A method and system for dynamic compensation control of mold processing errors

By acquiring the real-time tracking error sequence and decoupling high-frequency fluctuations and low-frequency deviations, combined with the dynamic stiffness feedforward image model and state switching logic, the problem of friction time-varying accumulation effect of multi-axis linkage platform under complex working conditions is solved, and high-precision dynamic compensation control of mold processing error is realized.

CN123113676APending Publication Date: 2026-11-17SHANDONG SHISHENG MASCH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202611448813.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies have limitations in handling multi-axis heterogeneous dynamic constraints, continuous cutting load fluctuations, and time-varying accumulation effects of transmission reverse nonlinear friction. This leads to the control system outputting compensation commands that violate the actual friction accumulation law under complex working conditions, making it difficult to achieve deep perception of the time-varying friction state across the entire trajectory and adaptive control of time-series disturbances.

Method used

By acquiring the real-time tracking error sequence of each axis position loop, using a discrete time-domain filter to decouple and extract high-frequency fluctuations and low-frequency deviation components, mapping and calculating transient position compensation, and forcing the weight to repeat to zero when the servo feed axis changes direction, using a dynamic stiffness feedforward mapping model for nonlinear mapping, and combining state switching logic and gating limit module, dynamic compensation control is achieved.

Benefits of technology

It effectively reduces trajectory tracking errors, lowers the risk of overcutting and undercutting at mold corners, improves the control stability and surface machining accuracy of multi-axis linkage machining, and ensures the smoothness and accuracy of the position control process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113676A_ABST
    Figure CN123113676A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of CNC machine tool control technology, and relates to a dynamic compensation control method and system for mold processing errors. The method includes: obtaining real-time tracking error sequences of each axis position loop using the interpolation cycle of the control system as a scale; processing the sequences using a discrete filter to extract high-frequency error fluctuations and low-frequency deviation trend components; inputting these components into a dynamic stiffness feedforward image model to calculate the transient spatial position compensation amount for the current cycle and writing it into a register; and adjusting the weight of the high-frequency components to zero for 15ms using a gated cutoff unit when the feed axis speed crosses the 0mm / s critical point, and calling the steady-state constant to correct the transient spatial position compensation amount. This invention eliminates the phase blind zone between command issuance and transient physical deformation caused by group delay, suppresses loop resonance during the load abrupt transition period in the loop gain resting state, and avoids surface overcutting defects caused by reverse overshoot.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of CNC machine tool control technology, and relates to a dynamic compensation control method and system for mold processing errors. Background Technology

[0002] Currently, trajectory tracking control of high-precision CNC machine tools and multi-axis linkage platforms has become a core path for precision manufacturing. However, existing technologies still have limitations in dealing with multi-axis heterogeneous dynamic constraints, continuous cutting load fluctuations, and the time-varying cumulative effects of transmission reverse nonlinear friction.

[0003] First, there is the feature point reverse friction compensation technology based on local spatiotemporal sampling. For example, Chinese invention patent CN107918356B discloses a method for cross-quadrant friction compensation in the axis linkage machining process of CNC machine tools. By discretely sampling the linkage machining data, the linkage error is converted into a single-axis following error sequence, the quadrant point waiting compensation area is determined, and the least squares method is used to fit discrete points to construct a segmented displacement, velocity, and acceleration compensation sequence. Its drawback is that this scheme is essentially an offline static fitting compensation based on local feature points. The decision logic lacks a continuous adaptive representation of the time-varying characteristics of the entire machining trajectory and the continuous variable curvature motion state on the time axis. It does not map the time-varying transient disturbances caused by dynamic cutting loads into dynamic boundary constraints of the global control space, which may cause the control system to output compensation commands that violate the actual friction accumulation law when facing complex, non-sinusoidal trajectory machining.

[0004] Second, online iterative control technology based on dynamic disturbance observation and state prediction; for example, Chinese invention patent CN116203895B discloses an online iterative compensation system and method for servo error, which treats nonlinearity and external disturbances in the processing process as lumped disturbances for observation and compensation, uses a dual-loop controller to suppress the light damping resonance of the flexible mechanism, and predicts the reference trajectory at future moments based on the linear model of the difference equation of the discrete system and multi-point historical trajectory interpolation to achieve online real-time iterative compensation of tracking error. Its shortcomings are: this technology focuses on a specific multi-level hardware architecture and open-loop state prediction under the master-slave control strategy. Although it improves the tracking performance in a specific frequency band, it fails to make a deep closed-loop correlation between high-order nonlinear dead zone friction and the heterogeneity of the dynamics of the multi-axis transmission chain drive source. Moreover, the polynomial interpolation prediction logic of the future trajectory is prone to prediction divergence due to high-frequency measurement noise of the sensor or sudden hard boundary, making it difficult to cope with the risk of real-time boundary compression of the computing bandwidth of the control system caused by drastic changes in large loads under complex processing conditions.

[0005] Third, position control technology based on geometric space constraint planning and position contour integration; for example, Chinese invention patent application CN111427308A discloses an error compensation integrated control method for trajectory planning of CNC platform. It establishes a feed rate regulator based on chord height error constraint to perform polynomial speed planning, and combines a real-time contour error estimation model. It uses a cross-coupled controller and an improved position error compensator to simultaneously calculate and compensate for the contour error and position error of each axis. Its shortcomings are: existing integrated control models of this kind usually regard the dynamic unit of each feed axis as an ideally matched symmetrical or idealized decision unit, ignoring the asymmetric continuous accumulation effect of transmission chain assembly error and uneven load distribution on long-term spatial motion; in long-cycle optimization or high-speed large curvature trajectory cutting, if there is a lack of correction and anti-disturbance forced convergence mechanism for single-axis low-speed reverse nonlinear dead zone friction and insufficient dynamic cutting stiffness, the predicted trajectory and the actual compensated cutting position are prone to time-series cumulative drift, causing the output control scheme to lose its practical value due to exceeding the stiffness limit and physical safety response bottom line of the servo drive system itself.

[0006] Therefore, the technical problem to be solved by this invention is how to construct an error compensation architecture with deep perception of time-varying friction states and adaptive control of time-series disturbances throughout the entire machining cycle, so that the control logic of the multi-axis linkage platform can converge at high speed to the real dynamic physical safety boundary. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and solve the technical problems of the limitations of the prior art in dealing with the time-varying accumulation effect of multi-axis heterogeneous dynamic constraints, continuous cutting load fluctuations and transmission reverse nonlinear friction, which leads to the dissipation of feedforward command timing or overtuning resonance. The invention provides a dynamic compensation control method and system for mold processing errors.

[0008] To achieve the above-mentioned objective, the present invention provides a dynamic compensation and control method for mold processing errors, comprising the following steps: Step S1, obtain the real-time tracking error sequence of each axis position loop: The control unit uses the current interpolation cycle of the CNC system as the reference time scale to synchronously obtain the real-time tracking error sequence of the position control loop of each servo feed axis; Step S2, filtering and decoupling to extract high-frequency fluctuation and low-frequency deviation components: The control unit uses a preset discrete time domain filter to process the real-time tracking error sequence of the position control loop, and decouples and extracts the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. Step S3, Mapping and Calculating Transient Position Compensation and Writing it to the Register: The control unit inputs the high-frequency error fluctuation component and the low-frequency deviation trend component into the dynamic stiffness feedforward mapping model constructed based on the control loop gain and the transmission chain stiffness coefficient. Through the nonlinear mapping of the high-frequency error fluctuation component and the low-frequency deviation trend component to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the index for the current interpolation period. This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. The quantitative relationship is as follows In the formula The cutoff frequency of the servo driver speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scale coefficients. The range of values ​​is ; Step S4, Force weight to repeat to zero and correct compensation amount when commutation crosses zero: When the axial movement direction of each servo feed axis changes and the current speed of the feed axis crosses the 0mm / s critical point, the control unit triggers the commutation gate cutoff unit and continues for 15ms. It outputs a gate cutoff signal to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and calls the stored historical steady-state stiffness constant to correct the transient spatial position compensation amount.

[0009] Step S3 of the present invention includes the following steps: Step S31, the control unit calculates the amplitude gradient change rate of the transient spatial position compensation; Step S32, when the amplitude gradient change rate is less than the set negative release threshold, the control unit switches to the smooth reduction state of the control quantity using state switching logic, the control unit closes the mapping path of the dynamic stiffness feedforward mapping model, and calls the discrete gradient decay calculation module to calculate the decreasing step adjustment coefficient based on the nonlinear residual in the current interpolation period. In the formula The nonlinear residual for the current interpolation period. Let be the system damping coefficient, such that the step adjustment coefficient multiplied by the transient spatial position compensation of the previous interpolation cycle satisfies ... The value of the position instruction register is overwritten to truncate the transient spatial position compensation.

[0010] Step S2 of the present invention includes the following steps: Step S21, the control unit configures a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module; Step S22, the control unit inputs the real-time tracking error sequence of the position control loop into the high-order low-pass filter matrix to output the low-frequency deviation trend component. ,in To track the error sequence in real time, To satisfy the preset filter weight coefficients for normalization, and to capture high-frequency disturbances using a high-pass differential module, according to... To output high-frequency error fluctuation components.

[0011] Step S4 of the present invention includes the following steps: Step S41, the control unit completes the construction of the forward prediction window to capture the velocity acceleration vector of each servo feed axis; Step S42, when the velocity acceleration vector indicates that the velocity of each servo feed axis passes through the 0 mm / s critical point, the commutation gating cutoff unit adjusts the weight of the high-frequency error fluctuation component to zero, blocks the transmission of commutation impact noise to the transient spatial position compensation amount, and during the gating period, uses the historical steady-state stiffness constant. Write the dynamic stiffness feedforward image model to reset the servo stiffness reference, and output the transient spatial position compensation amount consisting of the servo hysteresis correction vector. .

[0012] After calculating the transient spatial position compensation amount using the dynamic stiffness feedforward imaging model, the present invention further includes the following steps: Step S5, the control unit completes the mounting of the collaborative adjustment architecture consisting of the forward prediction window and the update loop; Step S6, the control unit increases the cutoff frequency of the discrete time domain filter by a preset proportional step size when the tracking error envelope variance is greater than a preset variance threshold, and decreases the cutoff frequency when it is less than a preset variance threshold, based on the tracking error envelope variance of five consecutive interpolation cycles, so as to calculate the transient spatial position compensation amount based on the high-frequency error fluctuation components.

[0013] The present invention further includes the following steps: Step S7, the control unit uses state switching logic to monitor and control the state of the interpolation buffer; Step S8, when the surface curvature in the interpolation buffer undergoes a step change and the rate of change of curvature exceeds a set abrupt change threshold, the state switching logic switches the control loop from the normal dynamic compensation mode to the edge condition energy level reduction mode, and multiplies the output weight of the transient spatial position compensation by a preset derating coefficient γ for reduction, wherein the value of γ ranges from 0.3 to 0.6.

[0014] The present invention further includes the following steps: Step S9, the control unit obtains the grating ruler feedback pulse of the servo driver to calculate the actual motion trajectory of each servo feed axis; Step S10, the control unit obtains the original interpolation position command, calculates the difference between the actual motion trajectory and the original interpolation position command to generate closed-loop steady-state accuracy verification data, and writes the closed-loop steady-state accuracy verification data into the non-volatile memory during the rest period of the control system to update the control reference.

[0015] Step S3 of the present invention includes the following steps: Step S33, the control unit configures a nonlinear stiffness mapping table in the dynamic stiffness feedforward image model; Step S34, the control unit retrieves the nonlinear stiffness mapping table based on the amplitude of the low-frequency deviation trend component to determine the servo stiffness coefficient corresponding to the current variable curvature load. And through the formula: Calculate the transient elastic offset vector of the tool tip In the formula For position loop control gain, The pulse equivalent scaling factor; the control unit calculates the current feed rate of the servo feed axis. Proportional gain of position control loop The corresponding feedforward time constant, according to the formula Calculate the servo hysteresis correction vector .

[0016] The present invention further includes the following steps: Step S11, the control unit synchronously acquires the decoupling components of each servo feed axis; Step S12, the control unit calculates the contour error vector between two adjacent servo feed axes, and injects the contour error vector as a cross-coupling constraint into the generation loop of the transient spatial position compensation quantity to offset the phase deviation between each servo feed axis.

[0017] This invention also provides a dynamic compensation control system for mold processing errors, which is used to implement a dynamic compensation control method for mold processing errors, including: The error acquisition module is used to synchronously acquire the real-time tracking error sequence of the position control loop of each servo feed axis using the current interpolation cycle of the CNC system as the reference time scale. The data processing module is connected to the error acquisition module and is used to process the real-time tracking error sequence of the position control loop using a preset discrete time domain filter, and decouple and extract the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. The feedforward calculation module, connected to the data processing module, is used to input high-frequency error fluctuation components and low-frequency deviation trend components into a dynamic stiffness feedforward mapping model constructed based on the control loop gain and transmission chain stiffness coefficient. Through the nonlinear mapping of high-frequency error fluctuation components and low-frequency deviation trend components to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. In the formula The cutoff frequency of the speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scaling coefficients; The gating limit module, connected to the feedforward calculation module, is used to trigger the reversing gating cutoff unit for 15ms when the axial movement direction of each servo feed axis changes and the current speed of the feed axis crosses the 0mm / s critical point. The gating cutoff signal is output to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and the stored historical steady-state stiffness constant is called to correct the transient spatial position compensation amount.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: 1. In the dynamic compensation control of mold processing error, by judging the changing trend of high-frequency error fluctuations and writing the position compensation command in advance when the cutting load increases rapidly, the compensation action can be intervened earlier, reducing the time delay caused by signal processing, register writing and motor response in the control system. As a result, the position compensation can better follow the transient elastic deformation of the mold's variable curvature surface, reduce the phase deviation between the compensation command and the actual deformation, and reduce trajectory tracking error.

[0019] 2. When the cutting load decreases rapidly, by closing the image path of the dynamic stiffness feedforward image model and gradually reducing the transient spatial position compensation, the residual compensation can be withdrawn in a timely and smooth manner, avoiding the compensation command from continuing to act after the load is released. This process can reduce reverse pulses and compensation overshoot, reduce the risk of machining defects such as overcutting and undercutting at the corners of the mold surface, and maintain the stability of the position control process.

[0020] 3. When the servo feed axis reverses direction and passes through the zero-speed position, the high-frequency error fluctuation component is temporarily stopped from participating in the compensation calculation by the reversal gating cutoff unit. The transient spatial position compensation amount is corrected by using the historical steady-state stiffness constant. This can avoid the high-frequency noise generated by reversal impact and transmission backlash being mistaken for changes in cutting load. As a result, the position correction during the reversal process is more stable, which can suppress the sudden change of transient spatial position compensation amount and the divergence of dynamic stiffness feedforward image model calculation, and improve the control stability and surface machining accuracy during multi-axis linkage machining. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the tracking error decoupling and compensation amount generation process of the present invention; Figure 2 This is a schematic diagram of the feedforward mapping and state gating logic module structure of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings.

[0023] Example 1: This embodiment discloses a dynamic compensation control method for mold processing errors, which runs within a control system and includes the following steps: Step S1, obtain the real-time tracking error sequence of each axis position loop: The control unit uses the current interpolation cycle of the CNC system as the reference time scale to synchronously obtain the real-time tracking error sequence of the position control loop of each servo feed axis; Step S2, filtering and decoupling to extract high-frequency fluctuation and low-frequency deviation components: The control unit uses a preset discrete time domain filter to process the real-time tracking error sequence of the position control loop, and decouples and extracts the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. Step S3, Mapping and Calculating Transient Position Compensation and Writing it to the Register: The control unit inputs the high-frequency error fluctuation component and the low-frequency deviation trend component into the dynamic stiffness feedforward mapping model constructed based on the control loop gain and the transmission chain stiffness coefficient. Through the nonlinear mapping of the high-frequency error fluctuation component and the low-frequency deviation trend component to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the index for the current interpolation period. This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. The quantitative relationship is as follows In the formula The cutoff frequency of the servo driver speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scale coefficients. The range of values ​​is ; Step S4, Force weight to repeat to zero and correct compensation amount when commutation crosses zero: When the axial movement direction of each servo feed axis changes and the current speed of the feed axis crosses the 0mm / s critical point, the control unit triggers the commutation gate cutoff unit and continues for 15ms. It outputs a gate cutoff signal to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and calls the stored historical steady-state stiffness constant to correct the transient spatial position compensation amount.

[0024] Step S3 in this embodiment includes the following steps: Step S31, the control unit calculates the amplitude gradient change rate of the transient spatial position compensation; Step S32, when the amplitude gradient change rate is less than the set negative release threshold, the control unit switches to the smooth reduction state of the control quantity using state switching logic, closes the mapping path of the dynamic stiffness feedforward mapping model, and calls the discrete gradient decay calculation module to calculate the decreasing step adjustment coefficient based on the nonlinear residual in the current interpolation period. In the formula The nonlinear residual for the current interpolation period. Let be the system damping coefficient, such that the step adjustment coefficient multiplied by the transient spatial position compensation of the previous interpolation cycle satisfies ... The value of the position instruction register is overwritten to truncate the transient spatial position compensation.

[0025] Step S2 in this embodiment includes the following steps: Step S21, the control unit configures a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module; Step S22, the control unit inputs the real-time tracking error sequence of the position control loop into the high-order low-pass filter matrix to output the low-frequency deviation trend component. ,in To track the error sequence in real time, To satisfy the preset filter weight coefficients for normalization, and to capture high-frequency disturbances using a high-pass differential module, according to... To output high-frequency error fluctuation components.

[0026] Step S4 in this embodiment includes the following steps: Step S41, the control unit completes the construction of the forward prediction window to capture the velocity acceleration vector of each servo feed axis; Step S42, when the velocity acceleration vector indicates that the velocity of each servo feed axis passes through the 0 mm / s critical point, the commutation gating cutoff unit adjusts the weight of the high-frequency error fluctuation component to zero, blocks the transmission of commutation impact noise to the transient spatial position compensation amount, and during the gating period, uses the historical steady-state stiffness constant. Write the dynamic stiffness feedforward image model to reset the servo stiffness reference, and output the transient spatial position compensation amount consisting of the servo hysteresis correction vector. .

[0027] After calculating the transient spatial position compensation amount using the dynamic stiffness feedforward image model, this embodiment further includes the following steps: Step S5, the control unit completes the mounting of the collaborative adjustment architecture consisting of the forward prediction window and the update loop; Step S6, the control unit increases the cutoff frequency of the discrete time domain filter by a preset proportional step size when the tracking error envelope variance is greater than a preset variance threshold, and decreases the cutoff frequency when it is less than a preset variance threshold, based on the tracking error envelope variance of five consecutive interpolation cycles, so as to calculate the transient spatial position compensation amount based on the high-frequency error fluctuation components.

[0028] The further steps described in this embodiment are as follows: Step S7, the control unit uses state switching logic to monitor and control the state of the interpolation buffer; Step S8, when the surface curvature in the interpolation buffer undergoes a step change and the rate of change of curvature exceeds the set abrupt change threshold, the state switching logic switches the control loop from the normal dynamic compensation mode to the edge condition energy level reduction mode, and multiplies the output weight of the transient spatial position compensation by a preset derating coefficient γ for reduction, wherein the value of γ ranges from 0.3 to 0.6.

[0029] The further steps described in this embodiment include: Step S9, the control unit obtains the grating ruler feedback pulse of the servo driver to calculate the actual motion trajectory of each servo feed axis; Step S10, the control unit obtains the original interpolation position command, calculates the difference between the actual motion trajectory and the original interpolation position command to generate closed-loop steady-state accuracy verification data, and writes the closed-loop steady-state accuracy verification data into the non-volatile memory during the rest period of the control system to update the control reference.

[0030] Step S3 in this embodiment includes the following steps: Step S33, the control unit configures a nonlinear stiffness mapping table in the dynamic stiffness feedforward image model; Step S34, the control unit retrieves the nonlinear stiffness mapping table based on the amplitude of the low-frequency deviation trend component to determine the servo stiffness coefficient corresponding to the current variable curvature load. And through the formula: Calculate the transient elastic offset vector of the tool tip In the formula For position loop control gain, The pulse equivalent scaling factor; the control unit calculates the current feed rate of the servo feed axis. Proportional gain of position control loop The corresponding feedforward time constant, according to the formula Calculate the servo hysteresis correction vector .

[0031] The further steps described in this embodiment include: Step S11, the control unit synchronously acquires the decoupling components of each servo feed axis; Step S12, the control unit calculates the contour error vector between two adjacent servo feed axes, and injects the contour error vector as a cross-coupling constraint into the generation loop of the transient spatial position compensation quantity to offset the phase deviation between each servo feed axis.

[0032] This embodiment also provides a dynamic compensation control system for mold processing errors, including: The error acquisition module is used to synchronously acquire the real-time tracking error sequence of the position control loop of each servo feed axis using the current interpolation cycle of the CNC system as the reference time scale. The data processing module is connected to the error acquisition module and is used to process the real-time tracking error sequence of the position control loop using a preset discrete time domain filter, and decouple and extract the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. The feedforward calculation module, connected to the data processing module, is used to input high-frequency error fluctuation components and low-frequency deviation trend components into a dynamic stiffness feedforward mapping model constructed based on the control loop gain and transmission chain stiffness coefficient. Through the nonlinear mapping of high-frequency error fluctuation components and low-frequency deviation trend components to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. In the formula The cutoff frequency of the speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scaling coefficients; The gating limit module, connected to the feedforward calculation module, is used to trigger the reversing gating cutoff unit for 15ms when the axial movement direction of each servo feed axis changes and the current speed of the feed axis crosses the 0mm / s critical point. The gating cutoff signal is output to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and the stored historical steady-state stiffness constant is called to correct the transient spatial position compensation amount.

[0033] Example 2: In this embodiment, under the condition of milling die-casting mold steel with a hardness of 48HRC on a high-speed multi-axis vertical machining center, each servo feed axis cuts into the variable curvature corner surface at a feed rate of 6000 mm / min. The tool system and mechanical transmission chain generate transient elastic offset under the action of nonlinear cutting load, resulting in contour deviation of the spatial trajectory. There is an inherent physical response delay of 15ms in the signal transformation, data register rewriting and motor electromagnetic response within the control loop. As a result, the feedforward compensation signal stream generates a group delay. The timing of the position compensation command register writing lags behind the node where the elastic offset occurs, forming a spatiotemporal phase misalignment, and causing reverse overshoot and overcut bulge defects at the corner position of the surface. The control unit uses the current interpolation cycle of the CNC system as the reference time scale to synchronously acquire the real-time tracking error sequence of the position control loop of each servo feed axis. Position control loop real-time tracking error sequence The discrete data stream generated by subtracting the motor encoder feedback position from the CNC system command position is used in the control unit, which is configured with a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module. Input a discrete-time filter; a high-order low-pass filter matrix performs matrix operations on the error values ​​of the current interpolation period and the two preceding interpolation periods according to preset filter coefficients, and outputs a low-frequency deviation trend component characterizing the low-frequency lag trend of the drivetrain. The high-pass differential module subtracts the current error value from the low-pass output value, captures high-frequency disturbances, and outputs a high-frequency error fluctuation component that characterizes the high-frequency fluctuations of the cutting load. .

[0034] The control unit also incorporates a collaborative adjustment architecture consisting of a forward prediction window and an update loop. The forward prediction window stores the tracking error envelope and the velocity and acceleration vectors of each servo feed axis according to the interpolation cycle. The tracking error envelope is formed by the absolute value of the real-time tracking error of each axis position control loop. The update loop calculates the variance of the tracking error envelope over five consecutive interpolation cycles, updates the cutoff frequency of the discrete-time filter based on this variance, and writes the updated cutoff frequency into the filter parameter register for extraction of high-frequency error fluctuation components in the next interpolation cycle. and low-frequency deviation trend components Then, the transient spatial position compensation is calculated; the control unit constructs a discrete time-domain sliding window containing three consecutive interpolation cycles, including the current interpolation cycle, and collects high-frequency error fluctuation components. The transient first derivative of the high-frequency error energy within this window is calculated to characterize the changing trend of the transient cutting load. Subsequently, the high-frequency error fluctuation components are... and low-frequency deviation trend components The input is a dynamic stiffness feedforward mapping model constructed based on the control loop gain and the transmission chain stiffness coefficient. The control loop gain is read from the position loop gain parameter of the servo driver, and the input is based on the cutoff frequency of the servo driver speed control loop. Proportional gain of position control loop Quantitative relationship The system's preset loop gain attenuation factor is obtained through conversion. ( The range of values ​​is , The frequency band matching scaling coefficients are then incorporated into the dynamic stiffness feedforward mapping model.

[0035] In the dynamic stiffness feedforward mapping model, the control unit is based on the low-frequency deviation trend component. The amplitude is retrieved from the nonlinear stiffness mapping table. This table stores the corresponding servo stiffness coefficients according to the amplitude range of the low-frequency deviation trend component. The control unit uses the servo stiffness coefficient corresponding to the current amplitude range as the denominator, and calculates the high-frequency error fluctuation component. and low-frequency deviation trend components Converting these values ​​to the spatial displacement compensation scale yields the transient elastic offset vector of the tool tip. and servo hysteresis correction vector ,in, and The units are all mm. The linear combination of the final output interfaces of the dynamic stiffness feedforward mapping model is: ,in, As a comprehensive position compensation vector, its mathematical properties are that of a three-dimensional vector. A dimensionless constant, the comprehensive position compensation vector. The transient spatial position compensation amount for the current interpolation cycle is constituted and written into the internal position instruction register.

[0036] The control unit calculates the rate of change of the amplitude gradient of the transient spatial position compensation based on the difference between the amplitude of the transient spatial position compensation in the current interpolation cycle and the amplitude of the transient spatial position compensation in the previous interpolation cycle, and the corresponding interpolation cycle duration. Its unit is mm / ms, and the state switching logic is based on... The threshold range in which the control path is selected is when The positive acceleration increases for two consecutive interpolation cycles and exceeds the preset positive acceleration threshold. At this time, the state switching logic switches to the accelerated deformation counterbalancing state, enables the register pointer pre-control flow, shifts the write address of the position command register forward by one interpolation cycle, so that the transient spatial position compensation is written into the position command register of the CNC system in advance, and uses the overload torque of the motor driver to build the forward counterbalancing force field in advance. This process advances the actual output time of the compensation command by one interpolation cycle, so that the deviation of the variable curvature trajectory tracking is transformed from in-phase synchronous blocking to advanced phase position counterbalancing; the nonlinear mapping of the dynamic stiffness feedforward mapping model is completed by the interval retrieval of the nonlinear stiffness mapping table and the threshold judgment of the state switching logic. The servo stiffness coefficient changes with the low-frequency deviation trend component. The amplitude range changes, and the state switching logic is based on the amplitude gradient change rate. Select the corresponding control path; Formula Only the transient elastic offset vector of the tool tip after spatial displacement scaling is considered. and servo hysteresis correction vector Vector combination is performed without changing the aforementioned nonlinear mapping relationship.

[0037] When the tool leaves the variable curvature corner and the cutting load drops sharply, the amplitude gradient change rate... Less than the preset negative release threshold At this time, the state switching logic switches to the state of smooth reduction of control quantity, executes damping control in the load release phase, closes the mapping path of the dynamic stiffness feedforward mapping model, and calls the discrete gradient decay calculation module. The nonlinear residual is determined by the difference between the real-time tracking error of the position control loop in the current interpolation cycle, converted by the spatial displacement compensation scale, and the transient spatial position compensation amount in the same cycle. The discrete gradient decay calculation module calculates the decreasing step adjustment coefficient based on the nonlinear residual in the current interpolation cycle, multiplies the step adjustment coefficient by the transient spatial position compensation amount of the previous interpolation cycle, and overwrites the position command register with the obtained value. In two consecutive interpolation cycles... During this period, the residual feedforward is forcibly truncated to zero, eliminating the dissipation and trailing lag of the feedforward command at locations where the cutting force drops sharply, thus reducing the risk of overcutting and undercutting at the corners of the profile. When the axial motion direction of each servo feed axis changes, the forward prediction window, according to the axis number, combines the speed and acceleration of the current interpolation cycle with the speed command of the next interpolation cycle to form a speed-acceleration vector. When this speed-acceleration vector indicates that the current speed of the feed axis is about to cross the 0 mm / s critical point, and the feed axis direction status register of the CNC system flips, the control unit triggers the reversing gating cutoff unit and maintains it for 15 ms, outputting a gating cutoff signal to suppress high-frequency error fluctuation components. The gain weight is adjusted to zero to block the transmission of commutation impulse noise to transient spatial position compensation.

[0038] During the gating cutoff period, the control unit retrieves the stored historical steady-state stiffness constants. Historical steady-state stiffness constant The milling stiffness coefficient of the machine tool, which is pre-calibrated and stored under steady-state milling conditions, is expressed in N / μm. The control unit will... The dynamic stiffness feedforward mapping model is written to reset the servo stiffness reference. After the weights of the high-frequency error fluctuation components are set to zero, the transient spatial position compensation amount composed of the servo hysteresis correction vector is output. The system obtains a smooth transition position correction. When the 15ms action time expires and the reversing impact noise frequency drops below the preset cutoff frequency, the reversing gating cutoff unit closes, and the system resumes adaptive phase counter-control. This decouples the nonlinear load change at the reversing position from the closed-loop adjustment channel of the position tracking control loop, suppressing the stiffness mapping calculation divergence caused by the impact noise generated by the transmission backlash and frequent reversing of the servo feed axis. The 15ms action time was determined through zero-crossing gap impact tests of the feed axis reversing at different cutting speeds on the test bench. When the action time is less than 10ms, the high-frequency mechanical collision noise at the moment of reversing has not completely decayed, and the residual impact noise will enter the dynamic stiffness feedforward mapping model and cause calculation divergence. When the action time is greater than 25ms, the weight zeroing time of the high-frequency error fluctuation component is too long, and the system's dynamic compensation capability for high-frequency load fluctuations at the corner position decreases. Setting the action time to 15ms can preserve the system's high-frequency response capability while blocking the reverse impact noise.

[0039] Under the alternating drive of the control flow in the above stages, the CNC system compensates for the transient spatial position. The original interpolation position command superimposed on the current interpolation cycle Generate correction position control commands The output is sent to the servo driver. Under the premise of keeping the gain parameters of each loop of the control loop in a safe resting state, the control process eliminates the spatiotemporal phase lag between identification and terminal action, reduces the overall machining surface contour error by more than 75%, eliminates the surface overcutting defect caused by reverse pulse overshoot, and enables the system to enter a stable control state with position self-healing and following performance.

[0040] Example 3: This embodiment was verified on a mold milling test bench employing the position-following characteristics of a multi-axis linkage CNC system. The position control loop sampling frequency of the test bench was 500Hz, the grating ruler feedback resolution was 0.1μm, and the motion control card had a multi-channel interpolation register overwrite function. The interpolation cycle was set to 2ms, and each servo feed axis cut into the variable curvature corner surface at a speed of 6000mm / min. This interpolation cycle is close to the lower limit of the allowable value range of the control loop, which can reduce the discrete aliasing of the trajectory following signal. The reference time scale of the CNC system was determined to be 2ms. The control unit selected three consecutive interpolation cycles, including the current interpolation cycle, as a sliding window to balance the error trend recognition delay and random interference smoothing. The test signal source introduced Gaussian white noise with a signal-to-noise ratio of 25dB, and simultaneously superimposed a high-frequency micro-vibration of the transmission chain with an amplitude of 0.005mm. The control unit calculated the difference between the motor encoder feedback position and the CNC system command position to form a real-time tracking error sequence for the position control loop. Its unit is mm, where the subscript is... Indicates the current sampling time, subscript This indicates the servo feed axis index.

[0041] The preset discrete-time filter adopts a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module. The control unit will... Input a high-order low-pass filter matrix to filter the error data for three consecutive interpolation periods, and output the low-frequency deviation trend component. The high-pass differential module extracts the high-frequency error fluctuation component based on the difference between the real-time tracking error and the low-frequency output. The experiment was conducted with three surface curvature change rates: 0.02 mm / s / ms, 0.08 mm / s / ms, and 0.25 mm / s / ms. The cutoff frequency of the servo drive's speed control loop was also tested. Set to 400Hz, position control loop proportional gain Set to 40 Band matching scaling coefficient Take 0.12, according to the formula The loop gain attenuation factor was calculated. The value is 1.2, which is a dimensionless constant; the historical steady-state stiffness constant. The milling stiffness coefficient of the machine tool, pre-calibrated on the test bench and set to 150 N / μm, was used as the stiffness reference for the dynamic stiffness feedforward mapping model, and the positive acceleration threshold was set. Set to 0.05 mm / ms, negative release threshold. Set to -0.03 mm / ms.

[0042] Under a medium-level curvature change rate of 0.08 mm / millisecond, the high-frequency error fluctuation component extracted by the experimental group The transient amplitude is 0.042 mm. The dynamic stiffness feedforward mapping model is based on the low-frequency deviation trend component. The amplitude is retrieved from the nonlinear stiffness mapping table to determine the servo stiffness coefficient corresponding to the current variable curvature load. This coefficient is then used as the denominator to complete the spatial displacement compensation scale conversion. The control unit then processes the transient elastic offset vector of the tool tip. With servo hysteresis correction vector Perform vector summation and scaling to calculate the comprehensive position compensation vector. : ,in, It is a three-dimensional vector and serves as the transient spatial position compensation for the current interpolation period; and The units are all mm; the control unit calculates the amplitude gradient rate of change based on the amplitude change of the transient spatial position compensation in adjacent interpolation cycles. , where subscript This represents the sliding window index, obtained from three consecutive interpolation cycles. The values ​​are 0.021 mm / ms, 0.058 mm / ms, and 0.064 mm / ms, respectively. The value of 0.058 mm / ms in the second interpolation cycle is greater than... At 0.05 mm / ms, the state switching logic switches to the accelerated deformation counterbalancing state, and the control unit will... The write address in the position instruction register is shifted forward by one interpolation cycle. Control group one disables the register pointer pre-control flow and maintains in-phase synchronization blocking mode. Measurements were taken under the same medium-level operating conditions. The transient amplitude is 0.078 mm. Since no timing shift was performed, The current interpolation cycle address is still written to the position instruction register; control group two will... Set to 0.40 mm / ms to deviate from the working window; the results obtained under the same abrupt load... The peak value was 0.092 mm / ms, which was less than 0.40 mm / ms, indicating that the pre-write operation of the position command register was not triggered.

[0043] Under a high curvature change rate of 0.25 per millisecond, the test group After the speed increases to 0.185 mm / ms, the tool moves away from the corner and enters the cutting load unloading phase. It dropped to -0.052 mm / ms, below The value is -0.03 mm / ms. The state switching logic switches to the state of smooth reduction of control quantity and closes the mapping path of the dynamic stiffness feedforward mapping model. The discrete gradient decay calculation module reads the ratio of the current cycle nonlinear residual to the previous cycle nonlinear residual, uses this ratio as the reference decay rate, and multiplies it with the preset system damping coefficient to obtain a step adjustment coefficient of 0.45 in the first interpolation cycle. In the second interpolation cycle, the coefficient of the previous cycle is subjected to quadratic power decay and truncation to obtain a step adjustment coefficient of 0.12. The step adjustment coefficients of 0.45 and 0.12 are multiplied by the transient spatial position compensation amount of the previous interpolation cycle in turn, and the position command register is overwritten to make the residual feedforward amount truncated and cleared to zero in two consecutive interpolation cycles. The control group 2 will also... Set to 0.3, this value is lower than The effective operating window is limited. Due to the small loop gain attenuation factor, the servo lag feedforward compensation is severely insufficient, leading to dynamic damping mismatch and phase lag in the control loop. Consequently, the feedback control quantity of the position control loop exhibits lag overshoot oscillation, causing closed-loop resonance divergence and a decrease in system stability.

[0044] During the reversal process of each servo feed axis, the forward prediction window captures the velocity-acceleration vector according to the interpolation cycle. When this vector indicates that the feed axis velocity crosses the 0mm / s critical point, and the feed axis direction status register flips, the control unit triggers the reversal gating cutoff unit and maintains it for 15ms, outputting a gating cutoff signal to... The gain weight is adjusted to zero to block the transmission of commutation impulse noise to the transient spatial position compensation. During the gating period, the control unit calls the stored historical steady-state stiffness constant. The servo stiffness reference is reset by writing it into the dynamic stiffness feedforward mapping model. After the weights of the high-frequency components are set to zero, the corrected transient spatial position compensation is output. The system generates a smooth transition position correction. After the 15ms action time expires and the commutation impact noise frequency drops below the preset cutoff frequency, the commutation gate cutoff unit is closed, and the system resumes phase counter-control. This decouples the nonlinear load change of the commutation position from the closed-loop adjustment channel of the position tracking control loop, suppressing the stiffness mapping calculation divergence caused by the transmission backlash and the impact noise generated by the frequent commutation of the servo feed axis.

[0045] Under the alternating action of control flow at each stage, the CNC system will integrate the position compensation vector. The original interpolation position command superimposed on the current interpolation cycle Generate correction position control commands And output to the servo driver, where, For three-dimensional vector coordinates, The position is a three-dimensional vector that changes over time. Under the premise of keeping the gain parameters of each loop of the control loop in a resting state, this process eliminates the spatiotemporal phase lag between identification and terminal action, reduces the overall machining surface contour error by more than 75%, and eliminates the surface overcutting defect caused by reverse pulse overshoot, so that the control system can maintain error convergence and position following state.

[0046] Example 4: This embodiment combines Figures 1 to 2 This paper describes a dynamic compensation control method and system for mold processing errors, such as... Figure 1 As shown, the process flow sequence includes step S1 of obtaining the real-time tracking error sequence of each axis position loop, step S2 of filtering and decoupling to extract high-frequency fluctuations and low-frequency deviation components, step S3 of mapping and calculating the transient position compensation amount and writing it into the register, and step S4 of forcing the weight to repeat to zero and correcting the compensation amount when the commutation crosses zero.

[0047] like Figure 2 As shown, the dynamic compensation control system for mold processing errors includes a cascaded filter structure, a dynamic stiffness feedforward mapping model, state switching logic, and a commutation gate cutoff unit. The cascaded filter structure includes a high-order low-pass filter matrix and a high-pass differential module. The dynamic stiffness feedforward mapping model includes a nonlinear stiffness mapping table and a discrete gradient attenuation calculation module. The state switching logic includes an accelerated deformation hedging state and a smooth reduction state of the control quantity. The commutation gate cutoff unit includes a forward prediction window and a historical steady-state stiffness constant.

[0048] Example 5: This embodiment enables the register data flow and dynamic control process of a multi-axis linkage CNC system in precision continuous milling of curved surfaces to be reproduced with the same parameters. This embodiment further explains the discrete-time domain filter, dynamic stiffness feedforward mapping model, and threshold calibration process. During continuous cutting, nonlinear abrupt changes in cutting load can cause deviations in the motion trajectory. When the calculation order of filtering, stiffness mapping, and gating criteria is unclear, the feedforward overwrite timing of position commands is prone to deviation and can cause overshoot oscillations. The control unit uses continuous interpolation cycles as the reference time scale, synchronously reads the motor encoder feedback position and the CNC system command position from the position loop register of each servo feed axis, calculates the difference between the two, and forms a real-time tracking error sequence for the position control loop. The discrete-time filter employs a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module. The high-order low-pass filter matrix performs matrix operations on the error values ​​of the current interpolation period and the two preceding interpolation periods according to preset filter coefficients, outputting a low-frequency deviation trend component. The high-pass differential module calculates the first-order backward difference between the current error value and the error value of the previous interpolation cycle, and uses the position control loop to track the error sequence in real time. Subtract low-frequency deviation trend component High-frequency error fluctuation components were obtained. Therefore, the low-frequency deviation trend component is used to characterize the low-frequency lag of the transmission chain, and the high-frequency error fluctuation component is used to characterize the high-frequency changes of the cutting load.

[0049] In the dynamic stiffness feedforward mapping model, the control unit is based on the low-frequency deviation trend component. The amplitude is retrieved from a nonlinear stiffness mapping table to determine the servo stiffness coefficient corresponding to the current variable curvature load. This mapping table uses historical steady-state stiffness constants stored in memory. As a stiffness reference, and configuring corresponding servo stiffness coefficients according to different amplitude ranges of low-frequency deviation trend components, the control unit uses the retrieved servo stiffness coefficients as the denominator, and combines them with the control loop gain to account for high-frequency error fluctuation components. Mapped to the spatial displacement compensation scale, a transient elastic offset vector of the tool tip is generated. Error data in the position loop register is stored and transmitted as discrete pulse counts. Before performing stiffness mapping, the control unit multiplies the pulse counts by the position feedback scaling factor through a scaling interface, converting them into μm-level displacement. The dynamic stiffness feedforward mapping model converts the converted high-frequency error fluctuation components into corresponding cutting load changes according to the control loop gain, then divides them by the current servo stiffness coefficient to obtain intermediate deformation in μm units. This deformation is then converted into a tool tip transient elastic offset vector in mm units through shift calibration. This process enables the discrete count values, servo stiffness coefficients, and spatial displacement compensation values ​​in the register to participate in the calculation on the same scale.

[0050] The control unit synchronously reads the low-frequency deviation trend component. The control unit reads the position loop gain parameters stored in the servo driver. The loop position feedforward time constant is taken from the time constant reference corresponding to the reciprocal of the position loop gain. The control unit multiplies the low-frequency deviation trend component and the current feed rate with this time constant to generate a servo hysteresis correction vector. The dynamic stiffness feedforward mapping model then applies the transient elastic offset vector of the tool tip. and servo hysteresis correction vector By combining the results, we obtain the comprehensive position compensation vector for the current interpolation period. The transient spatial position compensation value is written into the position command register. Before the CNC system enters the machining state, the control unit drives each servo feed axis to reciprocate in a no-cutting, unloaded state, and reads the first-order discrete differential data of the real-time tracking error of the position control loop. This differential data is converted to the amplitude change scale of the transient spatial position compensation value through a discrete time domain filter and a dynamic stiffness feedforward mapping model. The control unit then calculates its root mean square value within a preset sliding time window. And based on this, determine the positive acceleration threshold. : ,in, and The units are all mm / ms, and the constant 3 is a dimensionless multiplier factor. The control unit will calculate the positive acceleration threshold. Write to the decision register to calibrate the actual error fluctuation of the gate threshold under the no-load state of the random bed, and avoid the no-load noise triggering the pre-writing of the register.

[0051] In subsequent processing, the control unit calculates the amplitude gradient rate of change based on the transient spatial position compensation amplitude of adjacent interpolation cycles and the duration of the interpolation cycle. ,when Two consecutive interpolation cycles greater than the positive acceleration threshold At that time, the state switching logic switches to the accelerated deformation hedging state and integrates the position compensation vector. The write address in the position command register is shifted forward by one interpolation cycle, and the servo driver outputs the corresponding drive torque in advance accordingly, reducing the phase lag between the compensation command and the transient elastic offset caused by the 15ms loop delay. After running according to the above register data stream and control sequence, the spatial trajectory contour error at the variable curvature corner is reduced to below 0.011mm, the machining overcutting bulge defect caused by the spatial transmission backlash is eliminated, and the control loop maintains position error convergence and high-precision following state under dynamic nonlinear load disturbance.

[0052] Example 6: In this embodiment, after each servo feed axis stops and comes to a standstill, the control unit recalibrates the transmission chain stiffness. During calibration, the control unit injects a sweep excitation voltage signal with a frequency ranging from 10Hz to 500Hz into the servo circuit, simultaneously acquiring transmission chain electromagnetic torque data and grating ruler position feedback displacement data, and generating a corresponding frequency response curve. The control unit converts the sweep excitation voltage signal into a thrust value, extracts the reciprocal of the static gain from the low-frequency band of the frequency response curve, and calculates the transmission chain stiffness coefficient. : ,in, The thrust value is the converted value of the frequency sweep excitation voltage signal, in N; The grating ruler feedback displacement amplitude is in μm; the transmission chain stiffness coefficient is... The unit is N / μm; the control unit will use the calibrated transmission chain stiffness coefficient. The data is written to a dedicated non-volatile memory as a stiffness reference for the nonlinear stiffness mapping table and the dynamic stiffness feedforward mapping model. Under steady-state milling conditions, this stored value is used as a historical steady-state stiffness constant. During dynamic compensation, the control unit retrieves the nonlinear stiffness mapping table based on the amplitude of the low-frequency deviation trend component to determine the servo stiffness coefficient corresponding to the current variable curvature load. This servo stiffness coefficient is then used as the denominator to calculate the transient spatial position compensation. Thus, the transmission chain stiffness coefficient is used to establish and update the stiffness reference, the servo stiffness coefficient is used to characterize the actual mapping result under the current load condition, and the historical steady-state stiffness constant is used for compensation correction during feed axis reversal gating.

[0053] When the proportional loop gain of the servo controller is adjusted, the control unit reads the position loop gain and uses it as the control loop gain parameter of the dynamic stiffness feedforward mapping model; at the same time, it adjusts the loop gain attenuation factor according to the cutoff frequency of the speed loop inside the servo driver. When the speed loop cutoff frequency is 400Hz, the control unit calculates the ratio of the speed loop cutoff frequency to the position loop gain to obtain the reference gain attenuation coefficient, and writes this coefficient into the register as the current loop gain attenuation factor. When the servo driver switches the current loop sampling rate from 10kHz to 5kHz, the control unit reads the sampling rate change signal and adjusts the loop gain attenuation factor accordingly. The register values ​​are rewritten proportionally by doubling, and the dynamic stiffness feedforward mapping model is calculated synchronously using the updated control loop gain, transmission chain stiffness reference, and loop gain attenuation factor, so that the output amplitude of the transient spatial position compensation is adjusted according to the change of the current operating frequency band of the drive frequency converter circuit.

[0054] Example 7: In this embodiment, during long-term continuous milling, the electromagnetic coil continuously heats up, and high-speed mechanical friction causes the dynamic parameters of each servo feed axis to gradually drift. As the machining time increases, the working reference of the position control loop deviates from the initial calibration state, and the transient spatial position compensation calculated by the dynamic stiffness feedforward mapping model accumulates deviation. This type of position drift exhibits a slow, non-linear increasing trend, and dynamic compensation within a short interpolation cycle is insufficient to eliminate its long-term effects. It easily leads to accumulated position deviations in the middle and later stages of profile cutting, causing workpiece geometric dimensions to exceed tolerances and surface roughness to decrease. After each machining cycle enters the rest period of the control system, the control unit activates the online self-calibration loop. This loop establishes a sliding time window in the memory connected to the control unit's data bus, with an update step of 80ms. It continuously reads the grating ruler feedback pulses of each servo feed axis under no-load uniform speed conditions and calculates the actual motion trajectory of each servo feed axis based on the grating ruler feedback resolution. The control unit synchronously reads the original interpolation position command. Calculate the actual motion trajectory and the original interpolation position command. The difference between them is used to generate closed-loop steady-state accuracy verification data.

[0055] The control unit extracts the real-time tracking error sequence of the position control loop from the closed-loop steady-state accuracy verification data according to the axis number. The arithmetic mean of the sequence within the sliding time window is calculated and used as the thermal instability drift caused by long-term system operation. ,in, A scalar sequence in mm, with subscripts Indicates baseline drift resulting from accumulated operating time, subscript This indicates the axis number of the servo feed axis. During the resting period of the control system, the control unit stores the closed-loop steady-state accuracy verification data and the corresponding thermal instability drift of each axis. Write to non-volatile memory to update the control reference for the next processing cycle.

[0056] Before the start of the next processing cycle, the control unit reads the updated control reference from the non-volatile memory, including the thermal instability drift. As the inverse static correction, it is written into the input register of the dynamic stiffness feedforward mapping model. In the arithmetic accumulator, the model uses the tooltip transient elastic offset vector of the current interpolation cycle. Subtract the thermal instability drift corresponding to the same axis number The system rewrites the numerical boundary of the control gain constant based on the updated control reference, thereby periodically correcting the effects of long-term temperature rise and mechanical friction on the transient spatial position compensation. The online self-calibration loop, without changing the gain parameters of each feedback loop, updates the original interpolation position command. Feedforward correction is performed to change the position deviation caused by thermal deformation accumulation from continuous increase to gradual convergence. After adopting this control method, the cumulative position deviation and overcutting defects in the mid-to-late stage of surface cutting are suppressed, the overall machining surface contour error is reduced by more than 75%, and the multi-axis linkage CNC system maintains a long-term stable position following state.

Claims

1. A method for dynamic compensation and control of mold processing errors, characterized in that, Includes the following steps: Step S1: The control unit uses the current interpolation cycle of the CNC system as the reference time scale to synchronously acquire the real-time tracking error sequence of the position control loop of each servo feed axis; Step S2: The control unit uses a preset discrete time-domain filter to process the real-time tracking error sequence of the position control loop, and decouples and extracts the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. Step S3: The control unit inputs the high-frequency error fluctuation component and the low-frequency deviation trend component into the dynamic stiffness feedforward image model constructed based on the control loop gain and the transmission chain stiffness coefficient. Through the nonlinear mapping of the high-frequency error fluctuation component and the low-frequency deviation trend component to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the index for the current interpolation period. This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. The quantitative relationship is as follows In the formula The cutoff frequency of the servo driver speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scale coefficients. The range of values ​​is ; In step S4, when the axial movement direction of each servo feed axis changes and the current speed of the feed axis passes through the 0mm / s critical point, the control unit triggers the reversing gate cutoff unit and continues for 15ms. It outputs a gate cutoff signal to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and calls the stored historical steady-state stiffness constant to correct the transient spatial position compensation amount.

2. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Step S3 includes the following steps: Step S31, the control unit calculates the amplitude gradient change rate of the transient spatial position compensation; Step S32, when the amplitude gradient change rate is less than the set negative release threshold, the control unit switches to the smooth reduction state of the control quantity using state switching logic, closes the mapping path of the dynamic stiffness feedforward mapping model, and calls the discrete gradient decay calculation module to calculate the decreasing step adjustment coefficient based on the nonlinear residual in the current interpolation period. In the formula The nonlinear residual for the current interpolation period, Let be the system damping coefficient, such that the step adjustment coefficient multiplied by the transient spatial position compensation of the previous interpolation cycle satisfies ... The value of the position instruction register is overwritten to truncate the transient spatial position compensation.

3. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Step S2 includes the following steps: Step S21, the control unit configures a cascaded filter structure consisting of a high-order low-pass filter matrix and a high-pass differential module; Step S22, the control unit inputs the real-time tracking error sequence of the position control loop into the high-order low-pass filter matrix to output the low-frequency deviation trend component. ,in To track the error sequence in real time, To satisfy the preset filter weight coefficients for normalization, and to capture high-frequency disturbances using a high-pass differential module, according to... To output high-frequency error fluctuation components.

4. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Step S4 includes the following steps: Step S41, the control unit completes the construction of the forward prediction window to capture the velocity and acceleration vectors of each servo feed axis; Step S42, when the velocity and acceleration vectors indicate that the velocity of each servo feed axis crosses the 0 mm / s critical point, the commutation gating cutoff unit adjusts the weight of the high-frequency error fluctuation component to zero, blocks the transmission of commutation impact noise to the transient spatial position compensation, and during the gating period, uses the historical steady-state stiffness constant. Write the dynamic stiffness feedforward image model to reset the servo stiffness reference, and output the transient spatial position compensation amount consisting of the servo hysteresis correction vector. .

5. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, After calculating the transient spatial position compensation amount using the dynamic stiffness feedforward mapping model, the following steps are further included: Step S5, the control unit completes the mounting of the collaborative adjustment architecture consisting of the forward prediction window and the update loop; Step S6, the control unit increases the cutoff frequency of the discrete time domain filter by a preset proportional step size when the tracking error envelope variance is greater than a preset variance threshold, and decreases the cutoff frequency when it is less than a preset variance threshold, based on the tracking error envelope variance of five consecutive interpolation cycles, so as to calculate the transient spatial position compensation amount based on the high-frequency error fluctuation components.

6. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Further steps include: Step S7: The control unit uses state switching logic to monitor the state of the interpolation buffer in the control system; Step S8: When the surface curvature in the interpolation buffer undergoes a step change and the rate of change of curvature exceeds the set abrupt change threshold, the state switching logic switches the control loop from the normal dynamic compensation mode to the edge condition energy level reduction mode, and multiplies the output weight of the transient spatial position compensation by a preset derating coefficient. The adjustment was made, among which The range of values ​​is .

7. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Further steps include: Step S9: The control unit acquires the grating ruler feedback pulse of the servo driver to calculate the actual motion trajectory of each servo feed axis; Step S10: The control unit acquires the original interpolation position command, calculates the difference between the actual motion trajectory and the original interpolation position command to generate closed-loop steady-state accuracy verification data, and writes the closed-loop steady-state accuracy verification data into the non-volatile memory during the rest period of the control system to update the control reference.

8. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Step S3 includes the following steps: Step S33, the control unit configures a nonlinear stiffness mapping table in the dynamic stiffness feedforward image model; Step S34, the control unit retrieves the nonlinear stiffness mapping table based on the amplitude of the low-frequency deviation trend component to determine the servo stiffness coefficient corresponding to the current variable curvature load. And through the formula: Calculate the transient elastic offset vector of the tool tip In the formula For position loop control gain, The pulse equivalent scaling factor; the control unit calculates the current feed rate of the servo feed axis. Proportional gain of position control loop The corresponding feedforward time constant, according to the formula Calculate the servo hysteresis correction vector .

9. The method for dynamic compensation and control of mold processing errors according to claim 1, characterized in that, Further steps include: Step S11: The control unit synchronously acquires the decoupling components of each servo feed axis; Step S12: The control unit calculates the contour error vector between two adjacent servo feed axes and injects the contour error vector as a cross-coupling constraint into the generation loop of the transient spatial position compensation quantity to offset the phase deviation between each servo feed axis.

10. A dynamic compensation control system for mold processing errors, used to implement the dynamic compensation control method for mold processing errors as described in claim 1, characterized in that, include: The error acquisition module is used to synchronously acquire the real-time tracking error sequence of the position control loop of each servo feed axis, using the current interpolation cycle of the CNC system as the reference time scale. The data processing module, connected to the error acquisition module, is used to process the real-time tracking error sequence of the position control loop using a preset discrete time-domain filter, and decouple and extract the high-frequency error fluctuation component that characterizes the high-frequency fluctuation of the cutting load and the low-frequency deviation trend component that characterizes the low-frequency lag trend of the transmission chain. The feedforward calculation module, connected to the data processing module, is used to input high-frequency error fluctuation components and low-frequency deviation trend components into a dynamic stiffness feedforward mapping model constructed based on the control loop gain and transmission chain stiffness coefficient. Through the nonlinear mapping of high-frequency error fluctuation components and low-frequency deviation trend components to the spatial displacement compensation scale, according to the formula: Calculate the transient spatial position compensation amount for the current interpolation cycle. And write it into the internal position instruction register; where, This is the transient elastic offset vector of the blade tip. This is the servo hysteresis correction vector. This is the loop gain attenuation factor. In the formula The cutoff frequency of the speed control loop. For the proportional gain of the position control loop, These are the frequency band matching scaling coefficients; The gating limit module, connected to the feedforward calculation module, is used to trigger the reversing gating cutoff unit for 15ms when the axial movement direction of each servo feed axis changes and the current speed of the feed axis crosses the 0mm / s critical point. The gating cutoff signal is output to force the gain weight of the high-frequency error fluctuation component to be adjusted to zero, and the stored historical steady-state stiffness constant is called to correct the transient spatial position compensation amount.

Citation Information

Patent Citations

  • Methods for cross-quadrant friction compensation in CNC machine tool axis linkage machining processes

    CN107918356B

  • Error compensation comprehensive control method for numerical control platform trajectory planning

    CN111427308A

  • On-line iterative compensation system and method for servo errors

    CN116203895B