Low-latency Control System for Rare Earth Permanent Magnet Material Production Based on Digital Twin Technology

Abstract

The present invention belongs to the technical field of production control. The present invention discloses a low-latency control system for the production of rare earth permanent magnet materials based on digital twin technology, including: obtaining production sensing data of a rare earth permanent magnet material production line; using the production sensing data to construct a digital twin model of the production process of rare earth permanent magnet materials; using the digital twin model for simulation to obtain the change trends of key production parameters; calculating the sensitivity coefficient matrix of each key production parameter according to the change trends of the key production parameters; constructing a multi-objective optimization function based on the sensitivity coefficient matrix, and using an improved particle swarm algorithm to solve the multi-objective optimization function to obtain an optimal control parameter combination, which improves production efficiency and product quality, reduces energy consumption and raw material losses, improves resource utilization rate, and promotes green production.

Description

Technical Field

[0001] The present invention relates to the technical field of production control. More specifically, the present invention relates to a low-latency control system for rare earth permanent magnet material production based on digital twin technology. Background Art

[0002] The patent with the application publication number CN111091944A discloses a lanthanum-rich cerium-yttrium multi-main-phase fine-grained rare earth permanent magnet material and a preparation method thereof. Based on the characteristics of fast heating rate, short heating time and holding time of the spark plasma sintering technology, the interdiffusion and chemical inhomogeneity of elements inside the lanthanum-rich cerium-yttrium multi-main-phase magnet are controlled, various core-shell morphologies, magnetic hardening and pinning effects are precisely controlled, and the coercivity of the magnet is improved; while using rapid sintering to avoid abnormal grain growth, a stepwise heating is designed, the temperature is accurately controlled, the heating is uniform, overheating in local areas during the sintering process is avoided and the phenomenon of coexistence of coarse and fine grain regions is caused, the uniformity of the internal organizational structure of the magnet is realized, the grains are fine, the average grain size is less than 5 μm, the comprehensive magnetic properties of the lanthanum-rich cerium-yttrium multi-main-phase rare earth permanent magnet material are improved, meeting the commercial requirements. The preparation method is simple and convenient, easy to operate, and can accurately control the element distribution and organizational structure of the multi-main-phase magnet, reducing the preparation period and cost, and is suitable for large-scale production.

[0003] However, there are various challenges in the production control of existing rare earth permanent magnet materials. First, the ability to model and sense complex production processes is insufficient, resulting in the difficulty of meeting the requirements of high-quality production in terms of control accuracy and real-time performance; the data utilization efficiency is low, and it is difficult to fully explore and utilize valuable information in the production process, affecting the scientificity and timeliness of decision-making; secondly, the parameter optimization method is relatively rough, it is difficult to accurately identify and regulate key production parameters, resulting in poor control effects; in addition, it is often powerless to balance multiple production goals, and it is difficult to achieve comprehensive optimization in terms of performance, efficiency, energy consumption, etc.; at the same time, the system response speed is insufficient, and it is difficult to quickly adjust in the face of production fluctuations, increasing the risk of product quality fluctuations; finally, there is great room for improvement in energy management and raw material utilization, which not only affects production efficiency and product quality, but also leads to high production costs and resource waste.

[0004] In view of this, the present invention proposes a low-latency control system for rare earth permanent magnet material production based on digital twin technology to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention provides the following technical solution: A low-latency control system for rare earth permanent magnet material production based on digital twin technology, comprising: a data acquisition module for acquiring production sensing data of a rare earth permanent magnet material production line;

[0006] A model construction module that uses production sensing data to construct a digital twin model of the rare earth permanent magnet material production process;

[0007] A simulation module that uses the digital twin model for simulation to obtain the changing trends of key production parameters;

[0008] A matrix fitting module for calculating the sensitivity coefficient matrix of each key production parameter according to the changing trends of the key production parameters;

[0009] A parameter optimization module that constructs a multi-objective optimization function based on the sensitivity coefficient matrix, uses an improved particle swarm algorithm to solve the multi-objective optimization function, and obtains an optimal control parameter combination; Each module is connected by wired and / or wireless means.

[0010] Further, the production sensing data includes temperature data, pressure data, flow data, electrical parameters, material property data, fluid-related data, and diffusion-related data;

[0011] The temperature data includes the melting furnace temperature, sintering furnace temperature, mold temperature, and cooling temperature; The pressure data includes the press pressure, air pressure, and hydraulic pressure; The flow data includes the raw material input flow, coolant flow, and protective gas flow; The electrical parameters include the motor current, heating power, and control system voltage; The material property data includes the material density, material specific heat capacity, and material thermal conductivity; The fluid-related data includes the fluid viscosity and fluid density; The diffusion-related data includes the diffusion coefficient and concentration distribution of elements.

[0012] Further, the construction method of the digital twin model includes:

[0013] Preprocess the production sensing data by cleaning, denoising, imputing, and standardizing, and convert the preprocessed production sensing data into a unified format; Unify the timestamps of all data to the same frequency and starting point; And perform dimensionless processing on data with different dimensions; Perform one-hot encoding or ordinal encoding on categorical data; Obtain complete sensing data;

[0014] Construct a heat conduction model, a fluid dynamics model, and a diffusion model based on the complete production data; Establish interfaces between the heat conduction model, the fluid dynamics model, and the diffusion model, and couple the heat conduction model, the fluid dynamics model, and the diffusion model to obtain a digital twin model.

[0015] Further, the construction method of the heat conduction model includes:

[0016] Define a control function and the calculation domain of the control function, and the calculation domain includes the melting furnace, the press mold, the sintering furnace, and the cooling device;

[0017] The control function includes the heat conduction equation, boundary conditions, and initial conditions; the expression of the heat conduction equation is: , where is the density, is the specific heat capacity, is the temperature, is the time, is the heat conduction coefficient, is the internal heat source, is the gradient symbol;

[0018] The boundary conditions include known temperature conditions, known heat flux conditions, and a linear combination of known convection and radiation; the initial condition is the given initial temperature distribution;

[0019] The computational domain is discretized to generate a computational grid; the computational grid includes structured grids and unstructured grids; the structured grids and unstructured grids are refined to obtain a model processing grid; based on the model processing grid, the control function is integrated to obtain a heat conduction model.

[0020] Furthermore, the method of refinement includes:

[0021] Calculate the curvature distribution of the boundary of the computational domain and define a curvature-sensitive weight function ; the method of defining the curvature-sensitive weight function includes:

[0022] Based on the curvature distribution, obtain the curvature value of any node in the structured grid and unstructured grid , if the curvature value of any node is greater than 0, then the curvature-sensitive weight function ; where is the reference curvature, is the critical curvature value;

[0023] If the curvature value of any node is less than or equal to 0, then the curvature-sensitive weight function ; where and are both adjustable weight coefficients;

[0024] According to the curvature-sensitive weight function, construct a grid size field ; where is the desired reference grid size, and are the horizontal axis, vertical axis, and vertical axis of the three-dimensional coordinate system where the grid size field is located, respectively;

[0025] Use the grid size field to further divide the structured grid and unstructured grid; obtain the model processing grid.

[0026] Further, the method for further division includes:

[0027] Traverse all cells in the computational grid. For the centroid point of each cell, use the grid size field to calculate the corresponding expected size ;

[0028] If is less than , then subdivide the corresponding cell; if is greater than , then coarsen and merge the corresponding cell;

[0029] The method for subdividing includes:

[0030] Based on the expected size , calculate the minimum side length required for the cell , where is a constant determined by optimization; check all side lengths of the cell. If there is a side length greater than , then equally divide the side corresponding to the side length to generate new nodes; use the new nodes to re-mesh the original cell to generate sub-cells; recursively subdivide the generated sub-cells until the side lengths of all sub-cells are not greater than ;

[0031] The method for coarsening and merging includes:

[0032] Check the cells in the neighborhood of all faces of the cell. If the area difference between any two faces of two cells is less than a preset difference threshold, merge them; obtain the merged new cell;

[0033] Calculate the centroid size of the merged new cell , where is the area of any face of the cell before merging, is the volume of the cell before merging; if is greater than , then continue to merge with other cells in the neighborhood, otherwise stop;

[0034] Define a smoothing objective function. After coarsening and merging, smooth all cells by minimizing the smoothing objective function;

[0035] Smoothing objective function ; where is the th side length, is the target side length, is the The included angle between both sides is the ideal included angle and is the weight coefficient;

[0036] By optimizing the algorithm to solve the minimization of the smoothing objective function, the model processing grid is obtained.

[0037] Furthermore, the construction method of the hydrodynamic model includes:

[0038] Based on the model processing grid, a hydrodynamic control function is defined, and the expression of the hydrodynamic control function is: ; where is the mass fraction of the th substance, is the diffusion coefficient of the th substance, is the migration coefficient of the th substance, is the coupling weight factor applied to the th substance, is the source term of the th substance, is the energy function;

[0039] The energy function ; where is the specific heat capacity at constant pressure, is the total derivative of the mass fraction of the th substance, is the thermal conductivity;

[0040] Based on the model processing grid, the hydrodynamic control function is integrated to obtain the hydrodynamic model.

[0041] Furthermore, the construction method of the diffusion model includes:

[0042] Based on the model processing grid, a diffusion control function is defined, and the expression of the diffusion control function is: ; where is the concentration of the th element, is the diffusion coefficient tensor of the th element, is the source term of the th element;

[0043] ; where is the frequency factor of the th element, is the diffusion activation energy of the th element, is the universal gas constant is a concentration-dependent function, is the concentration of elements other than the th element, is a stress-dependent function, is the stress tensor; , where, is the activation volume, is the Boltzmann constant;

[0044] Concentration-dependent function , where, and are fitting parameters, is the reference concentration;

[0045] Discretize the diffusion control function to obtain a discretized diffusion function; integrate the discretized diffusion function on the basis of the model processing grid to obtain a preliminary hydrodynamic model;

[0046] Set diffusion boundary conditions on the basis of the preliminary hydrodynamic model to obtain a diffusion model; the diffusion boundary conditions include the first type of boundary condition, the second type of boundary condition, and the third type of boundary condition;

[0047] The first type of boundary condition is the concentration value on the specified boundary, and the second type of boundary condition is the concentration gradient or diffusion flux on the specified boundary; the third type of boundary condition is a linear combination of the concentration value and diffusion flux on the boundary.

[0048] Furthermore, the calculation method of the sensitivity coefficient matrix includes:

[0049] Define the output indexes of key production parameters as product performance indexes, production efficiency indexes, energy consumption indexes, and raw material utilization indexes;

[0050] Use the data points in the time series data as sample points, and define a sensitivity objective function; the expression of the sensitivity function is: ; where, is the vector composed of the key production parameters corresponding to the sample point , is a perturbation added to the sample point ; is the objective function; is the small perturbation function, is the difference function; is the sensitivity coefficient corresponding to the sample point ;

[0051] Small perturbation function ;

[0052] Difference function ; where, is the total number of sample points, is the sample point corresponding to the coefficient of the wavelet basis function, is the sample point corresponding to the perturbation point, is for the sample point , the constructed random perturbation sequence;

[0053] The objective function is the weighted sum of the performance index function, production efficiency function, energy consumption function, and raw material utilization rate function; the performance index function is defined as the weighted sum of remanence, coercivity, and energy product; the production efficiency function is defined as the product of production capacity and yield; the energy consumption function is the sum of the energy consumption of melting, molding, and sintering; the raw material utilization rate function is defined as the ratio of actual output to theoretical output;

[0054] Initialize a symmetric matrix, and use the sample point and the corresponding sensitivity coefficient as the elements of the symmetric matrix, that is, complete the calculation of the sensitivity coefficient matrix;

[0055] The multi-objective optimization function ; where is the vector composed of the control parameters to be optimized; is the process constraint matrix, is the expected product performance target vector, is the sensitivity coefficient matrix, is the regularization parameter.

[0056] Furthermore, the method for obtaining the optimal control parameter combination includes:

[0057] Initialize the particle swarm, define the number of particles , the maximum number of iterations , randomly generate the positions and velocities of particles, is the index of the particle; initialize the individual optimal position of each particle and the global optimal position

[0058] Use the multi-objective optimization function as the fitness function, and define the velocity update formula. The update formula is: ; where is the inertia weight, and are the acceleration factors; and are random numbers between is the time The particle position randomly selected at is the time when the particle speed, is the time when the particle speed, is the time when the particle individual optimal position, is the time when the particle global optimal position, is the time when the particle position;

[0059] Define the position update formula: ; where is the time when the particle position;

[0060] Define the search space. If exceeds the predefined search space, then pull it back to the boundary of the search space, calculate the fitness function value of the new position of each particle, compare and update the individual optimal position and global optimal position of each particle;

[0061] If the preset maximum number of iterations is reached, then end; otherwise continue to iterate, output the optimal solution, and the global optimal position corresponding to the optimal solution is the optimal control parameter combination.

[0062] Technical effects and advantages of the low-latency control system for rare earth permanent magnet material production based on digital twin technology in the present invention:

[0063] By constructing an accurate digital twin model, the present invention realizes the comprehensive perception and in-depth insight into the complex production process, greatly improving the accuracy and real-time performance of production control; using multi-source heterogeneous data and simulation, it can accurately predict the change trend of key production parameters, providing a reliable basis for timely adjusting control strategies; the introduced sensitivity analysis method effectively identifies the most critical control parameters and optimizes the control focus; the multi-objective optimization function and the improved particle swarm algorithm achieve the comprehensive balance of product performance, production efficiency, energy consumption, etc., improving production efficiency and product quality; the low-latency architecture ensures the rapid response of the system to production fluctuations, effectively reducing the generation of defective products; in addition, it also reduces energy consumption and raw material losses, improves resource utilization rate, and promotes green production. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1Schematic diagram of a low-latency control system for rare earth permanent magnet material production based on digital twin technology of the present invention;

[0065] Figure 2 Schematic diagram of a low-latency control method for rare earth permanent magnet material production based on digital twin technology of the present invention. Detailed implementation manners

[0066] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Embodiment

[0067] Please refer to Figure 1 As shown, the low-latency control system for rare earth permanent magnet material production based on digital twin technology in this embodiment includes:

[0068] A data acquisition module for acquiring production sensing data of the rare earth permanent magnet material production line;

[0069] A model construction module that uses the production sensing data to construct a digital twin model of the rare earth permanent magnet material production process;

[0070] A simulation module that uses the digital twin model for simulation to obtain the change trends of key production parameters;

[0071] A matrix fitting module for calculating the sensitivity coefficient matrix of each key production parameter according to the change trends of the key production parameters;

[0072] A parameter optimization module constructs a multi-objective optimization function based on the sensitivity coefficient matrix, uses an improved particle swarm algorithm to solve the multi-objective optimization function, and obtains an optimal control parameter combination; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0073] Furthermore, the production sensing data includes temperature data, pressure data, flow data, electrical parameters, material property data, fluid-related data, and diffusion-related data;

[0074] The temperature data includes the melting furnace temperature, sintering furnace temperature, mold temperature, and cooling temperature; the pressure data includes the press pressure, air pressure, and hydraulic pressure; the flow rate data includes the raw material input flow rate, coolant flow rate, and protective gas flow rate; the electrical parameters include the motor current, heating power, and control system voltage; the material property data includes the material density, specific heat capacity of the material, and thermal conductivity of the material; the fluid-related data includes the fluid viscosity and fluid density; the diffusion-related data includes the diffusion coefficient of the element and the concentration distribution.

[0075] It should be noted that the temperature data, pressure data, flow rate data, and electrical parameters are obtained using corresponding sensors; the material property data obtains standard values by referring to material manuals or databases. For standard fluids, the fluid-related data is obtained by referring to fluid property manuals. For specific fluids, instruments such as viscometers and densitometers are used for on-site measurement; the diffusion coefficient is obtained by referring to literature data, and the concentration distribution is obtained by sampling analysis or online monitoring equipment.

[0076] Furthermore, the construction method of the digital twin model includes:

[0077] Perform preprocessing on the production sensing data, including cleaning, denoising, imputation, and standardization, to ensure the integrity and quality of the data.

[0078] Convert the production sensing data into a unified format, such as CSV, Parquet, etc.; unify the timestamps of all data to the same frequency and starting point; perform dimensionless processing on data with different dimensions, such as Z-Score standardization; perform one-hot encoding or sequential encoding on categorical data (such as equipment names); obtain improved sensing data.

[0079] Based on the improved production data, construct a heat conduction model, a fluid dynamics model, and a diffusion model; establish interfaces between the heat conduction model, the fluid dynamics model, and the diffusion model to ensure that data can be transferred between different models, and couple the heat conduction model, the fluid dynamics model, and the diffusion model to obtain the digital twin model.

[0080] The construction method of the heat conduction model includes:

[0081] Define the control function and the computational domain of the control function. The computational domain includes the melting furnace, the pressing mold, the sintering furnace, and the cooling device;

[0082] The control function includes the heat conduction equation, boundary conditions, and initial conditions; the expression of the heat conduction equation is: , where is the density, is the specific heat capacity, is the temperature, is the time, is the thermal conductivity, is an internal heat source (the heat generated within a material or system during heat conduction), is the gradient symbol.

[0083] Boundary conditions include known temperature conditions (directly specifying the temperature value on the boundary), known heat flux conditions (specifying the heat flux density on the boundary), and a known linear combination of convection and radiation (heat exchange between the boundary and the surrounding environment, including convection and radiation); the initial condition is the given initial temperature distribution.

[0084] The computational domain is discretized to generate a computational grid; the computational grid includes structured grids and unstructured grids;

[0085] The ways of performing the discretization process include:

[0086] According to the geometric shape of the computational domain, the computational domain is divided into structured grids and unstructured grids;

[0087] If the geometric shape of the computational domain is simple and regular (such as a cuboid, cylinder, etc.), the computational domain is divided into N5 fan-shaped (2D) or annular (2D) cells in polar coordinates, that is, a structured grid is formed;

[0088] If the geometric shape of the computational domain is a complex geometric shape (such as having concave corners, multi-connected domains, etc.), the computational domain is divided into N6 irregular triangular (2D) or tetrahedral (3D) cells, that is, an unstructured grid is formed.

[0089] The structured grids and unstructured grids are encrypted, and the ways of encryption include:

[0090] Calculate the curvature distribution of the boundary of the computational domain, and the calculation methods include analytical methods, geometric fitting methods, discrete differential methods, etc.; the boundary of the computational domain is the outer edge surrounding the entire computational region.

[0091] Define a curvature-sensitive weight function , which is used to quantify the influence of curvature on the grid size; the ways of defining the curvature-sensitive weight function include:

[0092] Based on the curvature distribution, obtain the curvature value of any node in the structured grid and unstructured grid , if the curvature value of any node is greater than 0, then the curvature-sensitive weight function ; where, is the reference curvature, is the critical curvature value; the reference curvature is a benchmark curvature value for normalization calculation, and can take the average curvature or the maximum curvature in the computational domain; the critical curvature value is a preset threshold for judging whether the curvature is large enough to require special treatment.

[0093] If the curvature value of any node is less than or equal to 0, the curvature-sensitive weight function ; where and are both adjustable weight coefficients.

[0094] Based on the curvature-sensitive weight function, a mesh size field is constructed; where is the desired reference mesh size (determine the desired average element size for the entire computational domain according to the calculation accuracy requirements and hardware resource limitations), and are the horizontal, vertical, and vertical axes of the three-dimensional coordinate system where the mesh size field is located, respectively.

[0095] Using the mesh size field to further divide the structured mesh and the unstructured mesh; obtaining the model processing mesh.

[0096] The ways of further division include:

[0097] Traverse all the elements in the computational mesh. For the centroid point of each element, use the mesh size field to calculate the corresponding desired size ;

[0098] If is less than , then the corresponding element is encrypted and subdivided; if is greater than then the corresponding element is coarsened and merged.

[0099] The ways of encrypting and subdividing include:

[0100] Based on the desired size , calculate the minimum side length required for the element, where is a constant determined by optimization; check all the side lengths of the element. If there is a side length greater than, then divide the side corresponding to the side length equally to generate new nodes; use the new nodes to re-mesh the original element to generate sub-elements; specifically, generate new nodes on the long side of the original element to divide the original element into multiple sub-elements. The sub-elements are triangles (2D) or tetrahedrons (3D), and check the quality (such as shape, size, etc.) of the newly generated sub-elements; if the quality does not meet the requirements, further optimization is carried out, such as adjusting the node positions or re-meshing; The value of depends on the shape of the original element and the number of newly added nodes, usually 2-4 sub-elements.

[0101] Recursively encrypt and subdivide the generated sub-units until the side lengths of all sub-units are no greater than ;

[0102] The way of coarsening and merging includes:

[0103] Check the cells within the neighborhood of all faces of the cell (obtained by defining the neighborhood radius). If there are two cells whose area difference of any face is less than the preset difference threshold, merge them; obtain the new merged cell.

[0104] Calculate the centroid size of the new merged cell , where is the area of any face of the cell before merging, is the volume of the cell before merging; if is greater than , then continue to merge with other cells in the neighborhood (the merging method is the same as above), otherwise stop.

[0105] Define a smoothing objective function. After coarsening and merging, smooth all cells by minimizing the smoothing objective function.

[0106] Smoothing objective function ; where is the length of the th side, is the target side length, is the , the included angle between the two sides, is the ideal included angle,

[0107] Solve the minimization of the smoothing objective function through an optimization algorithm (such as the Gauss-Newton method) to make the grid side lengths and cell shapes reach the optimal, and obtain the model processing grid;

[0108] Integrate a control function on the basis of the model processing grid to obtain a heat conduction model.

[0109] It should be noted that integrating the control function means discretizing the continuous control function on each cell; applying the control function to each cell to form a series of algebraic equations; the specific steps are: applying the discretized heat conduction equation to each cell; applying the corresponding boundary conditions to the cells on the boundary of the computational domain; assembling the equations of all cells to form a global equation to obtain the heat conduction model.

[0110] Furthermore, the construction method of the hydrodynamic model includes:

[0111] Based on the model processing grid, a hydrodynamic control function is defined, and the expression of the hydrodynamic control function is: ; where is the specific heat capacity at constant pressure, is the total derivative of the mass fraction of the th substance, is the thermal conductivity;

[0112] Integrate the hydrodynamic control function on the basis of the model processing grid to obtain a hydrodynamic model.

[0113] Furthermore, the construction method of the diffusion model includes:

[0114] Based on the model processing grid, a diffusion control function is defined, and the expression of the diffusion control function is: ; where is the concentration of the th element, is the diffusion coefficient tensor of the th element, is the source term of the th element; ; where is the frequency factor of the th element (which reflects the attempt frequency of atomic jumps, is related to the lattice vibration frequency, and is obtained through experimental determination or theoretical calculation), is the diffusion activation energy of the th element (representing the energy barrier that atoms need to cross to occur diffusion), is the universal gas constant (8.314 J / (mol·K)), is the concentration-dependent function, is the concentration of other elements except the th element, is the stress-dependent function, is the stress tensor.

[0115] , where is the activation volume, is the Boltzmann constant;

[0116] Concentration-dependent function , where and are fitting parameters, is the reference concentration.

[0117] Use the finite difference method, finite element method or spectral method to discretize the diffusion control function to obtain a discretized diffusion function; integrate the discretized diffusion function on the basis of the model processing grid to obtain a preliminary hydrodynamic model.

[0118] Set diffusion boundary conditions based on the preliminary hydrodynamic model. The diffusion boundary conditions include the first kind of boundary condition, the second kind of boundary condition, and the third kind of boundary condition. The way to set the diffusion boundary conditions is also to apply the corresponding diffusion boundary conditions to the cells on the boundary of the computational domain, and obtain the diffusion model.

[0119] The first kind of boundary condition is the concentration value on the specified boundary, and its application scenarios are surface rapid equilibrium and constant source, such as the contact between the material surface and the gas phase or liquid phase with a fixed concentration; constant source, such as continuous surface penetration or plating process.

[0120] The second kind of boundary condition is the concentration gradient or diffusion flux on the specified boundary, and its application scenarios are insulating boundary and controlled surface reaction, such as surface oxidation or corrosion process.

[0121] The third kind of boundary condition is the linear combination of the concentration value and diffusion flux on the boundary, and its application scenario is the convective boundary condition. For example, it describes the interaction between the material surface and the flowing environment, and interfacial mass transfer. For example, the mass transfer process at the gas-solid interface or liquid-solid interface.

[0122] Furthermore, the obtaining method of the change trend of the key production parameters includes:

[0123] Define the key production parameters including temperature distribution, phase change rate, and element distribution; initialize the parameters of the digital twin model, set the time step and total simulation duration of the simulation; for each time step, input the actual production data at the current moment (with the same data type as the production sensing data), perform one-step simulation calculation, and store the simulation results (key production parameters); and use the ensemble Kalman filter to achieve data assimilation, regularly (such as every 10 time steps) incorporate the actual observed data into the model and adjust the parameters. After the total simulation duration ends, extract the time series data composed of the key production parameters from the stored simulation results, which is the change trend of the key production parameters.

[0124] Furthermore, the calculation method of the sensitivity coefficient matrix includes:

[0125] Define the output indicators of the key production parameters as product performance indicators, production efficiency indicators, energy consumption indicators, and raw material utilization indicators.

[0126] Take the data points on the time series data as sample points, and define the sensitivity objective function. The expression of the sensitivity function is: ; where is the vector composed of the key production parameters corresponding to the sample point , is a sufficiently small perturbation added to the sample point ; is the objective function; is the small perturbation function, is the difference function; is the sample point and the corresponding sensitivity coefficient.

[0127] Small perturbation function ;

[0128] Difference function ; where, is the total number of sample points, is the sample point and the coefficient of the corresponding wavelet basis function, is the sample point and the corresponding perturbation point, which satisfies a certain specific wavelet basis function, such as the Daubechies wavelet basis; is the randomly perturbed sequence constructed for the sample point , which follows a certain probability distribution, such as a uniform distribution or a Gaussian distribution.

[0129] The objective function is the weighted sum of the performance index function, production efficiency function, energy consumption function, and raw material utilization function; the performance index function is defined as the weighted sum of the remanence, coercivity, and energy product; the production efficiency function is defined as the product of production capacity and good product rate; the energy consumption function is the sum of the energy consumption of melting, molding, and sintering; the raw material utilization function is defined as the ratio of actual output to theoretical output.

[0130] Initialize a symmetric matrix, using the sample point and the corresponding sensitivity coefficient as the elements of the symmetric matrix, that is, the calculation of the sensitivity coefficient matrix is completed.

[0131] Furthermore, the multi-objective optimization function ; where, is the vector composed of the control parameters to be optimized; is the process constraint matrix, is the expected product performance target vector, is the sensitivity coefficient matrix, is the regularization parameter.

[0132] It should be noted that the control parameters to be optimized include melting temperature, melting time, molding pressure, holding pressure time, sintering temperature, sintering time, heat treatment temperature, heat treatment time, and cooling rate.

[0133] ​​​The process constraint matrix represents the relationship between control parameters and process variables. A mathematical model is established based on physical and chemical principles to obtain the relationship equation between control parameters and process variables. Using historical production data, machine learning algorithms (such as multiple linear regression, neural networks, etc.) are used to establish the mapping relationship between control parameters and process variables. Through orthogonal experiments or response surface methods, the influence coefficients of control parameters on process variables are obtained, that is, the process constraint matrix is constructed.

[0134] The expected product performance target vector includes the product performance indicators that are hoped to be achieved. Determining the key performance indicators includes remanence, coercivity, and maximum energy product. And target values are set for each product performance indicator, the different product performance indicators are normalized, and according to the importance of each product performance indicator, weights are assigned respectively, and the expected product performance target vector is calculated by weighted calculation.

[0135] Furthermore, the obtaining method of the optimal control parameter combination includes:

[0136] Initialize the particle swarm and define the number of particles , the maximum number of iterations , randomly generate the positions and velocities of particles, where is the index of the particle; initialize the individual optimal position of each particle as well as the global optimal position

[0137] The velocity of the particle represents the direction and magnitude of the particle's movement in the search space, which determines how the particle updates its position; for each particle, the individual optimal position is the position of the best solution found by the corresponding particle so far; the global optimal position is the position of the best solution found by all particles in the entire particle swarm; that is, the individual optimal position is the best production result generated by each specific parameter combination in history; the global optimal position is the best production result generated among all tried parameter combinations.

[0138] Take the multi-objective optimization function as the fitness function and define the update formula of the velocity. The update formula is: ; where is the inertia weight, and are acceleration factors; and are a random number between is the time and the randomly selected particle position at that time; is the time and the velocity of the particle at that time, is the time and the velocity of the particle at that time, is the time and the individual best position of the particle at that time, is the time and the global best position of the particle at that time, is the time and the position of the particle at that time.

[0139] Define the position update formula: ; where is the time and the position of the particle at that time;

[0140] Define the search space. If exceeds the preset search space, it is pulled back to the boundary of the search space, and the fitness function value of the new position of each particle is calculated, and the individual best position and the global best position of each particle are compared and updated.

[0141] If the preset maximum number of iterations is reached, end; otherwise, continue to iterate, output the optimal solution, and the global best position corresponding to the optimal solution is the optimal control parameter combination.

[0142] In this embodiment, by constructing an accurate digital twin model, a comprehensive perception and in-depth insight into the complex production process are realized, and the accuracy and real-time performance of production control are greatly improved; by using multi-source heterogeneous data and simulation, the change trend of key production parameters can be accurately predicted, providing a reliable basis for timely adjusting control strategies; the introduced sensitivity analysis method effectively identifies the most critical control parameters and optimizes the control focus; the multi-objective optimization function and the improved particle swarm algorithm achieve a comprehensive balance in multiple aspects such as product performance, production efficiency, and energy consumption, significantly improving production efficiency and product quality; the low-latency architecture ensures that the system quickly responds to production fluctuations and effectively reduces the generation of defective products; in addition, it also reduces energy consumption and raw material losses, improves resource utilization rate, and promotes green production. Embodiment

[0143] Please refer to Figure 2As shown in the figure, for the parts not described in detail in this embodiment, refer to the description in Embodiment 1. A low-latency control method for rare earth permanent magnet material production based on digital twin technology is provided, including:

[0144] S1. Obtain the production sensing data of the rare earth permanent magnet material production line;

[0145] S2. Use the production sensing data to construct a digital twin model of the rare earth permanent magnet material production process;

[0146] S3. Use the digital twin model for simulation to obtain the change trends of key production parameters;

[0147] S4. According to the change trends of key production parameters, calculate the sensitivity coefficient matrix of each key production parameter;

[0148] S5. Construct a multi-objective optimization function based on the sensitivity coefficient matrix, and use the improved particle swarm algorithm to solve the multi-objective optimization function to obtain the optimal control parameter combination. Embodiment

[0149] This embodiment publicly provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it realizes the operation mode of the above-provided low-latency control method for rare earth permanent magnet material production based on digital twin technology.

[0150] Since the electronic device introduced in this embodiment is the electronic device used to implement the low-latency control method for rare earth permanent magnet material production based on digital twin technology in the embodiments of the present application, based on the low-latency control method for rare earth permanent magnet material production based on digital twin technology introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners and various change forms of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device realizes the method in the embodiments of the present application will not be introduced in detail here. As long as those skilled in the art implement the electronic device used for the low-latency control method for rare earth permanent magnet material production based on digital twin technology in the embodiments of the present application, it falls within the scope of protection of the present application.

[0151] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain a formula that is closest to the real situation. The preset parameters and threshold selections in the formulas are set by those skilled in the art according to the actual situation.

[0152] The above are only the preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A low-latency control system for the production of rare earth permanent magnet materials based on digital twin technology, characterized in that, Including: A data acquisition module for obtaining production sensing data of a rare earth permanent magnet material production line; A model construction module for constructing a digital twin model of the production process of rare earth permanent magnet materials by using the production sensing data; A simulation module for performing simulation by using the digital twin model to obtain the change trends of key production parameters; A matrix fitting module for calculating the sensitivity coefficient matrix of each key production parameter according to the change trends of the key production parameters; A parameter optimization module for constructing a multi-objective optimization function based on the sensitivity coefficient matrix, solving the multi-objective optimization function by using an improved particle swarm algorithm, and obtaining an optimal control parameter combination; Each module is connected by wired and / or wireless means; The calculation method of the sensitivity coefficient matrix includes: Defining the output indexes of the key production parameters as product performance indexes, production efficiency indexes, energy consumption indexes, and raw material utilization rate indexes; Take the data points on the time series data as sample points and define a sensitivity objective function; the expression of the sensitivity function is: ; where is the vector composed of the key production parameters corresponding to the sample point , is a perturbation amount added to the sample point ; is the objective function; is the small perturbation function, is the difference function; is the sensitivity coefficient corresponding to the sample point ; Small perturbation function ; Difference function ; where is the total number of sample points, is the coefficient of the wavelet basis function corresponding to the sample point , is the perturbation point corresponding to the sample point , is the constructed random perturbation sequence for the sample point . The objective function is the weighted sum of a performance index function, a production efficiency function, an energy consumption function, and a raw material utilization rate function; the performance index function is defined as the weighted sum of remanence, coercivity, and energy product; the production efficiency function is defined as the product of production capacity and good product rate; the energy consumption function is the sum of the energy consumption of melting, molding, and sintering; the raw material utilization rate function is defined as the ratio of actual output to theoretical output; Initialize a symmetric matrix with the sample points and the corresponding sensitivity coefficients as the elements of the symmetric matrix, thus completing the calculation of the sensitivity coefficient matrix; The multi-objective optimization function ; where is a vector composed of control parameters to be optimized; is a process constraint matrix, is an expected product performance target vector, is a sensitivity coefficient matrix, is a regularization parameter.

2. The low-latency control system for the production of rare earth permanent magnet materials based on digital twin technology according to claim 1, wherein The production sensing data includes temperature data, pressure data, flow data, electrical parameters, material property data, fluid-related data, and diffusion-related data; The temperature data includes melting furnace temperature, sintering furnace temperature, mold temperature, and cooling temperature; the pressure data includes molding machine pressure, air pressure, and hydraulic pressure; the flow data includes raw material input flow, coolant flow, and protective gas flow; The electrical parameters include motor current, heating power, and control system voltage; the material property data includes material density, material specific heat capacity, and material thermal conductivity; the fluid-related data includes fluid viscosity and fluid density; the diffusion-related data includes the diffusion coefficient and concentration distribution of elements.

3. The low-latency control system for the production of rare earth permanent magnet materials based on digital twin technology according to claim 2, wherein The construction method of the digital twin model includes: Performing preprocessing such as cleaning, denoising, imputation, and standardization on the production sensing data, and converting the preprocessed production sensing data into a unified format; unifying the timestamps of all data to the same frequency and starting point; and performing dimensionless processing on data with different dimensions; performing one-hot encoding or ordinal encoding on categorical data; obtaining perfect sensing data; Constructing a heat conduction model, a fluid dynamics model, and a diffusion model based on the perfect production data; establishing interfaces between the heat conduction model, the fluid dynamics model, and the diffusion model pairwise, and coupling the heat conduction model, the fluid dynamics model, and the diffusion model to obtain a digital twin model.

4. The low-latency control system for rare earth permanent magnet material production based on digital twin technology according to claim 3, characterized in that, The construction method of the heat conduction model includes: Defining a control function and the calculation domain of the control function, and the calculation domain includes a melting furnace, a molding die, a sintering furnace, and a cooling device; The control function includes the heat conduction equation, boundary conditions, and initial conditions; the expression of the heat conduction equation is: , where is the density, is the specific heat capacity, is the temperature, is the time, is the heat conduction coefficient, is the internal heat source, is the gradient symbol; The boundary conditions include known temperature conditions, known heat flux conditions, and a linear combination of known convection and radiation; the initial condition is a given initial temperature distribution; Discretize the computational domain to generate a computational grid; the computational grid includes structured grids and unstructured grids; encrypt the structured grids and unstructured grids to obtain a model processing grid; integrate control functions on the basis of the model processing grid to obtain a heat conduction model.

5. The low-latency control system for rare earth permanent magnet material production based on digital twin technology according to claim 4, characterized in that, The encryption methods include: Calculate the curvature distribution of the boundary of the computational domain and define a curvature-sensitive weight function ; The ways to define the curvature-sensitive weight function include: Obtain the curvature values of any node in structured grids and unstructured grids based on the curvature distribution , if the curvature value of any node is greater than 0, then the curvature-sensitive weight function ; where is the reference curvature, is the critical curvature value; If the curvature value of any node is less than or equal to 0, the curvature-sensitive weight function ; where and are both adjustable weight coefficients; Construct a mesh size field according to the curvature-sensitive weight function ; where is the desired reference mesh size, and are the horizontal, vertical, and vertical axes of the three-dimensional coordinate system where the mesh size field is located, respectively; Using the grid size field Further divide the structured grid and the unstructured grid to obtain the model processing grid.

6. The low-latency control system for rare earth permanent magnet material production based on digital twin technology according to claim 5, wherein The further division methods include: Traverse all the cells in the computational grid, and for the centroid point of each cell, use the grid size field to calculate the corresponding expected size ; If less than , the corresponding unit is encrypted and subdivided; if greater than , the corresponding unit is coarsened and merged. The encryption and subdivision methods include: Based on the expected size , calculate the minimum side length required for the computing unit , where is a constant determined by optimization; check all side lengths of the unit. If there is a side length greater than , then equally divide the side corresponding to the side length to generate new nodes; use the new nodes to re-mesh the original unit to generate M1 sub-units; recursively perform encryption and subdivision on the generated sub-units until the side lengths of all sub-units are not greater than ; The coarsening and merging methods include: Check the cells in the neighborhood of all faces of the cell. If the area difference between any two faces of the two cells is less than the preset difference threshold, merge them; obtain the new merged cell; Calculate the centroid size of the new combined cell , where is the area of any face of the cell before combination, is the volume of the cell before combination; if is greater than , then continue to combine with other cells in the neighborhood, otherwise stop; Define a smoothing objective function. After coarsening and merging, smooth all cells by minimizing the smoothing objective function; Smoothing objective function ; where is the length of the th side, is the target side length, is the angle between the th two sides, is the ideal angle, is the weight coefficient; Solve the minimized smoothing objective function through an optimization algorithm to obtain a model processing grid.

7. The low-latency control system for rare earth permanent magnet material production based on digital twin technology according to claim 6, wherein The construction method of the hydrodynamic model includes: Based on the model processing grid, a hydrodynamic control function is defined, and the expression of the hydrodynamic control function is: ; where is the mass fraction of the th substance, is the diffusion coefficient of the th substance, is the migration coefficient of the th substance, is the coupling weight factor applied to the th substance, is the source term of the th substance, is the energy function; Energy function ; where is the specific heat capacity at constant pressure, is the total derivative of the mass fraction of the th substance, thermal conductivity; Integrate hydrodynamic control functions on the basis of the model processing grid to obtain a hydrodynamic model.

8. The low-latency control system for rare earth permanent magnet material production based on digital twin technology according to claim 7, wherein The construction method of the diffusion model includes: Based on the model processing grid, a diffusion control function is defined, and the expression of the diffusion control function is: ; where is the concentration of the -th element, is the diffusion coefficient tensor of the -th element, is the source term of the -th element; , Among them, is the frequency factor of the th element, is the diffusion activation energy of the th element, is the universal gas constant, is the concentration-dependent function, is the concentration of elements other than the th element, is the stress-dependent function, is the stress tensor; , where is the activation volume, is the Boltzmann constant; Concentration-dependent function , where and are fitting parameters, is the reference concentration; Discretize the diffusion control function to obtain a discretized diffusion function; integrate the discretized diffusion function on the basis of the model processing grid to obtain a preliminary hydrodynamic model; Set diffusion boundary conditions on the basis of the preliminary hydrodynamic model to obtain a diffusion model; the diffusion boundary conditions include the first kind of boundary condition, the second kind of boundary condition, and the third kind of boundary condition; The first kind of boundary condition is the concentration value on the specified boundary, the second kind of boundary condition is the concentration gradient or diffusion flux on the specified boundary; the third kind of boundary condition is a linear combination of the concentration value and diffusion flux on the boundary.

9. The low-latency control system for the production of rare earth permanent magnet materials based on digital twin technology according to claim 8, wherein The method for obtaining the optimal control parameter combination includes: Initialize the particle swarm, define the number of particles N3, the maximum number of iterations T3, and randomly generate the positions of N3 particles and velocities , where i is the index of the particle; initialize the individual best position of each particle as well as the global best position ; Take the multi-objective optimization function as the fitness function and define the update formula for velocity. The update formula is as follows: ; where is the inertia weight, and are acceleration factors; and are random numbers between [0, 1], is time is the particle position randomly selected at time is time at time the velocity of particle is time at time the velocity of particle is time at time the individual best position of particle is time at time the global best position of particle is time at time the position of particle Define the position update formula: ; where is the time when the particle is at the position; Define the search space. If it exceeds the predefined search space, pull it back to the boundary of the search space, calculate the value of the fitness function for the new position of each particle, compare and update the individual best position and the global best position of each particle; If the preset maximum number of iterations is reached, end; otherwise, continue the iteration and output the optimal solution. The global optimal position corresponding to the optimal solution is the optimal control parameter combination.

Images