A numerical control longitudinal cutting lathe thermal compensation dynamic self-adaptive control method and system

Abstract

The application belongs to the technical field of numerical control, and provides a numerical control longitudinal cutting lathe thermal compensation dynamic self-adaptive control method and system, which comprises the following steps: determining a rate difference critical range when a measurable offset occurs in a double-shaft butt joint position based on double-shaft historical thermal elongation rate data; when the rate difference continuously falls into the critical range, combining current thermal state parameters such as the speed of the main shaft, the cutting load and the cumulative thermal elongation, solving all feasible compensation value ordered pairs that satisfy the axial butt joint tolerance, the compensation change rate limitation and the direction constraint, forming an alternative set; analyzing historical persistent high-frequency chatter events, extracting the chatter pre-thermal elongation rate and the compensation value combination, and constructing a discriminant feature library containing positive and negative samples and derived features; comparing the alternative compensation ordered pairs with the feature library, eliminating the schemes that may induce chattering, and then selecting the most gentle compensation ordered pair through smoothness screening, so that the accurate triggering of thermal compensation, the active avoidance of chattering and the effective unification of smooth execution are realized.

Description

Technical Field

[0001] This invention belongs to the field of CNC technology, specifically a dynamic adaptive control method and system for thermal compensation of CNC longitudinal cutting lathe. Background Technology

[0002] In dual-spindle machining on CNC slitting lathes, the workpiece alignment accuracy directly affects the finished product quality. To eliminate positioning deviations caused by spindle thermal expansion, modern machine tools generally employ thermal compensation dynamic adaptive control. However, this technology faces more stringent challenges in dual-spindle scenarios than in single-spindle scenarios: due to differences in heat source distribution, heat dissipation conditions, and speed loads on both spindles, the thermal expansion rates of the Z1 and Z2 axes often differ significantly. Conventional thermal compensation strategies typically model each axis independently and output compensation values ​​separately. When the thermal deformation rates of the two axes are inconsistent, the compensation command for the Z2 axis is prone to dynamic conflict with the actual position of the Z1 axis. Even more problematic is that if the compensation algorithm fails to limit the rate of change of the output value, a step-like compensation mutation will occur instantaneously during sudden changes in operating conditions (such as speed jumps or coolant opening / closing). For the slender slide structure of a slitting lathe, this mutation is tantamount to an instantaneous force pulse, directly exciting the natural frequency of the mechanical system and causing continuous high-frequency chatter. On a microscopic level, the Z2 axis is in an unstable motion state of "jumping while advancing," causing periodic sawtooth-like vibration marks to form on the secondary clamping reference surface of the workpiece after docking. Once these vibration marks are generated, they cannot be eliminated in subsequent turning operations, resulting in out-of-tolerance coaxiality of the part or tool marks on the end face, rendering it unusable.

[0003] Therefore, the present invention provides a dynamic adaptive control method and system for thermal compensation of CNC longitudinal cutting lathe. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0005] The technical solution adopted by this invention to solve its technical problem is: a dynamic adaptive control method for thermal compensation of a CNC slitting lathe, comprising the following steps: Step S10: Based on the historical thermal elongation rate sampling data of the biaxial axis, the critical range of rate difference when a measurable offset occurs at the biaxial docking position is determined through statistical analysis. Step S20: Monitor the biaxial thermal elongation rate in real time. When the difference in biaxial thermal elongation rate reaches the critical range of rate difference, perform multi-objective solution based on the current thermal parameters and output an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. Step S30: Analyze persistent high-frequency flutter events in the historical records, extract the biaxial thermal elongation rate combination before each event and the ordered pairs of compensation values ​​performed, and construct a discriminative feature library of flutter-related compensation ordered pairs. Step S40: Compare the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values, filter out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter, and output the final dual-axis compensation ordered pairs through the smoothness optimal screening.

[0006] As a further aspect of the present invention: in step S10, the process of collecting historical thermal elongation rate sampling data is as follows: The timestamp, Z1 axis position value, and Z2 axis position value are collected at a fixed sampling period. The thermal elongation rate of the Z1 axis and the thermal elongation rate of the Z2 axis are calculated by dividing the position difference of adjacent sampling points by the sampling period, and then the difference in rates between the two axes is calculated.

[0007] As a further aspect of the present invention: in step S10, the process of statistically analyzing and determining the critical range of rate difference is as follows: Offset events that meet the conditions for measurable offset at the biaxial docking position are selected from historical thermal elongation rate sampling data. The conditions for measurable offset at the biaxial docking position are that the absolute value of the actual axial relative position deviation between the Z1 axis and the Z2 axis is greater than a preset threshold. For each offset event, extract all dual-axis rate differences within a fixed time window before the event occurs, and calculate the average rate difference within each window and the deviation between the maximum and minimum values ​​within the window as the fluctuation amplitude. Sort the average rate difference of all offset events from smallest to largest, and take the 5th percentile as the lower limit of the critical value. Sort the fluctuation amplitude of all offset events from smallest to largest, and take the 95th percentile as the upper limit of the critical value. The critical range of rate difference is obtained by constructing the upper and lower limits of the critical value.

[0008] As a further aspect of the present invention: in step S20, when monitoring the biaxial thermal elongation rate in real time, the current biaxial rate difference is calculated, and when the biaxial rate difference continuously falls within the rate difference critical range, multi-objective solution is triggered. The current thermal parameters include the current spindle speed, cutting load current, and the cumulative thermal expansion of the two axes from the start of this machining process to the current moment. The cumulative thermal expansion is calculated by the positional deviation between the current position and the initial position of the two axes.

[0009] As a further aspect of the present invention: in step S20, the process of performing multi-objective solution based on the current thermal parameters is as follows: Three constraints are set: axial fit tolerance, compensation value change rate limit, and compensation direction and rate difference sign opposite. A multi-objective particle swarm optimization algorithm is adopted. Within the preset range of Z1-axis compensation values ​​and Z2-axis compensation values, all combinations of Z1-axis compensation values ​​and Z2-axis compensation values ​​that simultaneously satisfy the three constraints are searched. Each combination is an ordered pair. All the ordered pairs obtained are integrated and saved as a set of ordered pairs of compensation values ​​that can satisfy the dual-axis axial docking constraints.

[0010] As a further aspect of the present invention: In step S30, a continuous high-frequency chatter event is defined as follows: when the root mean square value of the acceleration signal collected by the machine tool vibration sensor in the frequency band exceeds a preset threshold and reaches a preset duration, it is marked as a continuous high-frequency chatter event. For each high-frequency flutter event, extract the biaxial thermal elongation rate combination within a fixed time window before the event begins, as well as the ordered pair of compensation values ​​being executed at that time. The biaxial thermal elongation rate combination includes the Z1-axis thermal elongation rate and the Z2-axis thermal elongation rate, and the ordered pair of compensation values ​​being executed includes the Z1-axis compensation value and the Z2-axis compensation value being executed.

[0011] As a further aspect of the present invention: In S30, the process of constructing the discriminative feature library of flutter correlation compensation ordered pairs is as follows: Combine the Z1-axis thermal elongation rate, Z2-axis thermal elongation rate, Z1-axis compensation value, and Z2-axis compensation value into a quaternion; All quadruplets corresponding to persistent high-frequency flutter events are stored in the flutter positive example set. Processing periods without flutter, with an equal number of flutter events, are randomly selected from the historical records. Quadruplets within the same duration window are extracted and stored in the flutter negative example set. The two sets are merged, and a flutter label is added to each quadruplet. At the same time, the differences in thermal elongation rate and the differences in the combination of compensation values ​​in each quadruplet are calculated as derived features. Finally, it is determined that each record in the feature library contains the original quadruplet, the derived features, and the flutter label.

[0012] As a further aspect of the present invention: In S40, the process of selecting ordered pairs of compensation values ​​that satisfy the biaxial axial docking constraint and do not induce chatter is as follows: For each ordered pair in the set of ordered pairs of compensation values, the thermal elongation rates of the Z1 axis and Z2 axis monitored in real time at the current moment are combined to form the feature to be tested; the Euclidean distance between the feature to be tested and each quadruple in the discriminant feature library is calculated. If the Euclidean distance is less than a preset threshold, the corresponding ordered pair is removed, otherwise it is retained; after traversing all the features to be tested, the retained ordered pairs are summarized into a safe ordered pair set.

[0013] As a further aspect of the present invention: In S40, the process of outputting the final biaxially compensated ordered pairs through optimal smoothness screening is as follows: Based on Tikhonov regularization, a smoothness value is calculated for each ordered pair in the safe ordered pair set. The smoothness value is equal to the square of the difference between the current Z1 axis compensation value and the Z1 axis compensation value actually executed at the previous time step, plus the square of the difference between the current Z2 axis compensation value and the Z2 axis compensation value actually executed at the previous time step. Select the ordered pair with the smallest smoothness value as the final dual-axis compensation ordered pair; if there are multiple ordered pairs with the smallest smoothness value, select the ordered pair with the smallest sum of compensation absolute values; if there are still multiple ordered pairs, select any one of them as the final output.

[0014] A dynamic adaptive control system for thermal compensation of a CNC longitudinal cutting lathe includes the following modules: The difference threshold calibration module, based on the historical thermal elongation rate sampling data of the biaxial axis, determines the critical range of rate difference when a measurable offset occurs at the biaxial docking position through statistical analysis. The multi-objective compensation solution module monitors the biaxial thermal elongation rate in real time. When the difference in the biaxial thermal elongation rate reaches the critical range of the rate difference, it performs multi-objective solution based on the current thermal parameters and outputs an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. The flutter feature library module analyzes persistent high-frequency flutter events in historical records, extracts the biaxial thermal elongation rate combination and the ordered pairs of compensation values ​​performed before each event, and constructs a discriminative feature library of flutter-related compensation ordered pairs. The compensation optimization output module compares the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values. It then filters out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter. Finally, it outputs the dual-axis compensation ordered pairs through the smoothness optimization filter.

[0015] The beneficial effects of this invention are as follows: First, by statistically analyzing the average value and fluctuation amplitude of rate differences in historical offset events, an adaptive critical range is constructed using percentiles, avoiding frequent compensation triggers due to normal thermal fluctuations, while ensuring timely response when docking deviation is about to exceed tolerances, thus improving the accuracy and reliability of compensation triggering; Second, by introducing a multi-objective particle swarm optimization algorithm, under the constraints of axial docking tolerance, compensation rate of change limit, and direction constraint, an ordered set of all feasible compensation values ​​is output, rather than a single fixed solution, providing ample optimization space for subsequent safety and smoothness screening, and solving the problem that traditional methods cannot simultaneously satisfy accuracy, servo response, and... The problem of direction matching; third, by constructing a chatter correlation discrimination feature library (including positive and negative samples and derived features such as rate difference and compensation difference), compensation schemes that may induce high-frequency chatter in the range of 50Hz to 200Hz are actively identified and eliminated, thus realizing early avoidance of chatter risk and significantly improving the stability of the machining process; fourth, Tikhonov regularization smoothness screening is adopted to select the ordered pair with the smallest change in compensation value compared with the previous moment, and in the case of multiple solutions, the one with the smallest sum of absolute compensation values ​​is selected, which effectively avoids secondary interference caused by abrupt changes in compensation value to the machining surface quality and tool life, and ensures the smoothness and consistency of thermal compensation execution. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart of the steps of a dynamic adaptive control method for thermal compensation of a CNC slitting lathe as described in Embodiment 1 of the present invention; Figure 2 This is a logic judgment diagram of a dynamic adaptive control method for thermal compensation of a CNC longitudinal cutting lathe as described in Embodiment 2 of the present invention; Figure 3 This is a flowchart of a dynamic adaptive control system for thermal compensation of a CNC longitudinal cutting lathe, as described in Embodiment 3 of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0019] Example 1: Please refer to Figure 1-2 As shown in the embodiment of the present invention, a dynamic adaptive control method for thermal compensation of a CNC slitting lathe includes the following steps: Step S10: Based on the historical thermal elongation rate sampling data of the biaxial axis, the critical range of rate difference when a measurable offset occurs at the biaxial docking position is determined through statistical analysis. In step S10, the process of acquiring biaxial historical thermal elongation rate sampling data is as follows: In the machining state of a CNC longitudinal cutting lathe, the following data items are collected and recorded at a fixed sampling period (e.g., 0.1 seconds): timestamp, Z1 axis position value, and Z2 axis position value. For every two adjacent sampling points, the thermal expansion rate of the Z1 axis is calculated as (current Z1 position - previous Z1 position) / sampling period. Similarly, the thermal expansion rate of the Z2 axis is calculated. The difference between the two axes is calculated as Z1 axis thermal expansion rate - Z2 axis thermal expansion rate. Data from at least 100 complete machining cycles is collected continuously. For the collected data, outliers exceeding the mean ± 3 times the standard deviation are removed. The removed points are then filled using linear interpolation. The historical thermal expansion rate sampling data is output, including the timestamp, Z1 axis thermal expansion rate, Z2 axis thermal expansion rate, and difference between the two axes for each sampling moment. In step S10, the condition for a measurable offset in the dual-axis docking position is that the absolute value of the actual axial relative position deviation between the Z1 axis and the Z2 axis is greater than 3 micrometers. In step S10, the process of statistically analyzing and determining the critical range of rate difference when a measurable offset occurs at the biaxial docking position is as follows: Migration events that meet the measurable migration conditions of the biaxial docking position are selected from historical thermal elongation rate sampling data. For each migration event, all biaxial rate differences within the 10-second time window before the migration occur are extracted. The average value and fluctuation amplitude (the deviation between the maximum and minimum values ​​within the time window) of the biaxial rate differences within each time window are calculated. The average value of the biaxial rate differences of all migration events is sorted from smallest to largest, and the 5th percentile is taken as the lower limit of the critical value. The fluctuation amplitude of all migration events is sorted from smallest to largest, and the 95th percentile is taken as the upper limit of the critical value. The critical range of rate difference is constructed by the upper and lower limits of the critical values. It is understandable that the significance of step S10 is to: accurately define the critical range of rate difference when a measurable offset occurs at the biaxial docking position by statistically analyzing historical thermal elongation rate data, set an early warning threshold for the system, avoid frequent triggering of compensation due to normal thermal fluctuations, and ensure timely response when a significant docking deviation is about to occur, providing reliable triggering conditions for subsequent compensation control. Step S20: Monitor the biaxial thermal elongation rate in real time. When the difference in biaxial thermal elongation rate reaches the critical range of rate difference, perform multi-objective solution based on the current thermal parameters and output an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. In step S20, the current data collected in real time, namely the thermal elongation rate of the Z1 axis and the thermal elongation rate of the Z2 axis, are used to calculate the current difference between the two axes at a frequency of 0.1 seconds. When the difference between the two axes falls within the critical range of the rate difference for 2 consecutive seconds, multi-objective solution is triggered. In step S20, the current thermal parameters include the current spindle speed (in r / min), cutting load current (in A), and the cumulative thermal expansion of the two axes from the start of this machining operation to the current moment (calculated by the positional deviation between the current position and the initial position of the two axes; for example, if the initial position of axis Z1 is P_Z1_cold and the current position is P_Z1(t), then the cumulative thermal expansion is P_Z1(t) - P_Z1_cold)). In step S20, the process of performing multi-objective solution based on the current thermal parameters and outputting an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint is as follows: Three constraints are set: axial alignment tolerance (e.g., ±2μm), compensation value change rate limit (e.g., no more than 0.5μm per 0.1 seconds), and compensation direction and rate difference sign are opposite. A multi-objective particle swarm optimization algorithm is used to search for all combinations of (Z1 axis compensation value, Z2 axis compensation value) that simultaneously satisfy the three constraints within the Z1 compensation value range (e.g., -10μm to +10μm) and the Z2 compensation value range (e.g., -10μm to +10μm). These combinations form an ordered pair, where each ordered pair represents the amount of additional movement required for the Z1 axis and the Z2 axis. All the ordered pairs obtained from the solution are integrated and saved as a set of ordered pairs of compensation values ​​that can satisfy the biaxial axial docking constraint; It should be noted that the axial mating tolerance is set according to the drawing tolerance requirements or process specifications of the machined parts. For example, if the coaxiality of the two ends of the part is required to be ≤5μm, or the height of the tool mark on the end face is required to be ≤3μm, then the positional deviation during axial mating should be stricter than the above values, usually taken as 1 / 2 to 1 / 3; for example, if the coaxiality requirement is 5μm, then the mating tolerance is set to ±2μm; the compensation value change rate limit is set according to the machine tool mechanical structure and servo system design requirements. The compensation direction and the sign of the rate difference are opposite to those determined by the control target. For example, if the thermal elongation rate of the Z1 axis is greater than that of the Z2 axis (the difference is positive), it means that Z1 "expands" forward relative to Z2. At this time, the Z1 compensation value should be negative (retracted backward) or the Z2 compensation value should be positive (extended forward), or a combination of both, to offset the difference; the Z1 compensation value range (e.g., -10μm to +10μm) and the Z2 compensation value range (e.g., -10μm to +10μm) are based on the machine tool thermal deformation design limit requirements; It is understandable that the significance of step S20 is that when the difference in thermal elongation rate enters the critical range, combined with the current spindle speed, cutting load and cumulative thermal elongation and other thermal parameters, all feasible compensation schemes that meet the axial docking accuracy, compensation change rate limit and directional constraint are solved through multi-objective optimization. Step S20 solves the problem of how to compensate without exceeding the tolerance and outputs a set of candidate compensation ordered pairs, which reserves sufficient optimization space for subsequent screening. Step S30: Analyze persistent high-frequency flutter events in the historical records, extract the biaxial thermal elongation rate combination before each event and the ordered pairs of compensation values ​​performed, and construct a discriminative feature library of flutter-related compensation ordered pairs. In step S30, a continuous high-frequency chatter event is defined as follows: when the root mean square value of the acceleration signal collected by the machine tool vibration sensor in the frequency band of 50Hz to 200Hz exceeds a preset threshold (e.g., 0.5g) and the duration exceeds 0.5 seconds, it is marked as a continuous high-frequency chatter event. In step S30, for each continuous high-frequency flutter event, the following data within the 3-second time window before the start of the event is extracted: the combination of dual-axis thermal elongation rates and the ordered pair of compensation values ​​executed, wherein the combination of dual-axis thermal elongation rates includes: Z1-axis thermal elongation rate and Z2-axis thermal elongation rate, and the ordered pair of compensation values ​​executed includes the Z1-axis compensation value and Z2-axis compensation value being executed. In step S30, the process of constructing the discriminative feature library of flutter correlation compensation ordered pairs is as follows: The thermal elongation rate v1 of Z1 axis, the thermal elongation rate v2 of Z2 axis, the compensation value ΔZ1 of Z1 axis, and the compensation value ΔZ2 of Z2 axis are combined into a quaternion (v1, v2, ΔZ1, ΔZ2). Store the quadruplets corresponding to all persistent high-frequency flutter events into the flutter positive example set F_pos. Randomly select stable processing periods without flutter from the historical records that are equal in number to the number of persistent high-frequency flutter events, extract the quadruplets within the same duration window (3 seconds), and store them into the flutter negative example set F_neg. The set of positive flutter examples F_pos and the set of negative flutter examples F_neg are merged, and a label is added to each quadruple (1 indicates flutter, 0 indicates no flutter). At the same time, two derived features are calculated for each quadruple: the difference in thermal elongation rate Δv = v1-v2 and the difference in the combination of compensation values ​​Δ(ΔZ) = ΔZ1-ΔZ2. These features are stored in the discriminant feature library. Finally, each record in the discriminant feature library contains: the original quadruple, the derived features, and the flutter label. It is understandable that the significance of step S30 is: by analyzing historical flutter events, extracting the combination of thermal elongation rate and compensation value before the flutter occurred, and constructing a discriminative feature library containing positive and negative samples and derived features, step S30 is equivalent to establishing an experience risk memory library, enabling the system to identify which compensation schemes have induced high-frequency flutter, thereby actively avoiding it in subsequent screening and achieving the safety and stability of control. Step S40: Compare the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values, filter out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter, and output the final ordered pairs of dual-axis compensation values ​​through the optimal smoothness screening. In step S40, the process of selecting ordered pairs of compensation values ​​that satisfy the biaxial axial docking constraint and will not induce chatter is as follows: For each ordered pair in the set of ordered pairs of compensation values, the thermal elongation rate of the Z1 axis and the thermal elongation rate of the Z2 axis monitored in real time at the current moment are combined to form the feature to be measured. The feature to be measured includes the thermal elongation rate of the Z1 axis, the thermal elongation rate of the Z2 axis, the Z1 axis compensation value, and the Z2 axis compensation value. Based on any feature to be tested, calculate the Euclidean distance between the feature to be tested and each quadruple in the discriminant feature library. If the Euclidean distance is less than a preset threshold, the ordered pair corresponding to the feature to be tested is removed. Conversely, if the Euclidean distance is greater than or equal to the preset threshold, the ordered pair corresponding to the feature to be tested is retained. It should be noted that the preset threshold (Euclidean distance) is set as follows: take the 90th percentile of all Euclidean distances between all pairs of quadruples in the flutter positive example set; After traversing all the features to be tested and completing the elimination operation, the remaining ordered pairs are the ordered pairs of compensation values ​​that satisfy the biaxial axial docking constraint and will not induce chatter, and are summarized to obtain the safe ordered pair set. In step S40, the process of outputting the final biaxially compensated ordered pairs through optimal smoothness filtering is as follows: Based on Tikhonov regularization, calculate the smoothness value J for each ordered pair (ΔZ1, ΔZ2) in the set of safe ordered pairs: in, These are the Z1-axis compensation values ​​and Z2-axis compensation values ​​actually executed at the previous moment; For each ordered pair in the set of safe ordered pairs, calculate the smoothness value J, and select the ordered pair with the smallest smoothness value J as the final biaxial compensation ordered pair; If there is more than one ordered pair with the smallest smoothness value J, then the ordered pair with the smallest sum of compensation absolute values ​​(ΔZ1+ΔZ2) is selected as the final biaxial compensation ordered pair. If there are still multiple ordered pairs, then any one of them is selected as the final biaxial compensation ordered pair. Understandably, the significance of step S40 is as follows: First, the discriminative feature library is used to eliminate compensation schemes that may induce chatter. Then, from the remaining safe schemes, the smoothest change in the compensation ordered pair is selected for output through optimal smoothness screening (Tikhonov regularization). Step S40 integrates three objectives: docking accuracy, anti-chatter safety, and execution smoothness, ensuring that the final output compensation is both effective and stable, and avoiding secondary interference to the processing quality caused by abrupt compensation changes.

[0020] Example 2: Please refer to Figure 3As shown in the embodiment of the present invention, a dynamic adaptive control system for thermal compensation of a CNC slitting lathe includes the following modules: The difference threshold calibration module, based on the historical thermal elongation rate sampling data of the biaxial axis, determines the critical range of rate difference when a measurable offset occurs at the biaxial docking position through statistical analysis. The multi-objective compensation solution module monitors the biaxial thermal elongation rate in real time. When the difference in the biaxial thermal elongation rate reaches the critical range of the rate difference, it performs multi-objective solution based on the current thermal parameters and outputs an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. The flutter feature library module analyzes persistent high-frequency flutter events in historical records, extracts the biaxial thermal elongation rate combination and the ordered pairs of compensation values ​​performed before each event, and constructs a discriminative feature library of flutter-related compensation ordered pairs. The compensation optimization output module compares the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values. It then filters out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter. Finally, it outputs the dual-axis compensation ordered pairs through the smoothness optimization filter.

[0021] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic adaptive control method for thermal compensation of a CNC slitting lathe, characterized in that: Includes the following steps: Step S10: Based on the historical thermal elongation rate sampling data of the biaxial axis, the critical range of rate difference when a measurable offset occurs at the biaxial docking position is determined through statistical analysis. Step S20: Monitor the biaxial thermal elongation rate in real time. When the difference in biaxial thermal elongation rate reaches the critical range of rate difference, perform multi-objective solution based on the current thermal parameters and output an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. Step S30: Analyze persistent high-frequency flutter events in the historical records, extract the biaxial thermal elongation rate combination before each event and the ordered pairs of compensation values ​​performed, and construct a discriminative feature library of flutter-related compensation ordered pairs. Step S40: Compare the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values, filter out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter, and output the final dual-axis compensation ordered pairs through the smoothness optimal screening.

2. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 1, characterized in that: In step S10, the process of collecting historical thermal elongation rate sampling data is as follows: The timestamp, Z1 axis position value, and Z2 axis position value are collected at a fixed sampling period. The thermal elongation rate of the Z1 axis and the thermal elongation rate of the Z2 axis are calculated by dividing the position difference of adjacent sampling points by the sampling period, and then the difference in rates between the two axes is calculated.

3. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 2, characterized in that: In step S10, the process of statistically analyzing and determining the critical range of rate differences is as follows: Offset events that meet the conditions for measurable offset at the biaxial docking position are selected from historical thermal elongation rate sampling data. The conditions for measurable offset at the biaxial docking position are that the absolute value of the actual axial relative position deviation between the Z1 axis and the Z2 axis is greater than a preset threshold. For each offset event, extract all dual-axis rate differences within a fixed time window before the event occurs, and calculate the average rate difference within each window and the deviation between the maximum and minimum values ​​within the window as the fluctuation amplitude. Sort the average rate difference of all offset events from smallest to largest, and take the 5th percentile as the lower limit of the critical value. Sort the fluctuation amplitude of all offset events from smallest to largest, and take the 95th percentile as the upper limit of the critical value. The critical range of rate difference is obtained by constructing the upper and lower limits of the critical value.

4. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 1, characterized in that: In step S20, when the biaxial thermal elongation rate is monitored in real time, the current biaxial rate difference is calculated. When the biaxial rate difference continuously falls within the rate difference critical range, multi-objective solution is triggered. The current thermal parameters include the current spindle speed, cutting load current, and the cumulative thermal expansion of the two axes from the start of this machining process to the current moment. The cumulative thermal expansion is calculated by the positional deviation between the current position and the initial position of the two axes.

5. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 4, characterized in that: In step S20, the process of performing multi-objective solution based on the current thermal parameters is as follows: Three constraints are set: axial fit tolerance, compensation value change rate limit, and compensation direction and rate difference sign opposite. A multi-objective particle swarm optimization algorithm is adopted. Within the preset range of Z1-axis compensation values ​​and Z2-axis compensation values, all combinations of Z1-axis compensation values ​​and Z2-axis compensation values ​​that simultaneously satisfy the three constraints are searched. Each combination is an ordered pair. All the ordered pairs obtained are integrated and saved as a set of ordered pairs of compensation values ​​that can satisfy the dual-axis axial docking constraints.

6. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 1, characterized in that: In step S30, a continuous high-frequency chatter event is defined as follows: when the root mean square value of the acceleration signal collected by the machine tool vibration sensor in the frequency band exceeds a preset threshold and reaches a preset duration, it is marked as a continuous high-frequency chatter event. For each high-frequency flutter event, extract the biaxial thermal elongation rate combination within a fixed time window before the event begins, as well as the ordered pair of compensation values ​​being executed at that time. The biaxial thermal elongation rate combination includes the Z1-axis thermal elongation rate and the Z2-axis thermal elongation rate, and the ordered pair of compensation values ​​being executed includes the Z1-axis compensation value and the Z2-axis compensation value being executed.

7. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 6, characterized in that: In S30, the process of constructing the discriminative feature library of flutter correlation compensation ordered pairs is as follows: Combine the Z1-axis thermal elongation rate, Z2-axis thermal elongation rate, Z1-axis compensation value, and Z2-axis compensation value into a quaternion; All quadruplets corresponding to persistent high-frequency flutter events are stored in the flutter positive example set. Processing periods without flutter, with an equal number of flutter events, are randomly selected from the historical records. Quadruplets within the same duration window are extracted and stored in the flutter negative example set. The two sets are merged, and a flutter label is added to each quadruplet. At the same time, the differences in thermal elongation rate and the differences in the combination of compensation values ​​in each quadruplet are calculated as derived features. Finally, it is determined that each record in the feature library contains the original quadruplet, the derived features, and the flutter label.

8. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 1, characterized in that: In S40, the process of selecting ordered pairs of compensation values ​​that satisfy the biaxial axial docking constraint and do not induce chatter is as follows: For each ordered pair in the set of ordered pairs of compensation values, the thermal elongation rate of the Z1 axis and the thermal elongation rate of the Z2 axis monitored in real time at the current moment constitute the feature to be measured. Calculate the Euclidean distance between the feature to be tested and each quadruple in the discriminant feature library. If the Euclidean distance is less than a preset threshold, the corresponding ordered pair is removed; otherwise, it is retained. After traversing all the features to be tested, the retained ordered pairs are summarized into a safe ordered pair set.

9. The dynamic adaptive control method for thermal compensation of a CNC slitting lathe according to claim 8, characterized in that: In S40, the process of outputting the final biaxially compensated ordered pairs through optimal smoothness screening is as follows: Based on Tikhonov regularization, a smoothness value is calculated for each ordered pair in the safe ordered pair set. The smoothness value is equal to the square of the difference between the current Z1 axis compensation value and the Z1 axis compensation value actually executed at the previous time step, plus the square of the difference between the current Z2 axis compensation value and the Z2 axis compensation value actually executed at the previous time step. Select the ordered pair with the smallest smoothness value as the final dual-axis compensation ordered pair; if there are multiple ordered pairs with the smallest smoothness value, select the ordered pair with the smallest sum of compensation absolute values; if there are still multiple ordered pairs, select any one of them as the final output.

10. A dynamic adaptive control system for thermal compensation of a CNC slitting lathe, characterized in that: The system is used to execute any one of the control methods of claims 1-9 above, and includes the following modules: The difference threshold calibration module, based on the historical thermal elongation rate sampling data of the biaxial axis, determines the critical range of rate difference when a measurable offset occurs at the biaxial docking position through statistical analysis. The multi-objective compensation solution module monitors the biaxial thermal elongation rate in real time. When the difference in the biaxial thermal elongation rate reaches the critical range of the rate difference, it performs multi-objective solution based on the current thermal parameters and outputs an ordered set of compensation values ​​that can satisfy the biaxial axial docking constraint. The flutter feature library module analyzes persistent high-frequency flutter events in historical records, extracts the biaxial thermal elongation rate combination and the ordered pairs of compensation values ​​performed before each event, and constructs a discriminative feature library of flutter-related compensation ordered pairs. The compensation optimization output module compares the set of ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint with the discriminant feature library of ordered pairs of flutter-related compensation values. It then filters out ordered pairs of compensation values ​​that satisfy the dual-axis axial docking constraint and will not induce flutter. Finally, it outputs the dual-axis compensation ordered pairs through the smoothness optimization filter.

Images