An industrial operation method and system based on digital twin technology
By constructing a collaborative digital twin of geometric physical state, dynamic error, and process quality, and establishing a feedforward-feedback bidirectional closed-loop control, the problems of control lag and difficulty in ensuring process quality in existing technologies are solved, and the whole process optimization of industrial operations and product quality consistency are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-25
- Publication Date
- 2026-06-19
AI Technical Summary
Existing digital twin technology suffers from problems such as control lag, insufficient precision, and difficulty in ensuring process quality in industrial operations, and lacks a solution for intelligent closed-loop operation of the entire process.
A collaborative digital twin of geometric physical state, dynamic error, and process quality is constructed, and a feedforward-feedback two-way closed-loop control mechanism is established. The ideal operation trajectory is generated through the geometric physical state twin, the dynamic error twin is used for error prediction and correction, and the process quality twin is used for quality index adjustment, ultimately generating the optimal control command.
It achieves real-time optimization of the entire process, from equipment motion control to final process quality, improving operational accuracy, adaptability, and product quality consistency.
Smart Images
Figure CN122239467A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial automation control technology, and in particular to an industrial operation method and system based on digital twin technology. Background Technology
[0002] Digital twin technology, as a core enabling technology for the deep integration of cyber-physical systems, has gained widespread attention and application in the industrial manufacturing field in recent years. It enables real-time monitoring, simulation analysis, and optimization decision-making of production processes by constructing virtual mappings of physical entities.
[0003] Existing industrial operation methods based on digital twins can be roughly divided into the following levels according to their functional depth: The first level of methods primarily focuses on the perception and visualization of the state. For example, by deploying sensor networks to collect operational data (such as joint angles, motor current, and temperature) of physical equipment (e.g., industrial robots, CNC machine tools), a three-dimensional geometric model of the equipment is constructed in virtual space, and the state is synchronously mapped and displayed. This type of method can provide operators with an intuitive view of equipment operation, enabling fault alarms and basic data analysis. However, its digital twin model is essentially a passive mirror of the physical entity's state, lacking in-depth analysis and forward prediction capabilities of the physical entity's behavior, and thus unable to form an effective control loop.
[0004] The second level of approach, building upon state awareness, introduces domain-specific simulation and optimization. For example, some solutions utilize digital twins for production line layout simulation, logistics scheduling optimization, or predictive maintenance of equipment. In the control field, some research attempts to use twin models for offline simulation verification of control commands, or to perform hysteresis feedback correction based on the deviation between measured data and model output during operation. However, these methods have significant limitations: firstly, their optimization is mostly offline, based on fixed model open-loop simulations, unable to be dynamically adjusted according to real-time operating conditions during operation; secondly, their control strategies largely rely on traditional PID feedback, which has limited compensation effect and delayed response for complex time-varying dynamic errors caused by thermal deformation, servo hysteresis, mechanical wear, etc.
[0005] More critically, existing technologies generally suffer from a disconnect between geometric and physical models and process quality models. Most solutions focus only on the accuracy of the equipment's motion trajectory or treat process parameters (such as welding current and speed) as fixed inputs. However, in actual industrial operations (such as welding, spraying, and precision assembly), the final work quality (such as weld penetration, coating uniformity, and assembly stress) depends not only on the accuracy of the equipment's motion but also on the dynamically changing process. Existing methods lack an online prediction and optimization closed loop that can connect equipment motion, process, and final quality in real time, making it impossible to proactively adjust work instructions from the perspective of final product quality.
[0006] Therefore, existing digital twin solutions either remain at the monitoring and visualization level or can only achieve localized, lagging optimization control that is decoupled from the final quality target, making it difficult to meet the urgent need for intelligent closed-loop processes of perception, analysis, decision-making, and execution in high-precision and high-reliability industrial operations. Summary of the Invention
[0007] The purpose of this invention is to provide an industrial operation method and system based on digital twin technology, solving the problems of control lag, insufficient accuracy, and difficulty in guaranteeing process quality in existing digital twin technologies. By constructing three collaboratively working digital twins—geometric physical state, dynamic error, and process quality—a feedforward-feedback bidirectional closed-loop control mechanism is established to achieve real-time optimization of the entire process from equipment motion control to final process quality, significantly improving the accuracy, adaptability, and product quality consistency of industrial operations.
[0008] To achieve the above objectives, the present invention provides an industrial operation method based on digital twin technology, comprising the following steps: Step S1: Construct the geometric physical state twin, dynamic error twin, and process quality twin of the physical entity; Step S2: Input the target instruction sequence into the geometric physical state twin to obtain the ideal operation trajectory; input the target instruction sequence into the dynamic error twin to obtain the predicted error data, and correct the ideal operation trajectory based on the predicted error data to generate the preliminary optimization instruction; input the preliminary optimization instruction into the process quality twin to obtain the predicted quality index, and adjust the preliminary optimization instruction based on the predicted quality index to generate the final optimized control instruction. Step S3: Send the final optimization control command to the physical entity to drive the physical entity to perform the job task; Step S4: Collect the actual operation results and actual quality indicators after the physical entity performs the operation task, and update the dynamic error twin using the deviation data between the actual operation results and the ideal operation trajectory, while updating the process quality twin using the actual quality indicators.
[0009] Preferably, in step S1, the geometric-physical state twin is constructed based on a multibody system kinematics model, and its ideal pose of the end effector is obtained by solving the forward kinematics of the robot, specifically: For serial robots, their ideal pose The calculation is as follows: ; in, This indicates a joint described based on DH parameters. To the joint The transformation matrix, Indicates the first instruction in the target instruction sequence Each joint command angle This indicates the total number of joints in the robot.
[0010] Preferably, in step S1, the dynamic error twin is constructed based on a physical mechanism model that integrates thermodynamics and servo dynamics, and its predicted error data... Calculated using the following model: ; ; ; in, Represented in the machine tool coordinate system , , The predicted position error value in the axial direction, Indicates transpose. This represents the thermally induced error term. This represents the servo hysteresis error term. Represents the thermal error coefficient matrix. This represents the temperature field vector of the ball screw. This represents the reference temperature vector of the ball screw. Represents the servo gain matrix. Represents the command velocity vector. This represents the actual response velocity vector simulated by a second-order mass-damped-spring system.
[0011] Preferably, in step S1, the process quality twin is constructed based on a thermo-mechanical coupled multiphysics finite element model and is used to predict the weld penetration depth. It is achieved by solving the following governing equations: ; in, Indicates the density of the workpiece material. Indicates the specific heat capacity of the workpiece material. Indicates the thermal conductivity of the workpiece material. This represents the temperature field within the finite element computational domain. Indicates the simulation time. Indicates the heat source of the welding arc. Represents the vector differential operator; Predicted quality index: melting depth This represents the maximum penetration depth of the molten pool region in the thickness direction.
[0012] Preferably, in step S2, the specific process of inputting the target instruction sequence into the geometric-physical state twin to obtain the ideal operation trajectory is as follows: Perform on the target instruction sequence Spline curve interpolation generates smooth joint space trajectories. : ; in, Indicates the first Each joint at any time The angle of the instruction; Will The input is fed into the forward kinematics model, and the continuous ideal trajectory of the end effector in Cartesian space is calculated. : ; in, Indicates position coordinates, Represents Euler angles.
[0013] Preferably, in step S2, the specific process of inputting the target instruction sequence into the dynamic error twin to obtain prediction error data, and correcting the ideal operation trajectory based on the prediction error data to generate preliminary optimized instructions is as follows: Speed command in the target instruction sequence and position commands Input the dynamic error twin to obtain the predicted error trajectory ; The predicted error trajectory is combined with the ideal operation trajectory to generate a geometrically compensated trajectory. ,in, Position components are Position components and The attitude components are obtained by performing vector addition. The attitude components, after being represented by quaternions, are combined with those derived from... The derived attitude perturbation is obtained by performing quaternion multiplication. Trajectory after geometric compensation Perform inverse kinematics to find the solution that makes the end effector achieve The joint angle sequence, which is the initial optimization instruction. .
[0014] Preferably, in step S2, the preliminary optimization instructions are input into the process quality twin to obtain predicted quality indicators, and the preliminary optimization instructions are adjusted based on these predicted quality indicators to generate the final optimization control instructions. The specific process is as follows: The initial optimization instructions and their corresponding process parameters Input the process quality twin, perform transient simulation, and obtain the predicted melt depth. ; like Does not meet process requirements Then adjust the process parameters according to the following rules: ; in, This indicates the adjusted process parameters. This indicates the process parameters before adjustment. This indicates that the sensitivity coefficient is being adjusted. Indicates the target melting depth. Indicates the lower limit of melting depth. This indicates the upper limit of the melting depth.
[0015] Preferably, in step S4, the specific process of updating the dynamic error twin using the deviation data is as follows: The actual trajectory of the end of a physical entity is measured using a laser tracker. ; Calculate the actual error trajectory The position component of the actual error trajectory is obtained by subtracting the position coordinates of the ideal working trajectory from the position coordinates of the actual trajectory, and its attitude component is obtained by subtracting the attitude angle of the ideal working trajectory from the attitude angle of the actual trajectory. The thermal error coefficient matrix is updated using the recursive least squares method with the actual error trajectory as the observed value. and servo gain matrix .
[0016] Preferably, in step S4, the specific process of updating the process quality twin using actual quality indicators is as follows: After welding is completed, the actual weld penetration depth is measured using an ultrasonic flaw detector. ; Calculate the deviation of the melting depth prediction : ; Using gradient descent, To monitor the signal and update the arc thermal efficiency, the update formula is as follows: ; in, Indicates the learning rate. This indicates the updated arc thermal efficiency parameter value. This indicates the arc thermal efficiency parameter value before the update.
[0017] The present invention also provides an industrial operation system based on digital twin technology, comprising: The model building module is used to construct geometric physical state twins, dynamic error twins, and process quality twins of physical entities; The instruction optimization module, connected to the model building module, is used to input the target instruction sequence into the geometric physical state twin to obtain the ideal operation trajectory, input the target instruction sequence into the dynamic error twin to obtain the predicted error data and correct the ideal operation trajectory based on the data to generate preliminary optimization instructions, and input the preliminary optimization instructions into the process quality twin to obtain the predicted quality index and adjust the preliminary optimization instructions based on the index to generate the final optimized control instructions. The instruction execution module, connected to the instruction optimization module, is used to issue the final optimized control instructions to the physical entity; The model evolution module is connected to the instruction execution module and the model building module respectively. It is used to collect the actual operation results and actual quality indicators after the physical entity performs the operation task, and update the dynamic error twin with the deviation data between the actual operation results and the ideal operation trajectory. At the same time, it updates the process quality twin with the actual quality indicators.
[0018] Therefore, the present invention employs the above-mentioned industrial operation method and system based on digital twin technology, and the beneficial technical effects are as follows: (1) By constructing a dynamic error twin, the present invention can predict and compensate for the dynamic error of the system before the control command is issued, and transform the traditional lag feedback into advanced feedforward compensation, fundamentally overcoming the problem of insufficient accuracy caused by response delay in the existing technology, and improving the accuracy of trajectory tracking and positioning.
[0019] (2) By introducing a process quality twin, the present invention can adjust the control command in real time based on the predicted quality index (such as melting depth), realizing the leap from controlling how the equipment moves to ensuring what quality the operation achieves, ensuring the stability and consistency of the final product quality, and solving the pain point of motion control and process results not being related in the prior art. Attached Figure Description
[0020] Figure 1 This is a flowchart of an industrial operation method based on digital twin technology according to the present invention; Figure 2 This is an architecture diagram of an industrial operation system based on digital twin technology according to the present invention. Detailed Implementation
[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0022] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0023] Example 1 This embodiment uses a high-precision robotic welding system as an example, but the application of the present invention is not limited to this.
[0024] The configuration in this embodiment is as follows: Physical entity: A six-degree-of-freedom industrial welding robot with a welding torch at its end.
[0025] Data acquisition system: including robot controller (for reading joint angles and command speeds), laser tracker mounted on robot base (for measuring the actual trajectory of the end effector), and ultrasonic flaw detector used after welding (for measuring the actual weld penetration).
[0026] Computing platform: An industrial-grade edge computing server that hosts and runs three digital twins.
[0027] like Figure 1 As shown, an industrial operation method based on digital twin technology includes the following steps: Step S1, Model Building Steps.
[0028] This step includes: constructing a geometric physical state twin, a dynamic error twin, and a process quality twin of the physical entity.
[0029] (1) The geometric physical state twin is constructed based on the kinematic model of a multibody system, and the ideal pose of the end effector is obtained by solving the forward kinematics of the robot.
[0030] This twin is used to build a multibody system kinematics model in the MATLAB / Simulink environment based on the DH parameters provided by the robot manufacturer. For the serial robot, its ideal pose is determined. The calculation is as follows: ; in, This represents a description of the joint based on Denavit-Hartenberg (DH) parameters. To the joint The transformation matrix, Indicates the first instruction in the target instruction sequence Each joint command angle This indicates the total number of joints in the robot.
[0031] (2) The dynamic error twin is constructed based on a physical mechanism model that integrates thermodynamics and servo dynamics, and is implemented in the Python / TensorFlow environment. Its predicted error data... Calculated using the following model: ; ; ; in, Represented in the machine tool coordinate system , , The predicted position error value in the axial direction, Indicates transpose. This represents the thermally induced error term. This represents the servo hysteresis error term. Represents the thermal error coefficient matrix. This represents the temperature field vector of the ball screw. This represents the reference temperature vector of the ball screw. Represents the servo gain matrix. Represents the command velocity vector. This represents the actual response velocity vector simulated by a second-order mass-damped-spring system.
[0032] (3) Process quality twins are used to predict weld penetration depth. This twin is a thermo-mechanical coupled multiphysics finite element model built using COMSOL Multiphysics software. This model accurately describes the physical processes of a workpiece (low-carbon steel) under an electric arc heat source. Its core is solving the energy equation: ; in, Indicates the density of the workpiece material. Indicates the specific heat capacity of the workpiece material. Indicates the thermal conductivity of the workpiece material. This represents the temperature field within the finite element computational domain. Indicates the simulation time. Indicates the heat source of the welding arc. Represents the vector differential operator; Predicted quality index: melting depth This represents the maximum penetration depth of the molten pool region in the thickness direction, i.e. (1500℃) Indicates the melting point of the workpiece material.
[0033] Step S2, instruction optimization steps.
[0034] (1) Input the target instruction sequence into the geometric-physical state twin to obtain the ideal operation trajectory. In this embodiment, the target instruction is "weld a 200mm straight weld".
[0035] Perform on the target instruction sequence Spline curve interpolation generates smooth joint space trajectories. : ; in, Indicates the first Each joint at any time The angle of the instruction; Will The input is fed into the forward kinematics model, and the continuous ideal trajectory of the end effector in Cartesian space is calculated. : ; in, Indicates position coordinates, Represents Euler angles.
[0036] (2) Input the target instruction sequence into the dynamic error twin to obtain the prediction error data, and correct the ideal operation trajectory based on the prediction error data to generate preliminary optimization instructions.
[0037] Speed command in the target instruction sequence and position commands Input the dynamic error twin to obtain the predicted error trajectory ; The predicted error trajectory is combined with the ideal operation trajectory to generate a geometrically compensated trajectory. ,in, Position components are Position components and The attitude components are obtained by performing vector addition. The attitude components, after being represented by quaternions, are combined with those derived from... The derived attitude perturbation is obtained by performing quaternion multiplication. Trajectory after geometric compensation Perform inverse kinematics to find the solution that makes the end effector achieve The joint angle sequence, which is the initial optimization instruction. .
[0038] (3) Input the preliminary optimization instructions into the process quality twin to obtain the predicted quality index, and adjust the preliminary optimization instructions based on the predicted quality index to generate the final optimization control instructions; The initial optimization instructions and their corresponding process parameters (Including initial welding current of 250A and welding speed of 10mm / s) Input the process quality twin and perform transient simulation to obtain the predicted penetration depth. =4.8mm; like Does not meet process requirements Then adjust the process parameters according to the following rules: ; in, This indicates the adjusted process parameters. This indicates the process parameters before adjustment. This indicates that the sensitivity coefficient is being adjusted. Indicates the target melting depth. Indicates the lower limit of melting depth. This indicates the upper limit of the melting depth.
[0039] In this embodiment, Set to 5.0mm, tolerance ±0.1mm. This is based on the predicted melt depth. Less than the lower limit of melting depth (4.9mm), the system adjusts according to the above rules. Set to 10, the calculation is as follows It is 252A. Use =252A underwent process simulation again to obtain new results. =5.02mm, which meets the requirements. The system then generates the final optimized control command.
[0040] Step S3, Instruction Execution Steps.
[0041] The final optimized control commands and optimized process parameters are sent to the robot controller and welding machine to drive the physical entity to perform high-precision welding operations.
[0042] Step S4, Model Evolution Steps.
[0043] The system collects the actual operation results and actual quality indicators of the physical entity after it performs the operation task, and updates the dynamic error twin with the deviation data between the actual operation results and the ideal operation trajectory. At the same time, it updates the process quality twin with the actual quality indicators.
[0044] The specific process of updating the dynamic error twin using deviation data is as follows: The actual trajectory of the end of a physical entity is measured using a laser tracker. ; Calculate the actual error trajectory The position component of the actual error trajectory is obtained by subtracting the position coordinates of the ideal working trajectory from the position coordinates of the actual trajectory, and its attitude component is obtained by subtracting the attitude angle of the ideal working trajectory from the attitude angle of the actual trajectory. The thermal error coefficient matrix is updated using the recursive least squares method with the actual error trajectory as the observed value. and servo gain matrix Its parameter update law is: ; ; ; in, This represents the parameter vector to be identified. The regression vector is composed of the ball screw temperature field vector and the command velocity vector. Indicates the forgetting factor, Represents the gain vector. Represents the covariance matrix. Indices representing discrete time series.
[0045] The specific process of updating the process quality twin using actual quality indicators is as follows: After welding is completed, the actual weld penetration depth is measured using an ultrasonic flaw detector. =5.05mm; Calculate the deviation of the melting depth prediction (0.03mm): ; Using gradient descent, To monitor the signal and update the arc thermal efficiency, the update formula is as follows: ; in, This represents the learning rate, which is set to 0.01 in this embodiment. This indicates the updated arc thermal efficiency parameter value. This indicates the arc thermal efficiency parameter value before the update.
[0046] Example 2 like Figure 2 As shown, an industrial operation system based on digital twin technology includes: The model building module is used to construct geometric physical state twins, dynamic error twins, and process quality twins of physical entities; The instruction optimization module, connected to the model building module, is used to input the target instruction sequence into the geometric physical state twin to obtain the ideal operation trajectory, input the target instruction sequence into the dynamic error twin to obtain the predicted error data and correct the ideal operation trajectory based on the data to generate preliminary optimization instructions, and input the preliminary optimization instructions into the process quality twin to obtain the predicted quality index and adjust the preliminary optimization instructions based on the index to generate the final optimized control instructions. The instruction execution module, connected to the instruction optimization module, is used to issue the final optimized control instructions to the physical entity; The model evolution module is connected to the instruction execution module and the model building module respectively. It is used to collect the actual operation results and actual quality indicators after the physical entity performs the operation task, and update the dynamic error twin with the deviation data between the actual operation results and the ideal operation trajectory. At the same time, it updates the process quality twin with the actual quality indicators.
[0047] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.
[0048] Therefore, the present invention employs the above-mentioned industrial operation method and system based on digital twin technology, which can realize real-time feedforward compensation of dynamic errors and online optimization of process quality during industrial operations, effectively improving operation accuracy, adaptability and product quality consistency.
[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. 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 still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. An industrial operation method based on digital twin technology, characterized in that, Includes the following steps: Step S1: Construct the geometric physical state twin, dynamic error twin, and process quality twin of the physical entity; Step S2: Input the target instruction sequence into the geometric physical state twin to obtain the ideal operation trajectory; input the target instruction sequence into the dynamic error twin to obtain the predicted error data, and correct the ideal operation trajectory based on the predicted error data to generate the preliminary optimization instruction; input the preliminary optimization instruction into the process quality twin to obtain the predicted quality index, and adjust the preliminary optimization instruction based on the predicted quality index to generate the final optimized control instruction. Step S3: Send the final optimization control command to the physical entity to drive the physical entity to perform the job task; Step S4: Collect the actual operation results and actual quality indicators after the physical entity performs the operation task, and update the dynamic error twin using the deviation data between the actual operation results and the ideal operation trajectory, while updating the process quality twin using the actual quality indicators.
2. The industrial operation method based on digital twin technology according to claim 1, characterized in that, In step S1, the geometric-physical state twin is constructed based on a multibody system kinematics model. The ideal pose of the end effector is obtained by solving the forward kinematics of the robot, specifically: For serial robots, their ideal pose The calculation is as follows: ； in, This indicates a joint described based on DH parameters. To the joint The transformation matrix, Indicates the first instruction in the target instruction sequence Each joint command angle This indicates the total number of joints in the robot.
3. The industrial operation method based on digital twin technology according to claim 1, characterized in that, In step S1, the dynamic error twin is constructed based on a physical mechanism model that integrates thermodynamics and servo dynamics, and its predicted error data... Calculated using the following model: ； ； ； in, Represented in the machine tool coordinate system , , The predicted position error value in the axial direction, Indicates transpose. This represents the thermally induced error term. This represents the servo hysteresis error term. Represents the thermal error coefficient matrix. This represents the temperature field vector of the ball screw. This represents the reference temperature vector of the ball screw. Represents the servo gain matrix. Represents the command velocity vector. This represents the actual response velocity vector simulated by a second-order mass-damped-spring system.
4. The industrial operation method based on digital twin technology according to claim 3, characterized in that, In step S1, the process quality twin is constructed based on a thermo-mechanical coupled multiphysics finite element model and is used to predict the weld penetration depth. It is achieved by solving the following governing equations: ； in, Indicates the density of the workpiece material. Indicates the specific heat capacity of the workpiece material. Indicates the thermal conductivity of the workpiece material. This represents the temperature field within the finite element computational domain. Indicates the simulation time. Indicates the heat source of the welding arc. Represents the vector differential operator; Predicted quality index: melting depth This represents the maximum penetration depth of the molten pool region in the thickness direction.
5. An industrial operation method based on digital twin technology according to claim 4, characterized in that, In step S2, the specific process of inputting the target instruction sequence into the geometric-physical state twin to obtain the ideal operation trajectory is as follows: Perform on the target instruction sequence Spline curve interpolation generates smooth joint space trajectories. : ； in, Indicates the first Each joint at any time The angle of the instruction; Will The input is fed into the forward kinematics model, and the continuous ideal trajectory of the end effector in Cartesian space is calculated. : ； in, Indicates position coordinates, Represents Euler angles.
6. An industrial operation method based on digital twin technology according to claim 5, characterized in that, In step S2, the target instruction sequence is input into the dynamic error twin to obtain prediction error data, and the ideal operation trajectory is corrected based on the prediction error data to generate preliminary optimized instructions. The specific process is as follows: Speed command in the target instruction sequence and position commands Input the dynamic error twin to obtain the predicted error trajectory ; The predicted error trajectory is combined with the ideal operation trajectory to generate a geometrically compensated trajectory. ,in, Position components are Position components and The attitude components are obtained by performing vector addition. The attitude components, after being represented by quaternions, are combined with those derived from... The derived attitude perturbation is obtained by performing quaternion multiplication. Trajectory after geometric compensation Perform inverse kinematics to find the solution that makes the end effector achieve The joint angle sequence, which is the initial optimization instruction. .
7. An industrial operation method based on digital twin technology according to claim 6, characterized in that, In step S2, the preliminary optimization instructions are input into the process quality twin to obtain the predicted quality index. Based on the predicted quality index, the preliminary optimization instructions are adjusted to generate the final optimization control instructions. The specific process is as follows: The initial optimization instructions and their corresponding process parameters Input the process quality twin, perform transient simulation, and obtain the predicted melt depth. ; like Does not meet process requirements Then adjust the process parameters according to the following rules: ； in, This indicates the adjusted process parameters. This indicates the process parameters before adjustment. This indicates that the sensitivity coefficient is being adjusted. Indicates the target melting depth. Indicates the lower limit of melting depth. This indicates the upper limit of the melting depth.
8. An industrial operation method based on digital twin technology according to claim 7, characterized in that, In step S4, the specific process of updating the dynamic error twin using the deviation data is as follows: The actual trajectory of the end of a physical entity is measured using a laser tracker. ; Calculate the actual error trajectory The position component of the actual error trajectory is obtained by subtracting the position coordinates of the ideal working trajectory from the position coordinates of the actual trajectory, and its attitude component is obtained by subtracting the attitude angle of the ideal working trajectory from the attitude angle of the actual trajectory. The thermal error coefficient matrix is updated using the recursive least squares method with the actual error trajectory as the observed value. and servo gain matrix .
9. An industrial operation method based on digital twin technology according to claim 8, characterized in that, In step S4, the specific process of updating the process quality twin using actual quality indicators is as follows: After welding is completed, the actual weld penetration depth is measured using an ultrasonic flaw detector. ; Calculate the deviation of the melting depth prediction : ； Using gradient descent, To monitor the signal and update the arc thermal efficiency, the update formula is as follows: ； in, Indicates the learning rate. This indicates the updated arc thermal efficiency parameter value. This indicates the arc thermal efficiency parameter value before the update.
10. An industrial operation system based on digital twin technology, characterized in that, include: The model building module is used to construct geometric physical state twins, dynamic error twins, and process quality twins of physical entities; The instruction optimization module, connected to the model building module, is used to input the target instruction sequence into the geometric physical state twin to obtain the ideal operation trajectory, input the target instruction sequence into the dynamic error twin to obtain the predicted error data and correct the ideal operation trajectory based on the data to generate preliminary optimization instructions, and input the preliminary optimization instructions into the process quality twin to obtain the predicted quality index and adjust the preliminary optimization instructions based on the index to generate the final optimized control instructions. The instruction execution module, connected to the instruction optimization module, is used to issue the final optimized control instructions to the physical entity; The model evolution module is connected to the instruction execution module and the model building module respectively. It is used to collect the actual operation results and actual quality indicators after the physical entity performs the operation task, and update the dynamic error twin with the deviation data between the actual operation results and the ideal operation trajectory. At the same time, it updates the process quality twin with the actual quality indicators.