An electro-hydraulic bending machine slide synchronous precision self-adaptive calibration method based on oil temperature dynamic compensation

Abstract

The application relates to an electro-hydraulic bending machine slider synchronization precision self-adaptive calibration method based on oil temperature dynamic compensation, which comprises the following steps: injecting orthogonal pseudo-random micro-vibration excitation into Y1 and Y2 servo valves in a pressure maintaining stage; synchronously collecting pressure, synchronization error and temperature signals to determine effective temperature; taking cylinder internal leakage and valve zero leakage as temperature-related damping coefficients and zero leakage coefficients respectively in a double-input double-output model, and estimating respective coefficients of two cylinders in real time through recursive identification; correlating temperature to update two independent compensation curves of internal leakage and valve leakage; and inquiring the curves to obtain decoupling feedforward compensation in a subsequent movement stage and superimposing the compensation on position closed-loop control output. The application realizes decoupling identification and independent compensation of double-cylinder internal leakage and valve leakage in situ in the pressure maintaining stage, eliminates coupling compensation error, the compensation model can be evolved online with equipment aging, and the pressure maintaining synchronization precision and long-term maintenance-free nature are significantly improved.

Description

Technical Field

[0001] This application relates to the field of hydraulic synchronization control of electro-hydraulic bending machines, and in particular to an adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation. Background Technology

[0002] The synchronization accuracy of the slider in an electro-hydraulic bending machine is a key factor affecting the consistency of bending angle. Existing synchronization control methods generally use a grating ruler to detect the position difference between the two sliders in real time, and use closed-loop control combined with oil temperature compensation to correct the servo valve output. Oil temperature compensation is usually based on a static correction model of "oil temperature-hydraulic oil viscosity-flow gain" established by a single point temperature in the oil tank, and the output flow of the servo valve is uniformly scaled. Some solutions also include hydraulic cylinder leakage in the compensation, and use a fixed linear relationship calibrated at the factory for open-loop correction.

[0003] However, the above methods have fundamental flaws: Firstly, the zero-position leakage caused by the thermal expansion of the servo valve core and the internal leakage caused by the temperature change of the hydraulic cylinder seals are coupled and superimposed during the pressure holding stage. Existing solutions treat both as a unified flow error for compensation, failing to distinguish the source of error, leading to over-compensation or under-compensation, especially under high-pressure pressure holding conditions where the synchronization residual is significant. Secondly, the compensation model relies on factory calibration and cannot adaptively update with changes in the wear difference of the dual-cylinder seals and oil deterioration during equipment operation, resulting in a significant decrease in compensation accuracy after long-term operation. The bending machine slide operation has multi-stage characteristics of rapid descent, working feed, pressure holding, and return. During the pressure holding stage, the internal leakage of the hydraulic cylinder and the zero-position leakage of the servo valve are the dominant factors of synchronization error. However, due to the fact that the hydraulic system is in a low-flow or zero-flow state during this stage, conventional identification methods cannot separate the leakage temperature drift characteristics of the two cylinders in situ without affecting the process. This technological gap makes it difficult to further improve the synchronization accuracy of pressure holding.

[0004] Therefore, this invention proposes an adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation. Summary of the Invention

[0005] To address the aforementioned issues, this application provides an adaptive calibration method for the synchronization accuracy of the slider in an electro-hydraulic bending machine based on dynamic oil temperature compensation.

[0006] In a first aspect, this application provides an adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation, employing the following technical solution: An adaptive calibration method for the synchronization accuracy of the slider in an electro-hydraulic bending machine based on dynamic oil temperature compensation is applied to an electro-hydraulic bending machine equipped with a Y1 hydraulic cylinder, a Y2 hydraulic cylinder, Y1 servo valves and Y2 servo valves that drive the two hydraulic cylinders respectively, a slider displacement detection device, and a cylinder pressure sensor. The method includes the following steps: S1, Pressure Holding Trigger and Orthogonal Micro-vibration Excitation Injection: When the slider enters the pressure holding stage, a pseudo-random binary sequence micro-vibration excitation signal with limited amplitude and orthogonal phase is superimposed on the control signals of the Y1 servo valve and the Y2 servo valve respectively; the amplitude of the excitation signal is limited to make the pressure fluctuation of the rodless chamber of the two hydraulic cylinders less than a preset pressure threshold, and the slider synchronization error fluctuation less than a preset displacement threshold. S2. Synchronous acquisition of multi-source response signals and determination of effective temperature: During the excitation injection, the rodless chamber pressure of cylinder Y1, the rodless chamber pressure of cylinder Y2, the slider synchronization error obtained by the displacement detection device, and at least one temperature signal characterizing oil temperature are synchronously acquired, and the real-time effective temperature is determined by the acquired temperature signal. S3. Decoupling and Identification of Internal Leakage and Valve Leakage Characteristics of Dual Cylinders: Using the excitation signal as the model input and the collected pressure signal and slider synchronization error signal as the model output, a dual-input dual-output mathematical model describing the micro-dynamic characteristics of the hydraulic system in the pressure holding section is established. In this model, the internal leakage characteristics of the hydraulic cylinder are characterized as a nonlinear damping coefficient related to oil temperature, and the zero-position leakage characteristics of the servo valve are characterized as a zero-position leakage coefficient related to oil temperature. Using the input and output data, the internal leakage damping coefficient and valve zero-position leakage coefficient of cylinders Y1 and Y2 under the current temperature condition are estimated in real time through a recursive identification algorithm. S4. Dynamic update of dual-channel compensation curves: The two sets of internal leakage damping coefficients and valve zero-position leakage coefficients identified are associated with the corresponding real-time effective temperatures, and the cylinder internal leakage-oil temperature compensation curve and servo valve zero-position-oil temperature compensation curve are updated synchronously. S5. Decoupling feedforward compensation output: In each motion stage of the subsequent bending cycle, the updated compensation curve is queried based on the real-time effective temperature to obtain the decoupling feedforward compensation amount, and the feedforward compensation amount is superimposed on the position closed-loop control output of the corresponding servo valve.

[0007] Preferably, in step S3, the dual-input dual-output mathematical model also introduces the real-time working pressure of the rodless chambers of cylinders Y1 and Y2 as input variables to construct a two-dimensional mapping model of "oil temperature-pressure difference-leakage", so that the identified internal leakage damping coefficient and valve zero-position leakage coefficient are pure temperature-dependent coefficients stripped of the current working pressure difference effect.

[0008] Preferably, in step S1, before superimposing the micro-vibration excitation signal, the off-center load coefficient is calculated based on the real-time pressure difference between cylinder Y1 and cylinder Y2; when the off-center load coefficient exceeds a preset threshold, the amplitude of the excitation signal of the heavy-load side servo valve is increased, and the amplitude of the excitation signal of the light-load side servo valve is decreased.

[0009] Preferably, the internal leakage damping coefficient identified in each step S3 is continuously recorded, and the long-term drift velocity of the internal leakage damping coefficient on one side is monitored. When the drift velocity exceeds the preset aging threshold, the high-frequency component bandwidth of the superimposed pseudo-random binary sequence is automatically expanded when step S1 is executed next time. The upper limit of the expanded high-frequency component is 150-200Hz.

[0010] Preferably, the real-time effective temperature in step S2 is determined in the following way: Temperature sensors are arranged in the oil tank, servo valve block, rodless cavity wall of cylinder Y1, and rodless cavity wall of cylinder Y2 to acquire multi-node temperature data. The cylinder is discretized into multiple micro-element control bodies along the stroke direction. The flow path of the oil micro-elements is tracked according to the real-time output flow rate of the servo valve and the cross-sectional area of ​​the cylinder. The equivalent average temperature of the oil participating in leakage in the rodless cavity is output as the real-time effective temperature.

[0011] Preferably, the execution process of the recursive identification algorithm in step S3 includes: continuously recording the historical values ​​of the internal leakage damping coefficient of each cylinder obtained each time, and monitoring its long-term drift speed; when the drift speed of the internal leakage damping coefficient on one side exceeds the preset aging threshold, automatically reducing the forgetting factor of the recursive identification algorithm or increasing the process noise covariance to accelerate the model's tracking and convergence of the seal wear state.

[0012] Preferably, during the first N pressure holding identification cycles after the device cold start, the recursive identification algorithm in step S3 uses a model update gain higher than that in the normal state; when the relative deviation of two consecutive identification results is lower than the preset convergence threshold, it automatically switches to normal update gain; where N is a preset positive integer, and the value range is 3 to 5.

[0013] Preferably, in step S5, the decoupling feedforward compensation amount is weighted according to the following rules: During the rapid descent and return phases, the flow-temperature drift compensation component derived from the valve zero-position leakage coefficient is given primary weight, while the leakage compensation component derived from the internal leakage damping coefficient is given secondary weight. During the pressure holding stage, the leakage compensation component derived from the internal leakage damping coefficient is used as the main weight. During the bending stage, the flow rate temperature drift compensation component and the leakage compensation component are weighted and integrated, and the weighting coefficients are smoothly transitioned with the real-time speed of the slider.

[0014] Preferably, step S3 further includes an anomaly verification step: calculating the output residual of the identification model in real time and monitoring the transient fluctuations of the acquired signal; when the output residual exceeds the statistical threshold, or the transient fluctuations of pressure or displacement exceed the preset safety threshold, terminating the current parameter update, retaining the compensation parameters of the previous valid identification cycle, and triggering an alarm signal.

[0015] Preferably, in step S1, the preset pressure threshold is 0.10 to 0.15 MPa, and the preset displacement threshold is 0.003 to 0.005 mm; the recursive identification algorithm is the recursive least squares method or the extended Kalman filter algorithm.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. By actively injecting orthogonal micro-vibration excitation signals during the pressure holding stage, and decoupling and identifying the zero-position leakage of the servo valve and the internal leakage of the hydraulic cylinder as independent parameters in the dual-input dual-output mathematical model, this method achieves in-situ separation of the internal leakage temperature drift characteristics and valve leakage temperature drift characteristics of cylinders Y1 and Y2 under the pressure holding condition of the bending machine. This fundamentally eliminates the over-compensation or under-compensation problem caused by coupling the two into a unified flow gain for compensation in the existing scheme, and significantly improves the slider synchronization accuracy during the pressure holding stage.

[0017] 2. By introducing real-time working pressure as an input variable into the identification model, a two-dimensional mapping relationship of "oil temperature-pressure difference-leakage" is constructed, making the identified in-cylinder leakage damping coefficient and valve zero-position leakage coefficient pure temperature-dependent coefficients that are stripped of the current working pressure difference effect. This ensures the accuracy and universality of the compensation model under different bending process pressures and avoids the defect in existing technologies where the identification parameters fail as the working pressure changes.

[0018] 3. By continuously monitoring the long-term drift rate of the leakage damping coefficient in each cylinder and automatically adjusting the forgetting factor or process noise covariance of the recursive identification algorithm accordingly, this method enables the compensation model to proactively adapt to changes in hydraulic characteristics caused by wear and aging of the seals. It achieves online self-calibration throughout the entire lifecycle without downtime, overcoming the shortcomings of existing factory-calibrated compensation schemes where accuracy significantly decreases with increasing equipment operating time. Furthermore, combined with a rapid convergence mechanism during cold start-up, the equipment can quickly recover to a high-precision compensation state after prolonged downtime.

[0019] 4. By adaptively adjusting the amplitude distribution of the two-cylinder excitation signals under off-center load conditions, and introducing an abnormal data elimination mechanism for output residual verification and signal transient fluctuation monitoring during the identification process, this method can ensure the validity of the identification data and the reliability of the compensation parameters in asymmetric bending operations and common industrial interference environments, and has high industrial robustness. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for adaptive calibration of the slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation, as described in an embodiment of this application. Detailed Implementation

[0021] The following is in conjunction with the appendix Figure 1This application will be described in further detail.

[0022] This embodiment provides an adaptive calibration method for the slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation. This method is applied to an electro-hydraulic bending machine (i.e., a conventional electro-hydraulic bending machine) equipped with hydraulic cylinders Y1 and Y2, servo valves Y1 and Y2 respectively driving the two cylinders, a slider displacement detection device, and cylinder pressure sensors. The slider displacement detection device typically consists of two optical or magnetic scales mounted on both sides of the slider, used to acquire the slider positions on the Y1 and Y2 sides in real time and calculate the slider synchronization error. The cylinder pressure sensors are installed in the oil circuits of the rodless chambers of cylinders Y1 and Y2 respectively, used to acquire the working pressure of the two chambers in real time. The arrangement of the temperature sensors will be described in detail below in conjunction with the method for determining the real-time effective temperature. The controller uses an industrial-grade embedded system to execute all the identification algorithms and compensation logic described in this embodiment.

[0023] Reference Figure 1 An adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation includes the following steps: S1, Pressure Holding Trigger and Orthogonal Micro-vibration Excitation Injection: When the slider enters the pressure holding stage, a pseudo-random binary sequence micro-vibration excitation signal with limited amplitude and orthogonal phase is superimposed on the control signals of the Y1 servo valve and the Y2 servo valve respectively; the amplitude of the excitation signal is limited to make the pressure fluctuation of the rodless chamber of the two hydraulic cylinders less than a preset pressure threshold, and the slider synchronization error fluctuation less than a preset displacement threshold. S2. Synchronous acquisition of multi-source response signals and determination of effective temperature: During the excitation injection, the rodless chamber pressure of cylinder Y1, the rodless chamber pressure of cylinder Y2, the slider synchronization error obtained by the displacement detection device, and at least one temperature signal characterizing oil temperature are synchronously acquired, and the real-time effective temperature is determined by the acquired temperature signal. S3. Decoupling and Identification of Internal Leakage and Valve Leakage Characteristics of Dual Cylinders: Using the excitation signal as the model input and the collected pressure signal and slider synchronization error signal as the model output, a dual-input dual-output mathematical model describing the micro-dynamic characteristics of the hydraulic system in the pressure holding section is established. In this model, the internal leakage characteristics of the hydraulic cylinder are characterized as a nonlinear damping coefficient related to oil temperature, and the zero-position leakage characteristics of the servo valve are characterized as a zero-position leakage coefficient related to oil temperature. Using the input and output data, the internal leakage damping coefficient and valve zero-position leakage coefficient of cylinders Y1 and Y2 under the current temperature condition are estimated in real time through a recursive identification algorithm. S4. Dynamic update of dual-channel compensation curves: The two sets of internal leakage damping coefficients and valve zero-position leakage coefficients identified are associated with the corresponding real-time effective temperatures, and the cylinder internal leakage-oil temperature compensation curve and servo valve zero-position-oil temperature compensation curve are updated synchronously. S5. Decoupling feedforward compensation output: In each motion stage of the subsequent bending cycle, the updated compensation curve is queried based on the real-time effective temperature to obtain the decoupling feedforward compensation amount, and the feedforward compensation amount is superimposed on the position closed-loop control output of the corresponding servo valve.

[0024] The specific implementation details of each step are explained below, and corresponding explanations are provided for the additional technical features of each dependent claim.

[0025] Step S1 regarding pressure holding triggering and orthogonal micro-vibration excitation injection specifically includes: In each working cycle of the bending machine, the controller collects the slider position and the working pressure of the two hydraulic cylinders in real time to determine whether the slider has entered the pressure holding stage. The conditions for determining the pressure holding stage are: the slider reaches the set bending position and enters the pressure holding timer. At this time, the servo valve output is close to zero, and the hydraulic cylinder is in a high-pressure low-flow or no-flow state.

[0026] Once the pressure holding phase is determined, the controller superimposes a pair of amplitude-limited and phase-orthogonal pseudo-random binary sequence (PRBS) micro-vibration excitation signals onto the control signals of servo valves Y1 and Y2, respectively. The PRBS signals are generated by a linear feedback shift register (LFSR). In this embodiment, a 7th-order LFSR is used, with a clock frequency of 100Hz under normal operating conditions, corresponding to an effective bandwidth of approximately 0 to 50Hz, and a sequence period of 127 clock cycles. The orthogonal injection method is as follows: the control signal of servo valve Y1 is superimposed with the original PRBS sequence, and the control signal of servo valve Y2 is superimposed with the inverted PRBS sequence. That is, the excitation increments of the two are equal in magnitude and opposite in direction within each clock cycle. This differential orthogonal injection method can artificially create slight differences in the hydraulic characteristics of the two cylinders, allowing the slider synchronization error signal to carry richer identification information, while effectively suppressing common-mode interference and improving the signal-to-noise ratio of subsequent identification.

[0027] The reason for injecting micro-vibration excitation during the pressure holding stage is that the hydraulic cylinder is almost stationary during this stage, and the servo valve is near zero. At this time, the synchronization error drift is almost entirely caused by leakage within the cylinder and leakage at the valve zero position, without the coupling of interference factors such as flow gain, making it an ideal window for identifying leakage characteristics. However, in the field of electro-hydraulic bending machines, there has long been an operating standard that "absolute steady state must be maintained during the pressure holding stage," and those skilled in the art generally believe that any external disturbance will affect the forming accuracy. This method, by strictly limiting the excitation amplitude to a pressure fluctuation range of 0.10 to 0.15 MPa and a displacement fluctuation range of 0.003 to 0.005 mm, makes the impact of the excitation on the forming accuracy of the workpiece negligible, thus overcoming the aforementioned technical bias and transforming the pressure holding stage from an "unusable measurement dead zone" into a "diagnostic window with the most information." Another consideration for adopting the orthogonal injection method (Y1 in positive phase, Y2 in negative phase) is that when the excitation directions of the two cylinders are opposite, the response amplitude of the synchronization error signal is amplified and common-mode interference is suppressed, which is beneficial to improving the signal-to-noise ratio of subsequent identification.

[0028] The calibration method for the excitation signal amplitude is as follows: During the equipment commissioning phase, gradually increase the PRBS excitation amplitude from zero, and simultaneously monitor the peak pressure fluctuation of the rodless chambers of cylinders Y1 and Y2, as well as the peak fluctuation of the slider synchronization error; when the pressure fluctuation reaches 0.15MPa or the synchronization error fluctuation reaches 0.005mm, take 80% of the current excitation amplitude as the official usage value.

[0029] As a preferred implementation, before superimposing the excitation signal, the controller first calculates the off-center load coefficient based on the real-time pressure difference between the rodless chambers of cylinders Y1 and Y2 to assess whether the current bending operation is under an asymmetrical load condition. The formula for calculating the off-center load coefficient is α=|P1-P2| / max(P1,P2), where P1 and P2 are the real-time pressures of the rodless chambers of cylinders Y1 and Y2, respectively. When the off-center load coefficient α does not exceed a preset threshold (set to 0.2 in this embodiment), it indicates that the load on both cylinders is basically symmetrical, and the two cylinders use equal excitation amplitudes. When the off-center load coefficient α exceeds 0.2, it indicates that the current operation is under an off-center load condition, the seal on the heavy-load side is more compressed, and the nonlinearity of the internal leakage behavior is stronger. Stronger excitation energy is required to fully stimulate its leakage characteristics. Therefore, the excitation signal amplitude of the servo valve on the heavy-load side is multiplied by (1+α) to enhance it, and the excitation signal amplitude of the light-load side is multiplied by (1-α) to weaken it. The total excitation energy remains basically unchanged, ensuring that the leakage characteristics on the heavy-load side are fully stimulated, while ensuring that the light-load side does not produce excessive displacement.

[0030] Step S2 regarding the synchronous acquisition of multi-source response signals and determination of effective temperature specifically includes: During the micro-vibration excitation injection, the controller synchronously acquires four key signals at a high sampling rate (1kHz in this embodiment): the rodless chamber pressure of cylinder Y1, the rodless chamber pressure of cylinder Y2, the slider position difference between the Y1 and Y2 sides obtained by the slider displacement detection device (i.e., slider synchronization error), and at least one temperature signal characterizing the oil temperature to determine the real-time effective temperature. The acquisition duration covers at least one complete PRBS sequence cycle to ensure the sufficiency of the identification data.

[0031] Determining the real-time effective temperature is a crucial factor affecting compensation accuracy. Current technologies generally only collect the temperature at a single point in the oil tank as a representative of the oil temperature. However, the stroke of the cylinders in electro-hydraulic bending machines is typically 200 to 400 mm, with a large single-cylinder cavity and a long residence time of the oil within the cylinder. Heat exchange between the cylinder wall and the environment and the machine frame results in a difference of 3 to 8°C between the actual oil temperature inside the cylinder and the oil temperature in the oil tank, which can exceed 10°C during cold starts. Existing oil temperature compensation schemes generally use the single-point temperature in the oil tank directly for compensation calculations, neglecting this cavity thermal hysteresis effect. This invention preferably uses a multi-node temperature gradient field estimation method to determine the real-time effective temperature. The specific scheme is as follows: Temperature sensors are arranged at four locations: near the oil tank suction port, on the back of the servo valve block mounting surface, on the outer wall of the rodless cavity of cylinder Y1 near the sealing ring, and on the outer wall of the rodless cavity of cylinder Y2 near the sealing ring, to acquire multi-node temperature data; the cylinder is uniformly discretized into twenty to fifty micro-element control bodies along the stroke direction; the axial propulsion speed of the oil micro-particle in the cylinder is calculated based on the real-time servo valve output flow rate and the effective area of ​​the rodless cavity of the cylinder; the temperature nodes experienced sequentially by the oil micro-particle flowing from the oil tank through the valve block into the rodless cavity along the flow path are tracked; a one-dimensional axial fluid-structure interaction heat conduction model is established. The energy conservation equation of this model describes the convective heat transfer process between the oil and the cylinder wall and the environment when the oil flows axially; the oil temperature in the oil tank, the valve block temperature, the cylinder wall temperature, the ambient temperature, and the real-time flow rate are used as boundary conditions and input parameters; the finite difference method is used to iteratively solve the model in each sampling cycle of the controller; the equivalent average temperature of the oil participating in leakage in the rodless cavity is output as the real-time effective temperature. This method eliminates the need for direct measurement of the oil temperature inside the cylinder (which presents sealing and reliability challenges under high-pressure conditions by inserting a sensor into the cylinder), and instead uses externally measurable temperature nodes and fluid dynamics calculations to indirectly obtain a more accurate representative temperature value, fundamentally eliminating the compensation deviation caused by cavity thermal hysteresis.

[0032] Step S3, regarding the decoupling identification of internal leakage and valve leakage characteristics in dual cylinders, specifically involves: Step S3 is the core algorithm step of this invention. Using the PRBS excitation signal injected in step S1 as the model input, and the pressure signal and slider synchronization error signal synchronously acquired in step S2 as the model output, a dual-input dual-output mathematical model describing the micro-dynamic characteristics of the hydraulic system in the pressure-holding section is established.

[0033] In terms of model structure, the state vector contains four parameters to be identified: the internal leakage damping coefficient b1 of cylinder Y1, the internal leakage damping coefficient b2 of cylinder Y2, the zero-position leakage coefficient c1 of valve Y1, and the zero-position leakage coefficient c2 of valve Y2. The state equation adopts a random walk model, that is, the value of each parameter at the current moment is equal to the value of the previous moment plus the process noise. This model allows the algorithm to autonomously track the slow drift of parameters caused by oil temperature changes and seal wear. The observation equation is constructed based on the physical laws of gap flow and orifice flow: the rate of change of pressure in the rodless chamber of cylinder Y1 is related to the flow rate flowing in through the servo valve (including flow pulsation caused by the excitation signal), the leakage flow rate caused by the internal leakage damping coefficient b1 (proportional to the pressure difference between the two chambers and inversely proportional to the internal leakage damping coefficient), and the leakage flow rate caused by the zero-position leakage coefficient c1 of the valve (depending on the pressure difference and leakage coefficient on both sides of the valve port); the same applies to cylinder Y2; the rate of change of slider synchronization error is obtained by subtracting the net flow difference (inflow rate minus leakage flow rate) between the effective area of ​​the cylinder and the two cylinder rodless chambers.

[0034] As a preferred embodiment, the aforementioned dual-input dual-output mathematical model also incorporates the real-time working pressure of the rodless chambers of cylinders Y1 and Y2 as input variables, thereby constructing a two-dimensional mapping model of "oil temperature-pressure difference-leakage". In this two-dimensional model, the internal leakage flow rate is characterized as having a sublinear relationship with the pressure difference (an exponential range of 0.5 to 1.0, depending on the geometry of the leakage gap) and an inverse relationship with the internal leakage damping coefficient, which itself is only a function of oil temperature; the valve zero-position leakage flow rate is also determined by the pressure difference and the valve zero-position leakage coefficient, which is also only a function of oil temperature.

[0035] The physical significance of introducing differential pressure as an independent input variable into the identification model lies in the following: According to gap flow theory, the internal leakage of the hydraulic cylinder is not only related to the internal leakage damping coefficient of the seal (reflecting the state of the seal itself), but also has a sublinear relationship with the differential pressure acting on both sides of the seal. Similarly, the zero-position leakage of the servo valve also depends on the differential pressure on both sides of the valve port. If differential pressure is not explicitly introduced during the identification process, the coefficients obtained are actually coupling products of temperature and pressure under the "current specific operating conditions." Once the bending process changes and the working pressure deviates significantly from the pressure at the time of identification, the coefficients will no longer be accurate. By constructing a two-dimensional mapping model of "oil temperature-differential pressure-leakage," the identification algorithm can operate under the premise of known real-time differential pressure, separating the differential pressure effect from the coefficients, making the output internal leakage damping coefficient and valve zero-position leakage coefficient pure material characteristic parameters that are only related to temperature. In this way, no matter how the pressure of the subsequent bending process changes, the compensation curve formed by these coefficients can be consulted to obtain accurate compensation values.

[0036] The identification algorithm employs either recursive least squares or extended Kalman filtering. If recursive least squares is used, the recursive formula includes three steps: gain vector calculation, parameter vector update, and covariance matrix update. The forgetting factor λ is set to 0.95 to 0.99 during normal operation to balance tracking speed and smoothness. If extended Kalman filtering is used, the algorithm iterates alternately between prediction and update steps. The state transition matrix is ​​an identity matrix, and the observation matrix is ​​the Jacobian matrix of the observation equation with respect to each state variable. The values ​​of the process noise covariance Q and measurement noise covariance R are tuned based on sensor accuracy and system characteristics. The initial values ​​for both algorithms are derived from the factory-calibrated reference curve or valid parameters saved from the previous run.

[0037] Unlike existing methods that uniformly correct servo valve temperature drift and hydraulic cylinder internal leakage temperature drift using flow gain, this method achieves in-situ separation and extraction of the internal leakage damping coefficient and valve zero-position leakage coefficient during the pressure holding stage. Each of the two error sources obtains an independent compensation channel, fundamentally eliminating coupled compensation errors. The effectiveness of this technique stems from the physical nature of the identification logic, rather than empirical correction.

[0038] During the identification process, anomaly verification is also performed to ensure industrial robustness. The controller calculates the output residual of the identification model in real time—that is, the sum of the squares of the differences between the actual observed value and the model's predicted observed value at the current moment—and monitors the transient fluctuations of the collected signals. When the output residual exceeds the statistical threshold determined based on three times the standard deviation of the historical residual mean, or when the transient fluctuations of pressure and displacement exceed the preset safety thresholds (i.e., pressure fluctuation of 0.15 MPa and displacement fluctuation of 0.005 mm as mentioned in step S1), it is determined that the current identification cycle is subject to external abnormal interference (such as workpiece slippage, sudden pressure change, or external vibration), the parameter update for the current cycle is immediately terminated, the compensation parameters of the previous valid identification cycle are retained, and an alarm signal is output through the human-machine interface to prompt the operator to pay attention.

[0039] Step S4 regarding the dynamic update and aging adaptation of the dual-channel compensation curve specifically includes: After identification, the obtained in-cylinder leakage damping coefficient b1 of Y1, in-cylinder leakage damping coefficient b2 of Y2, zero-position leakage coefficient c1 of Y1 valve, and zero-position leakage coefficient c2 of Y2 valve are respectively associated with the current real-time effective temperature determined by step S2, and added as new data points to their respective compensation curve databases.

[0040] Each compensation curve uses the real-time effective temperature as the x-axis and the corresponding coefficient value (internal leakage damping coefficient or valve zero-position leakage coefficient) as the y-axis. When historical data points already exist near the same temperature, the curve is updated using a weighted average method, with the weight of the new data determined by the forgetting factor or covariance of the recursive identification algorithm. When the temperature point appears for the first time, a new data point is directly inserted, and the continuity of the curve is ensured by interpolation of adjacent known points. Thus, four independent compensation curves—Y1 cylinder internal leakage-oil temperature compensation curve, Y2 cylinder internal leakage-oil temperature compensation curve, Y1 valve zero-position-oil temperature compensation curve, and Y2 valve zero-position-oil temperature compensation curve—are updated synchronously, forming a dual-channel decoupled compensation database.

[0041] In addition, the controller continuously records the historical values ​​of the internal leakage damping coefficient of each cylinder obtained from each identification and monitors its long-term drift speed. The drift speed is calculated by performing linear regression on the sequence of the most recent twenty identification results, and using the slope of the regression line as the drift speed. When the drift speed of the internal leakage damping coefficient on one side exceeds the preset aging threshold, it indicates that the seal on that side has entered the accelerated wear stage, and the leakage characteristics are changing rapidly. At this time, the controller automatically adjusts the parameters of the identification algorithm to accelerate the model's tracking and convergence of the seal wear state—if the recursive least squares method is used, the forgetting factor λ is reduced from the normal value of 0.95 to 0.99 to 0.80 to 0.90; if the extended Kalman filter algorithm is used, the corresponding elements of the process noise covariance matrix are increased to 3 to 5 times the normal value. After the tracking is accelerated, the weight of the newly identified data is significantly increased, and the model can quickly converge to the new characteristics of the seal after wear in the subsequent identification cycles, realizing the proactive evolution from "periodic calibration" to "on-demand accelerated tracking".

[0042] It should be noted that the forgetting factor λ in the recursive least squares method and the process noise covariance Q in the extended Kalman filter essentially determine the algorithm's response speed to parameter changes. Under normal operating conditions, using a larger λ (0.95 to 0.99) and a smaller Q is beneficial for suppressing measurement noise and maintaining the smoothness of parameter estimation. However, when the seal enters the accelerated wear stage, the internal leakage characteristics of the hydraulic cylinder are undergoing rapid and unidirectional changes. If normal operating parameters are maintained at this time, the identification results will lag significantly behind the actual wear state, leading to a continuous accumulation of compensation errors. This method automatically determines whether the seal has entered the accelerated wear stage by continuously monitoring the long-term drift rate of the internal leakage damping coefficient, and proactively reduces the forgetting factor or increases the process noise covariance accordingly. This significantly enhances the weight of the identification algorithm on new data, thereby rapidly tracking the new state of the seal within several identification cycles. This "on-demand accelerated tracking" strategy embeds the entire lifecycle management of the equipment into the identification algorithm, which differs from the existing maintenance method of simple periodic calibration.

[0043] Furthermore, as another preferred measure in conjunction with aging monitoring, when the long-term drift rate of the unilateral internal leakage damping coefficient exceeds a preset aging threshold, the controller automatically expands the high-frequency component bandwidth of the superimposed PRBS signal during the next execution of step S1. Under normal operating conditions, the PRBS clock frequency is 100Hz, corresponding to an effective bandwidth of approximately 0 to 50Hz; after expansion, the clock frequency increases to 300 to 400Hz, and the upper limit of the high-frequency effective component increases to 150 to 200Hz. High-frequency pressure waves can penetrate deeper into the tiny leak gaps caused by seal wear, accurately capturing the characteristic changes in the leakage spectrum in the early wear stage, achieving an upgrade from "periodic detection" to "on-demand precision diagnosis."

[0044] Step S5 regarding the decoupling feedforward compensation output is as follows: After updating the compensation curve in step S4, the latest compensation curve can be used in all motion stages of the subsequent bending cycle. Based on the real-time effective temperature determined in step S2, the controller queries the four independent compensation curves to obtain the internal leakage damping coefficient and valve zero-position leakage coefficient of cylinders Y1 and Y2 under the current temperature condition, and then calculates the decoupled feedforward compensation amount, which is then superimposed on the position closed-loop control output of the corresponding servo valve.

[0045] The feedforward compensation consists of two parts: one is the leakage compensation component derived from the internal leakage damping coefficient, which physically represents the additional servo valve flow required to compensate for the flow loss caused by internal leakage under the current differential pressure and temperature; the other is the flow-temperature drift compensation component derived from the valve zero-position leakage coefficient, which physically represents the bias required to compensate for the control current bias caused by the servo valve zero-position temperature drift. These two compensation components are calculated and applied independently to the two servo valves, thus avoiding overcompensation or undercompensation caused by coupling valve leakage and cylinder leakage compensation in existing technologies.

[0046] The weights of the compensation components at different stages of motion are dynamically allocated based on the error-dominant mechanism. The specific rules are as follows: During the rapid descent and return phases, the slider moves at high speed, the servo valve is in a large-open state, and flow temperature drift is the dominant factor in synchronization error. Therefore, the flow temperature drift compensation component derived from the valve zero-position leakage coefficient is given the main weight (set to 0.85), and the leakage compensation component derived from the internal leakage damping coefficient is given the secondary weight (set to 0.15). During the pressure holding phase, the slider is almost stationary, the servo valve is near the zero position, and internal leakage and valve zero-position leakage together constitute the error source, but internal leakage usually accounts for a larger proportion. Therefore, the leakage compensation component derived from the internal leakage damping coefficient is given the main weight (set to 0.90), and the flow temperature drift compensation component is given the secondary weight (set to 0.10). During the working feed and bending phase, the slider speed varies within the working range, and the contribution ratio of valve leakage and internal leakage changes dynamically with the speed. Therefore, the flow temperature drift compensation component and the leakage compensation component are weighted and fused. The weight of the flow temperature drift compensation component smoothly transitions from 0.50 to 0.70 with the real-time slider speed (the higher the speed, the greater the weight), and the weight of the leakage compensation component is 1 minus this weight. The smooth transition of weighting coefficients can be achieved using linear interpolation or a sigmoid function to avoid secondary impacts on synchronization accuracy caused by sudden changes in compensation amount due to weight abrupt changes.

[0047] Regarding the rapid convergence of cold start: When the equipment is cold-started after a long period of shutdown, the temperature of the hydraulic system is close to the ambient temperature, and the identification algorithm in step S3 does not yet have valid historical data for the current equipment. At this time, the controller calls the factory-calibrated benchmark compensation curve as the initial compensation basis.

[0048] Factory calibration is completed before the equipment leaves the factory. The process is as follows: Under standard experimental conditions, the hydraulic oil is heated or cooled to multiple set temperature points (e.g., 20℃, 30℃, 40℃, 50℃). After stabilization at each temperature point, the controller controls the slider to move to a preset position and applies a load to the rodless chambers of the two hydraulic cylinders to establish an equivalent holding pressure, simulating the holding pressure condition in a normal bending cycle. Under this equivalent holding pressure state, the complete process described in steps S1 to S3, including orthogonal micro-vibration excitation injection, multi-source signal synchronous acquisition, and decoupling identification of internal leakage and valve leakage characteristics of the dual cylinders, is executed. The damping coefficients of the Y1 cylinder internal leakage, Y2 cylinder internal leakage, Y1 valve zero-position leakage coefficient, and Y2 valve zero-position leakage coefficient corresponding to each temperature point are recorded. The data points of each temperature point are fitted using least squares to obtain four benchmark compensation curves, which are stored in the controller's non-volatile memory for recall during cold starts.

[0049] During the first N pressure holding identification cycles after cold start (N is a preset positive integer, 3 to 5 in this embodiment), the recursive identification algorithm in step S3 uses a model update gain higher than that in the normal state. If the recursive least squares method is used, the forgetting factor λ is initially set to 0.90 to 0.95; if the extended Kalman filter algorithm is used, the elements of the process noise covariance matrix are set to 3 to 5 times the normal value. The high update gain enables the identification parameters to converge quickly from the factory reference value to the current actual state of the equipment. When the relative deviations of the internal leakage damping coefficients and valve zero-position leakage coefficients in two consecutive identification results are both lower than the preset convergence threshold (5% in this embodiment), the controller determines that the model has fully converged and automatically switches the update gain to the normal state value. After that, it enters the stable tracking stage, taking into account both compensation accuracy and parameter stability.

[0050] The method described in this embodiment, through the coordinated operation of the five core steps and multiple preferred implementation methods, achieves in-situ, non-destructive separation of the coupling characteristics between the internal leakage temperature drift of the dual cylinders and the zero-position leakage temperature drift of the servo valve during the pressure holding stage, and constructs a decoupled feedforward compensation model accordingly. Compared with the prior art, this method has the following significant technical effects: First, the pressure holding synchronization accuracy is significantly improved, especially under large-volume continuous bending and high-pressure pressure holding conditions, completely eliminating the residual caused by coupling compensation; Second, the compensation model self-evolves online as the equipment ages and the oil state changes, eliminating the maintenance cost of periodic shutdown for manual calibration; Third, the off-center load adaptive excitation, cold start rapid convergence, and abnormal fault tolerance mechanism jointly ensure the industrial robustness of the method, making it suitable for various sheet metal bending processing sites.

[0051] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. An adaptive calibration method for the synchronization accuracy of a slider in an electro-hydraulic bending machine based on dynamic oil temperature compensation, applied to an electro-hydraulic bending machine equipped with a Y1 hydraulic cylinder, a Y2 hydraulic cylinder, Y1 servo valves and Y2 servo valves respectively driving the two hydraulic cylinders, a slider displacement detection device, and a cylinder pressure sensor, characterized in that, Includes the following steps: S1, Pressure Holding Trigger and Orthogonal Micro-vibration Excitation Injection: When the slider enters the pressure holding stage, a pseudo-random binary sequence micro-vibration excitation signal with limited amplitude and orthogonal phase is superimposed on the control signals of the Y1 servo valve and the Y2 servo valve respectively; the amplitude of the excitation signal is limited to make the pressure fluctuation of the rodless chamber of the two hydraulic cylinders less than a preset pressure threshold, and the slider synchronization error fluctuation less than a preset displacement threshold. S2. Synchronous acquisition of multi-source response signals and determination of effective temperature: During the excitation injection, the rodless chamber pressure of cylinder Y1, the rodless chamber pressure of cylinder Y2, the slider synchronization error obtained by the displacement detection device, and at least one temperature signal characterizing oil temperature are synchronously acquired, and the real-time effective temperature is determined by the acquired temperature signal. S3. Decoupling and Identification of Internal Leakage and Valve Leakage Characteristics of Dual Cylinders: Using the excitation signal as the model input and the collected pressure signal and slider synchronization error signal as the model output, a dual-input dual-output mathematical model describing the micro-dynamic characteristics of the hydraulic system in the pressure holding section is established. In this model, the internal leakage characteristics of the hydraulic cylinder are characterized as a nonlinear damping coefficient related to oil temperature, and the zero-position leakage characteristics of the servo valve are characterized as a zero-position leakage coefficient related to oil temperature. Using the input and output data, the internal leakage damping coefficient and valve zero-position leakage coefficient of cylinders Y1 and Y2 under the current temperature condition are estimated in real time through a recursive identification algorithm. S4. Dynamic update of dual-channel compensation curves: The two sets of internal leakage damping coefficients and valve zero-position leakage coefficients identified are associated with the corresponding real-time effective temperatures, and the cylinder internal leakage-oil temperature compensation curve and servo valve zero-position-oil temperature compensation curve are updated synchronously. S5. Decoupling feedforward compensation output: In each motion stage of the subsequent bending cycle, the updated compensation curve is queried based on the real-time effective temperature to obtain the decoupling feedforward compensation amount, and the feedforward compensation amount is superimposed on the position closed-loop control output of the corresponding servo valve.

2. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, In step S3, the dual-input dual-output mathematical model also introduces the real-time working pressure of the rodless chambers of cylinders Y1 and Y2 as input variables to construct a two-dimensional mapping model of "oil temperature-pressure difference-leakage", so that the identified internal leakage damping coefficient and valve zero-position leakage coefficient are pure temperature-dependent coefficients stripped of the current working pressure difference effect.

3. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, In step S1, before superimposing the micro-vibration excitation signal, the off-center load coefficient is calculated based on the real-time pressure difference between cylinder Y1 and cylinder Y2. When the off-center load coefficient exceeds a preset threshold, the amplitude of the excitation signal of the heavy-load side servo valve is increased, and the amplitude of the excitation signal of the light-load side servo valve is decreased.

4. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 3, characterized in that, The internal leakage damping coefficient identified in each step S3 is continuously recorded, and the long-term drift velocity of the internal leakage damping coefficient on one side is monitored. When the drift velocity exceeds the preset aging threshold, the high-frequency component bandwidth of the superimposed pseudo-random binary sequence is automatically expanded in the next execution of step S1. The upper limit of the expanded high-frequency component is 150-200Hz.

5. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, The real-time effective temperature mentioned in step S2 is determined in the following way: Temperature sensors are arranged in the oil tank, servo valve block, rodless cavity wall of cylinder Y1, and rodless cavity wall of cylinder Y2 to acquire multi-node temperature data. The cylinder is discretized into multiple micro-element control bodies along the stroke direction. The flow path of the oil micro-elements is tracked according to the real-time output flow rate of the servo valve and the cross-sectional area of ​​the cylinder. The equivalent average temperature of the oil participating in leakage in the rodless cavity is output as the real-time effective temperature.

6. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, The execution process of the recursive identification algorithm in step S3 includes: continuously recording the historical values ​​of the internal leakage damping coefficient of each cylinder obtained each time, and monitoring its long-term drift speed; when the drift speed of the internal leakage damping coefficient on one side exceeds the preset aging threshold, automatically reducing the forgetting factor of the recursive identification algorithm or increasing the process noise covariance to accelerate the model's tracking and convergence of the seal wear state.

7. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, During the first N pressure holding identification cycles after the equipment cold start, the recursive identification algorithm in step S3 uses a model update gain higher than that in the normal state; when the relative deviation of two consecutive identification results is lower than the preset convergence threshold, it automatically switches to normal update gain; where N is a preset positive integer, and the value range is 3 to 5.

8. The adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, In step S5, the decoupling feedforward compensation amount is weighted according to the following rules: During the rapid descent and return phases, the flow-temperature drift compensation component derived from the valve zero-position leakage coefficient is given primary weight, while the leakage compensation component derived from the internal leakage damping coefficient is given secondary weight. During the pressure holding stage, the leakage compensation component derived from the internal leakage damping coefficient is used as the main weight. During the bending stage, the flow rate temperature drift compensation component and the leakage compensation component are weighted and integrated, and the weighting coefficients are smoothly transitioned with the real-time speed of the slider.

9. The adaptive calibration method for the synchronization accuracy of the slider of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, Step S3 also includes an anomaly verification step: real-time calculation of the output residual of the identification model and monitoring of transient fluctuations in the acquired signal; when the output residual exceeds the statistical threshold, or the transient fluctuations of pressure or displacement exceed the preset safety threshold, the current parameter update is terminated, the compensation parameters of the previous valid identification cycle are retained, and an alarm signal is triggered.

10. The adaptive calibration method for slider synchronization accuracy of an electro-hydraulic bending machine based on dynamic oil temperature compensation according to claim 1, characterized in that, In step S1, the preset pressure threshold is 0.10 to 0.15 MPa, and the preset displacement threshold is 0.003 to 0.005 mm; the recursive identification algorithm is either the recursive least squares method or the extended Kalman filter algorithm.

Images