Self-adaptive control system and method for forging and pressing aviation precision forgings

Abstract

The invention relates to the field of metal plastic forming control, and discloses a self-adaptive control system and method for forging and pressing aviation precision forgings, and the system comprises an online dynamic model identification module which identifies a process fingerprint representing the characteristics of a current blank at the initial stage of machining; the target curve self-adaptive regulator is used for regulating a reference energy dissipation rate curve according to the fingerprint and generating a personalized reference curve; the model prediction controller performs rolling optimization by using continuously updated model parameters and taking the reference curve as a tracking target, and outputs an optimal speed instruction; and the event triggering and constraint dynamic modulation module is used for actively tightening the operation constraint applied to the controller before a forging pressure sudden increase event is pre-judged through a parallel monitoring model prediction error. According to the method, through combination of target self-adaption and constraint dynamic modulation, control over the forging and pressing process is achieved, forging pressure overshoot can be effectively restrained, and the forming precision and batch-to-batch quality consistency of forge pieces are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal plastic forming control, in particular to an adaptive control system and method for forging processing of aviation precision forgings. Background Art

[0002] As a core process for precision forming, forging directly impacts the performance and service life of high-end equipment. In the aerospace sector, forgings, such as turbine disks and casings, made from difficult-to-deform materials like titanium alloys and high-temperature alloys, are subject to extremely stringent requirements for microstructure, performance, and dimensional accuracy. Precise control of the forging process is crucial for ensuring consistent batch-to-batch quality.

[0003] At present, the control methods of forging equipment mostly rely on pre-set process programs. The controller performs processing according to a fixed punch speed or forging force curve. This method can complete basic forming tasks when facing ideal and constant processing conditions. However, in actual industrial production, there is a profound contradiction between the rigidity of this control strategy and the high uncertainty of the process. On the one hand, there are unavoidable fluctuations in the initial physical properties of the billets entering the furnace, such as slight differences in chemical composition, uneven heating temperature or heat loss during the transfer process, which will cause its deformation resistance to deviate from the nominal value. The solidified control program cannot perceive and adapt to such differences between billets, resulting in significant fluctuations in the forming quality of the final forgings between different batches.

[0004] On the other hand, even for a single blank, forging itself is a highly nonlinear process with rapidly changing physical properties. In the milliseconds between punch presses, a variety of complex physical and metallurgical phenomena occur simultaneously within the material, including work hardening, dynamic recovery and recrystallization, and deformation thermal effects. These factors cause the material's deformation resistance to exhibit complex dynamic variations at different stages of the machining process. Traditional control methods, even those employing conventional PID closed-loop control, struggle to achieve continuous optimal tracking and regulation of this dynamic process throughout the entire stroke.

[0005] The more severe challenges occur in the final stage of the forging process. When the metal is about to completely fill the mold cavity, the flow space of the material decreases sharply, and its equivalent deformation resistance will increase exponentially, thus generating huge forging pressure in a very short time. The response of existing control systems often has a lag. They can only start to passively adjust after detecting that the pressure exceeds the threshold. However, the pressure increases at this time very quickly, enough to make the actual forging pressure seriously exceed the upper limit of the equipment or mold. This pressure overshoot not only damages the expensive mold and forging press body, shortening their service life, but also seriously affects the final dimensional accuracy and internal structure of the forging. It is a technical problem that has long been urgently needed to be solved in the field of precision forging. Therefore, the development of an advanced control method that can actively adapt to changes in working conditions and predictably avoid process risks is of great significance to enhancing the core competitiveness of my country's high-end forging manufacturing industry. Summary of the Invention

[0006] The technical problem to be solved by the present invention is that the existing forging control method for aviation precision forgings usually adopts a fixed process parameter program, which is difficult to adapt to the changes caused by uncertain factors in the processing process such as the initial state of the billet, lubrication conditions and mold temperature, thereby resulting in low consistency in the quality of the finished forgings and difficulty in optimizing the forging process.

[0007] In order to solve the above technical problems, the present invention provides an adaptive control system for forging processing of aviation precision forgings.

[0008] A first aspect of the present invention provides an adaptive control system for forging processing of aviation precision forgings.

[0009] The adaptive control system includes an online dynamic model identification module, a model prediction controller and an event triggering and constraint dynamic modulation module.

[0010] The online dynamic model identification module is configured to collect real-time forging force, punch velocity, and punch displacement data via connected sensors during the forging process. Based on the collected data, the online dynamic model identification module utilizes a pre-set identification algorithm to periodically and online update the model parameters of a lumped parameter physical model. This lumped parameter physical model is used to characterize the dynamic relationship between forging force, punch velocity, and punch displacement.

[0011] In a specific embodiment, the lumped parameter physical model may have the following form: ; in, is the forging pressure, is the punch displacement, is the punch speed, is the strain rate sensitivity index of the material, and the model parameters include the equivalent stiffness coefficient and the equivalent viscous damping coefficient .

[0012] The model predictive controller is communicatively connected to the online dynamic model identification module and is configured to receive the model parameters output by the online dynamic model identification module and updated in real time. The model predictive controller also stores a preset reference energy dissipation rate curve. Based on the received model parameters and the reference energy dissipation rate curve, the model predictive controller calculates a control instruction for controlling the punch speed by solving a constrained optimization problem within a limited prediction time domain, and outputs the control instruction to the actuator of the forging press.

[0013] In a specific embodiment, the optimization problem solved by the model predictive controller includes the following objective function: ; in, is the index of the current discrete time step; For discrete time steps Calculate the objective function value; is the length of the prediction time domain, To control the length of the time domain; For discrete time steps For the future The predicted value of the energy dissipation rate for each discrete time step; The reference energy dissipation rate curve in the future The target value for each discrete time step; For discrete time steps Calculated for the future The increment of punch velocity to be optimized in discrete time steps; is the weight matrix used to penalize the energy dissipation rate tracking error; is a weight matrix for suppressing the incremental change of the punch speed.

[0014] The event triggering and constraint dynamic modulation module is connected to the online dynamic model identification module and the model prediction controller. The event triggering and constraint dynamic modulation module is configured to calculate the prediction error between the predicted value of the model used by the online dynamic model identification module and the actual value collected by the sensor in real time. The event triggering and constraint dynamic modulation module continuously monitors the changing trend of the prediction error. When it is detected that the prediction error shows a preset abnormal trend (such as continuous unidirectional increase), the event triggering and constraint dynamic modulation module outputs a modulation signal to the model prediction controller. The modulation signal is used to enable it to dynamically adjust the constraints in its optimization problem, such as tightening the maximum forging force constraint. This design enables the system to pre-judge the occurrence of key physical events such as the end of mold filling, and adjust the control strategy in advance to ensure a smooth process.

[0015] Furthermore, the adaptive control system may also include a target curve adaptive regulator. The target curve adaptive regulator is connected to the online dynamic model identification module and the model prediction controller. The target curve adaptive regulator is configured to obtain the initial model parameters identified by the online dynamic model identification module in the initial stage of the forging process, and use the initial model parameters as the process fingerprint of the current workpiece. Subsequently, the target curve adaptive regulator performs mathematical operation and adjustment on a preset reference energy dissipation rate curve according to the process fingerprint to generate a new curve, and uses the newly generated curve as the reference energy dissipation rate curve for use by the model prediction controller. In this way, the control target can be adaptively adjusted according to the initial physical properties of each billet.

[0016] A second aspect of the present invention provides an adaptive control method for forging processing of aviation precision forgings.

[0017] The method comprises the following steps: a) in the initial stage of the forging process, identifying initial model parameters of a lumped parameter physical model using real-time collected forging force, punch velocity, and punch displacement data, and using the initial model parameters as a process fingerprint of the current workpiece; then, performing an adjustment operation on a preset baseline energy dissipation rate curve based on the process fingerprint to generate a reference energy dissipation rate curve; b) updating the model parameters of the lumped parameter physical model in real time during the entire forging process; c) adopting a model predictive control strategy, based on the model parameters continuously updated in step b) and the reference energy dissipation rate curve generated in step a), generating control instructions for controlling the punch speed by rolling solving a constrained optimization problem within a limited prediction time domain; d) While executing step c), the prediction error of the lumped parameter physical model is calculated in parallel, and the changing trend of the prediction error is monitored; when it is monitored that the prediction error shows a preset abnormal trend, the maximum forging force constraint in the optimization problem is tightened.

[0018] Through the above-mentioned technical solutions, the present invention constructs a system and method capable of adaptively controlling the forging process based on the real-time physical characteristics of the process. It no longer relies on fixed process procedures, but instead achieves closed-loop control of the forging process by online identification of real-time working conditions, adaptively adjusting control targets, and adopting a forward-looking optimization control strategy. At the same time, by monitoring the prediction error of the model to predict key events and adjust control constraints, the robustness of the control process is improved. As a result, the present invention can adapt to the initial state differences of different billets and disturbances during the processing process, thereby improving the forming quality and consistency of aviation precision forgings.

[0019] The present invention provides an adaptive control system and method for forging precision forgings for aviation, which has the following beneficial effects: 1. The present invention utilizes an online dynamic model identification module to acquire and update model parameters reflecting the current physical properties of the blank in real time, and utilizes a model predictive controller to control these updated model parameters. Furthermore, a target curve adaptive regulator allows personalized adjustment of the control target at the initial processing stage, enabling the entire control strategy to consistently adapt to the initial state differences of each blank and disturbances during processing. This allows for different batches of forgings to experience a more consistent thermodynamic processing history, contributing to improved quality stability of the final product.

[0020] 2. This invention utilizes the prediction error of the lumped parameter physical model by implementing an event-triggered and constrained dynamic modulation module. This module analyzes the changing trends of the prediction error to predict physical events prone to sudden pressure changes, such as those at the end of mold filling. It proactively and dynamically tightens the operating constraints of the model predictive controller before such events occur. This proactive adjustment mechanism avoids pressure overshoots caused by delayed response in traditional control methods, effectively protecting the mold and equipment and reducing the risk of related forming defects.

[0021] 3. This invention elevates the control objective from traditional kinematic or dynamic quantities to an energy dissipation rate that comprehensively reflects the physical nature of the machining process. Simultaneously, multiple modules within the system, including model identification, target adjustment, predictive control, and constraint modulation, form information synergy. For example, the results of model identification are used for both predictive control and for adjusting the control objective; the errors in model prediction are, in turn, used to modulate control constraints. This multi-level, information-coupled control architecture enables the system to execute a globally optimized control strategy that balances performance and safety, going beyond simple feedback control. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a functional block diagram of an adaptive control system for forging of aviation precision forgings according to an embodiment of the present invention; Figure 2 The present invention is a flowchart of an adaptive control method for forging processing of aviation precision forgings according to an embodiment of the present invention.

[0023] Among them, 10, online dynamic model identification module; 20, target curve adaptive regulator; 30, model prediction controller; 40, event triggering and constraint dynamic modulation module. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0025] Refer to the attached Figure 1 The adaptive control system for forging processing of aviation precision forgings can be deployed in computer equipment or dedicated industrial controllers, and communicated with a forging machine through a data interface.

[0026] The adaptive control system for forging precision aerospace forgings of the present invention can be implemented as one or more software modules running on a controller. The controller can be an industrial personal computer (IPC), an advanced functional module of a programmable logic controller (PLC), or a dedicated control unit equipped with a digital signal processor (DSP). The controller includes a processor, a memory for storing program instructions and data, and an input / output interface for receiving sensor data from the forging press and sending control instructions to the forging press.

[0027] The adaptive control system for forging processing of aviation precision forgings includes: an online dynamic model identification module 10 , a target curve adaptive regulator 20 , a model prediction controller 30 , and an event triggering and constraint dynamic modulation module 40 .

[0028] The input end of the online dynamic model identification module 10 is connected to the sensor configured on the forging machine to receive the forging force collected in real time. , punch speed and punch displacement The output of the online dynamic model identification module 10 is connected to the target curve adaptive regulator 20 and the model prediction controller 30 to provide them with real-time updated model parameter vectors. .

[0029] The input end of the target curve adaptive regulator 20 is connected to the online dynamic model identification module 10 to receive the model parameters of the initial stage of forging. The output end of the target curve adaptive regulator 20 is connected to the model prediction controller 30 to provide it with a reference energy dissipation rate curve generated through adjustment. .

[0030] The input end of the model prediction controller 30 is connected to the online dynamic model identification module 10, the target curve adaptive regulator 20 and the event triggering and constraint dynamic modulation module 40. The output end of the model prediction controller 30 is connected to the actuator of the forging press to output the calculated punch speed control instruction. .

[0031] The input of the event-triggered and constrained dynamic modulation module 40 is connected to the online dynamic model identification module 10 to receive the prediction error of the lumped parameter physical model. The output of the event-triggered and constrained dynamic modulation module 40 is connected to the model predictive controller 30 to send a modulation signal to the model predictive controller 30 when preset conditions are met.

[0032] In this embodiment, the overall working principle of the adaptive control system for forging processing of aviation precision forgings is as follows: The adaptive control system establishes and updates a lumped parameter physical model capable of characterizing the physical characteristics of the current forging process in real time through the online dynamic model identification module 10 .

[0033] In the initial forging stage, the target curve adaptive regulator 20 uses the initial model parameters of the lumped parameter physical model to adjust a baseline energy dissipation rate target to generate a control target that is suitable for the current specific billet.

[0034] Subsequently, throughout the entire machining process, the model predictive controller 30 utilizes the real-time updated lumped parameter physical model and the generated control target to generate punch speed control instructions that meet the physical constraints of the equipment through forward-looking optimization calculations.

[0035] At the same time, the event triggering and constraint dynamic modulation module 40 monitors the predictive performance of the model in parallel. When a critical physical event (such as the end of mold filling) is foreseen, the operating constraints of the model predictive controller 30 are actively adjusted to ensure the smoothness and safety of the processing process.

[0036] Refer to the attached Figure 1 The online dynamic model identification module 10 is used to establish and update in real time a lumped parameter physical model that can characterize the dynamic characteristics of the forging process. The online dynamic model identification module 10 receives the real-time forging force from the sensor. , punch speed and punch displacement The data is taken as input and the updated model parameter vector is periodically calculated and output .

[0037] In a specific embodiment, the online dynamic model identification module 10 uses a lumped parameter physical model to describe the relationship between the forging force and the punch motion state. The mathematical expression of the lumped parameter physical model is as follows: ; in, is the instantaneous forging pressure; is the punch displacement; is the punch speed; is the strain rate sensitivity index determined in advance based on the material properties, It is considered as a fixed constant in a machining process. The model parameters to be identified in the lumped parameter physical model are: equivalent stiffness coefficient and the equivalent viscous damping coefficient .

[0038] Equivalent stiffness coefficient It mainly characterizes the stiffness change effect caused by the increase of geometric constraints when the forging is filled in the mold cavity. It comprehensively characterizes the plastic deformation resistance of the material itself and the friction effect between the forging and the die. The equivalent stiffness coefficient and the equivalent viscous damping coefficient together constitute the model parameter vector , whose expression is: ; To achieve online identification, the online dynamic model identification module 10 uses the recursive least squares (RLS) algorithm with a forgetting factor. In the discretized adaptive control system, for each discrete time step , convert the above lumped parameter physical model into a linear regression form. Defined in discrete time steps The measurement value is , the regression vector It is composed of the motion state data of discrete time steps: ; The online dynamic model identification module 10 updates the model parameter vector by executing the following recursive calculation steps: : Step 1: Calculate the prediction error based on the model parameters at the previous moment : ; Step 2: Calculate the gain vector of the recursive least squares algorithm : ; in, is the covariance matrix of the recursive least squares method, It is the forgetting factor, which ranges from 0 to 1 and is used to adjust the influence of historical data on the current model parameter estimation.

[0039] Step 3: Based on the prediction error and gain vector , update the model parameter vector : ; Step 4: Update the covariance matrix For use in the next discrete time step: ; By repeatedly executing the above recursive calculation in each control cycle, the online dynamic model identification module 10 can continuously output the model parameter vector that reflects the current processing condition and is updated in real time. .

[0040] Refer to the attached Figure 1 The target curve adaptive regulator 20 is used to generate a one-time personalized control target for the forging process according to the initial physical properties of each blank. The target curve adaptive regulator 20 is activated at the initial stage of the forging process.

[0041] The target curve adaptive regulator 20 receives a set of model parameter vectors identified in the initial stage from the online dynamic model identification module 10, and defines the set of model parameter vectors as the process fingerprint of the current blank to be processed. The process fingerprint The comprehensive physical properties of the blank to be processed, such as initial temperature, dimensional tolerance and lubrication status, are quantitatively characterized.

[0042] The target curve adaptive regulator 20 pre-stores a reference energy dissipation rate curve obtained through offline finite element simulation or calibration experiment. The baseline energy dissipation rate curve represents the variation of the energy dissipation rate over time for a forging process completed under ideal and nominal working conditions.

[0043] In obtaining the process fingerprint Then, the target curve adaptive regulator 20 adjusts the target curve according to a preset adjustment function. , for the benchmark energy dissipation rate curve Perform mathematical operations to generate a new reference energy dissipation rate curve specific to the current process step The generation process of the reference energy dissipation rate curve can be described by the following expression: ; in, Based on process fingerprint A calculated adjustment factor. Adjustment function The purpose is to quantify the deviation of the physical characteristics reflected by the process fingerprint into a correction to the reference target. The adjustment function can be implemented in various forms.

[0044] In a specific embodiment, the regulatory factor The calculation method is only the same as the process fingerprint The initial equivalent viscous damping coefficient included in The mathematical expression of process fingerprint is as follows: ; in, From the process fingerprint The initial equivalent viscous damping coefficient value extracted from ; It is a nominal equivalent viscous damping coefficient value representing an ideal working condition. The nominal equivalent viscous damping coefficient value is pre-stored.

[0045] In another embodiment, the regulatory factor The calculation method can comprehensively consider the process fingerprint The initial equivalent stiffness coefficient included in and the initial equivalent viscous damping coefficient The mathematical expression of the adjustment factor can be a linear combination: ; in, From the process fingerprint The initial equivalent stiffness coefficient value extracted from ; is the nominal equivalent stiffness coefficient value representing the ideal working condition; and is a preset weight coefficient used to adjust the degree of influence of stiffness and viscosity, and .

[0046] In yet another embodiment, the adjustment function It can also be achieved through non-analytical methods. For example, a large number of experiments or simulations can be conducted in advance to establish a process fingerprint. To the regulatory factor In actual operation, the target curve adaptive regulator 20 identifies the , the final adjustment factor is determined by looking up the table and performing interpolation operations between adjacent data points in the table.

[0047] After the adjustment is completed, the reference energy dissipation rate curve generated It is transmitted to the model predictive controller 30 and serves as its tracking target in the subsequent entire processing process.

[0048] Refer to the attached Figure 1 The model predictive controller 30 is the core execution unit of the adaptive control system. The model predictive controller 30 is configured to , by solving a constrained finite time domain optimization problem to calculate an optimal punch speed control instruction and output it to the actuator of the forging press.

[0049] The model predictive controller 30 is used at each discrete time step Each time a rolling optimization is performed, the rolling optimization process is based on an optimization problem to be solved within a finite future time window (i.e., the forecast horizon). This optimization problem is defined by an objective function, a forecast model, and a set of constraints.

[0050] The objective function is used to quantify the quality of control performance. In this embodiment, the specific mathematical expression of the objective function is as follows: ; in, is the index of the current discrete time step; is the objective function value; is the length of the prediction time domain, To control the length of the time domain, and ; is the value at the current discrete time step For the future The predicted value of the energy dissipation rate for each discrete time step; The reference energy dissipation rate curve provided by the target curve adaptive regulator 20 is The target value for each discrete time step; is the value to be optimized from the current discrete time step The beginning of the future A sequence of punch velocity increments in discrete time steps; and is a preset, positive-definite weight matrix, the matrix The matrix used to penalize the deviation between the predicted energy dissipation rate and the reference target value is Used to suppress the change range of control instructions to ensure the smoothness of the control process.

[0051] The prediction model is used to calculate the predicted value in the objective function The model prediction controller 30 receives the latest model parameter vector provided by the online dynamic model identification module 10 Based on the current discrete time step The actual measurement status and a candidate control increment sequence The model predictive controller 30 predicts the future by forward integration or iteration of the adaptive control system dynamics. The state sequence of discrete time steps Then, using this state sequence and the model parameter vector , calculate the future forging force sequence Finally, according to the physical definition, the future energy dissipation rate prediction sequence is calculated: ; Constraints define the physical boundaries and process requirements that must be met when solving the optimization problem. These constraints are applied throughout the entire prediction horizon and can include: Punch speed constraint: ; in, and These are the minimum and maximum speeds allowed by the forging press actuator.

[0052] Punch speed change rate constraint: ; The punch speed change rate constraint is used to limit the drastic degree of speed change.

[0053] Forging force constraints: ; in, The maximum forging force allowed by the equipment can be dynamically adjusted by the event triggering and constraint dynamic modulation module 40.

[0054] Terminal displacement constraints: ; in, It is the preset forging end position.

[0055] At each discrete time step In the example, the numerical optimizer in the model predictive controller 30 solves the above optimization problem once. Since the prediction model is linear and the objective function is quadratic in this embodiment, the optimization problem is constructed as a quadratic programming (QP) problem. The numerical optimizer used to solve this QP problem in real time can use mature algorithms in the field, including active-set methods or interior-point methods, to solve and obtain an optimal control increment sequence. .

[0056] According to the rolling horizon control principle, only the first element of the control increment sequence The speed command output to the actuator is: .

[0057] At the next discrete time step , the adaptive control system will collect new sensor data and repeat the above optimization process. Figure 1 The event triggering and constraint dynamic modulation module 40 operates as a parallel monitoring and intervention unit. The event triggering and constraint dynamic modulation module 40 is configured to evaluate the prediction accuracy of the model used by the online dynamic model identification module 10 in real time. Based on the evaluation results, it predicts impending physical events that may cause a sharp increase in forging force, and proactively adjusts the operating constraints of the model predictive controller 30 before the event occurs.

[0058] The event trigger and constraint dynamic modulation module 40 continuously calculates the prediction error of the lumped parameter physical model The prediction error is defined as The actual forging force measured by the sensor The event-triggered and constrained dynamic modulation module 40 utilizes the latest model parameter vector and real-time measurement status The calculated model predicts the forging force The difference between: ; Among them, the model predicts the forging force The calculation formula is: ; in, and These are the latest model parameters received from the online dynamic model identification module 10 at the current moment.

[0059] Event triggering and constraint dynamic modulation module 40 analyzes the prediction error In one embodiment, the analysis of the change trend is performed by performing a fixed length This is achieved by accumulating the prediction error within a sliding time window. Its triggering logic is defined by the following conditions: ; in, is a pre-set, positive trigger threshold. When the absolute value of the cumulative prediction error within the sliding time window exceeds this trigger threshold, it indicates a persistent, unidirectional, systematic deviation in the model prediction. Such deviations are interpreted as precursors to critical physical events such as the impending complete mold filling and material flow obstruction.

[0060] Once the trigger conditions are met, the event trigger and constraint dynamic modulation module 40 outputs a modulation signal to the model predictive controller 30. The modulation signal is used to adjust the forging force constraint in the optimization problem. Dynamic tightening is performed. There are many specific implementation methods for this tightening operation.

[0061] In one embodiment, a fixed scaling factor is used Constraints on the original maximum forging force To reduce: ; in, is a preset constant with a value between 0 and 1. For example, it is 0.05.

[0062] In another more sophisticated embodiment, the tightening amount of the maximum forging force constraint is associated with the magnitude of the prediction error monitored during triggering. The mathematical expression thereof can be: ; in, It is a preset gain coefficient used to map the accumulated sum of prediction errors to the reduction of forging force.

[0063] After triggering and calculating the new maximum forging force constraint Afterwards, the new maximum forging force constraint will be sent to the model predictive controller 30 to replace the original maximum forging force constraint within it. , and is used in all subsequent optimization problems until the forging process is completed.

[0064] The following will refer to the attached Figure 2 , the embodiment of the adaptive control method for forging processing of aviation precision forgings provided by the present invention is described in detail. Figure 2The flowchart of the adaptive control method for forging of aviation precision forgings according to one embodiment of the present invention is shown. The method is executed on a forging device equipped with the adaptive control system of the present invention and includes the following steps: Step S100: System initialization.

[0065] Before a forging task begins, the system initialization step is performed first. The system initialization step includes loading and setting a series of preset parameters, including: strain rate sensitivity index in the lumped parameter physical model ; The prediction time domain length in the objective function of the model predictive controller 30 , control the time domain length , weight matrix and ; The forgetting factor of the recursive least squares algorithm used in the online dynamic model identification module 10 And the fixed length of the sliding time window of the event triggering and constraint dynamic modulation module 40 and trigger threshold At the same time, the baseline energy dissipation rate curve and various original constraint values ​​(maximum forging pressure of equipment and forging end position ) into system memory.

[0066] Step S200: Adaptive adjustment in the initial stage of processing.

[0067] When the forging punch begins to descend and contact the blank, the adaptive control system collects the forging force, punch speed and punch displacement data in real time. The online dynamic model identification module 10 continuously executes the recursive least squares algorithm within the preset initial stage of the forging stroke (for example, the first 5% of the stroke) to identify the model parameter vector representing the initial physical characteristics of the blank and solidify the model parameter vector into the process fingerprint. Subsequently, the target curve adaptive regulator 20 is triggered, and the target curve adaptive regulator 20 adjusts the target curve according to the acquired process fingerprint. and a preset adjustment function (such as one of the above embodiments), the reference energy dissipation rate curve Perform mathematical operations to generate a reference energy dissipation rate curve suitable for the current blank This adaptive adjustment step at the initial stage of the process is performed only once during a complete forging process.

[0068] Step S300: Scrolling execution of the main control loop.

[0069] After completing step S200, the adaptive control system immediately enters the main control loop phase, which is Repeat until the processing end condition is met. , perform the following operations in sequence: First, the online dynamic model identification module 10 uses the latest collected sensor data to perform a recursive calculation of the recursive least squares algorithm to transform the model parameter vector inside the online dynamic model identification module 10 from Updated to .

[0070] Then, the model predictive controller 30 receives the updated model parameter vector , and based on the model parameter vector , generated in step S200 As well as the current operating constraints, the constrained optimization problem within it is solved to calculate the optimal punch speed increment sequence.

[0071] Finally, according to the rolling time domain control principle, only the first element of the punch speed increment sequence is used to calculate the final punch speed control instruction of the current discrete time step and sends the final punch speed control instruction to the actuator of the forging machine.

[0072] Step S400: Parallel event monitoring and constraint modulation.

[0073] This step S400 is performed in parallel with the main control loop in step S300 at each discrete time step. Continuous execution. Event triggering and constraint dynamic modulation module 40 calculates the prediction error at each discrete time step , and updates the cumulative sum of the prediction errors within its internal sliding time window. The event trigger and constraint dynamic modulation module 40 continuously compares the absolute value of the cumulative sum with the preset trigger threshold If at a discrete time step , the absolute value of the accumulated sum exceeds , the event triggers the constraint dynamic modulation module 40 to immediately calculate a new, tightened maximum forging force constraint The new maximum forging force constraint is used to update and overwrite the original maximum forging force constraint stored in the model predictive controller 30 .

[0074] Step S500: processing is completed.

[0075] While executing steps S300 and S400, the adaptive control system continuously determines the current punch displacement. Has the preset forging end position been reached? Once the conditions are met , that is, the forging process is determined to be completed, and the control instructions to the actuator are stopped, and the entire control process ends.

[0076] To further illustrate the present invention, the application process of the adaptive control system and method of the present invention is described below with reference to a specific embodiment. This embodiment describes the complete process of isothermal forging a titanium alloy turbine disk forging using the technical solution of the present invention.

[0077] In a specific application scenario, the adaptive control system of the present invention is used to control a hydraulic press with a maximum forging force of 20,000 kN.

[0078] First, before the start of processing, the system is initialized in step S100. The following parameters are set in the model predictive controller: the strain rate sensitivity index of the material is set to 0.2; the forgetting factor of the recursive least squares algorithm is set to 0.98; the prediction time domain of the model predictive controller 30 For 20 discrete time steps, control the time domain There are 5 discrete time steps; the original maximum forging force constraint Set to 20000 kN; forging end position Set to 100 mm. At the same time, the adaptive control system has pre-stored the benchmark energy dissipation rate curve for the turbine disk forging obtained through simulation. And the model parameters under nominal conditions, where the nominal equivalent viscous damping coefficient is is 50.

[0079] Forging begins, and the process enters the adaptive adjustment phase of the initial processing stage, step S200. The punch descends and contacts the titanium alloy blank, whose initial temperature is slightly lower than the nominal temperature. During this phase, when the punch displacement increases from 0 mm to 5 mm, the online dynamic model identification module 10 identifies the process fingerprint of the blank by using the real-time collected forging force, punch speed, and punch displacement data. Since the blank temperature is low, its deformation resistance is high, and the identified initial equivalent viscous damping coefficient is is 55. The target curve adaptive regulator 20 is then enabled and and Calculate the adjustment factor The target curve adaptive regulator 20 multiplies this adjustment factor by the reference energy dissipation rate curve , generating a new reference energy dissipation rate curve with an overall reduced amplitude and transmits the new reference energy dissipation rate curve to the model predictive controller 30 .

[0080] Subsequently, the control process enters the main control loop of step S300 and the parallel event monitoring of step S400. During the process of the punch displacement moving from 5 mm to 100 mm, the adaptive control system repeats this loop in each control cycle (every 10 milliseconds). The online dynamic model identification module 10 continuously updates the model parameters and To reflect the changes in the physical properties of the material due to work hardening and temperature changes. The model predictive controller 30 uses the latest model parameter vector at each moment and the generated reference energy dissipation rate curve , continuously solve the optimization problem and output the optimal punch speed control instruction To accurately track energy dissipation rate targets.

[0081] When the punch displacement reaches 92 mm, the forging material begins to fill the fine structure of the die in large quantities, causing the actual forging force to begin to show a nonlinear growth faster than the model prediction. This makes the prediction error The event trigger and constraint dynamic modulation module 40 detects that the absolute value of the cumulative sum of the prediction errors exceeds the preset trigger threshold within its sliding time window of 10 discrete time steps. The event trigger and constraint dynamic modulation module 40 is triggered and immediately performs the constraint modulation operation. It is based on a fixed scale factor , calculate a new maximum forging force constraint The new maximum forging force constraint is sent to the model predictive controller 30, replacing its original 20,000 kN constraint.

[0082] Finally, the process enters the end stage of step S500. In the final stroke of the punch displacement from 92 mm to 100 mm, the predicted value of the forging force must meet the more stringent upper limit of 19000 kN when the model predictive controller 30 performs the optimization calculation. This enables the model predictive controller 30 to reduce the output punch speed command in advance, thereby smoothing the rising rate of the forging force when the die is fully filled, effectively avoiding pressure overshoot. When the punch displacement sensor measures When , it is determined that the processing is completed, the output of control instructions is stopped, and the forging process ends.

[0083] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive control system for forging processing of aviation precision forgings, characterized in that: include: An online dynamic model identification module is used to collect forging force, punch speed and punch displacement data in real time during the forging process, and to identify model parameters of a lumped parameter physical model that describes the relationship between forging force, punch speed and punch displacement based on the data; a model prediction controller, configured to receive the model parameters output by the online dynamic model identification module, and calculate and output a control instruction for controlling the punch speed by solving a constrained optimization problem based on the model parameters and a preset reference energy dissipation rate curve; An event triggering and constraint dynamic modulation module is used to monitor the prediction error of the model used by the online dynamic model identification module in real time, and when it is monitored that the prediction error shows a preset abnormal trend, it sends a modulation signal to the model prediction controller to adjust the constraint conditions in the optimization problem; The event triggering and constraint dynamic modulation module is specifically configured as follows: by calculating the cumulative sum of the prediction error in the sliding time window to determine whether the prediction error shows a trend of continuous unidirectional increase, and when the trend meets the trigger condition, sending the modulation signal to the model predictive controller to tighten the maximum forging force constraint in the optimization problem.

2. The adaptive control system for forging of aviation precision forgings according to claim 1 is characterized in that: The system further comprises: The target curve adaptive regulator is used to obtain the initial model parameters identified by the online dynamic model identification module as a process fingerprint in the initial stage of the forging process, and adjust a preset baseline energy dissipation rate curve according to the process fingerprint to generate the reference energy dissipation rate curve used by the model predictive controller.

3. The adaptive control system for forging of aviation precision forgings according to claim 2, characterized in that: The target curve adaptive regulator is specifically configured as follows: The equivalent viscous damping coefficient in the initial model parameters is compared with the nominal equivalent viscous damping coefficient, and an adjustment factor is generated based on the comparison result. The baseline energy dissipation rate curve is then scaled using the adjustment factor to generate the reference energy dissipation rate curve.

4. The adaptive control system for forging of aviation precision forgings according to claim 1, characterized in that: The lumped parameter physical model used by the online dynamic model identification module has the following form: ； in, is the forging pressure, is the punch displacement, is the punch speed, is the strain rate sensitivity index of the material, and the model parameters include the equivalent stiffness coefficient and the equivalent viscous damping coefficient .

5. The adaptive control system for forging of aviation precision forgings according to claim 1, characterized in that: The online dynamic model identification module uses a recursive least square method with a forgetting factor to identify the model parameters online.

6. The adaptive control system for forging of aviation precision forgings according to claim 1, characterized in that: The optimization problem solved by the model predictive controller includes the following objective function: ; in, is the index of the current discrete time step; For discrete time steps Calculate the objective function value; is the length of the prediction time domain, To control the length of the time domain; For discrete time steps For the future The predicted value of the energy dissipation rate for each discrete time step; The reference energy dissipation rate curve in the future The target value for each discrete time step; For discrete time steps Calculated for future The punch velocity increment to be optimized in discrete time steps is the weight matrix used to penalize the energy dissipation rate tracking error; is a weight matrix for suppressing the incremental change of the punch speed.

7. An adaptive control method for forging of aviation precision forgings, based on the system according to any one of claims 1 to 6, characterized in that: The following steps are involved: a) in an initial stage of a forging process, identifying in real time initial model parameters of a lumped parameter physical model as a process fingerprint, and adjusting a baseline energy dissipation rate curve according to the process fingerprint to generate a reference energy dissipation rate curve; b) updating the model parameters of the lumped parameter physical model in real time during the entire forging process; c) employing a model predictive control strategy, based on the model parameters continuously updated in step b) and the reference energy dissipation rate curve generated in step a), to generate a control instruction for controlling the punch speed by rolling solving a constrained optimization problem; d) While executing step c), the prediction error of the model is monitored in parallel, and when it is monitored that the prediction error shows a preset abnormal trend, the maximum forging force constraint in the optimization problem is dynamically tightened.

Images