore's physical state. This embodiment, however, first calculates the physical state parameters of each node, then transforms these parameters into a sampling weight using a formula. This weight is then used as the basis for determining the probability of each node being selected during sampling. The physical state parameters represent the node's wear resistance and residual energy level. Specifically, for each node Extract nodes Node damage variables and compressive strain energy Calculate the node sampling weight. The node sampling weight refers to the algebraic mapping of the priority interception probability assigned to particles that are physically in an undissociated state and have high internal residual mechanical energy in statistics. It is a dimensionless scalar. The mathematical expression for the node sampling weight is: ; in, Let be the dimension alignment constant, with a value of . This is used to make the exponential term dimensionless because The unit is joules per cubic meter; To prevent division by a small constant, the value is set to... Used to prevent When the value is 0, the division overflows; The upper limit of the preset exponent is set to 50 to prevent the exponent from overflowing. This truncation only affects the upper bound of the weights of extreme nodes and will not change the relative order of the node weights. When node Damage variables Near 0 and compressive strain energy When it is larger, it represents a node. The grain boundaries are not fractured and the residual mechanical energy is high, belonging to a difficult-to-grind, high-residual-energy intergrowth. Very small Larger Larger, via natural index After magnification, node sampling weights This will significantly increase the size of the node, making it a priority to be truncated in subsequent sampling; The node sampling weights of all nodes are normalized to obtain a node sampling weight distribution where the sum of all elements is 1.
[0037] Node sampling weights for all nodes Perform normalization processing, that is, each Divide by all nodes The sum of these values yields the node sampling weight vector. Node sampling weight vector The dimension is The elements are dimensionless scalars, and the sum of all elements is 1.
[0038] This invention utilizes the natural exponential function to directly map the damage variables of nodes and residual mechanical strain energy into sampling weights. By assigning a higher interception probability to high-risk intergrowth particles with high internal strain energy and which have not yet dissociated, the proportion of difficult-to-distinguish components in the sample is forcibly increased at the algorithm level. This ensures that the final intercepted sub-samples can expose the weak links in the crushing process of the current ore batch, maximizing the diagnostic value of the test data.
[0039] In the evaluation module, the process of constructing and updating the spatial scatter matrix includes: The elements in the node sampling weight distribution are sorted in descending order, and the nodes with the highest sorting ratio are extracted as high-risk nodes. If the intergrowth particles with high residual energy that are difficult to grind in the ore flow are spatially aggregated into clusters, then when the rotary divider is cut off at a fixed period, it may just miss or catch very few high-risk particles, resulting in the final test sample not being representative of the whole batch of ore. Therefore, it is necessary to quantify the spatial dispersion of particles in the ore flow. In traditional methods, the degree of dispersion is usually described by a covariance matrix or a scatter matrix. For particles in three-dimensional space, the x, y, and z coordinates of each particle need to be taken as samples. First, the mean of the coordinates of all particles is calculated, then the deviation of each particle from the mean is calculated, and finally the deviations are accumulated and divided by the total number of particles to obtain a 3×3 covariance matrix. The diagonal elements of this matrix reflect the variance in the three coordinate directions, and the off-diagonal elements reflect the correlation between the coordinates. To calculate the covariance matrix, the coordinates of all particles must first be saved. Since the mineral flow is continuous and unbounded, the number of particles is constantly increasing, and the number of coordinates saved in memory is also constantly increasing. Therefore, traditional methods cannot be directly used for online continuous mineral flow dispersion assessment. This embodiment solves the above problems by constructing and continuously updating a spatial scatter matrix. The spatial scatter matrix is used to retain the spatial distribution covariance information of high-risk nodes in the current window and historical windows in a weighted cumulative manner. Specifically, the node sampling weight vector All elements are sorted in descending order, and the top 2% of nodes are extracted as high-risk nodes. The 2% high-risk node ratio is determined based on the fact that the single interception mass of the divider in this embodiment accounts for about 1.5% to 2.5% of the ore mass in the window. 2% is taken as the upper limit of interception to ensure that the interception probability covers the main high-risk, difficult-to-grind, high residual energy intergrowth particles. Extract the three-dimensional spatial coordinates of high-risk nodes to construct a high-risk node coordinate matrix, calculate the mean vector of the high-risk node coordinate matrix, center the high-risk node coordinate matrix, and perform dimensionless processing based on the preset maximum grid spacing. The total number of high-risk nodes extracted is , , This indicates rounding up from the node coordinate matrix. Extract this The three-dimensional spatial coordinates of each high-risk node are combined to form a high-risk node coordinate matrix. High-risk node coordinate matrix The dimension is Each line records the three-dimensional spatial coordinates of a high-risk node; Calculate the coordinate matrix of high-risk nodes The mean vector of coordinates of all nodes in Coordinate mean vector for Row vectors, in millimeters, whose three elements represent the high-risk nodes at... , , The average of the coordinates in the three coordinate directions will form the coordinate matrix of the high-risk node. Subtract the mean vector of coordinates from each row The centralized high-risk coordinate matrix is obtained. Centralized high-risk coordinate matrix The dimension is The unit is millimeters; Centralize the high-risk coordinate matrix Divide by the preset maximum grid spacing The dimensionless high-risk coordinate matrix is obtained. ,Right now ; in, The value is 10.0 mm, and the dimensionless high-risk coordinate matrix is used. The dimension is Dimensionless; Based on the temporal forgetting factor, the dimensionless high-risk node coordinate matrix and its transpose matrix, and the total number of high-risk nodes, the spatial scatter matrix is calculated and updated incrementally. The mathematical expression for the spatial scatter matrix is: ; in, Let be the spatial scattering matrix. The temporal forgetting factor has a value of 0.95 and is dimensionless. It is used to gradually decay the historical spatial scattering matrix and avoid the accumulation of outdated mineral flow information from affecting the current judgment. This is the spatial scatter matrix of the previous calculation window; For dimensionless high-risk coordinate matrices, for transpose, This represents the total number of high-risk nodes. in, Represents a dimensionless high-risk coordinate matrix Multiplying the transpose of the matrix by itself yields a... A real symmetric matrix, where each element reflects the spatial distribution covariance information of high-risk interspersed nodes in the corresponding coordinate direction within the current calculation window. Dividing this matrix by... This ensures that the statistics are unaffected by changes in the total number of high-risk nodes. The addition operation accumulates the current covariance information into the weighted historical matrix from the previous time step, achieving a one-step incremental update of the spatial scatter matrix. Initialize to identity matrix; The process of calculating spatial segregation parameters includes: Find the inverse of the spatial scattering matrix, and calculate the sum of the main diagonal elements of the inverse matrix as the spatial segregation parameter.
[0040] Since the spatial scatter matrix is a matrix, it cannot be directly compared with the threshold and needs to be compressed into a scalar. This embodiment focuses on the spatial scatter matrix. Find the inverse matrix, and obtain the inverse matrix. Calculate the inverse matrix The sum of the main diagonal elements yields the spatial segregation parameter. ; Spatial segregation parameters It is a dimensionless positive scalar; the more concentrated the distribution of high-risk nodes, the more... The smaller certain eigenvalues are, The larger the sum of the elements on the main diagonal, the better. The larger, due to Furthermore, the initial matrix is the identity matrix, and each current term is positive semi-definite. It is always a positive definite matrix, and inversion is executable.
[0041] This invention retains spatial covariance information through weighted accumulation, enabling the acquisition of the segregation trend of hard-to-grind high residual energy particles agglomerating in space with extremely low memory overhead, providing a basis for judgment for subsequent hardware control of anti-segregation sampling.
[0042] In the control module, the process of generating the first control command includes: When the degree of dissociation is less than the preset dissociation threshold, the target discharge port gap is calculated. Based on the comparison between the target discharge port gap and the preset minimum discharge port gap, a first control command is generated. The first control command is used to control the discharge port gap. dissociation With preset dissociation threshold Comparison, preset lower limit of dissociation qualification The value is 0.75; when When the current batch of raw ore is determined to be abnormally dense, with insufficient mechanical energy and failure to penetrate grain boundaries, the target discharge port gap is calculated. The target discharge port gap is equal to the current discharge port gap minus the preset compensation step size of 3mm. If the target discharge port gap is greater than the preset minimum discharge port gap Send a closing control command to the jaw crusher, drive the hydraulic cylinder to push the moving jaw thrust plate, and adjust the discharge opening gap to the target discharge opening gap; If the target discharge port gap is less than or equal to the preset minimum discharge port gap In this embodiment The gap at the discharge port is adjusted to the mechanical limit position of the jaw crusher. And maintain, while triggering an alarm; After each first control command is sent, a 10-second action lockout period begins. During this lockout period, even if the subsequent calculation window meets the requirements again... It will not send closed control commands again; after the lock time expires, it will recalculate the latest window value. Determine whether to continue adjusting, when When the system determines that the current degree of dissociation is qualified, the gap at the crusher discharge port remains unchanged in its current physical state. Preset lower limit for dissociation The value was set at 0.75, and the experimental procedure for this value was as follows: for the same batch of heterogeneous gold ore samples, at different... Samples were prepared and tested under the selected values, and the pass rate of the tests and the frequency of crusher adjustment were statistically analyzed. The experimental results for these selected values are shown in Table 2. Table 2: Preset Lower Limit for Dissociation Completion Comparison table with representative pass rates of laboratory tests; The process of generating the second control command includes: The spatial segregation parameter is compared with the preset segregation threshold. When the spatial segregation parameter is greater than the preset segregation threshold, the time delay is calculated based on the ore transport distance and speed, and a second control command is generated. The second control command is used to control the sampling reducer.
[0043] Spatial segregation parameters Compared with the preset segregation threshold In comparison, this embodiment The value is 3.5; when At that time, it is determined that in the current ore flow, the refractory high residual energy intergrowths with high node sampling weights are spatially concentrated and are forming a concentrated material falling group. The fixed-period rotary divider action will lead to erroneous sampling. In this embodiment, the corresponding time delay is calculated based on the ratio of the physical distance of the ore from the crusher discharge port to the rotary divider inlet to the speed of the conveyor belt. After the corresponding time delay, it is determined that the batch of concentrated material feeding has reached the rotary divider. The motor of the rotary divider is then accelerated to perform high-frequency cutting, continuously rotating at a speed of 0.2 seconds per revolution for 4 revolutions. 0.2 seconds is the shortest single revolution cycle that the stepper motor of the divider can stably achieve under its limit torque. The duration of a single high-frequency cutting execution is 0.8 seconds. During the high-frequency cutting execution, even if the subsequent calculation window still meets the requirements... It will not be triggered repeatedly; after the current high-frequency interception is completed, it will be calculated based on the latest window. Determine if it should be triggered again; This high-frequency, intensive sampling method aims to capture as many high-risk, difficult-to-grind, and high-residual-energy intergrowths from concentrated material clusters as possible into the sample analysis container, thereby reducing the spatial segregation parameters of the subsequent calculation window. ,when When it is determined that there is no severe spatial segregation in the current particle flow, a hold command is sent to the rotary divider; The experimental procedure for determining the preset segregation threshold is as follows: For the same batch of heterogeneous gold ore samples, at different... Continuous sample reduction test was conducted under the given values. The number of abnormal high-frequency start commands of the rotary reducer motor and the relative standard deviation of the final output sub-sample were statistically analyzed. The experimental results are shown in Table 3. Table 3: Preset Segregation Thresholds Set up a comparison table of representative anti-segregation results from laboratory tests; Based on the results in Table 3, this embodiment will preset the segregation threshold. Fixed at 3.5; The first control command and the second control command are executed independently of each other, acting on the crusher actuator and the divider actuator respectively.
[0044] When faced with dense and difficult-to-grind ores, this invention can adaptively tighten the discharge port to ensure the dissociation effect. When high-risk particles aggregate in space, it can calculate the delay and drive the separator to intercept them at high speed and in a dense manner. This not only effectively prevents the leakage of segregated intergrowths into the sample and the resulting laboratory distortion, but also effectively avoids mechanical hardware wear caused by improper control frequency.
[0045] This invention overcomes the representativeness distortion problem of traditional offline static sampling when dealing with complex intergrowth particles in continuous ore flows by introducing multispectral and three-dimensional point cloud fusion, phase field damage partial differential equation calculation, and incremental update of spatial scatter matrix. This invention can capture the spectral gradient and geometric contour of the ore surface online in real time, and construct a heterogeneous mineral crystal tetrahedral topological basis based on it. Then, the measured crusher extrusion pressure is converted into nodal loads, and the tensile micro-damage evolution and compressive residual mechanical energy inside the ore are calculated. This invention gets rid of the dependence on pure macroscopic statistics and can directly quantify the grinding difficulty and dissociation state of each micro-node. Furthermore, the spatial scatter matrix is used to evaluate the spatial segregation degree of high-risk intergrowths in dynamic ore flows in real time, and adaptive closed-loop control commands are generated accordingly. This not only adjusts the crusher discharge port in time to ensure physical breakdown effect when the ore dissociation is insufficient, but also triggers high-frequency dense interception of the divider when high-risk particle segregation is identified. This greatly suppresses sampling deviation and ensures the authenticity and high reliability of the final test results in a high-flow continuous production environment.
[0046] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions can also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0047] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.
Claims
1. A digital twin control system for mineral processing, sampling, sample preparation, and testing, characterized in that, include: The data acquisition module is used to acquire ore point cloud data and crusher operating data; The identification module is used to extract boundary points based on the spatial gradient of spectral reflectance in the ore point cloud data, and to calculate the heterogeneous texture vector and heterogeneous phase boundary density to generate an interface metric tensor. The topology module is used to calculate the control repulsion radius based on the interface metric tensor and the heterogeneous phase boundary density, generate discrete nodes, and perform spatial partitioning and tetrahedral subdivision on the discrete nodes to generate a tetrahedral topological basis. The calculation module is used to obtain the outer contour nodes of the tetrahedral topological base, solve the node displacement of the outer contour nodes and calculate the node strain tensor based on the crusher working condition data, and decompose the node strain tensor to obtain the node tensile strain energy and node compressive strain energy. Based on the tensile strain energy of the nodes, the partial differential equation of phase field damage is constructed and solved to obtain the set of node damage variables, the degree of dissociation is calculated, and the compressive strain energy of the nodes is extracted as the set of residual mechanical strain energy. The sampling module is used to calculate the node sampling weight distribution based on the set of node damage variables and the set of residual mechanical strain energy. The evaluation module is used to extract high-risk nodes based on the node sampling weight distribution, construct and update the spatial scatter matrix, and calculate the spatial segregation parameters. The control module is used to compare the degree of dissociation with a preset dissociation threshold to generate a first control command, and to compare the spatial segregation parameter with a preset segregation threshold to generate a second control command.
2. The digital twin control system for mineral processing, sampling, preparation, and testing as described in claim 1, characterized in that, The ore point cloud data is a fused point cloud matrix obtained by fusing and reconstructing ore surface spectral data and ore surface three-dimensional spatial coordinates.
3. The digital twin control system for mineral processing, sampling, preparation, and testing as described in claim 2, characterized in that, The process of obtaining the heterogeneous texture vector and heterogeneous phase boundary density in the recognition module includes: Points with spectral reflectance values within the effective range are extracted from the fused point cloud matrix as valid spectral point cloud data; The absolute average of the difference between the spectral reflectance values of each point and its neighboring points in the effective spectral point cloud data is calculated as the spectral gradient of the point. Points with spectral gradients greater than a preset spectral gradient threshold are constructed as a boundary point set matrix. The principal component analysis algorithm is called to solve for the eigenvalues and eigenvectors of the covariance matrix of the boundary point set matrix. The eigenvectors corresponding to the minimum eigenvalues are extracted and normalized to obtain the heterogeneous texture vectors. The ratio of the total number of points in the boundary point set matrix to the total number of points in the effective spectral point cloud data is calculated as the heterogeneous phase boundary density. The process of obtaining the interface metric tensor includes: An interface metric tensor is constructed based on the heterogeneous texture vector and the heterogeneous phase boundary density.
4. The digital twin control system for mineral processing, sampling, sample preparation, and testing as described in claim 3, characterized in that, In the topology module, the process of obtaining the control repulsion radius includes: The control repulsion radius is calculated based on the heterogeneous phase boundary density, the preset maximum grid spacing, and the preset geometric scaling constant. The process of generating the discrete nodes includes: Using the three-dimensional convex hull corresponding to the fused point cloud matrix as the spatial envelope, and within the spatial envelope, with the control repulsion radius as the minimum node spacing parameter, the Poisson disk sampling algorithm is called to generate the node coordinate matrix. The process of generating the tetrahedral topological basis includes: Using the node coordinate matrix and the interface metric tensor as input, the anisotropic Thiessen polygon algorithm is invoked. The anisotropic Thiessen polygon algorithm uses the inverse matrix of the interface metric tensor as the metric tensor to define the anisotropic distance between nodes and performs spatial partitioning on the node coordinate matrix to obtain a polyhedral region. The constrained Delaunay tetrahedral partitioning algorithm is called to partition the polyhedral region to obtain topological connection data. The node coordinate matrix and the topological connection data together constitute the tetrahedral topological basis.
5. The digital twin control system for mineral processing, sampling, sample preparation, and testing as described in claim 4, characterized in that, The process of obtaining the outer contour nodes in the calculation module includes: Traverse the tetrahedral cells in the topology connection data, count the number of times each triangular face appears in all tetrahedral cells, extract and deduplicate the nodes on the triangular face that appears only once to obtain the outer contour nodes. The process of obtaining the nodal displacement and nodal strain tensors includes: Divide the magnitude of the extrusion force vector by the total number of the outer contour nodes to obtain the nodal force load of each outer contour node along the direction of the extrusion force vector. The units of the node coordinate matrix are uniformly converted to the International System of Units (SI). Based on the converted coordinate data and the topological connection data, the overall stiffness matrix is assembled. The element volume and strain displacement matrix are calculated using the tetrahedral shape function and its gradient. The elastic matrix is obtained by combining the linear elastic constitutive relation. Then, the element stiffness matrix of each tetrahedral element is calculated and accumulated into the overall stiffness matrix according to the node index. Solve the system of linear equations containing the overall stiffness matrix and the nodal force loads to obtain the nodal displacement column vector; For each node in the tetrahedral topological basis, the strain tensor of the tetrahedral element containing that node is extracted and averaged by element volume to obtain the nodal strain tensor of that node.
6. The digital twin control system for mineral processing, sampling, sample preparation, and testing as described in claim 5, characterized in that, The process of decomposing the nodal strain tensor in the calculation module includes: Based on the linear elastic constitutive relation, the nodal strain tensor is converted into elastic strain energy density. Then, using an asymmetric energy decomposition method based on the principal strain symbol, the elastic strain energy density is decomposed into nodal tensile strain energy and nodal compressive strain energy.
7. The digital twin control system for mineral processing, sampling, sample preparation, and testing as described in claim 6, characterized in that, The process of obtaining the set of node damage variables includes: Using the tensile strain energy of the nodes as energy input, a phase-field damage partial differential equation is constructed on each node of the tetrahedral topological substrate. The set of node damage variables for all nodes is obtained by solving the phase-field damage partial differential equation using the explicit Euler method. The degree of dissociation is the arithmetic mean of all elements in the set of node damage variables; The residual mechanical strain energy set is the nodal compressive strain energy of all nodes.
8. The digital twin control system for mineral processing, sampling, preparation, and testing as described in claim 7, characterized in that, The sampling module includes the following process for calculating the node sampling weight distribution: For each node, the node sampling weight is calculated based on the node damage variable of that node and the corresponding compressive strain energy in the set of residual mechanical strain energy. The node sampling weights of all nodes are normalized to obtain a node sampling weight distribution where the sum of all elements is 1.
9. The digital twin control system for mineral processing, sampling, preparation, and testing as described in claim 8, characterized in that, The process of constructing and updating the spatial scatter matrix in the evaluation module includes: The elements in the node sampling weight distribution are sorted in descending order, and the nodes with the highest sorting ratio are extracted as high-risk nodes. Extract the three-dimensional spatial coordinates of the high-risk nodes to construct a high-risk node coordinate matrix, calculate the mean vector of the high-risk node coordinate matrix, center the high-risk node coordinate matrix, and perform dimensionless processing based on the preset maximum grid spacing. Based on the temporal forgetting factor, the dimensionless high-risk node coordinate matrix and its transpose matrix, and the total number of high-risk nodes, the spatial scatter matrix is calculated and updated incrementally. The process of calculating the spatial segregation parameters includes: The inverse of the spatial scattering matrix is obtained, and the sum of the main diagonal elements of the inverse matrix is calculated as the spatial segregation parameter.
10. The digital twin control system for mineral processing, sampling, sample preparation, and testing as described in claim 9, characterized in that, In the control module, the process of generating the first control command includes: When the degree of dissociation is less than the preset dissociation threshold, the target discharge port gap is calculated, and based on the comparison result between the target discharge port gap and the preset minimum discharge port gap, the first control command is generated. The first control command is used to control the discharge port gap. The process of generating the second control command includes: The spatial segregation parameter is compared with a preset segregation threshold. When the spatial segregation parameter is greater than the preset segregation threshold, the time delay is calculated based on the ore transport distance and speed, and a second control command is generated. The second control command is used to control the sampling reducer.
Citation Information
Patent Citations
Tailing sampling test evaluation method, electronic equipment and storage medium
CN121805524A
