A high-precision feed control method for a milling machine

Abstract

This invention relates to the field of milling machine feed control, and discloses a high-precision feed control method for milling machines, comprising: acquiring the geometric parameters and rotation phase data of the face milling cutter and calculating the discrete sequence of the total cutting thrust load, separating it into a quasi-static low-frequency load and a high-frequency dynamic excitation component; injecting a test signal into the excitation coil to calculate the resistance and inductance of the reference and real-time tests, calculating the difference between Ohm and Henry to generate a thermomagnetic coupling gain factor; dividing the high-frequency dynamic excitation component by the equivalent stiffness of the feed axis to set it as the target micro-compensation displacement, calculating the initial compensation command current in combination with the thermomagnetic coupling gain factor, and generating a drive signal by performing a lead translation based on the real-time test impedance; the excitation coil receiving the signal drives the magnetostrictive material to generate endogenous stress, forcing the outer metal flexible diaphragm shell and the worktable saddle to generate high-frequency micro-displacements, which are superimposed and canceled out with the tool deformation in the same spatial coordinate system. This invention achieves high-frequency vibration smoothing and resistance to thermomagnetic attenuation.

Description

Technical Field

[0001] This invention relates to the field of milling machine feed control, specifically a high-precision feed control method for milling machines. Background Technology

[0002] Currently, in the field of high-precision axial feed machining for milling machines, the intermittent cutting of multi-tooth face milling cutters easily causes high-frequency forced resonance in the worktable drive chain. In order to suppress the resulting surface vibrations, modern precision feed systems typically embed magnetostrictive micro-motion units between the ball screw nut and the saddle, which, together with the spindle servo, form a macro-micro composite feed structure, using micron-level axial deformation to offset high-frequency machining deviations.

[0003] Regarding the aforementioned issues, existing feed error control schemes primarily rely on slip detection data from external grating rulers or static thrust mapping values ​​from the spindle servo drive. The controller inputs the acquired position displacement signal into a fixed feedback loop and directly outputs a drive control voltage to the internal excitation coil of the micro-motion unit. Through a fixed electromechanical conversion ratio, the material deformation pushes the slide of the worktable to complete the axial mechanical displacement tracking.

[0004] However, this control method has electromechanical matching defects when dealing with complex cutting conditions: First, because it directly uses the total error or total thrust as the control source, it fails to decouple the high-frequency dynamic excitation component that causes resonance by combining the geometric mapping of the face milling cutter's rotation angle, and the mechanical tool deformation lacks a precise algebraic quantitative basis; Second, the control coil heats up violently during continuous heavy-load machining, causing the magnetic permeability of the magnetic circuit to drift in real time. This, coupled with the multi-value hysteresis nonlinearity of the material itself, means that the fixed-gain drive cannot block the attenuation and distortion of the displacement amplitude; Third, the circuit inductance energy storage, bus communication, and switching transistor action together constitute physical time lag. The control current fails to correlate with the spindle absolute phase execution timing advance translation, and the phase misalignment on the time axis is prone to induce reverse miscompensation, ultimately limiting the improvement of feed accuracy.

[0005] Therefore, the present invention provides a high-precision feed control method for a milling machine to overcome the shortcomings of the prior art. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a high-precision feed control method for milling machines. This method solves the problems of insufficient mechanical tool deformation compensation accuracy and easy generation of high-frequency vibration marks on the machined surface caused by the inability to separate high-frequency cutting excitation loads, difficulty in overcoming the thermomagnetic drift of micro-motion mechanisms and material hysteresis nonlinearity, and physical lag in control timing of existing milling machine feed systems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A high-precision feed control method for a milling machine, wherein the feed system of the milling machine includes a feed axis, and the actuator includes an excitation coil, a sealed chamber containing magnetostrictive material, an outer metal flexible diaphragm shell, and a worktable saddle, comprising the following steps: Obtain the geometric characteristic parameters and rotation phase data of the face milling cutter, calculate the discrete time sequence of the total cutting thrust load on the feed bearing, and separate the discrete time sequence into a quasi-static low-frequency load component and a high-frequency dynamic excitation component; A test signal is injected into the excitation coil to acquire the waveform and calculate the reference resistance and reference inductance, as well as the real-time test resistance and real-time test inductance. The ohmic difference and Henry difference are calculated to generate the thermomagnetic coupling gain factor. The high-frequency dynamic excitation component is divided by the equivalent structural stiffness of the feed axis to obtain the tool deformation amount, and the tool deformation amount is inverted and set as the target micro-compensation displacement. The initial compensation command current is calculated in combination with the thermomagnetic coupling gain factor under the preset discrete time step. The drive signal is generated by performing a forward translation operation based on the real-time test inductor and the real-time test resistor. The excitation coil receives the driving signal to cause the magnetic flux in the sealed cavity to alternate, polarizing the magnetostrictive material in the sealed cavity to generate endogenous stress, which forces the outer metal flexible diaphragm shell and the worktable saddle to generate high-frequency micro-displacements. The high-frequency micro-displacements and the tool deformation are superimposed and canceled in the same spatial coordinate system, realizing the macro-micro composite motion of the feed axis.

[0008] By employing the above technical solution, a dynamic excitation source for structural resonance is extracted through discrete analysis and frequency domain separation based on the thrust of the face milling cutter's rotation phase. This is combined with data on the temperature rise and permeability drift of the excitation coil to establish the thermomagnetic coupling gain. The high-frequency load is mapped to an anti-phase micro-compensation displacement based on the transmission stiffness, and a precise driving current is generated by inputting a hysteresis differential evolution inverse model and a time-lead translation mechanism. Ultimately, this excites the magnetostrictive material to overcome the yield limit of the flexible mechanism and output micro-displacement. Therefore, a high-precision composite control effect is achieved in the same spatial coordinate system and time axis, which real-time offsets the tool deflection caused by intermittent cutting, eliminates high-frequency vibration marks on the machined surface, and maintains the control system's resistance to long-term thermomagnetic attenuation.

[0009] Preferably, the specific steps for calculating the discrete-time sequence of the total cutting thrust load on the feed bearing are as follows: Calculate the specific spatial coordinates of each tooth of the face milling cutter on the surface of the workpiece material at any sampling time, and determine whether each tooth is currently in the effective working range of cutting into the workpiece; Using the current spindle rotational angular velocity data and the feed axis linear feed rate data, the relative motion trajectory of the cutting tooth at the cutting point position is calculated, and the instantaneous cutting thickness of a single tooth modulated by the spindle phase angle is calculated. The instantaneous cutting thickness of a single tooth is continuously multiplied with the specific cutting force coefficient and the axial cutting depth to obtain the local cutting thrust generated by a single cutting tooth when it cuts into the workpiece. The local cutting thrust generated by all cutting teeth in the cutting state is discretely summed and accumulated at the current time node to generate a discrete time sequence of the total cutting thrust load on the feed bearing.

[0010] By employing the above technical solution, and utilizing the spatial kinematic geometry mapping of multi-tooth intermittent cutting, the relative trajectory and instantaneous thickness of a single tooth entering the working zone are extracted. Combined with the material cutting coefficient, a discrete accumulation operation of the local cutting thrust is performed. Therefore, a time-varying thrust theoretical sequence that highly matches the actual intermittent stress state in face milling is obtained, providing an accurate broadband load disturbance source input for subsequent micro-compensation strategies.

[0011] Preferably, based on the above-mentioned discrete-time sequence of the total cutting thrust load on the feed bearing, the specific steps for separating the discrete-time sequence into a quasi-static low-frequency load component and a high-frequency dynamic excitation component are as follows: Extract the discrete time series of the total cutting thrust load, perform spectral analysis on the frequency components of the discrete time series, and set the frequency segmentation boundary parameters of the digital filter bank based on the analysis results and the pre-calibrated natural frequency of the feed shaft drive train. The discrete-time series of the total cutting thrust load is simultaneously imported into the digital filter bank, which includes a low-pass digital filter and a high-pass digital filter. Discrete convolution filtering is performed in the low-pass digital filter operation link to filter out high-frequency periodic fluctuation signal components and output the quasi-static low-frequency load component that causes the overall force deformation of the transmission chain system. In the high-pass digital filter operation link, stopband attenuation operation is performed to eliminate DC bias components and slow evolution trend terms, and the high-frequency dynamic excitation component that causes tool forced resonance is extracted by rapidly fluctuating load components.

[0012] By adopting the above technical solution, a dual-channel digital filtering mechanism adaptively calibrated based on the natural frequency of the transmission chain foundation is used. This mechanism extracts the macroscopic deformation background through discrete convolution operations and extracts transient abrupt change peaks by performing stopband attenuation in the high-pass link. Therefore, it achieves the effect of stripping away steady-state cutting forces and precisely isolating the high-frequency forced resonance source of the tool system, avoiding miscompensation interference from micro-actuators on the machine tool foundation's propulsion load.

[0013] Preferably, the specific steps for calculating the reference resistance and reference inductance are as follows: During the cold operation phase before the system is cut, a weak AC voltage test signal with constant amplitude and frequency is injected into both ends of the excitation coil. The amplitude of the weak AC voltage test signal is set so that the excited alternating electromagnetic thrust cannot overcome the static friction dead zone of the mechanical structure. The current response waveform sequence generated in the synchronous acquisition loop is used to extract the amplitude attenuation rate and phase delay angle between the voltage input sequence and the current output sequence. Based on the amplitude attenuation rate and phase delay angle, the reference complex impedance characteristic parameters of the excitation coil in a cold environment are calculated, the reference resistor and the reference inductor are separated, and the reference resistor and the reference inductor are saved to a non-volatile memory.

[0014] By employing the above technical solution, and utilizing a weak AC voltage injection mode limited to the lower limit amplitude of the mechanical static friction dead zone, the amplitude and phase hysteresis deviation of the current waveform are captured in the unprocessed cold state. Therefore, the effect of extracting the background electromagnetic characteristics of the excitation mechanism without mechanical motion disturbance is achieved, establishing a reliable initial physical reference for subsequent evaluation of material heating and the evolution of internal magnetic domain structure.

[0015] Preferably, based on the above calculation of the reference resistor and the reference inductor, the specific steps for calculating the real-time test resistor and the real-time test inductor are as follows: After entering the formal cutting state, the online monitoring and extraction operation is started. According to a fixed sampling period, the frequency sweep sensor voltage signal of the high frequency sweep mode is superimposed on the existing drive control current command. Real-time synchronous acquisition of time series data of voltage and current observations returned from both ends of the excitation coil, including superimposed components; A digital bandpass filter is used to remove low-frequency main drive signals and high-frequency noise interference caused by cutting loads. Discrete Fourier transform is performed on the retained microsensor response frequency band data to calculate the real-time test resistance and real-time test inductance of the excitation coil as it changes with heating and magnetic field evolution.

[0016] By employing the above technical solution, and utilizing a hardware path that coaxially superimposes a weak high-frequency sweep signal and the main drive current, combined with bandpass filtering and discrete Fourier transform to separate the sensing feedback signal in the frequency domain, the effect of real-time tracking and capturing the Joule thermal resistance drift of the coil and the dynamic permeability decay characteristics of the magnetic conductor is achieved without interfering with the normal micro-displacement output process of cutting.

[0017] Preferably, based on the above calculation of the real-time test resistance and the real-time test inductance, the specific steps for calculating the ohmic difference and Henry difference to generate the thermomagnetic coupling gain factor are as follows: Extract the real-time test resistor and the real-time test inductor calculated in the current sampling period, and retrieve the reference resistor and the reference inductor stored in the non-volatile memory; The ohmic difference between the real-time test resistor and the reference resistor, and the Henry difference between the real-time test inductor and the reference inductor are calculated to characterize the coil heating temperature rise and the severity of magnetic permeability drift of the magnetic material. Obtain the preset reference coupling gain factor, resistive coupling correction coefficient, and inductive coupling correction coefficient, and substitute them, along with the ohmic difference and Henry difference, into the gain update model for calculation. Compensate for thermodynamic displacement decay and hysteresis performance decay by superimposing linear decay weights to generate the updated thermomagnetic coupling gain factor.

[0018] By adopting the above technical solution, and by using the difference mapping between online test electromagnetic parameters and cold-state reference parameters, combined with a fixed proportion of resistance and inductance attenuation correction weights to perform linear algebraic correction of the joint drift, the system achieves an adaptive update of the electromechanical conversion ratio, ensuring that the displacement amplitude output accuracy of the actuator unit does not degrade under continuous heavy load and heating conditions.

[0019] Preferably, the specific steps of dividing the high-frequency dynamic excitation component by the equivalent structural stiffness of the feed axis to obtain the tool deflection deformation, and inverting the tool deflection deformation to set as the target micro-compensation displacement are as follows: Extract the equivalent structural stiffness of the feed axis in the current machine tool transmission chain, and establish a mapping relationship model between the high-frequency dynamic excitation component and the elastic deformation of the structure; The high-frequency dynamic excitation component sequence is input into the mapping relationship model to continuously calculate the micro-scale displacement deviation of the feed axis under the current force state. The tool deflection deformation is obtained by dividing the high-frequency dynamic excitation component by the equivalent structural stiffness of the feed axis. Then, the target micro-compensation displacement with the opposite phase is generated by the inversion operation, which is used to require the active generation of reverse stroke compensation to offset the tool retraction caused by the cutting thrust.

[0020] By adopting the above technical solution, and utilizing the equivalent stiffness characteristic value of the overall axial resistance to deformation of the machine tool transmission components, the direct division of the dynamic excitation component's deviation from the mechanical microscopic level is performed, and the polarity is reversed to set the target trajectory. Therefore, the effect of accurately quantifying the high-frequency forced tool deflection depth at the physical structure level is obtained, and the basic algebraic reference surface for the equal-amplitude opposing stroke output of the drive compensation mechanism is established.

[0021] Preferably, based on the above-mentioned target micro-compensation displacement, the specific steps for calculating the initial compensation command current in conjunction with the thermomagnetic coupling gain factor at a preset discrete time step are as follows: A hysteresis inverse solution model based on differential equations is constructed, which decomposes the target micro-compensation displacement into a linearly reversible stiffness response part and a nonlinear, unmeasurable hysteresis hysteresis part. Extract the pre-calibrated linear stiffness ratio coefficient and hysteresis displacement weight coefficient, and use the linear stiffness ratio coefficient and the hysteresis displacement weight coefficient as the core weight parameters for inverse model solution; Substituting the target micro-compensation displacement into the nonlinear hysteresis displacement mapping relationship, and combining it with the received thermomagnetic coupling gain factor, the differential state evolution operation is performed by combining the current hysteresis intermediate variable state, and the high-frequency dynamic asymmetric pulse current used to offset the material's inherent hysteresis response error is solved in reverse as the initial compensation command current.

[0022] By employing the above technical solution, a displacement decomposition model is constructed by simultaneously extracting linear stiffness characteristics and nonlinear internal friction loss characteristics. Differential equations are used to track the time evolution derivatives of the hysteresis state variables and inversely deduce the electromagnetic excitation source input. Therefore, the effect of overcoming the multi-valued nonlinear memory barrier of magnetostrictive materials is achieved, ensuring that the generated high-frequency pulsed current drives the solid material to strictly conform to the anti-phase compensation trajectory evolution.

[0023] Preferably, based on the above-calculated initial compensation command current, the specific steps for generating the drive signal by performing a lead shift operation according to the real-time test inductor and the real-time test resistor are as follows: The electrical lag time constant under the current operating condition is calculated by dividing the obtained real-time test inductance by the real-time test resistance. The total advance time is obtained by summing the electrical lag time constant with the inherent digital communication delay of the system and the switching delay of the power amplifier. The total advance time is converted into the corresponding advance spindle angle. The time node of the next cutter tooth entering the workpiece is predicted by combining the rotation phase window of the spindle absolute encoder. The initial compensation command current is then shifted ahead on the time axis to generate the final drive signal.

[0024] By adopting the above technical solution, the time-varying inductance-to-resistance ratio is used to reflect the physical energy storage delay time. This value is then combined and accumulated with the digitally fixed delay scalar of the control link in the time dimension to convert it into the rotor phase angle advance. Therefore, a control effect that compensates for the energy charging delay and signal transmission misalignment of the electromagnetic system is achieved, ensuring a perfect match between the peak value of the micro-displacement output and the transient node of chatter caused by the actual contact of the tool with the material.

[0025] Preferably, based on the above-mentioned generation of the final drive signal, the specific steps for forcing the outer flexible metal diaphragm shell and the worktable saddle to generate high-frequency micro-displacements, wherein the high-frequency micro-displacements and the tool deformation are superimposed and canceled out in the same spatial coordinate system, are as follows: The magnetostrictive material in the sealed cavity undergoes instantaneous expansion and contraction of its spatial volume under the polarization effect of the high-frequency alternating magnetic field generated by the excitation coil, which is transformed into the outward expansion of the endogenous stress. The intrinsic stress forcibly overcomes the axial local elastic yield limit of the flexible metal diaphragm shell, forcing the worktable saddle to generate a controlled reverse high-frequency micro-displacement along the axial direction. The high-frequency micro-displacement and the tool deformation are superimposed and fused in the actual spatial coordinate system in the same frequency but opposite phase, and the in-situ structural cancellation is completed by algebraically adding the spatial dimensions.

[0026] By employing the above technical solution, the internal encapsulation material undergoes volume deformation and expansion using a closed electromagnetic field, which then pushes against the external rigid diaphragm shell, which has localized elastic weaknesses, through a stress-connected channel. This achieves a decoupled and combined effect between the macroscopic cutting load-bearing force flow and the microscopic compensating displacement dynamic flow at the same mechanical interface of the machine tool, completing a high-frequency structural closed-loop cancellation action in physical space at the machining contact point.

[0027] This invention provides a high-precision feed control method for a milling machine. It has the following advantages: 1. This invention obtains the geometric characteristic parameters and rotation phase data of the face milling cutter, calculates the total cutting thrust load, and separates the high-frequency dynamic excitation component. This component is then divided by the equivalent structural stiffness of the feed axis and set as the target micro-compensation displacement. This scheme transforms the high-frequency forced vibration source caused by intermittent cutting into a specific displacement compensation reference value. Furthermore, a magnetostrictive material drives the outer flexible metal diaphragm shell to generate a high-frequency micro-displacement with the same frequency but opposite phase, directly offsetting the tool deflection caused by cutting within the same spatial coordinate system. This eliminates the mechanical tool deflection caused by macroscopic cutting load at the physical structural level, improves the dynamic position accuracy of the feed axis, and reduces high-frequency vibration marks on the machined surface.

[0028] 2. This invention extracts the test resistance and inductance of the excitation coil in real time during operation, calculates the difference between Ohm and Henry's Law to generate a thermomagnetic coupling gain factor, and constructs a hysteresis inverse solution model based on differential equations to generate the initial compensation command current. This scheme performs real-time amplitude gain updates and material hysteresis nonlinearity corrections on the drive current at the control layer, actively compensating for the thermodynamic displacement attenuation of the actuator unit and the permeability drift of the magnetic material under continuous heating conditions. This maintains the micro-displacement output accuracy of the actuator in complex thermomagnetic coupling environments and prevents the degradation of compensation efficiency caused by long-term continuous operation of the machine tool.

[0029] 3. This invention calculates the electrical lag time constant by using the quotient of the real-time test inductance and resistance. This constant is then added to the system communication delay and the power amplifier switching delay to obtain the total advance time, which is then converted into an advance spindle rotation angle and used to perform a lead translation operation on the control signal. This scheme compensates for the time lag caused by the delayed charging of the electromagnetic circuit and the signal transmission through a feedforward prediction mechanism. This ensures that the small reverse displacement generated by the actuator is synchronized with the mechanical chatter caused by the moment the cutter teeth enter the workpiece, avoiding miscompensation and instability of the feed system due to control phase delay. Attached Figure Description

[0030] Figure 1 This is a diagram illustrating the architecture of the high-precision feed control system for the milling machine of the present invention. Figure 2 This is a flowchart illustrating the high-precision feed control method for the milling machine of the present invention. Figure 3 This is a flowchart illustrating the decoupling of spindle angle acquisition and cutting disturbance force in this invention. Figure 4 The flowchart of the present invention is shown below for real-time extraction of high-frequency characteristics of coils to correct thermomagnetic coupling gain. Figure 5 This is a flowchart of the anti-delay micro-compensation pulse instruction calculation based on the hysteresis inverse model of the present invention; Figure 6 This is a flowchart illustrating the electromechanical composite coordination of the main servo basic motion and micro-compensation current in this invention. Figure 7 This is a simulation comparison diagram of the microscopic displacement tracking curve of the present invention; Figure 8 These are comparison images of the microscopic contour morphology of the processed surfaces according to the present invention. Detailed Implementation

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] See attached document Figure 1 This invention provides a high-precision feed control method for a milling machine, which can be executed by the system and may include: a spindle synchronous observation module, an impedance sensing correction module, a hysteresis calculation feedforward module, and a rigid-flexible decoupling execution module.

[0033] The spindle synchronization observation module establishes a bidirectional data communication connection with the spindle servo driver and digital control system of the milling machine. The spindle synchronization observation module is used to acquire spindle speed data and absolute encoder rotation phase data in real time, and at the same time read preset face milling cutter geometric feature parameters and currently executed feed trajectory coordinate data from the digital control system.

[0034] The spindle synchronous observation module is equipped with a transient force analysis unit based on the characteristics of multi-tooth intermittent cutting. Utilizing the acquired spindle rotation phase data and the geometric parameters of the face milling cutter, the module calculates the instantaneous cutting thickness distribution during the sequential entry of each cutting tooth of the face milling cutter into the workpiece, and accordingly solves for the discrete-time sequence of the total cutting thrust load borne by the feed axis during its travel along the feed trajectory.

[0035] The output of the spindle synchronous observation module integrates a low-pass digital filter and a high-pass digital filter with independently adjustable cutoff frequencies. The spindle synchronous observation module simultaneously sends the calculated total cutting thrust load sequence into the filter bank, thereby separating the broadband total cutting thrust load into a quasi-static low-frequency load component that induces overall force deformation of the transmission chain system, and a high-frequency dynamic excitation component that induces forced resonance of the tool. The separated quasi-static low-frequency load component and high-frequency dynamic excitation component are then transmitted to the receiving end.

[0036] The test output of the impedance sensing correction module is electrically connected to the excitation coil contained in the rigid-flexible decoupling execution module, and its data output communicates with the hysteresis calculation feedforward module. When the milling machine is in the non-processing cold operating range, the impedance sensing correction module controls the built-in high-frequency signal generator to inject a weak AC voltage test signal with constant amplitude and frequency into the excitation coil. The corresponding current response waveform is acquired through the synchronous acquisition circuit, and the reference complex impedance characteristic parameters of the excitation coil in the cold environment are calculated accordingly.

[0037] The impedance sensing correction module activates its online monitoring mechanism after the milling machine enters the cutting process. Following a set sampling period, the module superimposes a small, sweep-frequency sensing voltage signal—which does not induce mechanical vibration in the system—onto the drive control current transmitted to the excitation coil. Simultaneously, it acquires the voltage and current time-series data returned from both ends of the excitation coil in real time, calculating the real-time test resistance and inductance of the excitation coil as it changes with heating and magnetic field evolution under this processing condition.

[0038] The impedance sensing correction module has a built-in gain correction calculation unit. Based on the calculated ohmic difference between the real-time test resistance and the reference resistance included in the reference complex impedance, and the Henry's law difference between the real-time test inductance and the reference inductance included in the reference complex impedance, the impedance sensing correction module calculates the joint drift deviation of the thermal and magnetic states. According to the pre-calibrated resistance attenuation weight and inductance attenuation weight, it calculates the new electromechanical conversion rate, generates the updated thermomagnetic coupling gain factor, and continuously sends it to the hysteresis solution feedforward module.

[0039] The data input terminals of the hysteresis calculation feedforward module are connected to the spindle synchronous observation module and the impedance sensing correction module, respectively. The hysteresis calculation feedforward module receives the high-frequency dynamic excitation component and the quasi-static low-frequency load component separated from the total cutting thrust load, and extracts the equivalent structural stiffness parameters of the milling machine feed drive train. The hysteresis calculation feedforward module uses the equivalent structural stiffness parameters to perform a division operation on the high-frequency dynamic excitation component to obtain the excitation deformation error, and sets its inverse phase as the target micro-compensation displacement required by the system.

[0040] The core arithmetic unit of the hysteresis solution feedforward module contains an operator for mapping the nonlinear characteristics of magnetostriction based on the hysteresis inverse equations. The hysteresis solution feedforward module uses the calculated target micro-compensation displacement as the input to this operator, and combines it with the received thermomagnetic coupling gain factor to perform differential state evolution calculations within a preset microsecond-level discrete time step. This results in a high-frequency dynamic asymmetric pulse current used to compensate for the inherent hysteresis response error of the material, which serves as the initial compensation command current.

[0041] The hysteresis calculation feedforward module is equipped with a timing control output interface. The hysteresis calculation feedforward module uses the real-time test inductance and resistance obtained by the impedance sensing correction module to calculate the electrical hysteresis time constant under the current operating condition. Simultaneously, it combines the spindle absolute encoder rotation phase window obtained by the spindle synchronization observation module to predict the tool tooth engagement time node, and performs a lead shift operation on the time axis on the initial compensation command current to generate the final drive signal output to the rigid-flexible decoupled execution module.

[0042] The rigid-flexible decoupled actuator module is installed at the mechanical connection point between the ball screw nut and the worktable saddle of the milling machine using a serial-parallel composite structure. The outer body of the rigid-flexible decoupled actuator module is covered by a flexible metal diaphragm housing made of high-strength alloy material. This flexible metal diaphragm housing is rigidly constrained in both radial and torsional directions, directly bearing the macroscopic feed thrust generated by the controlled rotation of the main servo motor and the reverse resistance generated by the cutting process.

[0043] The rigid-flexible decoupling actuator module has an absolutely sealed chamber constructed inside the flexible metal diaphragm housing, within which magnetostrictive material is solidified and encapsulated. A closed magnetic yoke, formed by stacking soft magnetic ferrite material, is tightly arranged around the outer perimeter of the sealed chamber, and the excitation coil is uniformly wound on the closed magnetic yoke. At the flange ends connecting the ball screw nut and the worktable saddle, austenitic stainless steel magnetic shielding plates are fitted to interrupt the path of alternating leakage magnetic flux to the surrounding machine tool castings.

[0044] After receiving the drive signal from the hysteresis calculation feedforward module, the excitation coil in the rigid-flexible decoupling execution module undergoes a high-frequency transient alternation of internal magnetic flux. Under the polarization of the alternating magnetic field, the magnetostrictive material within the sealed chamber experiences instantaneous expansion and contraction of its volume, generating endogenous stress sufficient to overcome the axial local elastic yield limit of the flexible metal diaphragm shell. This endogenous stress forces the worktable saddle to produce a high-frequency micro-displacement along the axial direction. This displacement, along with the tool deformation induced by the high-frequency dynamic excitation component, is superimposed in opposite directions with equal amplitude within the same spatial coordinate system, achieving a seamless integration of the macroscopic basic motion and microscopic deviation compensation motion of the feed axis.

[0045] See attached document Figure 2 The present invention provides a high-precision feed control method for a milling machine, which may include: Step S100 acquires the spindle rotation angle and decouples high- and low-frequency cutting disturbances. This is performed through the spindle synchronization observation module. Upon initiation of this step, a data communication link is established between the spindle servo drive and the digital control system. This data communication link is used to periodically acquire real-time spindle speed data and rotation angle phase data output from the absolute encoder. Simultaneously, the geometric feature parameters of the face milling cutter used in the current machining operation and the trajectory coordinates of the feed axis are extracted from the digital control system.

[0046] After data acquisition, transient force analysis is performed. Based on the spatial distribution characteristics of multi-tooth intermittent cutting, and combined with the acquired geometric feature parameters and rotation phase data of the face milling cutter, the instantaneous cutting thickness of each cutting tooth of the face milling cutter when it enters the workpiece material is calculated. Based on the instantaneous cutting thickness, a discrete-time sequence of the total cutting thrust load borne by the feed axis during its travel along the feed trajectory is generated.

[0047] The discrete-time sequence of the total cutting thrust load is subjected to frequency domain segmentation. The calculated total cutting thrust load sequence is then imported into a digital filter bank for computation. The digital filter bank outputs two signals: one is the quasi-static low-frequency load component that induces overall force deformation in the transmission chain system, and the other is the high-frequency dynamic excitation component that induces forced resonance in the tool. These two components are then output to the receiving port of the subsequent stage.

[0048] Step S200 involves real-time extraction of high-frequency coil characteristics to correct the thermomagnetic coupling gain. This is performed via an impedance sensing correction module. A reference state calibration is performed during the cold-state operation phase before the feed control method enters the cutting process. A weak AC voltage test signal with constant amplitude and frequency is injected into the excitation coil, and the response current waveform is recorded simultaneously. Based on this, the reference complex impedance characteristic parameters of the excitation coil under cold conditions are calculated, including the reference resistance and reference inductance.

[0049] After the feed control method enters the cutting state, online monitoring and extraction are performed. A small swept-frequency sensing voltage signal is superimposed onto the drive control current transmitted to the excitation coil according to a fixed sampling period. The amplitude of the small swept-frequency sensing voltage signal is set within a range that will not cause mechanical vibration of the system. Real-time synchronous acquisition of the voltage and current time-series data returned from both ends of the excitation coil is used to calculate the real-time test resistance and real-time test inductance of the excitation coil as it changes with heating and magnetic field evolution.

[0050] Gain correction calculations are performed based on the extracted feature parameters. The ohmic difference between the real-time test resistor and the reference resistor, and the Henry's law difference between the real-time test inductor and the reference inductor are calculated. Based on the pre-calibrated resistor attenuation weights and inductor attenuation weights within the system, the electromechanical conversion rate drift under the combined effects of thermal and magnetic states is calculated, an updated thermomagnetic coupling gain factor is generated, and the thermomagnetic coupling gain factor is continuously sent to the subsequent solution port.

[0051] Step S300 calculates the anti-delay micro-compensation pulse command based on the hysteresis inverse model. This is executed through the hysteresis calculation feedforward module. It receives the input high-frequency dynamic excitation component and the quasi-static low-frequency load component. It extracts the preset equivalent structural stiffness parameters of the milling machine feed drive chain. The high-frequency dynamic excitation component is divided by the equivalent structural stiffness parameters to obtain the micro-displacement deviation induced by the excitation disturbance, and this micro-displacement deviation is inverted and set as the target micro-compensation displacement to be generated by the system.

[0052] The target micro-compensation displacement is imported into the hysteresis inverse equation system operator for mapping solution. Combined with the received thermomagnetic coupling gain factor, differential state evolution calculation is performed within a preset microsecond-level discrete time step. The differential state evolution calculation yields a high-frequency dynamic asymmetric pulse current used to compensate for the inherent hysteresis response error of the magnetostrictive material; this high-frequency dynamic asymmetric pulse current serves as the initial compensation command current.

[0053] The initial compensation command current is subjected to timing-controlled feedforward operation. The electrical lag time constant under the current operating condition is calculated by dividing the acquired real-time test inductance by the real-time test resistance. The timing of the next cutter tooth entry into the workpiece is predicted by combining the rotation phase window of the spindle absolute encoder. The initial compensation command current is then shifted forward on the time axis according to the electrical lag time constant to generate the final drive signal for output.

[0054] Step S400 involves the superposition and coordination of the main servo's basic motion and the micro-compensation current. This is performed through a rigid-flexible decoupling execution module. During the controlled rotation of the main servo motor driving the feed axis, the flexible metal diaphragm shell covering the system maintains rigid constraints in the radial and torsional directions. The macroscopic feed thrust and the reverse resistance generated by cutting are directly borne and transmitted by the flexible metal diaphragm shell.

[0055] After receiving the final drive signal, the excitation coil causes a high-frequency transient alternation of magnetic flux inside the sealed cavity. Under the polarization of the alternating magnetic field, the magnetostrictive material within the sealed cavity undergoes instantaneous expansion and contraction of its volume. This change in volume is converted into outward-expanding endogenous stress. Simultaneously, austenitic stainless steel magnetic shielding plates mounted on the flange faces at both ends interrupt the path of alternating leakage magnetic flux to the surrounding machine tool castings.

[0056] The intrinsic stress overcomes the axial local elastic yield limit of the flexible metal diaphragm shell. The elastic yield forces the worktable saddle to generate a high-frequency micro-displacement along the axial direction. The generated high-frequency micro-displacement and the tool deformation caused by the high-frequency dynamic excitation component in the spatial coordinate system are in opposite directions and have the same amplitude. The two displacements are superimposed and cancel each other in the spatial dimension, completing the composite execution of the macroscopic basic motion and micro-deviation compensation motion of the feed axis.

[0057] See attached document Figure 3 Step S100 involves acquiring the spindle rotation angle and decoupling high- and low-frequency cutting disturbance forces. This is executed through the spindle synchronous observation module.

[0058] Step S101: System Initialization and Cutting Parameter Acquisition. The system initialization process is executed before the milling machine spindle starts and reaches the set machining speed. A high-speed industrial Ethernet communication connection is established between the milling machine's digital control system and the spindle servo drive. The data exchange link is confirmed to be in a stable transmission state with no packet loss and low latency, establishing the communication foundation for subsequent bidirectional transmission of real-time control commands and sensor data.

[0059] After the communication connection is established, the real-time rotational angular velocity data of the spindle is continuously read from the servo feedback loop inside the spindle servo driver according to the set microsecond-level data sampling period. Simultaneously, high-resolution angular phase data output from the absolute encoder installed at the tail end of the spindle is acquired. By analyzing the high-resolution angular phase data, the precise rotational angle of the spindle relative to the absolute reference zero point of the machine tool at the current sampling moment is obtained, tracking the transient position of the face milling cutter in the machining space.

[0060] While collecting real-time operating status data, a parameter call request is sent to the milling machine's digital control system to extract the pre-set machining process parameters and tool geometry information for the current operation. The acquired parameter information includes the total number of effective teeth mounted on the face milling cutter, the specific specific cutting force coefficient, and the given axial cutting depth. The acquired spindle rotational angular velocity data, high-resolution rotational phase data, and machining process parameters and tool geometry information are uniformly stored in the internal high-speed cache register.

[0061] Step S102: Time-space cutting disturbance prediction modeling. High-resolution rotational phase data and the total number of effective teeth are extracted from the cache register to construct the kinematic analytical mapping relationship for the multi-tooth discontinuous cutting process. Based on the total number of effective teeth, the circumferential distribution angle of the face milling cutter is calculated in equal parts to determine the fixed phase difference of each tooth relative to the spindle reference point in the circumferential direction. Combined with the real-time updated high-resolution rotational phase data, the specific spatial coordinates of each tooth on the workpiece material surface at any sampling time are calculated.

[0062] Based on the calculated spatial coordinates of each cutting tooth, it is determined whether each tooth is currently within its effective working range for cutting into the workpiece. For cutting edges within the effective working range, the relative motion trajectory of the tooth at the cutting point is calculated using the current spindle rotational angular velocity data and the feed axis linear feed rate data. Based on the projection of this relative motion trajectory in the direction perpendicular to the cutting surface, the instantaneous cutting thickness of a single tooth, modulated by the spindle phase angle, is calculated.

[0063] After obtaining the instantaneous cutting thickness of each effective working tooth, it is continuously multiplied by the specific cutting force coefficient and the axial cutting depth to obtain the local cutting thrust generated by a single tooth when cutting into the workpiece. The local cutting thrust generated by all teeth in the cutting state is discretely summed and accumulated at the current time point to obtain the theoretical time-varying thrust load on the feed bearing, generating and outputting the instantaneous total cutting force. The specific calculation formula for this instantaneous total cutting force is as follows: ; in, The instantaneous total cutting force represents the theoretical time-varying thrust load on the feed bearing when the tool enters the workpiece, and its unit is N; The total number of teeth represents the number of effective cutting edges mounted on the edge of the milling cutter head, and its value ranges from 2 to 30. The total number of teeth represents the number of effective cutting edges mounted on the edge of the milling cutter head, and its value ranges from 2 to 30. The specific cutting force coefficient represents the constant cutting force required to remove a unit area of ​​material from a particular material, and its unit is N / mm. 2 ; For the first The instantaneous cutting thickness of each cutting tooth represents the depth of penetration of a single tooth modulated by the spindle phase angle, in mm. The axial cutting depth represents the constant depth of cut into the workpiece matrix, expressed in mm. The instantaneous rotation phase of the main axis; This is a variable representing system runtime.

[0064] Step S103: High and low frequency excitation force signal filtering and separation. Extract the time-series data of the calculated instantaneous total cutting force and perform spectral analysis on its frequency components. Based on the analysis results and the pre-calibrated natural frequency of the feed axis drive train, set the frequency division boundary parameters of the digital filter bank. The digital filter bank contains independently configured and parallel-operating low-pass and high-pass digital filters. The time-series data of the instantaneous total cutting force is used as the input source and simultaneously imported into both the low-pass and high-pass digital filters.

[0065] In the low-pass digital filter operation chain, discrete convolution filtering is performed on the time-series data of the instantaneous total cutting force to filter out high-frequency periodic fluctuation signal components. Signal components related to the material basis removal rate and exhibiting low-frequency, slowly varying trends are retained, outputting a macroscopic quasi-static load. The output macroscopic quasi-static load characterizes the source of low-frequency thrust that induces overall structural bending and transmission elastic deformation in the feed axis system within the machining time span.

[0066] In the high-pass digital filter operation chain, stopband attenuation is performed on the time-series data of the instantaneous total cutting force to eliminate DC bias components and slowly varying evolution trend terms. The components causing rapid load fluctuations due to the intermittent periodic entry and exit movements of the multi-tooth workpiece are extracted, and high-frequency dynamic excitation components are output. These high-frequency dynamic excitation components characterize the load variables that cause forced high-frequency resonance in the tool system and leave periodic vibration marks on the workpiece surface. The separated macroscopic quasi-static load and high-frequency dynamic excitation components are sent to the next-stage receiving port via two independent data channels.

[0067] See attached document Figure 4 Step S200 involves real-time extraction of the high-frequency characteristics of the coil to correct the thermomagnetic coupling gain. This is performed via the impedance sensing correction module.

[0068] Step S201: Reference Complex Impedance Cold State Calibration. The system is in a static stage before cutting. The machine tool has reached a thermal equilibrium cold state consistent with the ambient temperature after a long period of shutdown. The main drive power supply to the feed axis actuator is cut off, keeping only the power supply to the data acquisition network and low-voltage control circuit active.

[0069] A high-frequency, weak AC voltage test signal with a specific frequency is injected into both ends of the excitation coil. The amplitude of this high-frequency, weak AC voltage test signal is set within a small, safe range. This ensures that the alternating electromagnetic thrust excited by this small voltage in the coil cannot overcome the static friction dead zone of the mechanical structure, thereby preventing actual mechanical displacement of the system.

[0070] During voltage injection, the current response time series generated in the excitation coil circuit is synchronously acquired. The amplitude attenuation rate and phase delay angle between the voltage input sequence and the current output sequence are extracted using a quadrature lock-in amplifier. Based on the amplitude attenuation rate and phase delay angle, the reference complex impedance is calculated, and the reference resistor and reference inductor are separated and stored in non-volatile memory. The specific calculation formula for this reference complex impedance is as follows: ; in, The reference complex impedance represents the initial blocking characteristic of the excitation coil in the complex domain under cold conditions and at a specific test frequency, and its unit is 1000 kJ / m². The angular frequency is used for testing, representing the rate of change of the rotating phasor of a small input test AC signal, and its unit is ω. The reference inductance represents the baseline value of the magnetic field energy stored in the coil under the initial permeability conditions, and its unit is 1. It is the imaginary unit.

[0071] Step S202: High-frequency sweep voltage injection and real-time impedance extraction. After the milling machine enters the formal cutting state, the online monitoring process is initiated. With the continuous application of a large current drive signal, the Joule heating effect inside the excitation coil and the nonlinear alternation of the surrounding magnetic circuit cause real-time drift in electrical parameters. A fixed millisecond-level sampling period is set for continuous extraction of internal parameters.

[0072] At a fixed sampling period, the micro-sensor voltage signal is superimposed on the existing drive current command of the system. The micro-sensor voltage signal adopts a high-frequency sweep mode, and its frequency bandwidth avoids the inherent resonant frequency point of the machine tool structure. The energy of the micro-sensor voltage is limited to an extremely low level to prevent interference with normal cutting feed movements.

[0073] Real-time acquisition of voltage and current observations, including superimposed components, is performed. A digital bandpass filter is used to remove low-frequency main drive signals and high-frequency noise interference caused by cutting loads. Discrete Fourier transform is performed on the retained microsensor response frequency band data to calculate the real-time test resistance and real-time test inductance of the excitation coil at the current moment.

[0074] Step S203 involves adaptive updating of the thermomagnetic coupling gain based on impedance drift. The real-time test resistance and inductance calculated for the current sampling period are extracted. The previously saved reference resistance and inductance are retrieved from non-volatile memory. The ohmic difference between the real-time test resistance and the reference resistance, and the Henry's law difference between the real-time test inductance and the reference inductance are calculated. These two differences characterize the coil heating temperature rise and the severity of permeability drift in the magnetic material.

[0075] The preset reference coupling gain factor, resistive coupling correction coefficient, and inductive coupling correction coefficient are obtained from the controller's internal parameter table. The obtained parameters, along with the calculated ohmic and Henry differences, are substituted into the gain update model for solution. Thermodynamic displacement decay and hysteresis performance decay are compensated for by a linear decay weighting method.

[0076] The latest real-time coupled gain factor is generated through a gain update model. This generated real-time coupled gain factor is transmitted to the high-speed data bus, overwriting the historical values ​​from the previous cycle. This real-time coupled gain factor serves as the basis for the conversion ratio in the feedforward compensation operation, ensuring that the micro-displacement output accuracy of the system does not degrade under long-term continuous cutting conditions. The specific calculation formula for this real-time coupled gain factor is as follows: ; in, The real-time coupling gain factor represents the actual conversion capability of the coil to excite micro-mechanical displacement per unit current under the current thermal conditions, and its unit is mm / A. The reference coupling gain factor represents the basic electromechanical conversion constant calibrated at the factory cold state, and its unit is mm / A; This is the resistance coupling correction factor, which represents the weight of thermodynamic displacement attenuation caused by the temperature rise due to resistance heating, and its unit is 1. ; For real-time resistance testing, it represents the ohmic resistance value of the coil under real-time heating conditions during processing, with units of Ω. ; The reference resistor represents the initial DC ohmic loss of the conductors inside the coil at the test frequency band, in units of... ; This is the inductive coupling correction factor, which represents the weight of hysteresis performance attenuation caused by the permeability drift of the magnetic material, and its unit is 1. ; For real-time testing of inductance, it represents the real-time magnetic field energy storage characteristic of the coil under changes in thermomagnetic environment, with units of . ; The reference inductance represents the background value of the magnetic field energy stored in the coil under the initial permeability environment, and its unit is H.

[0077] See attached document Figure 5Step S300 calculates the anti-delay micro-compensation pulse command based on the hysteresis inverse model. This is executed through the hysteresis calculation feedforward module.

[0078] Step S301: Calculation of high-frequency tool deflection elastic deformation. Receive the separated high-frequency dynamic excitation component data sequence from the buffer of the previous step. Establish a parameter reading channel with the feed axis servo controller and extract the static and dynamic stiffness characteristic parameters of the current feed axis system. The equivalent structural stiffness of the feed axis, as a key parameter, comprehensively reflects the elastic resistance characteristics of mechanical transmission components such as the ball screw, guide rail, and connecting bearings under axial force.

[0079] By utilizing the extracted equivalent structural stiffness of the feed axis, a mapping model between the high-frequency dynamic excitation component and the elastic deformation of the structure is established. Considering the rapidly fluctuating characteristics of the high-frequency dynamic excitation component, the feed axis system will experience microscopic axial elastic yielding under this high-frequency load, resulting in a high-frequency tool deflection phenomenon. This microscopic displacement deviation is directly superimposed on the basic feed motion, causing the tool cutting edge to deviate from its theoretical motion trajectory, thereby affecting the microscopic morphology of the machined surface.

[0080] The high-frequency dynamic excitation component sequence is input into the mapping relationship model, and the micro-displacement deviation of the feed axis under the current stress state is continuously calculated based on the stress-strain linear law. This micro-displacement deviation is represented on the time axis as a displacement fluctuation signal with the same frequency and phase as the high-frequency dynamic excitation component. The calculated micro-displacement deviation is smoothed in both amplitude and phase to provide a basic reference for setting the subsequent anti-phase target micro-compensation displacement.

[0081] Step S302: Setting the Inverse Target Micro-Compensation Displacement. The system acquires the calculated micro-displacement deviation time series and enters the compensation displacement target planning stage. To compensate for machining errors caused by high-frequency tool deflection, a compensation displacement command with the same amplitude but opposite direction to the micro-displacement deviation is set. The equivalent structural stiffness of the feed axis and the high-frequency dynamic excitation components used in calculating the micro-displacement deviation are extracted.

[0082] Based on the error cancellation principle, a target micro-compensation displacement calculation model is constructed. The high-frequency dynamic excitation component is divided by the equivalent structural stiffness of the feed axis to obtain the positive forced deformation. Then, an inverse target micro-compensation displacement is generated through an inversion operation. This target micro-compensation displacement command requires the actuator to actively generate a reverse stroke compensation amount to counteract the tool retraction caused by the cutting thrust at the tool cutting point in real time.

[0083] The target micro-compensation displacement calculation model continuously outputs a sequence of target micro-compensation displacements. This sequence is stored in a high-speed instruction buffer for subsequent nonlinear inverse calculation stages. The specific calculation formula for the target micro-compensation displacement is as follows: ; in, The target micro-compensation displacement represents the precise reverse stroke amount that the execution module needs to actively generate to offset the error, in mm. The high-frequency dynamic excitation component represents the high-frequency periodic load in the total cutting force that induces mechanical resonance and surface rippling, and its unit is N; The equivalent structural stiffness of the feed axis represents the comprehensive macroscopic mechanical characteristic value of the machine tool's overall transmission chain resisting axial elastic deformation, with units of N / mm.

[0084] Step S303: Hysteresis Inverse Model System State Solution. The target micro-compensation displacement sequence is continuously retrieved from the high-speed instruction buffer. Because the execution stage uses a material with magnetostrictive properties, this material exhibits significant nonlinear hysteresis during the conversion of electromagnetic energy into mechanical displacement. This nonlinear hysteresis manifests as a multi-valued mapping relationship between the input current and the output displacement; directly inputting a current proportional to the target displacement into the coil will cause the actual output displacement to deviate from the target trajectory.

[0085] To overcome the compensation error caused by the inherent hysteresis characteristics of materials, a hysteresis inverse solution model based on differential equations is constructed. This model takes the target micro-compensation displacement as the input variable and decomposes it into a linearly reversible stiffness response component and a nonlinear, unmeasurable hysteresis component. Pre-calibrated linear stiffness ratio coefficients and hysteresis displacement weight coefficients are extracted and used as the core weight parameters for the inverse model solution.

[0086] Within each control cycle, the target micro-compensation displacement is substituted into the nonlinear hysteresis displacement mapping relationship. The current state of the hysteresis intermediate variables is then combined with the equations to inversely solve for the required compensation command current. This compensation command current is an asymmetric pulse signal that has undergone nonlinear pre-distortion processing, enabling precise driving of the magnetostrictive material to produce the expected anti-phase micro-stroke. The specific calculation formula for this nonlinear hysteresis displacement mapping is as follows: ; in, The target micro-compensation displacement represents the precise reverse stroke amount that the execution module needs to actively generate to offset the error, in mm. is the linear stiffness ratio coefficient, which represents the conversion weight parameter of the mechanical displacement produced by the magnetostrictive effect in the linear section, and the unit is mm / A; To compensate for the command current, it represents the high-frequency dynamic asymmetric pulse current used to drive the coil, calculated by the hysteresis inverse system, and is measured in amperes (A). The hysteresis displacement weighting coefficient represents the proportion of displacement contribution caused by nonlinear factors such as internal magnetic domain friction loss, and is expressed in mm. As a hysteresis intermediate variable, it represents a mathematical intermediate quantity characterizing the unmeasurable hysteresis trajectory state inside the system.

[0087] Step S304: Hysteresis Internal State Calculus Evolution Update. After completing the inverse solution of the compensation command current at the current moment, the hysteresis intermediate variables are updated synchronously to ensure that the inverse model solution for the next control cycle can be deduced along the correct nonlinear hysteresis trajectory. The gradient of the compensation command current at the current moment is obtained, and the time derivative of the compensation command current is calculated. This time derivative reflects the transient surge rate or transient decay rate of the driving signal.

[0088] Preset hysteresis loop control parameters are retrieved from the controller's non-volatile memory. These parameters include shape control parameters that adjust the overall tilt slope, hysteresis loop shape adjustment parameter one that controls the hysteresis loop expansion width, hysteresis loop shape adjustment parameter two that determines the smoothness of the hysteresis curve inflection point, and loop smoothness control index that changes the order of the hysteresis loop corner transition sharpness. These static control parameters, along with the current hysteresis intermediate variable values ​​and the time derivative of the compensation command current, are input into the differential evolution model.

[0089] The time derivative of the hysteresis intermediate variable is obtained by performing differential evolution equation calculations. Numerical integration is then performed on this time derivative to obtain the updated value of the hysteresis intermediate variable required for the next control cycle. Through a recursive differential evolution mechanism, the control system tracks the changes in the state of the nonlinear magnetic domains within the magnetostrictive material in real time, ensuring the accurate output of the compensation command current. The specific calculation formula for the differential evolution of this hysteresis intermediate variable is as follows: ; in, The time derivative of the hysteresis intermediate variable represents the instantaneous rate at which the internal hysteresis state changes with the driving input, and its unit is 1 / 2 Å. ; For shape control parameters, it represents a system constant that indicates the magnitude of the overall tilt slope of the hysteresis loop on the coordinate axis, in units of... ; The time derivative of the command current is used to compensate for the transient rate of change of the driving command current, and its unit is A / s. This is parameter one for adjusting the hysteresis loop shape. It represents the system constant that controls the width of the hysteresis loop expansion, and its unit is 1. ; As a hysteresis intermediate variable, it represents a mathematical intermediate quantity characterizing the unmeasurable hysteresis trajectory state inside the system; This is parameter two for hysteresis loop shape adjustment. It represents the system constant that, together with parameter one for hysteresis loop shape adjustment, determines the smoothness of the hysteresis curve inflection point, and its unit is 1. ; The loop smoothness control index represents the value that changes the order of the transition sharpness of the corners of the hysteresis loop, and its value ranges from 1 to 3.

[0090] See attached document Figure 6 Step S400 involves the electromechanical composite coordination of the main servo base motion and the micro-compensation current. This is executed through a rigid-flexible decoupling execution module.

[0091] Step S401: Time advance feedforward control is initiated. The updated compensation command current sequence, along with the real-time test inductor and real-time test resistor, are obtained from the previous step. A dynamic response model of the excitation coil is established, and the time lag characteristic of the voltage drive signal converting into a stable coil current is analyzed. This lag characteristic is mainly determined by the coil's electrical time constant and the communication delay of the hardware system. Without compensation, it will cause a phase misalignment between the actual output displacement and the cutting disturbance on the time axis.

[0092] Based on the extracted real-time test inductance and resistance, the current electrical lag time constant of the electromagnetic coil is calculated. This electrical lag time constant is then summed with the system's inherent digital communication delay and the power amplifier's switching delay to obtain the total lead time. This total lead time represents the absolute time interval required from the controller issuing the command to the establishment of the target current inside the coil. The specific formula for calculating this total lead time is as follows: ; in, The total lead time is the absolute time interval required from the controller issuing the command to the establishment of the target current inside the coil, and is measured in seconds. To test the inductance in real time, it represents the real-time magnetic field energy storage parameter of the coil under changes in thermomagnetic environment, and the unit is H; For real-time resistance testing, it represents the ohmic resistance value of the coil under real-time heating conditions during processing, with units of Ω. ;

[0093] The inherent delay time of the system represents the constant time consumed by the bus communication and power electronic switch operation of the power amplifier module in the CNC system, and is measured in seconds.

[0094] After obtaining the total lead time, the real-time spindle speed is read, and the total lead time is converted into the corresponding lead spindle angle. During spindle rotation, a spindle angle window for triggering compensation is set. When the real-time spindle phase fed back by the absolute encoder reaches a specific position (the boundary of this angle window minus the lead spindle angle), the controller sends the compensation command current to the power drive module in advance. The power drive module outputs the corresponding drive voltage, causing the coil current to accurately reach the target peak value of the compensation command current at the moment the actual cutting teeth contact the workpiece.

[0095] In step S402, the flexible metal housing bears the macroscopic cutting force. During micro-compensation, the machine tool's main servo system continuously maintains the basic feed motion. The feed servo motor receives macroscopic position commands from the CNC system, drives the ball screw to rotate via a coupling, and propels the slide carrying the spindle or workpiece forward at a constant speed along the guide rail. The basic feed motion is responsible for achieving the set axial cutting depth, completing the removal of the main metal material.

[0096] Face milling cutters generate significant macroscopic quasi-static cutting loads when cutting workpiece material. This load reacts axially on the feed drive train. The system employs a parallel load-bearing structure design, using an external flexible metal shell as the primary load-bearing component. This flexible metal shell possesses high axial static stiffness, enabling it to directly transmit macroscopic cutting resistance to the machine tool bed's base casting, forming a closed mechanical transmission path.

[0097] Through the bypassing and load-sharing effect of the flexible metal shell, the internally encapsulated magnetostrictive material is effectively isolated from the main transmission chain of macroscopic cutting forces. This structural isolation prevents the internal precision conversion material from being subjected to overload compressive stress. It avoids excessive stress that could cause irreversible movement or stress demagnetization of the magnetic domains within the magnetostrictive material, thus ensuring the linearity and stability of its electromechanical conversion characteristics and providing a mechanical environment guarantee for the accurate execution of micro-compensation.

[0098] Step S403 involves the superposition of magnetostrictive micro-displacement. After receiving the pre-injected compensation command current, the excitation coil generates a high-frequency alternating magnetic field within the internal space. The magnetostrictive material at the center of the alternating magnetic field responds to the change in magnetic field strength, producing microscale magnetostrictive deformation along its axis. This deformation converts electromagnetic energy into linear mechanical expansion displacement, becoming the active power source to offset high-frequency tool deflection errors.

[0099] The elongation force of the magnetostrictive material acts on a pre-designed connection surface inside the flexible metal shell. The flexible metal shell is designed with a localized flexible hinge structure with specific stiffness. The expansion force overcomes the minute elastic yield limit of this flexible hinge structure, forcing the front end face of the flexible metal shell to produce a controlled, reverse, minute linear displacement. This minute linear displacement is precisely synchronized with the compensation command current in the time domain and covers the high-frequency component range of the excitation force in the frequency domain.

[0100] The resulting reverse micro-linear displacement acts directly on the cutting point of the tool. In the actual spatial coordinate system, this reverse micro-linear displacement and the tool retraction error displacement caused by the high-frequency dynamic excitation component are superimposed and merged in opposite phases. Through the algebraic addition of spatial dimensions, the actively generated compensation stroke exactly cancels out the retraction deformation caused by the cutting thrust, achieving structural cancellation of errors in situ, thereby suppressing the generation of high-frequency ripples on the machined surface.

[0101] Specific application examples: The spindle synchronization observation module establishes a communication connection with the machine tool control system and acquires preset machining parameters. It sets the total number of effective teeth on the face milling cutter. The value is 6, which is the specific cutting force coefficient. The value is 1500N per square millimeter, and the axial cutting depth is... The value is 2.5mm. The spindle synchronous observation module, combined with the spindle rotation phase feedback from the absolute encoder, calculates the instantaneous cutting thickness of each cutter tooth in real time, and then calculates the instantaneous total cutting force. The time series was analyzed, and the high-frequency dynamic excitation component with a peak value of 240N was separated by a high-pass digital filter. .

[0102] The impedance sensing and correction module injects a test signal into the excitation coil during the machine tool's cold shutdown phase. The impedance sensing and correction module then extracts the reference resistance of the coil. The value is 12.5 Reference inductor The value is 0.045H, and the cold-state reference coupling gain factor is recorded. The value is 0.005 mm per ampere. After entering the formal cutting state, the impedance sensing correction module extracts the electrical characteristics under heating conditions in real time through a small sweep frequency sensing voltage signal to obtain the real-time test resistance. The value rose to 15.2 Real-time inductance testing The value dropped to 0.041H. This is in conjunction with the system's preset resistive coupling correction factor. Value 0.015 With inductive coupling correction factor Value 0.8H -1 The impedance sensing correction module calculates and generates an updated real-time coupling gain factor. The value is 0.0046 mm per ampere, which compensates for the displacement attenuation caused by thermomagnetic drift.

[0103] The hysteresis solution feedforward module receives high-frequency dynamic excitation components from the data bus. And extract the equivalent structural stiffness of the feed shaft. The value is 120,000 N per millimeter. The hysteresis solution feedforward module will convert the high-frequency dynamic excitation components... Divide by the equivalent structural stiffness of the feed axis After phase inversion, a target micro-compensation displacement with an amplitude of approximately -0.002 mm is generated. The hysteresis calculation feedforward module is based on real-time test inductance. With real-time test resistor Calculate the dynamic electrical time constant, taking into account the inherent delay time of the digital control system. The value is 0.002s, and the total advance time is calculated. The value is 0.0047s. The hysteresis solution feedforward module will calculate the total advance time. The calculated spindle advance angle is converted, and the calculated compensation command current is output before the cutting tooth cutting time node.

[0104] The rigid-flexible decoupling execution module receives a compensation command current calculated through time advance translation and hysteresis inverse solution. This current then excites a high-frequency alternating magnetic field within a sealed cavity, driving the flexible metal diaphragm shell to produce a minute linear displacement. (See attached diagram.) Figure 7 , Figure 7 The horizontal axis represents the running time in milliseconds (ms), and the vertical axis represents the microscopic displacement in millimeters (mm) with values ​​ranging from the order of 10 to the power of -3. This figure illustrates the microscopic variation trajectories of the three displacement curves under different parameter conditions. The target micro-compensated displacement curve is represented by a solid line marked with a square. The actual displacement curve without feedforward calculation is represented by a dashed line marked with a cross, which, compared to the solid line, exhibits a significant time lag shift to the right and a decrease in amplitude with a reduced vertical peak value. The actual executed displacement curve using the method of this invention is represented by a dotted line marked with a circle. This dotted line highly overlaps with the solid line marked with a square in both time phase and peak amplitude, verifying that the feedforward compensation command current accurately overcomes the physical barriers of hysteresis nonlinearity and communication delay.

[0105] Experimental verification and effect comparison section: Under the same spindle speed and feed rate boundary conditions, a high-precision surface roughness profilometer was used to perform contact path scanning measurements on the surfaces of two identical metal workpieces machined using different control strategies. The measurement path extends linearly along the workpiece's basic feed direction, acquiring and outputting a continuous sequence of surface micro-undulation heights within the evaluation length range.

[0106] The first set of comparative experiments disconnected the compensation drive link of the rigid-flexible decoupling execution module. The machine tool relied solely on the main servo system to maintain macroscopic linear feed motion and withstand metal removal resistance. Under the continuous action of broadband cutting disturbance, the tool system generated high-frequency forced elastic deformation, and the machining process was accompanied by high-frequency tool deflection. The second set of comparative experiments activated the spindle synchronous observation module, impedance sensing correction module, hysteresis calculation feedforward module, and rigid-flexible decoupling execution module throughout the process. The excitation coil drove the magnetostrictive material according to the calculated feedforward command current, and the generated controlled reverse micro-displacement physically canceled the high-frequency forced tool deflection displacement at the cutting point in situ.

[0107] See attached document Figure 8 , Figure 8The horizontal axis represents the scanning measurement length in mm, and the vertical axis represents the contour undulation height in μm, illustrating the spatial distribution of the microscopic contour height of the workpiece surface obtained from two independent machining processes. The surface contour curve without activated micro-compensation is represented by a solid line marked with triangles, exhibiting high-amplitude, high-density vibrations modulated by the periodic excitation of the cutting teeth, with the height difference between peaks and troughs in a relatively high range. The surface contour curve with activated micro-compensation is represented by a dashed line marked with a pentagram, where the high-frequency spikes corresponding to this dashed line are flattened by the magnetostrictive micro-displacement composite superposition execution mechanism, and the absolute drop between peaks and troughs converges significantly. The curve statistics reflect a step-like decrease in the core roughness characteristic parameters towards lower numerical values, objectively confirming the deterministic physical effect of the control system architecture and calculation method of this invention in suppressing cutting vibrations and eliminating surface ripples.

Claims

1. A high-precision feed control method for a milling machine, wherein the feed system of the milling machine includes a feed axis, and the actuator includes an excitation coil, a sealed chamber containing magnetostrictive material, an outer metal flexible diaphragm shell, and a worktable saddle, characterized in that, Includes the following steps: Obtain the geometric characteristic parameters and rotation phase data of the face milling cutter, calculate the discrete time sequence of the total cutting thrust load on the feed bearing, and separate the discrete time sequence into a quasi-static low-frequency load component and a high-frequency dynamic excitation component; A test signal is injected into the excitation coil to acquire the waveform and calculate the reference resistance and reference inductance, as well as the real-time test resistance and real-time test inductance. The ohmic difference and Henry difference are calculated to generate the thermomagnetic coupling gain factor. The high-frequency dynamic excitation component is divided by the equivalent structural stiffness of the feed axis to obtain the tool deformation amount, and the tool deformation amount is inverted and set as the target micro-compensation displacement. The initial compensation command current is calculated in combination with the thermomagnetic coupling gain factor under the preset discrete time step. The drive signal is generated by performing a forward translation operation based on the real-time test inductor and the real-time test resistor. The excitation coil receives the driving signal to cause the magnetic flux in the sealed cavity to alternate, polarizing the magnetostrictive material in the sealed cavity to generate endogenous stress, which forces the outer metal flexible diaphragm shell and the worktable saddle to generate high-frequency micro-displacements. The high-frequency micro-displacements and the tool deformation are superimposed and canceled in the same spatial coordinate system, realizing the macro-micro composite motion of the feed axis.

2. The high-precision feed control method for a milling machine according to claim 1, characterized in that, The specific steps for calculating the discrete-time sequence of the total cutting thrust load on the feed bearing are as follows: Calculate the specific spatial coordinates of each tooth of the face milling cutter on the surface of the workpiece material at any sampling time, and determine whether each tooth is currently in the effective working range of cutting into the workpiece; Using the current spindle rotational angular velocity data and the feed axis linear feed rate data, the relative motion trajectory of the cutting tooth at the cutting point position is calculated, and the instantaneous cutting thickness of a single tooth modulated by the spindle phase angle is calculated. The instantaneous cutting thickness of a single tooth is continuously multiplied with the specific cutting force coefficient and the axial cutting depth to obtain the local cutting thrust generated by a single cutting tooth when it cuts into the workpiece. The local cutting thrust generated by all cutting teeth in the cutting state is discretely summed and accumulated at the current time node to generate a discrete time sequence of the total cutting thrust load on the feed bearing.

3. The high-precision feed control method for a milling machine according to claim 2, characterized in that, The specific steps for separating the discrete time series into quasi-static low-frequency load components and high-frequency dynamic excitation components are as follows: Extract the discrete time series of the total cutting thrust load, perform spectral analysis on the frequency components of the discrete time series, and set the frequency segmentation boundary parameters of the digital filter bank based on the analysis results and the pre-calibrated natural frequency of the feed shaft drive train. The discrete-time series of the total cutting thrust load is simultaneously imported into the digital filter bank, which includes a low-pass digital filter and a high-pass digital filter. Discrete convolution filtering is performed in the low-pass digital filter operation link to filter out high-frequency periodic fluctuation signal components and output the quasi-static low-frequency load component that causes the overall force deformation of the transmission chain system. In the high-pass digital filter operation link, stopband attenuation operation is performed to eliminate DC bias components and slow evolution trend terms, and the high-frequency dynamic excitation component that causes tool forced resonance is extracted by rapidly fluctuating load components.

4. The high-precision feed control method for a milling machine according to claim 1, characterized in that, The specific steps for calculating the reference resistance and reference inductance are as follows: During the cold operation phase before the system is cut, a weak AC voltage test signal with constant amplitude and frequency is injected into both ends of the excitation coil. The amplitude of the weak AC voltage test signal is set so that the excited alternating electromagnetic thrust cannot overcome the static friction dead zone of the mechanical structure. The current response waveform sequence generated in the synchronous acquisition loop is used to extract the amplitude attenuation rate and phase delay angle between the voltage input sequence and the current output sequence. Based on the amplitude attenuation rate and phase delay angle, the reference complex impedance characteristic parameters of the excitation coil in a cold environment are calculated, the reference resistor and the reference inductor are separated, and the reference resistor and the reference inductor are saved to a non-volatile memory.

5. A high-precision feed control method for a milling machine according to claim 4, characterized in that, The specific steps for calculating the real-time test resistance and real-time test inductance are as follows: After entering the formal cutting state, the online monitoring and extraction operation is started. According to a fixed sampling period, the frequency sweep sensor voltage signal of the high frequency sweep mode is superimposed on the existing drive control current command. Real-time synchronous acquisition of time series data of voltage and current observations returned from both ends of the excitation coil, including superimposed components; A digital bandpass filter is used to remove low-frequency main drive signals and high-frequency noise interference caused by cutting loads. Discrete Fourier transform is performed on the retained microsensor response frequency band data to calculate the real-time test resistance and real-time test inductance of the excitation coil as it changes with heating and magnetic field evolution.

6. The high-precision feed control method for a milling machine according to claim 5, characterized in that, The specific steps for calculating the ohmic difference and Henry difference to generate the thermomagnetic coupling gain factor are as follows: Extract the real-time test resistor and the real-time test inductor calculated in the current sampling period, and retrieve the reference resistor and the reference inductor stored in the non-volatile memory; The ohmic difference between the real-time test resistor and the reference resistor, and the Henry difference between the real-time test inductor and the reference inductor are calculated to characterize the coil heating temperature rise and the severity of magnetic permeability drift of the magnetic material. Obtain the preset reference coupling gain factor, resistive coupling correction coefficient, and inductive coupling correction coefficient, and substitute them, along with the ohmic difference and Henry difference, into the gain update model for calculation. Compensate for thermodynamic displacement decay and hysteresis performance decay by superimposing linear decay weights to generate the updated thermomagnetic coupling gain factor.

7. The high-precision feed control method for a milling machine according to claim 1, characterized in that, The specific steps for dividing the high-frequency dynamic excitation component by the equivalent structural stiffness of the feed axis to obtain the tool deflection deformation, and then inverting the tool deflection deformation and setting it as the target micro-compensation displacement are as follows: Extract the equivalent structural stiffness of the feed axis in the current machine tool transmission chain, and establish a mapping relationship model between the high-frequency dynamic excitation component and the elastic deformation of the structure; The high-frequency dynamic excitation component sequence is input into the mapping relationship model to continuously calculate the micro-scale displacement deviation of the feed axis under the current force state. The tool deflection deformation is obtained by dividing the high-frequency dynamic excitation component by the equivalent structural stiffness of the feed axis. Then, the target micro-compensation displacement with the opposite phase is generated by the inversion operation, which is used to require the active generation of reverse stroke compensation to offset the tool retraction caused by the cutting thrust.

8. A high-precision feed control method for a milling machine according to claim 7, characterized in that, The specific steps for calculating the initial compensation command current using the aforementioned thermomagnetic coupling gain factor at a preset discrete time step are as follows: A hysteresis inverse solution model based on differential equations is constructed, which decomposes the target micro-compensation displacement into a linearly reversible stiffness response part and a nonlinear, unmeasurable hysteresis hysteresis part. Extract the pre-calibrated linear stiffness ratio coefficient and hysteresis displacement weight coefficient, and use the linear stiffness ratio coefficient and the hysteresis displacement weight coefficient as the core weight parameters for inverse model solution; Substituting the target micro-compensation displacement into the nonlinear hysteresis displacement mapping relationship, and combining it with the received thermomagnetic coupling gain factor, the differential state evolution operation is performed by combining the current hysteresis intermediate variable state, and the high-frequency dynamic asymmetric pulse current used to offset the material's inherent hysteresis response error is solved in reverse as the initial compensation command current.

9. A high-precision feed control method for a milling machine according to claim 8, characterized in that, The specific steps for generating a drive signal by performing a lead shift operation based on the real-time test inductor and the real-time test resistor are as follows: The electrical lag time constant under the current operating condition is calculated by dividing the obtained real-time test inductance by the real-time test resistance. The total advance time is obtained by summing the electrical lag time constant with the inherent digital communication delay of the system and the switching delay of the power amplifier. The total advance time is converted into the corresponding advance spindle angle. The time node of the next cutter tooth entering the workpiece is predicted by combining the rotation phase window of the spindle absolute encoder. The initial compensation command current is then shifted ahead on the time axis to generate the final drive signal.

10. A high-precision feed control method for a milling machine according to claim 9, characterized in that, The specific steps for forcing the outer flexible metal diaphragm shell and the worktable saddle to generate high-frequency micro-displacements, and for these high-frequency micro-displacements to superimpose and cancel out the tool deformation in the same spatial coordinate system, are as follows: The magnetostrictive material in the sealed cavity undergoes instantaneous expansion and contraction of its spatial volume under the polarization effect of the high-frequency alternating magnetic field generated by the excitation coil, which is transformed into the outward expansion of the endogenous stress. The intrinsic stress forcibly overcomes the axial local elastic yield limit of the flexible metal diaphragm shell, forcing the worktable saddle to generate a controlled reverse high-frequency micro-displacement along the axial direction. The high-frequency micro-displacement and the tool deformation are superimposed and fused in the actual spatial coordinate system in the same frequency but opposite phase, and the in-situ structural cancellation is completed by algebraically adding the spatial dimensions.

Images