Virtual-real collaborative control method for rough and fine grinding integrated machine based on digital twinning

Abstract

The application belongs to the technical field of digital twinning, and particularly discloses a rough and fine grinding integrated machine virtual-real collaborative control method based on digital twinning, which comprises the following steps: S1, collecting real-time processing parameters of a rough grinding unit and a fine grinding unit, and constructing a grinding digital twinning body synchronized with an actual processing process; S2, constructing a long short-term memory network model to predict the state evolution trend of each unit in a future processing process; S3, performing process parameter optimization of a rough grinding to fine grinding whole process in the grinding digital twinning body; S4, searching for an optimal process parameter combination in the grinding digital twinning body environment by using a snake swarm optimization algorithm; S5, driving actual rough grinding and fine grinding process execution; and S6, if a processing abnormality or deviation is detected, automatically triggering re-optimization and correction control. The application combines a snake swarm optimization algorithm, a long short-term memory network model and an industrial field bus control technology, and realizes virtual-real collaborative grinding control of a full-automatic rough and fine grinding integrated machine based on digital twinning.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a virtual-real collaborative control method for an integrated rough and fine grinding machine based on digital twins. Background Technology

[0002] With the development of intelligent manufacturing and Industry 4.0, the machining field is placing increasingly higher demands on the quality and efficiency of grinding. As an important processing method, grinding is widely used in the surface finishing, dimensional control, and surface functional enhancement of metal parts. Especially in manufacturing scenarios with high precision and high consistency requirements, integrated rough and fine grinding equipment is gradually becoming the mainstream application trend. By integrating the rough grinding unit and the fine grinding unit into the same machine tool, it is possible to reduce workpiece transfer, improve processing efficiency, and enhance overall processing quality. However, traditional integrated rough and fine grinding machines still face many problems, especially in terms of control precision, process optimization, real-time response, and the degree of equipment intelligence, where there are still significant shortcomings.

[0003] Most fully automatic roughing and finishing grinding machines currently rely on preset, fixed process parameters for control, such as fixed feed rate, grinding depth, grinding wheel speed, and grinding pressure. This static parameter setting method lacks real-time response to changes in operating conditions during processing, easily leading to significant fluctuations in processing quality across different batches, materials, and environmental conditions. This can result in inconsistent surface roughness, dimensional accuracy deviations, localized burning, or excessive grinding wheel wear. Furthermore, traditional control methods lack effective predictive mechanisms and optimization strategies, often reacting passively after anomalies occur. This leads to increased energy consumption, decreased equipment utilization, and an inability to simultaneously optimize processing efficiency and quality.

[0004] In existing technologies, some studies have attempted to apply simple model predictive control or shallow neural networks to the adjustment of grinding parameters. However, due to the high complexity of the grinding process, its strong temporal dynamic characteristics, and spatial coupling relationships, traditional methods have significant limitations in capturing the evolution patterns of the processing and modeling the nonlinear dynamic relationships between complex process parameters. They often fail to accurately predict future processing state evolution trends. Furthermore, while traditional optimization methods such as genetic algorithms and particle swarm optimization algorithms can improve parameter optimization to some extent, they are insufficient in multi-objective optimization, local extremum escape, and dynamic convergence speed, making it difficult to meet the high requirements for real-time performance and precision in the combined roughing and finishing grinding process.

[0005] Digital twin technology, as an emerging method of virtual-real integration, has been widely used in equipment modeling, condition prediction, and process optimization in recent years. By establishing a virtual mirror that is dynamically synchronized with the actual equipment, digital twins can achieve real-time data synchronization, process simulation, fault prediction, and optimization decision-making. However, the application of digital twins for integrated rough and fine grinding machines is still in its early stages, mostly limited to equipment condition visualization. It lacks the ability to deeply model processing mechanisms, dynamically adjust fine-grained process parameters, and achieve closed-loop control through virtual-real collaboration, thus failing to fully unleash the potential of digital twins in adaptive optimization and control of the grinding process.

[0006] Therefore, how to provide a virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose a virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins. This invention combines snake swarm optimization algorithm, long short-term memory network model, and industrial fieldbus control technology. By constructing a grinding digital twin synchronized with the actual machining process, it captures the processing dynamics and key physical quantity changes of the roughing and finishing units in real time. Utilizing graph structure modeling and temporal feature extraction, it accurately predicts the state evolution trend of the machining units. The system employs a snake swarm optimization algorithm to search for the optimal combination of process parameters throughout the entire process in the digital twin environment. Through foraging and entanglement disturbance mechanisms, it improves optimization efficiency and the ability to escape local extrema, dynamically optimizing the process control commands during the roughing and finishing grinding linkage stages. The optimal process parameters are transmitted to the machining execution unit in real time via the industrial fieldbus, and adaptive corrections are made based on real-time feedback data, forming a closed-loop control system. This achieves virtual-real collaboration, autonomous optimization, and efficient execution of the roughing and finishing grinding processes, significantly improving the stability, efficiency, and intelligence level of grinding processing, and possessing high real-time performance, strong robustness, and continuous optimization capabilities.

[0008] A method for virtual-physical collaborative control of an integrated roughing and finishing grinding machine based on digital twins includes the following steps: S1. On the fully automatic coarse and fine grinding integrated machine, collect the real-time processing parameters of the coarse grinding unit and the fine grinding unit, and construct a grinding digital twin that is synchronized with the actual processing process; S2. Based on the grinding digital twin, a long short-term memory network model is constructed to transmit information through the connection relationship between nodes and predict the evolution trend of the state of each unit in the future processing process. S3. Based on the prediction results, optimize the process parameters of the entire process from rough grinding to fine grinding in the grinding digital twin, and output the optimal combination of process parameters obtained by optimization. S4. Based on the optimal combination of process parameters obtained through optimization, the snake swarm optimization algorithm is used to search for the best combination of process parameters in the grinding digital twin environment. Each snake is modeled as an individual containing a process parameter vector and an adjustment step size. It slides and searches according to the foraging direction and outputs the optimal process parameter scheme for the rough and fine grinding linkage stage. S5. Synchronously update the optimal process parameter scheme to the fully automatic coarse and fine grinding integrated machine to drive the actual coarse and fine grinding process; S6. During the actual grinding process, real-time processing parameters are continuously collected to update the grinding digital twin. If a processing abnormality or deviation is detected, re-optimization and correction control will be automatically triggered.

[0009] Optionally, S1 specifically includes: S11. Obtain the three-dimensional geometric data of the workpiece. The three-dimensional geometric data is a set of spatial coordinates of each measurement point of the workpiece, including the X coordinate, Y coordinate and Z coordinate of each measurement point. S12. Collect the physical parameters of the fully automatic coarse and fine grinding machine. The physical parameters include grinding wheel diameter, grinding wheel linear speed, spindle speed and feed speed. S13. Based on the three-dimensional geometric data of the workpiece and the physical parameters of the equipment, a digital twin data model of the grinding process is constructed through a time-synchronized data fusion method. S14. In the digital twin data model of the grinding process, the object state information is dynamically updated through a continuous time update mechanism so that the state of each object in the digital twin is synchronized with the state of the corresponding object in the actual processing process in time. S15. After the object state is synchronized, output the grinding digital twin.

[0010] Optionally, S2 specifically includes: S21. Based on the grinding digital twin, the rough grinding unit, the fine grinding unit, and the key processing components are defined as a set of nodes; S22. Construct a node feature vector for each node. The node feature vector includes the spatial coordinates of the component corresponding to the node, the processing state parameters, and the physical quantity change characteristic values. The processing state parameters include the current feed rate, grinding depth, and spindle speed. The physical quantity change characteristic values ​​include the grinding wheel load change rate, motor power change rate, grinding force change rate, and workpiece surface temperature change rate. S23. Based on the process flow logic and physical connection relationships, establish a set of edge connections between nodes; S24. Combine the set of nodes and the set of edges to form a graph structure, and use the graph structure as input to the long short-term memory network model for modeling. S25. Using a long short-term memory network model, based on the feature vector information propagation mechanism of neighbor nodes, the feature vector of each node in the previous time step is aggregated with the feature vector of neighbor nodes to output the updated node features. S26. Through iterative propagation and node state updates over multiple time steps, predict the future state evolution trend of each unit in the processing process: ; in, Let be the optimal processing state category predicted for the i-th processing unit of the integrated rough and fine grinding machine at time t. To select the category that maximizes the sum of the weighted features of its neighboring nodes among all possible processing state categories, For state categories, For the set of node state categories, The weighted sum of all neighboring units j directly connected to the i-th processing unit is calculated, where i is the i-th processing node in the Long Short-Term Memory network model, j are the neighboring nodes of processing node i, and t is the time step index. Let i be the set of neighboring processing units that have direct information interaction with the i-th processing unit in the process flow. The information propagation weight between the i-th processing unit and its j-th neighboring processing unit. Let be the characteristic state quantity of the j-th neighboring processing unit at time t.

[0011] Optionally, S3 specifically includes: S31. Extract the node status prediction values ​​of the coarse grinding unit and the fine grinding unit within a future predetermined time window. The node status prediction value is the comprehensive processing status of the node at the predetermined time. S32. Based on the predicted state values ​​of each node, determine the main process parameters that affect the processing quality and energy consumption of rough grinding and fine grinding; S33. Perform comprehensive optimization of process parameters for the entire process from rough grinding to fine grinding: ; in, To optimize the obtained optimal combination of process parameters, Let be the vector of process parameters to be optimized. This represents the total number of processing nodes in the integrated rough and fine grinding machine. For the first The current predicted surface roughness of the node. This is the current processing unit node. , , , To optimize the weighting coefficients of the indicators, Let be the target surface roughness of the i-th node. For absolute value operators, Let i be the currently predicted processing time for the i-th node. Let be the target processing time for the i-th node. For a moment Energy consumption per unit time The arithmetic mean roughness, The continuous physical time during the processing. For a moment The change in surface temperature of the workpiece per unit time. This is the minimize operator.

[0012] Optionally, S4 specifically includes: S41. Initialize the snake population in the grinding digital twin, and model each snake as an individual containing the process parameter vector P and the corresponding adjustment step size ΔP; S42. Set an initial position and initial step size for each individual snake. The initial position represents the combination of process parameters, and the initial step size represents the magnitude of a single parameter update. S43. Update the snake's position based on its foraging direction, and guide the snake to glide to the best position in the current snake population. S44. Combine the gliding stride length adjustment strategy and dynamically adjust the stride length according to the current fitness assessment value of the individual snake. S45. Compared to the snake swarm algorithm, the snake swarm optimization method introduces global direction guidance based on the difference in the optimal process parameter combination in the entanglement perturbation mechanism. It also combines multidimensional random perturbation to achieve dynamic adjustment of the perturbation amplitude in relation to the deviation from the process target. When a snake individual gets stuck in a local optimum, a perturbation search is performed based on the difference between the current snake individual's process parameter vector and the optimal process parameter combination P*. ; in, For the first The perturbation search results generated after the snake coils around the perturbation. This refers to the m-th process parameter in the optimal combination of process parameters. For the first The m-th process parameter value of the snake at its current position. This is the disturbance amplitude control coefficient. The number of dimensions in the process parameter vector. This introduces a small-amplitude random disturbance to the m-th dimension, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of each snake body; S46. Iteratively execute the foraging direction update, step size adjustment and entanglement disturbance process until the maximum number of iterations of 200 is met, and output the optimal process parameter scheme.

[0013] Optionally, S5 specifically includes: S51. Perform parameter analysis on the optimal process parameter scheme and extract the set values ​​corresponding to each process parameter. S52. Generate a process control instruction matrix based on each set value; S53. Encapsulate the process control instruction matrix according to the industrial fieldbus communication protocol to form a control instruction data frame that can be sent out. S54. The process control command data frame is sent to the processing execution module of the fully automatic coarse and fine grinding machine through the industrial fieldbus to drive the actual coarse and fine grinding process.

[0014] Optionally, S54 specifically includes: S541. The processing execution module parses the received process control instruction data frame and sets the actual execution parameters of the rough grinding unit and the fine grinding unit. S542. After completing the actual execution parameter setting, control the fully automatic coarse and fine grinding integrated machine to start the coarse grinding and fine grinding process according to the process parameter combination optimized by winding disturbance, and the coarse grinding unit and the fine grinding unit are executed synchronously. S543. During the processing, based on the real-time collected process parameter data and the disturbance search results generated after the snake-like entanglement disturbance, Perform difference calculation: ; in, This represents the cumulative error during the processing. For the m-th process parameter at time The actual collected values, To optimize the perturbation through snake swarm entanglement, the first The process parameter values ​​for the snake in the m-th dimension. The number of dimensions in the process parameter vector. This represents the total processing time. Let be the error weight of the m-th process parameter, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of the snake body, This refers to the continuous physical time during the processing.

[0015] Optionally, S6 specifically includes: S61. During the processing of the fully automatic coarse and fine grinding integrated machine, the actual execution value of each process parameter is collected in real time, and a real-time process parameter vector at the corresponding time point is generated. S62. Compare the real-time collected process parameter vector with the optimal process parameter scheme obtained through snake swarm optimization to generate a process parameter deviation vector. S63. Based on the process parameter deviation vector, accumulate the deviation of each process parameter and multiply it by the dynamic adjustment ratio factor to calculate the dynamic adjustment amount. S64. The dynamic adjustment amount is superimposed on the optimal process parameter scheme obtained by snake swarm optimization to generate a dynamically corrected combination of process parameters. S65. Based on the dynamically corrected combination of process parameters, regenerate the dynamically corrected process control instruction set. S66. The dynamically corrected process control instruction set is sent to the processing execution module of the fully automatic coarse and fine grinding machine in real time via the industrial fieldbus. S67. Control the rough grinding unit and the fine grinding unit to synchronously execute the rough and fine grinding processes according to the dynamically corrected combination of process parameters. S68. Periodically repeat S61 to S67.

[0016] The beneficial effects of this invention are: This invention proposes a virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins. It fully integrates deep learning modeling and intelligent optimization algorithms, effectively overcoming the problems of dynamic control lag, insufficient optimization capabilities, and low system intelligence levels in existing technologies during the roughing and finishing grinding linkage process. By constructing a grinding digital twin that is synchronized with the actual machining process in real time, and using a long short-term memory network model to model the state evolution laws of the roughing grinding unit, the finishing grinding unit, and key components, this invention can accurately predict the state evolution trends of each unit in the future machining process within a virtual twin environment, providing high-precision data support for subsequent process parameter optimization and dynamic control.

[0017] Furthermore, this invention employs an innovative snake swarm optimization algorithm to search for the optimal combination of process parameters within a grinding digital twin environment. Each snake not only contains a process parameter vector but also incorporates an adaptive step-size adjustment mechanism. This mechanism guides the sliding search through foraging direction and uses entanglement perturbations to escape local optima, significantly improving convergence speed and global optimization capabilities in complex multi-objective optimization problems. The optimization process directly correlates the predicted node state with the machining dynamics, ensuring that the final output process parameter scheme balances machining quality, energy consumption control, and machining time optimization, comprehensively enhancing overall machining performance.

[0018] In the process parameter issuance and actual execution phases, this invention proposes a virtual-physical integrated control mechanism based on synchronous updates of optimal parameter combinations to the equipment. Optimized process control commands are transmitted in real-time via an industrial fieldbus, driving the rough grinding and fine grinding units to execute processing operations according to dynamically optimal parameters. Key process status data is collected in real-time during the processing. Combining the differences between the real-time collected data and the optimized parameter results, this invention designs a large-scale, complex mathematical evaluation function based on cumulative error integration. This enables real-time monitoring of processing stability and deviation trends during the execution phase, effectively ensuring the consistency and stability of the rough and fine grinding linkage process.

[0019] Furthermore, this invention constructs a closed-loop mechanism for adaptive adjustment of dynamic process parameters based on real-time feedback. By periodically comparing the differences between real-time process parameters and the optimal process combination after snake swarm optimization perturbation, it dynamically calculates the adjustment amount and corrects the process control commands in real time, forming an intelligent closed-loop control link of virtual-real collaboration, autonomous optimization, and adaptive correction in the processing process. The closed-loop mechanism greatly enhances the system's robustness to external disturbances such as operating condition fluctuations, equipment wear, and environmental changes, ensuring the continuous stability of processing quality and maximizing overall production efficiency in the roughing and finishing grinding stages.

[0020] In summary, this invention overcomes the limitations of existing integrated rough and fine grinding machine grinding control technologies in terms of machining state modeling, process parameter optimization, dynamic adaptive adjustment, and virtual-real collaborative control. It achieves intelligent optimized grinding control driven by a digital twin throughout the entire process, significantly improving the comprehensive machining performance of fully automatic integrated rough and fine grinding machines in terms of precision, efficiency, energy consumption, and reliability. It has significant engineering application value and broad prospects for industrial promotion. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The flowchart shows the virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins proposed in this invention. Figure 2 This is a schematic diagram of the virtual-real collaborative control method for an integrated rough and fine grinding machine based on digital twins proposed in this invention; Figure 3 This is a data flow diagram of the virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins proposed in this invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0023] refer to Figure 1-3 The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins includes the following steps: S1. On the fully automatic coarse and fine grinding integrated machine, collect the real-time processing parameters of the coarse grinding unit and the fine grinding unit, and construct a grinding digital twin that is synchronized with the actual processing process; S2. Based on the grinding digital twin, a long short-term memory network model is constructed to transmit information through the connection relationship between nodes and predict the evolution trend of the state of each unit in the future processing process. S3. Based on the prediction results, optimize the process parameters of the entire process from rough grinding to fine grinding in the grinding digital twin, and output the optimal combination of process parameters obtained by optimization. S4. Based on the optimal combination of process parameters obtained through optimization, the snake swarm optimization algorithm is used to search for the best combination of process parameters in the grinding digital twin environment. Each snake is modeled as an individual containing a process parameter vector and an adjustment step size. It slides and searches according to the foraging direction and outputs the optimal process parameter scheme for the rough and fine grinding linkage stage. S5. Synchronously update the optimal process parameter scheme to the fully automatic coarse and fine grinding integrated machine to drive the actual coarse and fine grinding process; S6. During the actual grinding process, real-time processing parameters are continuously collected to update the grinding digital twin. If a processing abnormality or deviation is detected, re-optimization and correction control will be automatically triggered.

[0024] This invention combines a grinding digital twin with a long short-term memory network model. Based on information propagation between nodes, it accurately predicts the state evolution of machining units and performs process parameter optimization and snake swarm optimization global search within the twin environment based on the prediction results. By modeling process parameter vectors and dynamically adjusting step sizes, it searches for the optimal solution for roughing and finishing grinding in a sliding motion, ensuring stability and accuracy throughout the entire machining process. Optimization results are synchronized to the equipment in real time, and dynamic data acquisition and anomaly triggering mechanisms work together to form a closed-loop machining control system, achieving high-precision, high-reliability, and adaptive grinding control in complex grinding scenarios.

[0025] In this embodiment, S1 specifically includes: S11. Obtain the three-dimensional geometric data of the workpiece. The three-dimensional geometric data is a set of spatial coordinates of each measurement point of the workpiece, including the X coordinate, Y coordinate and Z coordinate of each measurement point. S12. Collect the physical parameters of the fully automatic coarse and fine grinding machine. The physical parameters include grinding wheel diameter, grinding wheel linear speed, spindle speed and feed speed. S13. Based on the three-dimensional geometric data of the workpiece and the physical parameters of the equipment, a digital twin data model of the grinding process is constructed through a time-synchronized data fusion method. S14. In the digital twin data model of the grinding process, the object state information is dynamically updated through a continuous time update mechanism so that the state of each object in the digital twin is synchronized with the state of the corresponding object in the actual processing process in time. S15. After the object state is synchronized, output the grinding digital twin.

[0026] This invention constructs a grinding digital twin data model that accurately reflects the actual processing state, based on the three-dimensional geometric data of the workpiece and the physical parameters of the equipment, using a time-synchronized data fusion method. Through a continuous-time dynamic update mechanism, the state changes of each object are synchronized in real time, ensuring a high degree of consistency between the digital twin and the actual processing process. This method effectively improves the accuracy and timeliness of grinding process modeling, providing highly reliable data support for subsequent process parameter optimization, process control, and anomaly detection, significantly enhancing the system's real-time perception and intelligent decision-making capabilities.

[0027] In this embodiment, S2 specifically includes: S21. Based on the grinding digital twin, the rough grinding unit, the fine grinding unit, and the key processing components are defined as a set of nodes; S22. Construct a node feature vector for each node. The node feature vector includes the spatial coordinates of the component corresponding to the node, the processing state parameters, and the physical quantity change characteristic values. The processing state parameters include the current feed rate, grinding depth, and spindle speed. The physical quantity change characteristic values ​​include the grinding wheel load change rate, motor power change rate, grinding force change rate, and workpiece surface temperature change rate. S23. Based on the process flow logic and physical connection relationships, establish a set of edge connections between nodes; S24. Combine the set of nodes and the set of edges to form a graph structure, and use the graph structure as input to the long short-term memory network model for modeling. S25. Using a long short-term memory network model, based on the feature vector information propagation mechanism of neighbor nodes, the feature vector of each node in the previous time step is aggregated with the feature vector of neighbor nodes to output the updated node features. S26. Through iterative propagation and node state updates over multiple time steps, predict the future state evolution trend of each unit in the processing process: ; in, Let be the optimal processing state category predicted for the i-th processing unit of the integrated rough and fine grinding machine at time t. To select the category that maximizes the sum of the weighted features of its neighboring nodes among all possible processing state categories, For state categories, For the set of node state categories, The weighted sum of all neighboring units j directly connected to the i-th processing unit is calculated, where i is the i-th processing node in the Long Short-Term Memory network model, j are the neighboring nodes of processing node i, and t is the time step index. Let i be the set of neighboring processing units that have direct information interaction with the i-th processing unit in the process flow. The information propagation weight between the i-th processing unit and its j-th neighboring processing unit. Let be the characteristic state quantity of the j-th neighboring processing unit at time t.

[0028] This invention, based on a grinding digital twin, constructs a graph structure representation of rough grinding units, fine grinding units, and key components. Node feature vectors describe spatial location, processing state, and physical change characteristics. A neighbor node feature propagation mechanism is introduced, utilizing a long short-term memory network to achieve dynamic aggregation of node states and prediction of time-step evolution. Through weighted feature summation and category selection, the future processing state change trend of each unit is accurately predicted, significantly improving the accuracy, timeliness, and global synergy of grinding process modeling, providing intelligent decision support for subsequent process optimization and anomaly early warning.

[0029] In this embodiment, S3 specifically includes: S31. Extract the node status prediction values ​​of the coarse grinding unit and the fine grinding unit within a future predetermined time window. The node status prediction value is the comprehensive processing status of the node at the predetermined time. S32. Based on the predicted state values ​​of each node, determine the main process parameters that affect the processing quality and energy consumption of rough grinding and fine grinding; S33. Perform comprehensive optimization of process parameters for the entire process from rough grinding to fine grinding: ; in, To optimize the obtained optimal combination of process parameters, Let be the vector of process parameters to be optimized. This represents the total number of processing nodes in the integrated rough and fine grinding machine. For the first The current predicted surface roughness of the node. This is the current processing unit node. , , , To optimize the weighting coefficients of the indicators, Let be the target surface roughness of the i-th node. For absolute value operators, Let i be the currently predicted processing time for the i-th node. Let be the target processing time for the i-th node. For a moment Energy consumption per unit time The arithmetic mean roughness, The continuous physical time during the processing. For a moment The change in surface temperature of the workpiece per unit time. This is the minimize operator.

[0030] This invention extracts key process parameter influencing factors based on the predicted node states within a predetermined time window of the rough and fine grinding units, and establishes a comprehensive objective function for optimizing processing quality and energy consumption. By comprehensively considering surface roughness, processing time, energy consumption, and temperature rise, a multi-index weighted optimization strategy is set, and the optimal combination of process parameters for the entire process is obtained through minimization computation. This method balances processing efficiency and quality control in both the rough and fine grinding stages, achieving simultaneous reduction in energy consumption, controlled temperature rise, and improved surface quality, demonstrating efficient, precise, and intelligent optimization capabilities.

[0031] In this embodiment, S4 specifically includes: S41. Initialize the snake population in the grinding digital twin, and model each snake as an individual containing the process parameter vector P and the corresponding adjustment step size ΔP; S42. Set an initial position and initial step size for each individual snake. The initial position represents the combination of process parameters, and the initial step size represents the magnitude of a single parameter update. S43. Update the snake's position based on its foraging direction, and guide the snake to glide to the best position in the current snake population. S44. Combine the gliding stride length adjustment strategy and dynamically adjust the stride length according to the current fitness assessment value of the individual snake. S45. Compared to the snake swarm algorithm, the snake swarm optimization method introduces global direction guidance based on the difference in the optimal process parameter combination in the entanglement perturbation mechanism. It also combines multidimensional random perturbation to achieve dynamic adjustment of the perturbation amplitude in relation to the deviation from the process target. When a snake individual gets stuck in a local optimum, a perturbation search is performed based on the difference between the current snake individual's process parameter vector and the optimal process parameter combination P*. ; in, For the first The perturbation search results generated after the snake coils around the perturbation. This refers to the m-th process parameter in the optimal combination of process parameters. For the first The m-th process parameter value of the snake at its current position. This is the disturbance amplitude control coefficient. The number of dimensions in the process parameter vector. This introduces a small-amplitude random disturbance to the m-th dimension, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of each snake body; S46. Iteratively execute the foraging direction update, step size adjustment and entanglement disturbance process until the maximum number of iterations of 200 is met, and output the optimal process parameter scheme.

[0032] This invention initializes a snake population within a grinding digital twin, models the dynamic evolution of process parameter vectors and step size, guides the snake's gliding search based on its foraging direction, and dynamically optimizes the search trajectory using a step size adjustment strategy. By introducing an entanglement perturbation mechanism based on the difference in optimal process parameters, combined with multidimensional random perturbation to achieve local extremum escape, it ensures that the snake can adaptively adjust when trapped in a local optimum. This method significantly improves the global search capability and local convergence accuracy of process parameter optimization, ultimately outputting the optimal process parameter scheme for the combined roughing and finishing grinding stages.

[0033] In this embodiment, S5 specifically includes: S51. Perform parameter analysis on the optimal process parameter scheme and extract the set values ​​corresponding to each process parameter. S52. Generate a process control instruction matrix based on each set value; S53. Encapsulate the process control instruction matrix according to the industrial fieldbus communication protocol to form a control instruction data frame that can be sent out. S54. The process control command data frame is sent to the processing execution module of the fully automatic coarse and fine grinding machine through the industrial fieldbus to drive the actual coarse and fine grinding process.

[0034] This invention analyzes the optimal process parameter scheme, extracts key process setpoints, and generates a standardized process control command matrix. The command data is encapsulated and transmitted via an industrial fieldbus communication protocol, enabling efficient and synchronous distribution of optimized parameters to the equipment's processing execution module. Based on the received commands, the processing unit precisely adjusts its execution strategies for the roughing and finishing grinding stages, ensuring that the process optimization results are quickly and accurately implemented during actual grinding. This effectively improves system response speed, execution consistency, and processing stability, significantly enhancing the level of processing intelligence and production efficiency.

[0035] In this embodiment, S54 specifically includes: S541. The processing execution module parses the received process control instruction data frame and sets the actual execution parameters of the rough grinding unit and the fine grinding unit. S542. After completing the actual execution parameter setting, control the fully automatic coarse and fine grinding integrated machine to start the coarse grinding and fine grinding process according to the process parameter combination optimized by winding disturbance, and the coarse grinding unit and the fine grinding unit are executed synchronously. S543. During the processing, based on the real-time collected process parameter data and the disturbance search results generated after the snake-like entanglement disturbance, Perform difference calculation: ; in, This represents the cumulative error during the processing. For the m-th process parameter at time The actual collected values, To optimize the perturbation through snake swarm entanglement, the first The process parameter values ​​for the snake in the m-th dimension. The number of dimensions in the process parameter vector. This represents the total processing time. Let be the error weight of the m-th process parameter, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of the snake body, This refers to the continuous physical time during the processing.

[0036] This invention uses a processing execution module to parse received process control commands and precisely set the actual execution parameters for the rough grinding and fine grinding units, ensuring the rapid implementation of optimized solutions. During processing, the cumulative error is calculated based on the difference between real-time collected process parameters and the optimization results of snake-like entanglement disturbance, dynamically monitoring changes in the execution status. By introducing a multi-dimensional error weighting and time integration mechanism, the system can evaluate processing consistency and stability in real time, promptly detect potential abnormal trends, and effectively improve the accuracy maintenance, energy consumption control, and adaptive optimization capabilities of the rough and fine grinding linkage process.

[0037] In this embodiment, S6 specifically includes: S61. During the processing of the fully automatic coarse and fine grinding integrated machine, the actual execution value of each process parameter is collected in real time, and a real-time process parameter vector at the corresponding time point is generated. S62. Compare the real-time collected process parameter vector with the optimal process parameter scheme obtained through snake swarm optimization to generate a process parameter deviation vector. S63. Based on the process parameter deviation vector, accumulate the deviation of each process parameter and multiply it by the dynamic adjustment ratio factor to calculate the dynamic adjustment amount. S64. The dynamic adjustment amount is superimposed on the optimal process parameter scheme obtained by snake swarm optimization to generate a dynamically corrected combination of process parameters. S65. Based on the dynamically corrected combination of process parameters, regenerate the dynamically corrected process control instruction set. S66. The dynamically corrected process control instruction set is sent to the processing execution module of the fully automatic coarse and fine grinding machine in real time via the industrial fieldbus. S67. Control the rough grinding unit and the fine grinding unit to synchronously execute the rough and fine grinding processes according to the dynamically corrected combination of process parameters. S68. Periodically repeat S61 to S67.

[0038] This invention generates a real-time process parameter vector by acquiring process parameters during machining execution in real time. This vector is then dynamically compared with the optimal process parameter scheme optimized by snake swarm optimization to calculate the process parameter deviation and generate a dynamic adjustment amount. Based on this deviation, the system adaptively corrects the process parameter combination, updates control commands in real time, and rapidly sends them to the machining execution module via an industrial fieldbus, driving the rough grinding and fine grinding units to execute synchronously. Through a periodic iterative correction mechanism, the machining process is ensured to remain in an optimal state, significantly improving the dynamic adaptability, machining accuracy, and system stability of the grinding process.

[0039] Example 1: To verify the feasibility of this invention in practice, it was applied to a high-precision turbine disk machining production line of an aerospace parts manufacturing company. The company's existing fully automatic integrated rough and fine grinding machine (model MTX-GMS800) was selected as the test object, with a Siemens 840Dsl control system. In the original production line process, the rough grinding stage used fixed parameters: feed rate 300 mm / min, grinding depth 0.05 mm, grinding wheel linear speed 35 m / s, and grinding pressure 120 N. The fine grinding stage was set with a feed rate of 150 mm / min, grinding depth 0.02 mm, grinding wheel linear speed 40 m / s, and grinding pressure 100 N. The original process suffered from large fluctuations in surface roughness, rapid grinding wheel wear, and thermal deformation during machining leading to decreased dimensional accuracy. To address these issues, the proposed digital twin-based virtual-real collaborative grinding control method was applied to the fully automatic integrated rough and fine grinding machine for system modification.

[0040] After applying the method of this invention, a digital twin of the turbine disk grinding process is first constructed through a remote data acquisition module. The acquired parameters include multi-dimensional data such as spindle load, grinding wheel wear, feed force variation, processing temperature distribution, and surface roughness evolution, establishing a twin model covering processing status, physical field changes, and key characteristic quantities. Next, a long short-term memory network model is used to model the roughing unit, fine grinding unit, and key nodes such as the spindle and grinding wheel as graph structure nodes with temporal characteristics. The input node features are position change, temperature change, and grinding force change data acquired once per second. During training, four months of historical processing data were used for pre-training, and a sliding window method was used for continuous updating.

[0041] During the optimization phase, a snake swarm optimization algorithm was used to search for process parameters. The initial snake swarm size was set to 50, with initial positions distributed within the range of feed rate 200-400 mm / min, grinding depth 0.03-0.07 mm, grinding wheel linear speed 30-45 m / s, and grinding pressure 90-130 N. Through 30 iterations, utilizing a foraging direction sliding search and entanglement perturbation local optimization mechanism, the optimal combination of process parameters was finally obtained: feed rate 328 mm / min, grinding depth 0.046 mm, grinding wheel linear speed 37.5 m / s, and grinding pressure 115 N for the rough grinding stage; and feed rate 165 mm / min, grinding depth 0.018 mm, grinding wheel linear speed 42.2 m / s, and grinding pressure 95 N for the fine grinding stage.

[0042] The optimized process parameters are transmitted to the integrated controller in real time via industrial Ethernet. The system dynamically adjusts the parameters based on real-time feedback and monitors execution stability using a cumulative error integral function. When the execution error exceeds a set threshold of 0.5%, a fine-tuning mechanism is automatically activated to locally correct the process instructions, maintaining dynamic consistency and stability in the processing.

[0043] Table 1. Comparison of the optimization effects of virtual and real collaborative grinding control in a fully automated roughing and finishing grinding machine based on digital twins. Table 1 shows that under the same process conditions, the equipment using the present invention improved processing efficiency by 9.7%, reduced the average surface roughness Ra by 16.5%, reduced the processing size fluctuation by 14.8%, reduced unit energy consumption by 8.3%, extended grinding wheel life by 11.2%, and improved the first-pass yield of finished products by 4.5%.

[0044] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for virtual-real collaborative control of an integrated roughing and finishing grinding machine based on digital twins, characterized in that, The steps include the following: S1. On the integrated rough and fine grinding machine, collect the real-time processing parameters of the rough grinding unit and the fine grinding unit to construct a grinding digital twin that is synchronized with the actual processing process; S2. Based on the grinding digital twin, a long short-term memory network model is constructed to transmit information through the connection relationship between nodes and predict the evolution trend of the state of each unit in the future processing process. S3. Based on the prediction results, optimize the process parameters of the entire process from rough grinding to fine grinding in the grinding digital twin, and output the optimal combination of process parameters obtained by optimization. S4. Based on the optimal combination of process parameters obtained by optimization, the snake swarm optimization algorithm is used to search for the optimal combination of process parameters in the grinding digital twin environment. Each snake is modeled as an individual containing process parameter vectors and adjustment step size. It slides and searches according to the foraging direction and outputs the optimal process parameter scheme for the rough and fine grinding linkage stage. S5. Update the optimal process parameter scheme to the integrated coarse and fine grinding machine in sync, and drive the actual coarse and fine grinding process to be executed. S6. During the actual grinding process, real-time processing parameters are continuously collected to update the grinding digital twin. If a processing abnormality or deviation is detected, re-optimization and correction control will be automatically triggered.

2. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins as described in claim 1, characterized in that, The real-time machining parameters specifically include grinding wheel load, spindle vibration, motor power, grinding force, and workpiece surface temperature.

3. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins as described in claim 1, characterized in that, The specific process parameters include feed rate, grinding depth, grinding wheel linear speed, grinding pressure, spindle speed, grinding wheel wear compensation, coolant flow rate, machining path trajectory, machining time setting, and grinding wheel-workpiece contact length.

4. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S1 specifically includes: S11. Obtain the three-dimensional geometric data of the workpiece. The three-dimensional geometric data is a set of spatial coordinates of each measurement point of the workpiece, including the X coordinate, Y coordinate and Z coordinate of each measurement point. S12. Collect the physical parameters of the coarse and fine grinding machine. The physical parameters include grinding wheel diameter, grinding wheel linear speed, spindle speed and feed speed. S13. Based on the three-dimensional geometric data of the workpiece and the physical parameters of the equipment, a digital twin data model of the grinding process is constructed through a time-synchronized data fusion method. S14. In the digital twin data model of the grinding process, the object state information is dynamically updated through a continuous time update mechanism so that the state of each object in the digital twin is synchronized with the state of the corresponding object in the actual processing process in time. S15. After the object state is synchronized, output the grinding digital twin.

5. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S2 specifically includes: S21. Based on the grinding digital twin, the rough grinding unit, the fine grinding unit, and the key processing components are defined as a set of nodes; S22. Construct a node feature vector for each node. The node feature vector includes the spatial coordinates of the component corresponding to the node, the processing state parameters, and the physical quantity change characteristic values. The processing state parameters include the current feed rate, grinding depth, and spindle speed. The physical quantity change characteristic values ​​include the grinding wheel load change rate, motor power change rate, grinding force change rate, and workpiece surface temperature change rate. S23. Based on the process flow logic and physical connection relationships, establish a set of edge connections between nodes; S24. Combine the set of nodes and the set of edges to form a graph structure, and use the graph structure as input to the long short-term memory network model for modeling. S25. Using a long short-term memory network model, based on the feature vector information propagation mechanism of neighbor nodes, the feature vector of each node in the previous time step is aggregated with the feature vector of neighbor nodes to output the updated node features. S26. Through iterative propagation and node state updates over multiple time steps, predict the future state evolution trend of each unit in the processing process: ； in, Let be the optimal processing state category predicted for the i-th processing unit of the integrated rough and fine grinding machine at time t. To select the category that maximizes the sum of the weighted features of its neighboring nodes among all possible processing state categories, For state categories, For the set of node state categories, The weighted sum of all neighboring units j directly connected to the i-th processing unit is calculated, where i is the i-th processing node in the Long Short-Term Memory network model, j are the neighboring nodes of processing node i, and t is the time step index. Let i be the set of neighboring processing units that have direct information interaction with the i-th processing unit in the process flow. The information propagation weight between the i-th processing unit and its j-th neighboring processing unit. Let be the characteristic state quantity of the j-th neighboring processing unit at time t.

6. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S3 specifically includes: S31. Extract the node status prediction values ​​of the coarse grinding unit and the fine grinding unit within a future predetermined time window. The node status prediction value is the comprehensive processing status of the node at the predetermined time. S32. Based on the predicted state values ​​of each node, determine the main process parameters that affect the processing quality and energy consumption of rough grinding and fine grinding; S33. Perform comprehensive optimization of process parameters for the entire process from rough grinding to fine grinding: ； in, To optimize the obtained optimal combination of process parameters, Let be the vector of process parameters to be optimized. This represents the total number of processing nodes in the integrated rough and fine grinding machine. For the first The current predicted surface roughness of the node. This is the current processing unit node. , , , To optimize the weighting coefficients of the indicators, Let be the target surface roughness of the i-th node. For absolute value operators, Let i be the currently predicted processing time for the i-th node. Let be the target processing time for the i-th node. For a moment Energy consumption per unit time The arithmetic mean roughness, The continuous physical time during the processing. For a moment The change in surface temperature of the workpiece per unit time. This is the minimize operator.

7. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S4 specifically includes: S41. Initialize the snake population in the grinding digital twin, and model each snake as an individual containing the process parameter vector P and the corresponding adjustment step size ΔP; S42. Set an initial position and initial step size for each individual snake. The initial position represents the combination of process parameters, and the initial step size represents the magnitude of a single parameter update. S43. Update the snake's position based on its foraging direction, and guide the snake to glide to the best position in the current snake population. S44. Combine the gliding stride length adjustment strategy and dynamically adjust the stride length according to the current fitness assessment value of the individual snake. S45. Compared to the snake swarm algorithm, the snake swarm optimization method introduces global direction guidance based on the difference in optimal process parameter combinations in the entanglement perturbation mechanism. It also combines multidimensional random perturbations to dynamically adjust the perturbation amplitude in relation to the deviation from the process target. When a snake individual gets trapped in a local optimum, the method adjusts the direction based on the current snake individual's process parameter vector and the optimal process parameter combination P. Perform a perturbation search based on the differences between them: ； in, For the first The perturbation search results generated after the snake coils around the perturbation. This refers to the m-th process parameter in the optimal combination of process parameters. For the first The m-th process parameter value of the snake at its current position. This is the disturbance amplitude control coefficient. The number of dimensions in the process parameter vector. This introduces a small-amplitude random disturbance to the m-th dimension, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of each snake body; S46. Iteratively execute the foraging direction update, step size adjustment and entanglement disturbance process until the maximum number of iterations of 200 is met, and output the optimal process parameter scheme.

8. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S5 specifically includes: S51. Perform parameter analysis on the optimal process parameter scheme and extract the set values ​​corresponding to each process parameter. S52. Generate a process control instruction matrix based on each set value; S53. Encapsulate the process control instruction matrix according to the industrial fieldbus communication protocol to form a control instruction data frame that can be sent out. S54. The process control command data frame is sent to the processing execution module of the coarse and fine grinding integrated machine through the industrial fieldbus to drive the actual coarse and fine grinding process.

9. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 8, characterized in that, Specifically, S54 includes: S541. The processing execution module parses the received process control instruction data frame and sets the actual execution parameters of the rough grinding unit and the fine grinding unit. S542. After completing the actual execution parameter setting, control the coarse and fine grinding integrated machine to start the coarse grinding and fine grinding process according to the process parameter combination optimized by the winding disturbance. The coarse grinding unit and the fine grinding unit are executed synchronously. S543. During the processing, based on the real-time collected process parameter data and the disturbance search results generated after the snake-like entanglement disturbance, Perform difference calculation: ； in, This represents the cumulative error during the processing. For the m-th process parameter at time The actual collected values, To optimize the perturbation through snake swarm entanglement, the first The process parameter values ​​for the snake in the m-th dimension. The number of dimensions in the process parameter vector. This represents the total processing time. Let be the error weight of the m-th process parameter, where m is the index number of the m-th parameter in the process parameter vector. The first in the snake population The number of the snake body, This refers to the continuous physical time during the processing.

10. The virtual-real collaborative control method for an integrated roughing and finishing grinding machine based on digital twins according to claim 1, characterized in that, S6 specifically includes: S61. During the processing of the integrated rough and fine grinding machine, the actual execution value of each process parameter is collected in real time, and a real-time process parameter vector is generated at the corresponding time point. S62. Compare the real-time collected process parameter vector with the optimal process parameter scheme obtained through snake swarm optimization to generate a process parameter deviation vector. S63. Based on the process parameter deviation vector, accumulate the deviation of each process parameter and multiply it by the dynamic adjustment ratio factor to calculate the dynamic adjustment amount. S64. The dynamic adjustment is superimposed on the optimal process parameter scheme obtained by snake swarm optimization to generate a dynamically corrected combination of process parameters. S65. Based on the dynamically corrected combination of process parameters, regenerate the dynamically corrected process control instruction set. S66. The dynamically corrected process control instruction set is sent to the processing execution module of the coarse and fine grinding integrated machine in real time via the industrial fieldbus. S67. Control the rough grinding unit and the fine grinding unit to synchronously execute the rough and fine grinding processes according to the dynamically corrected combination of process parameters. S68. Periodically repeat S61 to S67.

Images