Digital twinning method and system based on hotpot condiment grinding process

Abstract

The invention discloses a digital twinning method and system based on a hotpot condiment grinding process, and relates to the technical field of intelligent manufacturing, and the method comprises the steps: collecting the operation parameters of the grinding process and a discharge port condiment image in real time, constructing a geometric model of grinding equipment, and integrating a multi-physics field mechanism model; the operation parameters are input into a multi-physical field mechanism model, digital twin model simulation is conducted on the interior of the grinding equipment, and a simulation predicted value is output; a visual identification model based on deep learning extracts a quality characteristic value from the backing material image, and the quality characteristic value is compared with a simulation prediction value output by the digital twin model; and correcting physical boundary parameters in the digital twinborn model in real time according to the comparison deviation, and generating an optimal regulation and control instruction. According to the method, the problem that the prediction precision is attenuated along with time due to neglect of equipment abrasion and material fluctuation of an existing grinding process simulation model is solved, and a model self-calibration mechanism based on product quality actual measurement feedback is perfected.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a digital twin method and system based on the grinding process of hot pot base. Background Technology

[0002] In the preparation process of hot pot base, the grinding process is a crucial step that determines the smoothness of the base's texture, the state of oil emulsification, and the efficiency of flavor substance release. This process involves an extremely complex multi-physics coupling mechanism of heat, fluid, and solid. Solid spices (such as chili peppers and Sichuan peppercorns) undergo brittle fracture and fatigue breakage under high-speed shearing, while the carrier oil undergoes dynamic evolution of non-Newtonian fluid rheological properties under frictional heating. Currently, theoretical research on the grinding process mainly relies on single-physics simulations using the discrete element method or computational fluid dynamics.

[0003] Existing digital twin simulation technology for grinding processes has limitations in practical industrial applications. Current digital twin models lack real-time fidelity and adaptability. They are often limited to the acquisition of macroscopic scalar quantities such as temperature and current, lacking the ability to perceive the microscopic particle size distribution and morphological characteristics of the output material online. As a result, the control system cannot obtain feedback signals that directly characterize the grinding quality. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a digital twin method based on the grinding process of hot pot base to solve the problem that existing digital twin systems are unable to maintain high-precision process prediction under dynamically changing working conditions and cannot truly achieve intelligent closed-loop control of the grinding process.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a digital twin method based on the grinding process of hot pot base, which includes: real-time acquisition of the operating parameters of the grinding process and the image of the base material at the outlet, construction of a geometric model of the grinding equipment, and integration of a multi-physics mechanism model based on the coupling of discrete element method and computational fluid dynamics;

[0008] The operating parameters are input into the multiphysics mechanism model to perform digital twin model simulation of the inside of the grinding equipment, and the simulation prediction values ​​are output.

[0009] A deep learning-based visual recognition model extracts quality feature values ​​from the base material image and compares them with the simulation prediction values ​​output by the digital twin model.

[0010] The physical boundary parameters in the digital twin model are corrected in real time based on the comparison deviation. Based on the corrected digital twin model, the process trend is predicted and the optimal control command is generated.

[0011] As a preferred embodiment of the digital twin method based on the grinding process of hot pot base material described in this invention, the real-time acquisition of the operating parameters of the grinding process and the image of the base material at the outlet includes: installing temperature sensors and pressure sensors on the outer wall of the grinding chamber and at the inlet and outlet; installing current and torque sensors on the main shaft drive motor to calculate the grinding resistance; and installing a triaxial vibration sensor on the equipment base to monitor the contact state of the grinding disc.

[0012] An industrial high-frequency camera with self-cleaning function is installed at the discharge port to acquire macroscopic images of the bottom material at the discharge port.

[0013] As a preferred embodiment of the digital twin method based on the hot pot base grinding process described in this invention, the integration of the multiphysics mechanism model based on the coupling of discrete element method and computational fluid dynamics includes establishing a geometric model corresponding to the physical grinding equipment in virtual space. The geometric model includes a grinding chamber, a rotor grinding disc, a discharge port, and a material flow channel structure. The geometric boundary constraints of the multiphysics simulation are obtained based on the geometric model.

[0014] Under the geometric boundary constraints, a particle contact and crushing mechanism model based on the discrete element method (DEM) and a flow field and thermal field mechanism model based on computational fluid dynamics (CFD) are constructed and integrated. The particle contact and crushing mechanism model and the flow field and thermal field mechanism model are coupled to obtain a multi-physics mechanism model characterizing the temperature field distribution, shear force distribution and material crushing evolution process inside the grinding chamber.

[0015] As a preferred embodiment of the digital twin method based on the grinding process of hot pot base described in this invention, the digital twin model simulation of the inside of the grinding equipment includes: collecting the operating parameters of the grinding equipment as the real-time driving quantity of the digital twin model, and synchronizing the physical grinding process with the virtual simulation process.

[0016] The operating parameters include grinding speed, current load, feed flow rate, grinding chamber pressure, grinding chamber temperature, equipment vibration, and acoustic characteristics.

[0017] The multi-source operating parameters are time-stamp aligned and noise-reducing filtered, and an operating parameter sequence is formed based on the geometric model of the equipment and the mapping relationship between operating conditions.

[0018] The operating parameters are converted into boundary and initial conditions for a multiphysics mechanism model, including:

[0019] The grinding speed is mapped to the rotor's rotational boundary, the feed flow rate and the initial particle size distribution of the material are mapped to the DEM particle injection conditions, the current load is mapped to the contact friction power and shear energy terms, and the cooling medium flow rate and the cavity temperature are mapped to the convective heat transfer boundary on the CFD side.

[0020] The sequence of operating parameters is used as the real-time driving force for the digital twin model, and a collaborative solution is performed in each sampling period.

[0021] As a preferred embodiment of the digital twin method based on the hot pot base grinding process described in this invention, the deep learning-based visual recognition model extracts quality feature values ​​from the base image by performing a collaborative solution of the material-flow field-temperature field in the grinding chamber based on a multi-physics mechanism model that couples discrete element method (DEM) and fluid dynamics (CFD).

[0022] Specifically, DEM was used to calculate the force, collision, friction, agglomeration and breakage behavior of the substrate particles in the grinding chamber, and the particle velocity field, contact network and local shear rate were obtained; CFD was used to calculate the flow and heat exchange process of the carrier medium and volatile components in the chamber, and the flow velocity field, pressure field and temperature field distribution were obtained.

[0023] Based on the contact friction power and shear energy terms output by DEM, they are coupled into the CFD heat conduction equation to form a dynamic update of the temperature field. According to the spatiotemporal evolution results of the temperature field distribution and shear force distribution, combined with the material crushing criteria, the particle size distribution and crushing evolution state are updated. The grinding process evolution trajectory consistent with the physical entity is reconstructed in the digital twin model to achieve synchronous mapping between virtual and real.

[0024] The material crushing criteria are dynamically updated by the force evolution criterion, the temperature influence criterion, and the spatiotemporal evolution criterion.

[0025] Select a time window corresponding to the discharge port, and perform weighted fusion of the prediction results of the joint solution of DEM and CFD within the current time window to form a set of predicted quality feature values ​​that characterize the grinding quality trend of hot pot base.

[0026] As a preferred embodiment of the digital twin method based on the hot pot base grinding process described in this invention, the comparison with the simulation prediction value output by the digital twin model includes, during the continuous sampling process of the macroscopic morphology image at the outlet, using a deep learning visual recognition model to map the macroscopic morphology image into a real feature vector representing the real-time quality state of the base material.

[0027] The feature vector is time-aligned with the predicted quality feature vector output by the digital twin multiphysics mechanism model at the same time, and the Euclidean distance difference is calculated to obtain the virtual-real residual. Based on the virtual-real residual, the physical boundary parameters and interaction parameters in the mechanism model of the digital twin are corrected online, so that the granular evolution, fragmentation degree and morphological uniformity of the simulation prediction gradually converge to the visual image observation results.

[0028] As a preferred embodiment of the digital twin method based on the grinding process of hot pot base material according to the present invention, the step of real-time correction of the physical boundary parameters in the digital twin model according to the comparison deviation includes using a Bayesian optimization algorithm to search for physical variables that cause prediction deviation in the parameter space.

[0029] The grinding disc friction coefficient and material breakage rate coefficient in the mechanism model are used as optimization objectives, and the virtual and real residuals are minimized as the objective function for iterative optimization. The physical variables obtained by iterative optimization are hot-updated to the mechanism model of the digital twin through dynamic linking technology for parameter calibration.

[0030] The calibrated parameters drive the digital twin model to predict future short-term process trends, obtaining the predicted quality feature vector within the future window. The parameters corresponding to the predicted quality feature vector are converted into control variables and executed. After execution, the next sampling cycle begins, making the predicted quality feature vector approximate the true feature vector.

[0031] Secondly, the present invention provides a digital twin system based on the grinding process of hot pot base, including a multi-source data acquisition module for real-time acquisition of the operating parameters of the grinding equipment and acquisition of macroscopic morphological images of the hot pot base at the outlet.

[0032] The multiphysics digital twin simulation module is used to establish a geometric model corresponding to the physical grinding equipment in virtual space and integrate a multiphysics mechanism model coupled with DEM-CFD.

[0033] The virtual-real comparison module is used to extract feature vectors from the macroscopic shape image based on a deep learning visual recognition model and calculate the virtual-real residual.

[0034] The parameter calibration module is used to correct the physical boundary parameters and interaction parameters in the multiphysics mechanism model online based on the virtual and real residuals, generate the optimal control command and send it to the grinding equipment control unit.

[0035] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the digital twin method based on the hot pot base grinding process described in the first aspect of the present invention.

[0036] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the digital twin method based on the hot pot base grinding process as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: It utilizes a deep learning visual recognition model to extract true quality features from measured images and compares these features with the simulation predictions of a digital twin model. Based on the virtual-real residuals, it corrects the physical boundary parameters (such as grinding disc gap and breakage coefficient) in the mechanism model online, achieving adaptive calibration of the model. Finally, based on the corrected high-fidelity model, it predicts future process trends and generates optimal control commands. This invention effectively solves the problem of inaccurate predictions caused by equipment wear or material fluctuations in traditional simulation models, achieving precise and intelligent closed-loop control of the grinding process. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a digital twin method based on the grinding process of hot pot base.

[0040] Figure 2 This is a schematic diagram of a computer device for a digital twin method based on the grinding process of hot pot base. Detailed Implementation

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0042] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0044] Reference Figure 1 As one embodiment of the present invention, this embodiment provides a digital twin method based on the grinding process of hot pot base, including the following steps:

[0045] S1: Real-time acquisition of operating parameters and bottom material images of the grinding process, construction of geometric model of grinding equipment, and integration of multi-physics mechanism model based on discrete element method and computational fluid dynamics coupling.

[0046] Furthermore, the real-time acquisition of operating parameters and bottom material images of the grinding process includes installing temperature and pressure sensors on the outer wall of the grinding chamber and at the inlet and outlet; installing current and torque sensors on the main shaft drive motor to calculate the grinding resistance; and installing a triaxial vibration sensor on the equipment base to monitor the contact state of the grinding disc.

[0047] An industrial high-frequency camera with self-cleaning function is installed at the discharge port to acquire macroscopic images of the bottom material at the discharge port.

[0048] The integrated multiphysics mechanism model based on the coupling of discrete element method and computational fluid dynamics includes establishing a geometric model corresponding to the physical grinding equipment in virtual space. The geometric model includes the grinding chamber, rotor grinding disc, discharge port and material flow channel structure. The geometric boundary constraints of the multiphysics simulation are obtained based on the geometric model.

[0049] Under the geometric boundary constraints, a particle contact and crushing mechanism model based on the discrete element method (DEM) and a flow field and thermal field mechanism model based on computational fluid dynamics (CFD) are constructed and integrated. The particle contact and crushing mechanism model is coupled with the flow field and thermal field mechanism model to obtain a multi-physics mechanism model characterizing the temperature field distribution, shear force distribution and material crushing evolution process inside the grinding chamber.

[0050] Physics field initialization and boundary condition definition:

[0051] Establish a discrete element DEM mesh for the grinding chamber and define the stiffness coefficient and restitution coefficient of solid particles (such as chili peppers and Sichuan peppercorns).

[0052] The non-Newtonian properties of grease fluids are defined based on the Navier-Stokes equations, where dynamic viscosity η and shear rate are related. The relationship follows a power-law model:

[0053]

[0054] in, is the consistency coefficient, and n is the flow index.

[0055] A visual feature extraction network was constructed, using ResNet-50 as the backbone network and integrating a fully connected regression layer at the end to output the average particle size D of the substrate.

[0056] Loss function design: The network is trained using mean squared error (MSE) to enable it to invert the three-dimensional particle distribution features from the two-dimensional image texture.

[0057] Heterogeneous models are coupled and integrated, and a feature correlation matrix is ​​established to logically bind the shear stress S in the physical field with the texture roughness R in the visual model, thus establishing the mapping relationship between the models.

[0058] S2: Input the operating parameters into the multiphysics mechanism model, perform digital twin model simulation of the inside of the grinding equipment, and output the simulation prediction value.

[0059] The digital twin model simulation of the internal structure of the grinding equipment includes using the collected operating parameters of the grinding equipment as real-time driving parameters for the digital twin model to synchronize and align the physical grinding process with the virtual simulation process.

[0060] The operating parameters include grinding speed, current load, feed flow rate, grinding chamber pressure, grinding chamber temperature, equipment vibration, and acoustic characteristics.

[0061] The multi-source operating parameters are time-stamp aligned and noise-reducing filtered, and an operating parameter sequence is formed based on the geometric model of the equipment and the mapping relationship between operating conditions.

[0062] At sampling time The system acquires the sequence of operating parameters of physical entities in real time via an industrial bus.

[0063]

[0064] in This indicates the real-time rotational speed of the grinding disc, which determines the rotational angular velocity of the boundary walls in the discrete element model (DEM). This indicates the pressure inside the grinding chamber, used to set the pressure outlet boundary for fluid computational CFD. This represents the motor current load, which is converted to obtain the torque, and then the shear work applied to the particles is calculated. This represents the material feed temperature, which serves as the initial temperature field background value for the thermodynamic equations. The feed rate determines the particle injection rate within the virtual space.

[0065] The operating parameters are converted into boundary and initial conditions for a multiphysics mechanism model, including:

[0066] The grinding speed is mapped to the rotor's rotational boundary, the feed flow rate and initial particle size distribution are mapped to DEM particle injection conditions, the current load is mapped to contact friction power and shear energy terms, and the cooling medium flow rate and cavity temperature are mapped to the convective heat transfer boundary on the CFD side.

[0067] The sequence of operating parameters is used as the real-time driving force for the digital twin model, and a collaborative solution is performed in each sampling period.

[0068] S3: The deep learning-based visual recognition model extracts quality feature values ​​from the base material image and compares them with the simulation prediction values ​​output by the digital twin model.

[0069] The deep learning-based visual recognition model extracts quality feature values ​​from the base material image by performing a collaborative solution of the material-flow field-temperature field in the grinding chamber based on a multi-physics mechanism model that couples discrete element method (DEM) and fluid dynamics (CFD).

[0070] Based on the fourth-order Runge-Kutta multi-field collaborative solution, the system will As the driving boundary, the mechanism engine is invoked to solve for the energy and momentum balance:

[0071]

[0072] in, This represents the viscous dissipation term. It indicates the rate at which the mechanical energy of the base oil is converted into internal energy under high-speed shearing. The material flow velocity vector is represented by the CFD module based on pressure P and rotational speed. The calculation yielded the result. It represents the change in the temperature field. This indicates the real-time temperature at any point inside the grinding chamber. Indicates specific heat capacity. Indicates a unit of time. Density is the ratio of the mass to the volume of a material. It represents the heat absorption capacity of a unit volume of substance and is used for calculating heat changes.

[0073] The fourth-order Runge-Kutta algorithm is used to iterate the above differential equations over time steps to ensure the numerical stability of temperature evolution during the online non-grinding process.

[0074] Specifically, DEM was used to calculate the force, collision, friction, agglomeration and breakage behavior of the substrate particles in the grinding chamber, and the particle velocity field, contact network and local shear rate were obtained; CFD was used to calculate the flow and heat exchange process of the carrier medium and volatile components in the chamber, and the velocity field, pressure field and temperature field distribution were obtained.

[0075] Based on the contact friction power and shear energy terms output by DEM, they are coupled into the CFD heat conduction equation to form a dynamic update of the temperature field. According to the spatiotemporal evolution results of the temperature field distribution and shear force distribution, combined with the material crushing criteria, the particle size distribution and crushing evolution state are updated. The grinding process evolution trajectory consistent with the physical entity is reconstructed in the digital twin model to achieve synchronous mapping between virtual and real.

[0076] The material crushing criterion is dynamically updated based on the force evolution criterion, the temperature influence criterion, and the spatiotemporal evolution. The force evolution criterion includes: The stress distribution of a material is calculated based on the theory of mechanics of materials. It is assumed that the stress field of the material is caused by both external forces (such as shear force and pressure) and internal reaction forces. The stress state is described as follows:

[0077] in, It is the stress tensor. It is the strain tensor. It is the stiffness tensor of the material.

[0078] Calculate the critical stress for material fracture. For each small unit or particle, calculate its critical fracture stress. Typically, theories based on fracture mechanics are used:

[0079] in, It refers to the fracture toughness of the material. It is the crack area of ​​the material.

[0080] To determine whether a crack has occurred, the stress in a certain area... Exceeding the critical stress When the region begins to fracture, the model records the fracture event in that region and updates the stress field of the material.

[0081] Temperature influence criteria: based on the current temperature field of the material. and specific heat capacity The effect of temperature change on material crushing behavior is calculated. The temperature change is calculated using the heat conduction equation:

[0082] in, This is a source term, representing internal heat sources (such as friction, viscous dissipation, etc.). Temperature changes affect the brittleness and ductility of materials; generally, materials become more prone to fracture as temperature increases.

[0083] Calculate the effect of temperature on the stress state of materials. High temperatures may lead to a decrease in the elastic modulus of materials. Changes have occurred:

[0084] in, It is the temperature coefficient. It is the elastic modulus of the material at a reference temperature. Due to the influence of temperature on stress, changes in the temperature field will also dynamically adjust the fracture criteria of the material.

[0085] Spatiotemporal evolution dynamic update: based on the material's time... and space The system updates the stress, temperature, and displacement of each particle or unit based on its position. A fourth-order Runge-Kutta algorithm (RK4) is used for time-step iterations to ensure numerical stability of the material crushing process. In each time step, the temperature and stress fields of each unit are updated and corrected according to the dynamic changes in the material.

[0086] The mechanical behavior between particles is calculated using a Direct Emitter Reduction (DEM). In the DEM model, contact forces, collisions, friction, agglomeration, and breakup between particles are calculated through particle interactions. Each collision and friction between particles affects the local temperature and stress distribution of the material. The particle velocity field, contact network, and local shear rate are calculated as inputs for subsequent temperature and stress field updates. Fluid flow and heat exchange processes are calculated using Computational Fluid Dynamics (CFD). The flow, pressure field, and temperature field changes of fluids (such as gas or liquid media) within the grinding chamber are calculated using a CFD model. During material breakup, fluid flow affects the temperature field distribution of the material, while material breakup also affects the fluid flow state and heat exchange process.

[0087] A comprehensive evaluation of the fracture and temperature changes of each particle or material unit is conducted. For each particle, dynamic updates are performed based on its stress, temperature, and historical state to determine whether fracture conditions have been met. Considering the nonlinear characteristics of the crushing process, a dynamically updated thermodynamic and mechanical model is used to ensure accurate simulation of material crushing. Finally, a determination is made as to whether crushing has occurred. During each update, the state of all particles or units is checked, and when the fracture criteria are met, fracture processing is performed, and the corresponding model parameters are updated.

[0088] Select a time window corresponding to the discharge port, and perform weighted fusion of the prediction results of the joint solution of DEM and CFD within the current time window to form a set of predicted quality feature values ​​that characterize the grinding quality trend of hot pot base.

[0089] S4: Correct the physical boundary parameters in the digital twin model in real time based on the comparison deviation, predict the process trend based on the corrected digital twin model, and generate the optimal control command.

[0090] The comparison with the simulation prediction value output by the digital twin model includes, during the continuous sampling process of the macroscopic morphology image at the discharge port, using a deep learning visual recognition model to map the macroscopic morphology image into a real feature vector representing the real-time quality state of the bottom material.

[0091] The feature vector is time-aligned with the predicted quality feature vector output by the digital twin multiphysics mechanism model at the same time, and the Euclidean distance difference is calculated to obtain the virtual-real residual. Based on the virtual-real residual, the physical boundary parameters and interaction parameters in the mechanism model of the digital twin are corrected online, so that the granular evolution, fragmentation degree and morphological uniformity of the simulation prediction gradually converge to the visual image observation results.

[0092] The quantification calculation of the virtual and real residuals and the trigger determination system run in parallel with the deep learning visual recognition module. The system analyzes the images captured by the high-frequency industrial camera installed at the discharge port and inverts the measured average particle size of the bottom material to obtain the result. Timing alignment is necessary because there is a lag time between the material's arrival at the discharge port from the inside of the grinding chamber. The system will display the current time. True feature vector and Predictive quality eigenvector at time step Alignment is performed. Normalized residuals are calculated; to eliminate the influence of dimensions, normalized Euclidean distance is used to calculate the deviation index.

[0093]

[0094] Perform dead zone threshold determination and set a deviation tolerance threshold. (For example, 5%). Only when Error > Only when the deviation is within a certain threshold will the system determine that the model is inaccurate (possibly due to increased gap caused by grinding disc wear, or batch-to-batch fluctuations in the hardness of the raw materials). This triggers the subsequent parameter calibration process. If the deviation is within the threshold, the current parameters are maintained to save computational resources. This represents the predicted quality feature vector value.

[0095] The real-time correction of physical boundary parameters in the digital twin model based on the comparison deviation includes using a Bayesian optimization algorithm to search for physical variables that cause prediction deviations within the parameter space.

[0096] The grinding disc friction coefficient and material breakage rate coefficient in the mechanistic model are used as optimization objectives, and the virtual-real residual is minimized as the objective function for iterative optimization. The physical variables obtained by iterative optimization are then dynamically updated to the digital twin mechanistic model for parameter calibration.

[0097] Backward parameter optimization based on Bayesian optimization: After determining the space of dominant variables requiring adjustment, the system uses a Bayesian optimization algorithm to find the optimal parameter combination that minimizes the residuals. Compared to genetic algorithms or particle swarm optimization, Bayesian optimization uses a Gaussian process as a surrogate model, finding the global optimum with very few iterations, making it very suitable for computationally expensive digital twin scenarios. The surrogate model is constructed using historical computational data to establish the parameters. With residual The model employs a Gaussian process regression model. Guided by a sampling function, it utilizes an expectation improvement strategy to calculate the next sampling point. This strategy not only focuses on the region with the smallest current prediction residual (development) but also on regions with larger variance, i.e., higher uncertainty (exploration), thereby avoiding getting trapped in local optima. This indicates the deviation in the grinding disc clearance. The physical grinding disc may shrink due to thermal expansion or expand due to wear. This parameter corrects the boundary of the geometric model and directly affects the grinding fineness. This indicates the breakage probability coefficient, reflecting the brittleness of the raw material (such as chili peppers). This coefficient varies from batch to batch and needs to be dynamically adjusted to match the actual breakage rate.

[0098] The calibrated parameters drive the digital twin model to predict future short-term process trends, obtaining the predicted quality feature vector within the future window. The parameters corresponding to the predicted quality feature vector are converted into control variables and executed. After execution, the next sampling cycle begins, making the predicted quality feature vector approximate the true feature vector.

[0099] This embodiment also provides a digital twin system based on the hot pot base grinding process, including:

[0100] The multi-source data acquisition module is used to collect the operating parameters of the grinding equipment in real time and obtain the macroscopic morphological image of the hot pot base material at the discharge port.

[0101] The multiphysics digital twin simulation module is used to establish a geometric model corresponding to the physical grinding equipment in virtual space and integrate a multiphysics mechanism model coupled with DEM-CFD.

[0102] The virtual-real comparison module is used to extract feature vectors from the macroscopic shape image based on a deep learning visual recognition model and calculate the virtual-real residual.

[0103] The parameter calibration module is used to correct the physical boundary parameters and interaction parameters in the multiphysics mechanism model online based on the virtual and real residuals, generate the optimal control command and send it to the grinding equipment control unit.

[0104] This embodiment also provides a computer device, such as... Figure 2As shown, the applicable digital twin method based on the hot pot base grinding process includes: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the digital twin method based on the hot pot base grinding process proposed in the above embodiments.

[0105] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0106] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the digital twin method based on the hot pot base grinding process proposed in the above embodiments. 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 (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0107] In summary, this invention achieves the following: real-time acquisition of operating parameters and bottom material images from the grinding process to construct a geometric model of the grinding equipment; integration of a multiphysics mechanism model based on the coupling of discrete element method and computational fluid dynamics; input of operating parameters into the multiphysics mechanism model to perform digital twin model simulation of the grinding equipment's internal structure and output simulation prediction values; extraction of quality feature values ​​from the bottom material images using a deep learning-based visual recognition model and comparison with the simulation prediction values ​​output by the digital twin model; real-time correction of physical boundary parameters in the digital twin model based on the comparison deviation; prediction of process trends based on the corrected digital twin model; and generation of optimal control commands.

[0108] 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 scope of the claims of the present invention.

Claims

1. A digital twin method based on the grinding process of hot pot base, characterized in that: This includes real-time acquisition of operating parameters and bottom material images from the discharge port during the grinding process, construction of a geometric model of the grinding equipment, and integration of a multi-physics mechanism model based on the coupling of discrete element method and computational fluid dynamics. The operating parameters are input into the multiphysics mechanism model to perform digital twin model simulation of the inside of the grinding equipment, and the simulation prediction values ​​are output. A deep learning-based visual recognition model extracts quality feature values ​​from the base material image and compares them with the simulation prediction values ​​output by the digital twin model. The physical boundary parameters in the digital twin model are corrected in real time based on the comparison deviation. Based on the corrected digital twin model, the process trend is predicted and the optimal control command is generated.

2. The digital twin method based on the hot pot base grinding process as described in claim 1, characterized in that: The real-time acquisition of operating parameters and bottom material images of the grinding process includes installing temperature and pressure sensors on the outer wall of the grinding chamber and at the inlet and outlet; installing current and torque sensors on the main shaft drive motor to calculate the grinding resistance; and installing a triaxial vibration sensor on the equipment base to monitor the contact state of the grinding disc. An industrial high-frequency camera with self-cleaning function is installed at the discharge port to acquire macroscopic images of the bottom material at the discharge port.

3. The digital twin method based on the hot pot base grinding process as described in claim 2, characterized in that: The integrated multiphysics mechanism model based on the coupling of discrete element method and computational fluid dynamics includes establishing a geometric model corresponding to the physical grinding equipment in virtual space. The geometric model includes the grinding chamber, rotor grinding disc, discharge port and material flow channel structure. The geometric boundary constraints of the multiphysics simulation are obtained based on the geometric model. Under the geometric boundary constraints, a particle contact and crushing mechanism model based on the discrete element method (DEM) and a flow field and thermal field mechanism model based on computational fluid dynamics (CFD) are constructed and integrated. The particle contact and crushing mechanism model and the flow field and thermal field mechanism model are coupled to obtain a multi-physics mechanism model characterizing the temperature field distribution, shear force distribution and material crushing evolution process inside the grinding chamber.

4. The digital twin method based on the hot pot base grinding process as described in claim 3, characterized in that: The digital twin model simulation of the interior of the grinding equipment includes using the collected operating parameters of the grinding equipment as real-time driving quantities for the digital twin model to synchronize and align the physical grinding process with the virtual simulation process. The operating parameters include grinding speed, current load, feed flow rate, grinding chamber pressure, grinding chamber temperature, equipment vibration, and acoustic characteristics. The multi-source operating parameters are time-stamp aligned and noise-reducing filtered, and an operating parameter sequence is formed based on the geometric model of the equipment and the mapping relationship between operating conditions. The operating parameters are converted into boundary and initial conditions for a multiphysics mechanism model, including: The grinding speed is mapped to the rotor's rotational boundary, the feed flow rate and the initial particle size distribution of the material are mapped to the DEM particle injection conditions, the current load is mapped to the contact friction power and shear energy terms, and the cooling medium flow rate and the cavity temperature are mapped to the convective heat transfer boundary on the CFD side. The sequence of operating parameters is used as the real-time driving force for the digital twin model, and a collaborative solution is performed within each sampling period.

5. The digital twin method based on the hot pot base grinding process as described in claim 4, characterized in that: The deep learning-based visual recognition model extracts quality feature values ​​from the base material image by performing a collaborative solution of the material-flow field-temperature field in the grinding chamber based on a multi-physics mechanism model that couples discrete element method (DEM) and fluid dynamics (CFD). Specifically, DEM was used to calculate the force, collision, friction, agglomeration and breakage behavior of the substrate particles in the grinding chamber, and the particle velocity field, contact network and local shear rate were obtained; CFD was used to calculate the flow and heat exchange process of the carrier medium and volatile components in the chamber, and the flow velocity field, pressure field and temperature field distribution were obtained. Based on the contact friction power and shear energy terms output by DEM, they are coupled into the CFD heat conduction equation to form a dynamic update of the temperature field. According to the spatiotemporal evolution results of the temperature field distribution and shear force distribution, combined with the material crushing criteria, the particle size distribution and crushing evolution state are updated. The grinding process evolution trajectory consistent with the physical entity is reconstructed in the digital twin model to achieve synchronous mapping between virtual and real. The material crushing criteria are dynamically updated by the force evolution criterion, the temperature influence criterion, and the spatiotemporal evolution criterion. Select a time window corresponding to the discharge port, and perform weighted fusion of the prediction results of the joint solution of DEM and CFD within the current time window to form a set of predicted quality feature values ​​that characterize the grinding quality trend of hot pot base.

6. The digital twin method based on the hot pot base grinding process as described in claim 5, characterized in that: The comparison with the simulation prediction value output by the digital twin model includes, during the continuous sampling process of the macroscopic morphology image of the discharge port, using a deep learning visual recognition model to map the macroscopic morphology image into a real feature vector representing the real-time quality state of the bottom material. The feature vector is time-aligned with the predicted quality feature vector output by the digital twin multiphysics mechanism model at the same time, and the Euclidean distance difference is calculated to obtain the virtual-real residual. Based on the virtual-real residual, the physical boundary parameters and interaction parameters in the mechanism model of the digital twin are corrected online, so that the granular evolution, fragmentation degree and morphological uniformity of the simulation prediction gradually converge to the visual image observation results.

7. The digital twin method based on the hot pot base grinding process as described in claim 6, characterized in that: The real-time correction of physical boundary parameters in the digital twin model based on the comparison deviation includes using a Bayesian optimization algorithm to search for physical variables that cause prediction deviations in the parameter space. The grinding disc friction coefficient and material breakage rate coefficient in the mechanism model are used as optimization objectives, and the virtual and real residuals are minimized as the objective function for iterative optimization. The physical variables obtained by iterative optimization are hot-updated to the mechanism model of the digital twin through dynamic linking technology for parameter calibration. The calibrated parameters drive the digital twin model to predict future short-term process trends, obtaining the predicted quality feature vector within the future window. The parameters corresponding to the predicted quality feature vector are converted into control variables and executed. After execution, the next sampling cycle begins, making the predicted quality feature vector approximate the true feature vector.

8. A digital twin system based on the grinding process of hot pot base, based on the digital twin method based on the grinding process of hot pot base as described in any one of claims 1 to 7, characterized in that: This includes a multi-source data acquisition module, used to collect the operating parameters of the grinding equipment in real time and obtain macroscopic images of the hot pot base material at the discharge port; The multiphysics digital twin simulation module is used to establish a geometric model corresponding to the physical grinding equipment in virtual space and integrate a multiphysics mechanism model coupled with DEM-CFD. The virtual-real comparison module is used to extract feature vectors from the macroscopic shape image based on a deep learning visual recognition model and calculate the virtual-real residual. The parameter calibration module is used to correct the physical boundary parameters and interaction parameters in the multiphysics mechanism model online based on the virtual and real residuals, generate the optimal control command and send it to the grinding equipment control unit.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the digital twin method based on the hot pot base grinding process as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the digital twin method based on the hot pot base grinding process as described in any one of claims 1 to 7.

Images