A method for adjusting the machining path of a large vertical machine tool

By acquiring and processing multi-source machining status information of large vertical machine tools, establishing the correlation between spatial stiffness and tool position deviation, and generating correction amounts for each feed axis, the problem of real-time compensation for dynamic position deviation in heavy-duty deep cavity machining of large vertical machine tools is solved, thereby improving machining accuracy and stability.

CN122506987BActive Publication Date: 2026-11-17WUHAN JIA YOUJIA EQUIP MFG CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202610649390.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-05-12
Publication Date
2026-11-17
Estimated Expiration
2046-05-12

AI Technical Summary

Technical Problem

Large vertical machine tools have difficulty compensating for dynamic positional deviations in real time when machining deep cavities under heavy loads, resulting in poor machining accuracy and stability, and may even cause chatter, affecting workpiece quality and yield.

Method used

By acquiring information such as spindle drive status, ram overhang, crossbeam motion status, and tool installation parameters during automatic machining, the correlation between changes in machine tool spatial stiffness and tool position deviation is established, generating correction amounts for each feed axis, and switching to amplitude limiting correction state when the cutting load exceeds the threshold, thereby achieving real-time trajectory adjustment.

Benefits of technology

It effectively reduces tool marks in deep-machined areas, reduces cylindricity and coaxiality deviations, improves the dimensional accuracy and machining stability of long-stroke deep-cavity machining of heavy workpieces, and enhances the operational reliability and control safety of heavy-duty machining processes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122506987B_ABST
    Figure CN122506987B_ABST
Patent Text Reader

Abstract

This invention relates to the field of machine tool control technology and discloses a method for adjusting the machining path of a large vertical machine tool, which solves the problem of difficulty in real-time compensation of dynamic position deviation in traditional methods. The method first collects information on the spindle, slide, crossbeam and feed axis, and establishes the correlation between spatial stiffness change and tool position deviation in combination with cutting load. It generates trajectory correction amount and applies it to the machining path execution process. Under heavy load, it performs amplitude limiting control, recovery control and takeover control, thereby improving the form and position accuracy, machining stability and yield of long stroke deep cavity machining.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of machine tool control technology, specifically to a method for adjusting the machining path of a large vertical machine tool. Background Technology

[0002] Large vertical machine tools are widely used in the long-stroke deep cavity machining and large-diameter contour machining of heavy workpieces such as wind power, nuclear power and large aerospace components. In the prior art, error compensation methods are usually used to improve machining accuracy. For example, the published invention patent application CN101943896B discloses a trajectory regeneration compensation method for CNC machine tool errors, which mainly achieves trajectory compensation by establishing machine tool geometric error and thermal error models. Another example is the published invention patent application CN102513882B, which discloses a swingable ram flexure deformation compensation device, which mainly compensates for the flexure deformation after the ram extends. Although the above-mentioned prior art can improve the machining accuracy of machine tools to a certain extent, its compensation targets are mainly errors under no-load or quasi-static conditions of the machine tool, or only for the flexure deformation in a single direction of the ram. Therefore, it is difficult to adapt to the complex dynamic machining conditions of large vertical machine tools under heavy-load deep cavity machining conditions.

[0003] However, when large vertical machine tools perform long-stroke deep cavity machining or large-diameter contour machining, the overall bending stiffness and torsional stiffness of the machine tool will change significantly as the ram overhang increases. At the same time, the uneven distribution of blank allowance in large castings and forgings can easily cause fluctuations in cutting load, which in turn leads to flexural deformation and tool axis posture deviation during machining. Existing technologies usually focus on static error compensation or single component deformation compensation, which is difficult to decouple in real time from the dynamic posture deviation caused by the combined effects of structural stiffness changes and cutting load fluctuations, and it is also difficult to complete the feedforward adjustment of the machining trajectory within the interpolation cycle. As a result, deviations are easily generated between the actual machining trajectory and the theoretical trajectory, which can lead to tool marks, cylindricity, and coaxiality exceeding tolerances in deep-machined parts. In severe cases, it may even induce machining chatter, affecting the workpiece machining quality and even causing the workpiece to be scrapped. Therefore, it is urgent to propose a machining path adjustment method for large vertical machine tools to solve the problem of the difficulty in real-time compensation of dynamic posture deviation in existing technologies, thereby improving the form and position accuracy, machining stability, and yield of long-stroke deep cavity machining of heavy workpieces. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for adjusting the machining path of a large vertical machine tool, which solves the problem of difficulty in real-time compensation for dynamic pose deviations in traditional methods.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for adjusting the machining path of a large vertical machine tool includes:

[0007] S1. During the automatic machining process of the machine tool, the spindle drive status, ram overhang, crossbeam motion status, tool installation parameters and feedback information of each feed axis are acquired in a unified time sequence.

[0008] S2. Based on the changes in ram overhang and cutting load, determine the correlation between the changes in machine tool spatial stiffness and tool position deviation under the current machining state;

[0009] S3. Based on the correlation, determine the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory, and generate the corresponding correction amount for each feed axis.

[0010] S4. Perform constraint processing and timing matching on the correction amount of each feed axis to obtain the trajectory correction amount corresponding to the current interpolation command;

[0011] S5. Apply the trajectory correction amount to the current machining path during execution, and switch to amplitude limiting correction state when the cutting load exceeds the preset threshold.

[0012] Preferably, S1 includes:

[0013] When the processing and operating conditions are met, a unified time reference is established based on a distributed clock, and operating data is collected synchronously according to the interpolation cycle.

[0014] Combine the tool change status and spindle clamping status to load or update the tool installation parameters;

[0015] The system performs verification of the sampling window, smoothing of the sampled data, conversion of overhang, determination of beam motion status, and monitoring of abnormal rereading, rollback, and continuous abnormality.

[0016] Preferably, S2 includes:

[0017] The equivalent cutting load is calculated based on the spindle torque current, spindle speed, transmission parameters, and tool parameters.

[0018] Speed ​​regulation disturbances are screened out by combining changes in the spindle angular velocity;

[0019] Load jumps are determined based on the load change rate and load increment;

[0020] The stiffness state is determined based on the ratio of the ram overhang to the span of the reference support.

[0021] The coupling parameters and update frequency of the six-degree-of-freedom compliance matrix are adjusted in conjunction with the load conditions;

[0022] After aligning the tool cutting direction and spatial offset vector to the machine tool global coordinate system, a generalized load vector is constructed, the six-dimensional deviation trend is obtained, and attenuation backoff and freeze observation processing are performed on the correlation results that exceed the physical boundary.

[0023] Preferably, speed regulation disturbance screening is performed in conjunction with changes in the spindle angular velocity, including:

[0024] Information on spindle angular velocity variation is incorporated into the equivalent cutting load conversion process;

[0025] The judgment boundary is determined based on the fluctuation range of the angular velocity change rate during the no-load steady-speed operation of the spindle and the stable range of the speed loop of the drive. Accordingly, the corresponding period is divided into speed regulation transition period and associated update period. The speed regulation transition period is used for state tracking, and the associated update period is used for updating load associated parameters.

[0026] Preferably, S3 includes:

[0027] The spatial transfer relationship of the cutting point is established based on the theoretical tool tip position sequence, the theoretical tool axis direction sequence, and tool installation parameters.

[0028] The six-dimensional deviation trend is converted into a comprehensive spatial position deviation and attitude deviation.

[0029] For continuous deviation sequences, frequency band decomposition, risk band elimination, and phase consistency weighted screening are performed.

[0030] The filtered effective deviations are mapped to the correction amounts of each feed axis, and single-cycle constraints, cumulative constraints, recovery control, delayed release, saturation marking, associated invalidation retention, and status recording are performed on each feed axis correction amount.

[0031] Preferably, the six-dimensional deviation trend is converted into a comprehensive spatial position deviation and attitude deviation, including:

[0032] Based on the tool tip spatial offset, the effective tool overhang length, and the tool head radius, establish the spatial transfer relationship from the ram end reference point to the tool cutting point;

[0033] The six-dimensional deviation trend is converted into the spatial position deviation of the cutting point relative to the theoretical trajectory and the attitude deviation relative to the theoretical tool axis direction.

[0034] The normal additional displacement, tangential additional displacement, and contact position offset corresponding to the attitude deviation are superimposed with the translational deviation.

[0035] Preferably, S4 includes:

[0036] When the program segment is in the compensation-allowed state, the original correction amount is compressed, slope limited, direction reduced, and key direction prioritized allocation is performed based on the physical capabilities, stroke boundaries, and theoretical trajectory occupancy status of each feed axis.

[0037] Based on the interpolation point number and time index, perform forward matching or time-series delay alignment on the original correction amount;

[0038] Smoothing is performed on adjacent reverse corrections and transition corrections, and these are superimposed to form a trajectory correction with status markers, buffer release markers, and risk segment markers.

[0039] Preferably, based on the interpolation point number and time index, forward matching or temporal delay alignment is performed on the original correction amount, including:

[0040] Establish interpolation point numbers and time indexes for theoretical interpolation points, and establish corresponding correction point indexes to align the original correction values ​​to the interpolation point to be executed.

[0041] When the correction queue lags behind the interpolation queue by more than one interpolation cycle, and the original correction amounts of the current cycle and the next cycle are both valid and in the same direction, linear prediction is made based on the difference between the original correction amounts of the two cycles.

[0042] When the correction queue is ahead of the interpolation queue by more than one interpolation cycle, the corresponding correction amount is written to the timing buffer and superimposed on the output when the corresponding interpolation point is executed.

[0043] Preferably, S5 includes:

[0044] The trajectory correction is superimposed onto the execution trajectory using position and velocity feedforward methods;

[0045] The correction status is confirmed and the correction weights are adjusted based on position feedback and following error;

[0046] The system switches between pre-limited and strong-limited states based on two levels of cutting load thresholds, and performs critical direction retention, non-critical direction compression, attitude-related correction freezing, feed deceleration, heavy load observation, manual takeover judgment, recovery judgment, recovery release, and status recording processing.

[0047] Preferably, the correction status is confirmed and the correction weight is adjusted based on position feedback and following error, including:

[0048] In the next sampling cycle after servo execution, the actual position feedback and following error are read. The theoretical compensation direction is determined by combining the current position correction value and the velocity feedforward correction value. The current correction state is confirmed based on the consistency between the following error and the theoretical compensation direction and the amplitude change within the continuous cycle.

[0049] Maintain the current correction weight when the following error continues to decrease, and reduce the correction weight for the next cycle when the following error increases.

[0050] Compared with the prior art, the present invention provides a method for adjusting the machining path of a large vertical machine tool, which has the following beneficial effects:

[0051] 1. This invention, by uniformly acquiring spindle drive status, ram overhang, crossbeam motion status, tool installation parameters, and feedback information from each feed axis during automated machining, and establishing the correlation between machine tool spatial stiffness changes and tool position deviations in conjunction with cutting load variations, converts the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory into correction amounts for each feed axis. Then, by constraining, timing-matching, and limiting these correction amounts, it can correct dynamic position deviations caused by changes in ram overhang and cutting load fluctuations within the interpolation cycle. This reduces tool marks in deep-machined areas, lowers the probability of cylindricity and coaxiality deviations, mitigates chatter effects during heavy-load machining, and improves the form and position accuracy, machining stability, and yield of long-stroke deep-cavity machining of heavy workpieces.

[0052] 2. This invention, by loading the trajectory correction amount into the interpolation execution process and combining it with the cutting load threshold for pre-limiting switching, strong limiting switching, recovery release, and manual takeover judgment, can coordinate and control the correction amplitude, correction rhythm, and feed state when heavy load fluctuations, sudden changes in allowance, and local impact loads occur. This avoids sudden changes in correction amount, overcompensation, and continuous amplification of correction amount under abnormal working conditions, enhances the continuity and controllability of the trajectory correction process, and improves the operational reliability and control safety of heavy-load machining processes on large vertical machine tools. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of a method for adjusting the machining path of a large vertical machine tool according to the present invention;

[0054] Figure 2 This is a block diagram of the multi-source processing status information acquisition and unified timing synchronization of the present invention;

[0055] Figure 3 This is a schematic diagram illustrating the relationship between spatial stiffness variation and tool pose deviation in this invention.

[0056] Figure 4 This is a schematic diagram illustrating the conversion of the tool cutting point deviation to the correction amount of each feed axis in this invention.

[0057] Figure 5 This is a schematic diagram illustrating the correction constraints, cache release, and timing matching of the present invention;

[0058] Figure 6 This is a diagram showing the online loading of trajectory correction and the switching of amplitude limiting recovery states in this invention.

[0059] Figure 7 This is a schematic diagram of the heavy-duty deep cavity machining scenario and key quantity arrangement of the large vertical machine tool of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described 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.

[0061] Example 1: Figure 1 - Figure 7 A method for adjusting the machining path of a large vertical machine tool is presented, including:

[0062] S1. During the automatic machining process of the machine tool, the spindle drive status, ram overhang, crossbeam motion status, tool installation parameters and feedback information of each feed axis are acquired in a unified time sequence.

[0063] S2. Based on the changes in ram overhang and cutting load, determine the correlation between the changes in machine tool spatial stiffness and tool position deviation under the current machining state;

[0064] S3. Based on the correlation, determine the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory, and generate the corresponding correction amount for each feed axis.

[0065] S4. Perform constraint processing and timing matching on the correction amount of each feed axis to obtain the trajectory correction amount corresponding to the current interpolation command;

[0066] S5. Apply the trajectory correction amount to the current machining path during execution, and switch to amplitude limiting correction state when the cutting load exceeds the preset threshold.

[0067] In specific application scenarios, large vertical machine tools are used for heavy-duty, long-stroke deep cavity machining or large-difference contour machining of large castings and forgings such as wind turbine gear rings, nuclear power main pump housings, and aero-engine casings. The machine tool is in automatic continuous machining mode, with the spindle in a rotation-enabled state, and each feed axis performing linkage interpolation according to the CNC program. The data involved in the method operation includes at least spindle drive operation data, ram vertical displacement data, beam lateral displacement data, position feedback data for each feed axis, tool installation geometric parameters, theoretical trajectory data of the current program segment, and cutting force change information reflected by the spindle load. The above input information can come from the CNC system's underlying position and speed loop feedback, spindle driver, absolute linear encoder, rotary encoder, and tool management database. The processing objective of this method is to... This method enables online identification, correlation, correction, and control of machine tool spatial stiffness changes, tool pose deviations, and machining trajectory offsets caused by variations in ram overhang and cutting load fluctuations during machining. The output of this method includes at least the trajectory correction amounts for each feed axis corresponding to the current interpolation cycle, as well as information such as amplitude limiting correction commands and feed intervention commands triggered under abnormal heavy load conditions. The entire process revolves around acquiring multi-source operational information, establishing the correlation between stiffness changes and pose deviations, calculating tool cutting point deviations and converting axial correction amounts, handling correction amount constraints and timing matching, loading trajectory correction amounts, and switching control states under abnormal conditions. This ensures that intermediate results from the previous processing stage can be directly used as input for the next processing stage, thereby guaranteeing the continuity and closed-loop nature of the method chain in engineering implementation.

[0068] Specifically, such as Figure 2 As shown: After the machine tool completes power-on, homing, workpiece coordinate system setting, and tool compensation call, the CNC system continuously monitors the machining operation conditions. When the automatic machining mode is effective, the spindle speed is greater than 0, and the interpolation feed rate is greater than 0, the underlying high-frequency acquisition process synchronized with the CNC core interpolation cycle is initiated. A unified time reference is established through the distributed clock synchronization mechanism of industrial real-time Ethernet, enabling the CNC core, spindle driver, each feed axis driver, and position acquisition unit to output data under the same time reference. The basic sampling period can be selected as 250 microseconds, corresponding to a sampling frequency of 4kHz. This value matches the fast interpolation cycle commonly used in high-end CNC systems. Its function is to keep the sampling period consistent with the trajectory execution period, while covering the tracking requirements of rapid load fluctuations, rapid attitude changes, and slight position offsets during heavy-duty cutting, and controlling it within the processing capacity range of the industrial controller and real-time bus that can operate stably for a long time.

[0069] Within each sampling cycle, the following information is synchronously read: spindle drive torque current, actual spindle speed, absolute vertical position of the ram, absolute lateral position of the crossbeam, feedback position and actual speed of each feed axis involved in the current machining operation, as well as the current program segment number and interpolation point identifier. Tool installation parameters are loaded after tool change and spindle clamping confirmation, and remain valid during the current tool usage period. Tool installation parameters are reloaded when tool change, tool compensation number switching, or spindle clamping status changes. Tool installation parameters include at least the effective tool overhang length, tool head radius, tool principal cutting edge angle, cutting edge inclination angle, and spatial offset of the tool tip geometric center relative to the spindle reference point. To avoid data mismatch within the same sampling period caused by differences in the transmission sequence of data from different sources, the system adds timestamps, source identifiers, and validity identifiers to various types of data in the acquisition buffer. A sampling window is considered valid only if the timestamp difference between key data within the same sampling window is no greater than 1 / 10 of the basic sampling period and all validity identifiers are valid; otherwise, the sampling window is considered invalid and triggers supplementary sampling or rollback processing. When the basic sampling period is 250 microseconds, the upper limit of the timestamp difference can be selected as 25 microseconds. This value ensures that multi-source data corresponds to the same physical processing state, while allowing for slight jitter during bus transmission to prevent valid data from being mistakenly judged as mismatched.

[0070] To ensure that the spindle load information and position feedback information accurately reflect the current machining status, denoising and validity checks are performed on the original sampled values. Within a sliding time window consisting of multiple consecutive sampling periods, outlier removal and weighted smoothing are performed on the spindle torque current sequence, ram position sequence, and crossbeam position sequence. The sliding time window length can be selected from 8 to 24 sampling periods, with 16 sampling periods being preferred. When the basic sampling period is 250 microseconds, 16 sampling periods correspond to a 4-millisecond window length. This value is based on a comprehensive balance between the requirements for no-load noise suppression and dynamic response: if the window length is too small, the number of samples participating in outlier detection and smoothing calculation is insufficient, which can easily reduce the stability of denoising; if the window length is too large, additional hysteresis can be introduced, affecting the response speed of subsequent trajectory correction. By adopting the above-mentioned value range, it is possible to maintain the ability to track changes in machining status while suppressing instantaneous electromagnetic spikes, communication jitter, and mechanical micro-vibration interference. Outlier removal is judged based on the magnitude of deviation from the window value. The outlier judgment boundary is determined based on historical fluctuation data collected during the machine tool's no-load operation phase, and three times the corresponding signal standard deviation can be selected as the outlier judgment boundary. This value is derived from the baseline noise level under no-load conditions, and its function is to remove electromagnetic spikes, occasional code disturbances, and single-point sampling anomalies, while preserving as much effective fluctuation as possible caused by changes in the actual cutting state. The remaining sampled values ​​after outlier removal are weighted and summed according to the principle of time proximity. Sampled values ​​closer to the current sampling period are assigned higher weights, and sampled values ​​farther from the current sampling period are assigned lower weights, thus obtaining the smoothed value of the current period. The smoothed spindle torque current is used as a characterizing quantity of the current spindle drive state, and the smoothed ram vertical absolute position and crossbeam lateral absolute position are used as the basis for judging the ram overhang and crossbeam motion state.

[0071] The ram overhang is obtained by subtracting the reference position when the ram is fully retracted from the current absolute vertical position of the ram, ensuring that subsequent stiffness change judgments are based on the actual overhang length, rather than on the overall machine tool coordinates. The crossbeam motion state is determined by both the crossbeam position change and the crossbeam velocity. Within a continuous predetermined sampling window, when the cumulative crossbeam displacement does not exceed the position holding threshold and the crossbeam velocity is lower than the micro-motion threshold, the crossbeam is determined to be in a position holding state. When the cumulative crossbeam displacement exceeds the position holding threshold or the crossbeam velocity is higher than the micro-motion threshold, the crossbeam is determined to be in a motion transition state. The position holding threshold is determined by the position sensor resolution and the allowable drift under no-load thermal stability. The position holding threshold can be selected as 3 to 5 times the minimum resolution unit, and preferably not less than the allowable drift under no-load thermal stability. The above values ​​are based on... The position sensor output is easily affected by quantization errors near the minimum resolution unit. The crossbeam also experiences slight thermal drift during no-load thermal stabilization, and the machine tool structure itself exhibits minor vibrations. Setting the position holding threshold within the aforementioned range helps avoid misinterpreting quantization errors, thermal drift, and structural micro-vibrations as crossbeam motion. The micro-motion threshold is determined jointly based on the crossbeam axis's rated speed, the speed fluctuation range during no-load low-speed operation, and the servo's low-speed stability range. The micro-motion threshold can be selected from 0.5% to 1% of the crossbeam axis's rated speed. This selection is based on the fact that the crossbeam axis's speed feedback fluctuates slightly during low-speed holding. A threshold that is too low can amplify minor disturbances, while a threshold that is too high will delay the identification of the true motion transition state. Setting the micro-motion threshold within the aforementioned range helps distinguish between the position holding state and the true motion transition state.

[0072] To ensure the data acquisition link can still provide usable input for subsequent processing under abnormal operating conditions, the system continuously monitors the data update status. If the feedback message timestamp is not updated within any sampling period, or if the feedback position remains unchanged despite changes in the sliding ram command position, a reread is performed. If the feedback message timestamp is not updated for three consecutive sampling periods, it is determined that there is an instantaneous update anomaly in the current data link. Three sampling periods correspond to 0.75 milliseconds, a value used to identify data acquisition synchronization failures or message interruptions as early as possible without significantly amplifying occasional delay misjudgments. At this time, the corresponding data source is marked as abnormal, and the smoothed value of the previous valid period is temporarily used as a temporary alternative input. The anomaly log table records information such as device identifier, anomaly start time, current program segment number, and interpolation point identifier. If the anomaly lasts for more than 10 milliseconds, it is determined to be a continuous anomaly state. 10 milliseconds corresponds to 40 basic sampling cycles. This value is used to distinguish between instantaneous communication jitter and continuous data anomalies, and to reserve processing time for short-term bus retransmission, buffer recovery, and data link self-recovery. After entering a continuous anomaly state, the system suspends the generation of new trajectory correction values ​​and only maintains the trajectory correction state corresponding to the previous valid cycle. When a continuous anomaly state occurs and the current program segment is still in the cutting state and the feed rate is greater than 0, the manual takeover judgment process is initiated, and an anomaly prompt is output to the upper-level monitoring interface.

[0073] After unified timing alignment, outlier removal, and smoothing, a multi-source machining status data packet is formed. This data packet includes at least the current spindle drive status, ram overhang, crossbeam motion status, tool installation parameters, feedback information of each feed axis, current program segment number, interpolation point identifier, etc., and serves as the input for subsequently determining the correlation between changes in machine tool spatial stiffness and tool pose deviation under machining conditions.

[0074] Specifically, such as Figure 3 As shown: After receiving the multi-source machining status data packet generated above, the CNC system establishes the correlation between the changes in machine tool spatial stiffness and tool posture deviation under the current machining state based on the ram overhang and cutting load changes. The cutting load change does not rely on independent cutting force sensors for acquisition, but is characterized online based on the spindle smooth torque current, actual spindle speed, effective cutting radius of the tool, and tool angle parameters. Specifically, the equivalent cutting torque of the spindle in the current cycle is first calculated based on the spindle torque current, spindle motor torque constant, main drive transmission ratio, and mechanical efficiency, and then the equivalent load intensity at the tool cutting point is calculated in combination with the effective cutting radius of the tool. Using the above method, load characterization results consistent with the current cutting state can be obtained using the existing signals of the CNC system and spindle driver, thereby reducing the need for additional sensor configuration.

[0075] To avoid misjudging additional disturbances caused by active spindle speed regulation or acceleration / deceleration transitions as workpiece load changes, spindle angular velocity change information is synchronously introduced during load conversion. The spindle angular velocity change rate threshold is jointly determined based on the standard deviation of the angular velocity change rate during the spindle's no-load steady-speed operation, the maximum fluctuation envelope, and the stable range of the driver speed loop. It can be selected as 3 to 5 times the standard deviation of the angular velocity change rate in the no-load steady-speed state, or 1% to 3% of the change rate corresponding to the rated angular velocity. The former method is used to adapt to the noise levels of different spindle systems, while the latter method is used to provide an engineering-based alternative boundary when the drive parameters are known but the no-load statistical sample is insufficient. If the threshold value is too small, normal fluctuations during the spindle's own speed regulation process may be easily identified as load anomalies. If the value is too large, it will weaken the ability to distinguish changes in the actual cutting state. When the spindle angular velocity change rate exceeds the threshold, the current cycle is only used for state tracking and does not participate in the update of spatial stiffness correlation parameters, thereby avoiding interference from the spindle speed adjustment process on load judgment and subsequent pose correlation.

[0076] To identify transient changes in cutting load, the equivalent load intensity is differentially calculated for multiple consecutive sampling periods. The differential window can be selected as three consecutive sampling periods. When the base sampling period is 250 microseconds, the three sampling periods correspond to a 0.75 millisecond time window. This time scale can suppress the amplification effect of single-point sampling fluctuations on the differential results, and will not weaken the response speed to heavy load changes due to an excessively long window. It is suitable for identifying load jumps in heavy-load deep cavity machining scenarios. After differential calculation, the load change rate and load increment are obtained. When the load change rate is higher than the reference load change rate and the load increment is higher than the load increment threshold, it is determined that a load jump has occurred in the current machining state.

[0077] The reference load change rate is determined based on historical load fluctuation data of the current tool during the stable cutting phase in the uniform allowance range. It can be obtained by multiplying the sliding average value by a proportionality coefficient. The proportionality coefficient can be selected from 20% to 40%, with 30% being preferred. The value of this proportionality coefficient is based on the need to strike a balance between false alarm control and abrupt change recognition sensitivity: when the proportionality coefficient is below 20%, normal load fluctuations during the stable cutting phase are easily misjudged as load jumps; when the proportionality coefficient is above 40%, the real load surges caused by abrupt changes in allowance, material hard spots, or uneven microstructure are not easily identified in time; a value of 30% can better balance false alarm control and abrupt change recognition capabilities.

[0078] The load increment threshold is determined jointly based on the spindle's rated output torque, the tool's effective cutting radius, machine tool specifications, tool type, and workpiece material, and can be selected as 10% to 20% of the rated equivalent cutting load. The rated equivalent cutting load is calculated by converting the spindle's rated output torque and the tool's effective cutting radius. The 10% to 20% value is based on the following: when the proportion is below 10%, normal load fluctuations under stable cutting conditions are easily identified as abnormal impacts; when the proportion is above 20%, it reduces the sensitivity to identifying significant load jumps, making it difficult to capture sudden heavy load changes in a timely manner; for rated output torques close to 2... For typical large heavy-duty vertical machine tools with a cutting radius of 0.8m to 1.0m and a cutting radius of 25kN·m, the load increment threshold calculated in the above manner can be, for example, 3kN to 5kN. This range matches the common cutter head size, spindle capacity, and heavy-duty cutting conditions of heavy-duty vertical machine tools, and is used to distinguish between normal fluctuations under stable cutting conditions and significant impact loads caused by factors such as hard spots, sudden changes in allowance, and uneven structure inside the workpiece. The above threshold can be automatically retrieved from the process database based on tool specifications, material type, and cutting parameters, or it can be corrected by combining on-site test cutting data.

[0079] After obtaining the changes in ram overhang and cutting load, a spatial stiffness variation characterization is constructed under the current machining state. The spatial stiffness variation characterization is segmented based on the ratio of ram overhang to the reference support span. The reference support span is the effective support length of the ram within the crossbeam guide support area, which can be determined based on the machine tool structural dimensions, factory static stiffness test results, and assembly / calibration data. When the ratio of overhang to the reference support span is less than 0.7, it can be identified as the linear stiffness zone; when the ratio is between 0.7 and 1.0, it can be identified as the transition stiffness zone; when the ratio is greater than 1.0, it can be identified as the attenuated stiffness zone. Here, 1.0 represents the ram overhang reaching the reference support span. The geometric critical ratio at the span is used to characterize the critical state when the ram's overhang reaches the upper limit of the effective support length; 0.7 is determined jointly based on the endpoint of the linear segment of the overhang and deflection response curve in the machine tool's factory static stiffness test and the starting inflection point of stiffness nonlinear enhancement in the modal test, and is used to characterize the initial interval where stiffness changes from approximately linear to significant nonlinear change; for different specifications of machine tools, the specific value of 0.7 can be corrected according to the corresponding machine tool's overhang and deflection calibration curve; by adopting the above partitioning method, the influence of overhang changes on bending stiffness and torsional stiffness can be distinguished according to different stages, thereby improving the accuracy of spatial stiffness change characterization;

[0080] To characterize the spatial compliance features in different zones, a 6-DOF compliance matrix is ​​maintained in memory, including at least 3 translational and 3 rotational degrees of freedom. The main diagonal terms characterize stiffness variations in the principal directions, while the off-diagonal terms characterize cross-directional coupling. In the linear stiffness region, the off-diagonal terms can be initialized at 5% to 15% of the main diagonal terms to reflect weaker cross-influence. In the transition stiffness region, the off-diagonal terms increase proportionally with the overhang ratio. In the attenuation stiffness region, the correction factor for the off-diagonal terms can be increased to twice that of the linear stiffness region. The ratio is set at 4 times to enhance the characterization of cross-effects such as vertical offset caused by lateral force, tool axis yaw caused by linear thrust, and attitude rotation caused by tangential load. The ratio ranges of 5% to 15% and 2 to 4 times are determined jointly based on the results of machine tool static load test, modal test, and finite element calibration, and are used to reflect the relative magnitude relationship between the main direction compliance and the coupling compliance. If the ratio value is too small, it will lead to insufficient characterization of cross-directional posture deviation. If the ratio value is too large, it will amplify the cross-coupling under light load or small overhang conditions, thereby affecting the stability of the correlation results.

[0081] The update frequency of the compliance matrix is ​​determined jointly based on the overhang state and the load state. When the current period simultaneously satisfies the condition that the overhang is in the attenuation stiffness region and the load jump is valid, the period is marked as a strongly coupled state, and the compliance matrix is ​​switched to be refreshed every sampling period to reflect the rapid changes in spatial stiffness under dangerous working conditions. When only one of the conditions is met, the normal update frequency is maintained. The normal update frequency can be selected as once every 4 sampling periods. When the basic sampling period is 250 microseconds, the corresponding update period is 1 millisecond. Using a 1-millisecond update period can reduce the computational burden and parameter jitter caused by high-frequency updates in the stable cutting region, while maintaining the basic tracking ability of stiffness and load changes. Switching to periodic refresh after entering the strongly coupled state is beneficial to improving the real-time performance of the correlation relationship and avoiding correlation lag caused by using low-frequency update results under sudden heavy load conditions.

[0082] After characterizing the spatial stiffness change, the current tool cutting direction and tool spatial offset vector are aligned with the machine tool global coordinate system, forming a transformation relationship between the tool local coordinate system and the machine tool global coordinate system. Specifically, the X-axis displacement deviation trend, Y-axis displacement deviation trend, and Z-axis displacement deviation trend correspond to the three orthogonal translational directions in the machine tool global coordinate system, while the pitch deviation trend, yaw deviation trend, and roll deviation trend correspond to the attitude change trends around these three orthogonal directions. Subsequently, the current equivalent cutting load is mapped to the machine tool global coordinate system, forming a transformation relationship between the translational load and the rotational load. The generalized load vector composed of the loads is used to calculate the spatial translational deviation trend and attitude deviation trend at the current moment by operating the compliance matrix on the generalized load vector. The above results are used to characterize the offset trend of the tool pose under the combined action of the current stiffness change and load change, and are used for subsequent trajectory correction generation and execution boundary judgment. The associated results include at least the following fields: X-axis displacement deviation trend, Y-axis displacement deviation trend, Z-axis displacement deviation trend, pitch deviation trend, yaw deviation trend, roll deviation trend, compliance status label, load jump status label, and association validity label.

[0083] To ensure that the correlation results are within the machine tool's executable range and the process's allowable range, a physical reasonable boundary is set for each deviation trend. The upper limit of the single-cycle allowable change of the translation deviation trend is determined jointly based on the maximum allowable following error of each feed axis, the current cutting thickness variation tolerance, and the interpolation cycle length, and can be selected as 50% to 80% of the maximum allowable following error. When the value is lower than 50%, the effective deviation trend will be compressed too early, reducing the sensitivity of subsequent corrections. When the value is higher than 80%, the single-cycle deviation change may approach the servo's tolerable boundary, which is not conducive to the stability of subsequent execution. The upper limit of the allowable angle of the attitude deviation trend is calculated based on the effective tool overhang length, the tool head radius, and the allowable contour error. It can be calculated by dividing the allowable contour error by the effective tool overhang length, and if necessary, the end amplification effect can be checked in conjunction with the tool head radius. Using the above boundaries helps to match the deviation trend with the machine tool's execution capability and process tolerance, and avoids directly introducing instantaneous abnormal correlation results into subsequent trajectory corrections.

[0084] When any deviation trend calculated in a certain period exceeds the corresponding physical reasonable boundary, that period is marked as an abnormal correlation state. For abnormal correlation states, new correlation results are not directly output, but a tiered rollback process is performed. The first-level rollback calls the correlation results of the previous valid period and adds a decay factor. The decay factor can be selected from 0.8 to 0.95, with 0.9 being preferred. If the decay factor is too small, the correlation results will decay rapidly, which is not conducive to maintaining the continuity of the results. If the value is too large, it will amplify the inheritance effect of the abnormal period on the historical results. A value of 0.9 is preferred to balance continuity and stability. If the first-level rollback still cannot restore the result to the physical reasonable range, the second-level rollback process is performed, i.e., freezing. The current correlation update is performed, and the next 4 to 8 sampling periods are set as the observation interval, preferably 4 sampling periods; when the base sampling period is 250 microseconds, 4 sampling periods correspond to 1 millisecond; this time interval is used to allow short recovery time for the compliance matrix and load characterization, while avoiding excessive observation time from affecting the subsequent response speed; within the observation interval, the system continues to receive new overhang, load change status, beam motion status and tool installation parameters, but only performs validity checks and does not update the correlation results; if the data recovers to normal, the freeze is lifted and the correlation update is resumed; if the anomaly continues, the current program segment is registered as an abnormal segment, and the status flag limiting the correction range is passed to the subsequent processing process;

[0085] After the above processing, an association result dataset is formed; this dataset includes at least the aforementioned six-dimensional deviation trend, current compliance state label, load jump state label, association validity label, whether it is in a strong coupling state, whether it is in the observation interval, etc., and serves as the direct input for subsequent calculation of the spatial position deviation and attitude deviation of the tool cutting point.

[0086] Specifically, such as Figure 4As shown: After receiving the associated result dataset generated above, the system determines the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory, and generates the corresponding corrections for each feed axis. The theoretical trajectory is taken from the theoretical tool tip position sequence and theoretical tool axis direction sequence generated by the interpolator in the current CNC program segment, and the current deviation is taken from the six-dimensional deviation trend generated above. During processing, the spatial transfer relationship from the slide end reference point to the tool cutting point is first established based on the tool tip spatial offset, effective tool overhang length, and tool head radius in the tool installation parameters. Then, the spatial position deviation of the tool cutting point relative to the theoretical trajectory and the tool axis relative to the theoretical trajectory are calculated based on the six-dimensional deviation trend. The system considers the orientation deviations; spatial position deviations include at least the deviations in the X, Y, and Z directions, and the orientation deviations include at least the pitch deviation, yaw deviation, and roll deviation. Considering that the orientation deviations will be amplified at the outer edge of the tool after being transmitted through the long tool holder and large-diameter tool head, the orientation deviations are processed together with the tool size parameters during conversion. For example, pitch deviation can cause additional normal displacement at the lower edge of the tool head, yaw deviation can cause additional tangential displacement at the circumferential side edge of the tool, and roll deviation can cause uneven offset of the contact positions on both sides of the tool. The system superimposes the above additional displacements with the translational deviations to obtain the comprehensive spatial position deviation of the cutting point in the current cycle.

[0087] Since the overall spatial position deviation includes both slowly varying deviations and high-frequency disturbances, the system performs frequency band filtering on the continuous deviation sequence to avoid directly converting high-frequency spurious deviations caused by structural flutter or electrical noise into axis corrections. The system buffers the continuous deviation sequence within a fixed-length sliding window, which can be selected from 400 to 600 interpolation cycles, with 500 interpolation cycles being preferred. When the interpolation cycle is 250 microseconds, 500 interpolation cycles correspond to a 125-millisecond time window. This time length is determined jointly based on the machine tool's low-frequency deflection mode, feed axis following response characteristics, and the interpolation cycle. When the minimum frequency of interest is between 8Hz and 10Hz, the 125-millisecond time window can cover the main variation cycle near this frequency, thus satisfying the requirements for slowly varying deviations. The sample length required for deviation identification and frequency band decomposition should be carefully considered, while avoiding excessively long windows that could significantly weaken the dynamic response speed of subsequent corrections. For the comprehensive spatial position deviation and attitude deviation within the window, frequency band decomposition can be performed using methods such as bandpass filtering, fast Fourier transform combined with frequency band division, short-time spectrum analysis, or variational mode decomposition. The decomposition results should include at least the following information: low-frequency fundamental component, transient response component, resonance risk component, and noise component. Among these, the low-frequency fundamental component reflects the slowly varying deviation caused by changes in overhang and average load; the transient response component reflects the followable deviation caused by feed acceleration / deceleration and machining allowance fluctuations; the resonance risk component reflects dangerous deviations close to the machine tool structure's natural frequency; and the noise component reflects electromagnetic interference and communication jitter.

[0088] The risk frequency band is determined based on the natural frequencies obtained from the machine tool modal tests. A frequency range of 10% to 15% above and below the center value of the natural frequency can be selected as the risk frequency band. When the modal frequency is less affected by temperature rise, load, or overhang changes, the risk frequency band can be taken as 10% above and below the center value of the natural frequency. When the modal frequency drift is significant, the risk frequency band can be taken as 15% above and below the center value of the natural frequency. These values ​​are used to cover frequency shifts caused by modal frequency drift, changes in operating conditions, and measurement errors, while avoiding misjudging excessively wide frequency bands as resonance risk areas. If the center frequency of a certain deviation component falls into the risk frequency band, that deviation component will not participate in the generation of shaft correction.

[0089] If a certain deviation component has a high consistency with the phase change of the corresponding feed axis acceleration curve, then the deviation component is considered to be a stable response of the corresponding feed axis and participates in the generation of the correction amount with a high weight; if the phase consistency is low, it participates with a low weight; the phase consistency can be calculated based on the phase difference, correlation coefficient, normalized cross-correlation peak value, or waveform overlap between the deviation component and the corresponding feed axis acceleration sequence; the matching degree threshold is set based on the statistical results of feed axis no-load following test, constant load following test, and historical machining samples; when the matching degree is higher than 80%, it can be judged as high consistency, and the corresponding weight can be selected from 0.7 to 0.9; when the matching degree is lower than 50%, it can be judged as low consistency, and the corresponding weight can be selected from 0.1 to 0.3; matching degree When the error rate is between 50% and 80%, the weights are determined using a linear transition method. The weight range of 0.7 to 0.9 is determined based on the statistical results of the responses of stable-following deviation components in feed axis no-load following tests, constant-load following tests, and historical machining samples. This range is used to retain most of the effective deviation components while avoiding amplifying the correction amount to near the execution boundary. The weight range of 0.1 to 0.3 is determined based on the statistical results of the responses of unstable-following disturbance components. This range is used to suppress high-frequency disturbances and low-consistency deviation components, preventing them from directly participating in axis correction generation. The aforementioned matching degree thresholds and weight ranges are used to distinguish between deviation components that can be stably responded to by the feed axis and unstable-following disturbance components, thereby achieving a balance between correction effectiveness and execution stability.

[0090] After completing the frequency band selection, the system converts the effective spatial position deviation and attitude deviation into correction amounts for each feed axis. For translational deviations with corresponding physical feed axes, they are directly projected onto the corresponding feed axes according to the machine tool coordinate mapping relationship. For attitude deviations without independent attitude compensation axes, they are converted into combined correction amounts for two or more linear feed axes through the spatial mapping relationship of the tool cutting point. For example, pitch deviation can be converted into a combined displacement fine adjustment of the vertical axis and the forward and backward feed axes, yaw deviation can be converted into a combined displacement fine adjustment of the transverse axis and the forward and backward feed axes, and roll deviation is mapped into joint compensation of the transverse axis and the vertical axis according to the deviation trend of the contact positions on both sides of the tool head. By using the above conversion method, the tool attitude change can be uniformly converted into an axial fine adjustment amount that the machine tool can execute, thereby maintaining the ability to compensate for attitude deviations even in the absence of independent attitude compensation axes.

[0091] To ensure that the generated correction values ​​can be stably executed by the servo system, three types of constraint parameters are set for each axis correction value: single-cycle variation limit, cumulative variation limit, and recovery slope. The single-cycle variation limit is determined by the maximum allowable following error, maximum allowable acceleration, and interpolation cycle of the corresponding feed axis. It can be selected as 30% to 60% of the maximum allowable following error, or calculated by converting the maximum allowable acceleration and the interpolation cycle. This value is used to limit the abrupt change of the correction value within a single interpolation cycle, avoiding a sudden increase in the following error due to excessive jumps in the correction value. The cumulative variation limit is based on the allowable contour error of the current process segment, the tool entry tolerance, and the axis travel safety margin. The value is determined jointly and can be selected from 50% to 100% of the allowable contour error. This value is used to limit the total magnitude of the correction after multiple consecutive cycles are superimposed, so as to avoid the correction process encroaching on the allowable error band of the process or approaching the axis travel boundary. The recovery slope is determined jointly based on the maximum allowable acceleration, maximum allowable jump and servo response bandwidth of the corresponding feed axis. It can be determined according to the allowable reduction in correction within a unit interpolation cycle, preferably 10% to 30% of the upper limit of the single cycle change, or calculated by converting the maximum allowable acceleration and the interpolation cycle. This value is used to control the correction to return to zero at an executable speed after the deviation disappears, so as to avoid reverse sudden change or secondary impact.

[0092] If the correction amount for a certain axis exceeds the change limit in a single cycle, the excess portion is not directly discarded but written into a delayed release buffer and gradually released according to the recovery slope in subsequent interpolation cycles. The delayed release interval can be selected from 4 to 20 interpolation cycles, with the specific value determined based on the response bandwidth of the corresponding feed axis and the allowable delay of the current process. When the interpolation cycle is 250 microseconds, the delayed release interval corresponds to 1 to 5 milliseconds. This interval is used to avoid excessive single-cycle output while maintaining the continuity of correction. The delayed release process ends when the correction amount to be released in the buffer is cleared to zero or the current correction state is desaturated. If the cumulative change limit is exceeded, the current correction state is marked as saturated, and the saturation mark is passed to the subsequent trajectory correction execution process. If the correlation validity label in the previous correlation result is invalid, the system only retains the correction amount of the previous valid cycle and does not generate new correction amounts based on the current invalid correlation result to avoid the abnormal correlation result from continuing to amplify.

[0093] After the above processing, a sequence of original correction values ​​for each feed axis under the current interpolation cycle is formed. This result includes at least the original correction value for each feed axis, deviation validity label, frequency band status label, saturation status label, and delayed release buffer status, and serves as the direct input for subsequent execution trajectory correction and constraint coordination.

[0094] Specifically, such as Figure 5 As shown: The system performs constraint processing and timing matching on the original correction values ​​of each feed axis formed above to generate the trajectory correction value corresponding to the current interpolation command; when the current interpolation cycle has completed the generation of the original correction value and the current program segment is in a machining state that allows compensation, the trajectory correction processing is started; if the current program segment has been marked as an abnormal segment, tool change segment, idle segment, or rapid positioning segment, the trajectory correction value will not be output in this cycle, only the status field and cache status will be updated; using the above triggering conditions helps to limit the trajectory correction action to the actual cutting machining range and avoids introducing invalid intervention in the non-cutting range;

[0095] The constraint handling process first performs boundary checks on the physical capabilities of each feed axis. The boundary checks are based on data including at least the axis driver's rated maximum speed, rated maximum acceleration, rated maximum jerk, maximum permissible following error, travel margin near the current position, and the proportion of the axis speed and acceleration capabilities currently occupied by the theoretical trajectory. The system converts the original correction amount for this cycle into additional speed and acceleration requirements, and compares them with the corresponding axis's upper limit for corrected speed and upper limit for permissible acceleration. The upper limit for corrected speed is determined based on the axis's rated maximum speed, the proportion of the current theoretical trajectory speed occupied, and the servo margin, and can be selected from 10% to 30% of the axis's rated maximum speed. The value of % is determined based on the fact that the base trajectory usually occupies the main speed capability, and the correction action can only use the remaining dynamic margin. When the value is below 10%, the correction capability is insufficient and it is difficult to effectively eliminate the deviation. When the value is above 30%, it is easy to squeeze the base interpolation speed and increase the risk of servo fluctuation. The upper limit of allowable acceleration is determined by the axis's rated maximum acceleration, the acceleration requirement of the current interpolation segment, and the servo dynamic margin. It can be selected from 10% to 40% of the axis's rated maximum acceleration. This range is used to limit the proportion of servo dynamic capability occupied by the correction action. If the value is too small, it will cause the correction response to be sluggish. If the value is too large, it may cause overcurrent, amplification of following error, or mechanical shock.

[0096] When the additional speed demand exceeds the upper limit of the correction speed, the system compresses the current original correction amount proportionally according to the ratio of the additional speed demand to the upper limit of the correction speed. When the additional acceleration demand exceeds the upper limit of the allowable acceleration, the system implements slope limitation based on the upper limit of the allowable acceleration and the upper limit of the allowable jump, so that the correction amount smoothly approaches the target value within multiple interpolation cycles. If the current position is less than the travel boundary distance, the correction amount pointing towards the boundary is unidirectionally reduced. The safety boundary distance is determined jointly based on the maximum braking distance of the corresponding axis, the maximum allowable following error, the interpolation step size near the current position, and the reserved margin before the mechanical limit. It can be selected as 1.2 to 2 times the sum of the maximum braking distance and the maximum allowable following error, and preferably not less than the sum of the maximum braking distance and the maximum allowable following error. This value is used to ensure that the axis still has sufficient braking and buffer space during sudden deceleration, continuous accumulation of correction amount, or state switching, thereby avoiding the risk of overtravel.

[0097] For the original correction amount that is already saturated, the system switches the correction strategy from correcting all directions to prioritizing the correction of critical directions. The critical direction is determined by the contour error sensitive direction, assembly datum control direction, and normal direction of the machined surface in the current process segment. Priority can be given to the normal intrusion direction, the main feed following direction, and the radial direction that affects the coaxiality of the hole system. The correction amount of non-critical directions can be compressed to 20% to 50% of the original value in the saturated state, or kept frozen for several consecutive cycles. This compression range is determined based on the process error sensitivity ranking, shaft dynamic margin, and current machined surface quality requirements. It is used to prioritize the control effect of the main error directions when the correction capability is limited.

[0098] After completing the physical capability constraints, the system performs timing matching on the correction values ​​for each axis to resolve the synchronization issue between the correction value generation time and the interpolation command execution time. The system adds an interpolation point number and time index to each theoretical point in the current interpolation queue and establishes a corresponding correction point index in the correction queue. The correction value is aligned to the interpolation point to be executed according to the interpolation point number, rather than the interpolation point that has already been executed. When it is detected that the correction queue lags behind the interpolation queue by more than one interpolation cycle, it is determined that there is a timing mismatch between the current correction value and the execution point. When the basic interpolation cycle is 250 microseconds, this threshold corresponds to a timing deviation of 250 microseconds, which is used to identify the same-cycle mismatch between the correction value and the actual execution point. At this time, if the original correction values ​​of the current cycle and the next cycle are both valid and the correction directions of the two cycles are consistent, then forward matching is performed; if the correction directions of the two cycles are inconsistent, forward matching is not performed, and a smooth transition process is initiated.

[0099] Forward matching does not simply copy the correction amount of the next cycle, but performs linear estimation based on the difference between the original correction amount of the current cycle and the next cycle, making the current output trajectory correction amount closer to the actual deviation state that is about to occur. If the correction queue is detected to be ahead of the interpolation queue by more than one interpolation cycle, timing delay alignment is performed, that is, the correction amount generated ahead is temporarily stored in the timing buffer and then superimposed and output when the corresponding interpolation point arrives. By adopting two processing methods, forward matching and timing delay alignment, the consistency between the correction amount and the actual execution point can be improved without changing the generation order of the basic interpolation trajectory, thereby reducing the risk of delayed correction or overcompensation caused by timing misalignment.

[0100] For continuous curve interpolation segments, the system further checks the consistency of correction directions between adjacent interpolation points. If the correction directions of adjacent points are opposite, and the correction amplitudes of both points exceed 50% to 80% of the single-cycle change limit, an overshoot risk is identified. This value is based on the fact that when adjacent correction amounts simultaneously reach more than half of the single-cycle change limit and their directions reverse, a local reversal trend is more likely to form on the trajectory. By controlling the upper limit of the threshold below 80%, a smooth transition can be performed in advance before overshoot occurs. For the connection between linear interpolation segments and circular interpolation segments, the system establishes a transition segment marker and increases the weight of tangential continuity constraints. The tangential continuity weight of the transition segment can be selected as 1.2 to 2 times that of the ordinary interpolation segment. This value is determined based on the sensitivity of tool marks at the transition between linear and circular segments, the magnitude of interpolation direction changes, and the requirements for the continuity of the machined surface contour. When the value is less than 1.2 times, the continuity enhancement of the transition segment is insufficient; when the value is greater than 2 times, it is easy to cause excessive suppression of corrections in other directions.

[0101] After completing constraint processing and timing matching, the system superimposes the correction values ​​of each axis with the current theoretical interpolation displacement to form the trajectory correction value. The trajectory correction value includes at least the interpolation point index, time index, position correction value of each axis, velocity feedforward correction value of each axis, correction status field, correction saturation status field, recovery slope field, and buffer release ratio field. If there is a timing buffer or amplitude buffer in this cycle, the corresponding buffer release ratio is recorded in the trajectory correction value. If only the previous valid correction value is retained in this cycle due to an anomaly, the correction status field is set to "retain". If the correction value is output after compression in this cycle, the correction status field is set to "limited". If the complete correction value is output in this cycle, the correction status field is set to "normal". The above status fields facilitate the direct identification of the generation method and constraint status of the current correction value in subsequent execution processes.

[0102] If the correction amount of a certain axis exceeds the change limit in a single cycle, the excess portion is not directly discarded, but written into the amplitude buffer and gradually released according to the recovery slope in subsequent interpolation cycles. The amplitude release interval can be selected from 4 to 20 interpolation cycles, and the specific value is determined according to the response bandwidth of the corresponding feed axis and the allowable delay of the current process. When the interpolation cycle is 250 microseconds, the amplitude release interval corresponds to 1 to 5 milliseconds. This interval is used to avoid excessive single-cycle output while maintaining the continuity of correction. When the interval is too short, the correction amount may still change abruptly. When the interval is too long, it will weaken the real-time performance of the correction response. When the correction amount to be released in the buffer is cleared to zero, or the current correction state is desaturated, the amplitude release process ends. If the cumulative change limit is exceeded, the current correction state is marked as saturated, and the saturation mark is passed to the subsequent trajectory correction execution process. If the correlation validity label in the previous correlation result is invalid, the system only retains the correction amount of the previous valid cycle and does not generate a new correction amount based on the current invalid correlation result, so as to avoid the abnormal correlation result from continuing to amplify.

[0103] When the correction status field remains in a held state for 8 to 40 consecutive interpolation cycles without returning to a normal or limited state, the system registers the current processing segment as a risk segment. When the basic interpolation cycle is 250 microseconds, the corresponding risk judgment time is 2 to 10 milliseconds. This range is determined based on process sensitivity, execution layer intervention response time limit, and short-term disturbance elimination requirements. For finishing or highly sensitive surfaces, 8 to 16 interpolation cycles can be used. For roughing or conditions with large disturbances, 20 to 40 interpolation cycles can be used. Through the above settings, short-term held states and continuous risk states can be distinguished. The final output can be directly loaded into the trajectory correction queue in the current interpolation execution process. This queue corresponds one-to-one with the interpolator's theoretical trajectory and serves as the direct input for subsequent trajectory execution and feed intervention judgment.

[0104] Specifically, such as Figure 6 As shown: The system loads the trajectory correction value generated earlier into the current machining path execution process, and switches to amplitude limiting correction state when the cutting load exceeds the threshold; the loading position of the trajectory correction value is located between the theoretical trajectory output by the CNC core interpolator and the axis position given by the servo controller, that is, the trajectory correction value is superimposed on the execution trajectory of the current cycle as position feedforward and speed feedforward, rather than achieving compensation by modifying the machining program code or post-coordinate offset; during loading, the system reads the corresponding item in the current trajectory correction value queue, superimposes the position correction value of each axis onto the theoretical position command, superimposes the speed feedforward correction value of each axis onto the theoretical speed feedforward, and writes the correction status field into the servo status record area in the same cycle; after servo execution, the system reads the actual position feedback and following error in the next sampling cycle to confirm the correction execution effect online; the theoretical compensation direction is jointly determined by the current position correction value and the speed feedforward correction value, and is used to characterize the direction in which the error should be reduced after the current correction effect;

[0105] If the actual following error is consistent with the theoretical compensation direction and the amplitude continues to decrease within 4 to 16 consecutive interpolation cycles, the current correction is deemed effective. When the basic interpolation cycle is 250 microseconds, the corresponding confirmation time is 1 to 4 milliseconds. This time scale matches the servo's short-time response time and the requirement for filtering out single-point disturbances, and is used to avoid misjudging the correction effect due to fluctuations in a single sampling point, while also preventing significant lag in the confirmation result. If the following error does not decrease but increases instead, the correction weight for the next cycle is reduced by 10% to 30% of the current correction weight, and the adjustment result is written to the correction status record field. When the error increase is small and the duration is short, the reduction can be 10%. When the error continues to increase or the following is significantly insufficient, the reduction can be 20% to 30%. The above reduction is determined based on the servo dynamic margin, the error sensitivity of the current process segment, and the correction response result of the previous cycle, and is used to suppress the continued amplification of ineffective corrections.

[0106] The cutting load threshold adopts a two-level structure. The first-level threshold is used to indicate the entry into the heavy load risk zone, and the second-level threshold is used to trigger forced limiting. The first-level threshold is determined based on the average stable cutting load under the current tool and workpiece material combination, and can be obtained by multiplying the average stable cutting load by 1.3 to 1.5. This coefficient range corresponds to the upper envelope of load fluctuations in the stable cutting stage, material fluctuations, and the normal load fluctuation caused by tool wear. When the machining state is relatively stable, a value of 1.3 can be used, and when material fluctuations are large or tool wear is significant, a value of 1.5 can be used. The second-level threshold is preferentially determined based on the equivalent cutting load corresponding to the rated allowable torque of the spindle. When the rated allowable torque cannot be directly referenced, it can be determined by multiplying the historical average stable load by 1.8 to 2.2. This range is used to distinguish between loads that can be passed through the spindle and those that cannot be passed through the spindle. The system handles moderate heavy loads and sustained heavy loads requiring forced suppression, with the amplitude limiting correction digesting the load. When the equivalent cutting load exceeds the first-level threshold but not the second-level threshold, the system switches to a pre-limited state. In the pre-limited state, critical direction correction is prioritized, non-critical direction correction is reduced, and the recovery slope is lowered to 50% to 80% of the normal recovery slope. This adjustment range is determined based on the current load exceeding the threshold, servo dynamic margin, and machining surface quality requirements. It is used to slow down the release speed of correction capability and avoid amplified correction fluctuations at the edge of heavy load. The critical direction is determined based on the contour error sensitive direction, assembly datum control direction, and machining surface normal direction of the current process segment. The normal intrusion direction, main feed following direction, and radial direction affecting the coaxiality of the hole system can be prioritized.

[0107] When the equivalent cutting load exceeds the secondary threshold, the system enters a strong limiting state. In this state, all correction amplitudes are immediately compressed, attitude-related corrections are frozen, and only translational basic corrections are retained. Simultaneously, a deceleration request is sent to the feed control channel. The deceleration request is determined in a tiered manner, with deceleration ratios of 10%, 20%, or 30%. When the equivalent cutting load is above the primary threshold but below the secondary threshold, the deceleration ratio can be 10%. When the equivalent cutting load exceeds the secondary threshold but the over-threshold ratio is no more than 10%, the deceleration ratio can be 20%. When the equivalent cutting load exceeds the secondary threshold by more than 10%, the deceleration ratio can be 30%. This tiered method is determined based on the degree of over-threshold, the spindle dynamic margin, and the current process segment's ability to withstand changes in the cut-off rate. It is used to suppress cutting impact while avoiding unnecessary machining cycle time losses.

[0108] To distinguish between instantaneous impact loads and continuous heavy load conditions, the system sets up a heavy load observation window. The heavy load observation window can select 40 to 200 interpolation cycles. When the basic interpolation cycle is 250 microseconds, the corresponding observation time is 10 to 50 milliseconds. This time range is adapted to the spindle torque filtering time scale, the duration of the cutting load, and the impact continuity characteristics under heavy-load cutting scenarios of heavy vertical machine tools, in order to avoid misjudging a single impact as continuous overload. If the equivalent cutting load is continuously higher than the secondary threshold within the heavy load observation window, and the spindle torque change rate is continuously greater than 0 for 4 to 16 consecutive sampling cycles, or continuously higher than the reference load change rate determined above, then the spindle torque change rate is determined to still be in an upward state. The judgment interval of 4 to 16 consecutive sampling cycles is jointly determined based on the requirements for filtering out short-term fluctuations in spindle torque and the sensitivity for identifying continuous upward trends, in order to avoid misjudging an instantaneous increase at a single sampling point as continuous growth. When the above conditions are met, the system can mark the current program segment as a heavy load abnormal segment and enter the manual takeover judgment process.

[0109] Manual intervention employs a composite judgment method, considering not only the absolute value of the load but also the duration of correction saturation, servo follow-up error status, and the status of critical data sources. The system triggers a processing halt and requests manual confirmation only when the following conditions are simultaneously met: the equivalent cutting load remains consistently above the secondary threshold within the aforementioned heavy load observation window; the correction status remains saturated for 8 to 40 consecutive interpolation cycles; and the actual follow-up error remains consistently above the manual intervention judgment threshold. The judgment interval of 8 to 40 consecutive interpolation cycles follows the time scale set for the continuous risk status mentioned earlier, used to distinguish between transient saturation and continuous saturation. The manual intervention judgment threshold is determined based on the servo alarm threshold, the process allowable follow-up error, and the current machining surface accuracy requirements, and can be selected as 50% to 80% of the servo alarm threshold, provided that this threshold does not exceed the upper limit of the current process allowable follow-up error. These composite judgment conditions are used to avoid unnecessary downtime due to a single abnormal fluctuation, while ensuring timely manual intervention when continuous heavy load, correction failure, and execution anomalies occur simultaneously. If necessary, the abnormal status of critical data sources can also be considered in conjunction with the judgment.

[0110] Regarding the release of the amplitude limiting correction state, the system sets up a recovery judgment logic. The recovery judgment requires that the equivalent cutting load continuously falls back to below the first-level threshold, and that the load change rate is lower than the reference load change rate determined above within the recovery judgment window, while the servo following error recovers to the allowable range. The recovery judgment window can use the time scale of the aforementioned heavy load observation window, or it can be configured independently according to the same magnitude. The allowable range of servo following error follows the setting principle of the manual takeover judgment threshold, and preferably is not greater than the current process allowable following error. After the above conditions are met, the system does not immediately recover to full amplitude correction, but gradually releases the previously compressed correction capability according to the recovery slope. The recovery order can be selected as recovering the critical direction correction first, then the attitude-related correction, and finally the non-critical direction correction. Using this recovery order is beneficial to prioritize ensuring the control continuity of the main process error directions, while avoiding premature recovery of attitude-related corrections that cause secondary fluctuations. If the load rises again during the recovery process and exceeds the first-level threshold, the recovery process is terminated and the system returns to the pre-limited state.

[0111] To ensure subsequent verification and parameter tuning capabilities, the system synchronously establishes execution and rollback records. Each program segment records at least the trajectory correction loading start time, amplitude limiting state switching time, deceleration ratio, recovery start time, recovery completion time, current correction status field, and current saturation status field. If manual intervention occurs, the system further records the current program segment number, tool number, workpiece coordinate position, alarm reason, and manual confirmation time. These records are used to reproduce the actual machining control process and also for subsequent tuning of thresholds, observation windows, recovery slopes, and correction weights. After the above processing, the trajectory correction execution results and their status records are formed during the current machining path execution process, serving as the direct basis for subsequent trajectory execution and feed intervention judgments.

[0112] Example 2: Based on Example 1, the specific application process of a method for adjusting the machining path of a large vertical machine tool is further explained:

[0113] Specifically, such as Figure 7The example shown illustrates a scenario where a large vertical machine tool is used for heavy-duty, long-stroke, deep-cavity semi-finishing of a large cast steel annular shell. The machine tool is in automatic continuous machining mode, with the spindle driving a large-diameter tool head to perform interpolation along the workpiece's inner cavity contour. The ram continuously extends downwards during machining, and the crossbeam adjusts laterally according to program requirements. The workpiece blank allowance is unevenly distributed along the circumference, with significant abrupt changes in allowance in local areas. Before machining, the CNC system has completed zeroing, workpiece coordinate system setting, tool compensation activation, and tool installation parameter loading. The tool installation parameters include at least the tool tip space offset and the effective tool overhang. The parameters include length, tool head radius, tool principal cutting edge angle, and cutting edge inclination angle. The input information for this operation process includes at least the spindle drive status, ram overhang, crossbeam motion status, position feedback of each feed axis, theoretical trajectory data, and cutting load change information reflected by the spindle load. The output results include at least the correction amount of each feed axis, trajectory correction amount, limit status, deceleration request, and execution record. The processing objective of this operation process is to convert the dynamic pose deviation caused by the changes in ram overhang and cutting load fluctuations during the machining process into trajectory correction actions that can be executed within the bottom interpolation cycle.

[0114] After machining begins, the system acquires the spindle drive status, ram overhang, crossbeam motion status, tool installation parameters, and feedback information from each feed axis according to a unified timing sequence. Once the machine tool enters automatic machining mode, and the spindle speed and interpolation feed rate are both greater than 0, a low-level high-frequency acquisition process is triggered. The basic sampling period can be selected as 250 microseconds, corresponding to a sampling frequency of 4kHz. The CNC core, spindle driver, feed axis driver, and position acquisition unit share the same time reference via a distributed clock. Within each sampling period, the system synchronously reads information such as spindle torque current, actual spindle speed, ram vertical absolute position, crossbeam lateral absolute position, feedback position of each linked feed axis, actual speed of each linked feed axis, current program segment number, and interpolation point identifier. If the timestamp difference of key data within the same sampling window is no greater than 1 / 10 of the basic sampling period, the sampling window is considered valid. When the basic sampling period is 250 microseconds, the upper limit of the timestamp difference can be 25 microseconds; otherwise, supplementary sampling or rollback processing is performed.

[0115] Subsequently, within a sliding time window consisting of multiple consecutive sampling periods, the system performs outlier removal and weighted smoothing on the spindle torque current sequence, ram position sequence, and crossbeam position sequence; the window length can be 16 sampling periods, corresponding to 4 milliseconds; the ram overhang is obtained by subtracting the reference position when the ram is fully retracted from the current absolute position of the ram; the crossbeam motion state is determined by the cumulative displacement and velocity of the crossbeam; when the cumulative displacement of the crossbeam does not exceed the position holding threshold and the crossbeam velocity is lower than the micro-motion threshold, the crossbeam is determined to be in the position holding state; when the cumulative displacement of the crossbeam exceeds the position holding threshold, or the crossbeam velocity is higher than the micro-motion threshold, the crossbeam is determined to be in the motion transition state; after unified timing alignment, outlier removal, and smoothing, a multi-source machining state data packet is formed; this data packet includes at least the spindle drive state, ram overhang, crossbeam motion state, tool installation parameters, feedback information of each feed axis, current program segment number, interpolation point identifier, etc., and serves as the input for subsequent spatial stiffness change characterization;

[0116] After acquiring multi-source machining status data packets, the system establishes a correlation between changes in machine tool spatial stiffness and tool pose deviation under the current machining state based on ram overhang and cutting load changes. Cutting load changes do not rely on independent cutting force sensors but are calculated online based on spindle smoothing torque current, actual spindle speed, effective tool cutting radius, and tool angle parameters. Specifically, the system first calculates the equivalent cutting torque of the spindle in the current cycle based on spindle torque current, spindle motor torque constant, main drive transmission ratio, and mechanical efficiency, and then calculates the equivalent load intensity at the tool cutting point based on the effective tool cutting radius. To avoid... Additional disturbances caused by active spindle speed regulation affect load judgment, and the system synchronously incorporates spindle angular velocity change information. When the spindle angular velocity change rate exceeds a preset threshold, this period is only used for state tracking and does not participate in stiffness-related parameter updates. Subsequently, the system performs differential calculations on the equivalent load intensity for multiple consecutive sampling periods. For example, the differential window can be three consecutive sampling periods. When the base sampling period is 250 microseconds, the corresponding time window is 0.75 milliseconds, which is used to identify transient changes in cutting load. When the load change rate is higher than the reference load change rate and the load increment is higher than the load increment threshold, a load jump is determined to have occurred.

[0117] Meanwhile, the system partitions the spatial stiffness state based on the ratio of the ram overhang to the reference support span. When the ratio is less than 0.7, it is determined to be in the linear spatial stiffness zone; when the ratio is between 0.7 and 1.0, it is determined to be in the spatial stiffness transition zone; when the ratio is greater than 1.0, it is determined to be in the spatial stiffness decay zone. For example, if the ram overhang gradually increases from 620 mm to 870 mm, and the reference support span is 900 mm, the overhang ratio increases from 0.689 to 0.967, corresponding to the spatial stiffness state transitioning from the linear zone to the transition zone. The system maintains a six-degree-of-freedom flexibility matrix in memory. The main diagonal terms of the matrix are used to characterize the change in compliance in the main direction, while the off-diagonal terms are used to characterize cross-directional coupling. As the overhang ratio increases and load jumps occur, the off-diagonal terms are increased by a predetermined ratio to enhance the ability to characterize the cross effects such as vertical offset caused by lateral force and attitude yaw caused by tangential load. After the above calculations, the system obtains the tool pose deviation trend result, which includes at least the following fields: X-direction displacement deviation trend, Y-direction displacement deviation trend, Z-direction displacement deviation trend, pitch deviation trend, yaw deviation trend, roll deviation trend, compliance status label, load jump status label, and association validity label.

[0118] After obtaining the correlation results mentioned above, the system calculates the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory based on the tool installation parameters and the six-dimensional deviation trend, and generates the corresponding correction amount for each feed axis. Specifically, the spatial transfer relationship from the reference point at the end of the ram to the tool cutting point is first established, and then the spatial position deviation of the tool cutting point relative to the theoretical trajectory and the attitude deviation of the tool axis relative to the theoretical direction are converted according to the six-dimensional deviation trend. Since the tool head radius and the effective overhang length of the tool will amplify the attitude deviation, the pitch deviation, yaw deviation and roll deviation are further converted into the additional normal displacement of the lower edge of the tool head, the additional tangential displacement of the circumferential side edge, and the uneven offset of the contact positions on both sides of the tool, and are superimposed with the translation deviation to obtain the comprehensive spatial position deviation. For example, in a certain machining section, the system calculates that the X-direction displacement deviation trend is 0.018 mm, the Z-direction displacement deviation trend is 0.026 mm, and the additional normal displacement of the lower edge of the tool head corresponding to the pitch deviation trend is 0.012 mm. Then the comprehensive cutting point normal deviation is about 0.038 mm.

[0119] To avoid directly converting high-frequency flutter or noise components into axis correction values, the system performs frequency band decomposition on the integrated spatial position and attitude deviations within a fixed-length sliding window. The window length can be 500 interpolation cycles; when the interpolation cycle is 250 microseconds, the corresponding time window is 125 milliseconds. The decomposition results include at least the low-frequency fundamental component, transient response component, resonance risk component, and noise component. If the center frequency of a deviation component falls into the risk frequency band, that deviation component will not participate in the correction value generation. If a deviation component has a high consistency with the phase change of the corresponding feed axis acceleration curve, it will participate in the correction value generation with a higher weight; if the phase consistency is low, it will participate in the correction value generation with a lower weight. Through the above processing, while retaining effective slow-varying deviations and followable deviations, the interference of resonance risk components and noise components on the subsequent correction value generation can be suppressed.

[0120] Subsequently, the system maps effective spatial position and attitude deviations to the corresponding feed axes. For translational deviations with physically corresponding axes, they are directly projected onto the corresponding feed axes. For attitude deviations without independent attitude compensation axes, they are converted into combined corrections for two or more linear feed axes through the spatial mapping relationship of the tool cutting point. For example, pitch deviation can be converted into a combined displacement fine-tuning of the vertical axis and the forward / backward feed axis, and yaw deviation can be converted into a combined displacement fine-tuning of the transverse axis and the forward / backward axis. To ensure that the corrections are executable, the system applies single-cycle variation upper limit, cumulative variation upper limit, and recovery slope constraints to the corrections of each axis. If the correction of a certain axis exceeds the variation upper limit in a single cycle, the excess part is written into the delayed release buffer and gradually released according to the recovery slope in subsequent interpolation cycles. After the above processing, the original corrections for each feed axis corresponding to the current machining state are formed, providing input for the generation of subsequent trajectory corrections.

[0121] After obtaining the original correction values ​​for each feed axis, the system performs physical capability constraints and timing matching to form the trajectory correction value corresponding to the current interpolation command. The system first performs boundary checks on the original correction values ​​based on the rated maximum speed, rated maximum acceleration, rated maximum jerk, maximum allowable following error, stroke margin near the current position, and the proportion of the current theoretical trajectory's occupation of the axis speed and acceleration capabilities. If the additional speed requirement exceeds the correction speed limit, the original correction value is compressed proportionally. If the additional acceleration requirement exceeds the allowable acceleration limit, the correction process is smoothed by slope limiting. If the current position is less than the safe boundary distance from the stroke boundary, the correction value pointing towards the boundary direction is unidirectionally reduced. For the original correction value that is already saturated, the system further switches the correction strategy from correcting all directions to prioritizing correction in critical directions, prioritizing the retention of the normal intrusion direction, the main feed following direction, and the radial direction correction that affects the coaxiality of the hole system, so as to ensure that the limited correction capability is preferentially applied to the error-sensitive direction.

[0122] Subsequently, the system performs timing matching on the correction amount; the correction amount is aligned to the interpolation point to be executed according to the interpolation point number, rather than the interpolation point that has already been executed; if the correction queue is detected to be lagging behind the interpolation queue by more than one interpolation cycle, forward matching is performed when the correction direction of the current cycle and the next cycle are consistent; if the correction queue is detected to be ahead, timing delay alignment is performed, the correction amount generated ahead is temporarily stored in the timing buffer, and superimposed and output when the corresponding interpolation point arrives; for continuous curve interpolation segments, the system also checks the consistency of the correction direction between adjacent interpolation points; if the correction directions of adjacent points are opposite and the correction magnitudes are both large, a smooth transition is performed to avoid forming local backtracking on the trajectory; after completing the above processing, the trajectory correction amount is formed; the trajectory correction amount includes at least the interpolation point index, time index, position correction value of each axis, feedforward correction value of velocity of each axis, correction status field, correction saturation status field, recovery slope field, and buffer release ratio field.

[0123] After the trajectory correction is generated, the system loads it into the current machining path execution process and switches to amplitude limiting correction state when the cutting load exceeds the threshold. The trajectory correction is loaded between the interpolator's output theoretical trajectory and the servo controller's received axis position setting. That is, the trajectory correction is superimposed on the current cycle's execution trajectory as position feedforward and velocity feedforward, rather than being compensated by modifying the machining program code or post-processing coordinate offset. After servo execution, the system reads the actual position feedback and following error in the next sampling cycle to confirm the correction effect online. The theoretical compensation direction is determined by the current position correction value and velocity feedforward correction value, and is used to characterize the direction in which the error should decrease after the correction. If the actual following error is consistent with the theoretical compensation direction and the amplitude continues to decrease within 4 to 16 consecutive interpolation cycles, the current correction is deemed effective. If the following error does not decrease but increases instead, the correction weight for the next cycle is reduced by 10% to 30% of the current correction weight.

[0124] The cutting load threshold adopts a two-level structure. The first-level threshold is used to indicate the entry into the heavy load risk zone, and the second-level threshold is used to trigger forced amplitude limiting. If the equivalent cutting load exceeds the first-level threshold but does not reach the second-level threshold, the system enters a pre-limiting state. In this state, critical direction corrections are prioritized, non-critical direction corrections are reduced, and the recovery slope is lowered to 50% to 80% of the normal recovery slope. If the equivalent cutting load exceeds the second-level threshold, the system enters a strong amplitude limiting state. In this state, all correction amplitudes are immediately compressed, attitude-related corrections are frozen, only translational basic corrections are retained, and a deceleration request is sent to the feed control channel. For example, when the current stable average cutting load is 20kN, the first-level threshold can be 26kN to 3kN. 0kN; if the secondary threshold is determined as 2.0 times the historical average stable load, it can be taken as 40kN; when the local margin change causes the equivalent cutting load to rise to 28kN, the system switches to the pre-limiting state; when the equivalent cutting load further rises to 42kN and continues to be higher than the secondary threshold, the system enters the strong limiting state and sends a 20% or 30% deceleration request according to the over-threshold ratio; if the equivalent cutting load in the heavy load observation window continues to be higher than the secondary threshold, and the spindle torque change rate continues to rise, while the correction state is in saturation for several consecutive interpolation cycles, and the actual following error continues to be higher than the manual takeover judgment threshold, the system marks the current program segment as a heavy load abnormal segment and triggers the manual takeover judgment process;

[0125] If the load continuously falls below the first-level threshold, and the load change rate within the recovery judgment window is lower than the baseline load change rate, while the servo following error returns to the allowable range, the system gradually releases the previously compressed correction capability according to the recovery slope. The recovery order prioritizes critical direction correction, followed by attitude-related correction, and finally non-critical direction correction. Simultaneously, the system records information such as the trajectory correction loading start time, amplitude limiting state switching time, deceleration ratio, recovery start time, recovery completion time, current correction status field, and current saturation status field. If manual intervention occurs, the system further records the program segment number, tool number, and workpiece. Information such as coordinate position and alarm cause is processed as described above, and trajectory correction loading, amplitude limiting switching, recovery control, and status recording are completed. Through the above operation mode, the dynamic pose deviation caused by the change of slide overhang and cutting load fluctuation during heavy-duty long-stroke deep cavity machining or large-difference contour machining of large vertical machine tools can be transformed into trajectory correction actions that can be generated and loaded online. Constraint control, amplitude limiting control, recovery control, and backtracking control are performed on the trajectory correction actions, so that the tool cutting point continuously approaches the theoretical trajectory within the bottom interpolation cycle, thereby improving the trajectory consistency, form and position accuracy, and operation stability of the machining process.

[0126] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0127] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0128] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0129] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for adjusting the machining path of a large vertical machine tool, characterized in that, include: S1. During the automatic machining process of the machine tool, the spindle drive status, ram overhang, crossbeam motion status, tool installation parameters and feedback information of each feed axis are acquired in a unified time sequence. S2. Based on the changes in ram overhang and cutting load, determine the correlation between the changes in machine tool spatial stiffness and tool pose deviation under the current machining state; adjust the coupling parameters and update frequency of the six-degree-of-freedom compliance matrix in combination with the load state; construct a generalized load vector after aligning the tool cutting direction and spatial offset vector to the machine tool global coordinate system, obtain the six-dimensional deviation trend, and perform attenuation backoff and freeze observation processing on the correlation results that exceed the physical boundary. S3. Based on the correlation, determine the spatial position deviation and attitude deviation of the tool cutting point relative to the theoretical trajectory, and generate the corresponding correction amount for each feed axis; perform frequency band decomposition, risk band elimination and phase consistency weighted screening on the continuous deviation sequence; establish the spatial transfer relationship from the slide end reference point to the tool cutting point based on the tool tip spatial offset, the effective tool overhang length and the tool head radius; S4. Perform constraint processing and timing matching on the correction amount of each feed axis to obtain the trajectory correction amount corresponding to the current interpolation command; For theoretical interpolation points, establish interpolation point numbers and time indices, and corresponding correction point indices. Align the original correction values ​​to the interpolation point to be executed. When the correction queue lags behind the interpolation queue by more than one interpolation cycle, and the original correction values ​​of the current cycle and the next cycle are both valid and in the same direction, perform linear prediction based on the difference between the original correction values ​​of the two cycles. When the correction queue leads the interpolation queue by more than one interpolation cycle, write the corresponding correction value into the timing buffer and superimpose it when executing at the corresponding interpolation point. S5. Load the trajectory correction amount into the current machining path during execution, and switch to amplitude limiting correction state when the cutting load exceeds the preset threshold; superimpose the trajectory correction amount onto the execution trajectory in the form of position feedforward and speed feedforward; confirm the correction state and adjust the correction weight based on position feedback and following error; In the next sampling cycle after servo execution, the actual position feedback and following error are read. The theoretical compensation direction is determined by combining the current position correction value and the velocity feedforward correction value. The current correction state is confirmed based on the consistency between the following error and the theoretical compensation direction and the amplitude change within the continuous cycle. Maintain the current correction weight when the following error continues to decrease, and reduce the correction weight for the next cycle when the following error increases; The system switches between pre-limited and strong-limited states based on two levels of cutting load thresholds, and performs critical direction retention, non-critical direction compression, attitude-related correction freezing, feed deceleration, heavy load observation, manual takeover judgment, recovery judgment, recovery release, and status recording processing.

2. The method for adjusting the machining path of a large vertical machine tool according to claim 1, characterized in that, S1 includes: When the processing and operating conditions are met, a unified time reference is established based on a distributed clock, and operating data is collected synchronously according to the interpolation cycle. Combine the tool change status and spindle clamping status to load or update the tool installation parameters; The system performs verification of the sampling window, smoothing of the sampled data, conversion of overhang, determination of beam motion status, and monitoring of abnormal rereading, rollback, and continuous abnormality.

3. The method for adjusting the machining path of a large vertical machine tool according to claim 1, characterized in that, S2 includes: The equivalent cutting load is calculated based on the spindle torque current, spindle speed, transmission parameters, and tool parameters. Speed ​​regulation disturbances are screened out by combining changes in the spindle angular velocity; Load jumps are determined based on the load change rate and load increment; The stiffness state is determined based on the ratio of the ram overhang to the span of the reference support.

4. The method for adjusting the machining path of a large vertical machine tool according to claim 3, characterized in that, Speed ​​regulation disturbances are screened out by combining changes in spindle angular velocity, including: Information on spindle angular velocity variation is incorporated into the equivalent cutting load conversion process; The judgment boundary is determined based on the fluctuation range of the angular velocity change rate during the no-load steady-speed operation of the spindle and the stable range of the speed loop of the drive. Accordingly, the corresponding period is divided into speed regulation transition period and associated update period. The speed regulation transition period is used for state tracking, and the associated update period is used for updating load associated parameters.

5. The method for adjusting the machining path of a large vertical machine tool according to claim 1, characterized in that, S3 includes: The spatial transfer relationship of the cutting point is established based on the theoretical tool tip position sequence, the theoretical tool axis direction sequence, and tool installation parameters. The six-dimensional deviation trend is converted into a comprehensive spatial position deviation and attitude deviation. The filtered effective deviations are mapped to the correction amounts of each feed axis, and single-cycle constraints, cumulative constraints, recovery control, delayed release, saturation marking, associated invalidation retention, and status recording are performed on each feed axis correction amount.

6. The method for adjusting the machining path of a large vertical machine tool according to claim 5, characterized in that, The six-dimensional deviation trend is converted into a comprehensive spatial position deviation and attitude deviation, including: The six-dimensional deviation trend is converted into the spatial position deviation of the cutting point relative to the theoretical trajectory and the attitude deviation relative to the theoretical tool axis direction. The normal additional displacement, tangential additional displacement, and contact position offset corresponding to the attitude deviation are superimposed with the translational deviation.

7. The method for adjusting the machining path of a large vertical machine tool according to claim 1, characterized in that, S4 includes: When the program segment is in the compensation-allowed state, the original correction amount is compressed, slope limited, direction reduced, and key direction prioritized allocation is performed based on the physical capabilities, stroke boundaries, and theoretical trajectory occupancy status of each feed axis. Based on the interpolation point number and time index, perform forward matching or time-series delay alignment on the original correction amount; Smoothing is performed on adjacent reverse corrections and transition corrections, and these are superimposed to form a trajectory correction with status markers, buffer release markers, and risk segment markers.

Citation Information

Patent Citations

  • Trajectory regeneration compensation method of numerical control machine error

    CN101943896B

  • Swingable deflection deformation compensation device for ram

    CN102513882B

  • Compensation method for tool vector error generated by linear shaft rolling angle error of multi-shaft machine tool

    CN111077847A

  • Ram deflection compensation system and method based on double-lead-screw load change rate

    CN117733654A

  • Anti-overload control method and system for numerical control machine tool cutter

    CN121232706A