Dynamic regulation and control method for spatial attitude angle of boring cutter

Abstract

The invention discloses a dynamic regulation and control method for a space attitude angle of a boring cutter, which comprises the following steps: calculating pitching and deflection angle deviations of a cutter attitude and a target path, generating a preliminary control instruction through a multi-axis linkage control model, and optimizing a track sequence in combination with a path planning algorithm; a pre-trained tool posture deviation prediction model is used for conducting feedforward pre-compensation on the optimized track, and a calibrated motion control sequence is obtained; in the execution process, an instruction is corrected online through real-time feedback data and a feedback compensation algorithm, and parameters of the control model and the compensation algorithm are adjusted in a self-adaptive mode according to comprehensive performance indexes obtained through fusion of vibration and load data. By means of a cooperative control mechanism combining feedforward prediction, real-time feedback and self-adaptive optimization, high-precision and real-time dynamic regulation and control and multi-axis motion coordination of the space attitude of the tool are achieved, and the machining precision, the dynamic response capacity and the process stability of boring of complex parts are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of boring tool posture control, in particular to a dynamic control method for the spatial posture angle of a boring tool. BACKGROUND

[0002] In the field of precision machining, boring technology is the key process to realize high-precision hole machining and complex cavity forming, and its machining precision directly determines the performance and reliability of the parts. With the increasing complexity of product design, the machining demand for parts such as multi-angle inclined holes, deep cavity special-shaped structures and the like is growing, which puts unprecedented high requirements on the dynamic control ability of the tool spatial posture in the boring process.

[0003] A core bottleneck currently faced by the industry is that the existing boring equipment and control methods are difficult to realize high-precision, real-time collaborative adjustment of the tool pitch and yaw angles in continuous machining. When the machining path involves frequent posture transformation, the traditional method often relies on pre-set fixed programs or lagging error feedback, which leads to the fact that the tool axis cannot timely and accurately track the ideal direction when dealing with dynamic changes and non-standard angle machining tasks, thereby causing the machining path to deviate, the surface quality to decrease and even the parts to be scrapped.

[0004] To solve this problem, the existing technology attempts to introduce more complex multi-axis numerical control systems and offline trajectory simulation, however, these schemes still have obvious limitations. Offline simulation cannot respond to real-time disturbances that occur during machining, such as tool wear, uneven workpiece material or machine tool thermal deformation; and simply relying on error feedback after movement for correction has system delay and cannot pre-inhibit the deviation that is about to occur, especially in high-speed and high-dynamic machining, this hysteresis greatly reduces the precision control effect. SUMMARY

[0005] Therefore, the technical problem to be solved by the present application is to overcome the defects of insufficient machining precision and low efficiency caused by tool posture adjustment lag, poor multi-axis motion coordination and lack of forward-looking error compensation when the existing boring machining surface faces complex paths and dynamic working conditions, and to provide a dynamic control method for the spatial posture angle of a boring tool, which can realize high-precision, real-time dynamic control and multi-axis motion coordination of the tool pitch and yaw angles through a collaborative control mechanism combining feedforward prediction, real-time feedback and adaptive optimization, thereby significantly improving the machining precision, dynamic response capability and process adaptive stability of complex part boring.

[0006] To solve the above technical problems, the present application provides a dynamic control method for the spatial posture angle of a boring tool, comprising the following steps: Collecting current posture data of the boring tool and parameters of the target machining path in real time; calculating a pitch angle deviation value and a yaw angle deviation value of the tool axis according to the current posture data and the parameters of the target machining path; Based on the pitch angle deviation value and the yaw angle deviation value, calculating adjustment amounts required for driving each motion axis of the tool through a multi-axis linkage control model, and generating a preliminary tool direction control instruction according to the adjustment amounts; According to the preliminary tool direction control instruction and the target machining path, generating an optimized machining trajectory sequence that coordinates multi-directional motion of the tool through a path planning algorithm; Inputting the optimized machining trajectory sequence into a pre-trained tool posture deviation prediction model to predict a posture angle drift trend that may occur in the tool during motion along the optimized machining trajectory sequence; and pre-compensating and correcting the optimized machining trajectory sequence according to the predicted posture angle drift trend to obtain a calibrated motion control sequence; Converting the calibrated motion control sequence into real-time control instructions and issuing the real-time control instructions for execution; at the same time, acquiring motion feedback data of the machining equipment in real time, calculating a deviation between the real-time control instructions and the motion feedback data, and correcting subsequent real-time control instructions using a feedback compensation algorithm; In the machining process, collecting vibration data and load data reflecting the machining state in real time; performing fusion processing on the vibration data and the load data to obtain a comprehensive performance index; and adaptively adjusting control parameters of the multi-axis linkage control model and / or the feedback compensation algorithm according to changes in the comprehensive performance index.

[0007] In an embodiment of the present application, the current posture data of the boring tool is collected in real time, including: Continuously reading original electric signals generated by three mutually orthogonal axial sensors in an inertial measurement device directly fixedly installed on a boring tool shank or a machine tool spindle end, the original electric signals representing instantaneous motion states; including: three axial sensor output three original angular velocity signals for sensing rotational motion, and three axial sensor output three original acceleration signals for sensing linear motion; Conducting continuous integral operation on the three original angular velocity signals with respect to time to calculate a spatial angle change history of the tool axis; Using steady-state gravity field direction information contained in the three original acceleration signals to periodically correct and align the preliminarily calculated spatial angle change history; Finally outputting a group of angle values representing real-time orientations of the tool axis in a machine tool fixed coordinate system, i.e. current posture data, the data including a current pitch angle measured value and a current yaw angle measured value.

[0008] In one embodiment of the present application, according to the current posture data and the parameters of the target machining path, the pitch angle deviation value and the yaw angle deviation value of the tool axis are calculated, including: From the machining program stored in the numerical control system, the program segment instruction corresponding to the current machining time is read, the program segment instruction is decoded, the target position coordinates of the tool tip point specified by the program segment and the theoretical pointing information that the tool axis should have when reaching the target position are separated and extracted, and the theoretical pointing information is the parameter of the target machining path, including a theoretical pitch angle setting value and a theoretical yaw angle setting value; The current pitch angle measured value and the theoretical pitch angle setting value are subjected to algebraic difference operation to obtain a first deviation value as the pitch angle deviation value; The current yaw angle measured value and the theoretical yaw angle setting value are subjected to algebraic difference operation to obtain a second deviation value as the yaw angle deviation value.

[0009] In one embodiment of the present application, the multi-axis linkage control model is based on the geometric mapping relationship between the tool space position and posture and the position of each motion axis predefined based on the physical structure of the machine tool, and based on the pitch angle deviation value and the yaw angle deviation value, the adjustment amount is calculated through the multi-axis linkage control model, including: The current calculated pitch angle deviation value and yaw angle deviation value are added to the current pitch angle and yaw angle obtained through the sensor to obtain the target pitch angle and the target yaw angle of the next control period that the tool needs to reach; The target pitch angle and the target yaw angle are combined with the tool length to calculate the tool tip point target position or the tool direction vector in the machine tool coordinate system required to realize the posture; Based on the geometric series or parallel relationship between the motion axes of the machine tool, the tool tip point target position or the direction vector is uniquely mapped to a set of target position coordinates of each motion axis by using the inverse equation set containing the trigonometric function and the inter-axis offset amount; The current position feedback values of each motion axis are read, and the difference between each axis target position coordinates is obtained, which is the adjustment amount required for each motion axis to drive the tool.

[0010] In one embodiment of the present application, an optimized machining trajectory sequence is generated through a path planning algorithm, including: Based on the position points of each axis arranged in time sequence contained in the preliminary tool direction control instruction, a curve fitting technique is used to generate a smooth transition curve segment between each adjacent position point to ensure the continuous change of the motion speed and acceleration of each axis; All the curve segments are connected to form a trajectory describing the continuous and smooth movement of each axis, and the trajectory is discretized into a position point sequence according to a control period, as an optimized machining trajectory sequence.

[0011] In an embodiment of the present application, the establishment of the tool posture deviation prediction model comprises: A test machining program containing various posture changes is executed on the machine tool, and the optimized machining trajectory sequence, the spindle load data, and the measured posture deviation data obtained through the sensor are recorded synchronously; The optimized machining trajectory sequence and the spindle load data are taken as input samples, and the measured posture deviation data are taken as target output samples, which are input into a data processing network with multiple layers of nodes; The connection weights between the nodes in the data processing network are adjusted iteratively, so that the data processing network can learn and memorize the statistical correlation between the input samples and the target output samples, and the trained data processing network constitutes the tool posture deviation prediction model.

[0012] In an embodiment of the present application, the tool posture deviation prediction model is used to pre-compensate and correct the optimized machining trajectory sequence to obtain a calibrated motion control sequence, which comprises: The currently generated optimized machining trajectory sequence and the real-time collected spindle load data are input into the tool posture deviation prediction model, and according to the learned statistical correlation, the prediction results of the tool posture angle drift amount and the change direction that may occur in the future machining time are calculated and output; The angle drift amount in the prediction results is converted into a compensation amount for the position coordinates of each axis in the optimized machining trajectory sequence, and the compensation amount is superimposed on the original trajectory coordinates in the form of negative feedback, thereby generating the calibrated motion control sequence.

[0013] In an embodiment of the present application, the vibration data and the load data are fused to obtain a comprehensive performance index, which comprises: The original vibration signal in the machining process is collected in real time by the vibration sensor installed on the spindle box or the tool holder; the original vibration signal is analyzed in the frequency domain, the total vibration energy value of the signal in the pre-set characteristic frequency band associated with the machining stability is calculated, the vibration energy value is compared and converted with the equipment vibration energy safety value, and a vibration evaluation value representing the current vibration level relative to the equipment vibration energy safety value is obtained; The real-time current value or real-time power value of the machine tool spindle driving motor is read in real time as the original load data; the original load data is compared and converted with the equipment rated load value, and a load evaluation value representing the current load level relative to the equipment rated load value is obtained; The vibration evaluation value and the load evaluation value are weighted and fused according to different working conditions to give priority to the influence on the overall stability, that is, a comprehensive performance index reflecting the overall stability of the current machining system.

[0014] In an embodiment of the present application, the comprehensive performance index is monitored in real time, if it is determined that it is continuously higher than a preset stability threshold in a plurality of continuous control periods and shows an upward trend at the same time, it is determined that the current machining stability is deteriorating, and the control parameters of the multi-axis linkage control model and / or the feedback compensation algorithm are adaptively adjusted.

[0015] In an embodiment of the present application, in response to the determination of the deterioration of the stability, the following are included: Adaptively adjusting the control parameters of the multi-axis linkage control model, reducing the feedforward response strength coefficient in the multi-axis linkage control model for converting the angle instruction into the shaft position instruction, to reduce the response speed of the instruction change; Adaptively adjusting the control parameters of the feedback compensation algorithm, increasing the proportional action strength coefficient and the integral action strength coefficient in the feedback compensation algorithm for calculating the correction amount according to the position error, to enhance the correction ability of the deviation.

[0016] The above technical solutions of the present application have the following advantages compared with the prior art: The dynamic regulation method of the spatial posture angle of the boring tool provided by the present application can significantly improve the dynamic regulation precision and response speed of the spatial posture of the boring tool in the boring process, effectively avoid the deviation of the machining path and the decline of the surface quality, and is especially suitable for precision machining tasks of high dynamic and complex cavity. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to make the content of the present application more easily understood, the present application will be further described in detail below according to specific embodiments of the present application and in conjunction with the drawings, in which: Figure 1 is a step flow chart of the dynamic regulation method of the spatial posture angle of the boring tool of the present application; Figure 2 is a step flow chart of the method for collecting the current posture data of the boring tool in real time of the present application; Figure 3 is a step flow chart of the multi-axis linkage control model of the present application; Figure 4is a step flow chart of the present application for generating an optimized machining trajectory sequence through a path planning algorithm; Figure 5 is a step flow chart of the present application for establishing and using a tool posture deviation prediction model; Figure 6 is a step flow chart of the present application for obtaining a comprehensive performance index by fusing vibration data and load data; Figure 7 is a step flow chart of the present application for adaptively adjusting the control parameters of the multi-axis linkage control model and feedback compensation algorithm. DETAILED DESCRIPTION

[0018] The present application will be further described below in conjunction with the accompanying drawings and specific embodiments, so that those skilled in the art can better understand the present application and implement it.

[0019] Referring to Figure 1 illustrated, the present application proposes a dynamic regulation method for the spatial posture angle of a boring tool, comprising the following steps: Real-time acquisition of the current posture data of the boring tool and the parameters of the target machining path; calculation of the pitch angle deviation value and the yaw angle deviation value of the tool axis according to the current posture data and the parameters of the target machining path; starting from real-time data acquisition, first accurately calculate the pitch and yaw angle deviations between the tool axis and the target path, to lay a accurate input foundation for subsequent regulation.

[0020] Based on the pitch angle deviation value and the yaw angle deviation value, calculate the adjustment amount required for driving each motion axis of the tool through a multi-axis linkage control model, and generate a preliminary tool direction control instruction according to the adjustment amount; through the multi-axis linkage control model, intelligently solve these angle deviations into specific adjustment amounts of each motion axis, thereby generating a preliminary direction control instruction, which ensures the accuracy of the tool posture adjustment instruction in theory.

[0021] According to the preliminary tool direction control instruction and the target machining path, generate an optimized machining trajectory sequence that coordinates the multi-directional motion of the tool through a path planning algorithm; further introduce the path planning algorithm to globally coordinate and optimize the preliminary instruction, generating a smooth and efficient tool motion trajectory sequence, which effectively reduces the vibration and error caused by motion mutations.

[0022] The optimized machining trajectory sequence is input into the pre-trained tool posture deviation prediction model to predict the posture angle drift trend that may occur during movement of the tool along the optimized machining trajectory sequence; the optimized machining trajectory sequence is pre-compensated and corrected according to the predicted posture angle drift trend to obtain a calibrated motion control sequence; the pre-trained tool posture deviation prediction model is embedded into the control process, which is not an offline simulation tool but a prediction engine trained based on historical machining data and capable of online operation. The prediction engine performs a preview of the optimized trajectory sequence before execution, predicts the posture angle drift trend that may occur when the tool follows the trajectory, and pre-compensates and corrects the trajectory based on the prediction result, so that the deviation is offset before it actually occurs, realizing a fundamental change from “lagging correction” to “forward suppression”.

[0023] The calibrated motion control sequence is converted into real-time control instructions and executed; at the same time, motion feedback data of the machining equipment are acquired in real time, the deviation between the real-time control instructions and the motion feedback data is calculated, and the subsequent real-time control instructions are corrected by using a feedback compensation algorithm; during the instruction execution stage, real-time motion feedback monitoring is performed synchronously, and the subsequent instructions are fine-tuned by using the feedback compensation algorithm to form an inner loop correction for rapid response to real-time disturbances. The feedback compensation algorithm can use a PID controller with fixed gain to generate a correction amount based on the deviation between the real-time control instructions and the motion feedback data.

[0024] During the machining process, vibration data and load data reflecting the machining state are acquired in real time; the vibration data and the load data are fused to obtain a comprehensive performance index; the control parameters of the multi-axis linkage control model and / or the feedback compensation algorithm are adaptively adjusted according to the change of the comprehensive performance index; by continuously acquiring vibration and load data, fusing the data into a comprehensive performance index, and adaptively adjusting the core model and the control parameters accordingly, the entire system can adapt to changes in the machining conditions, such as tool gradual wear or machine tool state fluctuations, and maintain high precision and high stability in the long term.

[0025] In principle, the above technical features work together to systematically solve the problems pointed out in the background art. The multi-axis linkage control model and the path planning algorithm jointly ensure the accuracy and coordination of multi-directional movement, theoretically eliminating the trajectory error caused by mismatched movement axes. The application of the tool posture deviation prediction model essentially incorporates the dynamic characteristics (such as structural flexibility, inertial delay, and nonlinear friction) of the machining system into the feedforward control. By predicting and compensating potential deviations in advance, the method effectively overcomes the lagging defect of traditional feedback control and significantly improves the dynamic accuracy. Real-time feedback compensation and adaptive parameter adjustment constitute a double-layer closed loop, with the former quickly suppressing instantaneous disturbances and the latter slowly optimizing the long-term performance of the system, thereby ensuring the robustness of the method under complex and time-varying working conditions.

[0026] Overall, the method in this application achieves precise, stable, and intelligent dynamic control of tool space attitude through a fusion architecture of "feedforward prediction - real-time feedback - adaptive optimization". The beneficial effects include: firstly, significantly improved machining accuracy, especially dynamic contour accuracy, as major error sources are proactively suppressed; secondly, improved machining efficiency for complex paths due to more coordinated motion and no need for speed reduction for error correction; and finally, the system's adaptive capability extends the stable machining time window, reducing manual intervention and downtime caused by changes in process conditions, thereby improving the overall reliability, consistency, and intelligence level of boring complex parts.

[0027] In some of the embodiments described above in this application, real-time acquisition of the boring tool's current attitude data is a fundamental step in the entire dynamic control method. Its accuracy and real-time performance directly affect the effectiveness of subsequent deviation calculations and control command generation. However, in actual high-speed, high-precision machining environments, the boring tool's attitude changes are complex and rapid, and may be accompanied by vibration and impact. This makes it difficult for traditional attitude measurement methods to provide sufficiently accurate and stable real-time attitude data, easily leading to problems such as measurement lag, noise interference, or long-term drift, thereby affecting the performance of the entire control system.

[0028] In this regard, refer to Figure 2 As shown, this application further proposes a method for real-time acquisition of the current attitude data of a boring tool, specifically including: continuously reading the raw electrical signals representing the instantaneous motion state generated by three mutually orthogonal axial sensors in an inertial measurement unit (IMU) directly fixed to the boring tool holder or the end of the machine tool spindle; wherein, the three axial sensors used to sense rotational motion output three raw angular velocity signals, and the three axial sensors used to sense linear motion output three raw acceleration signals; continuously integrating the three raw angular velocity signals with respect to time to calculate the spatial angle change history of the tool axis; periodically correcting and aligning the initially calculated spatial angle change history using the steady-state gravity field direction information contained in the three raw acceleration signals; and finally outputting a set of angle values ​​representing the real-time orientation of the tool axis in the machine tool fixed coordinate system, i.e., the current attitude data, which includes a measured value of the current pitch angle and a measured value of the current yaw angle.

[0029] Specifically, the method ensures the close synchronization of the sensor and the boring tool movement by directly fixing the inertial measurement device to the boring tool shank or the machine tool spindle end, thereby directly and accurately reflecting the instantaneous movement state of the boring tool. The inertial measurement device usually integrates a three-axis gyroscope and a three-axis accelerometer. Among them, three mutually orthogonal gyroscope sensors are used to perceive the rotational movement of the boring tool in space and output three raw angular velocity signals. These angular velocity signals represent the instantaneous rotational rate of the boring tool in each axis. By continuously integrating the time operation on these raw angular velocity signals, the total rotation angle of the boring tool axis in a period of time can be accumulated, thereby calculating the angle change history of the tool axis in space at a high frequency. This angular velocity integration-based method can provide fast attitude update and real-time capture of dynamic attitude changes of the boring tool.

[0030] However, pure integration operation is susceptible to sensor noise and zero drift, resulting in cumulative error after long-time integration. To solve this problem, the method further utilizes the three raw acceleration signals output by the three mutually orthogonal accelerometer sensors. In the absence of external non-gravitational acceleration (or external acceleration can be identified and compensated), the accelerometer mainly perceives gravitational acceleration. The direction of gravitational acceleration is relatively stable in the earth coordinate system and can be used as a reference direction for attitude calculation. By comparing the gravitational direction measured by the accelerometer with the gravitational direction corresponding to the attitude calculated by the gyroscope integration, the cumulative error generated by the gyroscope integration can be periodically corrected, especially the drift of the pitch angle and the yaw angle, so that the attitude calculation result is aligned with the actual gravitational direction, thereby improving the long-term stability and accuracy of the attitude data. After the fusion algorithm processing of the above-mentioned angular velocity integration and acceleration correction, the inertial measurement device can finally output a set of angle values accurately representing the real-time orientation of the tool axis in the machine tool fixed coordinate system. For boring tool attitude regulation, the key lies in its pointing in the machining plane, therefore, the output current attitude data specifically includes a current pitch angle measured value and a current yaw angle measured value, which directly reflects the inclination angle of the tool axis relative to the machine tool coordinate system X-Y plane and Z axis, providing accurate input for subsequent deviation calculation and attitude regulation.

[0031] Furthermore, the method for calculating the pitch angle deviation and yaw angle deviation of the tool axis based on the current attitude data and the parameters of the target machining path includes: reading the program segment instruction corresponding to the current machining moment from the machining program stored in the CNC system; decoding the program segment instruction; separating and extracting the target position coordinates that the tool tip should reach as specified by the program segment, as well as the theoretical pointing information that the tool axis should have when reaching the target position. The theoretical pointing information is the parameter of the target machining path, including a theoretical pitch angle setting value and a theoretical yaw angle setting value; performing an algebraic difference operation between the current measured pitch angle value and the theoretical pitch angle setting value to obtain the first deviation value, which is used as the pitch angle deviation value; and performing an algebraic difference operation between the current measured yaw angle value and the theoretical yaw angle setting value to obtain the second deviation value, which is used as the yaw angle deviation value.

[0032] Specifically, the CNC system is the core control unit for automated machining on a machine tool, storing detailed machining programs. These programs consist of a series of program segments arranged chronologically, each precisely specifying the position and orientation the tool should achieve at a specific machining moment. By reading and decoding the program segment instructions for the current machining moment, the system can extract the precise target position coordinates that the tool tip should reach, as well as the theoretical orientation information that the tool axis should possess when reaching that target position. This theoretical orientation information is the core basis for guiding tool orientation adjustments, specifically including the theoretical pitch angle setting value of the tool axis in the vertical plane and the theoretical yaw angle setting value in the horizontal plane.

[0033] After obtaining the theoretical attitude setting, the pitch angle deviation value measures the degree to which the tool axis deviates from the theoretical attitude in the vertical plane. By performing a simple algebraic difference between the current measured pitch angle value obtained in real time from the inertial measurement unit and the theoretical pitch angle setting value extracted from the CNC machining program, the actual deviation of the tool in the pitch direction can be directly and quantitatively obtained. This first deviation value is the direct input for subsequent pitch direction adjustments in the multi-axis linkage control model. Similarly, the yaw angle deviation value reflects the degree to which the tool axis deviates from the theoretical attitude in the horizontal plane. By performing an algebraic difference between the current measured yaw angle value obtained in real time and the theoretical yaw angle setting value, the actual deviation of the tool in the yaw direction can be obtained. This second deviation value is another key input for subsequent yaw direction adjustments in the multi-axis linkage control model.

[0034] In actual machining, how to accurately and efficiently convert abstract angular deviation values ​​into specific adjustment amounts for each motion axis of the machine tool in order to achieve precise control of the tool posture is a challenge, especially in multi-axis linkage systems. The complex geometric relationship between the tool posture and the position of each motion axis makes direct calculation of adjustment amounts challenging and may lead to insufficient control accuracy or low calculation efficiency.

[0035] In this regard, refer to Figure 3 As shown, this application further proposes a multi-axis linkage control model, which is based on the geometric mapping relationship between the tool's spatial position and attitude and the positions of each motion axis, predefined according to the machine tool's physical structure. Essentially, the multi-axis linkage control model is a machine tool kinematics model, its core being the precise description of how the relative motions of each motion axis (e.g., X, Y, Z linear axes and A, B, C rotary axes) jointly determine the tool's position and attitude in space. This geometric mapping relationship is pre-established based on the machine tool's mechanical design drawings, assembly parameters, and the motion range of each axis, among other physical structural information. It can be a forward kinematics model (mapping each axis position to the tool's attitude and position) or a reverse kinematics model (mapping the tool's attitude and position to each axis position). This model ensures a precise mathematical description of the tool's spatial motion and is the foundation for achieving precise attitude control. Based on the pitch angle deviation and yaw angle deviation values, the adjustment amount is calculated through this multi-axis linkage control model, specifically including the following steps: First, the calculated pitch and yaw deviation values ​​are added to the current pitch and yaw angles obtained from the sensors to obtain the target pitch and yaw angles for the tool in the next control cycle. This step aims to determine the ideal attitude the tool should achieve in the next control cycle. By superimposing the actual tool attitude (current pitch and yaw angles) detected at the current moment with the attitude deviation values ​​(pitch and yaw angle deviations) calculated based on the target machining path, a corrected target attitude closer to the ideal state can be obtained. This superposition operation can be understood as a feedforward control strategy, that is, before actual execution, the target is pre-adjusted based on known deviation information in order to eliminate or reduce the current deviation in the next control cycle. The sensor can be an inertial measurement unit (IMU) or other attitude measurement device.

[0036] Secondly, by combining the target pitch angle and target yaw angle with the tool length, the target position of the tool tip or the tool direction vector required to achieve this posture in the machine coordinate system is calculated. After determining the target pitch angle and target yaw angle of the tool, they need to be converted into specific geometric quantities in the fixed coordinate system of the machine tool that the machine tool control system can understand. Considering that boring tools usually have a certain length, the position of their tool tip is affected by the tool posture. Therefore, by combining the known tool length information, the target position of the tool tip in the machine coordinate system under the target posture can be accurately calculated, or the target direction vector of the tool axis in the machine coordinate system can be directly calculated. This provides a clear input for the subsequent inverse kinematics solution.

[0037] Secondly, based on the geometric series or parallel relationships between the machine tool's motion axes, the inverse kinematics equations, including trigonometric functions and inter-axis offsets, uniquely map the tool tip target position or tool direction vector to a set of target position coordinates for each motion axis. This is a crucial step in converting the tool's spatial attitude and position commands into specific motion commands for each machine tool axis. The inverse kinematics equations are a mathematical model established based on the machine tool's mechanical structure (e.g., whether it's a serial or parallel robot structure, and the relative positions and directions between axes). These equations typically contain complex trigonometric functions (to describe the motion of rotary axes) and inter-axis offsets (to describe fixed distances or offsets between axes). By solving these equations, the tool tip target position or tool direction vector can be uniquely converted into the target position coordinates that each motion axis (such as X, Y, Z, A, B, and C axes) of the machine tool should achieve. This uniqueness of mapping ensures the explicitness of the control commands.

[0038] Finally, the current position feedback value of each motion axis is read and subtracted from the target position coordinates of each axis. The difference for each axis is the adjustment amount required to drive each motion axis of the tool. To achieve precise control, the calculated target position coordinates of each axis need to be compared with the actual positions of the machine tool axes. The current position feedback value of each motion axis can be obtained in real time through encoders or other position sensors inside the machine tool. Subtracting the corresponding current position feedback value from the target position coordinates of each axis yields the displacement adjustment amount required for that axis. This adjustment amount directly indicates how the drive system should move the axis to reach the target position, thereby achieving precise tool orientation adjustment.

[0039] In the dynamic control method of boring tool spatial attitude angle, after calculating the required adjustment amount for each motion axis to drive the tool based on the pitch angle deviation value and yaw angle deviation value through a multi-axis linkage control model, and generating preliminary tool direction control commands based on the adjustment amount, directly using these discrete preliminary commands as motion trajectories may cause abrupt changes in velocity and acceleration between adjacent command points for each motion axis. This discontinuous motion will cause mechanical shock and vibration, which will not only accelerate the wear of machine tool components and reduce equipment life, but more importantly, it will seriously affect the accuracy of boring and the quality of the machined surface, especially in precision machining scenarios where attitude stability is extremely important.

[0040] In this regard, refer to Figure 4 As shown, this application further proposes a step of generating an optimized machining trajectory sequence through a path planning algorithm. Specifically, based on the position points of each axis arranged in chronological order contained in the initial tool direction control command, curve fitting technology is used to generate smooth transition curve segments between adjacent position points to ensure that the motion speed and acceleration of each axis change continuously. All curve segments are connected end to end to form a trajectory describing the continuous and smooth motion of each axis. The trajectory is then discretized into a position point sequence according to the control cycle, which serves as the optimized machining trajectory sequence.

[0041] The initial tool orientation control commands are generated based on the pitch and yaw angle deviations of the tool axis, calculated by a multi-axis linkage control model to determine the necessary adjustments for each motion axis driving the tool. These commands are essentially a series of discrete target position coordinates that each motion axis should reach at a specific point in time. These position points constitute the initial path skeleton for the tool's adjustment from its current posture to the target posture, serving as the foundational data for generating a smooth trajectory.

[0042] Curve fitting is a mathematical method used to construct a continuous curve that passes through or approximates a set of discrete data points. In this application, this technique is used to connect adjacent axis position points in the initial tool direction control command to eliminate potential motion discontinuities between points. Commonly used curve fitting techniques include, but are not limited to, B-spline curves, NURBS curves, and cubic spline interpolation. By selecting appropriate curve types and fitting parameters, it can be ensured that the generated curve segments have good geometric and kinematic smoothness.

[0043] The continuity of motion speed and acceleration is crucial for achieving smooth machining. In mechanical motion, discontinuities in speed can lead to impacts, and discontinuities in acceleration can cause vibrations. By using curve fitting techniques, high-order continuous curve segments can be generated, mathematically ensuring that the speed and acceleration of each motion axis do not change abruptly as it moves along the trajectory. This avoids mechanical impacts and vibrations and improves the smoothness of the motion.

[0044] After generating smooth transition curve segments between adjacent points, these independent curve segments need to be precisely connected end-to-end according to their temporal sequence and spatial connection relationship. This connection operation ensures that the transition from one curve segment to the next is also smooth, thus forming a complete trajectory that runs through the entire machining process and describes the continuous and smooth motion of each motion axis. This trajectory is the ideal path for the tool to move in space, taking into account both geometric accuracy and kinematic stability.

[0045] Although the generated trajectory is continuous, the CNC system of a machine tool typically performs motion control with discrete control cycles. Therefore, the continuous trajectory needs to be sampled or interpolated according to a preset control cycle (e.g., every millisecond or microsecond) to convert it into a series of discrete position points. This sequence of discrete position points is the final optimized machining trajectory sequence, which can be directly input into the machine tool's motion controller as instructions to drive each motion axis. This discretization process must be performed while ensuring trajectory accuracy and motion smoothness.

[0046] In some embodiments described above, a tool attitude deviation prediction model is proposed to predict the potential attitude angle drift trend of the tool during its movement along an optimized machining trajectory sequence. However, in its implementation, the accuracy and reliability of this prediction model directly affect the effectiveness of subsequent pre-compensation correction. If the prediction model fails to fully capture the complex physical phenomena and nonlinear factors in actual machining, its prediction results may deviate significantly from the actual situation, leading to inaccurate pre-compensation correction or even introducing new errors, thus affecting the final machining accuracy of the boring tool.

[0047] In this regard, refer to Figure 5As shown, this application further proposes a method for establishing a tool attitude deviation prediction model, including: executing a test machining program containing various attitude changes on a machine tool to obtain multi-dimensional, highly correlated data required for training the prediction model. This typically involves designing a series of representative machining paths that should cover various complex attitude changes that the boring tool may encounter in actual machining, such as large pitch or yaw angle changes, high-speed motion, and sudden stops and starts, to ensure the diversity and coverage of the training data. Simultaneously recording the optimized machining trajectory sequence, spindle load data, and attitude deviation data obtained through sensor measurements requires a high-precision, high-sampling-rate data acquisition system to ensure that the optimized machining trajectory sequence (output by the CNC system), spindle load data (obtained through a spindle current or power sensor), and attitude deviation data (obtained through an inertial measurement unit or other high-precision attitude sensors) are strictly aligned in time. The attitude deviation data obtained through sensor measurements is typically acquired through a high-precision external measurement system, such as a laser tracker, an optical measurement system, or a high-precision inertial measurement unit directly mounted on the tool or spindle. The inertial measurement unit can provide real-time angular velocity and acceleration information. The actual attitude angle of the tool is obtained through integration and filtering, and then compared with the theoretical attitude angle to obtain the deviation.

[0048] The optimized machining trajectory sequence and spindle load data are used as input samples, and the corresponding measured attitude deviation data are used as the target output samples, both input to a multi-layered data processing network. The optimized machining trajectory sequence in the input samples can be represented as a series of time-series data, including information such as the position, velocity, and acceleration of each axis. The spindle load data can be real-time current, power, or torque values. These data typically require preprocessing before being input into the network, such as normalization and feature extraction, to adapt to the input format of the data processing network. The target output samples are the measured attitude deviation data, usually time series of pitch and yaw angle deviations. The multi-layered data processing network typically refers to an artificial neural network, such as a feedforward neural network, a recurrent neural network (e.g., a long short-term memory network, a gated recurrent unit), or a convolutional neural network. The choice of network structure depends on the characteristics of the data and the complexity of the prediction task; for example, for time-series data, a recurrent neural network or a long short-term memory network may be more suitable for capturing time dependencies. This network typically contains an input layer, one or more hidden layers, and an output layer, with each layer consisting of multiple nodes (neurons).

[0049] By iteratively adjusting the connection weights between nodes within the data processing network, the network learns and memorizes the statistical correlations between input samples and target output samples. This typically employs a backpropagation algorithm combined with an optimizer to minimize the error between the predicted and target outputs. In each iteration, the network processes a batch of training data, calculates the prediction error, and then adjusts the connection weights and bias terms between nodes based on the error. After sufficient training, the data processing network can automatically discover and encode the complex nonlinear relationship between input features (trajectory sequences, loads) and tool attitude deviations from a large number of input-output sample pairs. This relationship may include the influence of factors such as mechanical structure deformation, thermal drift, and changes in cutting force on tool attitude. Ultimately, the adjusted training data processing network constitutes the tool attitude deviation prediction model.

[0050] Specifically, in this embodiment, considering that tool attitude deviation is a dynamic process strongly correlated with time history, and its changes have a complex nonlinear mapping relationship with historical and current motion states, in a specific embodiment of the present invention, the data processing network of the multi-layer nodes preferentially adopts a Long Short-Term Memory (LSTM) network suitable for time series analysis. The LSM network structure, through its internal gating mechanism, can effectively learn and memorize key temporal features in long-sequence data, thereby accurately predicting the attitude drift trend caused by factors such as machine tool dynamics and cutting force changes.

[0051] Training a Long Short-Term Memory (LSTM) network is an iterative optimization process aimed at making the network output as close as possible to the actual pose deviation. Sufficient training means that the model's training process reaches a preset convergence criterion, thereby ensuring that the model acquires stable and reliable predictive ability. To achieve this goal, a clear training termination condition is needed. In this embodiment, the training termination condition is defined when the prediction error (e.g., mean squared error) calculated on the reserved validation dataset decreases by less than 1 × 10⁻⁶ within 20 consecutive training epochs. -5 When the total number of training cycles reaches a preset upper limit (such as 1000 cycles), it can be determined that the network has been sufficiently trained. At this point, the iteration is stopped, and the network state at this time is fixed as the final tool attitude deviation prediction model.

[0052] To enable those skilled in the art to implement this more clearly, a workable construction example is provided here. A predictive model is built that can be directly applied to a five-axis CNC boring machine to predict tool attitude angle drift within a future interpolation cycle. First, data preparation and preprocessing are performed: the model input is defined as historical time-series data of 10 control cycles, i.e., using the state of the current moment and the previous 9 consecutive cycles to predict the deviation at the next moment. The input feature vector for each time step specifically includes motion state features from the optimized machining trajectory sequence, totaling 15 dimensions: target position, velocity, and acceleration of the X, Y, and Z linear axes; target angle, angular velocity, and angular acceleration of the A and C rotary axes; and load features from process monitoring (1 dimension, i.e., spindle motor current percentage). Thus, each sample constitutes a tensor of dimension [10, 16]. Before inputting the data into the network, each feature dimension needs to undergo min-max normalization preprocessing. This involves calculating the maximum and minimum values ​​of each feature dimension based on the training set and linearly scaling the original value x to the [0,1] interval using the formula (x-min) / (max-min). The pitch and yaw deviation data to be predicted are processed in the same way. Next, a sequence-to-point long short-term memory network prediction model is constructed, with the following architecture: the input layer receives the tensor of the above shape; the first long short-term memory network layer is configured with 64 memory units and is set to return the entire time step sequence to retain complete temporal information; the second long short-term memory network layer is configured with 32 memory units and only outputs the hidden state of the final time step as a summary of the entire sequence; finally, a fully connected output layer with 2 neurons is connected, and a linear activation function is used to directly output the inversely normalized pitch and yaw deviation prediction values. The model training configuration is as follows: Mean squared error is used as the loss function, and an adaptive moment estimator optimizer with an initial learning rate of 0.001 is used for optimization; the batch size is set to 32 during training, and the historical dataset is divided into training and validation sets in an approximately 8:2 ratio; the training process continues for multiple epochs, and the validation set loss is evaluated after each epoch. The validation set loss is considered valid if the loss value decreases by less than 1 × 10⁻⁶ within 20 consecutive epochs. -5 Alternatively, training can be stopped when the total number of cycles reaches 1000, and the optimal weights can be saved.

[0053] Specifically, based on the established tool attitude deviation prediction model, this application further proposes using the tool attitude deviation prediction model to pre-compensate and correct the optimized machining trajectory sequence, obtaining a calibrated motion control sequence. This includes: inputting the currently generated optimized machining trajectory sequence and real-time acquired spindle load data into the tool attitude deviation prediction model. The optimized machining trajectory sequence is generated by a path planning algorithm based on the initial tool direction control command and the target machining path, describing the sequence of position points for continuous and smooth motion of each motion axis, representing the ideal motion path that the tool should follow. The real-time acquired spindle load data, such as the real-time current or power value of the spindle drive motor, can reflect the magnitude and changes of the cutting force between the tool and the workpiece. Changes in cutting force are one of the important factors causing dynamic drift in tool attitude. By acquiring this data in real time, key information about the current machining state can be provided to the prediction model. The tool attitude deviation prediction model is pre-trained using a large amount of test machining data, and it internally stores the statistical correlation between the optimized machining trajectory sequence, spindle load data, and actual attitude deviation. Providing the above input data to the model is to activate its prediction function, enabling it to infer possible deviations in tool attitude based on current and future machining conditions.

[0054] Based on the learned statistical correlations, the tool attitude deviation prediction model calculates and outputs prediction results. After receiving the optimized machining trajectory sequence and real-time spindle load data, it processes these input data using the complex nonlinear mapping relationships learned during the training phase. This processing typically involves calculations of multi-layer neural networks, converting input features into output predicted values ​​through multiplication and addition operations of activation functions and weight matrices. The prediction results specifically represent the possible pitch and yaw angle drifts of the tool axis over a future machining period, as well as the directions of these drifts. For example, the model might predict that at a certain point in time, the tool pitch angle will drift upwards by 0.05 degrees, and the yaw angle will drift to the left by 0.03 degrees.

[0055] The predicted angular drift is converted into compensation for the coordinates of each axis in the optimized machining trajectory sequence. The predicted angular drift represents the change in tool posture in space, while the machine tool's motion control system controls the positions of each motion axis (such as the X, Y, and Z axes, and rotary axes A, B, and C) to achieve tool movement. Therefore, the predicted angular drift needs to be converted into specific position compensation for each motion axis. This conversion process typically utilizes the inverse kinematics model of the machine tool. This model can inversely calculate the coordinate values ​​of each motion axis required to achieve the target position and posture of the tool tip (such as the tool tip point) in space. When a tool posture drift is predicted, it can be considered a small change in the target posture, and then the inverse kinematics model can be used to calculate the corresponding position adjustments required for each motion axis to counteract this small change. For example, if the tool is predicted to deflect upwards, it may be necessary to fine-tune the position of the Z-axis or B-axis (if present) to counteract this deflection.

[0056] The compensation amount is superimposed onto the original trajectory coordinates in the form of negative feedback, thereby generating a calibrated motion control sequence. Negative feedback superposition means that the calculated compensation amount is applied to the axis position coordinates of the original optimized machining trajectory sequence in the opposite direction to the predicted drift. For example, if the tool is predicted to drift upwards, the compensation amount will cause the tool to adjust downwards to counteract the expected upward drift. Through this pre-emptive, proactive compensation, the original optimized machining trajectory sequence is corrected, generating a calibrated motion control sequence. When this calibrated sequence is executed, the actual tool trajectory will be closer to the ideal state because the expected attitude drift has been considered and counteracted in advance.

[0057] In some of the above implementations, it was proposed to collect vibration and load data in real time and adjust control parameters based on changes in comprehensive performance indicators. Accurately and comprehensively assessing the overall stability of the machining process to guide subsequent adaptive adjustments is crucial to ensuring the effectiveness of the control system. Relying solely on vibration or load data may not adequately reflect the complex machining conditions, leading to inaccurate or delayed adjustments.

[0058] In this regard, refer to Figure 6As shown, this application further proposes to fuse vibration data and load data to obtain comprehensive performance indicators. Specifically, a vibration sensor installed on the spindle box or tool holder collects the raw vibration signal during the machining process in real time. The raw vibration signal is a direct physical quantity reflecting the dynamic characteristics of the interaction between the tool and the workpiece during machining. By converting it into an electrical signal through a vibration sensor (such as an accelerometer), mechanical vibrations caused by cutting forces, tool wear, chatter, and other phenomena can be captured. Vibration sensors typically employ piezoelectric accelerometers, which can be directly fixed to the boring bar holder, machine tool spindle box, or structural components near the cutting area to maximize the acquisition of vibration information related to the cutting process. The sensor must possess high sensitivity, a wide frequency response range, and good anti-interference capabilities to ensure the authenticity and accuracy of the acquired signals.

[0059] Subsequently, frequency domain analysis is performed on the original vibration signal to calculate the total vibration energy value within a pre-defined characteristic frequency band associated with machining stability. Frequency domain analysis converts a time-domain signal into a frequency-domain signal, revealing the various frequency components and their intensities. In machining stability assessment, specific frequency ranges (characteristic frequency bands) are often associated with specific instability phenomena (such as chatter and resonance). Calculating the total vibration energy value within these frequency bands quantifies the vibration intensity directly related to stability. Frequency domain analysis typically employs the Fast Fourier Transform (FFT) algorithm. Pre-defined characteristic frequency bands can be determined experimentally, theoretically, or empirically, such as the machine tool's natural frequency or tool chatter frequency. The total vibration energy value can be obtained by integrating or summing the squares of the amplitudes of each frequency component within the characteristic frequency band.

[0060] Next, the vibration energy value is compared and converted with the equipment's safe vibration energy value to obtain a vibration evaluation value representing the current vibration level relative to the equipment's safe vibration energy value. The equipment's safe vibration energy value is the maximum permissible vibration energy threshold under normal operation or stable processing conditions. By comparison, the original vibration energy value can be standardized into a dimensionless evaluation value, intuitively reflecting the "dangerous" degree of the current vibration level, facilitating subsequent integration and decision-making. The equipment's safe vibration energy value is usually provided by the equipment manufacturer or derived through statistical analysis of extensive experimental data under stable processing conditions. Comparison and conversion can be performed using ratios, percentages, or graded scoring methods; for example, dividing the current vibration energy value by the safe value, or assigning different scores based on its position within the safe value range.

[0061] Simultaneously, the real-time current or power value of the machine tool spindle drive motor is read as raw load data. The current or power value of the machine tool spindle drive motor directly reflects the load on the spindle during cutting. Excessive cutting load may lead to tool deformation, accelerated wear, and even machine tool vibration, which is an important factor affecting machining stability. Modern CNC machine tools usually have built-in current or power sensors for the spindle drive motor, which can directly acquire this data in real time through the CNC system interface or a dedicated data acquisition module.

[0062] Then, the original load data is compared and converted with the equipment's rated load value to obtain a load evaluation value representing the current load level relative to the equipment's rated load value. The equipment's rated load value is the maximum permissible load specified during the design and manufacture of the machine tool spindle. By comparing it with the rated load value, the original load data can be standardized into an evaluation value, intuitively reflecting the "load" level of the current load, facilitating subsequent integration and decision-making. The equipment's rated load value is usually explicitly given in the machine tool's technical parameters. The comparison and conversion methods are similar to those for vibration evaluation values, and can use ratios, percentages, or graded scores, for example, dividing the current load value by the rated load value and expressing it as a percentage.

[0063] Finally, the vibration evaluation value and load evaluation value are weighted and fused according to the different working conditions' emphasis on the overall stability, resulting in a comprehensive performance index reflecting the overall stability of the current machining system. This comprehensive performance index aims to comprehensively and accurately reflect the overall stability of the machining system. A single vibration or load index may not capture all unstable factors; weighted fusion combines the advantages of both to more comprehensively evaluate the machining state. The degree of influence of vibration and load on stability may differ under different working conditions, therefore weighting is necessary based on the actual situation. Weighted fusion can be achieved through linear weighted summation, i.e.: Comprehensive performance index = Wv * Vibration evaluation value + Wl * Load evaluation value, where Wv and Wl are the weighting coefficients for vibration and load, respectively, and Wv + Wl = 1. The weighting coefficients can be dynamically adjusted based on different machining materials, tool types, cutting parameters, and other working conditions using methods such as expert experience, historical data analysis, and machine learning algorithms.

[0064] Reference Figure 7 As shown, during the processing, the comprehensive performance index is monitored in real time. If it is determined that the index is continuously higher than the preset stable threshold for multiple consecutive control cycles and shows an upward trend, it is considered that the current processing stability is deteriorating, and the control parameters of the multi-axis linkage control model and / or feedback compensation algorithm are adaptively adjusted.

[0065] Specifically, real-time monitoring of the comprehensive performance index refers to the system continuously acquiring and analyzing the comprehensive performance index obtained by fusing vibration and load data within each control cycle. This index serves as a quantitative representation of the current processing state, providing a basis for subsequent stability assessment. When determining processing stability, the system continuously tracks changes in the comprehensive performance index. A preset stability threshold is a critical value determined based on experience, experimentation, or simulation, used to distinguish between stable and unstable processing states. When the comprehensive performance index is higher than this preset stability threshold for multiple consecutive control cycles, it indicates that the processing state has deviated from the ideal stable region. Simultaneously, the system analyzes the trend of this index; if it shows a continuous upward trend, it further confirms that processing stability is continuously deteriorating, rather than experiencing occasional fluctuations. This determination mechanism based on persistence and trend effectively avoids misjudgments, ensuring that adjustments are triggered only when intervention is truly necessary. Once it is determined that the current processing stability is deteriorating, the system will activate an adaptive adjustment mechanism to dynamically modify the control parameters of the multi-axis linkage control model and / or feedback compensation algorithm. Adaptive adjustment means that parameter modifications are not fixed but rather based on intelligent decisions made according to the degree and trend of the current processing state's deterioration. The multi-axis linkage control model is responsible for converting tool posture commands into actual motion quantities for each axis. Its control parameters may include response gain, filtering coefficients, etc. Adjusting these parameters can change the system's response characteristics to commands. The feedback compensation algorithm corrects the deviation between the actual motion and the command. Its control parameters typically include proportional, integral, and derivative gains, etc. Adjusting these parameters can change the system's ability and speed to correct deviations. Through adaptive adjustment of these parameters, the aim is to counteract or mitigate the trend of deteriorating machining stability, allowing the system to return to a stable operating state.

[0066] Specifically, this application further proposes a response to stability deterioration determination, including: adaptively adjusting the control parameters of the multi-axis linkage control model, reducing the feedforward response strength coefficient used to convert angle commands into axis position commands in the multi-axis linkage control model, so as to reduce the response speed of command changes; and adaptively adjusting the control parameters of the feedback compensation algorithm, increasing the proportional action strength coefficient and integral action strength coefficient used to calculate the correction amount based on the position error in the feedback compensation algorithm, so as to enhance the deviation correction capability.

[0067] Specifically, the feedforward response strength coefficient in a multi-axis linkage control model is a key parameter determining how quickly the system responds to changes in external angular commands. This coefficient typically exists as a gain in the feedforward control path, directly affecting the strength of the adjustment amount calculated from the pitch and yaw angle deviations and converted into the required position commands for each motion axis. When machining stability deteriorates, the system may be more susceptible to external disturbances or internal vibrations. In this case, if the feedforward response strength coefficient is too high, the system's rapid response to commands may lead to overcompensation or resonance, thereby exacerbating instability. Therefore, by adaptively reducing this coefficient, the system's response speed to command changes can be effectively reduced, making the system behavior smoother and avoiding oscillations or instability caused by rapid responses. For example, this coefficient can be an adjustable proportional gain, whose value is reduced by a preset step or proportionally according to the degree of deterioration when stability deteriorates.

[0068] Meanwhile, feedback compensation algorithms typically employ a PID (Proportional-Integral-Derivative) controller, where the proportional and integral action strength coefficients are core parameters. The proportional action strength coefficient determines the system's immediate response to the current position error; that is, the larger the error, the greater the correction force. The integral action strength coefficient is used to eliminate the system's steady-state error, providing continuous correction force by accumulating historical errors. When machining stability deteriorates, the tool may more easily deviate from the predetermined trajectory, generating continuous or cumulative errors. In this case, by adaptively increasing the proportional and integral action strength coefficients, the feedback compensation algorithm's ability to correct these deviations can be significantly enhanced. An increased proportional action strength coefficient enables the system to react more strongly to instantaneous errors, quickly pulling the tool back to the target trajectory; while an increased integral action strength coefficient effectively eliminates long-standing small errors, ensuring that the tool can ultimately follow the trajectory accurately, thereby effectively suppressing the decrease in machining accuracy caused by stability deterioration. For example, these coefficients can be dynamically adjusted according to the degree of deterioration of the overall performance index through table lookup or preset functions to ensure that appropriate correction force is provided under different degrees of deterioration.

[0069] Through the above technical solution, when the system determines that machining stability is deteriorating, this application can adopt a synergistic and complementary control strategy. On the one hand, by reducing the feedforward response strength coefficient in the multi-axis linkage control model, the system's response to new command changes becomes smoother and more cautious, effectively avoiding oscillations or instability that may be caused by excessively rapid response in unstable states, thereby suppressing further aggravation of the instability trend. On the other hand, by increasing the proportional action strength coefficient and integral action strength coefficient in the feedback compensation algorithm, the system exhibits a stronger ability to correct deviations between the actual motion and the target trajectory, and can quickly and continuously eliminate various errors generated during machining. This "feedforward deceleration, feedback acceleration" strategy enables the system to avoid exacerbating problems due to aggressive response when facing deteriorating machining stability, while actively and effectively correcting existing deviations, thereby quickly restoring and maintaining machining accuracy, significantly improving the machining stability and reliability of the boring tool under complex working conditions, and effectively preventing problems such as decreased machining quality, increased tool wear, and even workpiece scrap caused by stability issues.

[0070] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for dynamically controlling the spatial attitude angle of a boring tool, characterized in that, Includes the following steps: The current attitude data of the boring tool and the parameters of the target machining path are collected in real time; based on the current attitude data and the parameters of the target machining path, the pitch angle deviation and yaw angle deviation of the tool axis are calculated. Based on the pitch angle deviation and yaw angle deviation, the adjustment amount required for each motion axis to drive the tool is calculated through a multi-axis linkage control model, and preliminary tool direction control commands are generated according to the adjustment amount. Based on the initial tool direction control commands and the target machining path, an optimized machining trajectory sequence that coordinates the multi-directional movement of the tool is generated through a path planning algorithm. The optimized machining trajectory sequence is input into a pre-trained tool attitude deviation prediction model to predict the attitude angle drift trend that the tool may experience during its movement along the optimized machining trajectory sequence. Based on the predicted attitude angle drift trend, the optimized machining trajectory sequence is pre-compensated and corrected to obtain the calibrated motion control sequence. The calibrated motion control sequence is converted into real-time control commands and issued for execution. Simultaneously, motion feedback data of the processing equipment is acquired in real time, the deviation between the real-time control commands and the motion feedback data is calculated, and the subsequent real-time control commands are corrected using a feedback compensation algorithm. During the processing, vibration and load data reflecting the processing status are collected in real time; the vibration and load data are fused to obtain comprehensive performance indicators; and the control parameters of the multi-axis linkage control model and / or feedback compensation algorithm are adaptively adjusted according to the changes in the comprehensive performance indicators.

2. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: Real-time acquisition of the boring tool's current attitude data, including: The system continuously reads the raw electrical signals representing instantaneous motion states generated by three mutually orthogonal axial sensors within the inertial measurement unit, which is directly fixed to the boring tool holder or the end of the machine tool spindle. This includes three raw angular velocity signals output by the three axial sensors used to sense rotational motion, and three raw acceleration signals output by the three axial sensors used to sense linear motion. By continuously integrating the three original angular velocity signals over time, the spatial angle change history of the tool axis is calculated. Using the steady-state gravitational field direction information contained in the three original acceleration signals, the initially calculated spatial angle change history is periodically corrected and aligned; The final output is a set of angle values ​​representing the real-time orientation of the tool axis in the machine tool's fixed coordinate system, i.e., the current attitude data. This data includes a measured value of the current pitch angle and a measured value of the current yaw angle.

3. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 2, characterized in that: Based on the current attitude data and the parameters of the target machining path, the pitch angle deviation and yaw angle deviation of the tool axis are calculated, including: Read the program segment instruction corresponding to the current machining moment from the machining program stored in the CNC system, decode the program segment instruction, separate and extract the target position coordinates that the tool tip should reach as specified by the program segment, and the theoretical pointing information that the tool axis should have when reaching the target position. The theoretical pointing information is the parameter of the target machining path, including a theoretical pitch angle setting value and a theoretical yaw angle setting value. The first deviation value is obtained by performing an algebraic difference between the current measured pitch angle value and the theoretical pitch angle set value, and is used as the pitch angle deviation value. The second deviation value is obtained by algebraically subtracting the measured value of the current yaw angle from the theoretical yaw angle setting value, and is used as the yaw angle deviation value.

4. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: The multi-axis linkage control model is based on the geometric mapping relationship between the predefined spatial position and attitude of the tool and the positions of each motion axis according to the machine tool's physical structure. Based on the pitch angle deviation and yaw angle deviation values, the adjustment amount is calculated through the multi-axis linkage control model, including: The calculated pitch angle deviation and yaw angle deviation values ​​are added to the current pitch angle and yaw angle obtained by the sensor to obtain the target pitch angle and target yaw angle that the tool needs to achieve in the next control cycle. By combining the target pitch angle and target yaw angle with the tool length, the target position of the tool tip or the tool direction vector required to achieve this posture in the machine tool coordinate system can be calculated. Based on the geometric series or parallel relationship between the motion axes of the machine tool, the inverse equation system containing trigonometric functions and inter-axis offset is used to uniquely map the target position or direction vector of the tool tip to a set of target position coordinates of each motion axis. Read the current position feedback value of each motion axis, and subtract it from the target position coordinate of each axis. The difference obtained for each axis is the adjustment amount required for each motion axis to drive the tool.

5. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: Optimized processing trajectory sequences are generated using path planning algorithms, including: Based on the position points of each axis arranged in chronological order contained in the initial tool direction control command, curve fitting technology is used to generate smooth transition curve segments between adjacent position points to ensure that the motion speed and acceleration of each axis change continuously. Connect all curve segments end to end to form a trajectory describing the continuous and smooth motion of each axis. Then, discretize the trajectory into a sequence of position points according to the control cycle, which serves as the optimized machining trajectory sequence.

6. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: The establishment of a tool attitude deviation prediction model includes: The test machining program, which includes multiple posture changes, is executed on the machine tool, and the optimized machining trajectory sequence, spindle load data, and posture deviation data obtained by actual measurement through sensors are recorded simultaneously. The optimized machining trajectory sequence and spindle load data are used as input samples, and the corresponding measured attitude deviation data are used as target output samples, which are then input into a data processing network with multiple nodes. By iteratively adjusting the connection weights between nodes within the data processing network, the network can learn and memorize the statistical correlation between input samples and target output samples. The adjusted training data processing network then constitutes the tool attitude deviation prediction model.

7. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 6, characterized in that: Using a tool attitude deviation prediction model, the optimized machining trajectory sequence is pre-compensated and corrected to obtain the calibrated motion control sequence, including: The currently generated optimized machining trajectory sequence and the real-time collected spindle load data are input into the tool attitude deviation prediction model. Based on the statistical correlation it has learned, the model calculates and outputs the prediction results of the tool attitude angle drift and direction of change that may occur in the future machining time. The angular drift in the prediction results is converted into compensation for the position coordinates of each axis in the optimized machining trajectory sequence, and this compensation is superimposed on the original trajectory coordinates in the form of negative feedback, thereby generating the calibrated motion control sequence.

8. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: The vibration data and load data are fused to obtain comprehensive performance indicators, including: Vibration sensors installed in the spindle box or tool holder are used to collect raw vibration signals in real time during the machining process. Frequency domain analysis is performed on the raw vibration signals to calculate the total vibration energy value of the signal in a pre-set characteristic frequency band that is associated with machining stability. The vibration energy value is compared and converted with the equipment vibration energy safety value to obtain a vibration evaluation value that represents the current vibration level relative to the equipment vibration energy safety value. The real-time current or power value of the machine tool spindle drive motor is read in real time as the raw load data; the raw load data is compared and converted with the rated load value of the equipment to obtain a load evaluation value that represents the current load level relative to the rated load value of the equipment. By weighting and fusing the vibration evaluation value and the load evaluation value according to the different working conditions and their respective impacts on the overall stability, a comprehensive performance index reflecting the overall stability of the current processing system is obtained.

9. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 1, characterized in that: Real-time monitoring of comprehensive performance indicators; if it is determined that the indicators are continuously higher than the preset stability threshold for multiple consecutive control cycles and show an upward trend, it is considered that the current processing stability is deteriorating, and the control parameters of the multi-axis linkage control model and / or feedback compensation algorithm are adaptively adjusted.

10. The method for dynamically controlling the spatial attitude angle of a boring tool according to claim 9, characterized in that: The criteria for determining stability degradation include: The control parameters of the multi-axis linkage control model are adaptively adjusted, and the feedforward response strength coefficient used to convert angle commands into axis position commands in the multi-axis linkage control model is reduced to reduce the response speed of command changes. The control parameters of the feedback compensation algorithm are adaptively adjusted, and the proportional and integral action strength coefficients used to calculate the correction amount based on the position error are increased to enhance the deviation correction capability.

Images