Ring piece precision control method and system of numerical control full-automatic radial-axial ring rolling machine

Abstract

The application belongs to the technical field of automatic control, and particularly relates to a ring piece precision control method and system for a numerical control full-automatic radial-axial ring rolling machine, which comprises the following steps: discretizing single-piece rolling time length, and initializing a feedforward compensation sequence of a first ring piece; collecting an error sequence, a resistance sequence and an initial resistance average of a kth ring piece rolling process; dividing a full-process control beat into a high-gradient sensitive area and a low-gradient stable area; in a k+1th ring piece rolling start-up section, calculating a ratio of the initial resistance average of the current piece to that of the previous piece, linearly compensating a learning increment of the high-gradient sensitive area, keeping a learning increment of the low-gradient stable area updated according to a standard law, obtaining an adaptive feedforward amount of the k+1th ring piece, and then obtaining a final control current. The application overcomes the compensation mismatch problem caused by blank temperature fluctuation in batch production, accurately senses the working condition difference and performs time domain correction, can inhibit ring piece wall thickness oscillation, and improves forming quality.

Description

Technical Field

[0001] This invention belongs to the field of automatic control technology, specifically relating to a method and system for controlling the accuracy of rings in a fully automatic CNC radial axial rolling machine. Background Technology

[0002] Radial and axial ring rolling is a key pressure processing technology used to manufacture high-performance ring forgings, widely applied in aerospace, energy, and transportation industries. In the mass production of precision components such as aerospace flange rings, ensuring dimensional consistency and uniform circumferential wall thickness across batches is a core process objective. As the manufacturing industry transforms towards high-end and intelligent manufacturing, the industry places higher demands on the CNC precision and dynamic stability of ring rolling machines during continuous processing. Especially when facing fluctuations in raw material conditions, ensuring long-term consistency in processing quality has become a critical requirement for improving product yield.

[0003] To improve trajectory tracking accuracy in the ring rolling process, the most commonly used traditional technique in the industry is to introduce iterative learning control. This technique fully utilizes the periodic characteristics of ring rolling, collecting and analyzing the feed speed error of the previous ring piece to correct the current control command. This historical learning approach, under ideal conditions of completely constant operating conditions, can effectively compensate for the fixed response time delay of the hydraulic servo system and repetitive mechanical load disturbances, thereby gradually bringing the feed trajectory of subsequent workpieces closer to the command curve and improving the accuracy of single-piece machining.

[0004] However, traditional iterative learning control has serious limitations in actual mass production environments. Since billets are sequentially discharged from the furnace, significant temperature differences often exist between adjacent billets due to furnace temperature uniformity or discharge cycle time. This temperature fluctuation directly alters the metal's deformation resistance, causing the equivalent gain of the hydraulic feed system to drift. Existing technologies lack online awareness of the current batch's operating conditions, mechanically inheriting historical feedforward data, leading to a severe physical mismatch between the feedforward compensation and the actual hardness of the billet. This mismatch not only fails to eliminate the deviation but also introduces directional feed deviations between batches, causing repeated oscillations in the ring wall thickness, resulting in poor product dimensional consistency and making it difficult to meet the requirements of high-precision mass production. Summary of the Invention

[0005] This invention provides a method and system for controlling the accuracy of ring parts in a fully automatic radial and axial ring rolling machine under CNC control, in order to solve the technical problem that the feedforward compensation amount is mismatched with the actual working conditions due to the temperature difference between adjacent billets exiting the furnace in actual mass production, causing repeated oscillations in the wall thickness difference of the ring parts in the circumferential direction.

[0006] In a first aspect, the present invention provides a method for controlling the accuracy of rings in a fully automatic radial-axial ring rolling machine, comprising the following steps: The rolling time of a single piece is discretized according to the control cycle, and the feedforward compensation sequence of the first ring piece is initialized; In the During the rolling process, the feed speed error and feed resistance of each control cycle are collected synchronously to obtain the error sequence, resistance sequence and initial resistance mean; the first gradient of the feedforward compensation sequence between adjacent control cycles is calculated, and the arithmetic mean of the absolute value of the gradient is defined as the partition threshold to divide the entire control cycle into a high gradient sensitive area and a low gradient stable area. In the The feed resistance in the starting section of the rolling mill is collected and the current average initial resistance is calculated. The ratio of the average initial resistance of the current piece to that of the previous piece is calculated to obtain the inter-batch resistance ratio. Based on the inter-batch resistance ratio, linear compensation is performed on the learning increment in the high-gradient sensitive region, while the learning increment in the low-gradient stable region is updated according to the standard law, thus obtaining the first... The adaptive feedforward quantity of the component loop; The adaptive feedforward quantity is superimposed with the real-time feedback quantity to obtain the final control current applied to the hydraulic servo valve, so as to eliminate the wall thickness deviation caused by batch-to-batch temperature fluctuations through the final control current.

[0007] Its effect is as follows: By implementing inter-batch working condition sensing and sensitive zone control, this invention maps the macroscopic difference in billet exit temperature into microscopic control current step size compensation. This method enables the generated adaptive feedforward quantity to accurately match the actual hardness of the current billet, eliminating feedback control lag and feedforward control overshoot caused by working condition drift, significantly improving the dynamic tracking accuracy of CNC ring rolling machines during continuous processing, and achieving efficient elimination of wall thickness deviation.

[0008] Furthermore, the method for obtaining the initial resistance mean includes: obtaining the first... The measured duration of the first revolution after the start of the rolling mill is determined; based on the measured duration, the total number of control cycles corresponding to the first five rotation cycles after the start of the rolling mill is determined as the initial resistance acquisition window; the feed resistance sequence within the initial resistance acquisition window is arithmetically averaged to obtain the average initial resistance value.

[0009] Furthermore, the method for calculating the feed rate error includes: obtaining the real-time displacement of the hydraulic cylinder piston at the end of the current control cycle, and calculating the ratio of the difference between the real-time displacement and the displacement at the end of the previous control cycle to the sampling period to obtain the actual feed rate; using the difference between the commanded feed rate stored in the CNC system and the actual feed rate as the feed rate error; wherein, a positive feed rate error indicates that the current pressing amount is insufficient, and a negative feed rate error indicates that the current pressing amount is too large.

[0010] Furthermore, the method for obtaining the feed resistance includes: collecting the real-time oil pressure of the hydraulic cylinder in each control cycle; and using the product of the real-time oil pressure and the effective working area of ​​the hydraulic cylinder as the feed resistance. The magnitude of the feed resistance is negatively correlated with the local temperature of the billet and is used to characterize the metal deformation resistance.

[0011] Furthermore, the inter-batch resistance ratio and the speed gain of the hydraulic feed system satisfy an inverse relationship: when the inter-batch resistance ratio is greater than 1, it is determined that the equivalent speed gain of the current piece is lower than that of the previous piece, and the learning increment of the high gradient sensitive area is amplified to compensate for the attenuation of the feed correction amount; when the inter-batch resistance ratio is less than 1, it is determined that the equivalent speed gain of the current piece is higher than that of the previous piece, and the learning increment of the high gradient sensitive area is contracted to suppress the feed correction overshoot.

[0012] Its effect is as follows: based on the inter-batch resistance ratio, the learning increment of the high gradient sensitive area is linearly compensated. This invention establishes a proportional mapping between the feedback compensation amount and the physical load in the region where control efficiency is most sensitive. In this way, when the billet temperature decreases and the deformation resistance increases, the current increment can be automatically amplified to offset the attenuation of the actuator response, preventing the tendency of the ring to be too thick due to insufficient feed, and ensuring the control stability under complex dynamic processes.

[0013] Furthermore, we obtain the first The adaptive feedforward amount of the component includes: in the first... After the rolling start-up phase of the workpiece is completed, the adaptive feedforward quantity is calculated once and stored in the CNC system cache for the remaining control cycles; for the control cycles within the start-up phase, the first cycle is directly used. The feedforward compensation sequence of the component loop is used as the adaptive feedforward quantity.

[0014] Furthermore, the acquisition of learning increments also involves learning gains. Identification: Multiple sets of different candidate learning gain values ​​are preset and applied in the production process of the first three consecutive ring parts; the root mean square value of the feed rate error after the end of each ring part is calculated, and the candidate learning gain value corresponding to the minimum root mean square value of the third ring part is determined as the target learning gain.

[0015] The effect is that by pre-setting candidate gains and selecting them based on minimizing the root mean square value, this method achieves the optimal match between the learning gain and specific mechanical properties and billet material. This technical approach avoids the randomness and instability of manual parameter tuning, ensuring that the system can achieve the optimal error convergence rate with minimal tuning costs, providing a highly reliable control parameter benchmark for subsequent mass production.

[0016] Furthermore, in the After the rolling of the ring part is completed, the control method also includes: calculating the root mean square value of the feed speed error during the entire rolling process of the current part, and comparing the root mean square value with the preset convergence threshold; if the root mean square value is less than the preset convergence threshold, the current adaptive feedforward amount is frozen for use in the production of subsequent ring parts of the same specification.

[0017] Its effect is as follows: by determining convergence and freezing the feedforward compensation sequence after the error reaches the target, this scheme effectively filters out the interference of high-frequency noise from the sensor on the control commands. This freezing mechanism prevents unnecessary fluctuations in the feedforward amount caused by minor disturbances during the stable production stage, reduces the computational load of the CNC system while ensuring machining accuracy, and extends the service life of the hydraulic servo actuator under stable operating conditions.

[0018] Furthermore, after the adaptive feedforward quantity is frozen, the control method also includes a warning restart step: continuously monitoring the inter-batch resistance ratio of each subsequent component's start-up segment; calculating the sliding standard deviation of the initial resistance mean of the most recent preset number of batches; and in response to the deviation between the initial resistance mean and the sliding mean of the current component being greater than twice the sliding standard deviation, unfreezing the state and re-executing the adaptive update iteration.

[0019] Its effectiveness lies in the following: by monitoring the sliding standard deviation of wall thickness error in real time and executing early warning and restart logic, this solution endows the system with long-term self-healing and adaptive capabilities. When faced with sudden changes in operating conditions such as material batch switching or abrupt environmental changes, the system can autonomously identify the failure of the control sequence and restart the learning process, ensuring that the control system always has the ability to find the optimal processing state in complex industrial production environments.

[0020] Secondly, the present invention provides a precision control system for ring parts of a fully automatic radial axial ring rolling machine, including a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned precision control method for ring parts of a fully automatic radial axial ring rolling machine is realized.

[0021] The beneficial effects are as follows: This invention achieves adaptive correction of the control step size by quantitatively sensing the inter-batch resistance ratio in the rolling start-up stage and combining it with the gradient division of the feedforward sequence into sensitive zones. This scheme can capture the fluctuation of metal deformation resistance caused by the temperature control deviation of the heating furnace in real time and convert it into a linear compensation factor for the equivalent gain of the hydraulic feed system. The technical means solve the problem of working condition mismatch in the feedforward quantity inheritance process of traditional iterative learning, so that the feed correction quantity maintains physical consistency under different temperature environments, thereby significantly suppressing the oscillation phenomenon of ring wall thickness in mass production and ensuring the dimensional accuracy and consistency of large-scale aerospace flange rings and other components. Attached Figure Description

[0022] Figure 1This is a flowchart of the method for controlling the accuracy of rings in a fully automatic CNC radial and axial ring rolling machine.

[0023] Figure 2 This is a comparison chart showing the effect of controlling the thickness difference of the circumferential wall of the ring component.

[0024] Figure 3 The feed rate error convergence curve and feedforward sequence iteration diagram are shown. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] An embodiment of the method for controlling the accuracy of rings in a fully automatic radial and axial ring rolling machine provided by this invention: like Figure 1 As shown, the method for controlling the accuracy of rings in a fully automatic CNC radial and axial ring rolling machine includes the following steps: S1 discretizes the rolling time of a single piece according to the control cycle and initializes the feedforward compensation sequence of the first ring piece.

[0027] Specifically, the CNC system uses a fixed sampling period as the control cycle and divides the pre-set rolling time for a single piece into equal parts. At the start of the production task, for the first piece of ring to be processed, since there is no inheritable auxiliary correction data, all control points in its feedforward compensation sequence are initialized to zero values. That is, during the processing of the first piece, the feed is driven solely by the basic feedback control law.

[0028] In this step, the rolling time per piece is... With sampling period Satisfying the relation:

[0029] In the formula, The total number of control cycles included in the entire rolling process for a single piece is a positive integer. The preset total rolling time for a single piece is preferably 120 seconds in the aviation flange ring scenario; The sampling period is preferably set to 50 milliseconds.

[0030] Understandably, by cutting the continuous rolling process into... The discrete time nodes provide a data carrier for subsequent error tracking and precise updates of feedforward quantities at each specific moment, ensuring the closed loop of control logic in the time domain.

[0031] In this way, the definition of the initial state of the production task and the division of the time-domain space were completed, establishing a unified reference index for the subsequent transmission of control data between various components.

[0032] S2, in the During the rolling process, the feed speed error and feed resistance of each control cycle are collected synchronously to obtain the error sequence, resistance sequence and initial resistance mean. The first gradient of the feedforward compensation sequence between adjacent control cycles is calculated, and the arithmetic mean of the absolute value of the gradient is defined as the partition threshold to divide the entire control cycle into a high gradient sensitive region and a low gradient stable region.

[0033] Specifically, to achieve a quantitative description of the dynamic characteristics of the rolling process, it is necessary to obtain the kinematic characteristics representing the feed state and the dynamic characteristics representing the deformation resistance. The calculation method for the feed speed error satisfies the following relationship:

[0034] In the formula, For the first The ring component in the first The feed rate error of each control cycle; The preferred feed rate for the instruction stored in the CNC system, in the case of aerospace flange ring rolling, is 0.5 mm / s to 2.0 mm / s; For the first The real-time displacement of the component at the end of the current control cycle; This is the displacement at the end of the previous control cycle; The sampling period is 50ms, with the preferred value being 50ms.

[0035] Understandably, by obtaining the actual feed rate through the first-order difference of the real-time displacement and then subtracting it from the preset commanded feed rate, the motion deviation caused by external load fluctuations under the current cycle time can be accurately extracted. A positive feed rate error indicates that the current compression is insufficient, while a negative feed rate error indicates that the current compression is excessive.

[0036] Subsequently, in order to sense the change in the deformation resistance of the billet, the method for obtaining the feed resistance satisfies the following relationship:

[0037] In the formula, For the first The ring component in the first The feed resistance of each control cycle; The real-time oil pressure of the hydraulic cylinder is collected by a pressure sensor; The effective working area of ​​the hydraulic cylinder is preferably 0.03 square meters.

[0038] Understandably, the magnitude of the feed resistance is negatively correlated with the local temperature of the billet, directly characterizing the metal's resistance to deformation. When the billet temperature decreases, the metal hardness increases, and the feed resistance increases accordingly. This physical quantity can be linearly mapped by multiplying the oil pressure by the area of ​​action.

[0039] Furthermore, in order to establish a benchmark for comparing operating conditions, the method for obtaining the initial average resistance includes: obtaining the first... The measured duration of the first revolution after the start of the rolling mill is determined; based on the measured duration, the total number of control cycles corresponding to the first five rotational cycles after the start of the rolling mill is determined, which serves as the initial resistance acquisition window; the feed resistance sequence within the initial resistance acquisition window is arithmetically averaged to obtain the average initial resistance. The average initial resistance satisfies the following relationship:

[0040] In the formula, For the first The average initial resistance of the component; The total number of control beats corresponding to the first five rotation cycles is obtained by multiplying the measured duration of the first rotation by 5 and then dividing by the sampling period. To collect the feed resistance for each cycle within the acquisition window.

[0041] It should be noted that using the average resistance value of the first five revolutions in the starting stage as a characteristic value can effectively avoid single-point noise interference, thereby stably locking the average furnace exit temperature characteristics of this batch of billets.

[0042] Subsequently, to classify the spatial sensitivity of the control sequence, the first-order gradient of the feedforward compensation sequence between adjacent control beats is calculated, and the arithmetic mean of the absolute values ​​of the gradients is defined as the partition threshold. The partition threshold satisfies the following relationship:

[0043] In the formula, For the first The partition threshold of the component ring; For the feedforward compensation sequence in the th Feedforward amount per beat; This represents the total number of beats.

[0044] Understandably, if the feedforward quantity changes drastically within a certain period, i.e., the absolute value of the first gradient is greater than the partition threshold, then that cycle is classified into the high gradient sensitive region; otherwise, it is classified into the low gradient stable region. This partitioning logic is based on the physical mechanism that the control gain is more sensitive to operating condition fluctuations when dynamic changes are drastic.

[0045] In this way, the processing data can be deconstructed in multiple dimensions, including dynamics, kinematics, and gradients in the control domain, laying a data foundation for subsequent implementation of differentiated compensation strategies for regions with different physical characteristics.

[0046] S3, in the The feed resistance is collected in the starting section of the rolling mill and the current average initial resistance is calculated. The ratio of the average initial resistance of the current piece to that of the previous piece is calculated to obtain the inter-batch resistance ratio.

[0047] Specifically, in the first After the workpiece begins processing, the acquisition and calculation logic in step S2 is repeated to obtain the feed resistance corresponding to each control cycle during the initial phase (the first five rotational cycles). Based on the feed resistance at each cycle, the feed resistance of the first cycle is calculated. The average initial resistance of each component was determined. Subsequently, to quantitatively characterize the change in the equivalent gain of the hydraulic feed system caused by inter-batch temperature fluctuations, the inter-batch resistance ratio was calculated.

[0048] It should be noted that the inter-batch resistance ratio Satisfying the relation:

[0049] In the formula, For the first relative to the first ring component The ratio of batch resistance between components; For the first The average initial resistance of the component during the starting phase; For the previous workpiece, that is, the first The average initial resistance value corresponding to the component ring is calculated and stored in step S2.

[0050] Understandably, the inter-batch resistance ratio The reason for its construction lies in the fact that, in hydraulic servo feed control, changes in the metal deformation resistance directly alter the actual mapping relationship between the control current and the feed speed. Since the feed resistance is negatively correlated with the local temperature of the billet, by calculating the resistance ratio of two adjacent rings at the same rolling progress, it is possible to capture in real time the operating condition drift caused by furnace temperature control deviations. A value greater than 1 indicates that the temperature of the current workpiece is lower than that of the previous one, increasing the metal's resistance to deformation. This results in a slower actual response speed of the feed axis under the same control current input, meaning the system gain has physically decreased. This ratio transforms the complex temperature drop disturbance into a clear linear compensation factor, providing a quantitative basis for subsequent control step size adjustments.

[0051] In this way, by obtaining the inter-batch resistance ratio within the rolling start-up section, online quantitative sensing of inter-batch operating condition deviations is achieved, ensuring that subsequent compensation actions have clear physical feedback guidance.

[0052] S4, based on the inter-batch resistance ratio, performs linear compensation on the learning increment in the high-gradient sensitive region, while maintaining the learning increment in the low-gradient stable region updated according to the standard law, thus obtaining the first... The adaptive feedforward of the component.

[0053] Specifically, in order to accurately map the processing experience accumulated from the previous piece to the current blank with temperature deviation, it is necessary to determine the learning gain. Learning Gain The acquisition method includes: before the start of the batch production task, a set of candidate learning gain values ​​is preset. For example, the preferred set of candidate learning gain values ​​is 0.4 mA / s / mm, 0.6 mA / s / mm, and 0.8 mA / s / mm; during the debugging phase, three ring parts are continuously rolled, and each ring part is assigned a candidate learning gain value; the root mean square value of the feed speed error of each test ring part during the entire rolling process is calculated, and the candidate learning gain value corresponding to the minimum root mean square value is taken as the target learning gain. .

[0054] It should be noted that the timing of the generation of adaptive feedforward quantities includes: in the first... After the start-up phase of the component cycle is completed, i.e., after the first five rotational cycles are finished and the batch-to-batch resistance ratio is calculated, an adaptive feedforward quantity is calculated and generated for all subsequent control cycles and stored in the CNC system's cache. During the cycles within the start-up phase, the first... The component ring directly adopts the first one. The feedforward compensation sequence of the component loop is used as the control input.

[0055] Furthermore, the first The adaptive feedforward of the component satisfies the following relationship:

[0056] In the formula, For the first The ring component in the first The adaptive feedforward amount for each control cycle; For the first The feedforward compensation sequence value of components in the same cycle time; The inter-batch resistance ratio determined in step S3; The learning gain determined by the above identification method; For the first The ring component in the first The feed rate error of each control cycle; This refers to the high gradient sensitive region defined in step S2; This refers to the low-gradient stable region defined in step S2.

[0057] Understandably, this formula is constructed because the high gradient sensitive region typically corresponds to critical stages where the cross-section of the ring changes drastically or the feed rate adjusts rapidly. At this time, the control performance is extremely sensitive to fluctuations in operating conditions. The inter-batch resistance ratio is inversely proportional to the actual speed gain of the hydraulic system. When the inter-batch resistance ratio is greater than 1, it means that the current billet temperature is low and the hardness is high. The speed compensation effect produced by the same current increment will be reduced. In this case, by learning the increment... Multiplying by the inter-batch resistance ratio for linear amplification effectively compensates for the attenuation of correction due to gain reduction. Conversely, when the inter-batch resistance ratio is less than 1, the increment is correspondingly contracted to prevent correction overshoot caused by billet overheating and gain increase. In the low-gradient steady region, the system is robust due to the gradual change in feedforward. By maintaining the standard learning law update, the stability of the control sequence in the global range can be ensured.

[0058] In this way, by combining the macroscopic inter-batch operating condition characteristics with the microscopic time-domain control gradient through a differentiated partition compensation strategy, the physical consistency of the feedforward correction amount under different temperature conditions is achieved, thus eliminating wall thickness fluctuations caused by inherited data mismatch from the source.

[0059] S5 superimposes the adaptive feedforward quantity with the real-time feedback quantity to obtain the final control current applied to the hydraulic servo valve, so as to eliminate the wall thickness deviation caused by batch temperature fluctuations through the final control current.

[0060] Specifically, the CNC system synthesizes the adaptive feedforward quantity generated in step S4 with the real-time feedback quantity generated by the real-time sensor in real time, step by step, to form a synthesized command to drive the hydraulic actuator; the final control current satisfies the following relationship:

[0061] In the formula, For the first The ring component in the first The final control current applied to the hydraulic servo valve by each control cycle is preferably between 4 mA and 20 mA. This is the adaptive feedforward quantity pre-calculated and stored in the cache for this tick; This refers to the real-time feedback quantity generated by the conventional proportional-integral-derivative control algorithm based on the current cycle time.

[0062] Understandably, this formula is constructed because the adaptive feedforward quantity carries the working condition experience corrected based on the inter-batch resistance ratio, and is responsible for pre-compensating for systematic feed deviations caused by changes in metal deformation resistance; while the real-time feedback quantity is responsible for handling random disturbances during the processing. The final control current formed by the superposition of the two enables the opening degree of the hydraulic servo valve to accurately match the actual hardness of the current billet, ensuring that the mandrel can still strictly follow the commanded feed trajectory under dynamically fluctuating temperature conditions, thereby eliminating wall thickness oscillations in the circumferential direction of the ring at the source.

[0063] Among them, after completing the first After the rolling of the ring parts, in order to further improve production efficiency and ensure the stability of the control system, the control method also includes convergence determination and sequence freezing steps: calculating the first step using the standard root mean square formula. The root mean square (RMS) value of the feed rate error of the workpiece throughout the entire process is calculated. This RMS value is then compared to a preset convergence threshold. If the RMS value is less than the convergence threshold, the adaptive feedforward update is stopped during subsequent workpiece processing, and the current feedforward compensation sequence is frozen as a fixed input for subsequent workpieces. Understandably, the convergence threshold is preferably 0.1 mm / s. When the error drops below this threshold, it indicates that the current control sequence can perfectly adapt to the average working conditions of the current batch of billets. The freezing operation can avoid unnecessary fluctuations in the control sequence caused by sensor noise.

[0064] It should be noted that, to cope with sudden changes in operating conditions that may occur during production, the control method also includes a warning restart step: during batch production after freezing the feedforward compensation sequence, the sliding standard deviation of the ring component wall thickness error is monitored in real time; in response to the sliding standard deviation exceeding the warning threshold in three consecutive processed parts, the freeze state of the feedforward compensation sequence is lifted, and steps S2 to S4 are executed again for adaptive learning. Understandably, the sliding standard deviation characterizes processing stability by calculating the dispersion of wall thickness deviation within the most recent control cycles, and the warning threshold is preferably 0.5 mm. When a sudden change in ambient temperature or batch switching of billet material causes the frozen sequence to become inapplicable, the automatic restart learning mechanism ensures that the system has long-term self-healing capabilities.

[0065] like Figure 2 As shown, a comparative diagram illustrates the difference in circumferential wall thickness of ring parts produced in batches under conditions of significant fluctuations in billet furnace temperature, using a traditional fixed-step feedforward control method and the method of this invention. It can be clearly seen that the traditional method, due to feedforward compensation mismatch, causes significant oscillations in wall thickness difference between batches that cannot converge; while the adaptive feedforward control method provided by this invention accurately senses the temperature difference and implements compensation from the second piece onwards, significantly suppressing wall thickness oscillations and enabling the wall thickness difference of subsequent batches of ring parts to quickly stabilize within the engineering tolerance range.

[0066] like Figure 3 As shown, the iterative evolution of the feedforward sequence in the time domain and its corresponding velocity error convergence curve throughout the process from the first piece to convergence and freezing, using the method of this invention, are illustrated. The dashed box in the figure marks the high gradient region. It can be seen that as the batch increases, the feedforward sequence undergoes targeted working condition compensation correction at high gradient moments with large differences. The amplitude and root mean square value of the corresponding velocity error sequence show a rapid decreasing trend, reaching the convergence threshold and successfully freezing at the 5th piece.

[0067] In this way, by organically combining the adaptive feedforward quantity of working condition perception with the real-time feedback quantity, and supplementing it with the closed-loop guarantee logic of convergence freeze and early warning restart, the radial and axial ring rolling machine achieves high-precision and high-stability automated production under complex batch-to-batch temperature difference disturbances, which significantly improves the dimensional consistency of key components such as aerospace flange rings.

[0068] An embodiment of the CNC fully automatic radial and axial ring rolling machine ring precision control system provided by the present invention: The precision control system for rings of a fully automatic CNC radial axial rolling machine includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned precision control method for rings of a fully automatic CNC radial axial rolling machine is implemented.

[0069] The precision control system for the rings of the fully automatic radial and axial rolling machine also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0070] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0071] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for controlling the accuracy of rings in a fully automatic CNC radial and axial ring rolling machine, characterized in that, Includes the following steps: The rolling time of a single piece is discretized according to the control cycle, and the feedforward compensation sequence of the first ring piece is initialized; In the During the rolling process of the workpiece, the feed speed error and feed resistance of each control cycle are collected simultaneously to obtain the error sequence, resistance sequence and initial resistance mean. Calculate the first gradient of the feedforward compensation sequence between adjacent control beats, and define the arithmetic mean of the absolute values ​​of the gradients as the partition threshold to divide the entire control beat into a high gradient sensitive region and a low gradient stable region. In the The feed resistance in the starting section of the rolling mill is collected and the current average initial resistance is calculated. The ratio of the average initial resistance of the current piece to that of the previous piece is calculated to obtain the inter-batch resistance ratio. Based on the inter-batch resistance ratio, linear compensation is performed on the learning increment in the high-gradient sensitive region, while the learning increment in the low-gradient stable region is updated according to the standard law, thus obtaining the first... The adaptive feedforward quantity of the component loop; The adaptive feedforward quantity is superimposed with the real-time feedback quantity to obtain the final control current applied to the hydraulic servo valve, so as to eliminate the wall thickness deviation caused by batch-to-batch temperature fluctuations through the final control current.

2. The method for controlling the accuracy of ring parts in a fully automatic radial and axial ring rolling machine according to claim 1, characterized in that, Methods for obtaining the initial resistance mean include: obtaining the first... The measured duration of the first revolution after the start of the rolling mill is determined; based on the measured duration, the total number of control cycles corresponding to the first five rotation cycles after the start of the rolling mill is determined as the initial resistance acquisition window; the feed resistance sequence within the initial resistance acquisition window is arithmetically averaged to obtain the average initial resistance value.

3. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 1, characterized in that, The method for calculating the feed rate error includes: obtaining the real-time displacement of the hydraulic cylinder piston at the end of the current control cycle, and calculating the ratio of the difference between the real-time displacement and the displacement at the end of the previous control cycle to the sampling period to obtain the actual feed rate; the difference between the commanded feed rate stored in the CNC system and the actual feed rate is taken as the feed rate error; among which, a positive feed rate error indicates that the current pressing amount is insufficient, and a negative feed rate error indicates that the current pressing amount is too large.

4. The method for controlling the accuracy of ring parts in a fully automatic radial and axial ring rolling machine according to claim 1, characterized in that, The method for obtaining the feed resistance includes: collecting the real-time oil pressure of the hydraulic cylinder in each control cycle; and using the product of the real-time oil pressure and the effective working area of ​​the hydraulic cylinder as the feed resistance. The magnitude of the feed resistance is negatively correlated with the local temperature of the billet and is used to characterize the metal deformation resistance.

5. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 1, characterized in that, The inter-batch resistance ratio and the speed gain of the hydraulic feed system satisfy an inverse relationship: when the inter-batch resistance ratio is greater than 1, it is determined that the equivalent speed gain of the current piece is lower than that of the previous piece, and the learning increment of the high gradient sensitive area is amplified to compensate for the attenuation of the feed correction amount. When the inter-batch resistance ratio is less than 1, it is determined that the equivalent speed gain of the current piece is higher than that of the previous piece, and the learning increment of the high gradient sensitive area is reduced accordingly to suppress feed correction overshoot.

6. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 1, characterized in that, Get the first The adaptive feedforward amount of the component includes: in the first... After the rolling start-up phase of the workpiece is completed, the adaptive feedforward quantity is calculated once and stored in the CNC system cache for the remaining control cycles; for the control cycles within the start-up phase, the first cycle is directly used. The feedforward compensation sequence of the component loop is used as the adaptive feedforward quantity.

7. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 1, characterized in that, The acquisition of learning increments also involves learning gains. Identification: Multiple sets of different candidate learning gain values ​​are preset and applied in the production process of the first three consecutive ring parts; the root mean square value of the feed rate error after the end of each ring part is calculated, and the candidate learning gain value corresponding to the minimum root mean square value of the third ring part is determined as the target learning gain.

8. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 1, characterized in that, In the After the rolling of the ring part is completed, the control method also includes: calculating the root mean square value of the feed speed error during the entire rolling process of the current part, and comparing the root mean square value with the preset convergence threshold; if the root mean square value is less than the preset convergence threshold, the current adaptive feedforward amount is frozen for use in the production of subsequent ring parts of the same specification.

9. The method for controlling the accuracy of ring parts in a fully automatic CNC radial and axial ring rolling machine according to claim 8, characterized in that, After the adaptive feedforward quantity is frozen, the control method also includes an early warning restart step: continuously monitor the inter-batch resistance ratio of each subsequent component in the starting segment; calculate the sliding standard deviation of the initial resistance mean of the most recent preset number of batches; and, in response to the deviation between the initial resistance mean and the sliding mean of the current component being greater than twice the sliding standard deviation, unfreeze the state and re-execute the adaptive update iteration.

10. A precision control system for ring parts of a fully automatic CNC radial and axial ring rolling machine, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for controlling the precision of ring parts of a fully automatic radial and axial rolling machine as described in any one of claims 1-9 is implemented.

Images