Time-varying precision customized fine self-healing method based on virtual-real homeomorphism evolution

Abstract

The application discloses a time-varying precision customized fine self-recovery method based on virtual-real isomorphism evolution, relates to the technical field of machine tool precision control, and comprises the following steps: firstly, according to the positional relationship of each component of a machine tool, a homogeneous transformation matrix is constructed, and a unified mathematical model of errors is obtained based on a small error theory; secondly, a measuring device is used to obtain small errors of movement and rotation angle and reverse gap error of a translational axis; then, a virtual-real isomorphism digital twin platform is constructed, and a model and errors are integrated to realize virtual-real error isomorphism evolution; real-time sensing data is acquired, time series prediction is combined to evaluate precision time variation, a self-recovery path is generated by driving virtual-real synchronous evolution; finally, positioning deviation is determined according to error variation at the beginning and end of the path and the error model, and the path is transmitted when the error exceeds a preset value; in addition, the time lag defect of traditional real-time compensation is relieved through advanced prediction / deduction. The method realizes unified error modeling of a multi-axis machine tool, makes up for the defect of ignoring reverse gap, solves errors caused by compensation lag and blind compensation, and guarantees machining precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine tool precision control, in particular to a time-varying precision customized fine self-healing method based on virtual-real isomorphism evolution. BACKGROUND

[0002] In the machine tool machining process, key components such as spindles and tools are easily affected by factors such as thermal expansion and wear, which can cause dynamic errors and lead to a decrease in machining precision. At present, the industry mainly deals with this problem through two ways: error avoidance and error compensation. Error avoidance can alleviate the error formation rate through means such as thermal symmetry design and cooling pipe optimization, but it cannot completely eliminate errors and is difficult to cope with dynamic precision deficiency in continuous machining. Error compensation, which is the core means of time-varying precision control, offsets existing dynamic errors by artificially creating new errors. Precise real-time error prediction is the key.

[0003] Existing error prediction is mainly divided into two categories: mechanism-driven and data-driven. Mechanism-driven methods are based on complex boundary conditions such as bearing heat generation and heat conduction, and use finite element methods to simplify error analysis. However, they are not suitable for complex and variable machining processes. Data-driven methods are based on machine learning technology, using key component sensor data as input and error as output to establish a model. With the development of big data and artificial intelligence, they have become a research hotspot. However, data-driven error compensation still has significant defects in actual industrial scenarios: first, existing methods can only calculate the amount of dynamic error, and lack of error compensation strategy research. When the amount to be compensated is small, the positioning accuracy of the machine tool itself is insufficient, which can easily introduce more errors. Second, there is a time cost in the compensation process. Existing real-time prediction does not consider the advance of compensation time, which leads to lag in compensation, and there is no fine design of compensation step size, which can easily damage the surface quality of the workpiece. SUMMARY

[0004] The purpose of the present application is to provide a time-varying precision customized fine self-healing method based on virtual-real isomorphism evolution, which can realize accurate prediction and fine compensation of real-time error.

[0005] To achieve the above purpose, the present application provides the following solutions:

[0006] A time-varying precision customized fine self-healing method based on virtual-real isomorphism evolution, comprising the following steps:

[0007] According to the positional relationship between the bed of the numerical control machine tool, each motion axis, the tool and the workpiece, a homogeneous transformation matrix from the workpiece coordinate system to the tool coordinate system is constructed, and based on the small error theory, a unified mathematical model of machine tool error is obtained. The unified mathematical model of machine tool error covers three-axis, four-axis and five-axis numerical control machine tools.

[0008] The positioning accuracy of each motion axis of the numerical control machine tool is tested by using an accuracy measuring device to determine the micro error data of the numerical control machine tool; the micro error data includes the movement, rotation micro error of each motion axis at different positions, and the reverse gap error of each translational axis.

[0009] A virtual-real isomorphism digital twin platform is constructed, a unified mathematical model of machine tool error is integrated, and micro error data is stored in the platform database to perform isomorphic evolution of the geometric error state of the virtual digital twin and the physical numerical control machine tool.

[0010] Sensing data during the machining process of the numerical control machine tool is obtained, the time series prediction algorithm is combined to evaluate the machining precision time-varying condition during the machining process, and the synchronous evolution of the virtual and physical machine tool machining precision is driven to generate a precision self-healing path; the precision self-healing path is a set of motion trajectories and operation parameters for guiding the machine tool to perform precision compensation.

[0011] Based on the virtual-real isomorphism digital twin platform, the change of each motion axis micro error component at the beginning and end of the precision self-healing path is determined, and the positioning accuracy deviation during the self-healing process is determined based on the unified mathematical model of machine tool error. When the positioning accuracy deviation exceeds the preset precision self-healing value, the precision self-healing path is transmitted to the physical numerical control machine tool.

[0012] Optionally, the three-axis numerical control machine tool is a numerical control machine tool involving only three translational axes; the four-axis and five-axis numerical control machine tools are numerical control machine tools that, in addition to the X, Y, and Z three-direction translational axes, also include rotating axes A, B, and C rotating around the X / Y / Z axis movement direction; the unified mathematical model of machine tool error of the three-axis numerical control machine tool is a basic error mathematical model, which contains four types, and the specific type is determined by the position of the bed reference coordinate system R relative to the X, Y, and Z three-direction translational axis coordinate system, i.e. RXYZ type, XRYZ type, XYRZ type, and XYZR type.

[0013] Optionally, according to the positional relationship between the rotating axis and the translational axis in the workpiece-tool coordinate system, it is determined that the four-axis and five-axis numerical control machine tools respectively contain 24 and 48 types; among them, the mainstream types of the five-axis numerical control machine tool include 12 types: RXYZAB type, XRYZAB type, XYRZAB type, XYZRAB type, CARXYZ type, CAXRYZ type, CAXYRZ type, CAXYZR type, ARXYZB type, AXRYZB type, AXYRZB type, and AXYZRB type.

[0014] Optionally, taking the RXYZ type three-axis machine tool as an example, the homogeneous transformation matrix from the workpiece coordinate system to the tool coordinate system is as follows:

[0015]

[0016] wherein, is the worktable W relative to the toolT The homogeneous transformation matrix, For workbench W Compared to the bed frame R The homogeneous transformation matrix, bed frame R Relative to the translation axis X The homogeneous transformation matrix, Translation axis X Relative to the translation axis Y The homogeneous transformation matrix, Translation axis Y Relative to the translation axis Z The homogeneous transformation matrix, For translation axis Z Relative to spindle S The homogeneous transformation matrix, Main axis S Compared to cutting tools T The homogeneous transformation matrix, x , y and z These represent the ideal conditions of a CNC machine tool along... X , Y and Z The distance moved in a direction.

[0017] Optionally, the sensing data during the CNC machine tool machining process includes temperature sensing data, stress sensing data, vibration sensing data, acoustic emission sensing data, and real-time coordinate trajectory data of the machine tool cutting tool for key components of the CNC machine tool.

[0018] Optionally, the time-series prediction algorithm employs a long short-term memory network or a temporal convolutional network. By extracting the long and short-term temporal dependencies in multi-source sensor data such as temperature, stress, vibration, and acoustic emission, it mines the dynamic evolution trend of time-series data and combines attention mechanisms or causal convolution to enhance the expressive power of key features, thereby achieving accurate prediction of thermal errors and tool wear.

[0019] Optionally, based on the virtual-physical isomorphic digital twin platform, the changes in the minute error components of each motion axis at the beginning and end of the accuracy self-healing path are determined, and the positioning accuracy deviation during the self-healing process is determined in combination with the unified mathematical model of machine tool error. When the positioning accuracy deviation exceeds the preset accuracy self-healing value, the accuracy self-healing path is transmitted to the physical CNC machine tool, specifically including the following steps:

[0020] Record the spatial coordinates of the initial point and the end point on the precision self-healing path, and use the difference between the spatial coordinates of the initial point and the end point on the precision self-healing path as the preset precision self-healing value.

[0021] The virtual-real homeomorphic digital twin platform is used to obtain the micro error components of each motion axis at the start and end positions, the positive and negative directions of the errors are determined based on each local coordinate system, and the reverse gap error term is additionally superimposed when the parallel motion axis moves in the reverse direction.

[0022] The calculated micro error component change value is combined with the unified mathematical model of machine tool errors to determine the positioning deviation existing in the process of the numerical control machine tool moving along the precision self-healing path.

[0023] The positioning deviation value is compared with the preset precision self-healing value, and when the positioning deviation value does not exceed the preset precision self-healing value, the precision self-healing strategy is not started.

[0024] When the positioning deviation value exceeds the preset precision self-healing value, the precision self-healing instruction is issued to the entity numerical control machine tool through the OPC UA protocol, the time-varying precision self-healing of the machine tool is realized in the form of G code, the customized strategy is provided in combination with the positioning precision of the machine tool itself, and the machining capacity of the machine tool is ensured.

[0025] Optionally, the sensing data is real-time sensing data based on the OPC UA protocol or future sensing data obtained through forward deduction; the forward deduction includes multi-scale time feature extraction, horizontal and vertical attention fusion, and time sequence feature dynamic modeling.

[0026] The multi-scale time feature extraction includes real-time sensing data for transient windows, medium-range windows and long-range windows, and the alignment of time sequences with different window lengths is realized through non-uniform Fourier embedding.

[0027] The horizontal and vertical attention fusion includes cross-window vertical attention and window-in horizontal attention; the cross-window vertical attention dynamically establishes the key point correlation between the features of different time scale windows through a deformable convolution kernel; the window-in horizontal attention enhances the capture of local dynamic characteristics of sensing features by introducing a data change rate feature.

[0028] The time sequence feature dynamic modeling is based on a state space model, including state space parameter dynamicization and bidirectional state transmission; the state space parameter dynamicization adjusts the state transition matrix in real time according to the current time sequence feature change rate; the bidirectional state transmission dynamically accumulates the process through forward modeling and captures the system response lag characteristics through backward transmission.

[0029] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0030] The application provides a time-varying precision customized fine self-repairing method based on virtual-real isomorphism evolution. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0032] Figure 1 A flowchart of a time-varying precision customized fine self-repairing method based on virtual-real isomorphism evolution provided by an embodiment of the present application.

[0033] Figure 2 A flowchart of step A5 in a time-varying precision customized fine self-repairing method based on virtual-real isomorphism evolution provided by an embodiment of the present application. DETAILED DESCRIPTION

[0034] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0035] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0036] The time-varying precision customized fine self-healing method based on virtual-real isomorphism evolution provided by the embodiments of the present application includes the following steps in an exemplary embodiment, as shown in Figure 1 FIG. 1.

[0037] A1, according to the position relationship among the bed of the numerical control machine tool, each motion axis, the tool and the workpiece, a homogeneous transformation matrix from the workpiece coordinate system to the tool coordinate system is constructed, and a unified mathematical model of machine tool error is obtained based on small error theory; the motion axes include X, Y and Z three-direction translational axes, A and B two-direction rotational axes and the machine tool spindle; the unified mathematical model of machine tool error covers three-axis, four-axis and five-axis numerical control machine tools. The bed of the machine tool is determined as the reference coordinate system R, and local coordinate systems are created on the X, Y and Z three-direction translational axes, the spindle S, the tool T and the workpiece W in turn. On this basis, the movement error and the angle error of each motion axis in the real state are represented in combination with the small error theory (linear superposition of geometric error terms, smallness assumption of angle error, etc.), so as to obtain a basic error mathematical model involving only three translational axes.

[0038] Specifically, the three-axis numerical control machine tool is a numerical control machine tool involving only three translational axes; the four-axis and five-axis numerical control machine tools are numerical control machine tools including rotational axes A, B and C rotating around the X / Y / Z axis movement direction in addition to the X, Y and Z three-direction translational axes; the unified mathematical model of machine tool error of the three-axis numerical control machine tool is the basic error mathematical model, which contains four types, and the specific type is determined by the position of the bed reference coordinate system R relative to the X, Y and Z three-direction translational axis coordinate systems, i.e. RXYZ type, XRYZ type, XYRZ type and XYZR type.

[0039] For four-axis and five-axis machine tools including rotating axes, on the basis of the mathematical model of the base error of three-axis machine tools, the local coordinate systems of rotating axes rotating around the movement directions of X / Y / Z axes are determined as A, B or C respectively, and the unified mathematical model of the machine tool error embedded with the error module of the rotating axes is obtained in combination with the position relationship between the rotating axes and the translating axes in the workpiece-tool coordinate system. In the embodiment, according to the position relationship between the rotating axes and the translating axes in the workpiece-tool coordinate system, it is determined that the four-axis and five-axis numerical control machine tools respectively include 24 and 48 types; among them, the mainstream types of the five-axis numerical control machine tools include 12 types: RXYZAB type, XRYZAB type, XYRZAB type, XYZRAB type, CARXYZ type, CAXRYZ type, CAXYRZ type, CAXYZR type, ARXYZB type, ACRYZB type, ACRYZB type and ACRYZB type.

[0040] Taking the RXYZ type three-axis machine tool as an example, the homogeneous transformation matrix from the workpiece coordinate system to the tool coordinate system is shown in the following formula:

[0041]

[0042] Among them, is the homogeneous transformation matrix of the worktable W relative to the tool, T is the homogeneous transformation matrix of the worktable relative to the bed, W is the homogeneous transformation matrix of the bed R relative to the translating axis, is the homogeneous transformation matrix of the bed R relative to the translating axis, X is the homogeneous transformation matrix of the translating axis relative to the translating axis, X is the homogeneous transformation matrix of the translating axis Y relative to the main shaft, is the homogeneous transformation matrix of the main shaft Y relative to the tool, Z , and Z respectively represent the distances of the numerical control machine tool moving along S , and S directions in the ideal case. T x y z X Y Z

[0043] ​​​​​​​A2, use the precision measuring device to test the positioning accuracy of each motion axis of the numerical control machine tool, and determine the micro error data of the numerical control machine tool; the micro error data includes the movement, angle micro error of each motion axis at different positions, and the reverse gap error of each translational axis.

[0044] Specifically, one-way and two-way positioning accuracy tests can be performed on each motion axis of the machine tool using a laser interferometer. The laser interferometer has high precision measurement characteristics and can accurately collect linear displacement errors, angular displacement errors (such as pitch angle errors, yaw angle errors, and roll angle errors) of each axis at different motion positions. At the same time, through two-way motion testing, the gap error when the translational axis moves in reverse is captured, providing basic data support for subsequent error compensation.

[0045] In an exemplary embodiment, a laser interferometer is used to perform one-way positioning accuracy tests on the X, Y, Z, A, and B axes. The X-axis test range is 0-1000mm, with a sampling interval of 50mm, and the linear error ( delta x ) and angular error ( theta x 、 、 psi x ) of each sampling point are recorded; the Y-axis test range is 0-800mm, with a sampling interval of 40mm; the Z-axis test range is 0-600mm, with a sampling interval of 30mm; the A-axis test range is -180°-180°, with a sampling interval of 10°; and the B-axis test range is 0°-90°, with a sampling interval of 5°.

[0046] Two-way positioning accuracy tests are performed: the X-axis moves from 0→1000mm→0mm, the Y-axis moves from 0→800mm→0mm, and the Z-axis moves from 0→600mm→0mm, and the gap error of each axis when moving in reverse is recorded (e.g., the X-axis reverse gap error is 0.0008mm, the Y-axis is 0.0007mm, and the Z-axis is 0.0006mm).

[0047] A3, construct a virtual-real identical digital twin platform, integrate the machine tool error unified mathematical model, and store the micro error data to the platform database, to ensure that the virtual digital twin and the geometric error state of the physical numerical control machine tool evolve identically. The virtual digital twin is a virtual numerical control machine tool corresponding to the physical numerical control machine tool.

[0048] Specifically, a three-dimensional geometric model of the numerical control machine tool is constructed based on three-dimensional modeling software, and the structural dimensions and assembly relationship of the machine tool bed, guide rail, spindle, tool, and workpiece are restored. In the digital twin platform, the error unified mathematical model constructed in step A1 is integrated, the micro error data obtained by step A2 is stored to the platform database, and the association mapping of error data and machine tool three-dimensional model is established (such as the error data of a sampling point of X axis corresponds to the position of X axis in the three-dimensional model). An OPC UA communication server is built to realize data interaction between the digital twin platform and the entity numerical control machine tool, and to ensure that the error state and motion position of the virtual machine tool are real-time synchronized with the entity machine tool.

[0049] A4, acquiring sensing data in the machining process of the numerical control machine tool, evaluating the time-varying situation of machining precision in the machining process in combination with a time series prediction algorithm, and driving the synchronous evolution of the machining precision of the virtual and real machine tools to generate a precision self-healing path; the precision self-healing path is a set of motion trajectories and operation parameters for guiding the machine tool to perform precision compensation; the sensing data is real-time sensing data read based on the OPC UA protocol or future sensing data obtained through forward deduction.

[0050] The sensing data in the machining process of the numerical control machine tool includes temperature sensing data, stress sensing data, vibration sensing data, acoustic emission sensing data for monitoring the dynamic error of key components of the numerical control machine tool, and real-time running coordinate trajectory data of the machine tool cutter. In one embodiment, the temperature sensing data is the temperature of the spindle bearing and the temperature of the guide rail, the sampling frequency is 1 Hz, the vibration sensing data is the vibration acceleration of the spindle, the sampling frequency is 1000 Hz, the acoustic emission sensing data is the acoustic emission signal of the cutter during cutting, the sampling frequency is 2000 Hz, and the sampling frequency of the real-time coordinate trajectory data of the cutter is 100 Hz.

[0051] The time series prediction algorithm adopts a long short-term memory network or a time convolution network, extracts long-term and short-term time series dependencies in temperature, stress, vibration, and acoustic emission multi-source sensing data, mines the dynamic evolution trend of time series data, and combines an attention mechanism or a causal convolution to enhance the expression ability of key features, thereby achieving accurate prediction of thermal error and cutter wear.

[0052] In one exemplary embodiment, a long short-term memory network (LSTM) is used to construct a time series prediction model: the input layer is temperature, vibration, and acoustic emission data (time window is 60s, step is 1s), the hidden layer includes 2 layers of LSTM units (64 neurons per layer), and the output layer is the predicted value of thermal error and cutter wear error in the next 10s; the model training uses an Adam optimizer, the loss function is mean square error (MSE), and the training data set is the sensing data and actual error measurement data of the machine tool during continuous machining for 8 hours.

[0053] The error time-varying condition obtained by time series prediction is input into the digital twin platform, driving the precision state of the virtual machine tool to evolve synchronously with the actual machine tool. Through digital twin technology and time series prediction algorithm, the actual machining data of the physical entity machine tool and the simulation data of the virtual machine tool are interacted and optimized in real time to realize the dynamic improvement and consistency guarantee of the machining precision. That is, if the actual machine tool spindle temperature rises, causing the thermal error to increase, the virtual machine tool synchronously updates the thermal error value, and generates a precision self-healing path based on the error evolution result. The precision self-healing path is used for error compensation. The principle of error compensation is to add a value in the opposite direction to the predicted time-varying error. For example, if the spindle expands by 3 microns (Z direction), the compensation is to make the Z direction retract by 3 microns in the opposite direction through the numerical control system. In an embodiment, the generated precision self-healing path includes: "X-axis compensation displacement 0.0012 mm → Y-axis compensation displacement 0.0009 mm → A-axis compensation angle 0.005° → …".

[0054] To solve the problem that the existing real-time prediction does not consider compensation time advance, leading to compensation lag, and without fine design of compensation step, which easily damages the surface quality of the workpiece, an advanced deduction process is proposed in this embodiment, including multi-scale time feature extraction, horizontal and vertical attention fusion, and dynamic modeling of time series features.

[0055] Multi-scale time feature extraction includes real-time sensing data for transient window (second level), medium window (minute level) and long window (hour level), and realizes the alignment of different window length time series through non-uniform Fourier embedding. This design effectively solves the problem that traditional single time series modeling cannot consider multi-scale features such as transient impact, gradual change and long-term degradation in machine tool operation, and ensures the effective fusion of different time scale features through frequency domain feature alignment.

[0056] Horizontal and vertical attention fusion includes cross-window vertical attention and window-in horizontal attention; cross-window vertical attention dynamically establishes key point correlation between different time scale window features through deformable convolution kernel; window-in horizontal attention enhances the ability to capture local dynamic characteristics of sensing features by introducing data change rate features. This design significantly improves the model's adaptability to complex machine conditions, which can capture the coupling relationship between multi-scale time windows and accurately capture the dynamic change characteristics of single window sensing features.

[0057] The dynamic modeling of the time sequence characteristics is based on a state space model, including dynamic state space parameterization and bidirectional state transmission. The dynamic state space parameterization adjusts the state transition matrix in real time according to the current time sequence characteristic change rate, and the bidirectional state transmission dynamically accumulates the process through forward modeling and captures the system response lag characteristics through backward transmission. This design enables the model to completely describe the time-asymmetric physical process, fully considers the memory effect and path dependence in the dynamic response of the system, and realizes the high-precision advance deduction of the sensing data.

[0058] On the basis of obtaining high-precision advance sensing data, the time sequence prediction algorithm for predicting thermal error and tool wear is input into the time sequence prediction algorithm for predicting thermal error and tool wear, and a future time precision self-healing path containing multiple-step micro-adjustments is generated in advance, which ensures the precision self-healing timeliness and ensures the workpiece surface quality from being affected by the self-healing process.

[0059] A5, determine the change of the micro error component of each motion axis of the precision self-healing path based on the virtual-real homomorphism digital twin platform, and determine the positioning accuracy deviation in the self-healing process combined with the unified mathematical model of machine tool error, and transmit the precision self-healing path to the physical numerical control machine tool when the positioning accuracy deviation exceeds the preset precision self-healing value.

[0060] In the embodiment, as shown in Figure 2 , step A5 specifically includes the following steps:

[0061] A51, record the spatial coordinates of the initial point and the terminal point on the precision self-healing path, and take the spatial coordinate difference of the initial point and the terminal point on the precision self-healing path as the preset precision self-healing value.

[0062] A52, obtain the micro error component of each motion axis at the start and end positions through the virtual-real homomorphism digital twin platform, determine the positive and negative directions of the error based on each local coordinate system, and additionally superimpose the reverse gap error term when the parallel motion axis moves in the reverse direction.

[0063] A53, combine the calculated micro error component change value with the unified mathematical model of machine tool error to determine the positioning deviation existing in the movement of the numerical control machine tool along the precision self-healing path.

[0064] A54, judge whether the positioning deviation value exceeds the preset precision self-healing value, and obtain a first judgment result. If the first judgment result is yes, execute step A55; and if the first judgment result is no, do not start the precision self-healing strategy, and return to step A4 to continue to evaluate the machining precision time-varying condition in the machining process, and drive the synchronous evolution of the virtual-real machine tool machining precision to generate the precision self-healing path.

[0065] A55. By issuing accuracy self-healing commands to physical CNC machine tools through the OPC UA protocol, the machine tool's time-varying accuracy self-healing is realized in the form of G-code. Combined with the machine tool's own positioning accuracy, customized strategies are provided to ensure the machine tool's machining capabilities.

[0066] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0067] This application constructs a unified mathematical model of errors covering three-axis, four-axis and five-axis machine tools. By combining the testing and superposition of backlash errors, it solves the problems of poor adaptability and neglect of backlash influence in traditional error models, and provides unified support for geometric error modeling of multiple types of machine tools.

[0068] A customized self-healing strategy is proposed, which determines whether to activate self-healing by comparing the positioning deviation with the accuracy self-healing value. This avoids the problem of insufficient tool positioning accuracy introducing new errors when the amount to be compensated is small, and ensures the effectiveness of error compensation.

[0069] The design of the sensor data advance prediction module enables the advance prediction of sensor data through multi-scale feature extraction, horizontal and vertical attention fusion, and temporal dynamic modeling. It pre-generates a self-healing path with multiple small adjustments, which not only solves the compensation lag problem, but also avoids the damage to the workpiece surface quality caused by a single large adjustment, thus ensuring the dual requirements of processing accuracy and quality.

[0070] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A time-varying precision customized fine self-healing method based on virtual-real homeomorphism evolution, characterized in that, The application relates to a method for realizing precision self-recovery of a numerical control machine tool. According to the position relationship among the bed body of the numerical control machine tool, each movement axis, a tool and a workpiece, a homogeneous transformation matrix from a workpiece coordinate system to a tool coordinate system is constructed, and a machine tool error unified mathematical model is obtained based on a small error theory; the machine tool error unified mathematical model covers three-axis, four-axis and five-axis numerical control machine tools; The positioning accuracy of each movement axis of the numerical control machine tool is tested by using an accuracy measuring device, and small error data of the numerical control machine tool are determined; the small error data include movement, rotation small error of each movement axis at different positions and reverse gap error of each translational axis; A virtual-real homomorphism digital twin platform is constructed, the machine tool error unified mathematical model is integrated, and the small error data are stored in a platform database, so that geometric error state homomorphism evolution of a virtual digital twin and a physical numerical control machine tool is carried out; Sensing data in a machining process of the numerical control machine tool are acquired, machining precision time-varying conditions in the machining process are evaluated by combining a time sequence prediction algorithm, synchronous evolution of virtual-physical machine tool machining precision is driven, and a precision self-recovery path is generated; the precision self-recovery path is a set of movement trajectories and operation parameters for guiding the machine tool to carry out precision compensation; Based on the virtual-real homomorphism digital twin platform, small error component changes of each movement axis at the beginning and end of the precision self-recovery path are determined, positioning accuracy deviation in the self-recovery process is determined by combining the machine tool error unified mathematical model, and when the positioning accuracy deviation exceeds a preset precision self-recovery value, the precision self-recovery path is transmitted to the physical numerical control machine tool, and the method specifically comprises the following steps: The spatial coordinates of initial points and terminal points on the precision self-recovery path are recorded, and a spatial coordinate difference value of the initial points and the terminal points on the precision self-recovery path is taken as the preset precision self-recovery value; Small error components of each movement axis at the beginning and end positions are acquired through the virtual-real homomorphism digital twin platform, the positive and negative directions of the errors are determined based on local coordinate systems, and reverse gap error items are additionally superimposed when the translational axes move reversely; The calculated small error component change values are combined with the machine tool error unified mathematical model, and positioning accuracy deviation existing in the movement process of the numerical control machine tool along the precision self-recovery path is determined; The value of the positioning accuracy deviation is compared with the preset precision self-recovery value, and when the value of the positioning accuracy deviation does not exceed the preset precision self-recovery value, the precision self-recovery strategy is not started; When the value of the positioning accuracy deviation exceeds the preset precision self-recovery value, a precision self-recovery instruction is issued to the physical numerical control machine tool through an OPC UA protocol, time-varying precision self-recovery of the machine tool is realized in the form of G code, customized strategies are provided by combining the positioning accuracy of the machine tool itself, and the machining capacity of the machine tool is ensured.

2. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 1, characterized in that, The three-axis numerical control machine tool is a numerical control machine tool involving only three translational axes; the four-axis and five-axis numerical control machine tools are numerical control machine tools including rotating axes A, B and C rotating around X / Y / Z axis movement directions in addition to X, Y and Z three-direction translational axes; the machine tool error unified mathematical model of the three-axis numerical control machine tool is a basic error mathematical model, which contains four types, and the specific types are determined by the position of a bed body reference coordinate system R relative to X, Y and Z three-direction translational axis coordinate systems, namely RXYZ type, XRYZ type, XYRZ type and XYZR type.

3. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 2, characterized in that, According to the position relationship between the rotation axis and the translation axis in the workpiece-tool coordinate system, it is determined that four-axis and five-axis CNC machine tools respectively contain 24 and 48 types; among them, the mainstream types of five-axis CNC machine tools include 12 types: RXYZAB type, XRYZAB type, XYRZAB type, XYZRAB type, CARXYZ type, CAXRYZ type, CAXYRZ type, CAXYZR type, ARXYZB type, ACRYZB type, AXYRZB type and AXYZRB type.

4. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 2, characterized in that, Taking the RXYZ type three-axis machine tool as an example, the homogeneous transformation matrix from the workpiece coordinate system to the tool coordinate system is as follows: wherein is the homogeneous transformation matrix of the worktable W relative to the tool T , is the homogeneous transformation matrix of the worktable W relative to the bed R , is the homogeneous transformation matrix of the bed R relative to the linear axis X , is the homogeneous transformation matrix of the linear axis X relative to the linear axis Y , is the homogeneous transformation matrix of the linear axis Y relative to the linear axis Z , is the homogeneous transformation matrix of the linear axis Z relative to the spindle S , is the homogeneous transformation matrix of the spindle S relative to the tool T , x , y and z denote the distances that the CNC machine tool moves along the X , Y and Z directions in the ideal case, respectively.

5. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 1, characterized in that, The sensing data in the machining process of the numerical control machine tool includes temperature sensing data, stress sensing data, vibration sensing data, acoustic emission sensing data and real-time running coordinate track data of the machine tool cutter of the key components of the numerical control machine tool.

6. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 4, characterized in that, The time series prediction algorithm adopts a long short-term memory network or a time convolution network, extracts long short-term time series dependence in temperature, stress, vibration and acoustic emission multi-source sensing data, mines the dynamic evolution trend of time series data, and combines an attention mechanism or a causal convolution to enhance the expression ability of key features, to realize accurate prediction of thermal errors and tool wear.

7. The time-varying precision customized fine self-healing method based on virtual-real homeomorphous evolution according to claim 1, characterized in that, The sensing data is real-time sensing data read based on the OPC UA protocol or future sensing data obtained through forward deduction; the forward deduction includes multi-scale time feature extraction, horizontal and vertical attention fusion and time series feature dynamic modeling; The multi-scale time feature extraction includes real-time sensing data for transient windows, medium windows and long windows, and realizes the alignment of different window length time series through non-uniform Fourier embedding; The horizontal and vertical attention fusion includes cross-window vertical attention and window-in horizontal attention; The cross-window vertical attention dynamically establishes the key point correlation between different time scale window features through a deformable convolution kernel; the window-in horizontal attention enhances the capture of local dynamic characteristics of sensing features by introducing data change rate features; The time series feature dynamic modeling is based on a state space model, including state space parameter dynamics and bidirectional state transmission; the state space parameter dynamics adjusts the state transition matrix in real time according to the current time series feature change rate; the bidirectional state transmission dynamically accumulates the process through forward modeling and captures the system response lag characteristics through backward transmission.

Images