A multi-axis equipment automatic coordination control method for intelligent manufacturing

Abstract

The application discloses a kind of multi-axis equipment automation coordination control methods for intelligent manufacturing, and the application relates to the technical field of industrial automation control, contains three cooperation modules: multi-physical field strong coupling error online decoupling and real-time compensation module, heat drift disturbance and cutting dynamic disturbance are taken into joint state space framework as extended state, adaptive extended state observer is designed, online decoupling compensation of multi-physical field coupling error is realized, face flexible production change free calibration multi-axis coordination adaptive reconfiguration module, submillimeter level multi-frequency sinusoidal detection signal is injected in the idle gap of production change, recursive least square method with forgetting factor is used to identify dynamics parameters online, feedback control quantity, disturbance feedforward compensation quantity and variable gain coordination quantity are superimposed to synthesize control instruction, the multi-axis linkage profile error of the application is reduced by more than 50% compared with prior art, and production change calibration downtime is shortened to within 10 seconds.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to a method for automated coordinated control of multi-axis equipment for intelligent manufacturing. Background Technology

[0002] Precision coordinated control of multi-axis CNC equipment is a core foundational technology for intelligent manufacturing equipment. The closest existing technical solution to this invention is a multi-axis cross-coupled coordinated control system based on a fixed model. This solution consists of a CNC interpolation module, independent PID position controllers for each axis, a fixed-gain cross-coupled controller, a thermal error compensation module, and a geometric error compensation module, with each module operating independently.

[0003] In the existing solution, the entire stroke of the equipment is calibrated using a laser interferometer at the factory to generate a geometric error compensation table. A static thermal error compensation model is established using a temperature sensor group. The above data is then stored in the CNC system. During the machining process, the independent PID controllers of each axis calculate the control quantities of each axis based on the actual position feedback from the encoder. The fixed gain cross-coupled controller calculates the contour error in real time and outputs a coordinated compensation quantity that is superimposed on the control input of each axis. The thermal error compensation quantity and the geometric error compensation quantity are independently superimposed on the reference instruction by the lookup table module. When changing production, if the quality and processing technology of the new workpiece are significantly different from the original, the machine must be stopped and a complete set of laser interferometry calibrations must be performed again. The controller gain must be manually adjusted by the engineer before production can continue.

[0004] The aforementioned existing technology has the following three technical defects:

[0005] First, the independent compensation architecture makes it impossible to accurately eliminate coupling errors. In actual cutting, thermal errors, geometric errors and cutting dynamic errors influence and superimpose each other, exhibiting strong nonlinear coupling characteristics. They are not independent error sources. Existing solutions model each error separately and linearly superimpose them, ignoring the cross-coupling terms between errors. When the cutting force surges, it is impossible to separate high-frequency instantaneous structural deformation and low-frequency thermal drift online, resulting in inaccurate compensation calculations. The machining contour error cannot converge to the ideal level, making it difficult to meet the micron-level machining accuracy requirements of aerospace and high-end equipment manufacturing.

[0006] Second, the fixed-gain cross-coupled controller cannot balance strong robustness and high synchronization accuracy. The coordination gain is manually set before machining begins and remains unchanged throughout the process. When a sudden heavy cutting causes a large disturbance to a certain axis, a small gain leads to insufficient coordination compensation capability and increased contour error, while a large gain introduces overcompensation during the stable machining phase, causing system oscillation and reducing following accuracy. The fixed-gain scheme cannot simultaneously achieve strong robustness and high synchronization accuracy across the entire working range, and there is an inherent contradiction between the two.

[0007] Third, the strong dependence on calibration leads to long downtime during production changeovers, which severely restricts the efficiency of flexible production. Changes in workpiece quality and fixture configuration after production changeovers cause significant changes in the equivalent inertia matrix, resulting in a serious mismatch between the original control law parameters and the actual system. After production changeovers, it is necessary to stop the machine and recalibrate it using a laser interferometer. This process usually takes 2 to 8 hours. In addition, the time for manual parameter tuning greatly reduces the effective production time of the production line, leading to a serious decline in the overall efficiency of the equipment. To address this, we propose an automated coordinated control method for multi-axis equipment in intelligent manufacturing. Summary of the Invention

[0008] The purpose of this invention is to provide an automated coordinated control method for multi-axis equipment in intelligent manufacturing.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an automated coordination control method for multi-axis equipment in intelligent manufacturing, applicable to multi-axis equipment with n linkage axes. The method includes three functional modules that cooperate with each other: an online decoupling and real-time compensation module for strong coupling errors in multi-physics fields, a calibration-free multi-axis coordinated adaptive reconfiguration module for flexible production changeover, and a nonlinear variable gain global synchronization control law execution module.

[0010] The collaborative relationship among the three functional modules is as follows: The calibration-free multi-axis coordinated adaptive reconfiguration module is triggered and activated during the non-cutting transition phase of equipment changeover. The dynamic parameters obtained online are respectively passed to the multi-physics field strong coupling error online decoupling and real-time compensation module and the nonlinear variable gain global synchronous control law execution module to complete parameter initialization. During the machining process, the multi-physics field strong coupling error online decoupling and real-time compensation module continuously runs the adaptive extended state observer to decouple thermal drift disturbances and cutting dynamic disturbances online. It passes the disturbance estimates and tracking error state quantities after decoupling of each axis to the nonlinear variable gain global synchronous control law execution module. The nonlinear variable gain global synchronous control law execution module dynamically adjusts the coordination gain according to the real-time load disturbance intensity of each axis, synthesizes the control torque commands of each axis and sends them to the servo drive end. The encoder and current sensor data of the servo drive end are simultaneously fed back to the multi-physics field strong coupling error online decoupling and real-time compensation module, forming a complete closed loop.

[0011] The multi-physics field strong coupling error online decoupling and real-time compensation module integrates the position tracking error, velocity tracking error, thermal drift disturbance component, and cutting dynamic disturbance component of each axis into a joint state space framework to construct a joint state space model. Based on this, an adaptive extended state observer is designed, using the L2 norm of the information vector as the driving signal. The observer gain matrix is ​​adaptively adjusted through a Sigmoid-type smooth saturation function. Taking advantage of the significant separation characteristics of thermal drift disturbance and cutting dynamic disturbance in the frequency domain, the decoupling estimates of the two types of disturbances are extracted by low-pass digital filters and high-pass digital filters, respectively.

[0012] The calibration-free multi-axis coordinated adaptive reconfiguration module injects multi-frequency sinusoidal superposition detection displacement signals with sub-millimeter amplitudes into each axis during the non-cutting transition phase of production changeover. It uses a recursive least squares method with a forgetting factor to identify the dynamic parameters of the multi-axis system online, and automatically reconfigures the control law parameters required by each module after convergence.

[0013] The nonlinear variable gain global synchronous control law execution module calculates the normalized load disturbance intensity index of each axis based on the cutting dynamic disturbance estimate output by the adaptive extended state observer. Based on this index, it adjusts the coordination gain of each axis in real time according to the linear proportional amplification rule, and combines the independent feedback control quantity, disturbance feedforward compensation quantity and variable gain cross-coupling coordination quantity to synthesize the control command of each axis.

[0014] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0015] 1. This invention constructs a joint state-space model and introduces an adaptive extended state observer to achieve precise online decoupling and synchronous compensation of thermal error, geometric error and cutting dynamic error under cutting conditions. This fundamentally eliminates the coupling residuals introduced by the independent compensation architecture, and reduces the contour error of multi-axis linkage machining by more than 50% compared with the prior art.

[0016] 2. This invention proposes a nonlinear variable gain global synchronization control law based on the normalized load disturbance intensity index. Under sudden load conditions, it automatically enhances the coordinated compensation gain of the disturbed axis and reduces the gain under stable machining conditions to improve the following accuracy of each axis, thus completely solving the inherent contradiction that strong robustness and high synchronization accuracy cannot be achieved at the same time.

[0017] 3. This invention proposes a calibration-free multi-axis coordinated adaptive reconfiguration mechanism for flexible production changeover. By utilizing the no-load gap during production changeover, the latest dynamic parameters are identified online and the control law is automatically reconfigured within 10 seconds, reducing the calibration downtime of several hours to less than 10 seconds, thus significantly improving the overall equipment efficiency of intelligent production lines. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall process in an embodiment of the present invention;

[0019] Figure 2 This is a schematic diagram of the independent process of the calibration-free adaptive reconfiguration module in the production changeover stage in this embodiment of the invention;

[0020] Figure 3 This is a schematic diagram of the normal single-cycle closed-loop control process in an embodiment of the present invention. Detailed Implementation

[0021] The following is in conjunction with the appendix Figure 1-3The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0022] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0023] I. System Overall Architecture:

[0024] The automated coordination control method proposed in this invention adopts a three-layer hierarchical control architecture of cloud-edge-device, which deploys computing tasks with different real-time requirements on corresponding computing power platforms to form a complete closed-loop control system.

[0025] The end layer is a servo drive layer, which includes a multi-axis servo drive group, a high-resolution encoder, a current sensor and a temperature sensor array. Each end layer device is connected to the edge layer via an EtherCAT bus with a communication cycle of 125μs to 1ms. The edge layer is a high real-time industrial controller with an FPGA / DSP coprocessing module, which deploys an adaptive extended state observer core computing module and a variable gain coordinated control law execution module with a control cycle of 125μs to 500μs. The cloud layer is the main controller, which deploys a thermal error evolution slow dynamic model update module, a calibration-free online identification and parameter reconstruction module and a global task scheduling module with a refresh cycle of 10ms to 100ms. The end layer uploads raw sensor data to the edge layer, the edge layer sends control torque commands for each axis to the end layer, the edge layer uploads filtered state estimates to the cloud layer, and the cloud layer sends updated model parameters and gain matrices to the edge layer. The overall design reflects the hierarchical design concept of high-frequency control at the edge and low-frequency optimization in the cloud.

[0026] The core technical solution of this invention consists of three mutually cooperating functional modules that operate in a closed loop: a multi-physics field strong coupling error online decoupling and real-time compensation module, a calibration-free multi-axis coordinated adaptive reconfiguration module for flexible production changeover, and a nonlinear variable gain global synchronous control law execution module. The cooperative relationship of the three modules is as follows: the calibration-free multi-axis coordinated adaptive reconfiguration module is responsible for identifying the latest dynamic parameters during the production changeover interval and transmitting the parameters to the first two modules to complete the initialization. During the processing, the multi-physics field strong coupling error online decoupling and real-time compensation module continuously runs an adaptive extended state observer to perform error decoupling and disturbance estimation, and inputs the decoupled disturbance state quantity into the nonlinear variable gain global synchronous control law execution module. The nonlinear variable gain global synchronous control law execution module dynamically adjusts the control gain of each axis according to the current load state, generates control commands, and feeds back the state quantity of each axis to the multi-physics field strong coupling error online decoupling and real-time compensation module to form a closed loop.

[0027] II. Online Decoupling and Real-time Compensation Module for Strong Coupling Errors in Multiphysics Fields:

[0028] (I) Establishment of the joint state-space model:

[0029] Consider a machine with For multi-axis equipment with multiple linkage axes, the position error, speed error, thermal drift disturbance component, and cutting dynamic disturbance component of each axis are uniformly incorporated into a joint state space framework, defining... Joint state vector of the shaft system for:

[0030] ;

[0031] in, For the first Shaft position tracking error, in meters. , For the first Axis reference desired position, For the first Actual measured position of the shaft For the first The speed tracking error of the shaft, in meters per second. For the first Thermal drift disturbance component of the shaft, in meters. For the first The cutting dynamic disturbance component of the shaft, in meters, superscript Indicates transpose. ;

[0032] Based on engineering assumptions: geometric error The axis position is a slowly varying function, approximately constant over one control cycle, and can be incorporated into the thermal drift disturbance in the design of an adaptive extended state observer. When processed together, the continuous-time form of the joint state-space model is:

[0033] ;

[0034] ;

[0035] in, The system state matrix is ​​structured by the dynamic equations of each axis (including inter-axis coupling inertia terms), and the 2×2 main diagonal blocks of each axis sub-block are... The self-decay term of the thermal perturbation substate reflects the thermal time constant. To control the input matrix, the control torque of each axis is mapped to the corresponding state equation. Input torque vectors for each axis control. The perturbation input matrix, For unmodeled perturbation vectors, For the dimension of the unmodeled perturbation, To observe the output matrix, encoder position and driver current are selected as measurable outputs. For measurable output vectors, To measure the noise vector, The measurable output dimension;

[0036] (II) Design of Adaptive Extended State Observer:

[0037] The core idea of ​​the adaptive extended state observer is to incorporate thermal drift perturbations. and cutting dynamic disturbance As two extended states with significantly different frequency band characteristics, the thermal drift disturbance generally has a frequency below 0.01 Hz, while the cutting dynamic disturbance generally has a frequency above 10 Hz. The two have obvious separation characteristics in the frequency domain, which is the physical basis for realizing online decoupling.

[0038] Discrete-time form of the adaptive extended state observer (sampling period is) )for:

[0039] ;

[0040] in, for At any given time, the joint state vector The estimated value, and These are continuous state matrices. and The corresponding matrix after discretization by the zero-order hold, for The adaptive observer gain matrix at time step 1. To output the predicted residuals (information), drive the observer state correction;

[0041] To ensure that the adaptive extended state observer maintains good tracking performance for both low-frequency thermal disturbances and high-frequency cutting disturbances, the following measures are taken: Adopting a frequency band adaptive adjustment strategy:

[0042] ;

[0043] in, For the innovation vector, The L2 norm of the innovation vector is used to measure the current system disturbance strength. The slope parameter is pre-tuned. This is a saturation threshold determined by the system sensor noise level, used to prevent over-excitation of the gain caused by noise. To ensure the fundamental stability of the observer, the fundamental gain matrix is ​​solved using the pole placement method. This is the additional gain matrix for the high-frequency disturbance channel, with the main non-zero elements corresponding to the cutting dynamic disturbance state. The observation equations, As a natural constant, when the cutting force increases abruptly, the new innovation 2-norm... As the signal increases rapidly, the tracking bandwidth of the observer for high-frequency disturbances automatically increases. When the processing enters a stable phase, the signal decreases and the gain falls back to the baseline value, thus avoiding amplifying sensor noise.

[0044] (III) Execution steps of a single control cycle for the adaptive extended state observer:

[0045] In the At the start of each control cycle, the edge controller synchronously acquires all data via the EtherCAT bus. absolute position signal of the shaft encoder ( The driver phase current signal (converted to equivalent axial torque via Clarke-Park transformation) (and temperature sensor readings located at key thermal nodes of the machine tool spindle box, guide rail slide, and worktable) ( , (The number of temperature sensors is used to perform mean filtering on the raw data with a window length of 3 to 5 sampling points to remove impulse noise, obtaining smoothed measurement values, and constructing a measurable output vector.) ;

[0046] Using the state estimate obtained at the end of the previous control cycle As a priori estimate for the current period, the information vector is calculated:

[0047] ;

[0048] And calculate the L2 norm of the new information. As a real-time quantitative indicator of the current system disturbance intensity, the observer gain matrix is ​​updated according to the above-mentioned Sigmoid-type band adaptive strategy based on the new innovation norm 2. Execute the discrete state update equation of the adaptive extended state observer and calculate the first... Posterior state estimator of step:

[0049] ;

[0050] In this equation, For prior predictions based on a known model, The correction term is based on the observed residuals, and the two together constitute the posterior estimate;

[0051] from Based on the block structure defined by the state vector, the thermal drift perturbation estimate of each axis is extracted separately. and cutting dynamic disturbance estimate To further purify the frequency components of the two types of perturbations, The output of a first-order low-pass digital filter with a cutoff frequency of 0.1Hz is used to... The decoupled thermal disturbance estimate is obtained by passing the output of a first-order high-pass digital filter with a cutoff frequency of 5Hz. and dynamic disturbance estimates The data is then passed to the nonlinear variable gain global synchronization control law execution module for feedforward compensation calculation. After the current control cycle ends... As the initial estimate for the next cycle, the algorithm returns the sensor data acquisition steps and executes them cyclically.

[0052] III. Calibration-free multi-axis coordinated adaptive reconfiguration module for flexible production changeover:

[0053] (I) Design of production change triggering and detection signals:

[0054] When the production line scheduling system issues a changeover command to the multi-axis equipment, the calibration-free multi-axis coordinated adaptive reconfiguration module is activated. During the subsequent clamping / tool ​​changing transition phase (when the equipment is in a no-cutting, no-load motion state), this module automatically intervenes to execute the online identification process without stopping the machine for special laser interference calibration.

[0055] The excitation signal is identified using a multi-frequency sinusoidal superposition micro-amplitude test trajectory, with the displacement signal detected along the i-th axis. Designed as follows:

[0056] ;

[0057] in, This represents the number of superimposed frequencies, ranging from 5 to 10. For the first The amplitude of each frequency component is in the sub-millimeter range (0.05mm to 0.5mm) to ensure no interference with the workpiece and fixture. For the first Each frequency component is logarithmically uniformly selected from the device's motion bandwidth (range 1Hz to 50Hz). For the first Axis No. The initial phase of each frequency component is staggered between axes to satisfy the decoupling identification requirements for simultaneous multi-axis excitation, i.e., to satisfy the continuous excitation condition. The time unit is seconds. The detection signal is superimposed on the idle jogging trajectory during the production transition phase, and the duration is... For 5 to 10 seconds;

[0058] (ii) Online identification of recursive least squares with forgetting factor:

[0059] During the application of the detection signal, with a sampling period Continuously acquire position output of each axis Speed ​​output (Calculated from position difference) and actual driving torque The matrix form of the dynamic equations of a multi-axis system is:

[0060] ;

[0061] in, The inertia matrix, Here is the damping matrix. Here is the stiffness matrix. For each axis driving torque vector, the parameter vector to be identified is... Defined as , , The column vectors of each independent element are used to construct a regression matrix with acceleration, velocity, and position measurements along each axis as elements. The equation is transformed into standard linear parameterized form:

[0062] ;

[0063] in, The residual vector represents the acceleration. It is calculated from the second-order difference of the position signal and smoothed by the Savitzky-Golay filter to suppress noise in the differential amplification;

[0064] Online estimation of parameter vectors using recursive least squares with forgetting factor The recursive formula is as follows:

[0065] ;

[0066] ;

[0067] ;

[0068] in, For the first The Kalman gain matrix of the step, The total number of parameters to be identified. To estimate the error covariance matrix, the initial value is set to... , Large positive numbers (values) This reflects the uncertainty of initial parameters. The forgetting factor (ranging from 0.97 to 0.999) controls the rate at which the algorithm forgets historical data. for identity matrix Initialize to the parameter estimate stored after the last production change, and initialize to the factory nominal value on the first run;

[0069] (III) Identifying convergence judgments:

[0070] The convergence criterion for parameter estimation is defined as: continuity Within each control cycle (take) The relative rate of change of the estimated values ​​of each parameter component is less than the preset threshold. :

[0071] ;

[0072] in, For the first The parameter components in the first... The estimated value of the step, To prevent small positive numbers with a denominator of zero (values) ), The convergence threshold (ranging from 0.5% to 1%) is used. If the convergence condition is met, the identification phase ends, and the current value is extracted. Reconstructing the inertia matrix and damping matrix If the identification time exceeds If the problem still does not converge, the previous valid identification result will be retained and an alarm signal will be issued. The problem will be re-identified during the next production change.

[0073] (iv) Automatic reconfiguration of control laws:

[0074] After convergence, the results obtained from the identification are used. and Automatically refactor the parameters of each module:

[0075] Update the system matrix of the joint state-space model in the online decoupling and real-time compensation module for strong multiphysics coupling errors. and The submatrix blocks that depend on the inertia parameter are replaced with those based on and The new value is calculated, and the fundamental gain matrix is ​​resolved using the pole placement method. ;

[0076] by Based on this, the initial gain matrix of the cross-coupled control law in the nonlinear variable gain global synchronization control law execution module is recalculated. The calculation method is as follows:

[0077] ;

[0078] in, This is the reference gain matrix under the factory-specified operating conditions. The factory-specified inertia matrix is ​​used to write the reconstructed parameters into the register of the corresponding control module. A parameter update completion signal is sent, the production line resumes normal processing mode, and the calibration-free multi-axis coordinated adaptive reconstruction module enters a dormant state to wait for the next production changeover trigger.

[0079] IV. Execution Module of Nonlinear Variable Gain Global Synchronization Control Law:

[0080] (I) Overall structure of the control law:

[0081] No. The axis's total control input It consists of three superimposed parts:

[0082] ;

[0083] in, For the first The independent feedback control quantity of the axis is achieved using PD control. This is the perturbation feedforward compensation amount based on the decoupling results of the adaptive extended state observer. This is a variable gain cross-coupling coordination control quantity;

[0084] (ii) Independent feedback control items:

[0085] ;

[0086] in, For the first The proportional gain of the axis, For the first The differential gain of both axes is automatically updated by the calibration-free multi-axis coordinated adaptive reconfiguration module during post-production parameter reconfiguration. and All data are taken from the state estimation output of the adaptive extended state observer of the multi-physics field strong coupling error online decoupling and real-time compensation module, and its noise filtering effect is better than that of direct difference calculation.

[0087] (iii) Disturbance feedforward compensation term:

[0088] ;

[0089] in, Inertia matrix The Middle The equivalent inertia of the shaft (i.e.) The Line number (Column elements), used to convert the equivalent position disturbance output by the adaptive extended state observer into an equivalent compensation torque. The values ​​are thermal disturbance estimates extracted using a low-pass digital filter with a cutoff frequency of 0.1 Hz. The dynamic disturbance estimate is extracted by a 5Hz cutoff frequency high-pass digital filter. This feedforward term achieves active cancellation of multi-physics coupling disturbances.

[0090] (iv) Nonlinear variable gain cross-coupling coordination term:

[0091] First, based on the position errors of each axis unit vector of the tangent direction to the current reference trajectory Calculate the Cartesian space contour error vector :

[0092] ;

[0093] in, For each axis position error vector, The following error (scalar) is the tracking error along the trajectory direction. The contour error component is perpendicular to the trajectory direction. Provided in real time by the interpolation module;

[0094] Secondly, the dynamic disturbance estimate output by the multiphysics strong coupling error online decoupling and real-time compensation module is used. Calculate the normalized load disturbance intensity index for each axis :

[0095] ;

[0096] in, Characterizing the first The current degree of disturbance of the axis relative to The normalized ratio of the axis most severely disturbed among the axes. To prevent small positive numbers with a denominator of zero;

[0097] based on Define a nonlinear adaptive adjustment rule for the coordination gain of each axis:

[0098] ;

[0099] in, For the first The cross-coupled basic gain (scalar) of the axis, reconstructed by the calibration-free multi-axis coordinated adaptive reconstructing module, corresponds to the tuning value under stable machining conditions. The variable gain amplification factor (ranging from 1 to 5) determines the amplification factor of the coordination gain relative to the base value under the most severe disturbance conditions. When the shaft is subjected to the strongest disturbance, Approaching 1, Zoom in This doubles the coordination and compensation capability of the axis, when all axes are free from significant disturbances. As it approaches zero, the gain falls back to the base value. To avoid overcompensation that could introduce oscillations;

[0100] The variable gain cross-coupling coordinated control quantity is calculated as follows:

[0101] ;

[0102] in, Contour error vector The The component, i.e., the first component The contribution component of the shaft to the total profile error;

[0103] (v) Single control cycle execution steps:

[0104] Read the estimated position error values ​​of each axis output by the multiphysics strong coupling error online decoupling and real-time compensation module in the current cycle. and speed error estimate Calculate the independent PD feedback control quantity for each axis according to the independent feedback control term formula. PD gain , The value is taken from the register value written after the most recent production changeover reconfiguration of the calibration-free multi-axis coordinated adaptive reconfiguration module;

[0105] Read the decoupled perturbation estimates of each axis output by the online decoupling and real-time compensation module for strong multiphysics coupling errors. and The equivalent inertia identified by the calibration-free multi-axis coordinated adaptive reconfiguration module. Calculate the disturbance feedforward compensation amount for each axis according to the disturbance feedforward compensation term formula. This enables the active cancellation of thermal drift and cutting dynamic disturbances;

[0106] Read the unit tangent direction vector at the current trajectory reference point from the interpolation module. Construct the position error vector Calculate the Cartesian contour error vector at the current moment according to the contour error formula, and extract the components of each axis. Using the calculated Calculate the normalized load disturbance intensity index for each axis using the normalized load disturbance intensity index formula. And calculate the adaptive variable gain coordination coefficient of each axis at the current moment according to the coordination gain adaptive adjustment rule. ;

[0107] Combine the control components according to the total control quantity formula. Axis control variables, control variables for all axes ( The data is synchronously sent to the corresponding servo driver via the EtherCAT bus, driving the servo motor to perform motion. After completing one control cycle, the algorithm returns to the sensor data acquisition step.

[0108] V. Overall System Control Closed-Loop Flow:

[0109] The overall control closed-loop process of the system is divided into two major stages: the production change triggering stage and the normal processing stage.

[0110] Changeover Trigger Phase: When a changeover command arrives, the calibration-free multi-axis coordinated adaptive reconfiguration module execution process is initiated. Within 10 seconds, calibration-free online identification and parameter reconfiguration are completed. After reconfiguration, the updated parameters are written into the parameter registers of the multi-physics field strong coupling error online decoupling and real-time compensation module and the nonlinear variable gain global synchronous control law execution module. A parameter update completion signal is issued, and the production line immediately enters the normal processing phase. When the changeover command has not arrived, the system continues to operate in the normal processing phase.

[0111] During normal machining: the interpolator outputs reference position commands for each axis in real time. and reference trajectory tangent vector The multi-physics field strong coupling error online decoupling and real-time compensation module sequentially performs sensor data acquisition, information calculation, adaptive gain update, state observation and disturbance decoupling, and outputs thermal disturbance estimate. Dynamic disturbance estimates Position error and speed error The nonlinear variable gain global synchronization control law execution module sequentially performs feedback control, disturbance feedforward, contour error calculation, variable gain coordination, and control quantity synthesis, outputting control torque commands for each axis. The control input is sent to the servo motor via the EtherCAT bus. The actual position output by the servo motor is fed back to the multi-physics field strong coupling error online decoupling and real-time compensation module via the encoder, forming a complete digital control closed loop. The control cycle of the large cycle during normal processing is 125μs to 500μs, and it continues to run until the next production change trigger.

[0112] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. An automated coordinated control method for multi-axis equipment in intelligent manufacturing, applicable to equipment with... Multi-axis equipment with multiple linkage axes, among which For integers greater than or equal to 2, the characteristic is that, This method comprises three cooperative functional modules: an online decoupling and real-time compensation module for strong multi-physics coupling errors, a calibration-free multi-axis coordinated adaptive reconstruction module for flexible production changeover, and a nonlinear variable gain global synchronous control law execution module. The nonlinear limitation is reflected in The nonlinearity is reflected in the piecewise saturation, normalization processing, and product coupling with contour error, as well as in the piecewise characteristics of the normalized load disturbance intensity index and the cross-coupling of the coordination quantities between axes. The collaborative relationship among the three functional modules is as follows: the calibration-free multi-axis coordinated adaptive reconfiguration module is triggered and activated during the non-cutting transition phase of equipment changeover. The dynamic parameters identified online are respectively passed to the multi-physics field strong coupling error online decoupling and real-time compensation module and the nonlinear variable gain global synchronous control law execution module to complete parameter initialization. During the machining process, the multi-physics field strong coupling error online decoupling and real-time compensation module continuously runs the adaptive extended state observer to decouple thermal drift disturbance and cutting dynamic disturbance online. It passes the disturbance estimate and tracking error state quantity after decoupling of each axis to the nonlinear variable gain global synchronous control law execution module. The nonlinear variable gain global synchronous control law execution module dynamically adjusts the coordination gain according to the real-time load disturbance intensity of each axis, synthesizes the control torque command of each axis and sends it to the servo drive end. The sensor data of the servo drive end is simultaneously fed back to the multi-physics field strong coupling error online decoupling and real-time compensation module, forming a complete closed loop. The multi-physics field strong coupling error online decoupling and real-time compensation module integrates the position tracking error, velocity tracking error, thermal drift disturbance component, and cutting dynamic disturbance component of each axis into a joint state space framework, constructs a joint state space model, designs an adaptive extended state observer, uses the L2 norm of the information vector as the driving signal, adaptively adjusts the observer gain matrix through a Sigmoid-type smooth saturation function, and utilizes the significant separation characteristics of thermal drift disturbance and cutting dynamic disturbance in the frequency domain to extract the decoupling estimates of the two types of disturbances through low-pass digital filters and high-pass digital filters respectively. The calibration-free multi-axis coordinated adaptive reconfiguration module injects multi-frequency sinusoidal superposition detection displacement signals with sub-millimeter amplitudes into each axis during the non-cutting transition phase of production changeover. It uses a recursive least squares method with a forgetting factor to identify the dynamic parameters of the multi-axis system online, and automatically reconfigures the control law parameters required by each module after convergence. The nonlinear variable gain global synchronous control law execution module calculates the normalized load disturbance intensity index of each axis based on the cutting dynamic disturbance estimate output by the adaptive extended state observer. Based on this index, it adjusts the coordination gain of each axis in real time according to the linear proportional amplification rule, and combines the independent feedback control quantity, disturbance feedforward compensation quantity and variable gain cross-coupling coordination quantity to synthesize the control command of each axis.

2. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: In the joint state-space model Joint state vector of the shaft system Including the axis( Position tracking error Speed ​​tracking error Thermal drift perturbation component and cutting dynamic disturbance components ,total Each state component has a geometric error that is approximately constant within one control cycle, and thermal drift disturbance components are incorporated. In a combined manner, the joint state-space model uses encoder position and driver current signals as measurable outputs. The structure of the main diagonal sub-blocks of each axis in the system state matrix is ​​determined by the dynamic equations of each axis (including the inter-axis coupling inertia term). The self-decay term of the thermal disturbance sub-state reflects the thermal time constant.

3. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The observer gain matrix of the adaptive extended state observer Basic gain matrix The sum of the adaptive correction and the fundamental gain matrix is ​​obtained by the pole placement method to ensure observer stability. The adaptive correction is driven by the L2 norm of the innovation vector through a sigmoid-type smoothing saturation function and is combined with the gain matrix of the high-frequency perturbation channel. The multiplication results in the main non-zero elements of the high-frequency disturbance channel dedicated gain matrix corresponding to the observation equation row of the cutting dynamic disturbance state. When the innovation norm 2 exceeds the saturation threshold determined by the sensor noise level, the high-frequency channel component in the gain matrix automatically increases, improving the tracking bandwidth for high-frequency cutting dynamic disturbances. When the innovation norm 2 is lower than the saturation threshold, the gain matrix falls back to the basic value, avoiding amplification of sensor noise.

4. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The multi-physics field strong coupling error online decoupling and real-time compensation module performs the following steps sequentially in each control cycle: synchronously acquire the absolute position signal of each axis encoder, the driver current signal (converted to equivalent axial torque by Clarke-Park transformation), and the temperature sensor readings via EtherCAT bus; after mean filtering, construct a measurable output vector; calculate the innovation vector and its L2 norm using the state estimate of the previous cycle; update the observer gain matrix according to the innovation L2 norm using a Sigmoid-type frequency band adaptive strategy; execute the discrete state update equation to obtain the posterior state estimate; extract the thermal disturbance decoupling estimate from the posterior state estimate through a low-pass digital filter with a cutoff frequency of 0.1Hz; extract the dynamic disturbance decoupling estimate through a high-pass digital filter with a cutoff frequency of 5Hz; and output it to the nonlinear variable gain global synchronous control law execution module.

5. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The amplitude of each frequency component of the multi-frequency sinusoidal superposition displacement detection signal is in the range of 0.05mm to 0.5mm. Each superposition frequency is logarithmically selected from the equipment motion bandwidth range, with a frequency range of 1Hz to 50Hz. The initial phases of the detection signals between each axis are staggered to meet the continuous excitation conditions when multiple axes are simultaneously excited. The detection signals are superimposed on the idle jogging trajectory during the production changeover transition phase, with a duration not exceeding 10 seconds.

6. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The recursive least squares method with a forgetting factor transforms the dynamic equations of a multi-axis system into a standard linear parameterized form by multiplying a regression matrix containing the acceleration, velocity, and position measurements of each axis with the vector of parameters to be identified. The recursive estimate includes the parameter vector containing independent elements of the inertia matrix, damping matrix, and stiffness matrix. The forgetting factor ranges from 0.97 to 0.

999. The identification convergence criterion is that within 50 consecutive control cycles, the relative rate of change of the estimated values ​​of all parameters to be identified is less than a preset convergence threshold. If the identification duration exceeds a preset upper limit and the convergence criterion is not met, the previous valid identification result is retained and an alarm signal is issued.

7. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The automatic reconstruction of the control law parameters after identification and convergence includes the following steps: Based on the identified inertia matrix and damping matrix, the system matrix and basic gain matrix of the joint state space model of the multi-physics field strong coupling error online decoupling and real-time compensation module are updated by using the pole placement method. The cross-coupled basic gain matrix of the nonlinear variable gain global synchronous control law execution module is updated by multiplying the identified inertia matrix by the inverse of the factory nominal inertia matrix and then by the factory nominal operating condition reference gain matrix. After writing all reconstruction parameters into the registers of the corresponding modules, the production line resumes normal processing mode, and the calibration-free multi-axis coordinated adaptive reconstruction module enters a dormant state to wait for the next production change trigger.

8. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: The normalized load disturbance intensity index For the first The absolute value of the estimated dynamic disturbance of shaft cutting, divided by all The ratio of the peak value of the absolute value of the estimated dynamic disturbance of the cutting axis to the sum of the zero-prevention small positive numbers, ranging from 0 to 1, represents the coordination gain of each axis. equal to the base gain and The product of the variable gain amplification factor The value range is 1 to 5. The Cartesian space contour error vector is the component perpendicular to the trajectory direction obtained by subtracting the projection component along the unit tangent direction of the reference trajectory from the position error vector of each axis. The required unit tangent direction vector is provided in real time by the interpolation module.

9. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: No. The disturbance feedforward compensation for the axis is equal to the product of the axis's equivalent inertia and the sum of the thermal disturbance decoupling estimate and the dynamic disturbance decoupling estimate, wherein the equivalent inertia is taken from the identified inertia matrix. Line number The nonlinear variable gain global synchronous control law execution module sequentially calculates the independent PD feedback control quantity, disturbance feedforward compensation quantity, Cartesian space contour error vector, normalized load disturbance intensity index, adaptive variable gain coordination coefficient, and variable gain cross-coupling coordination quantity for each axis in each control cycle. After superimposing the above three control quantities, they are synchronously sent to each axis servo driver via the EtherCAT bus.

10. The automated coordinated control method for multi-axis equipment in intelligent manufacturing according to claim 1, characterized in that: A three-layer hierarchical control architecture of cloud-edge-device is adopted: The device layer includes a multi-axis servo drive group, a high-resolution encoder, a current sensor and a temperature sensor array. Each device in the device layer is connected to the device layer via an EtherCAT bus with a communication cycle of 125μs to 1ms. The device layer is a high real-time industrial controller with an FPGA / DSP coprocessing module, and deploys an adaptive extended state observer core computing module and a variable gain coordinated control law execution module with a control cycle of 125μs to 500μs. The cloud layer is the main controller and deploys a thermal error evolution slow dynamic model update module, a calibration-free online identification and parameter reconstruction module and a global task scheduling module with a refresh cycle of 10ms to 100ms. The device layer uploads raw sensor data to the device layer, the device layer sends control torque commands to the device layer, the device layer uploads filtered state estimates to the cloud layer, and the cloud layer sends updated model parameters and gain matrices to the device layer.

Images