Precise compensation method of composite numerical control machine tool

Abstract

The invention discloses a precision compensation method of a composite numerical control machine tool, and particularly relates to the technical field of precision control of the numerical control machine tool, which comprises the following steps: synchronously acquiring temperature, vibration and force information under the driving of a unified clock through multiple types of sensors deployed at key nodes of the machine tool, and generating a synchronous multi-source sensing data set with aligned timestamps; inputting the data set into a multivariable coupling error model obtained through data driving training, and calculating and generating a comprehensive space volume error predicted value of a tool nose point of the tool; performing inverse calculation based on a machine tool kinematic chain model, and generating a multi-dimensional micro-compensation instruction set containing the compensation amount of each motion axis and the sequential relationship; and finally, dynamically writing the instruction set into a servo control ring of the numerical control system through a real-time data interface, and driving each motion shaft to execute synchronous compensation motion. According to the method, comprehensive compensation of multi-physics field coupling errors such as temperature, vibration and force load is achieved, and the machining precision and stability of the numerical control machine tool under complex working conditions are improved.

Description

Technical Field

[0001] This invention relates to the field of precision control technology for CNC machine tools, and more specifically, to a precision compensation method for composite CNC machine tools. Background Technology

[0002] As the core of high-end manufacturing equipment, the machining accuracy of composite CNC machine tools directly affects the quality of precision parts. Currently, common error compensation methods to improve the accuracy of CNC machine tools mainly include thermal error compensation, geometric error compensation, and cutting force compensation. Existing technologies mostly use single-type sensors for data acquisition, and predict and compensate for specific types of errors by establishing empirical formulas, regression models, or finite element models. For example, by placing temperature sensors in key parts of the machine tool, using regression models to predict thermal deformation, and reducing errors by offset compensation of specific axes; or by measuring geometric errors using a laser interferometer and performing reverse compensation based on an error table.

[0003] However, in practical use, it still has some shortcomings. It only compensates for a single error source independently and fails to fully consider the comprehensive impact of the coupling effect of multiple physical fields such as temperature, vibration, and force load on spatial volume error, resulting in limited compensation effect under complex working conditions. Sensor data is mostly acquired asynchronously, lacking a unified time reference, making it difficult to establish an accurate mapping relationship between multi-source data and spatial error. Traditional compensation models rely on mechanism modeling or empirical formulas, which have poor generalization ability and are complicated to debug, making it difficult to adapt to the dynamic changes in the machining process. The integration of existing compensation command generation and servo control links is not high, and the real-time performance and multi-axis synchronization of compensation actions are insufficient, which can easily introduce secondary errors or excite machine tool vibration. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a precision compensation method for composite CNC machine tools, which solves the problems mentioned in the background art through the following scheme.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a precision compensation method for a composite CNC machine tool, comprising: S1. Generate a synchronous multi-source sensor dataset: By deploying multiple types of sensors at key nodes of the machine tool, temperature, vibration and force information are collected synchronously under the drive of a unified clock, and a set of timestamp-aligned synchronous multi-source sensor datasets are generated. S2: Generate a comprehensive spatial volume error prediction value: Input the synchronous multi-source sensor dataset into the multivariate coupled error model obtained through data-driven training, calculate and generate a comprehensive spatial volume error prediction value that characterizes the accuracy drift of the tool tip. S3: Generate a multi-dimensional micro-compensation instruction set for each motion axis: The predicted value of the comprehensive spatial volume error is calculated in reverse based on the motion chain model of the machine tool, and a multi-dimensional micro-compensation instruction set containing the compensation amount and timing relationship of each motion axis is generated. S4: Execute compensation motion: The multi-dimensional micro-compensation instruction set is dynamically written into the servo control loop of the machine tool CNC system through a real-time data interface, driving each motion axis to perform the final compensation motion to offset the comprehensive spatial volume error.

[0006] Preferably, the synchronous multi-source sensor dataset includes: deploying temperature sensors, vibration sensors, and force sensors, all of which are connected to the central acquisition unit via shielded cables; employing a multi-channel synchronous data acquisition instrument with an internal high-stability crystal oscillator as the clock reference; after the CNC system triggers the acquisition, the data is stored in a structured database containing timestamps, sensor IDs, and values ​​according to a unified timestamp, and the acquisition continues until the processing is completed.

[0007] Preferably, the key nodes of the machine tool include: spindle system, feed system and structural components, and the sensor arrangement is determined based on finite element analysis or experimental modal analysis.

[0008] Preferably, the comprehensive spatial volume error prediction value includes: preprocessing according to the same process as the training stage of the multivariate coupled error model; inputting the preprocessed synchronous multi-source sensor dataset into the offline trained multivariate coupled error model deployed in the real-time computing unit of the machine tool control system in real time; and outputting the comprehensive spatial volume error prediction value characterizing the accuracy drift of the tool tip through the forward propagation calculation of the input data by the model.

[0009] Preferably, the multivariate coupled error model includes: a black box or gray box model that is trained on historical data using a machine learning algorithm and is capable of capturing the nonlinear mapping relationship between temperature, vibration, force load and spatial volume error; the model adopts a neural network structure, uses mean square error as the loss function during training, and optimizes the model parameters through a backpropagation algorithm.

[0010] Preferably, the kinematic chain model includes: a mathematical model based on multibody system theory and homogeneous transformation matrix, which describes the geometric and kinematic coupling relationship of each motion axis; the kinematic chain model is used to inversely decompose the spatial error of the tool tip into the compensation displacement of each motion axis.

[0011] Preferably, the multi-dimensional micro-compensation instruction set includes: a sequence of micro-step compensation amounts for each axis after discretization according to the interpolation cycle of the machine tool CNC system, and each instruction includes a timestamp and the compensation amount that each motion axis should execute at that moment; the instruction set is planned using trapezoidal or S-shaped speed curves to ensure synchronous start-up, speed matching, and synchronous termination of multiple axes.

[0012] Preferably, the real-time data interface includes: a real-time communication protocol based on industrial Ethernet, with a communication cycle synchronized with the servo cycle of the CNC system, and the communication cycle being [missing information]. to The bit error rate is lower than Periodic jitter is less than .

[0013] Preferably, the compensation motion includes: superimposing the micro-compensation command and the original position command of the CNC system in the servo control loop to generate the final target position command; and monitoring the tracking error of each axis and the motor torque in real time. If the tracking error exceeds the safety threshold or the torque saturates, the compensation pause mechanism is triggered.

[0014] The technical effects and advantages of this invention are as follows: Comprehensive dynamic compensation for multi-physics coupling errors has been achieved: by synchronously collecting multi-source sensor data such as temperature, vibration, and force load, and using a data-driven multivariate coupling error model for real-time prediction, the limitations of compensating only a single error source are overcome, and the comprehensive impact of multi-physics coupling on machine tool accuracy under complex working conditions is fully reflected and offset, thus improving the comprehensiveness and accuracy of compensation. It ensures the synchronization and mapping accuracy of multi-source data: a unified clock based on a high-stability crystal oscillator is used to drive all sensors to collect data synchronously, and data is aligned by a unified timestamp, which solves the problem of data timing disorder caused by asynchronous acquisition from the root. It enhances the adaptability and generalization ability of the compensation model: Based on machine learning algorithms, a data-driven black box or gray box model is constructed, which can autonomously learn the complex nonlinear mapping relationship between multiple variables and errors. This overcomes the shortcomings of traditional mechanism models, such as poor generalization ability and complex debugging, and makes the model more adaptable and maintains its accuracy. The real-time performance and multi-axis synchronization of the compensation execution are enhanced: the micro-compensation instruction set is seamlessly integrated and written into the servo control loop through a real-time data interface based on industrial Ethernet; the use of inverse calculation based on the kinematic chain model and micro-step instructions with speed planning ensures the synchronous start, speed matching and synchronous end of the multi-axis compensation motion, effectively avoiding secondary errors and structural vibration. An intelligent safety monitoring and fault-tolerance mechanism has been introduced: real-time monitoring of tracking error and motor torque of each axis, and setting dynamic safety thresholds. If the threshold is exceeded, a compensation pause mechanism is immediately triggered and the system is notified to issue an early warning, ensuring the safety and reliability of the compensation process and preventing equipment damage or further deterioration of processing quality. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0016] Figure 2 This is a schematic diagram of the synchronous multi-source sensor dataset generation structure of the present invention.

[0017] Figure 3 This is a schematic diagram of the structure for generating the comprehensive spatial volume error prediction value according to the present invention.

[0018] Figure 4 This is a schematic diagram of the multidimensional micro-compensation instruction generation and execution structure of the present invention. Detailed Implementation

[0019] 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.

[0020] refer to Figure 1 - Figure 4 The precision compensation method for a composite CNC machine tool shown includes: S1: Generate a synchronous multi-source sensor dataset: By deploying multiple types of sensors at key nodes of the machine tool, temperature, vibration and force information are collected synchronously under the drive of a unified clock, and a set of timestamp-aligned synchronous multi-source sensor datasets are generated. S2: Generate a comprehensive spatial volume error prediction value: Input the synchronous multi-source sensor dataset into the multivariate coupled error model obtained through data-driven training, calculate and generate a comprehensive spatial volume error prediction value that characterizes the accuracy drift of the tool tip. S3: Generate a multi-dimensional micro-compensation instruction set for each motion axis: The predicted value of the comprehensive spatial volume error is calculated in reverse based on the motion chain model of the machine tool, and a multi-dimensional micro-compensation instruction set containing the compensation amount and timing relationship of each motion axis is generated. S4: Execute compensation motion: The multi-dimensional micro-compensation instruction set is dynamically written into the servo control loop of the machine tool CNC system through a real-time data interface, driving each motion axis to perform the final compensation motion to offset the comprehensive spatial volume error.

[0021] S1: Sensor selection and placement: Temperature sensor: A PT100 platinum resistance temperature sensor is used, with a measurement range of 0-120℃ and an accuracy of ±0.1℃. A total of 12 measuring points are arranged at key heat sources and easily deformable points of the machine tool, including: 2 mounted on the front bearing housing of the spindle, 1 mounted on the spindle motor housing, 2 each mounted on the ball screw nut pairs of the XYZ axes, 1 each mounted on the servo motor housing of the AC axis, and 1 mounted on the machine tool bed base. Vibration sensor: A triaxial ICP accelerometer is used, with a frequency range of 0.5~5000Hz and a sensitivity of 100mV / g. A total of 8 measuring points are arranged at the main vibration source and structural response points, including: 1 on the spindle box, 1 on each of the X / Y / Z axis slides, 1 on each of the A / C axis rotary tables, and 1 on the tool holder. Force sensor: A miniature piezoelectric triaxial force sensor with a measurement range of ±5kN is adopted. It is installed inside the spindle broaching mechanism to indirectly monitor the changes in cutting force. At the same time, the current feedback signals of each axis motor provided by the servo driver of the CNC system are used as auxiliary observations of the cutting load after calibration.

[0022] All sensors are connected to the central acquisition unit via shielded cables to reduce signal interference.

[0023] Construction of a synchronous data acquisition system: A multi-channel synchronous data acquisition instrument is used; the acquisition instrument has a built-in high-precision constant current source and anti-aliasing filter. Configure a uniform sampling rate for all sensor channels to satisfy the Nyquist sampling theorem for the highest frequency component of the vibration signal, and fully acquire slowly varying signals such as temperature. The acquisition unit uses an internal high-stability crystal oscillator to provide a clock reference, ensuring that the sampling clocks of all channels are strictly synchronized, thus eliminating time deviations between channels at the hardware level.

[0024] Data collection and timestamp alignment: A dedicated data acquisition and triggering module is embedded in the CNC system of the machine tool. When the CNC program starts, this module sends a start trigger signal (TTL pulse) to the synchronous acquisition instrument. Upon receiving the trigger signal, the synchronous acquisition instrument immediately initiates synchronous parallel acquisition of all channels; Each acquired data sample point is marked with a high-precision timestamp generated by the acquisition instrument's clock; this timestamp is aligned with the absolute time of the machine tool's CNC system through the trigger moment. The acquisition program stores data from different physical quantities and channels into a structured database according to a unified timestamp; each data record contains a timestamp field, a sensor ID field, and a sensor value field. This results in a timestamp-aligned, synchronous multi-source sensor dataset.

[0025] Data acquisition must be continuous throughout the entire processing to ensure the continuity of the dataset.

[0026] It should be further explained that unified clock driving refers to providing a unified sampling clock signal to all sensors through a high-precision clock source, ensuring strict time alignment of multi-source data and avoiding error coupling caused by time deviations. Key nodes include machine tool spindle systems, feed systems, structural components, and other parts sensitive to thermal deformation and vibration response. Sensor placement needs to be based on finite element analysis or experimental modal analysis to determine the optimal measurement points.

[0027] S2: The synchronous multi-source sensor dataset generated in S1 is input into a pre-trained multivariate coupled error model obtained through data-driven training. Through real-time calculation by this model, a comprehensive spatial volume error prediction value is generated that can accurately characterize the accuracy drift of the machine tool tip point relative to its theoretical position during actual machining. This prediction value is a three-dimensional vector, typically represented as... , respectively representing the tool tip point in the machine tool coordinate system The deviation in three directions.

[0028] Construction and training preparation of a multivariate coupled error model: The multivariate coupled error model is the core, which adopts a data-driven approach and learns from a large amount of historical data through machine learning algorithms. This results in a black box or gray box model that can capture the complex nonlinear mapping relationship between multiple physical quantities such as temperature, vibration, and force load and the final spatial volume error.

[0029] The dataset required for model training is obtained through a series of carefully designed experiments on the machine tool, which need to cover the expected working range of the machine tool, including different ambient temperatures, spindle speeds, feed rates, depths of cut, widths of cut, and different combinations of tool and workpiece materials; Input the aforementioned synchronous multi-source sensor dataset; during the training phase, this dataset needs to be strictly aligned with high-precision actual error measurement data; High-precision measuring equipment is used to measure the actual positional deviation of the machine tool tip in three-dimensional space in real time. The measurement process must be synchronized with the acquisition of multi-source sensor data under a unified time reference to ensure that the sensor data at each moment corresponds to a true spatial volume error value. This value is the target value (label) for model training.

[0030] Preprocess the collected raw training data: Data cleaning: Remove outliers caused by sensor malfunctions or measurement interference; Normalization: Normalize sensor data of different physical dimensions and orders of magnitude to accelerate model convergence and improve training performance. Feature selection: Analyze the correlation between the signals of each sensor and the target error, and may eliminate redundant features or construct new comprehensive features to reduce model complexity and improve its generalization ability.

[0031] Offline training of multivariate coupled error models: After preparing a high-quality training dataset, the model is trained in an offline computing environment; The preprocessed dataset is divided into training, validation, and test sets in a certain ratio (70%:15%:15%). The training set is used to update the model weights; the validation set is used to monitor model performance, adjust hyperparameters, and detect overfitting early during training; and the test set is used to evaluate the generalization ability of the trained model. Loss function: Choose mean squared error or mean absolute error to measure the difference between the model's predicted value and the true label; Optimization algorithm: Optimizers such as Adam and SGD are used to iteratively update network weights through backpropagation algorithm to minimize the loss function; Iterative training continues until the model's performance on the validation set no longer shows significant improvement or reaches the preset number of iterations, while early stopping, regularization, and other techniques are used to prevent overfitting. After training, the final performance of the model is evaluated using a test set. Evaluation metrics include root mean square error and coefficient of determination. Only when the model accuracy meets the preset requirements is it saved as the final multivariate coupled error model for deployment in the online prediction phase.

[0032] The offline trained and validated multivariate coupled error model is integrated and deployed onto the real-time computing unit in the machine tool control system. During processing, the continuously generated, timestamp-aligned synchronous multi-source sensor data streams from S1 are fed into the deployed multivariate coupled error model in real time; before input, the real-time sensor data must be processed using the same preprocessing procedure as in the training phase. The model performs forward propagation calculations on the input real-time sensor data samples. Based on the learned complex mapping relationships, the neural network instantly calculates the predicted comprehensive spatial volume error value most likely to occur at the tool tip at the current moment. .

[0033] It should be further explained that the loss function, used to quantify the difference between the model's predicted values ​​and the true labels, is crucial in guiding the direction of model optimization. The predicted value of the comprehensive spatial volume error is a three-dimensional vector. The characteristic is that the mean squared error is used as the loss function. Its definition is as follows: in: The number of samples in a training batch; For the model to the first The spatial error vector predicted from each sample input; To obtain through high-precision measuring equipment, and the first Each sample inputs a strictly aligned true spatial error vector (i.e., label); Batch size S3: The generated comprehensive spatial volume error prediction value is reverse-calculated based on the machine tool's precise kinematic chain model, and converted into a multi-dimensional micro-compensation instruction set that can be directly executed by the CNC system, containing the compensation amount of each motion axis and its precise timing relationship.

[0034] Establish the geometric and kinematic model of the machine tool's motion chain: Based on the structural layout and motion axis configuration of the composite CNC machine tool, establish a model that includes all motion axes (such as...). linear axis and A complete kinematic chain model of the rotation axis; the model is mathematically described using multibody system theory and homogeneous transformation matrix, and the geometric and kinematic coupling relationships between each kinematic axis are clearly defined. For the aforementioned composite CNC machine tool, its kinematic chain model can be represented as: ,in Represents the total transformation matrix from the base to the tip. Indicates the first Each motion axis is in its current position The homogeneous transformation matrix under the given conditions; At the same time, establish the Jacobian matrix corresponding to this kinematic chain. This matrix represents the minute displacements of each motion axis. The small displacement of the tool tip in the machine tool coordinate system Mapping relationship between them:

[0035] Inverse calculation of the theoretical compensation amount for each motion axis: using the predicted value of the comprehensive spatial volume error. The target deviation that needs to be compensated for at the blade tip. Based on the aforementioned kinematic chain model, inverse kinematics calculations are performed. By solving the inverse kinematics or applying an error allocation algorithm, the spatial deviation of the tool tip is decomposed and allocated to each motion axis, and the compensation displacement required by each axis is calculated. This ensures that the spatial error can be accurately offset after the composite of the compensated motions of each axis. For kinematic chains with analytical inverse solutions, the compensation amounts for each axis are calculated directly using the inverse transformation formula; for complex coupled structures, numerical iterative methods (such as the Newton-Raphson method) are used to solve the following equations: in, Based on the current actual position of each axis The calculated Jacobian matrix. During the solution process, the geometric constraints and motion limits of the machine tool must be considered to ensure the accuracy of the calculated compensation displacement. Within the executable range of each axis.

[0036] Generate a multi-dimensional micro-compensation instruction set with synchronized timing: based on the interpolation cycle of the machine tool CNC system. In conjunction with the servo control cycle, the compensation displacement of each motion axis is adjusted. Discretize and schedule according to time series, and calculate the total compensation displacement. Decomposed into The sequence of microstep compensation amounts executed within each interpolation cycle Microstep compensation amount The size needs to be planned according to the dynamic response characteristics of the servo axis (such as maximum acceleration and jerk). It is usually optimized using a trapezoidal or S-shaped speed curve to avoid exciting machine tool structure vibration and ensure smooth movement. Simultaneously, the start time and duration of the compensation actions of each motion axis are strictly coordinated to ensure synchronization between multiple axes, generating a multi-dimensional micro-compensation instruction set that is strictly aligned with the absolute timestamp of the machine tool CNC system. This instruction set is a data queue, and each instruction in the queue contains: a timestamp. and each motion axis at that moment Microstep compensation amount to be performed .

[0037] Instruction set encapsulation and format conversion: The micro-compensation instruction set is encapsulated into a data format that can be recognized by the CNC system, ensuring that it can be seamlessly written into the servo control loop through the real-time data interface in S4.

[0038] It should be further clarified that the timing relationship is not simply ordered by timestamps, but must follow the principle of synchronous start-up to rate matching to synchronous end-up; synchronous start-up: all micro-compensation commands for motion axes must be initiated at the start of the same servo cycle, and the start signal is triggered by the global synchronization pulse of the CNC system to ensure that there is no time difference in multi-axis compensation; rate matching: the rate of change of micro-step compensation is adjusted according to the servo response speed of each axis, such as... The axis servo response time is The maximum change rate of compensation per microstep is , The axis servo response time is The maximum change rate of compensation per microstep is To avoid asynchronous compensation due to differences in response speed; Synchronization completion: all axis micro-compensation commands must be completed at the end of the same servo cycle. If an axis cannot be completed in a single cycle due to a large compensation amount, it needs to be decomposed into micro-step compensation amounts of multiple consecutive cycles to ensure that multi-axis compensation ends simultaneously.

[0039] S4: The multi-dimensional micro-compensation instruction set is dynamically written into the servo control loop of the machine tool CNC system through a real-time data interface to drive each motion axis to perform compensation motion.

[0040] The real-time data interface is started and configured to establish a stable communication connection with the servo control loop of the machine tool CNC system. This interface adopts a real-time communication protocol based on industrial Ethernet, and the communication protocol cycle is strictly synchronized with the CNC system servo cycle. The communication cycle is configured as follows: to This is to ensure the timeliness of compensation commands. After establishing a connection, a communication quality self-check is performed to confirm that the data transmission error rate is below [a certain threshold]. And the periodic jitter is less than This meets the requirements for high-reliability control.

[0041] The generated multidimensional micro-compensation instruction set is loaded into a real-time memory buffer, which is managed by a high-priority real-time task that maintains strict synchronization with the CNC system's interpolator cycle. At the start of each servo cycle, this task retrieves and extracts the compensation data points (including those for each motion axis) to be executed within the current cycle from the buffer based on the global timestamp provided by the CNC system. Microstep compensation amount of the axis ).

[0042] The extracted micro-compensation values ​​for each axis are dynamically written to the servo driver of the corresponding motion axis in digital or analog form through the real-time data interface; the writing mechanism is as follows: the compensation value is... ( The original position commands (representing each axis) output by the CNC system interpolator. The final target position command is generated by superimposing the commands within the feedforward channel or position loop of the servo control loop. : This fusion process is completed in real time within each cycle of the servo controller, ensuring that the compensation command is seamlessly superimposed on the original motion command without affecting the original planned trajectory and motion logic of the CNC system.

[0043] Each axis servo driver receives the final fused command and drives the servo motor to perform precise micro-compensation motion. During execution, the actual position feedback, tracking error, and motor torque of each axis are monitored in real time. If the tracking error of any axis exceeds a preset safety threshold, or the motor torque becomes saturated, the compensation pause mechanism is immediately triggered and reported to the CNC system to prevent system oscillation or overload, ensuring the safety and stability of the compensation process.

[0044] It should be further explained that the safety threshold is set as follows: If a tracking error of more than 10 micrometers occurs on a single axis, it is likely that the servo system has a serious malfunction. In this case, continuing to perform compensation will not only fail to correct the original spatial error, but may also exacerbate machining defects due to the superposition of errors, or even damage the tool and workpiece. If the tracking error reaches 1.5 times the compensation command itself, it means that the servo system has completely failed to execute the command effectively, and the compensation behavior has failed.

[0045] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. 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 precision compensation method for a composite CNC machine tool, characterized in that, include: S1. Generate a synchronous multi-source sensor dataset: By deploying multiple types of sensors at key nodes of the machine tool, temperature, vibration and force information are collected synchronously under the drive of a unified clock, and a set of timestamp-aligned synchronous multi-source sensor datasets are generated. S2: Generate a comprehensive spatial volume error prediction value: Input the synchronous multi-source sensor dataset into the multivariate coupled error model obtained through data-driven training, calculate and generate a comprehensive spatial volume error prediction value that characterizes the accuracy drift of the tool tip. S3: Generate a multi-dimensional micro-compensation instruction set for each motion axis: The predicted value of the comprehensive spatial volume error is calculated in reverse based on the motion chain model of the machine tool, and a multi-dimensional micro-compensation instruction set containing the compensation amount and timing relationship of each motion axis is generated. S4: Execute compensation motion: The multi-dimensional micro-compensation instruction set is dynamically written into the servo control loop of the machine tool CNC system through a real-time data interface, driving each motion axis to perform the final compensation motion to offset the comprehensive spatial volume error.

2. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The synchronous multi-source sensor dataset includes: temperature sensors, vibration sensors, and force sensors, all of which are connected to the central acquisition unit via shielded cables; a multi-channel synchronous data acquisition instrument is used, with an internal high-stability crystal oscillator as the clock reference; after the CNC system triggers the acquisition, the data is stored in a structured database containing timestamps, sensor IDs, and values ​​according to a unified timestamp, and the acquisition continues until the processing is completed.

3. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The key nodes of the machine tool include: spindle system, feed system and structural components, and the sensor arrangement is determined based on finite element analysis or experimental modal analysis.

4. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The comprehensive spatial volume error prediction value includes: preprocessing according to the same process as the training stage of the multivariate coupled error model; inputting the preprocessed synchronous multi-source sensor dataset into the offline trained multivariate coupled error model deployed in the real-time computing unit of the machine tool control system in real time; and outputting the comprehensive spatial volume error prediction value characterizing the accuracy drift of the tool tip through the forward propagation calculation of the input data by the model.

5. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The multivariate coupled error model includes: a black box or gray box model that captures the nonlinear mapping relationship between temperature, vibration, force load and spatial volume error, obtained by training historical data through machine learning algorithms; the model adopts a neural network structure, uses mean square error as the loss function during training, and optimizes the model parameters through backpropagation algorithm.

6. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The kinematic chain model includes: a mathematical model based on multibody system theory and homogeneous transformation matrix that describes the geometric and kinematic coupling relationship of each motion axis; the kinematic chain model is used to inversely decompose the spatial error of the tool tip into the compensation displacement of each motion axis.

7. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The multidimensional micro-compensation instruction set includes: a sequence of micro-step compensation amounts for each axis after discretization according to the interpolation cycle of the machine tool CNC system. Each instruction includes a timestamp and the compensation amount that each motion axis should execute at that moment. The instruction set is planned using trapezoidal or S-shaped speed curves to ensure synchronous start-up, speed matching, and synchronous termination of multiple axes.

8. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The real-time data interface includes: a real-time communication protocol based on industrial Ethernet, with a communication cycle synchronized with the servo cycle of the CNC system, and a communication cycle of [missing information]. to The bit error rate is lower than Periodic jitter is less than .

9. The precision compensation method for a composite CNC machine tool according to claim 1, characterized in that, The compensation motion includes: superimposing the micro-compensation command and the original position command of the CNC system in the servo control loop to generate the final target position command; real-time monitoring of the tracking error of each axis and the motor torque, and triggering the compensation pause mechanism if the tracking error exceeds the safety threshold or the torque saturates.

Images