Forging press adjusting control method and system based on digital twinning

Abstract

The application relates to the technical field of industrial process control and discloses a forging press adjustment control method and system based on digital twinning. The method comprises the following steps: carrying out hierarchical behavior decomposition on a forging operation process to obtain a forging control state space, obtaining a predicted value by driving a digital twin body simulation with sensor data, comparing the predicted value with a measured value to generate a deviation sequence, and inversely calculating and updating adjustment parameters, superimposing a deviation compensation amount, verifying the closed-loop adjustment control instruction through process constraints, and issuing the closed-loop adjustment control instruction to a servo control system according to an execution time sequence for execution. The application solves the problems that the existing forging press control system cannot locate the deviation source of each basic action unit due to the lack of hierarchical behavior decomposition, and the digital twin body cannot directly participate in real-time closed-loop control due to model parameter drift.

Description

Technical Field

[0001] This application relates to the field of industrial process control technology, and in particular to a method and system for regulating and controlling a forging press based on digital twins. Background Technology

[0002] Large free-form forging hydraulic presses undertake the forming and processing of key forgings in aerospace, energy equipment, and shipbuilding industries. Their operation involves the coordinated operation of multiple stages, including rapid slide descent, pressure forging, pressure holding, return stroke, and forging rotation. In current industrial practice, the control system of forging presses generally adopts a single-loop structure based on PID feedback. The measured signals from displacement and pressure sensors are used as feedback quantities, and the PID controller adjusts the opening of the hydraulic servo valve to track preset displacement and pressure target values. In recent years, the application of digital twin technology in the forging field has gradually increased. By constructing a virtual simulation model of the forging press, the simulation model is driven by data collected from sensors for offline optimization of process parameters, equipment status monitoring, and forging quality prediction.

[0003] At the control level, the PID controller treats the entire forging process as a single-layer feedback control, failing to distinguish the sources of motion deviations in different actuators within each stage. When forging pressure deviates, the controller cannot determine whether the deviation is caused by hydraulic system leakage or changes in material deformation resistance, and can only make uniform adjustments to the total output, lacking specificity. At the twin level, existing digital twin applications position the twin as an offline optimization tool or condition monitoring platform. The twin outputs optimized process parameter suggestions or equipment health scores, without forming a real-time closed loop with the controller. When unexpected changes occur in the forging temperature field during forging, causing material constitutive parameters to deviate from the initial set value, the deviation between the twin's simulation output and the actual behavior of the physical equipment gradually accumulates, but this deviation is neither detected online nor fed back to the control loop.

[0004] The fundamental reason why PID controllers cannot distinguish the source of deviation is that the continuous operation process of the forging press is not quantified and decomposed according to the degrees of freedom of the actuators, and there is a lack of fine mapping relationship between the control variables and the physical components. If a hierarchical behavior decomposition is introduced to establish this mapping, although each decomposed basic action unit can be independently set with a control target, its internal intermediate motion state (such as the real-time strain rate during the plastic deformation process of the material) cannot be directly measured by sensors under high temperature and high pressure conditions, which introduces a new problem of the unobservable intermediate state. While using digital twins to perform hierarchical real-time simulation of each basic action unit can fill the measurement blind spot, the parameters of the twin model will drift with the forging passes. If the simulation output is directly used in the control decision without online verification, the model error will be introduced into the control loop, causing systematic deviation. Summary of the Invention

[0005] This application provides a digital twin-based adjustment and control method and system for forging presses, which solves the problems of existing forging press control systems being unable to locate the source of deviation of each basic action unit due to the lack of hierarchical behavior decomposition, and the inability of the digital twin to directly participate in real-time closed-loop control due to model parameter drift.

[0006] In a first aspect, this application provides a digital twin-based adjustment and control method for a forging press, the digital twin-based adjustment and control method for a forging press comprising: The operation process of forging passes is decomposed into hierarchical behavior, and each basic action unit is mapped as a control state variable and constraint boundary to obtain the forging control state space. The simulation sub-modules corresponding to the forging control state space within the digital twin are driven by the real-time data stream from the sensors to obtain the simulation control prediction values ​​of each of the basic action units. The simulated control prediction value and the sensor measured value are compared periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters in the twin are updated by back-calculation using the corresponding calibration data, and the deviation compensation amount is calculated in the opposite direction of deviation growth. The simulated control prediction value and the deviation compensation amount are superimposed and verified by process constraints to obtain the closed-loop regulation control command. The closed-loop regulation and control command is sent to the servo control system of the forging press for execution according to the execution sequence of the basic action unit.

[0007] Secondly, this application provides a digital twin-based forging press adjustment and control system, which includes: The decomposition module is used to decompose the operation process of the forging and pressing stages into hierarchical behavior, and map each basic action unit into control state variables and constraint boundaries to obtain the forging and pressing control state space. The running module is used to drive the simulation sub-modules corresponding to the forging control state space within the digital twin to run by the real-time data stream from the sensor, so as to obtain the simulation control prediction value of each of the basic action units; The verification module is used to compare the simulation control prediction value with the sensor measured value periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters in the twin body are updated by back calculation using the corresponding calibration data, and the deviation compensation amount is calculated in the opposite direction of deviation growth. After superimposing the simulation control prediction value and the deviation compensation amount, the closed-loop regulation control command is obtained through process constraint verification. The execution module is used to send the closed-loop regulation and control command to the servo control system of the forging press for execution according to the execution sequence of the basic action unit.

[0008] Thirdly, a digital twin-based forging press adjustment and control device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the digital twin-based forging press adjustment and control device to execute the aforementioned digital twin-based forging press adjustment and control method.

[0009] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned digital twin-based forging press regulation and control method.

[0010] In the technical solution provided in this application, the forging control state space is obtained by hierarchically decomposing the operation process of the forging press and mapping each basic action unit to control state variables and constraint boundaries. This refines the control object of the forging press from the overall macro output of the machine under the traditional PID method to the basic action unit level control quantity corresponding to each hydraulic actuator and manipulator transmission mechanism. When the control state variable of a certain basic action unit deviates, the physical component and action link where the deviation occurs can be directly locked without making a general adjustment to the entire forging press. This hierarchical structure allows the constraint boundaries of each dimension in the forging control state space to be set independently and without interference. Different basic action units are applicable to different allowable deviation ranges due to differences in process accuracy requirements (such as the tolerance requirements of the material plastic deformation unit for forging pressure being much stricter than the tolerance requirements for the slider displacement in the rapid descent stage). Therefore, the allocation of control resources is more in line with the actual process requirements. The technical approach of using real-time data streams from sensors to drive the operation of various simulation sub-modules within a digital twin to obtain simulation control prediction values ​​transforms the digital twin from an offline optimization tool into an online simulation engine in the real-time control link. Each simulation sub-module operates in a hierarchical structure of behavior decomposition tree, and the simulation output final state of the previous primitive action unit is directly passed to the next primitive action unit as the simulation input initial state. This cascaded transmission mechanism enables the twin to provide continuous simulation prediction outputs in intermediate motion states that cannot be directly measured by sensors (such as the instantaneous strain rate distribution inside the forging during the plastic deformation of materials), thus compensating for the information loss caused by the blind spot of physical measurement in closed-loop control. The mechanism of comparing the simulated control predictions with the sensor measurements periodically to generate a twin deviation sequence and using the TIC coefficient as the deviation metric provides an online quantitative monitoring method for the simulation accuracy of the twin. The TIC coefficient directly reflects the normalized deviation between the simulation and the actual measurement and is not affected by the differences in the physical dimensions of each control state variable. Therefore, a unified judgment can be made using the same deviation threshold of 0.15 for three different dimensions: displacement, velocity, and force, avoiding the parameter inflation problem caused by setting an absolute deviation threshold for each physical quantity separately. When the TIC coefficient exceeds the deviation threshold, the adjustment parameters in the twin are updated by back-calculation using calibration data. Essentially, this involves converting the model parameters in the control equations of the simulation submodule from known coefficients to unknowns and substituting the sensor measurements as true values ​​into the equations for back-solving. This operation is automatically executed at each sliding window update time, ensuring that the constitutive parameters, hydraulic parameters, and kinematic parameters of the twin always track the current real state of the physical equipment, fundamentally suppressing the degradation of simulation accuracy caused by the drift of model parameters as the forging passes progress.

[0011] The technical approach involves calculating the deviation compensation amount in the opposite direction of deviation growth and superimposing it with the simulation control prediction value, followed by process constraint verification to obtain the closed-loop regulation control command. This constructs a dual-channel control command generation structure that combines simulation prediction and deviation compensation. The simulation control prediction value provides a feedforward prediction of the optimal control quantity for each subsequent basic action unit based on the latest parameters from the twin. The deviation compensation amount provides a correction and supplement to the residual error that may exist in the feedforward prediction based on historical deviation trends. The control command resulting from the superposition of both includes model knowledge and data-driven correction in terms of information sources, exhibiting stronger robustness compared to control methods that rely solely on simulation prediction or sensor feedback. In the process constraint verification, the forging pressure, slide speed, and flipping angle are truncated item by item. This ensures that any control quantity exceeding the limit introduced by the deviation compensation amount will be truncated to the safe boundary value range. The truncation operation only applies to the individual component exceeding the limit, not the entire command, ensuring that an anomaly in a single parameter will not disrupt the control flow of the entire forging pass, achieving a reasonable balance between safety and continuity. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of an embodiment of the digital twin-based forging press adjustment and control method in this application. Figure 2 This is a schematic diagram illustrating the consistency verification between the simulated predicted value of forging pressure and the measured value of forging pressure in the embodiments of this application. Figure 3 This is a schematic diagram showing the variation of the weighted average TIC coefficient of each forging pass with the pass number in the embodiments of this application. Detailed Implementation

[0014] This application provides a digital twin-based method and system for regulating and controlling a forging press. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the digital twin-based forging press regulation and control method in this application includes: Step S1: Decompose the hierarchical behavior of the forging process and map each basic action unit into control state variables and constraint boundaries to obtain the forging control state space. Specifically, the hierarchical behavior decomposition targets the complete work cycle of a 185MN free-form forging hydraulic press, from slide start to forging turnover within a single forging pass. A basic motion unit refers to the smallest segment of motion in which one or more hydraulic actuators of the forging press complete a continuous displacement or force output in a single direction of motion. For example, the process of the master cylinder rising from zero pressure to working pressure constitutes an independent basic motion unit. Each basic motion unit is assigned three types of control state variables: displacement corresponds to the spatial stroke of the actuator in the motion segment, velocity corresponds to the motion rate along the stroke, and force corresponds to the hydraulic pressure or operating torque required to drive the motion. The constraint boundaries are directly taken from the finished product dimensional tolerances and equipment safety limits specified in the forging process documents, where the upper and lower allowable deviation limits correspond to the positive and negative limits of the process tolerance, respectively. The set of state variables and constraint boundaries of all basic action units constitutes the forging control state space. The dimension of this space is equal to the total number of basic action units multiplied by the number of state variables of each unit. In this scheme, with nine basic action units and three state variables per unit, the forging control state space is 27-dimensional.

[0016] Step S2: Drive the simulation sub-modules corresponding to the forging control state space within the digital twin to run using the real-time data stream from the sensor, and obtain the simulation control prediction values ​​of each of the basic action units; Specifically, the simulation sub-modules within the digital twin correspond one-to-one with the basic action units in the forging control state space, and each simulation sub-module contains a set of control equations. The sampling period for the real-time sensor data stream is set to 5 milliseconds in this scheme. This value is determined based on the relationship between the maximum movement speed of the 185MN forging press slide during the rapid descent phase (approximately 300 mm / s) and the upper limit of the allowable displacement deviation (1.5 mm). Within a 5-millisecond sampling interval, the maximum displacement increment of the slide is 1.5 mm, which is exactly equal to the upper limit of the allowable deviation. A larger sampling interval would cause the displacement increment within a single cycle to exceed the deviation tolerance, resulting in control lag. After the sensor data enters the digital twin, it is routed to the corresponding simulation sub-module according to the topological relationship of the behavior decomposition tree. Each simulation sub-module independently calculates the simulation control prediction value of its own basic action unit. Simultaneously, adjacent modules maintain state continuity through a final-state-initial-state transfer mechanism; that is, the final displacement, final velocity, and final force values ​​at the end of the previous module's calculation are directly used as the initial input conditions for the next module.

[0017] Step S3: Compare the simulation control prediction value with the sensor measured value periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, use the corresponding calibration data to back-calculate and update the adjustment parameters in the twin body, and calculate the deviation compensation amount in the opposite direction of deviation growth. After superimposing the simulation control prediction value and the deviation compensation amount, and verifying it through process constraints, a closed-loop regulation control command is obtained. Specifically, the TIC coefficient is an abbreviation for Theil's inequality coefficient. Mathematically, it is the root mean square of the difference between the simulation and measured sequences divided by the sum of their respective root mean squares. A TIC coefficient of 0 indicates complete consistency between the simulation and the measurement, while a value of 1 indicates complete incompatibility. The deviation threshold is set to 0.15. This value originates from the precision requirements of the forging process for controlling critical dimensions. When the TIC coefficient reaches 0.15, the deviation between the simulated prediction and the measured value accounts for approximately 15% of the signal amplitude. Under typical forging pressure levels (80MN to 150MN) of a 185MN forging press, the corresponding force deviation is approximately 12MN to 22.5MN. This deviation level is close to the critical tolerance of the forging quality; therefore, 0.15 is used as the threshold for triggering parameter back-calculation and updating. The sliding window length N is set to 20 sampling periods, i.e., a 100-millisecond observation window. This length covers the typical delay time (60 to 80 milliseconds) from command issuance to pressure response stabilization in the hydraulic system of the forging press, ensuring that the calculation of the TIC coefficient includes a complete control response cycle. When the measured values ​​of the sensors in the calibration data pair are substituted into the control equations of the corresponding simulation submodule for back-calculation, the essence of back-calculation is to replace the controlled variables (forging pressure, slider displacement, etc.) in the equations with known quantities, and replace the original model parameters (deformation resistance coefficient, leakage coefficient, etc.) with unknown quantities, obtaining the current true values ​​of the parameters through algebraic solutions. The calculation of the deviation compensation amount adopts the bounded gradient projection method. The projection direction is taken as the opposite direction of the deviation growth trend shown by the twin deviation sequence in the most recent sliding window. The projection result is truncated within the feasible region defined by the constraint boundary to avoid excessive compensation amount causing the control command to go out of bounds.

[0018] Step S4: Send the closed-loop adjustment control command to the servo control system of the forging press according to the execution sequence of the basic action unit; Specifically, closed-loop control commands must pass three process constraint verifications before being issued. The forging pressure verification criterion is that the target forging pressure value in the command must not exceed 166.5MN (the equipment's rated load of 185MN) multiplied by a safety factor of 0.9. The safety factor of 0.9 is based on the structural fatigue safety margin requirements commonly used in the large forging press industry. The slide speed verification criterion is that the slide speed setpoint in the command must not exceed the limit speed corresponding to the maximum flow rate that the hydraulic system can output at the current oil temperature. This limit speed varies with oil temperature and is calculated by the hydraulic system simulation submodule in the twin based on real-time oil temperature data. The tilt angle verification criterion is that the target tilt angle of the manipulator must not exceed the manipulator's mechanical limit value, which is a fixed value directly taken from the equipment nameplate parameters. If any verification item fails, the corresponding command component is truncated to the safety boundary value, rather than the entire command being invalidated. The truncated command is still issued and executed. This approach ensures that an anomaly in a single parameter will not disrupt the control flow of the entire forging pass. After the command is issued, the servo control system continuously receives sensor feedback during the execution of each basic motion unit. When the measured value enters the allowable deviation range of the constraint boundary, it triggers the issuance of the command for the next basic motion unit, forming a unit-by-unit progressive execution logic within the track.

[0019] In one specific embodiment, the forging process is decomposed hierarchically, and each basic action unit is mapped to a control state variable and a constraint boundary to obtain the forging control state space, including: Based on the structural parameters of the forging press and the forging process documents, the complete forging pass is decomposed into the first layer according to the functional stages, resulting in a stage-level action sequence including the rapid descent stage, the pressure forging stage, the pressure holding stage, the return stage, and the forging flipping stage. The action sequence of each stage is decomposed into a second layer according to the motion relationship of the components to obtain the basic action unit corresponding to each stage. Based on the equipment structure parameters, a set of control state variables containing displacement, velocity and force is established for each of the basic action units, and a set of constraint boundaries containing the upper limit and lower limit of allowable deviation for each of the control state variables is established based on the tolerance requirements in the forging process document. The set of control state variables of all the basic action units is combined with the set of constraint boundaries to obtain the forging control state space.

[0020] Specifically, the forging pass operation process refers to the sum of all mechanical movements within the complete cycle of a 185MN free-form forging hydraulic press, from the moment the slide begins its descent to the moment the forging is flipped back into place. The first layer of decomposition is based on functional stage division. That is, according to the sequence of operations specified in the forging process document, the entire pass is divided into five non-overlapping stages on the time axis: the rapid descent stage corresponds to the slide falling freely from its initial height to contact the upper surface of the forging; the pressurizing forging stage corresponds to the main cylinder establishing working pressure and driving the slide to apply plastic deformation to the forging; the pressure holding stage corresponds to the main cylinder maintaining constant pressure to fully deform the forging; the return stage corresponds to the return cylinder driving the slide upward to detach from the forging; and the forging flipping stage corresponds to the manipulator clamping the forging and rotating it to the next forging angle. The second layer of decomposition is based on component motion relationships. That is, within each stage, according to the independent degrees of freedom of the hydraulic actuators or mechanical transmission mechanisms involved in the motion, the stage is further divided into indivisible basic motion units. Taking the rapid descent phase as an example, the main cylinder rapid descent start unit describes the hydraulically driven motion of the main cylinder filling valve opening, oil flowing into the main cylinder cavity, and the slider accelerating from rest. The slider free descent unit describes the motion of the slider falling at a constant or near-constant speed to contact the forging under its own weight and inertia. The physical driving mechanisms of the two are different (the former is hydraulically driven, the latter is gravity-driven), therefore they are decomposed into two independent basic action units. In the pressurization forging phase, the main cylinder pressurization unit corresponds to the gradual opening of the main cylinder servo valve and the transition process of oil pressure climbing from the rapid descent pressure to the working pressure. The material plastic deformation unit corresponds to the process of the slider continuously pressing down after the working pressure is established, causing plastic rheology in the forging. In the return phase, the return cylinder pressurization unit corresponds to the process of pressurizing the return cylinder oil cavity and overcoming the slider's own weight. The slider upward unit corresponds to the process of the slider rising at a constant speed to the initial height under the thrust of the return cylinder. In the forging flipping stage, the manipulator clamping unit corresponds to the process of the manipulator jaws closing and clamping the forging, while the rotary positioning unit corresponds to the process of the manipulator drive motor driving the forging to rotate around the axis to the target angle and then braking and locking it. In the pressure holding stage, since the entire process only involves the single action of maintaining the main cylinder pressure, a second layer of disassembly is not performed, and a pressure holding unit is directly formed.

[0021] The control state variable set for each basic action unit is established as follows: Displacement is taken from the stroke range of the corresponding actuator in the equipment structural parameters. For example, the initial displacement value of the slider free-falling unit is the initial height position value of the slider (2800 mm on a 185MN forging press, referring to the distance from the bottom surface of the slider to the worktable surface), and the target value is the height position value of the slider contacting the upper surface of the forging (depending on the current placement height of the forging, specified for each pass in the forging process document). The initial and target values ​​of velocity are also extracted from the process document. For example, the target velocity value of the material plastic deformation unit is the forging speed specified in the process document (30 mm / s to 80 mm / s in typical processes for large alloy steel forgings). The target force value is taken from the forging force value specified in the process document for that pass (typically between 80MN and 150MN on a 185MN forging press, depending on the forging material and cross-sectional area). The upper and lower limits of the allowable deviations in the constraint boundary set directly correspond to the tolerance requirements in the process document. For example, the upper limit of the allowable deviation of the displacement of the slider free-falling unit is +1.5 mm and the lower limit is -1.5 mm. This value is taken from the accuracy requirement of the end position of the empty stroke of a large free forging. The upper limit of the allowable deviation of the force of the material plastic deformation unit is +5% of the forging pressure value specified in the process document and the lower limit is -5%. This percentage comes from the tolerance of the dimensional consistency of the large forging after forging to the fluctuation of forging pressure. Each of the nine basic motion units has three control state variables: displacement, velocity, and force. Together with the initial value, target value, upper limit of allowable deviation, and lower limit of allowable deviation for each variable, all the data are organized in matrix form to form a 27-dimensional forging control state space. Here, 27 dimensions are the result of multiplying the nine basic motion units by the three state variables of each unit. Each dimension carries the initial value, target value, and constraint boundary information of the variable.

[0022] In one specific embodiment, the simulation sub-modules corresponding to the forging control state space within the digital twin are driven by real-time data streams from sensors to obtain the simulation control prediction values ​​of each of the primitive action units, including: Pressure sensors, displacement sensors, infrared temperature sensors, and angle encoders are respectively deployed on the main cylinder, return cylinder, slide guide rail, forging surface, and manipulator rotating shaft of the forging press. The sensors are synchronously collected at a fixed sampling period to obtain the real-time data stream of the sensors. Based on the correspondence between each of the basic action units and the sensors in the forging control state space, the corresponding sensor data in the real-time data stream of the sensors is used as input boundary conditions and allocated to each of the simulation sub-modules in the digital twin. Each of the input boundary conditions is input into the corresponding simulation submodule, and the simulation output value of the control state variable of the primitive action unit is obtained by the control equation calculation built into the simulation submodule. The simulation sub-modules are cascaded according to the hierarchical structure of the behavior decomposition tree. The final state of the simulation output value of the previous primitive action unit is passed as the initial state of the simulation input of the next primitive action unit to obtain the simulation control prediction value of each primitive action unit.

[0023] Specifically, the acquisition of real-time sensor data streams relies on the coordinated acquisition of five types of sensors on the physical hardware of the forging press. A set of pressure sensors, covering a range of 0 to 200 MPa, is deployed on both the main cylinder and the return cylinder to acquire instantaneous hydraulic pressure values ​​within the cylinder cavity. A magnetostrictive displacement sensor, with a range of 0 to 3000 mm and a resolution of 0.05 mm, is deployed along the stroke direction on the slider guide rail to acquire the current position and speed of the slider (the speed is obtained by dividing the displacement difference between two adjacent sampling periods by the sampling period). An infrared temperature sensor, with a temperature measurement range of 200 to 1300 degrees Celsius, is installed above the forging surface for non-contact acquisition of the forging surface temperature field data. An absolute angle encoder, with a resolution of 0.01 degrees, is installed on the rotating shaft of the manipulator to acquire the real-time rotation angle and angular velocity of the manipulator. The fixed sampling period is set to 5 milliseconds, a value whose engineering basis has been given in the previous implementation section. Five types of sensors are synchronized via industrial Ethernet using the IEEE 1588 precision time protocol, ensuring that the time stamp deviation of each sensor's data at the same sampling moment does not exceed 100 microseconds. All sensor values ​​collected in each sampling period are packaged into a single data frame and output continuously in chronological order, forming a real-time sensor data stream. After the data frame arrives at the digital twin, it is routed according to the pre-established primitive action unit-sensor mapping table in the forging control state space. The main cylinder rapid descent start unit and the main cylinder booster unit receive data from the main cylinder pressure sensor and the slider displacement sensor; the material plastic deformation unit receives data from the slider displacement sensor and the infrared temperature sensor; the pressure maintenance unit receives data from the main cylinder pressure sensor; the return cylinder pressurization unit and the slider upward unit receive data from the return cylinder pressure sensor and the slider displacement sensor; and the manipulator clamping unit and the rotary positioning unit receive data from the angle encoder. Data from the same sensor can be shared by multiple simulation sub-modules. For example, slider displacement sensor data can simultaneously enter four simulation sub-modules: the main cylinder rapid descent start unit, the slider free descent unit, the main cylinder booster unit, and the material plastic deformation unit.

[0024] The control equations built into each simulation submodule are divided into three categories according to the physical characteristics of the basic motion units. The basic motion units related to hydraulic actuation (main cylinder rapid descent start unit, main cylinder boosting unit, pressure maintenance unit, return cylinder pressurization unit, slider upward unit, and slider free downward unit) adopt the hydraulic system flow-pressure dynamic equation. This equation describes the dynamic balance relationship between oil flow rate, cylinder chamber pressure, and piston movement speed. The key parameters in the equation include the equivalent bulk elastic modulus and leakage coefficient. The input is the cylinder pressure value collected by the pressure sensor, and the output is the simulated output value of the slider acceleration, velocity, and displacement of the basic motion unit in the current sampling period. The material plastic deformation unit employs the material constitutive equation, which describes the nonlinear relationship between the deformation resistance and plastic strain of the forging material under specific temperature and strain rate conditions. Key parameters in the equation include the temperature-dependent deformation resistance coefficient and the strain rate sensitivity index. The inputs are the measured values ​​of the slider displacement (used to calculate the reduction and strain rate) and the surface temperature values ​​of the forging collected by the infrared temperature sensor. The outputs are the simulated output values ​​of the forging pressure and the forging deformation in the current sampling period. The rotary positioning unit and the manipulator clamping unit employ the manipulator kinematic equation, which describes the relationship between the output torque of the manipulator drive motor, the moment of inertia of the forging, and the rotational angular acceleration. Key parameters in the equation include the friction coefficient and the transmission clearance value. The input is the real-time angle value collected by the angle encoder, and the outputs are the simulated output values ​​of the flipping angular velocity and the time to reach the target angle. Each simulation submodule independently completes one calculation within each 5-millisecond sampling period, and the calculation result is the simulated output value of the control state variable of the basic unit at the current moment. Cascading transmission occurs at the handover moment between adjacent basic unit motion modules. Specifically, the transmitted physical quantities are the final values ​​of displacement, velocity, and force of the preceding unit's simulation output at the instant the preceding unit's action ends. These three values ​​are directly written into the initial condition register of the subsequent unit's simulation submodule, replacing the original initial value settings of the subsequent unit. In actual forging passes, the basic unit motion modules are strictly arranged sequentially on the time axis; there is no situation where the preceding module has not finished while the following module has started. When the displacement or force in the preceding unit's simulation output value reaches the intersection of the target value and the allowable deviation range, cascading transmission is immediately triggered, and the subsequent unit's simulation submodule immediately starts its calculation. After all nine basic unit motion modules complete the cascading calculation sequentially from left to right according to the behavior decomposition tree, the complete sequence of simulation output values ​​output by each unit in each sampling period constitutes the simulation control prediction value.

[0025] In one specific embodiment, the simulated control predicted value and the sensor measured value are compared periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters within the twin are updated using the corresponding calibration data, including: Within each sampling period, the simulated control prediction value of each of the basic action units is subtracted from the corresponding measured value of the sensor item by item to obtain the single-cycle deviation value of each of the basic action units. The deviation time series is formed by taking the single-cycle deviation values ​​of the most recent N sampling periods in a sliding window manner. The TIC coefficient of each primitive action unit is calculated by taking the root mean square of the difference between the simulation control prediction value series and the sensor measured value series in the deviation time series as the numerator and the sum of the root mean square of the simulation control prediction value series and the root mean square of the sensor measured value series as the denominator. Each TIC coefficient is compared with a preset deviation threshold. The basic action unit whose TIC coefficient exceeds the deviation threshold is marked as a deviation over-limit unit. The sensor measured value and simulation control prediction value corresponding to the deviation over-limit unit at the over-limit moment are recorded to obtain the calibration data pair. The measured values ​​of the sensors in the calibration data pair are substituted into the control equations used in the simulation submodule corresponding to the deviation over-limit unit for back calculation. The original adjustment parameters in the simulation submodule are replaced with the back-calculated parameter values ​​to obtain the updated adjustment parameters.

[0026] Specifically, the calculation of the single-cycle deviation value is triggered at the end of each 5-millisecond sampling period. The calculation object is the difference between the simulated control prediction value and the sensor measured value within the same basic element action unit on the same control state variable dimension. Taking the material plastic deformation unit as an example, the simulated control prediction value of this unit in a certain sampling period includes three components: the simulated value of forging pressure, the simulated value of forging deformation, and the simulated value of slider speed. The sensor measured value includes three corresponding components: the measured value of forging pressure collected by the pressure sensor, the measured value of deformation calculated differentially by the displacement sensor, and the measured value of speed. The item-by-item subtraction is performed between three pairs of similar physical quantities: displacement and displacement, speed and speed, and force and force. Each pair of subtraction produces a scalar deviation value. The three scalar deviation values ​​together constitute the single-cycle deviation value vector of this basic element action unit in the current sampling period. The sliding window length N is set to 20 sampling periods (corresponding to a 100-millisecond time span). The window is updated in a period-by-period sliding manner. At each new sampling period, the deviation value of the earliest period is discarded at the front of the window, and the deviation value of the current period is included at the back of the window, always maintaining exactly 20 sets of single-period deviation values ​​within the window. The TIC coefficient is calculated independently for each control state variable dimension, rather than being calculated for all three dimensions together. Taking the forging pressure dimension as an example, the simulated control prediction values ​​of forging pressure in 20 sampling periods within the window constitute a simulation sequence of length 20, and the measured values ​​of forging pressure from the sensor in 20 sampling periods constitute a measured sequence of length 20. The two are subtracted element by element, and the square of each difference is taken, summed, divided by 20, and then the square root is taken to obtain the root mean square of the difference as the numerator of the TIC coefficient. The square of each element in the simulation sequence is taken, summed, divided by 20, and then the square root is taken to obtain the simulation root mean square. The same operation is performed on the measured sequence to obtain the measured root mean square. The two are added together as the denominator of the TIC coefficient. The numerator divided by the denominator is the TIC coefficient of the basic action unit in the forging pressure dimension. The velocity and displacement dimensions are calculated independently, so each primitive action unit generates three TIC coefficient values ​​at each sliding window update time.

[0027] The deviation threshold comparison is performed on three dimensions. When the TIC coefficient of any dimension exceeds 0.15, the entire basic unit is marked as a deviation-over-limit unit. The engineering consideration for this logic is that there is a physical coupling relationship between the control state variables of the forging press (for example, forging pressure deviation is inevitably accompanied by deformation deviation). Exceeding the limit in a single dimension often means that the internal parameters of the simulation submodule corresponding to the unit have deviated from the true value, and the overall parameters of the submodule need to be updated rather than just correcting a single dimension. At the same time as marking a deviation-over-limit unit, all sensor measured values ​​(not just the measured values ​​of the exceeding dimension) and all simulation control prediction values ​​of that period are recorded for the sampling period that triggered the over-limit. The two sets of data are paired and packaged to form a calibration data pair. The number of sensor measured values ​​included in the calibration data pair depends on the type of sensor associated with the basic unit. The calibration data pair of the material plastic deformation unit includes three quantities: forging pressure measured value, forging surface temperature measured value, and slider displacement measured value. The calibration data pair of the main cylinder rapid descent start unit includes two quantities: cylinder pressure measured value and slider displacement measured value. The mathematical essence of inverse calculation update is to transform the model parameters, which were originally known coefficients in the control equations used by the simulation submodule, into unknowns, and replace the controlled variables (from the simulation control predictions) that were originally outputs with the measured values ​​from the sensors (i.e., the true values) as knowns. Then, the equations are solved algebraically to obtain the current true values ​​of the model parameters. Taking the material plastic deformation element as an example, the forward operation of the material constitutive equation involves inputting the slider displacement (used to calculate strain and strain rate) and the forging temperature, using the deformation resistance coefficient and strain rate sensitivity index as known coefficients, and outputting the simulated forging pressure value. In the inverse calculation, the measured values ​​of forging pressure and temperature from the calibration data pair are substituted into the corresponding positions on the left and right sides of the equation. The deformation resistance coefficient and strain rate sensitivity index are treated as unknowns, and the roots are found through nonlinear equations (at least two sets of equations are needed in the case of two parameters; this scheme uses calibration data from multiple sampling periods within the same sliding window to construct an overdetermined set of equations, which is then solved using the least squares method) to obtain the updated values ​​of these two parameters. Once the updated value is obtained, it is directly written to the parameter register of the simulation submodule, overwriting the original value. The simulation submodule will run with the updated adjustment parameters from the next sampling period, without the need to restart or reinitialize the entire twin.

[0028] Figure 2 This diagram illustrates the consistency verification between the simulated and measured forging pressure values ​​in this embodiment. The horizontal axis represents the measured forging pressure collected by the pressure sensor, and the vertical axis represents the simulated forging pressure value output by the digital twin material plastic deformation unit. The solid line is the ideal consistency line (y=x), the dashed line is the linear regression fitting line for all data points, the gray shading band is the confidence band extending 8MN outwards from the ideal consistency line, and the scatter plots represent the simulated and measured forging pressure data pairs for each sampling period in multiple forging passes. The coefficient of determination for the linear regression is also shown. The value is 0.987, and the vast majority of the scatter points fall within the ±8MN confidence band. A small number of scatter points deviating from the confidence band correspond to the sampling period in which the twin adjustment parameters of the initial forging pass have not yet been updated. As the update mechanism continues to run, these deviations are eliminated in subsequent passes.

[0029] In one specific embodiment, the measured values ​​of the sensors in the calibration data pair are substituted into the control equations used by the simulation submodule corresponding to the deviation exceeding the limit unit for back calculation. The original adjustment parameters in the simulation submodule are replaced with the back-calculated parameter values ​​to obtain updated adjustment parameters, including: When the deviation exceeding the limit unit is a material plastic deformation unit, the measured value of forging pressure and the measured value of forging surface temperature in the calibration data pair are substituted into the material constitutive equation for back calculation to obtain the current deformation resistance coefficient and strain rate sensitivity index. The original corresponding values ​​in the simulation submodule are replaced with the deformation resistance coefficient and the strain rate sensitivity index. When the deviation exceeds the limit unit is the main cylinder fast-descent start unit or the return cylinder pressurization unit, the measured values ​​of the cylinder pressure change rate and the measured values ​​of the slider acceleration in the calibration data are substituted into the hydraulic system flow-pressure dynamic equation for back calculation to obtain the current equivalent bulk elastic modulus and leakage coefficient. The original corresponding values ​​in the simulation submodule are replaced with the equivalent bulk elastic modulus and the leakage coefficient. When the deviation-over-limit unit is a rotary positioning unit, the measured angular velocity value in the calibration data pair is substituted into the kinematic equation of the manipulator for back calculation to obtain the current friction coefficient and transmission clearance value, and the original corresponding value in the simulation submodule is replaced with the friction coefficient and transmission clearance value. The simulation submodule of the deviation-over-limit unit is rerun with the updated adjustment parameters. The TIC coefficient after rerun is compared with the deviation threshold again. If it is lower than the deviation threshold, the correction is confirmed to be effective. If it is higher than the deviation threshold, the deviation-over-limit unit is marked as requiring manual intervention.

[0030] Specifically, the constitutive equation used in the material plastic deformation unit is the Arrhenius-type hyperbolic sine model, which is widely used in high-temperature, large-deformation forging scenarios. Its general form is that stress equals the deformation resistance coefficient multiplied by the strain rate sensitivity index raised to the hyperbolic sine function value. In the equation, the deformation resistance coefficient reflects the material's baseline strength against plastic deformation at a given temperature, while the strain rate sensitivity index reflects the sensitivity of the material's rheological stress to changes in strain rate. These two parameters together determine the actual rheological stress value of the forging material under specific temperature and strain rate conditions. During back-calculation, the measured forging pressure in the calibration data pair is divided by the current compressive cross-sectional area of ​​the forging (calculated from the measured slider displacement and the initial geometric parameters of the forging) to convert it into the actual rheological stress value. Simultaneously, the current strain rate value is obtained by dividing the difference in slider displacement between two adjacent sampling periods by the sampling period and then by the current height of the forging. The temperature conditions are determined by the measured surface temperature of the forging collected by an infrared temperature sensor. At this point, only the deformation resistance coefficient and the strain rate sensitivity index remain as unknowns in the equation. Using calibration data from 20 sampling periods within the same sliding window, each period provides a set of measured ternary equations for rheological stress-strain rate-temperature. Substituting these into the equations forms an overdetermined system of 20 equations and two unknowns. The optimal solutions for the deformation resistance coefficient and strain rate sensitivity index are obtained using the least squares method. The hydraulic system flow-pressure dynamic equation describes the rate of change of oil pressure in the cylinder cavity as equal to the equivalent bulk modulus divided by the effective volume of the cylinder cavity, multiplied by the difference between the net flow rate entering the cavity and the volume change caused by piston movement. In the equation, the net flow rate entering the cavity is determined by the servo valve opening and the oil supply pressure (both quantities are known by the control system). The rate of change of volume caused by piston movement is equal to the effective piston area multiplied by the slider speed (the slider speed is obtained differentially from the displacement sensor). The effective volume of the cylinder cavity is calculated by multiplying the piston area by the current slider position. The cylinder pressure change rate is not directly output by the sensor, but is obtained by performing a first-order backward difference on the pressure values ​​collected by the pressure sensor in two adjacent sampling periods and dividing by the sampling period. Similarly, the slider acceleration is obtained by subtracting the velocity sequence from the displacement sensor again and dividing it by the sampling period. After substituting the known quantities into the equation, only two unknown parameters remain: the equivalent bulk elastic modulus and the leakage coefficient. These are then used to construct an overdetermined system of equations within the sliding window and solved using the least squares method.

[0031] The kinematic equations of the manipulator describe the force as follows: the output torque of the drive motor minus the frictional torque, and then minus the impact torque caused by the transmission backlash, equals the moment of inertia of the forging about its rotational axis multiplied by the angular acceleration. In the equations, the output torque of the drive motor is determined by the product of the motor current and the torque constant (the motor current is recorded in real-time by the control system), the moment of inertia is calculated from the mass and geometric dimensions of the forging (these parameters are extracted from the process documents), and the angular acceleration is obtained by taking the first difference of the angular velocity sequence collected by the angle encoder and dividing it by the sampling period. The frictional torque equals the friction coefficient multiplied by the bearing normal load multiplied by the bearing radius. The normal load equals the component of the forging's weight in the direction of the bearing support reaction force (determined by the forging's mass and attitude angle). Therefore, the frictional torque can be expressed as a linear function of the friction coefficient. The impact torque caused by the transmission backlash only occurs when the manipulator's rotation direction changes. Its magnitude is related to the transmission backlash value and the rate of change of angular velocity. When the angular velocity changes from a positive value to zero and then to a negative value (or in the opposite direction), the impact torque generated by the transmission gear undergoing backlash elimination is proportional to the transmission backlash value. Therefore, during the continuous rotation phase of the manipulator (without direction switching), the equation simplifies to a form containing only the friction coefficient as an unknown, which can be directly solved from a single set of data. In the data at the moment of rotation direction switching, the friction coefficient is known (determined by the solution results from the continuous rotation phase), leaving only the transmission clearance value as an unknown in the equation, which can also be directly solved. This phased solution strategy avoids the underdetermined problem of two unknowns appearing simultaneously in one equation. After all the back-calculated parameter values ​​are written into the parameter register of the corresponding simulation submodule, the simulation submodule reruns a complete simulation with the updated adjustment parameters for all 20 sampling cycles of sensor measured data within the current sliding window. This involves re-inputting the input boundary conditions for each of the 20 cycles, recalculating the simulation output value cycle by cycle, and then recalculating the TIC coefficient using the recalculated simulation output value sequence and the original sensor measured value sequence. If the TIC coefficients in all three control state variable dimensions drop below 0.15, the deviation exceeding the limit marker of the deviation exceeding the limit unit is cleared, and the simulation submodule continues to operate normally with the updated adjustment parameters. If, after rerun, the TIC coefficient for any dimension is still higher than 0.15, it indicates a structural deviation in the simulation model of that unit (rather than simple parameter drift). In this case, the primitive action unit is marked as requiring manual intervention, and an alarm message is sent to the visual interaction layer. The alarm message includes the specific dimension where the deviation exceeds the limit, the current TIC coefficient value, and the key values ​​in the calibration data pair. Primitive action units marked as requiring manual intervention will no longer trigger the automatic back-calculation update process in subsequent sampling cycles. Their simulation submodules will continue to operate with the adjustment parameters updated most recently. In the corresponding closed-loop control command, the component of this unit will switch to direct feedback control based on sensor measurements (i.e., degenerate into a traditional single-loop feedback control mode), without relying on the simulation prediction output of the twin, until the operator manually removes the manual intervention mark after troubleshooting the cause of the fault.

[0032] In one specific embodiment, the deviation compensation amount is calculated in the opposite direction of deviation growth. The simulation control prediction value is then superimposed with the deviation compensation amount and verified under process constraints to obtain a closed-loop regulation control command, including: Based on the updated adjustment parameters, feedforward simulation calculations are performed on each of the basic action units that have not yet been executed in the current forging pass. The final state measured values ​​of the basic action units that have been executed are used as initial conditions to obtain the simulation control prediction values ​​of each subsequent basic action unit. Using the simulation control prediction value as the center point and the upper and lower limits of the allowable deviation in the constraint boundary as the feasible region boundary, the gradient projection value is calculated in the opposite direction of the deviation growth direction indicated by the twin deviation sequence. The gradient projection value is then truncated to the range of the feasible region boundary to obtain the deviation compensation amount. The simulation control prediction value is added to the deviation compensation amount to obtain the command value to be verified; The process constraints of the command values ​​to be verified are verified one by one. The verification items include whether the forging pressure exceeds the safety factor multiple of the rated load of the equipment, whether the slide speed exceeds the limit speed of the hydraulic system, and whether the flipping angle exceeds the mechanical limit of the manipulator. The values ​​of the verification items that fail are truncated to the corresponding safety boundary values ​​to obtain the closed-loop regulation and control command.

[0033] Specifically, the feedforward simulation is triggered during the gap between the completion of the adjustment parameter update and the generation of the closed-loop control command. Unlike real-time drive simulation, the object of feedforward simulation is not the currently executing primitive action unit, but all subsequent primitive action units in the current forging pass that have not yet started execution. The initial conditions of the feedforward simulation do not come from the real-time data stream of the sensors, but from the sensor final-state measured values ​​at the moment the action of the most recently completed primitive action unit ends. For example, when the main cylinder boosting unit in the pressurization stage has just finished executing, the current displacement value recorded by the slider displacement sensor, the current cylinder pressure value recorded by the pressure sensor, and the current velocity value obtained by displacement difference are the three measured final-state values ​​used as the initial condition input for the feedforward simulation of the material plastic deformation unit. Starting from this initial condition, the simulation submodule uses the updated adjustment parameters (deformation resistance coefficient, strain rate sensitivity index, and other parameter values ​​that have been back-calculated and updated) to gradually deduce the simulation output value trajectory of each control state variable of the material plastic deformation unit throughout the entire action duration at a sampling period of 5 milliseconds. This trajectory ends at the target value. The feedforward simulation of subsequent elementary element action units also follows the cascaded transfer rule. That is, the final state simulation value of the material plastic deformation unit is passed to the pressure maintenance unit of the holding unit as the initial condition for its feedforward simulation, and so on until the rotation positioning unit in the forging flipping stage completes the feedforward simulation. The simulation output values ​​of all subsequent elementary element action units at their respective target values ​​constitute the simulation control prediction values ​​obtained by the feedforward simulation. These prediction values ​​are essentially the twin's estimation of the optimal control quantity for each subsequent action link based on the current latest parameter state.

[0034] The direction of deviation growth is extracted through linear regression of the twin deviation sequences within the current sliding window. Specifically, the single-cycle deviation values ​​for each control state variable dimension are taken within 20 sampling periods of the sliding window. A univariate linear regression is performed with the sampling period number as the independent variable and the deviation value as the dependent variable. The slope obtained from the regression represents the trend direction of the deviation over time in that dimension. A positive slope indicates that the simulated value is consistently higher than the measured value and the deviation is increasing; a negative slope indicates that the simulated value is consistently lower than the measured value and the deviation is increasing. The direction of the deviation compensation is the opposite of this slope direction. That is, if the slope is positive, the compensation is in the negative direction (requiring downward correction of the control quantity); if the slope is negative, the compensation is in the positive direction (requiring upward correction of the control quantity). The magnitude of the compensation is equal to the absolute value of the regression slope multiplied by the sliding window length N, and then multiplied by a compensation gain coefficient (set to 0.5 in this scheme; the engineering basis for this value is to avoid excessive compensation causing control oscillations. A gain coefficient of 1.0 means completely offsetting the predicted deviation trend, which would lead to overcompensation due to execution lag in actual hydraulic servo systems. The value of 0.5 ensures that the compensation magnitude is half of the deviation trend, gradually converging within one control response cycle rather than in one step). After the compensation is calculated, the feasible region needs to be truncated. The truncation rule is as follows: the result of adding the simulation control prediction value and the compensation is compared with the upper and lower limits of the allowable deviation in the constraint boundary. If the sum exceeds the target value of the simulation control prediction plus the upper limit of the allowable deviation, the compensation is truncated to the upper limit of the allowable deviation minus the difference between the simulation control prediction value and the target value; if the sum is lower than the target value of the simulation control prediction minus the absolute value of the lower limit of the allowable deviation, the compensation is truncated to the negative absolute value of the lower limit of the allowable deviation plus the negative number of the difference between the simulation control prediction value and the target value. The truncation operation is performed independently on each control state variable dimension, and the truncated compensation amount is the final deviation compensation amount. After the simulation control prediction value and the deviation compensation amount are added dimension by dimension to obtain the command value to be verified, the three process constraint verifications are performed one by one. The safety boundary value for forging pressure verification is 185MN multiplied by a safety factor of 0.9, which equals 166.5MN. This safety factor is set according to the requirements for structural fatigue margin in the industry standard for large forging presses. When the forging pressure component in the command value to be verified exceeds 166.5MN, this component is forcibly truncated to 166.5MN. The safety boundary value for slider speed verification is not a fixed constant, but a limit speed value calculated by the hydraulic system simulation submodule in the twin based on the current hydraulic oil temperature. The hydraulic oil temperature affects the oil viscosity and thus the maximum flow capacity of the servo valve. In the typical working oil temperature range of 40 degrees Celsius to 60 degrees Celsius for a 185MN forging press, the slider limit speed is approximately 280 mm / s to 320 mm / s. When the slider speed component in the command value to be verified exceeds this limit speed, it is truncated to this value.The safety boundary value for overturning angle check is the mechanical limit value of the manipulator, which is ±90° on the manipulator equipped for a 185MN forging press. When the absolute value of the overturning angle component in the instruction value to be checked exceeds 90°, it is truncated to +90° or -90° (depending on the rotation direction). After all three checks are passed or the truncation processing is completed, the final values of each control state variable in the instruction value to be checked constitute the closed-loop regulation control instruction.

[0035] In a specific embodiment, issuing the closed-loop regulation control instruction to the servo control system of the forging press for execution according to the execution sequence of the primitive action units comprises: According to the hierarchical structure of the behavior decomposition tree, decomposing the closed-loop regulation control instruction into sub-instructions corresponding one-to-one to each primitive action unit, and arranging them according to the sequence of the primitive action units in the forging pass to obtain a time-series instruction queue; Issuing the sub-instructions belonging to the hydraulic action category in the time-series instruction queue to the hydraulic servo control system of the forging press, and the hydraulic servo control system adjusts the opening of the servo valve according to the target oil pressure set value and the slider speed set value in the sub-instruction; Issuing the sub-instructions belonging to the overturning action category in the time-series instruction queue to the manipulator control system, and the manipulator control system adjusts the rotating speed of the driving motor and the braking time according to the target overturning angle set value in the sub-instruction; During the execution of the sub-instruction of each primitive action unit, comparing the real-time feedback value of the corresponding sensor with the target set value in the sub-instruction, and when the real-time feedback value enters the allowable deviation range of the constraint boundary, triggering the issuing of the sub-instruction of the next primitive action unit in the time-series instruction queue.

[0036] Specifically, the closed-loop regulation control command, when generated, already includes the target control quantities of all subsequent basic action units in the current forging pass across all control state variable dimensions. The splitting operation extracts the corresponding control quantum sets one by one according to the leaf node order of the basic action units in the behavior decomposition tree, and each subset constitutes a sub-instruction. The data fields of each sub-instruction are determined by the type of the basic action unit. The sub-instructions of hydraulic action units (seven in total: main cylinder rapid descent start unit, slide free downward unit, main cylinder boosting unit, material plastic deformation unit, pressure maintenance unit, return cylinder pressurization unit, and slide upward unit) include two fields: target oil pressure setpoint and slide speed setpoint. The sub-instructions of flipping action units (two: manipulator clamping unit and rotary positioning unit) include two fields: target flipping angle setpoint and clamping force setpoint. The nine sub-instructions, after being split, are arranged in a fixed sequence into a timing instruction queue, namely: main cylinder rapid descent start, slide free descent, main cylinder pressurization, material plastic deformation, pressure maintenance, return cylinder pressurization, slide upward movement, manipulator clamping, and rotational positioning. The head of the queue is the sub-instruction to be executed first, and the tail of the queue is the sub-instruction to be executed last in the current pass. After receiving the hydraulic action sub-instructions, the hydraulic servo control system operates its internal instruction-execution mapping logic as follows: the pressure deviation is obtained by subtracting the target oil pressure setpoint from the real-time cylinder pressure value fed back by the pressure sensor; the speed deviation is obtained by subtracting the target slide speed setpoint from the real-time speed value calculated differentially by the displacement sensor. The two deviations are calculated by the existing PID controller (proportional-integral-derivative controller, an existing configuration of the hydraulic servo system of the forging press) and then superimposed to output the servo valve opening command. The servo valve adjusts the hydraulic oil flow according to the opening command to drive the slide movement. The closed-loop regulation control command in this solution replaces the target setpoint input of the PID controller, rather than replacing the PID controller itself. The PID controller is still responsible for the fast response at the command tracking level. This solution provides a more accurate target setpoint after twin prediction and deviation compensation.

[0037] After receiving a flipping action sub-instruction, the manipulator control system calculates the angle difference from the current angle (feedback from the angle encoder in real time) to the target angle based on the target flipping angle setpoint. Combining this with the forging's moment of inertia (extracted from the process document), a trapezoidal angular velocity curve is planned. The time allocation for the acceleration, constant speed, and deceleration phases is determined by the angle difference and the rated torque of the manipulator's drive motor. The braking moment is set at the end of the deceleration phase, when the remaining distance between the angle encoder feedback value and the target angle equals the transmission clearance value (which has been updated in the aforementioned back-calculation). This lead time ensures that after the motor brakes, the gears undergo a clearance elimination phase, allowing the forging to stop precisely at the target angle position. The unit-by-unit triggering judgment is executed once per sampling cycle. The judgment logic is as follows: the real-time feedback value of the sensor corresponding to the currently executing basic action unit is subtracted from the target setpoint in the sub-instruction, and the absolute value is taken. If this absolute value is simultaneously less than the absolute values ​​of the upper and lower limits of the allowable deviation for the corresponding control state variable in the constraint boundary set, it is determined to be within the allowable deviation range, triggering a dequeue operation in the timing instruction queue, and the next sub-instruction is sent to the corresponding control system. When a basic action unit fails to bring its real-time feedback value within the allowable deviation range during continuous execution, a timeout protection mechanism intervenes. Each basic action unit carries a maximum allowable execution time field in its sub-instruction. The value of this field is equal to the typical action duration of the basic action unit under normal operating conditions multiplied by a timeout coefficient of 1.5 (for example, if the typical duration of a material plastic deformation unit is 3 seconds, then the maximum allowable execution time is 4.5 seconds). When the actual execution time exceeds the maximum allowable execution time and the real-time feedback value still fails to bring it within the allowable deviation range, the current sub-instruction is forcibly terminated, the basic action unit is marked as having timed out, and an alarm message is sent to the visual interaction layer. The timing instruction queue suspends dequeueing, waiting for the operator to confirm the subsequent processing method. If the operator chooses to continue execution, the queue resumes dequeueing and triggers the next sub-instruction. If the operator chooses to terminate the cycle, the queue is cleared and the sub-instructions of the return phase are executed directly to safely return the slider to its position.

[0038] Figure 3This diagram illustrates the variation of the weighted average TIC coefficient of each forging pass in this embodiment of the application, along with the pass number. The horizontal axis represents the forging pass number, and the vertical axis represents the weighted average TIC coefficient calculated by weighting all nine basic motion units of that pass according to their accuracy level. The dashed line indicates a pass-level reliability threshold of 0.10. Dark bars correspond to passes with a weighted average TIC coefficient higher than this threshold (i.e., passes that trigger the full parameter recalibration process), while light bars correspond to passes with a weighted average TIC coefficient lower than this threshold (i.e., passes where the twin parameters maintain their current values ​​and directly proceed to the next pass). A broken line connects the values ​​of each pass to show a decay trend. The diagram shows that the weighted average TIC coefficient of the first pass is approximately 0.178. After six consecutive passes of full parameter recalibration, it drops to 0.088 in the seventh pass and stabilizes below the threshold, indicating that the consistency between the twin's internal adjustment parameters and the physical entity of the forging press gradually converges after iterative correction between passes.

[0039] The forging press adjustment and control method based on digital twins in the embodiments of this application has been described above. The forging press adjustment and control system based on digital twins in the embodiments of this application is described below. One embodiment of the forging press adjustment and control system based on digital twins in the embodiments of this application includes: The decomposition module is used to decompose the operation process of the forging and pressing stages into hierarchical behavior, and map each basic action unit into control state variables and constraint boundaries to obtain the forging and pressing control state space. The running module is used to drive the simulation sub-modules corresponding to the forging control state space within the digital twin to run by the real-time data stream from the sensor, so as to obtain the simulation control prediction value of each of the basic action units; The verification module is used to compare the simulation control prediction value with the sensor measured value periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters in the twin body are updated by back calculation using the corresponding calibration data, and the deviation compensation amount is calculated in the opposite direction of deviation growth. After superimposing the simulation control prediction value and the deviation compensation amount, the closed-loop regulation control command is obtained through process constraint verification. The execution module is used to send the closed-loop regulation and control command to the servo control system of the forging press for execution according to the execution sequence of the basic action unit.

[0040] This invention also provides a digital twin-based forging press adjustment and control device, which can be a server. The digital twin-based forging press adjustment and control device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computing and control capabilities. The memory of the digital twin-based forging press adjustment and control device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the digital twin-based forging press adjustment and control device stores the data corresponding to this embodiment. The network interface of the digital twin-based forging press adjustment and control device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0041] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the digital twin-based forging press regulation and control method.

[0042] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0043] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a digital twin-based forging press adjustment and control device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0044] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A forging press adjustment and control method based on digital twin, characterized in that, The method includes: The forging process flow is decomposed hierarchically, mapping each basic action unit to control state variables and constraint boundaries to obtain the forging control state space. This includes: firstly, decomposing the complete forging process according to functional stages based on the forging press equipment structural parameters and forging process documents, resulting in a stage-level action sequence including a rapid descent stage, a pressurizing stage, a holding stage, a return stage, and a forging flipping stage; secondly, decomposing each stage-level action sequence according to component motion relationships, resulting in basic action units corresponding to each stage; establishing a set of control state variables including displacement, velocity, and force for each basic action unit based on the equipment structural parameters, and establishing a set of constraint boundaries including the upper and lower limits of allowable deviations for each control state variable based on the tolerance requirements in the forging process documents; and finally, combining the set of control state variables of all basic action units with the set of constraint boundaries to obtain the forging control state space. The simulation sub-modules within the digital twin, corresponding to the forging control state space, are driven by real-time sensor data streams to obtain the simulation control prediction values ​​of each basic action unit. This includes: deploying pressure sensors, displacement sensors, infrared temperature sensors, and angle encoders on the main cylinder, return cylinder, slide guide, forging surface, and rotating shaft of the forging press, respectively; synchronously acquiring data from each sensor at a fixed sampling period to obtain the real-time sensor data stream; based on the correspondence between each basic action unit and the sensor in the forging control state space, allocating the corresponding sensor data from the real-time sensor data stream as input boundary conditions to each simulation sub-module within the digital twin; inputting each input boundary condition into the corresponding simulation sub-module, and performing control equation calculations built into the simulation sub-module to obtain the simulation output value of the control state variable of the basic action unit; The simulation sub-modules are cascaded according to the hierarchical structure of the behavior decomposition tree. The final state of the simulation output value of the previous primitive action unit is passed as the initial state of the simulation input of the next primitive action unit to obtain the simulation control prediction value of each primitive action unit. The simulated control prediction value and the sensor measured value are compared periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters in the twin are updated by back-calculation using the corresponding calibration data, and the deviation compensation amount is calculated in the opposite direction of deviation growth. The simulated control prediction value and the deviation compensation amount are superimposed and verified by process constraints to obtain the closed-loop regulation control command. The closed-loop regulation and control command is sent to the servo control system of the forging press for execution according to the execution sequence of the basic action unit.

2. The forging press adjustment and control method based on digital twin according to claim 1, characterized in that, The process of comparing the simulated control prediction value with the sensor measured value periodically to generate a twin deviation sequence, and when the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, using the corresponding calibration data to back-calculate and update the adjustment parameters within the twin, includes: Within each sampling period, the simulated control prediction value of each of the basic action units is subtracted from the corresponding measured value of the sensor item by item to obtain the single-cycle deviation value of each of the basic action units. The deviation time series is formed by taking the single-cycle deviation values ​​of the most recent N sampling periods in a sliding window manner. The TIC coefficient of each primitive action unit is calculated by taking the root mean square of the difference between the simulation control prediction value series and the sensor measured value series in the deviation time series as the numerator and the sum of the root mean square of the simulation control prediction value series and the root mean square of the sensor measured value series as the denominator. Each TIC coefficient is compared with a preset deviation threshold. The basic action unit whose TIC coefficient exceeds the deviation threshold is marked as a deviation over-limit unit. The sensor measured value and simulation control prediction value corresponding to the deviation over-limit unit at the over-limit moment are recorded to obtain the calibration data pair. The measured values ​​of the sensors in the calibration data pair are substituted into the control equations used in the simulation submodule corresponding to the deviation over-limit unit for back calculation. The original adjustment parameters in the simulation submodule are replaced with the back-calculated parameter values ​​to obtain the updated adjustment parameters.

3. The forging press adjustment and control method based on digital twin according to claim 2, characterized in that, The step involves substituting the measured sensor values ​​from the calibration data pair into the control equations used in the simulation submodule corresponding to the deviation exceeding the limit unit for back calculation. The back-calculated parameter values ​​are then used to replace the original adjustment parameters in the simulation submodule, resulting in updated adjustment parameters, including: When the deviation exceeding the limit unit is a material plastic deformation unit, the measured value of forging pressure and the measured value of forging surface temperature in the calibration data pair are substituted into the material constitutive equation for back calculation to obtain the current deformation resistance coefficient and strain rate sensitivity index. The original corresponding values ​​in the simulation submodule are replaced with the deformation resistance coefficient and the strain rate sensitivity index. When the deviation exceeds the limit unit is the main cylinder fast-descent start unit or the return cylinder pressurization unit, the measured values ​​of the cylinder pressure change rate and the measured values ​​of the slider acceleration in the calibration data are substituted into the hydraulic system flow-pressure dynamic equation for back calculation to obtain the current equivalent bulk elastic modulus and leakage coefficient. The original corresponding values ​​in the simulation submodule are replaced with the equivalent bulk elastic modulus and the leakage coefficient. When the deviation-over-limit unit is a rotary positioning unit, the measured angular velocity value in the calibration data pair is substituted into the kinematic equation of the manipulator for back calculation to obtain the current friction coefficient and transmission clearance value, and the original corresponding value in the simulation submodule is replaced with the friction coefficient and transmission clearance value. The simulation submodule of the deviation-over-limit unit is rerun with the updated adjustment parameters. The TIC coefficient after rerun is compared with the deviation threshold again. If it is lower than the deviation threshold, the correction is confirmed to be effective. If it is higher than the deviation threshold, the deviation-over-limit unit is marked as requiring manual intervention.

4. The forging press adjustment and control method based on digital twin according to claim 3, characterized in that, The step of calculating the deviation compensation amount in the opposite direction of deviation growth, superimposing the simulation control prediction value with the deviation compensation amount, and then verifying it through process constraints yields a closed-loop regulation control command, including: Based on the updated adjustment parameters, feedforward simulation calculations are performed on each of the basic action units that have not yet been executed in the current forging pass. The final state measured values ​​of the basic action units that have been executed are used as initial conditions to obtain the simulation control prediction values ​​of each subsequent basic action unit. Using the simulation control prediction value as the center point and the upper and lower limits of the allowable deviation in the constraint boundary as the feasible region boundary, the gradient projection value is calculated in the opposite direction of the deviation growth direction indicated by the twin deviation sequence. The gradient projection value is then truncated to the range of the feasible region boundary to obtain the deviation compensation amount. The simulation control prediction value is added to the deviation compensation amount to obtain the command value to be verified; The process constraints of the command values ​​to be verified are verified one by one. The verification items include whether the forging pressure exceeds the safety factor multiple of the rated load of the equipment, whether the slide speed exceeds the limit speed of the hydraulic system, and whether the flipping angle exceeds the mechanical limit of the manipulator. The values ​​of the verification items that fail are truncated to the corresponding safety boundary values ​​to obtain the closed-loop regulation and control command.

5. The forging press adjustment and control method based on digital twin according to claim 1, characterized in that, The step of sending the closed-loop adjustment control command to the servo control system of the forging press according to the execution sequence of the basic action unit includes: Based on the hierarchical structure of the behavior decomposition tree, the closed-loop regulation control instruction is decomposed into sub-instructions corresponding to each of the basic action units, and arranged according to the order of the basic action units in the forging and pressing passes to obtain the timing instruction queue. The sub-instructions belonging to the hydraulic action category in the timing instruction queue are sent to the hydraulic servo control system of the forging press. The hydraulic servo control system adjusts the opening of the servo valve according to the target oil pressure setting value and the slide speed setting value in the sub-instruction. The sub-instructions belonging to the flipping action category in the timing instruction queue are sent to the manipulator control system, which then adjusts the drive motor speed and braking timing according to the target flipping angle setting value in the sub-instruction. During the execution of sub-instructions of each of the primitive action units, the real-time feedback value of the corresponding sensor is compared with the target setting value in the sub-instruction. When the real-time feedback value enters the allowable deviation range of the constraint boundary, the next sub-instruction of the primitive action unit in the timing instruction queue is triggered.

6. A digital twin-based adjustment and control system for a forging press, characterized in that, For implementing the digital twin-based forging press regulation and control method as described in any one of claims 1-5, the digital twin-based forging press regulation and control system comprises: The decomposition module is used to hierarchically decompose the operation process of forging passes, mapping each basic action unit to control state variables and constraint boundaries to obtain the forging control state space. This includes: firstly, decomposing the complete forging pass according to functional stages based on the forging press equipment structural parameters and forging process documents, obtaining a stage-level action sequence including a rapid descent stage, a pressurizing stage, a holding stage, a return stage, and a forging flipping stage; secondly, decomposing each stage-level action sequence according to component motion relationships, obtaining the basic action units corresponding to each stage; establishing a set of control state variables including displacement, velocity, and force for each basic action unit based on the equipment structural parameters, and establishing a set of constraint boundaries including the upper and lower limits of allowable deviations for each control state variable based on the tolerance requirements in the forging process documents; and finally, combining the set of control state variables of all basic action units with the set of constraint boundaries to obtain the forging control state space. The execution module is used to drive the operation of each simulation sub-module corresponding to the forging control state space within the digital twin, driven by the real-time data stream from the sensors, to obtain the simulation control prediction values ​​of each basic action unit. This includes: deploying pressure sensors, displacement sensors, infrared temperature sensors, and angle encoders on the main cylinder, return cylinder, slide guide, forging surface, and rotating shaft of the forging press, respectively; synchronously acquiring data from each sensor at a fixed sampling period to obtain the real-time data stream from the sensors; allocating the corresponding sensor data from the real-time data stream to each simulation sub-module within the digital twin as input boundary conditions according to the correspondence between each basic action unit and the sensors in the forging control state space; inputting each input boundary condition into the corresponding simulation sub-module, and obtaining the simulation output value of the control state variable of the basic action unit through the control equations built into the simulation sub-module; cascading each simulation sub-module according to the hierarchical structure of the behavior decomposition tree, passing the final state of the simulation output value of the previous basic action unit as the initial state of the simulation input of the next basic action unit, to obtain the simulation control prediction values ​​of each basic action unit. The verification module is used to compare the simulation control prediction value with the sensor measured value periodically to generate a twin deviation sequence. When the TIC coefficient in the twin deviation sequence exceeds the deviation threshold, the adjustment parameters in the twin body are updated by back calculation using the corresponding calibration data, and the deviation compensation amount is calculated in the opposite direction of deviation growth. After superimposing the simulation control prediction value and the deviation compensation amount, the closed-loop regulation control command is obtained through process constraint verification. The execution module is used to send the closed-loop regulation and control command to the servo control system of the forging press for execution according to the execution sequence of the basic action unit.

7. A forging press adjustment and control device based on digital twin, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the digital twin-based forging press adjustment and control method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the digital twin-based forging press regulation and control method as described in any one of claims 1 to 5.

Images