A multi-variable cooperative control system for complex motion trajectories

Abstract

The application relates to the technical field of machine control systems, and discloses a multivariable cooperative control system for complex motion trajectories, which comprises an interpolation instruction processing unit, a reference load power spectrum storage unit, an interference source identification unit, a feedforward compensation control unit and an output instruction filtering unit. The system collects real-time stator currents of main shaft motors and calculates power spectrum densities, determines the chip load interference state by comparing the residuals of real-time and reference power spectrum densities, generates compensation instructions of each motion shaft according to a multibody dynamics model, and matches the mechanical resonance frequency of a servo transmission chain by using low-pass filtering. The application accurately identifies chip shearing interference and captures nonlinear coupling forces generated by changes in machining postures, effectively solves the problem that multi-axis following is not synchronized, and guarantees the fidelity of a synthesized trajectory and the running stability during complex trajectory machining.

Description

Technical Field

[0001] This invention relates to a multivariable cooperative control system for complex motion trajectories, belonging to the field of industrial control system technology. Background Technology

[0002] Currently, the industry generally adopts multi-axis linkage CNC systems for the manufacturing process of aerospace titanium alloy components or irregular medical implants. These systems utilize time-axis discrete displacement interpolation commands, combined with a cascaded control architecture consisting of position, velocity, and current, to track theoretical trajectory points through feedback signals from position sensors. However, when dealing with nonlinear and complex trajectories, multi-axis linkage exhibits obvious dynamic coupling characteristics. Tool posture switching induces centrifugal force and Coriolis force interference. These dynamic torques are transformed into cross-interference loads between servo axes in the mechanical transmission chain. Since the adjustment logic of each axis is independent, the system has difficulty in real-time offsetting the synthetic errors caused by cross-axis coupling, resulting in limited machining trajectory fidelity.

[0003] To reduce tracking errors, conventional approaches attempt to increase the servo loop gain or add a fixed feedforward compensation. However, the spindle system undergoes microscopic elastic deformation under transient load impacts. This deformation causes a phase lag in the feedback signal relative to the physical real position. Simply increasing the gain not only fails to compensate for the phase shift but also easily excites the first-order resonance of the mechanical transmission chain, resulting in physical excitation. The industry is constantly trying to optimize the rigidity of the mechanical transmission chain to overcome hardware limitations. However, simply relying on physical structural reinforcement cannot fundamentally eliminate the dynamic disturbances of multi-axis linkage. There are also shortcomings in software control methods. For example, Chinese invention patent application CN120335386A discloses a CNC machine tool linkage operating system. By deploying high-precision grating rulers and accelerometers on the linkage axis to construct an error-coupled dynamic model and adjusting the PID gain parameters in real time, the underlying objective logic of the pure post-compensation mechanism based on position feedback and vibration monitoring is highly dependent on the occurrence of macroscopic mechanical displacement and physical oscillation. Under complex nonlinear working conditions such as deep cavity high-speed cutting, sudden changes in cutting load cause microscopic elastic deformation of the tool and transient phase lag before the sensor captures macroscopic displacement deviation. The control system receives vibration or position deviation signals and performs closed-loop adjustment, which is difficult to offset the transient cross-axis coupling torque. It is very easy to accidentally superimpose due to signal feedback time lag and excite high-frequency mechanical resonance of the servo drive chain. The existing technology has a fundamental mismatch between the core preset premise and the actual boundary conditions.

[0004] Therefore, how to combine the dynamic characteristics of the spindle with the multi-axis spatial decoupling model to achieve advance compensation for tool deformation phase lag and real-time decoupling of cross-axis interference has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A multivariable cooperative control system for complex motion trajectories, comprising:

[0006] The interpolation instruction processing unit is used to acquire interpolation point data for multi-axis linkage trajectory;

[0007] The reference load power spectrum storage unit is used to store the reference no-load power spectral density of the machine tool under different trajectory curvatures;

[0008] The interference source identification unit is used to acquire the real-time stator current signal of the spindle motor, calculate the real-time power spectral density of the real-time stator current signal, calculate the residual spectrum between the real-time power spectral density and the reference no-load power spectral density, and determine the debris load interference state based on the energy integral value of the residual spectrum in the frequency band from 300Hz to 800Hz.

[0009] The feedforward compensation control unit is used to convert the chip load disturbance state into the equivalent disturbance torque of each motion axis based on the multibody dynamics model, and generate compensation commands in the control loop of each motion axis. The polarity of the compensation command is opposite to the polarity of the nonlinear coupling force generated by the change of machining posture.

[0010] The output command filtering unit is used to select a cutoff frequency that matches the first-order mechanical resonant frequency of the servo drive train and perform low-pass filtering on the compensation command.

[0011] Preferably, the interference source identification unit determines the debris load interference state through the following steps: Step S1, acquire the real-time stator current signal, and map the time-domain current signal into a frequency-domain feature vector through fast Fourier transform; Step S2, calculate the difference between the frequency-domain feature vector and the reference value of the corresponding trajectory curvature in the reference load power spectrum storage unit, and extract the residual power spectral density; Step S3, calculate the energy integral of the residual power spectral density in the frequency band of 300Hz to 800Hz, and obtain the residual spectral feature quantity; Step S4, compare the residual spectral feature quantity with the preset physical source identification threshold, and when the residual spectral feature quantity exceeds the physical source identification threshold, it is determined to be a debris load interference state.

[0012] Preferably, the feedforward compensation control unit extracts the real-time displacement components and angular acceleration components of each motion axis based on the multibody dynamics model, calculates the cross-coupling intervention amount generated by the linkage of each motion axis, and nonlinearly weights and superimposes the cross-coupling intervention amount with the equivalent disturbance torque to construct a dynamic contour error compensation model.

[0013] Preferably, the reference load power spectrum storage unit stores a load feature mapping table of irregularly shaped workpieces. This load feature mapping table is based on the spatial geometric vector of the interpolation point data and has a pre-set load distribution model under different feed speeds.

[0014] Preferably, the system also includes a real-time status monitoring unit, which is used to acquire the feedback torque of each motion axis servo motor and calculate the transient deviation between the feedback torque and the theoretical torque corresponding to the interpolation point data. The feedforward compensation control unit corrects the compensation command based on the transient deviation.

[0015] Preferably, the interpolation command processing unit responds to the debris load interference state determined by the interference source identification unit, adjusts the topology of the control command, and suppresses the torque load fluctuation of the spindle motor by changing the feed amount within the interpolation cycle.

[0016] Preferably, when generating compensation commands, the feedforward compensation control unit calculates a phase lag compensation operator, which is used to compensate for the displacement response delay of the motion axis caused by the change in spindle torque.

[0017] Preferably, the reference load power spectrum storage unit records the power spectrum feature evolution data throughout the entire life cycle of the machine tool, and the interference source identification unit updates the reference no-load power spectrum density in real time based on the power spectrum feature evolution data to filter out background noise generated by bearing wear.

[0018] Preferably, the system adopts a real-time industrial Ethernet communication architecture, and the feedforward compensation control unit synchronously sends compensation commands to each servo drive through a distributed clock synchronization mechanism.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. In the multivariable collaborative control of motion trajectory, by acquiring the time first derivative of the high-frequency torque ripple signal of the spindle motor and the time first derivative of the actual speed feedback of the feed driver, the cross transient phase deviation angle of the two in the polar coordinate system is calculated, so as to realize the accurate extraction of the transient mechanical impedance state of the machining area. This mechanism changes the limitation of the traditional control logic that only relies on the load scalar amplitude as the feedback source, so that the system can identify the micro-elastic deflection phenomenon of the tool induced by chip accumulation in the early stage before the torque amplitude reaches the judgment threshold. By mapping the phase deviation angle to the virtual stiffness damping coefficient and generating the phase-leading decoupling feedforward command, the system actively counteracts the trajectory distortion caused by the elastic deformation of the tool body and eliminates the risk of workpiece overcutting caused by physical response lag.

[0021] 2. The multivariable collaborative decoupling control unit utilizes Jacobi inverse matrix operations to map the angular acceleration component in the reference deflection command output by the theoretical interpolation module into an equivalent coupled disturbance torque acting on each linear feed axis in real time. This cross-variable correlation compensation mechanism generates feedforward commands with opposite polarities within the control loop, offsetting the centrifugal force and Coriolis force interference induced by drastic changes in tool holder posture during the five-axis linkage process. Compared to the independent closed-loop adjustment method for each axis, this scheme ensures the synchronization of the dynamic response of multiple servo axes under complex nonlinear trajectories through deep coupling of the multibody dynamics model, maintaining the fidelity of the synthetic trajectory under high-speed, high-curvature machining conditions.

[0022] 3. The system performs spectrum analysis on the high-frequency torque ripple signal of the spindle motor and subtracts the pre-stored machine tool reference no-load power spectrum to calculate the energy integral value of the residual spectrum in the chip interference frequency band. This constructs a physical feature-driven working condition identification logic. This logic uses spectrum stripping technology to remove background noise caused by uneven workpiece material hardness and spindle bearing wear, accurately locking the specific physical source of chip shear interference. This intervention mechanism based on frequency domain energy distribution avoids frequent fluctuations in feed rate caused by false interference signals, improves the operational stability of the CNC system in extreme environments such as deep cavity and irregular shape machining, and ensures the consistency of workpiece surface quality. Attached Figure Description

[0023] Figure 1 This is a flowchart of the interference identification and feedforward compensation process of the multi-axis linkage control system of the present invention.

[0024] Figure 2 This is a diagram of the communication network and physical drive architecture of the multivariable collaborative control system of the present invention.

[0025] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0027] A multivariable cooperative control system for complex motion trajectories includes:

[0028] The interpolation instruction processing unit is used to acquire interpolation point data for multi-axis linkage trajectory;

[0029] The reference load power spectrum storage unit is used to store the reference no-load power spectral density of the machine tool under different trajectory curvatures;

[0030] The interference source identification unit is used to acquire the real-time stator current signal of the spindle motor, calculate the real-time power spectral density of the real-time stator current signal, calculate the residual spectrum between the real-time power spectral density and the reference no-load power spectral density, and determine the debris load interference state based on the energy integral value of the residual spectrum in the frequency band from 300Hz to 800Hz.

[0031] The feedforward compensation control unit is used to convert the chip load disturbance state into the equivalent disturbance torque of each motion axis based on the multibody dynamics model, and generate compensation commands in the control loop of each motion axis. The polarity of the compensation command is opposite to the polarity of the nonlinear coupling force generated by the change of machining posture.

[0032] The output command filtering unit is used to select a cutoff frequency that matches the first-order mechanical resonant frequency of the servo drive train and perform low-pass filtering on the compensation command.

[0033] Preferably, the interference source identification unit determines the debris load interference state through the following steps: Step S1, acquire the real-time stator current signal, and map the time-domain current signal into a frequency-domain feature vector through fast Fourier transform; Step S2, calculate the difference between the frequency-domain feature vector and the reference value of the corresponding trajectory curvature in the reference load power spectrum storage unit, and extract the residual power spectral density; Step S3, calculate the energy integral of the residual power spectral density in the frequency band of 300Hz to 800Hz, and obtain the residual spectral feature quantity; Step S4, compare the residual spectral feature quantity with the preset physical source identification threshold, and when the residual spectral feature quantity exceeds the physical source identification threshold, it is determined to be a debris load interference state.

[0034] Preferably, the feedforward compensation control unit extracts the real-time displacement components and angular acceleration components of each motion axis based on the multibody dynamics model, calculates the cross-coupling intervention amount generated by the linkage of each motion axis, and nonlinearly weights and superimposes the cross-coupling intervention amount with the equivalent disturbance torque to construct a dynamic contour error compensation model.

[0035] Preferably, the cutoff frequency of the output command filtering unit The following quantitative constraints must be met: ,in, The first-order mechanical resonant frequency of the machine tool servo drive chain is the pre-determined frequency, and α is the preset frequency domain matching coefficient, with the value of α ranging from 0.3 to 0.4.

[0036] Preferably, the reference load power spectrum storage unit stores a load feature mapping table of irregularly shaped workpieces. This load feature mapping table is based on the spatial geometric vector of the interpolation point data and has a pre-set load distribution model under different feed speeds.

[0037] Preferably, the system also includes a real-time status monitoring unit, which is used to acquire the feedback torque of each motion axis servo motor and calculate the transient deviation between the feedback torque and the theoretical torque corresponding to the interpolation point data. The feedforward compensation control unit corrects the compensation command based on the transient deviation.

[0038] Preferably, the interpolation command processing unit responds to the debris load interference state determined by the interference source identification unit, adjusts the topology of the control command, and suppresses the torque load fluctuation of the spindle motor by changing the feed amount within the interpolation cycle.

[0039] Preferably, when generating compensation commands, the feedforward compensation control unit calculates a phase lag compensation operator, which is used to compensate for the displacement response delay of the motion axis caused by the change in spindle torque.

[0040] Preferably, the reference load power spectrum storage unit records the power spectrum feature evolution data throughout the entire life cycle of the machine tool, and the interference source identification unit updates the reference no-load power spectrum density in real time based on the power spectrum feature evolution data to filter out background noise generated by bearing wear.

[0041] Preferably, the system adopts a real-time industrial Ethernet communication architecture, and the feedforward compensation control unit synchronously sends compensation commands to each servo drive through a distributed clock synchronization mechanism.

[0042] Example 1: In the five-axis linkage precision CNC machining of aerospace titanium alloy blades, the interpolation command processing unit acquires the interpolation point data of the multi-axis linkage trajectory. At this time, with the nonlinear change of the curvature of the cutting trajectory, chip removal is obstructed in the deep cavity machining area, causing a sudden surge in physical resistance in the tool contact area and triggering local elastic deformation of the spindle system. Traditional single-axis independent closed-loop adjustment logic cannot counteract the interference of centrifugal force and Coriolis force induced by tool attitude switching. Furthermore, under transient load impact, the feedback signal exhibits high-frequency phase lag relative to the physical real position, forming cross-coupling interference. This results in asynchronous following responses of multiple servo axes under transient physical disturbances, leading to... Dynamic contour error; When the system is dealing with this multivariate cross-coupling condition, the reference load power spectrum storage unit pre-stores the reference no-load power spectrum density of the machine tool under different trajectory curvatures. The interference source identification unit obtains the real-time stator current signal of the spindle motor and converts the time-domain current signal into a frequency-domain feature vector through fast Fourier transform. The difference between the frequency-domain feature vector and the reference no-load power spectrum density of the corresponding trajectory curvature is calculated to extract the residual power spectrum density. The energy integral of the residual spectrum feature obtained by the energy integral exceeds the preset physical source identification threshold, and it is determined that the current processing area is in a chip load interference state.

[0043] Based on the physical nature of cutting dynamics—where the regular cutting force exhibits discrete harmonics while random chip compression presents a continuous broadband spectrum—the interference source identification unit executes a broadband feature decoupling procedure to obtain the current spindle speed command and the number of cutting teeth to calculate the main cutting fundamental frequency. A digital notch filter is then used to remove the energy in the frequency band containing the main cutting fundamental frequency and the first three harmonics from the aforementioned energy integral, outputting only the net residual energy integral E representing random resistance fluctuations. Subsequently, a disturbance quantization mapping procedure is executed. According to the electromechanical energy conversion principle, the square of the active component of the motor stator current is linearly positively correlated with the output electromagnetic torque fluctuation. The feedforward compensation control unit then uses the formula... Calculate the basic equivalent disturbance moment ,in, η represents the nominal torque constant determined by the nameplate parameters of the spindle servo motor, and η represents the dimensionless coefficient of the mechanical efficiency of the spindle drive train, pre-calibrated through offline dynamometer experiments. This formula converts the frequency-domain electrical energy deterministically into mechanical-domain physical torque data. The multibody dynamics model, based on the Newton-Euler equations, abstracts the kinematic topology of the five-axis machine tool into a rigid body system node network containing three translational degrees of freedom and two rotational degrees of freedom. Its internal mapping operation constructs the Jacobian inverse matrix in the link coordinate system, decomposing the equivalent disturbance torque at the spindle end into the driving direction of each linear feed axis in real time. Specifically, the calculation mechanism is that the driving compensation torque vector of each axis is equal to the transpose of the Jacobian inverse matrix multiplied by the vector product of the end-effector disturbance torque and the spatial attitude angular acceleration, thus completing the transformation from disturbance state... To achieve a precise quantitative conversion of the physical axis compensation value, the feedforward compensation control unit extracts the real-time displacement and angular acceleration components of each motion axis based on the multibody dynamics model to calculate the cross-coupling intervention amount generated by the linkage of each motion axis. At the same time, it calculates the phase lag compensation operator used to compensate for the displacement response delay of the motion axis caused by the change of spindle torque. The cross-coupling intervention amount is superimposed with the equivalent disturbance torque amount converted from the chip load disturbance state. A compensation command with the opposite polarity to the nonlinear coupling force generated by the change of machining posture is generated in the control loop of each motion axis. By converting the frequency domain energy distribution characteristics into phase and torque compensation parameters in the control logic, the compensation command output by the feedforward compensation control unit solves the technical contradiction in the control system that increasing the gain of the servo loop will inevitably excite the first-order resonance of the mechanical transmission chain.

[0044] To maintain the fidelity of the synthesized trajectory during high-speed machining and eliminate the risk of control commands exciting inherent vibrations in the machine tool structure, the output command filtering unit selects a low-pass filter compensation command with a cutoff frequency that matches the pre-determined first-order mechanical resonance frequency of the servo drive train. The cutoff frequency is then set. Satisfying the relation ,in, The cutoff frequency is α, which is a preset frequency domain matching coefficient with a specific value ranging from 0.3 to 0.4. The compensation command, which is the first-order mechanical resonant frequency and has been bandwidth-matched filtered, is synchronously sent to each servo driver through a distributed clock synchronization mechanism, thereby filtering out the high-order harmonic components contained in the intervention logic. After the aforementioned multivariable state space decoupling and frequency domain feature stripping operations, the multivariable cooperative control system eliminates the background noise caused by uneven workpiece material hardness and spindle bearing wear by comparing the residual between the real-time power spectral density and the reference no-load power spectral density, separates the chip shearing interference as a specific physical source, and actively offsets the synthetic error caused by cross-axis coupling before the torque amplitude reaches the physical threshold that causes destructive overcutting, maintaining the dynamic response synchronization of multiple servo axes under complex nonlinear trajectories and adaptability to time-varying machining load disturbances.

[0045] Example 2: In the five-axis linkage precision CNC machining of aerospace titanium alloy blades, the curvature of the tool machining trajectory changes abruptly continuously, and the spindle is in a cutting environment with time-varying loads. The experiment relies on a machining center test platform with a five-axis linkage mechanism to obtain physical data. The maximum feed rate of the linear axis of the platform is set to 30 m / min, the rated spindle speed is set to 12000 rpm, and the current sampling frequency is set to 10 kHz to cover high-frequency ripple signals. To construct the industrial test conditions, Gaussian white noise with a signal-to-noise ratio of 21.5 dB is superimposed on the test signal source, and power frequency interference signals with a frequency of 50 Hz and its odd harmonics are injected to reproduce the physical impact of workshop electromagnetic interference on the stator current acquisition link. The selection of the fast Fourier transform time window length required to extract the residual power spectral density involves engineering considerations of data processing real-time performance and frequency domain resolution. When the time window length increases, the frequency domain resolution increases accordingly, which will lead to an extension of the calculation cycle of the control system and cause phase of the servo loop. The time window length is determined based on the mapping relationship between the spindle's rated speed and the fundamental frequency. When the spindle speed reaches 12,000 rpm and the corresponding fundamental frequency is 200 Hz, the time window length must include 10 complete fundamental cycles to avoid spectral leakage. Based on this, the specific value of the time window length is determined to be 51.2 ms. This satisfies the real-time interpolation cycle constraint of the CNC system while ensuring the accuracy of energy integration calculation in the 300 Hz to 800 Hz frequency band. The aforementioned absolute characteristic frequency band of 300 Hz to 800 Hz is not randomly set, but is a frequency domain statistical boundary established based on a large number of generalized cutting dynamics experiments of aerospace titanium alloy materials. A large amount of cutting force experimental data confirms that, since the secondary shearing and transient extrusion fracture events of chips in the narrow chip groove of the tool have a specific microsecond-level duration, the broadband excitation signal reflected on the spindle torque ripple is not directly affected by the macroscopic feed speed and is always constantly concentrated in this specific high-frequency band.

[0046] An experimental intensity gradient control system was constructed to verify the adaptability of the feedforward compensation logic. The curvature radii of the cutting trajectory were set to 50.5 mm, 20.2 mm, and 8.1 mm, simulating low, medium, and high nonlinear coupling intensity gradients. In the control group without the feedforward compensation control unit activated, the extracted real-time stator current signal contained broadband noise, with its energy integral in the 300-800 Hz frequency band reaching 1.25 A at low curvature. 2 It climbs to 8.72 A under high curvature. 2 At this point, the following error of the linear feed axis climbed to 45.6 μm, causing trajectory distortion. In the test group for launching a multivariable cooperative control system for complex motion trajectories, the interference source identification unit calculated the residual between the real-time power spectral density and the pre-stored reference no-load power spectral density. Under the high curvature condition of 8.1 mm, the net debris load interference energy integral value after residual stripping was stably extracted to be 6.85 A. 2 This intermediate characteristic quantity removes 21.4% of the background noise energy. Based on this residual characteristic quantity and the multibody dynamics model, the feedforward compensation control unit calculates the equivalent disturbance torque as 15.4 N⋅m, with an output phase lead angle of 12.5°. ∘ The compensation command generates compensation data to offset transient physical disturbances.

[0047] To verify the technical boundary of the frequency domain matching coefficient α in the output command filtering unit, an out-of-range control group was set up for this parameter, and the first-order mechanical resonant frequency of the servo drive train was measured. With a frequency domain matching coefficient α of 0.2, the cutoff frequency is calculated based on the formula, assuming a frequency of 450 Hz. At 90 Hz, the system suppresses mechanical resonance, but the low-pass filter effect exacerbates the phase lag of the compensation command. The dynamic profile error under high curvature remains at 28.4 μm. When the set frequency domain matching coefficient α crosses 0.4 and reaches 0.6, the cutoff frequency is calculated. At 270 Hz, the experimental data showed a nonlinear degradation inflection point. The servo loop gain excited the second-order resonance of the drive train, causing the spindle to produce a continuous oscillation with an amplitude of 12.8 μm. When the experimental group set α to 0.35, the cutoff frequency was determined. At 157.5 Hz, the system suppresses the dynamic response synchronization error of each servo axis to 4.2 μm under the condition of 8.1 mm high curvature, and makes the attenuation amplitude of the high harmonics of the compensation command reach 32.5 dB. By introducing a frequency domain feature extraction mechanism based on residual power spectral density and a phase compensation logic of bandwidth matching filtering, the multivariable collaborative control system for complex motion trajectories enables the five-axis linkage platform to cancel the cross-coupling interference induced by tool posture switching in the trajectory curvature gradient and high-noise electromagnetic environment. The control system converts nonlinear physical deformation into current frequency domain compensation signal, maintaining the synchronization of multi-servo axis linkage response and the cutting fidelity of complex geometric contours.

[0048] Example 3: In the five-axis linkage precision CNC machining of aerospace titanium alloy blades, the multivariable collaborative control system for complex motion trajectories starts the reference state calibration program before the cutting task begins. The interpolation command processing unit sends a sweep excitation current command covering the 0 to 1000Hz frequency band to each servo axis. The output command filtering unit synchronously collects the mechanical vibration feedback signal of the servo drive chain, extracts the frequency point corresponding to the peak amplitude of the mechanical vibration feedback signal in the frequency domain, and determines the frequency point as the first-order mechanical resonant frequency of the servo drive chain. The machine tool runs unloaded according to a predetermined complex nonlinear trajectory. The interference source identification unit collects the stator current signal of the spindle motor under no-load conditions at a sampling frequency of 10kHz, and calculates the residual power spectral density energy integral in the 300Hz to 800Hz frequency band within multiple consecutive sampling periods. The statistical mean and standard deviation of the multiple energy integral data are extracted, and the physical source identification threshold is set to the statistical mean plus three times the standard deviation. To eliminate the slow time-varying drift of mechanical impedance caused by structural thermal expansion and physical wear of spindle bearings during continuous machine tool operation, the baseline adaptive update procedure is activated. When the machine tool performs non-cutting tool change or trajectory fast traverse idle travel clearance, the stator current of the spindle motor is re-acquired. The residual energy integral of the current independent time period is calculated according to the no-load calibration process. The statistical mean and standard deviation are updated in real time using the forgetting factor recursive least squares method. Once the mean of three consecutive non-cutting time periods deviates from the initial calculation benchmark by more than 5% tolerance boundary, the historical data in the benchmark load power spectrum storage unit is overwritten with built-in logic instructions, and the physical source identification threshold is re-determined based on the latest acquired statistical parameters.

[0049] When the machining system enters a load cutting state and the extracted residual spectral features exceed the physical source identification threshold, the interference source identification unit determines that the machining area is in a chip load interference state. The feedforward compensation control unit obtains the reference deflection command output by the interpolation command processing unit, extracts the angular acceleration component from the reference deflection command, and applies the Jacobian inverse matrix operation to map the angular acceleration component into an equivalent disturbance torque acting on each linear feed axis. The feedforward compensation control unit calculates the phase lead compensation time based on the extracted spindle torque change, and superimposes the equivalent disturbance torque with the phase lead compensation time to generate a phase lead compensation time within the control loop of each linear feed axis. The compensation command has the opposite polarity to the equivalent disturbance torque. The output command filtering unit calculates the cutoff frequency based on the pre-determined first-order mechanical resonance frequency and the set frequency domain matching coefficient. Based on the cutoff frequency, the low-pass filter compensation command is sent to the servo drives of each linear feed axis through a distributed clock synchronization mechanism. The multivariable cooperative control system for complex motion trajectories establishes the interference judgment boundary by extracting the statistical characteristics of the current spectrum of the reference no-load state. The dynamic kinematic parameters of the rotary axis are converted into the torque compensation of the linear axis through the Jacobian inverse matrix to maintain the synchronization of the multi-servo axis linkage response.

[0050] Example 4: When the system faces the on-site deployment of a five-axis CNC machine tool, the multivariable collaborative control system starts the offline parameter calibration program before the machining task begins. The interpolation command processing unit sends test motion commands containing standard orthogonal trajectories and different radii of curvature to each servo driver. The interference source identification unit synchronously collects the no-load stator current signal of each axis motor within the full working envelope range at a set sampling frequency. The collected no-load stator current signal is converted into frequency domain features through fast Fourier transform. The reference power spectral density energy integral of the 300Hz to 800Hz frequency band under each curvature trajectory is extracted. The series of energy integral data and the corresponding trajectory geometric parameters are used to construct a mapping matrix. This mapping matrix is ​​written into the reference load power spectrum storage unit to construct the initial comparison baseline. The control system performs feedforward compensation. The control unit applies a step torque excitation and collects transient displacement response data of each linear feed axis under this step excitation. Based on the transient displacement response data, the actual moment of inertia and stiffness damping coefficient of each linkage axis are calculated. The collected physical and mechanical parameters are updated to the node parameters of the multibody dynamics model, so that the nodes of the dynamics model are anchored to the physical properties of the current machine tool. The mapping matrix internally encapsulates the load distribution model under different feed speeds. The data structure of the model is specifically represented as a three-dimensional discrete lookup table with the inverse of curvature of the spatial geometric vector and the angle between the normal vector as two-dimensional input indices and the corresponding instantaneous cutting load force as the output node weights. During online operation, the bilinear interpolation algorithm enables the independently sampled geometric parameters and feed speed to achieve strong spatial coupling calculation, and directly looks up the table to output the matching reference no-load value.

[0051] Based on the mapping matrix and physical and mechanical parameters obtained from the aforementioned offline parameter calibration program, the multivariable cooperative control system continuously calls and updates the multibody dynamics model during the cutting task. When the interference source identification unit obtains the real-time stator current signal and calculates the residual spectral energy integral of the real-time power spectral density, which is greater than the physical source identification threshold set based on the statistical characteristics of the no-load data, the feedforward compensation control unit, based on the updated moment of inertia and stiffness damping coefficient in the multibody dynamics model, applies the Jacobian inverse matrix operation to convert the extracted spindle angular acceleration component into an equivalent disturbance torque acting on the linear feed axis. Based on this equivalent disturbance torque, it generates compensation commands with opposite polarities within the control loop of each linear feed axis. The output command filtering unit performs low-pass filtering of the compensation command according to the cutoff frequency of the first-order mechanical resonance frequency of the matched servo drive chain. Through a distributed clock synchronization mechanism, it synchronously sends the filtered compensation command to the corresponding servo driver. The multivariable cooperative control system outputs a feedforward compensation control signal based on the calibrated physical and mechanical parameters and mapping matrix to adjust the dynamic following displacement of the linear feed axis driven by each servo driver.

[0052] Example 5: When the system faces the transient mechanical impedance state extraction condition during high-speed cutting, the interference source identification unit simultaneously acquires the high-frequency torque ripple signal of the spindle motor and the actual speed feedback signal of the linear feed axis. To suppress the broadband noise amplification effect induced by direct differentiation, the multivariable cooperative control system for complex motion trajectories sets a sliding sampling window with a time span covering the basic rotation cycle of the spindle. Within the sliding sampling window, the least squares method is used to linearly fit the high-frequency torque ripple signal of the spindle motor and the actual speed feedback signal, respectively, and extract the corresponding time first derivatives. The extracted time first derivative of the high-frequency torque ripple signal of the spindle motor is set as the vertical axis physical component of the orthogonal coordinate system, and the time first derivative of the actual speed feedback signal is set as the horizontal axis physical component. The quantity is calculated using the arctangent function to determine the transient phase deviation angle between the two components in their corresponding coordinate systems. This signal transformation process filters out high-frequency electromagnetic noise interference and outputs quantified physical parameters related to the local elastic deflection state of the tool. Before substituting these parameters into the arctangent function to perform trigonometric operations, the system's built-in physical conversion interface pre-multiplies the vertical axis physical component representing mechanical characteristics by a compliance transformation operator derived from the nominal stiffness of the spindle, and simultaneously multiplies the horizontal axis physical component representing kinematic characteristics by an inertia transformation operator derived from the equivalent mass of the feed system. This unifies the torque change rate and acceleration, which originally belonged to different dimensions, into a dimensionless instantaneous reactive power fluctuation ratio coefficient, eliminating the dimensional differences of cross-scale physical variables and ensuring that the extraction of the geometric phase angle conforms to rigorous and effective mathematical and geometric premises.

[0053] The feedforward compensation control unit is pre-loaded with a stiffness-damping mapping matrix obtained from offline cutting calibration tests. This mapping matrix stores multiple sets of discrete phase deviation angle interval boundary values ​​and corresponding measured values ​​of virtual stiffness-damping coefficients. After the machining task starts, the feedforward compensation control unit obtains the cross transient phase deviation angles from the real-time calculation output. It uses a linear interpolation algorithm in the stiffness-damping mapping matrix to calculate the target virtual stiffness-damping coefficient corresponding to the current control cycle. The product of the target virtual stiffness-damping coefficient and the real-time feed rate is calculated to obtain the decoupling feedforward compensation amount. This decoupling feedforward compensation amount is superimposed with the reference deflection command output by the interpolation command processing unit to generate a decoupling feedforward command. Relying on the distributed clock synchronization mechanism, the decoupling feedforward command is sent to the servo drivers of each linear feed axis to drive the servo motor to output compensation torque. The multivariable collaborative control system relies on offline calibration data interpolation and continuous matrix mapping process to transform the physical correlation quantity characterizing the mechanical deformation of the tool body into a feedforward compensation data stream for adjusting the servo position loop.

[0054] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multivariable cooperative control system for complex motion trajectories, characterized in that, include: The interpolation instruction processing unit is used to acquire interpolation point data for multi-axis linkage trajectory; The reference load power spectrum storage unit is used to store the reference no-load power spectral density of the machine tool under different trajectory curvatures; The interference source identification unit is used to acquire the real-time stator current signal of the spindle motor, calculate the real-time power spectral density of the real-time stator current signal, calculate the residual spectrum between the real-time power spectral density and the reference no-load power spectral density, and determine the debris load interference state based on the energy integral value of the residual spectrum in the frequency band from 300Hz to 800Hz. The feedforward compensation control unit is used to convert the chip load disturbance state into the equivalent disturbance torque of each motion axis based on the multibody dynamics model, and generate compensation commands in the control loop of each motion axis. The polarity of the compensation command is opposite to the polarity of the nonlinear coupling force generated by the change of machining posture. The output command filtering unit is used to select a cutoff frequency that matches the first-order mechanical resonant frequency of the servo drive train and perform low-pass filtering on the compensation command.

2. The multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The interference source identification unit determines the debris load interference state through the following steps: Step S1, acquire the real-time stator current signal, and map the time-domain current signal into a frequency-domain feature vector through fast Fourier transform; Step S2, calculate the difference between the frequency-domain feature vector and the reference value of the corresponding trajectory curvature in the reference load power spectrum storage unit, and extract the residual power spectral density; Step S3, calculate the energy integral of the residual power spectral density in the frequency band from 300Hz to 800Hz, and obtain the residual spectral feature quantity; Step S4, compare the residual spectral feature quantity with the preset physical source identification threshold. When the residual spectral feature quantity exceeds the physical source identification threshold, it is determined to be a debris load interference state.

3. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The feedforward compensation control unit extracts the real-time displacement and angular acceleration components of each motion axis based on the multibody dynamics model, calculates the cross-coupling intervention amount generated by the linkage of each motion axis, and nonlinearly weights and superimposes the cross-coupling intervention amount with the equivalent disturbance torque to construct a dynamic contour error compensation model.

4. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The reference load power spectrum storage unit stores a load feature mapping table for irregularly shaped workpieces. This load feature mapping table is based on the spatial geometric vector of the interpolation point data and has a preset load distribution model under different feed speeds.

5. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The system also includes a real-time status monitoring unit, which is used to acquire the feedback torque of each motion axis servo motor and calculate the transient deviation between the feedback torque and the theoretical torque corresponding to the interpolation point data. The feedforward compensation control unit corrects the compensation command based on the transient deviation.

6. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The interpolation command processing unit responds to the debris load interference state determined by the interference source identification unit, adjusts the topology of the control command, and suppresses the torque load fluctuation of the spindle motor by changing the feed amount within the interpolation cycle.

7. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, When generating compensation commands, the feedforward compensation control unit calculates the phase lag compensation operator, which is used to compensate for the displacement response delay of the motion axis caused by the change in spindle torque.

8. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The reference load power spectrum storage unit records the power spectrum feature evolution data throughout the machine tool's entire life cycle. The interference source identification unit updates the reference no-load power spectrum density in real time based on the power spectrum feature evolution data to filter out background noise generated by bearing wear.

9. A multivariable cooperative control system for complex motion trajectories according to claim 1, characterized in that, The system adopts a real-time industrial Ethernet communication architecture, and the feedforward compensation control unit sends compensation commands to each servo drive synchronously through a distributed clock synchronization mechanism.

Images