A three-layer progressive smooth control method and system for a hydraulic servo feed system

By employing a three-layer progressive smooth control method, combined with friction state observation, self-disturbance rejection control, and embedded active damping, the crawling and vibration problems of the hydraulic servo feed system under low-speed and heavy-load conditions are solved, achieving high-precision motion smooth control, which is suitable for the upgrading and transformation of CNC grinding machines.

CN123103525APending Publication Date: 2026-11-17NANJING NANTE PRECISE MASCH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611601572.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-16
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Hydraulic servo feed systems are prone to creeping and impact vibration during steady-state reversal under low-speed and heavy-load conditions. Existing technologies are difficult to effectively suppress these phenomena, and traditional methods have poor adaptive capabilities and cannot adjust damping characteristics in real time.

Method used

A three-layer progressive smooth control method is adopted, including a collaborative mechanism of friction state observation and feedforward compensation, active disturbance rejection control and embedded active damping. By estimating nonlinear friction force and system disturbance in real time, high-precision smooth control of the feed process is achieved.

Benefits of technology

It effectively eliminates crawling excitation, achieves robust position tracking and efficient residual vibration suppression, ensures smooth motion under all working conditions, reduces vibration marks and micro-unevenness on the machined surface, and is suitable for upgrading and retrofitting existing CNC grinding machines.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123103525A_ABST
    Figure CN123103525A_ABST
Patent Text Reader

Abstract

This invention discloses a three-layer progressive smooth control method and system for a hydraulic servo feed system. The method includes initialization and parameter setting; real-time data acquisition and preprocessing; state observation, vibration extraction, and friction state estimation, real-time estimation of the nonlinear friction force of the feed system to achieve the first layer of friction feedforward compensation processing; active damping force calculation and servo command mapping; obtaining the correction amount of additional active damping force and servo valve control current based on the system preset parameters and the vibration components of the feed system, and obtaining the final output control correction amount after amplitude limiting, as the third layer of smooth control; position loop active disturbance rejection control and friction feedforward compensation to generate the basic control amount, as the second layer of smooth control, the basic control amount is superimposed with the feedforward compensation amount estimated by the friction state observer to form the final basic control amount of the position loop; total control command synthesis and output; cyclic iteration to achieve high-precision smooth control of the feed process and suppress creeping and impact vibration.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of precision manufacturing technology for CNC machine tools, and in particular to a three-layer progressive smooth control method and system for a hydraulic servo feed system. Background Technology

[0002] The performance of a CNC grinding machine directly determines the final machining quality of precision parts. Gantry grinding machines require machining errors of less than 5 micrometers and surface roughness of less than 0.8 micrometers within a 1-meter length, placing extremely high demands on the smoothness and anti-interference capabilities of the feed system. Hydraulic servo feed systems are widely used due to their high rigidity and load capacity; however, the inherent low-damping characteristics of hydraulic systems make them prone to creeping at low speeds, and vibrations caused by hydraulic shocks during startup and reversal severely affect the machining accuracy and surface quality of the workpiece.

[0003] Traditional solutions often employ passive damping methods, such as adding throttles or accumulators to hydraulic pipelines for physical filtering. These methods have fixed parameters, poor adaptability, and cannot dynamically adjust damping characteristics according to real-time operating conditions, often at the cost of sacrificing system response speed. With the development of embedded technology, some controllers can achieve algorithm compensation, but still have the following shortcomings: (1) lack of embedded hardware-algorithm co-design for active damping at the bottom layer of hydraulic servo systems; (2) existing active damping schemes mainly focus on the suppression after vibration occurs, and have not yet observed and compensated for friction, the source of vibration excitation. It should be noted that one of the physical roots of low-speed crawling in hydraulic systems is the Stribeck effect of nonlinear friction—static friction is greater than dynamic friction, causing motion to alternate between "stickiness" and "sliding," resulting in intermittent nonlinear oscillations. Summary of the Invention

[0004] Technical Objective: To address the shortcomings of existing hydraulic servo systems in low-speed, heavy-load conditions such as crawling and impact vibration during steady-state reversal, this invention discloses a three-layer progressive smooth control method and system for hydraulic servo feed systems. Through a three-layer collaborative mechanism of friction state observation and feedforward compensation, active disturbance rejection control, and embedded active damping, high-precision smooth control of the feed process is achieved, suppressing crawling and impact vibration.

[0005] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution.

[0006] This invention discloses a three-layer progressive smooth control method for a hydraulic servo feed system, comprising the following steps:

[0007] Step S1: Initialization and parameter setting; obtain the system preset parameters, including the effective working area of ​​the hydraulic cylinder, the system's expected damping ratio, stiffness coefficient, mass coefficient, and the initial parameters of the Stribeck friction model;

[0008] Step S2: Real-time data acquisition and preprocessing; real-time acquisition of operating parameters, including displacement signal, rodless cavity pressure, rod cavity pressure and vibration acceleration signal, at a preset sampling frequency;

[0009] Step S3: State observation, vibration extraction, and friction state estimation; Based on the received displacement and acceleration signals, the actual motion velocity and vibration components of the feed system are calculated in real time using the state observer; The nonlinear friction force of the feed system is estimated in real time using the friction state observer based on the improved Stribeck model, and used as a feedforward compensation quantity for the position loop control in step S5, realizing the first layer of friction feedforward compensation processing; The four parameters in the improved Stribeck model are identified and updated using the recursive least squares method, including Coulomb friction force, maximum static friction force, Stribeck velocity, and viscous friction coefficient;

[0010] Step S4: Calculate the active damping force and map the servo command; obtain the correction amount of the additional active damping force and the servo valve control current based on the system preset parameters and the vibration components of the feed system; perform amplitude limiting processing on the correction amount to obtain the final output control correction amount, which serves as the smooth control of the third layer.

[0011] Step S5: Position loop active disturbance rejection control and friction feedforward compensation; The position closed-loop controller adopts an active disturbance rejection control algorithm, which generates a basic control quantity through a tracking differentiator, an extended state observer, and a nonlinear state error, as the smooth control of the second layer. The basic control quantity is superimposed with the feedforward compensation quantity estimated by the friction state observer in step S3 to form the final basic control quantity of the position loop.

[0012] Step S6: Synthesis and output of the overall control command; The final basic control quantity of the position loop calculated in step S5 is superimposed with the active damping correction quantity calculated in step S4 to form the overall control command;

[0013] Step S7: Iterate repeatedly; repeat steps S2 to S6, in each sampling period, acquire signals in real time, update state observations, calculate active damping force, and refresh the overall control command to form closed-loop control until the feeding process ends.

[0014] This invention discloses a three-layer progressive smooth control system for a hydraulic servo feed system, used to implement the three-layer progressive smooth control method for a hydraulic servo feed system described above, including a hydraulic actuator, a signal acquisition unit, an embedded active damping controller, and a hydraulic servo valve group;

[0015] The hydraulic actuator includes a hydraulic cylinder and a piston rod that drives the hydraulic cylinder, the piston rod being fixedly connected to the worktable of the CNC grinding machine;

[0016] The signal acquisition unit includes a pressure sensor group and an acceleration sensor. The pressure sensor group is used to acquire real-time pressure data of the rodless chamber and the rod chamber of the hydraulic cylinder. The acceleration sensor is installed on the worktable and is used to acquire vibration acceleration signals of the feed system.

[0017] A hydraulic servo valve assembly includes a servo valve and a hydraulic pipeline connecting an oil source and a hydraulic cylinder. The servo valve is used to control the flow rate and direction of the oil entering the hydraulic cylinder.

[0018] An embedded active damping controller is electrically connected to the signal acquisition unit and the hydraulic servo valve group, respectively; it is used to receive the raw signal output by the signal acquisition unit and perform filtering and amplification processing; it executes a three-layer progressive smooth control method to output control commands, thereby controlling the valve core opening of the servo valve.

[0019] Beneficial Effects: This invention constructs a three-layer progressive smooth control architecture: "First-layer friction feedforward compensation (source vibration reduction) → Second-layer self-disturbance suppression (process vibration damping) → Third-layer embedded active damping absorption (residual vibration absorption)". The three-layer mechanism operates sequentially according to the vibration energy transmission path, forming a complete suppression chain from vibration cause to vibration result. Compared with existing technologies, it has the following outstanding effects:

[0020] (1) Eliminate crawling excitation at the source. By identifying the Stribeck model parameters online through the friction state observer, the nonlinear friction force is estimated in real time and compensated in a feedforward manner, directly eliminating the excitation source of the "viscous-slipping" crawling phenomenon caused by the difference between static and dynamic friction forces, thus making up for the shortcomings of existing active damping schemes that only "treat the symptoms but not the root cause".

[0021] (2) Robust position tracking. Active disturbance rejection control replaces traditional PID control. By using an extended state observer to estimate and compensate for the total disturbance of the system (including load changes, oil temperature drift, elastic modulus changes, etc.) in real time, high-precision tracking of position commands can be achieved without the need for an accurate mathematical model. This is especially advantageous under heavy load and variable operating conditions.

[0022] (3) Highly efficient residual vibration suppression. The embedded active damper module is embedded in the microprocessor in the form of software code. It generates active damping force in real time through a virtual "mass-spring-damping" dynamic model, and can achieve differentiated suppression of vibrations at specific frequencies without modifying the existing mechanical structure.

[0023] (4) Ensuring Smooth Motion Under All Working Conditions. The three-layer mechanism works in concert to systematically ensure the smoothness of the feed system's motion under all working conditions from three dimensions: "source vibration reduction," "disturbance suppression," and "real-time vibration absorption." This significantly reduces the vibration marks and micro-irregularities on the machined surface, providing a reliable guarantee for achieving a surface roughness Ra<0.8μm. This design requires no additional physical components, is low in cost, highly reliable, and easy to upgrade on existing CNC grinding machines. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the hardware structure of the system of the present invention;

[0025] Figure 2 This is a flowchart of the control method of the present invention;

[0026] Figure 3 This is a schematic diagram of the overall functional modules of the smooth control method for the hydraulic servo feed system of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0028] Example 1: As shown in the attached document Figure 1 To be continued Figure 3 As shown in this embodiment, a three-layer progressive smooth control method and system for a hydraulic servo feed system achieves high-precision smooth control of the feed process and suppresses crawling and impact vibration through a three-layer collaborative mechanism of friction state observation and feedforward compensation, active disturbance rejection control, and embedded active damping.

[0029] The three-layer progressive smooth control architecture of this invention is as follows:

[0030] The first layer is the friction feedforward compensation layer: the Stribeck model parameters are identified online through the friction state observer, the nonlinear friction force is estimated in real time and injected into the position loop in a feedforward manner, and the excitation source of the crawling phenomenon is eliminated from the physical level.

[0031] The second layer is the active disturbance rejection layer: the total system disturbance (load change, oil temperature drift, etc.) is estimated and compensated in real time by the extended state observer, and the smoothing of the command is combined with the tracking differentiator to achieve highly robust tracking of position commands and suppress the influence of external and internal uncertainties on motion smoothness.

[0032] The third layer is an embedded active damping absorption layer: through a virtual "mass-spring-damping" model embedded in the microprocessor, additional active damping force is calculated based on the real-time extracted vibration components and superimposed on the control command to quickly absorb residual vibration energy.

[0033] The three-layer mechanism works collaboratively according to the progressive logic of "source elimination → process suppression → residual absorption", covering the frictional origin domain, disturbance effect domain and vibration response domain of the feeding system respectively, forming a complete smooth control link.

[0034] This embodiment discloses a three-layer progressive smooth control method for a hydraulic servo feed system, including the following steps:

[0035] Step S1: Initialization and parameter setting;

[0036] Upon system power-on, the embedded microprocessor 14 performs a self-test and reads system preset parameters stored in the memory, including the effective working area of ​​the hydraulic cylinder and the system's desired damping ratio. Stiffness coefficient quality coefficient The initial parameters of the Stribeck friction model, where the effective working area of ​​the hydraulic cylinder includes the effective working area of ​​the rodless chamber. Effective working area of ​​the rod cavity The initial parameters of the Stribeck friction model include Coulomb friction force. Maximum static friction Stribeck speed viscous friction coefficient .

[0037] Step S2: Real-time data acquisition and preprocessing;

[0038] The signal acquisition unit collects operating parameters, including displacement signals, in real time at a preset sampling frequency. rodless chamber pressure Rod chamber pressure and vibration acceleration signals Among them, displacement signal Obtained through the worktable of a CNC grinding machine, such as through a machine tool linear encoder.

[0039] The signal conditioning circuit 13 performs hardware filtering on the acquired raw signal to remove high-frequency noise interference; the embedded microprocessor 14 performs analog-to-digital conversion on the processed signal to obtain a digital quantity.

[0040] Step S3: State observation, vibration extraction, and friction state estimation;

[0041] The embedded microprocessor 14 receives the displacement signal. Using the acceleration signal 'a', the actual velocity of the feed system is calculated in real time through the built-in state observer algorithm. and vibration components .

[0042] The embedded microprocessor 14 has a built-in state observer that employs a frequency-segmented sensor fusion algorithm. Specifically, to overcome the limitations of single-sensor measurements, the microprocessor fuses the displacement signal... After low-pass filtering and differentiation, the low-frequency velocity characterizing macroscopic motion is obtained. Simultaneously, the acceleration signal After high-pass filtering and integration, the high-frequency velocity characteristic of the dynamic response is obtained. Subsequently, and By superimposing and fusing the data in the time domain, a precise and phase-lag-free actual feed rate can be obtained, which is the actual motion rate of the feed system. In this embodiment, a frequency-segmented sensor fusion algorithm is implemented by dividing the frequency bands of the low-frequency and high-frequency speed channels, i.e., by setting the frequency filtering of low-pass and high-pass filters.

[0043] Using a digital bandpass filter (BPF) with an adjustable center frequency, the fused speed signal is processed. Filtering is performed. The passband of this bandpass filter is set according to the structural resonant frequency of the grinding machine feed system. After bandpass filtering, the macroscopic feed motion component is filtered out, and the output is the vibration velocity component, which characterizes the micro-vibration of the system. .

[0044] To further evaluate system stability, the microprocessor performs a short-time Fourier transform on the original acceleration signal *a*. Using a sliding time window, the dominant frequency *ω* of the vibration at the current moment and its corresponding energy amplitude *E* are calculated in real time. *E* is then compared with a preset stability threshold. Compare. When E exceeds When the system is about to become unstable, the active damping enhancement mode is triggered, and the subsequent damping control parameters are dynamically adjusted, which are steps S4 to S6.

[0045] Meanwhile, the friction state observer built into the embedded microprocessor 14, based on an improved Stribeck model, estimates the nonlinear friction force of the feed system in real time. The model expression is:

[0046]

[0047] in, Coulomb friction, For the maximum static friction force, Let σ be the Stribeck velocity and σ be the coefficient of viscous friction. These four parameters are identified and updated online using the recursive least squares (RLS) method to adapt to frictional characteristic drift caused by changes in oil temperature and guideway lubrication conditions. The estimated frictional force... It will be used as a feedforward compensation amount in the position loop control of step S5 to eliminate the "sticky-slipping" crawling phenomenon caused by the difference between static and dynamic friction from a physical perspective.

[0048] The above four parameters are identified and updated online using the recursive least squares (RLS) method, and the recursive process is as follows:

[0049] Step 1: Prediction

[0050]

[0051] Step 2: Calculate the prediction error

[0052]

[0053] Step 3: Calculate the gain

[0054]

[0055] Step 4: Update parameters

[0056]

[0057] Step 5: Update covariance

[0058] covariance matrix

[0059] in, The parameter vector to be identified, For the regression vector, The measured driving force at the current moment is λ∈(0.95, 1], which is the forgetting factor used to balance the weights of historical and current data to accommodate the drift in friction characteristics caused by changes in oil temperature and guide rail lubrication conditions. k represents the current cycle, and k-1 represents the previous cycle. Predict the driving force for the current moment;

[0060] The physical meaning of the above recursive formula is: within each control cycle, the algorithm first uses the parameters estimated in the previous cycle. and the current regression vector Predict the current driving force and compare it with the measured driving force. The prediction error is obtained by comparison; then, based on the current covariance matrix... Calculate the gain from the regression vector. This gain determines the contribution of the current error to parameter correction; finally, the parameters are corrected using this gain, and the covariance matrix is ​​updated to reduce the uncertainty of parameter estimation. The introduction of the forgetting factor λ causes the weight of historical data to decay over time, enabling the algorithm to track the slow drift of friction characteristics caused by changes in oil temperature and lubrication conditions.

[0061] Estimated friction force This will be used as the feedforward compensation amount in the position loop control of step S5. To avoid overcompensation under high-speed conditions, a compensation enable condition is set: when... ( When the speed threshold is set to 10 mm / s, Effective injection location loop; when At that time, the friction feedforward compensation gradually decreases to zero.

[0062] After the first layer of friction feedforward compensation, the system eliminates the main excitation source of creep at the physical level. However, external disturbances such as load changes and oil temperature drift, as well as hydraulic shocks during the reversing process, may still induce residual vibrations. Therefore, this invention further suppresses dynamic disturbances during the feed process through a second layer of active disturbance rejection control, while simultaneously using embedded active damping in a third layer to rapidly absorb residual vibration energy. Steps S4 and S5 correspond to the specific implementations of the third and second layers, respectively.

[0063] Step S4: Calculate the active damping force and map the servo command; obtain the correction amount of the additional active damping force and the servo valve control current based on the system preset parameters and the vibration components of the feed system; perform amplitude limiting processing on the correction amount to obtain the final output control correction amount;

[0064] Step S4.1 Calculate the virtual damping force;

[0065] The embedded active damper module extracts the vibration velocity components in step S3. Simulate the dynamic characteristics of a "mass-spring-damped" mechanical system and calculate the additional active damping force required to suppress vibration. Its core algorithm model is:

[0066]

[0067] in, This is a virtual mass coefficient used to suppress sudden changes in system acceleration; This is the active damping coefficient, used to absorb vibration energy; This is a virtual stiffness coefficient used to suppress position drift. This is the initial additional active damping force.

[0068] The above , , The three virtual model parameters are tuned as follows: based on the equivalent second-order dynamic model of the grinding machine feed system, with the system's desired damping ratio... For the objective, the pole placement method is used to determine The initial value; Based on the equivalent conversion of the total mass of the workbench and load. The parameters are set based on the desired system stiffness and position loop bandwidth. Once tuned, the three parameters remain constant during the feed process and are only recalibrated after major system maintenance or when the workpiece type is changed.

[0069] Calculated initial additional active damping force It also needs to undergo saturation limiting treatment to obtain the additional active damping force: ,in The preset maximum additional damping force limit is used to prevent excessive damping force commands from causing servo valve saturation or system instability.

[0070] Step S4.2 Feedforward compensation based on force / pressure difference conversion; obtain the current pressure of the rodless chamber and rod chamber of the hydraulic cylinder, adopt a control strategy of adjusting one chamber and holding the other chamber, obtain the actual output additional damping force, and use real-time compensation to ensure that the actual output additional damping force tracks the additional active damping force.

[0071] In a hydraulic system, the actual force output by the piston rod Determined by the pressure in both chambers:

[0072]

[0073] To achieve the additional damping force calculated in step S4.1 A corresponding additional pressure difference needs to be established between the two chambers of the hydraulic cylinder. According to the force balance equation of the hydraulic cylinder:

[0074]

[0075] in, , These represent the changes in pressure in the rodless chamber and the rod chamber relative to the current pressure, respectively. , This represents the effective working area of ​​the two cavities.

[0076] Because the system is equipped with pressure sensors, it can obtain the current pressure of the two chambers in real time. , To achieve rapid response and avoid control coupling, a control strategy of one chamber for regulation and another for holding is adopted. It is assumed that the rod chamber is connected to the return oil, and the pressure is essentially constant. The additional pressure difference mainly acts on the rodless cavity:

[0077]

[0078] Therefore, to achieve the desired additional damping force, the target pressure of the rodless cavity needs to be set as follows:

[0079]

[0080] Although it is assumed in the control strategy To simplify calculations, the above derivation assumes a constant rod cavity pressure. However, in actual operation, the rod cavity pressure... The pressure is monitored in real time by the second pressure sensor 8, and compensation and correction are performed in the pressure feedback closed loop according to the following formula: The formula for calculating the actual output additional damping force is as follows:

[0081]

[0082] in, The reference pressure for the rod cavity, The target pressure in the rodless chamber is given. When the pressure in the rod chamber fluctuates, real-time compensation is performed using the above formula to ensure that the actual output force accurately tracks the desired additional damping force. .

[0083] In addition, the target pressure in this step The calculation constitutes the "feedforward path", while the pressure closed-loop controller in step S4.3 constitutes the "feedback path". Together, they form a composite control structure of feedforward + feedback, which can significantly improve the response speed and tracking accuracy of damping force compared to simple feedback control.

[0084] Step S4.3: Servo valve control quantity calculation; compare the target pressure of the rodless chamber with the current pressure of the rodless chamber to obtain the pressure deviation signal; after the pressure deviation signal is input into the pressure closed-loop controller, the incremental PID control algorithm with integral separation strategy is used to output the correction amount of the servo valve control current;

[0085] The target pressure obtained in step S4.2 Compared with the current measured pressure By comparison, the pressure deviation signal is obtained:

[0086]

[0087] Where k is the discrete sampling time.

[0088] Deviation signal Input pressure closed-loop controller. The pressure closed-loop controller adopts an incremental PID control algorithm with integral separation strategy, and its output is the correction amount of the servo valve control current. The specific calculation process is as follows:

[0089] First, calculate the proportional, integral, and derivative control components respectively:

[0090]

[0091]

[0092]

[0093] in, , , The proportional, integral, and derivative coefficients of the pressure closed-loop controller are tuned using system identification and pole placement methods. The preset pressure deviation threshold is used. When the deviation exceeds the threshold, the integral term is cut off (i.e., integral separation) to prevent large overshoot of the control quantity caused by integral saturation. At the same time, PD control is used to achieve fast response. When the deviation falls back to within the threshold, the complete PID control is restored to eliminate steady-state error.

[0094] The above three components are summed to obtain the initial control correction:

[0095]

[0096] Finally, to prevent hydraulic shock caused by severe vibration of the servo valve spool, [the following measures are taken]. The rate of change is limited to obtain the final output control correction value. :

[0097]

[0098] in, The preset maximum rate of change limit is used to ensure that the change in control command between two adjacent sampling cycles does not exceed this threshold, thereby ensuring the smoothness of the servo valve core movement.

[0099] The active damping force calculation in step S4 relies on the real-time extraction of vibration components, while the position loop control in step S5 provides the basic driving force for the macroscopic feed of the system. The two are calculated in parallel within the control cycle and finally superimposed for output, ensuring the smoothness of the system's motion from the two time scales of "macroscopic position tracking" and "microscopic vibration suppression".

[0100] To prevent significant overshoot of the control quantity caused by integral saturation, this pressure closed-loop controller adopts an integral separation strategy: when ( When the pressure deviation threshold is preset, the integral term coefficient is... Set to zero, using only PD control for fast response; when At that time, full PID control is restored to eliminate steady-state error. Furthermore, It also needs to be subject to rate of change limiting: This is to prevent the servo valve core from vibrating violently.

[0101] The active damping force calculation in step S4 relies on the real-time extraction of vibration components, while the position loop control in step S5 provides the basic driving force for the macroscopic feed of the system. The two are calculated in parallel within the control cycle and finally superimposed for output, ensuring the smoothness of the system's motion from the two time scales of "macroscopic position tracking" and "microscopic vibration suppression".

[0102] Step S5: Position loop active disturbance rejection control and friction feedforward compensation; The position closed-loop controller adopts an active disturbance rejection control algorithm, which generates a basic control quantity through a tracking differentiator, an extended state observer, and a nonlinear state error, as the smooth control of the second layer. The basic control quantity is superimposed with the feedforward compensation quantity estimated by the friction state observer in step S3 to form the final basic control quantity of the position loop.

[0103] This step uses an Active Disturbance Rejection Control (ADRC) algorithm for the position closed-loop controller, replacing the traditional PID control, to achieve strong robust suppression of internal and external disturbances such as nonlinear friction and load changes. Specifically, it includes the following three parts:

[0104] (1) Tracking Differentiator (TD): For external position commands Smoothing is performed, and a reasonable transition process is arranged to avoid the impact caused by sudden changes in instructions. Its discrete form is:

[0105]

[0106]

[0107] in, To track signals, It is a differential signal. For velocity factor, For the filter factor, This is the fastest synthesis function.

[0108] (2) Extended State Observer (ESO): Using the actual control input u and the actual position X as inputs, it estimates the system's state variables (displacement estimate z1, velocity estimate z2) and total disturbance estimate z3 (including internal and external disturbances such as nonlinear friction, load changes, and changes in the elastic modulus of the oil) in real time. Its discrete form is:

[0109]

[0110]

[0111]

[0112] in, , , , For observer gain, As a compensation factor, It is a nonlinear function. , , This is an adjustable parameter.

[0113] The above nonlinear function The function is defined as:

[0114]

[0115] Where δ is the width of the linear interval and α∈(0,1) is the nonlinear factor, which is an adjustable parameter. This function uses the nonlinear characteristics of "large error, small gain; small error, large gain" to improve the convergence speed while suppressing high-frequency oscillations.

[0116] The observer gains β1, β2, and β3 are tuned according to the following rules: Let the observer bandwidth be ω0 (generally ω0 = 3~5 times the bandwidth of the control system), then β1 = 3ω0, β2 = 3ω0², and β3 = ω0³. Increasing ω0 can improve the disturbance estimation speed of the observer, but an excessively large ω0 will amplify the measurement noise, and a trade-off must be made based on the signal-to-noise ratio of the actual system.

[0117] (3) Nonlinear State Error Feedback (NLSEF): Based on the error between the tracking signal and the differential signal output by TD and the state estimated by ESO, a basic control quantity is generated. :

[0118]

[0119] in, and It is the feedback gain of NLSEF. and yes The nonlinear factor of the function;

[0120] Based on this, the frictional force feedforward compensation amount estimated by the frictional state observer in step S3 is superimposed. This forms the final basic control quantity of the position loop:

[0121]

[0122] in, For NLSEF output, This represents the friction feedforward gain coefficient. As a compensation factor, For the total disturbance compensation item, This is a feedforward term for the friction force specific project.

[0123] Compared with traditional PID, ADRC does not require an accurate mathematical model of the system. It estimates and compensates for the total disturbance of the system in real time through ESO. Combined with the targeted suppression of friction feedforward, it has stronger robustness to parameter perturbations and external disturbances. It is especially suitable for complex working conditions such as load (workpiece weight) changes and hydraulic system characteristic drift during the grinding machine feed process.

[0124] It should be noted that the friction feedforward compensation and the disturbance observation compensation of ADRC are not redundant, but complementary and synergistic: the friction feedforward (first layer) actively compensates for the modeled Stribeck friction force with well-defined physical laws, exhibiting phase lead characteristics and the ability to "predictively" eliminate the effects of friction; while the extended state observer of ADRC (second layer) is responsible for estimating and compensating for all unmodeled dynamics and external disturbances (such as load abrupt changes, oil temperature drift, unmodeled nonlinearities, etc.) other than the modeled friction. The clear division of labor avoids the estimation burden of the extended state observer on large-amplitude friction disturbances, thereby improving the overall disturbance rejection performance and transient response speed of the system.

[0125] Step S6: Synthesis and output of overall control commands;

[0126] The position loop basic control quantity calculated in step S5 The control correction amount of the final output calculated in step S4.3 The commands are superimposed to form the overall control command:

[0127]

[0128] Among them, the final basic control quantity of the position loop The friction feedforward compensation amount of the first layer has been included. The second layer of ADRC control quantity .

[0129] The overall control command is converted into an analog drive current via the drive circuit (15), which acts on the valve core of the servo valve (10) to regulate the flow rate into the hydraulic cylinder. Through closed-loop regulation, the actual pressure in the rodless chamber is adjusted. Real-time tracking of target pressure This allows for the precise superposition of damping force into the output force of the hydraulic cylinder. This enables active suppression of vibration.

[0130] Step S7: Iterate through the loop;

[0131] Repeat steps S2 to S6, in real time acquiring signals, updating state observations, calculating active damping force, and refreshing the overall control command in each sampling period to form closed-loop control until the feeding process ends.

[0132] This embodiment also discloses a three-layer progressive smooth control system for a hydraulic servo feed system, including a hydraulic actuation unit, a signal acquisition unit, an embedded active damping controller (12), and a hydraulic servo valve group. The above units work together to provide a hardware support platform for the first-layer friction feedforward compensation, the second-layer self-disturbance suppression, and the third-layer embedded active damping absorption of this invention.

[0133] The hydraulic actuator includes a hydraulic cylinder 1 and a piston rod 2 that drives the hydraulic cylinder 1. The piston rod 2 is fixedly connected to the worktable 3 of the CNC grinding machine.

[0134] The signal acquisition unit includes a pressure sensor group 4 and an acceleration sensor 9. The pressure sensor group 4 includes a first pressure sensor 7 and a second pressure sensor 8, which are respectively installed on the oil lines of the rodless chamber 5 and the rod chamber 6 of the hydraulic cylinder 1, for acquiring real-time pressure data of the two chambers. The acceleration sensor 9 is installed on the worktable 3 for acquiring vibration acceleration signals of the feed system.

[0135] The hydraulic servo valve assembly includes a servo valve 10 and a hydraulic pipeline connecting the oil source 11 and the hydraulic cylinder 1. The servo valve 10 is used to control the flow rate and direction of the hydraulic cylinder 1.

[0136] An embedded active damping controller 12 is electrically connected to the signal acquisition unit and the hydraulic servo valve group, respectively. The embedded active damping controller 12 includes a signal conditioning circuit 13, an embedded microprocessor 14, and a drive circuit 15. The signal conditioning circuit 13 receives the raw signal output from the signal acquisition unit and performs filtering and amplification processing. The embedded microprocessor 14 has a built-in non-volatile memory storing a friction state observation algorithm, an active disturbance rejection control algorithm, and an active damping control algorithm. The friction state observation algorithm includes a friction state observer, and the active disturbance rejection control algorithm includes an extended state observer. The three algorithms correspond to the calculation tasks of the first layer of friction feedforward compensation, the second layer of active disturbance rejection suppression, and the third layer of embedded active damping absorption, respectively. The drive circuit 15 receives the control commands output by the embedded microprocessor 14 and converts them into drive current to control the valve core opening of the servo valve 10.

[0137] The embedded active damping controller 12 also includes an embedded active damper module. This module is not an independent mechanical damper, but rather a piece of embedded code embedded in the embedded microprocessor 14, which works in conjunction with the processor hardware to form a virtual "mass-spring-damping" active adjustment system.

[0138] Embedded microprocessor 14 with a fixed sampling period =200μs (corresponding to a control frequency of 5kHz) to execute the control algorithm. Within each control cycle, the timing allocation of the three-layer algorithm is as follows: signal acquisition and analog-to-digital conversion are triggered at the beginning of the cycle (0μs); the first-layer friction state observation and the second-layer ADRC state estimation are completed within 20μs (sharing state observation data); the third-layer embedded active damping force calculation and the second-layer ADRC control quantity calculation are executed in parallel, completed within 80μs and 100μs respectively; the first-layer friction feedforward compensation quantity is superimposed after the ADRC calculation is completed, and the control command synthesis and D / A output are completed before the end of the cycle (180μs), reserving 20μs as an execution margin.

[0139] Example 2: This example details the specific steps of implementing the three-layer progressive smooth control method based on the above system. (See reference...) Figure 2 The flowchart and Figure 3 The algorithm principle block diagram is shown below. In this embodiment, a gantry CNC surface grinder is used as an example. Its table travel is 2 meters, the maximum workpiece weight is 3 tons, the target machining accuracy is an error of less than 5 micrometers within a 1-meter length, and the surface roughness is... <0.8μm. Sampling period of the control system. The setting is 200μs, which means the control frequency is 5kHz.

[0140] Step S1: Initialization and parameter setting;

[0141] After the grinding machine performs a power-on self-test, the embedded microprocessor (14) reads the system preset parameters from the memory. For the gantry grinding machine in this embodiment, the key preset parameters are as follows:

[0142] The effective working area of ​​the rodless chamber of the hydraulic cylinder is: = π×(0.04m)² ≈ 5.0265×10⁻³ m²;

[0143] The effective working area of ​​the rod chamber of the hydraulic cylinder is: = π×(0.04² - 0.0225²) ≈ 3.468×10⁻³m²;

[0144] Virtual quality coefficient = 50 kg, used to suppress sudden changes in system acceleration;

[0145] Active damping coefficient = 5000 N·s / m, used to absorb vibration energy;

[0146] Virtual stiffness coefficient = 1×10 5 N / m, used to suppress position drift;

[0147] Initial parameters for the Stribeck friction model: Coulomb friction force = 500 N, maximum static friction force = 1200 N, Stribeck speed = 2 mm / s, coefficient of viscous friction = 800 N·s / m;

[0148] Stability region energy threshold = 0.01 m² / s², this value was determined based on previous modal analysis experiments;

[0149] In the pressure loop PID coefficient, the proportional coefficient = 2.5 × 10⁻ 7 Integral coefficient = 5×10⁻ 9 Differential coefficients = 1×10⁻ 7 ;

[0150] In the ADRC parameters, the observer bandwidth ω0 = 150 rad / s (taken as 3 times the control system bandwidth), and the velocity factor... =100, filter factor = 0.01, nonlinear factors α1 = 0.5, α2 = 0.25, α3 = 0.5, α4 = 0.25, linear interval width δ = 0.01, compensation factor b0 = 1, friction feedforward gain coefficient = 0.8;

[0151] Step S2: Real-time data acquisition and preprocessing;

[0152] Within each 200μs sampling period, the controller synchronously acquires the following signals through the signal acquisition unit:

[0153] The displacement signal X comes from the machine tool linear encoder and has a resolution of 0.1 μm;

[0154] Pressure signals P1 and P2 come from the first pressure sensor 7 and the second pressure sensor 8;

[0155] The acceleration signal a comes from the acceleration sensor 9 installed on the worktable 3.

[0156] The signal conditioning circuit 13 first performs hardware filtering on these raw signals, for example, using a high-order Butterworth low-pass filter with a cutoff frequency of 500Hz to remove high-frequency noise interference. Subsequently, the built-in ADC of the embedded microprocessor 14 converts the conditioned analog signal into a 24-bit digital value for subsequent algorithm calls.

[0157] Step S3: State observation, vibration extraction, and friction state estimation;

[0158] The embedded microprocessor 14 receives the displacement signal. and acceleration signal The actual speed of the feed system is calculated in real time using the built-in state observer algorithm. and vibration components .like Figure 3 As shown on the left, the state observer uses a sensor fusion algorithm based on frequency segmentation:

[0159] The displacement signal X is passed through a 5th-order IIR low-pass filter with a cutoff frequency of 20Hz to filter out high-frequency measurement noise, and then backward differential is performed to obtain the low-frequency velocity characterizing the macroscopic motion. Simultaneously, the acceleration signal 'a' is passed through a second-order IIR high-pass filter with a cutoff frequency of 15Hz to remove the DC component and low-frequency drift before integration to obtain the high-frequency velocity characterizing the dynamic response. Subsequently, and The actual feed rate is obtained by superimposing and fusing the data in the time domain. .

[0160] The fused velocity signal V is filtered using a digital bandpass filter (BPF) with an adjustable center frequency. Based on the modal analysis of the grinding machine's table-hydraulic cylinder system, its dominant resonant frequency is approximately 40Hz; therefore, the BPF's center frequency is set to 40Hz, and its bandwidth to 10Hz. After bandpass filtering, the macroscopic feed motion component is removed, and the output is the vibration velocity component characterizing the system's micro-vibrations. .

[0161] To further evaluate system stability, the microprocessor performs a short-time Fourier transform on the original acceleration signal 'a'. Using a sliding time window, the dominant frequency ω = 40Hz and its corresponding energy amplitude E are calculated in real time. It is assumed that at a certain commutation moment, the calculated value of E is 0.05 m² / s², exceeding the preset stability threshold. = 0.01 m² / s². At this point, the system determines that it is about to become unstable and immediately triggers the active damping enhancement mode, providing a basis for subsequent dynamic adjustment of damping control parameters.

[0162] Meanwhile, the friction state observer built into the embedded microprocessor 14, based on an improved Stribeck model, estimates the nonlinear friction force of the feed system in real time. The model expression is:

[0163]

[0164] At a certain sampling moment in this embodiment, assuming the actual velocity V = 5 mm / s, the current estimated friction parameter value is... = 480 N, = 1150 N, = 2.1 mm / s, = 820 N·s / m, then the calculation yields:

[0165]

[0166] = [480 + 670·e^(-5.67)]·1 + 4.1

[0167] ≈ (480 + 670×0.0035) + 4.1 ≈ 486.4 N

[0168] The four parameters mentioned above are identified and updated online using the recursive least squares (RLS) method. In this embodiment, the forgetting factor λ = 0.98 and the initial value of the covariance matrix P(0) = 10 are taken. 6 Let I (where I is the identity matrix), the initial parameter vector θ(0) = [500, 1200, 2.0, 800]ᵀ, and the regression vector. Taking the current sampling time as an example, if φ = [-0.0035, 1, 0.005]ᵀ, the predicted output ŷ(k) = φᵀ·θ(k-1), and the measured driving force y(k) = P1·A1 -P2·A2, then the prediction error ε(k), gain L(k), and updated parameter vector θ(k) can be calculated, realizing the online update of friction parameters cycle by cycle.

[0169] To avoid overcompensation of feedforward compensation under high-speed conditions, a compensation enable condition is set: when hour, Effective injection location loop; when At this time, the friction feedforward compensation gradually decreases to zero. In this embodiment, the current velocity V = 5 mm / s, which is less than 10 mm / s, therefore... = 486.4 N will be injected as feedforward compensation into the position loop control in step S5.

[0170] Step S4: Calculation of active damping force and mapping of servo commands;

[0171] S4.1 Calculation of virtual damping force;

[0172] The embedded active damper module extracts the vibration velocity components in step S3. Simulate the dynamic characteristics of a "mass-spring-damped" mechanical system and calculate the additional active damping force required to suppress vibration. Its core algorithm model is:

[0173]

[0174] Assuming at the current sampling time k, the calculated = 0.005 m / s, its differential = 1.5 m / s², integral = 0.0002 m. Substituting the preset parameters, the calculated additional active damping force is:

[0175] = 50 × 1.5 + 5000 × 0.005 + 1 × 10 5 × 0.0002 = 75 + 25 + 20 =120 N

[0176] Calculated additional active damping force Saturation limiting processing is applied. This embodiment is preset. = 500 N, since 120 N does not exceed the amplitude limit, therefore Maintain at 120N.

[0177] S4.2 Feedforward compensation based on force / pressure difference conversion;

[0178] In a hydraulic system, the actual force output by the piston rod Determined by the pressure in both chambers:

[0179]

[0180] To achieve the calculated additional damping force A corresponding additional pressure difference needs to be established between the two chambers of the hydraulic cylinder. According to the force balance equation:

[0181]

[0182] Because the system is equipped with pressure sensors, it can obtain the current pressure of the two chambers in real time. , To achieve rapid response and avoid control coupling, this embodiment employs a control strategy of one chamber for regulation and one chamber for holding. Let... The additional pressure difference mainly acts on the rodless cavity:

[0183]

[0184] Read the current rodless chamber pressure = 8.5 MPa, rod chamber pressure = 2.0 MPa. Therefore, to achieve the desired additional damping force, the target pressure of the rodless cavity needs to be set as follows:

[0185]

[0186] It should be noted that the above derivation assumes a constant rod cavity pressure. However, in actual operation, the rod cavity pressure... The pressure is monitored in real time by the second pressure sensor 8, and compensation and correction are performed in the pressure feedback closed loop according to the following formula:

[0187]

[0188] in This is the reference pressure for the rod-shaped cavity. In this embodiment... = 2.0 MPa. When the pressure in the rod chamber fluctuates, the above formula is used for real-time correction to ensure that the actual output force accurately tracks the expected additional damping force. .

[0189] S4.3 Servo valve control quantity calculation;

[0190] The target pressure obtained in step S4.2 Compared with the current measured pressure By comparison, the pressure deviation signal is obtained:

[0191]

[0192]

[0193] Deviation signal Input pressure closed-loop controller. The pressure closed-loop controller adopts an incremental PID control algorithm with integral separation strategy, and its output is the correction amount of the servo valve control current. The specific calculation process is as follows:

[0194] First, calculate the proportional, integral, and derivative control components respectively:

[0195]

[0196]

[0197]

[0198] In this embodiment, the proportional coefficient of the pressure closed-loop controller = 2.5 × 10⁻ 7 Integral coefficient = 5×10⁻ 9 Differential coefficients = 1×10⁻ 7 Pressure deviation threshold = 30000 Pa. Assuming the current time... = 23870 Pa, previous moment = 20000 Pa, all less than Therefore, the integral term is valid.

[0199] The above three components are summed to obtain the initial control correction:

[0200]

[0201] In this embodiment, a maximum single-step rate of change limit is preset. = 0.05V. Assuming the control correction value from the previous cycle... = 0.25V, the initial value obtained from the current calculation. = 0.30V, then the rate of change is +0.05V, which just reaches the limit value. After the rate of change is limited, the final output control correction is... = 0.30V, to ensure the smooth operation of the servo valve spool.

[0202] Step S5: Position loop active disturbance rejection control and friction feedforward compensation;

[0203] This step uses an Active Disturbance Rejection Control (ADRC) algorithm in the position closed-loop controller, replacing the traditional PID control. Assume the current position command is input from the external CNC. = 100.000 mm, actual position = 99.998 mm.

[0204] (1) Tracking Differentiator (TD) for External Position Commands Smoothing is performed. Its discrete form is:

[0205]

[0206]

[0207] In this embodiment, = 100, = 0.01, and after TD smoothing, the tracking signal V1 and the differential signal V2 are obtained.

[0208] (2) The Extended State Observer (ESO) takes the actual control input u and the actual position X as inputs and estimates the system's state variables and total disturbance in real time:

[0209]

[0210]

[0211]

[0212] In this embodiment, the observer bandwidth ω0 = 150 rad / s, then β1 = 3ω0 = 450, β2 = 3ω0² = 67500, β3 = ω0³ = 3375000, and the compensation factor b0 = 1.

[0213] (3) Nonlinear State Error Feedback (NLSEF) generates a basic control quantity u0 based on the error between the tracking signal and the differential signal output by TD and the state estimated by ESO:

[0214]

[0215]

[0216] In this embodiment, β 10 = 100, β 20 = 20, α3 = 0.5, α4 = 0.25, δ = 0.01.

[0217] Based on this, the frictional force feedforward compensation amount estimated by the frictional state observer in step S3 is superimposed. This forms the final basic control quantity of the position loop:

[0218]

[0219] In this embodiment, step S3 estimates the following: = 486.4 N, = 0.8, assuming the current time = 300N, = 2.0V, then:

[0220] = 2.0 - (300 + 486.4×0.8) / 1 = 2.0 - (300 + 389.1) = 2.0 -689.1 = -687.1 (dimensionless control quantity)

[0221] Step S6: Synthesis and output of overall control commands;

[0222] The position loop basic control quantity calculated in step S5 The control correction amount of the final output calculated in step S4.3 The commands are superimposed to form the overall control command:

[0223]

[0224] In this embodiment, The corresponding analog output is 5.0V. = 0.30V, then:

[0225] = 5.0V + 0.30V = 5.30V

[0226] Among them, the position loop basic control quantity The friction feedforward compensation amount of the first layer has been included. The second layer of ADRC control quantity In this embodiment, the analog contribution corresponding to the friction feedforward compensation is (486.4 × 0.8) / 1 × ≈ 0.15V, where (As the control quantity to voltage conversion coefficient), the ADRC control quantity contribution is approximately 4.85V, and the final output control correction quantity is 0.30V. The three layers of control quantities are superimposed in the final instruction for output.

[0227] The overall control command is converted into an analog drive current by the drive circuit 15, which acts on the valve core of the servo valve 10 to fine-tune the valve core opening, allowing more pressurized oil to enter the rodless chamber and push the piston rod to output greater force. Through this closed-loop regulation, the actual pressure in the rodless chamber is adjusted. Real-time tracking of target pressure This allows an additional damping force of 120N to be precisely superimposed on the output force of the hydraulic cylinder. This enables active suppression of vibration.

[0228] Thus, the three-layer progressive smooth control of this invention completes the coordinated operation of one control cycle: the first layer (friction feedforward compensation in step S3) eliminates the excitation source of crawling at the physical level, reducing the disturbance burden on subsequent control layers; the second layer (active disturbance rejection control in step S5) suppresses the influence of external disturbances and system uncertainties on motion smoothness at the macroscopic position tracking level; the third layer (embedded active damping in step S4) rapidly absorbs residual vibration energy at the microscopic vibration level, ensuring the smoothness of the output force. The three mechanisms act on the friction origination domain, disturbance effect domain, and vibration response domain respectively, forming a complementary relationship in the frequency domain—friction compensation mainly targets low-frequency stick-slip phenomena, ADRC suppresses mid-frequency disturbances, and active damping targets high-frequency resonance components, jointly achieving a systematic guarantee of motion smoothness across the entire frequency band of the feed system.

[0229] Step S7: Iterate

[0230] The system repeats steps S2 to S6 in 200μs cycles, acquiring signals, updating state observations, calculating active damping force, and refreshing the overall control command in real time during each sampling cycle to form a closed-loop control until the feeding process ends.

[0231] In the embodiments of this application, the terms "first" and "second" (if they exist) are used only as name identifiers and do not represent the order of first and second.

[0232] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. The components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment solution according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0233] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A three-layer progressive smooth control method for a hydraulic servo feed system, characterized in that: Includes the following steps: Step S1: Initialization and parameter setting; obtain the system preset parameters, including the effective working area of ​​the hydraulic cylinder, the system's expected damping ratio, stiffness coefficient, mass coefficient, and the initial parameters of the Stribeck friction model; Step S2: Real-time data acquisition and preprocessing; real-time acquisition of operating parameters, including displacement signal, rodless cavity pressure, rod cavity pressure and vibration acceleration signal, at a preset sampling frequency; Step S3: State observation, vibration extraction, and friction state estimation; Based on the received displacement and acceleration signals, the actual motion velocity and vibration components of the feed system are calculated in real time using the state observer; The nonlinear friction force of the feed system is estimated in real time using the friction state observer based on the improved Stribeck model, and used as a feedforward compensation quantity for the position loop control in step S5, realizing the first layer of friction feedforward compensation processing; The four parameters in the improved Stribeck model are identified and updated using the recursive least squares method, including Coulomb friction force, maximum static friction force, Stribeck velocity, and viscous friction coefficient; Step S4: Calculate the active damping force and map the servo command; obtain the correction amount of the additional active damping force and the servo valve control current based on the system preset parameters and the vibration components of the feed system; perform amplitude limiting processing on the correction amount to obtain the final output control correction amount, which serves as the smooth control of the third layer. Step S5: Position loop active disturbance rejection control and friction feedforward compensation; The position closed-loop controller adopts an active disturbance rejection control algorithm, which generates a basic control quantity through a tracking differentiator, an extended state observer, and a nonlinear state error, as the smooth control of the second layer. The basic control quantity is superimposed with the feedforward compensation quantity estimated by the friction state observer in step S3 to form the final basic control quantity of the position loop. Step S6: Synthesis and output of the overall control command; The final basic control quantity of the position loop calculated in step S5 is superimposed with the final output control correction quantity calculated in step S4 to form the overall control command; Step S7: Iterate repeatedly; repeat steps S2 to S6, in each sampling period, acquire signals in real time, update state observations, calculate active damping force, and refresh the overall control command to form closed-loop control until the feeding process ends.

2. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: Based on the received displacement and acceleration signals, the actual velocity and vibration components of the feed system are calculated in real time by the state observer, including: The state observer employs a frequency-segmented sensor fusion algorithm to calculate the actual motion velocity and vibration components of the feeding system. Specifically, the displacement signal is differentiated after low-pass filtering to obtain the low-frequency velocity; the acceleration signal is integrated after high-pass filtering to obtain the high-frequency velocity; the low-frequency and high-frequency velocities are superimposed and fused in the time domain to obtain the actual motion velocity of the feeding system; and the vibration components are obtained by filtering the actual motion velocity using a digital bandpass filter.

3. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: The nonlinear friction force is used as a feedforward compensation quantity for the position loop control in step S5. It also includes setting compensation enable conditions: when the actual motion speed is less than the speed threshold, the nonlinear friction force is effectively injected into the position loop; when the actual motion speed is greater than or equal to the speed threshold, the nonlinear friction force gradually decays to zero.

4. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: Step S4 includes: Step S4.1, Virtual damping force calculation; Calculate the additional active damping force based on the vibration components obtained in step S3; Step S4.2: Feedforward compensation based on force / pressure difference conversion; obtain the current pressure of the rodless chamber and rod chamber of the hydraulic cylinder, calculate the target pressure of the rodless chamber; adopt a control strategy of adjusting one chamber and holding the other chamber to obtain the actual output additional damping force, and use real-time compensation to ensure that the actual output additional damping force tracks the additional active damping force. Step S4.3: Servo valve control quantity calculation; compare the target pressure of the rodless chamber with the current pressure of the rodless chamber to obtain the pressure deviation signal; after the pressure deviation signal is input into the pressure closed-loop controller, the incremental PID control algorithm with integral separation strategy is used to output the correction amount of the servo valve control current; the correction amount is limited to obtain the final output control correction amount.

5. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 4, characterized in that: The formulas for calculating the additional active damping force include: ; in, This is a virtual quality coefficient. This is the active damping coefficient. This is the virtual stiffness coefficient; For the initial additional active damping force, For vibration components; After saturation limiting, the additional active damping force is obtained: ,in This is the preset maximum additional damping force limit value.

6. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 4, characterized in that: The formula for calculating the target pressure of the rodless cavity includes: ; in, The current pressure in the rodless cavity. To add active damping force, The effective working area of ​​the rodless cavity; The formulas for calculating the additional damping force in actual output include: ; in, The reference pressure for the rod cavity, The current pressure in the rod chamber. This represents the effective working area of ​​the rod cavity.

7. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: The tracking differentiator smooths the external position pointer and calculates the tracking signal and the differential signal; The extended state observer takes the actual control input and actual position as inputs and estimates the system's state variables and total disturbance in real time; the state variables include displacement and velocity estimates. The nonlinear state error feedback generates a basic control quantity based on the error between the tracking signal and the differential signal output by the tracking differentiator and the state estimated by the extended state observer. This basic control quantity is then superimposed with the feedforward compensation quantity estimated by the friction state observer in step S3 to form the final basic control quantity of the position loop.

8. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: The formulas for calculating the final basic control values ​​in the position loop include: ; in, The output is the nonlinear state error feedback. This represents the friction feedforward gain coefficient. As a compensation factor, For the total disturbance estimate, This is friction.

9. The three-layer progressive smooth control method for a hydraulic servo feed system according to claim 1, characterized in that: The calculation formula for the general control command includes: ; in, This is the final basic control variable for the position loop. This is the control correction amount for the final output.

10. A three-layer progressive smooth control system for a hydraulic servo feed system, used to implement the three-layer progressive smooth control method for a hydraulic servo feed system as described in any one of claims 1-9, characterized in that: It includes a hydraulic actuator, a signal acquisition unit, an embedded active damping controller (12), and a hydraulic servo valve assembly; The hydraulic actuator includes a hydraulic cylinder (1) and a piston rod (2) that drives the hydraulic cylinder (1), the piston rod (2) being fixedly connected to the worktable (3) of the CNC grinding machine; The signal acquisition unit includes a pressure sensor group (4) and an acceleration sensor (9). The pressure sensor group (4) is used to acquire real-time pressure data of the rodless chamber (5) and the rod chamber (6) of the hydraulic cylinder (1). The acceleration sensor (9) is installed on the worktable (3) and is used to acquire vibration acceleration signals of the feed system. The hydraulic servo valve assembly includes a servo valve (10) and a hydraulic line connecting an oil source (11) and a hydraulic cylinder (1). The servo valve (10) is used to control the flow rate and direction of the oil entering the hydraulic cylinder (1). An embedded active damping controller (12) is electrically connected to the signal acquisition unit and the hydraulic servo valve group respectively; it is used to receive the raw signal output by the signal acquisition unit and perform filtering and amplification processing; it executes the three-layer progressive smooth control method to output control commands, thereby controlling the valve core opening of the servo valve (10).