A multi-axis linkage processing method for chamfering and polishing end of stud

Abstract

The present application relates to the technical field of numerical control machining, and discloses a multi-axis linkage machining method for chamfering and polishing of the end of a stud, which comprises: establishing a latch mechanism of servo operating state parameters and spindle phase between a servo driver and an edge computing unit; performing rigid chamfering, constructing a comprehensive rigid load phase mapping function based on servo parameters, and extracting an interference allowance peak interval; performing flexible polishing, converting the comprehensive rigid load phase mapping function into a radial position offset and a pre-compensation feed speed to perform machining, and synchronously constructing a flexible friction load mapping function; through cross-decoupling analysis of the load mapping functions of rigid and flexible processes, calculating a wear index to determine a polishing tool physical failure mode; and based on a load gradient and a rotation axis kinematics boundary, reconstructing a feed speed instruction to perform deflection avoidance. The method dynamically adjusts polishing parameters by using prior load data, and realizes online monitoring of tool wear and smooth movement.

Description

Technical Field

[0001] This invention relates to the field of CNC machining technology, specifically to a multi-axis linkage machining method for chamfering and polishing the end of a stud. Background Technology

[0002] In mechanical manufacturing, the machining quality of the stud end directly affects the assembly accuracy and service life of the connector. Conventional machining processes typically include rigid chamfering at the front end followed by flexible polishing. In actual production, due to machining errors in previous processes or dimensional fluctuations in the raw material itself, the stud end often exhibits uneven distribution of interference allowance. Traditional polishing processes usually rely on fixed interpolation trajectories and uniform feed parameters, making it impossible for the system to detect and adapt to these random variations in allowance distribution. When the flexible polishing tool moves to an area with a large allowance, the machining contact force undergoes localized abrupt changes. This passive force fluctuation makes it difficult to ensure consistent end-face polishing quality and also leads to irregular, accelerated wear on the polishing tool surface.

[0003] Since the wear of flexible polishing tools directly affects the final surface quality, monitoring their failure state is typically required in industrial settings. Existing monitoring solutions mainly rely on periodic shutdowns for manual spot checks or the installation of external vision sensors and high-precision torque sensors inside the machine tool. Shutdown inspections interrupt the machining cycle, reducing production efficiency; while adding external sensors not only increases hardware costs but also easily leads to signal distortion in the complex machining chamber with cutting fluid splashes and metal chips, resulting in poor reliability. Current CNC machining methods lack a decoupling mechanism to separate material allowance variables from tool wear variables from the underlying servo data, making it difficult to achieve online monitoring of tool life through pure software algorithms.

[0004] Furthermore, during multi-axis machining, when the tool end encounters a sudden local load change or allowance interference, if the control system relies solely on physical force to passively yield or force cutting along the original trajectory, it can easily lead to instantaneous overload of the feed axis. Existing adaptive adjustment technologies lack comprehensive consideration of the kinematic boundaries of the machine tool's underlying rotary and linear axes when generating avoidance trajectories. When performing tool axis posture deflection adjustments, the lack of synchronous reconstruction of the feed speed to conform to machine tool dynamics limitations easily leads to mismatches between speed and position commands, resulting in uneven servo motion, machine tool vibration, and significant position following errors. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-axis linkage machining method for chamfering and polishing the end of a stud. This method solves the problems of existing machining methods, such as difficulty in adaptively adjusting polishing parameters when faced with uneven material allowance distribution, leading to increased wear of polishing tools and the inability to achieve online status monitoring, and the tendency of fixed trajectory machining to cause interference and uneven movement when faced with sudden load changes.

[0006] To address the above problems, the present invention provides the following technical solution:

[0007] This invention provides a multi-axis linkage machining method for chamfering and polishing the end of a stud, comprising the following steps:

[0008] Execute system initialization and establish a real-time latching mechanism between the underlying servo driver and the edge computing unit inside the CNC system for servo operating status parameters and spindle phase angle;

[0009] Rigid chamfering and feature extraction are performed, and the rigid chamfering tool is invoked to perform machining. Based on the synchronously acquired servo running status parameters, a comprehensive rigid load phase mapping function is constructed, and the peak intervals characterizing the material interference allowance distribution are extracted.

[0010] The feedforward pre-compensation and flexible friction feature acquisition are performed, and a flexible polishing tool is replaced. The comprehensive rigid load phase mapping function is converted into radial position offset and pre-compensation feed speed to execute the pre-compensation polishing trajectory, and polishing feedback data is collected simultaneously to construct the flexible friction load phase mapping function.

[0011] Furthermore, the step of constructing the comprehensive rigid load phase mapping function includes: obtaining the net value of the spindle torque current, the net value of the linear feed axis composite thrust current, and the spatial composite position following error of the linear feed axis collected by the underlying servo driver, and performing low-pass filtering processing;

[0012] An extreme value normalization algorithm is used to map the filtered multidimensional bottom net parameters to a dimensionless interval of 0 to 1;

[0013] Based on the spindle synchronization phase angle, and combined with preset spindle current weights, feed current weights, and follower error weights, a weighted summation is performed on the normalized parameters to construct a comprehensive rigid load phase mapping function. The step of extracting the peak intervals characterizing the material interference margin distribution includes:

[0014] Calculate the arithmetic mean and standard deviation of the comprehensive rigid load phase mapping function within the basic analysis period, sum the arithmetic mean and the standard deviation of a preset multiple, and use the calculated sum as the dynamic peak value judgment threshold.

[0015] The load values ​​are compared with the peak value determination threshold point by point along the main axis phase angle to extract continuous candidate interference intervals;

[0016] The area integral of the load function sequence within each candidate interference interval is calculated, and the interval with the largest integral area is selected as the peak interval. The starting phase angle and ending phase angle corresponding to the peak interval are then output.

[0017] Furthermore, the step of converting the comprehensive rigid load phase mapping function into radial position offset and pre-compensated feed speed includes: extracting the load value of the comprehensive rigid load phase mapping function at the target phase, multiplying the load value with a preset force-position equivalent conversion coefficient, and using the calculation result as the radial position offset compensated to the normal direction of the basic interpolation trajectory.

[0018] Extract the set nominal polishing feed rate, subtract the product of the load value and the speed suppression gain coefficient, and generate a pre-compensated feed rate by taking the maximum value function in combination with the preset minimum feed rate threshold.

[0019] The radial position offset and the pre-compensated feed rate are superimposed on the original polishing toolpath data model to generate multi-axis linkage drive commands, which are then issued and executed.

[0020] Furthermore, this method also includes: performing cross-decoupling calculations of dual-process load characteristics, comparing and analyzing the phase mapping function of the comprehensive rigid load and the phase mapping function of the flexible friction load, and calculating the macroscopic wear index and the local work transfer coefficient to determine the physical failure mode of the flexible polishing tool. The step of calculating the macroscopic wear index includes:

[0021] The phase mapping function of flexible friction load and the phase mapping function of comprehensive rigid load are respectively calculated by definite integral in the phase interval from 0 to 2π to obtain the global friction work and the rigid integral area.

[0022] The macroscopic wear index is generated by performing a ratio calculation using the global friction work as the numerator and the product of the rigid integral area and the reference friction conversion coefficient and the sum of the positive bias constant as the denominator.

[0023] When the macro wear index is found to be lower than the set wear judgment threshold for multiple consecutive processing cycles, the flexible polishing tool is determined to have reached the macro wear limit.

[0024] Furthermore, the step of determining the physical failure mode includes: extracting local rigid cutting load sequences and local flexible friction load sequences based on the peak interval, calculating the covariance using the Pearson correlation coefficient algorithm, and dividing the covariance by the product of the standard deviations of the two local sequences to generate a local morphological correlation index;

[0025] Discrete integrals are performed on the local rigid cutting load sequence and the local flexible friction load sequence respectively to obtain the local rigid cutting interference and the local flexible friction work.

[0026] The local work transfer coefficient is generated by dividing the amount of work done by the local flexible friction by the product of the local rigid cutting interference and the preset proportional coefficient.

[0027] When the local morphological correlation index is lower than the preset correlation diagnosis threshold and the local work transfer coefficient is lower than the preset transfer coefficient diagnosis threshold, it is determined that the flexible polishing tool has a microscopic failure in the corresponding phase region.

[0028] Furthermore, this method also includes: performing trajectory adaptive reconstruction based on dynamic constraints, generating tool axis vector deflection based on the load gradient of the comprehensive rigid load phase mapping function, evaluating the physical limit time in combination with the kinematic boundary parameters of the machine tool rotary axis, reconstructing and issuing the feed rate command after maintenance correction to complete the deflection avoidance machining. The step of generating the tool axis vector deflection includes:

[0029] The load gradient sequence is obtained by performing differential operation on the integrated rigid load phase mapping function; the absolute value of the load gradient sequence is compared with the set load gradient dead zone threshold, and the difference exceeding the load gradient dead zone threshold is multiplied by the attitude conversion gain coefficient to generate the basic deflection angle.

[0030] Based on the maximum permissible angular velocity and maximum permissible angular acceleration of the machine tool's rotary axis, the basic deflection angle is filtered and its amplitude is truncated to generate a smooth deflection angle. The smooth deflection angle and the basic tool axis vector are then subjected to a matrix rotation transformation to synthesize a dynamic tool axis deflection vector. The steps of reconstructing and issuing the feed rate command after maintenance correction include:

[0031] Extract the linear displacement increment of the linear feed axis within the current spatial interpolation micro-segment, calculate the shortest time to complete the synchronous motion based on the underlying dynamic constraints of the rotation axis, and use it as the physical limit time;

[0032] By using the linear displacement increment as the numerator and the sum of the physical limit time and the time offset constant as the denominator, a ratio calculation is performed to obtain the forced dimension reduction reconstruction feed rate.

[0033] A first-order inertial low-pass filter algorithm is used to smooth the forced dimensionality reduction and reconstruction feed rate in the time domain, and the result is encapsulated with the synthesized dynamic tool axis deflection vector into a synchronous data frame and sent to the underlying servo driver.

[0034] This invention provides a multi-axis linkage machining method for chamfering and polishing the end of a stud. It has the following beneficial effects:

[0035] 1. This invention uses the comprehensive rigid load phase mapping function constructed in the rigid chamfering stage as the prior parameter of the polishing stage. By extracting the load value of the corresponding phase, it is equivalently converted into radial position offset and pre-compensated feed speed, and superimposed on the original polishing toolpath. This method uses the load data of the rigid stage to characterize the physical distribution of material interference allowance, and performs position and feed speed compensation adjustment in advance in the flexible polishing process. This solves the problem of polishing force fluctuation caused by uneven material allowance in the front-end process, and ensures the consistency of polishing.

[0036] 2. This invention cross-compares the phase mapping function of the comprehensive rigid load and the phase mapping function of the flexible friction load, uses integral calculation to obtain the global and local work done, and combines the Pearson correlation coefficient to calculate the macroscopic wear index and the local work transfer coefficient. This data decoupling calculation mechanism directly uses the CNC underlying servo data to separate the remaining variable of the machining object from the consumption variable of the polishing tool without adding external physical sensors, realizing online monitoring of the overall wear limit and local micro-failure of the flexible polishing tool.

[0037] 3. When dealing with local abrupt interference, this invention performs differential calculations on the load mapping function to generate the tool axis vector deflection. Simultaneously, it combines the kinematic boundary conditions such as the maximum allowable angular velocity and angular acceleration of the machine tool's underlying rotary axis to recalculate the physical limit time and reconstruct the feed rate command after the reduction and correction. This control logic, while triggering the tool axis deflection avoidance action, ensures that the multi-axis synchronous motion command conforms to the dynamic constraints of the underlying servo system, avoids machine tool vibration and following errors caused by command abrupt changes, and guarantees the smoothness of the linkage machining process. Attached Figure Description

[0038] Figure 1 This is a diagram of the multi-axis linkage machining system architecture of the present invention;

[0039] Figure 2 This is a flowchart of the multi-axis linkage machining method of the present invention;

[0040] Figure 3 This is a characteristic curve diagram of the dual-process load phase mapping function of the present invention;

[0041] Figure 4 This is a performance comparison chart of different processing methods of the present invention. Detailed Implementation

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] See attached document Figure 1 , Figure 1 This is a diagram illustrating the architecture of a multi-axis linkage machining system according to an embodiment of the present invention. The present invention provides a multi-axis linkage machining method for chamfering and polishing the end of a stud, which is implemented based on a multi-axis linkage machining system.

[0044] like Figure 1 As shown, the multi-axis linkage machining system includes a multi-axis linkage CNC machine tool, a CNC system, and an edge computing unit. The multi-axis linkage CNC machine tool is equipped with linear feed axes for spatial translation and positioning, and rotary axes for attitude deflection control. The spindle of the multi-axis linkage CNC machine tool is configured as a spindle rotary axis. The multi-axis linkage CNC machine tool is internally equipped with an automatic tool changer, which selectively clamps rigid chamfering tools or flexible polishing tools.

[0045] The CNC system integrates low-level servo drivers and position encoders to control the movement of each interpolation axis. The CNC system establishes a data interaction connection with the edge computing unit via an industrial communication bus. The low-level servo drivers transmit servo electrical control parameters acquired during machining to the edge computing unit via the industrial communication bus for algorithm processing and instruction reconstruction.

[0046] See attached document Figure 2 , Figure 2 This is a flowchart of a multi-axis linkage machining method according to an embodiment of the present invention. Based on the above-described multi-axis linkage machining system, the multi-axis linkage machining method for chamfering and polishing the end of a stud according to the present invention includes the following steps:

[0047] Step S10: Perform system initialization. The CNC system reads the physical and dynamic boundary parameters of the multi-axis linkage CNC machine tool and establishes a hard real-time latching mechanism between the servo parameters and the phase angle of the spindle rotation axis between the underlying servo driver and the edge computing unit.

[0048] Step S20: Perform rigid chamfering and feature extraction. The multi-axis CNC machine tool calls the rigid chamfering tool to execute the basic contour interpolation trajectory. The edge computing unit constructs a comprehensive rigid load phase mapping function based on multi-source low-level servo data synchronously acquired by the low-level servo driver, and extracts the peak intervals characterizing the material interference margin distribution.

[0049] Step S30: Perform feedforward pre-compensation and flexible friction feature acquisition. The multi-axis CNC machine tool is switched to a flexible polishing tool via an automatic tool changer. The edge computing unit converts the comprehensive rigid load phase mapping function into the radial position offset and speed modulation of the multi-axis linkage interpolation trajectory and sends it to the CNC system. The multi-axis CNC machine tool executes the pre-compensated polishing trajectory, while the edge computing unit synchronously collects polishing feedback data to construct the flexible friction load phase mapping function.

[0050] Step S40: Perform cross-decoupling calculation of dual-process load characteristics. Within the processing auxiliary time window, the edge computing unit performs integration and comparison analysis on the integrated rigid load phase mapping function and the flexible friction load phase mapping function, calculates the global wear index and local transmission coefficient, and determines the physical failure mode of the flexible polishing tool.

[0051] Step S50: Perform trajectory adaptive reconstruction based on dynamic constraints. When the flexible polishing tool is determined to have experienced local yielding failure, the edge computing unit generates the tool axis vector deflection based on the phase gradient of the rigid load, and solves the physical time bottleneck by combining the kinematic limits of the rotary axes of the multi-axis CNC machine tool. The edge computing unit reconstructs and issues the feed rate command after the maintenance correction, driving the multi-axis CNC machine tool to complete the closed-loop compliant deflection machining in the next machining cycle.

[0052] As a specific implementation method, step S10 of the present invention is implemented by the following sub-steps to initialize the dynamic parameters and acquire and latch the underlying data.

[0053] During the actual machining environment configuration phase, the CNC system internally stores machine tool parameters and servo configuration data for multi-axis CNC machine tools. The edge computing unit sends read commands to the CNC system via an industrial communication bus to obtain the dynamic boundary parameters of the rotary axes. In this embodiment, the specific dynamic boundary parameters include the maximum permissible angular velocity of the rotary axes. and maximum permissible angular acceleration The aforementioned dynamic boundary parameters are typically determined by the machine tool manufacturer during the servo debugging phase, taking into account the characteristics of the physical motor, and preset in the system register. These parameters serve as input constraints for subsequently determining the trajectory reconstruction time bottleneck.

[0054] To establish a unified coordinate system for spatial data analysis, the edge computing unit defines one revolution of the main spindle rotation axis as the basic analysis period, and sets the phase angle variable of the main spindle rotation axis as... Its domain is set as .

[0055] During the machining process, the underlying servo driver acquires the servo operating status parameters of the motor at a set high-frequency sampling period. For the configuration of the underlying servo driver's data acquisition bus and the establishment of the communication handshake protocol, those skilled in the art can perform corresponding register mapping operations according to the specific hardware device model used. The interface configuration process is well-known in the field and will not be elaborated upon here.

[0056] The edge computing unit acquires the underlying state parameters through the industrial communication bus, performs data preprocessing, and calculates and generates underlying net parameters for feature analysis. These underlying net parameters include the net value of the spindle torque current, the net value of the linear feed axis composite thrust current, and the spatial composite position following error of the linear feed axis.

[0057] To eliminate interference from the machine tool's own mechanical friction and the gravity of moving parts on the load characteristics, the system needs to acquire reference data under no-load conditions beforehand. Preferably, in a no-load running state without a workpiece clamped, the multi-axis CNC machine tool executes a reference no-load cycle according to the set machining program. During this cycle, the system records the steady-state no-load current values ​​of the spindle and each linear feed axis at the current speed and feed rate. When extracting data for formal cutting, the edge computing unit extracts the real-time total torque current of the spindle's rotating axis and subtracts the spindle no-load running current measured under the corresponding operating conditions from this value, thereby obtaining the net spindle torque current value, denoted as [missing value]. .

[0058] Considering that the mechanical inertia of the machine tool spindle during rigid cutting can smooth out minor fluctuations in cutting resistance to some extent, single spindle current data often cannot fully reflect the load changes caused by local material interference. Therefore, the edge computing unit synchronously extracts the real-time thrust current of each linear feed axis participating in the spatial contour interpolation motion. After deducting the no-load thrust current corresponding to each axis, the system performs a square root operation on the sum of the squares of the net thrust current components of each feed axis (such as the X-axis and Y-axis) participating in the interpolation to complete the spatial vector synthesis, and finally obtains the net composite thrust current of the linear feed axes, denoted as […]. .

[0059] Furthermore, at the physical machining level, when the tool encounters abrupt changes in material, such as the thread termination zone, the instantaneous increase in cutting resistance can easily lead to a slight dynamic lag in the actual trajectory of the feed axis. This lag manifests as an instantaneously amplified difference in the servo system's position closed loop. Based on this physical characteristic, the edge computing unit calculates the following error of each linear feed axis participating in the interpolation motion by subtracting the commanded position value from the actual position value fed back by the position encoder. The error components in each direction are then vectorized using the square root of the sum of squares to obtain the spatial composite position following error of the linear feed axis, denoted as […]. .

[0060] After acquiring the aforementioned multi-dimensional parameters, the edge computing unit establishes a hardware-level hard real-time alignment mechanism. The edge computing unit uses the spindle phase angle pulse signal fed back by the position encoder as the trigger clock. When the spindle rotation axis rotates to a specific phase angle... At that time, the edge computing unit synchronously latches the corresponding time according to the trigger clock.

[0061] This mechanism can effectively reduce the time delay deviation generated during communication transmission, accurately map and store the servo parameters collected in the time domain into the corresponding spatial phase domain matrix, and complete the real-time latching of the underlying net parameters, thereby providing a reliable data foundation for subsequent spatial feature analysis and trajectory reconstruction.

[0062] After acquiring the underlying net parameters of the synchronous latch, considering that the original data is accompanied by high-frequency electrical noise and that there are differences in the physical dimensions between different parameters, the edge computing unit executes the corresponding data processing mechanism to construct a comprehensive rigid load phase mapping function characterizing the cutting allowance. The specific processing procedure in this embodiment includes the following steps.

[0063] Step S221, the edge computing unit latches the net value of the spindle torque current in real time. Net value of combined thrust current of linear feed axis and linear feed axis spatial composite position following error Low-pass filtering is performed. In actual machining environments, the pulse-width modulation (PWM) drive process of servo motors inevitably introduces high-frequency switching noise. Directly using the unprocessed raw sequence can easily lead to misjudgment of local extrema during subsequent optimization. Under this condition, the edge computing unit uses a discrete digital filtering algorithm to smooth the data sequence in the spatial phase domain. For the selection of specific digital filter types and the setting of cutoff frequencies, those skilled in the art can perform conventional matching based on the motor drive frequency and the system sampling period. The filtering implementation process is well-known in the field and will not be elaborated here.

[0064] In step S222, the edge computing unit performs dimensionless normalization mapping on the filtered multidimensional bottom-level net parameters. Torque current is typically measured in amperes or rated percentages, while position following error is measured in micrometers. Not only are their physical dimensions different, but their numerical ranges also differ by orders of magnitude. Direct arithmetic synthesis would cause parameters with larger numerical ranges to mask changes in parameters with smaller numerical ranges, leading to distortion of physical characteristics. As a preferred approach, the edge computing unit employs an extreme value normalization algorithm. This algorithm maps the data from each dimension to a uniform value by subtracting the minimum value of the sequence within the current basic analysis period from the measured value of the current phase, and then dividing by the difference between the maximum and minimum values ​​of the sequence. Within the dimensionless range. After the scaling transformation in this step, the edge computing unit obtains the normalized principal shaft current accordingly. Normalized feed current and normalized following error .

[0065] In step S223, the edge computing unit introduces the system-preset machine tool rigidity weight coefficient and performs weighted synthesis of phase domain multidimensional features. Different models of multi-axis CNC machine tools have differences in their mechanical structure design, resulting in varying degrees of energy transmission of cutting resistance on the spindle and each linear feed axis. To accurately reproduce the actual interference situation in the machining area, the CNC system is pre-configured with dimensionless spindle current weights. Feed current weighting and following error weights The aforementioned weighting coefficients satisfy the allocation condition that their sum equals 1. The edge computing units are synchronized according to the phase angle of the main axis rotation. A weighted summation operation is performed on the three normalized sets of parameters to finally construct the comprehensive rigid load phase mapping function.

[0066] The mathematical relationship it follows is:

[0067] ;

[0068] In the formula:

[0069] This is a phase mapping function for a comprehensive rigid load; The weight of the dimensionless principal current; This is the normalized principal spindle current; Weighted by feed current; Normalized feed current; For following error weights; This is the normalized following error.

[0070] This mapping function records the asymmetric material allowance distribution characteristics along the circumference caused by the thread termination during rigid chamfering in a univariate form. It eliminates the adverse effects of servo noise and dimensional deviation, providing a smooth and unified data representation for subsequent identification and contour analysis of material interference peak intervals.

[0071] After constructing the comprehensive rigid load phase mapping function, the edge computing unit needs to identify the high-margin material abrupt change region caused by thread finishing. This region corresponds to the phase segment where the chamfering tool experiences a sharp increase in force during physical machining. The edge computing unit extracts the phase boundary of this region by executing an asymmetric material interference peak interval optimization mechanism. The specific processing procedure in this embodiment includes the following steps.

[0072] In step S231, the edge computing unit generates a dynamic peak determination threshold based on statistical characteristics. Different batches of stud blanks exhibit individual differences in material hardness and basic cutting allowance; using a fixed absolute value as the threshold can easily lead to missed or false judgments. As a preferred approach, the edge computing unit calculates the arithmetic mean and standard deviation of the comprehensive rigid load phase mapping function over the entire basic analysis cycle. The edge computing unit sums the arithmetic mean with the standard deviation of a preset multiple, and directly uses the calculated sum as the dynamic peak determination threshold. The sensitivity coefficient corresponding to the aforementioned preset multiple is usually empirically determined by process engineers based on the signal-to-noise ratio of historical processing samples and preset into the system. The introduction of the dynamic threshold enables the optimization algorithm to adapt to the fluctuations of the baseline load under different processing conditions.

[0073] In step S232, the edge computing unit performs a circumferential phase domain scan and candidate interval localization. Along the phase angle growth direction of the main rotation axis, the edge computing unit compares the real-time value of the integrated rigid load phase mapping function with the peak value threshold point by point. When the load value is detected to cross from below the threshold to above the threshold, the edge computing unit marks the corresponding phase angle as the interval start point; when the load value falls back from above the threshold to below the threshold, it is marked as the interval end point. The system pairs the start point with the corresponding end point to obtain one or more continuous candidate interference intervals.

[0074] In step S233, the edge computing unit establishes the main peak interval and outputs the phase boundary. In the cutting environment, tiny chips or localized hard spots remaining on the machined surface may trigger brief load pulses, leading to the extraction of multiple candidate interference intervals in step S232. To accurately locate the actual thread-end hard burr area, the edge computing unit performs area integration calculations on the load function sequence within each candidate interference interval to characterize the locally accumulated material interference. The edge computing unit selects the candidate interference interval with the largest integration area as the final peak interval. For the specific discrete numerical calculation method of the interval area integration, those skilled in the art can use the trapezoidal integral rule or Simpson's rule to solve it; the numerical analysis process is well-known in the art and will not be elaborated here.

[0075] After the above optimization and selection, the edge computing unit outputs the starting phase angle and ending phase angle corresponding to the main peak interval, which are denoted as... This interval accurately defines the physical location of the asymmetric material surplus concentration at the stud end, providing a spatial phase reference for multi-axis trajectory pre-compensation and micro-failure diagnosis in subsequent flexible polishing processes.

[0076] After completing the rigid cutting process and extracting the material allowance distribution characteristics, the system enters the pre-compensation polishing stage. The multi-axis CNC machine tool unloads the rigid chamfering tool from the spindle via an automatic tool changer and replaces it with a flexible polishing tool. Since the material removal mechanism of the flexible polishing tool relies on the physical elastic deformation of the brush or abrasive unit to establish contact pressure, if a conventional fixed trajectory is used for constant-speed machining, when the tool passes through the thread-end hard-burr area with significant material interference, the local compliant elements may yield due to instantaneous overload, or the burr removal may be incomplete due to insufficient dwell time. To overcome this physical defect, the edge computing unit, based on the force-position equivalence principle, converts the previously acquired rigid load into a pre-compensation amount for the flexible machining trajectory. The specific processing procedure in this embodiment includes the following steps.

[0077] In step S311, the edge computing unit calculates the radial position offset based on the comprehensive rigid load phase mapping function. Physically, the high load during the rigid cutting stage corresponds to a significant material abrupt change in allowance in that region. To maintain sufficient polishing contact stress in these high-allowance regions, the system needs to actively increase the radial compression depth of the flexible polishing tool. The edge computing unit extracts the current value of the comprehensive rigid load phase mapping function according to the synchronous phase angle of the spindle, and performs a proportional product operation with a preset force-position equivalent conversion coefficient. The product result is directly used as the radial position offset to compensate for the normal direction of the basic interpolation trajectory. In this embodiment, the value of the force-position equivalent conversion coefficient depends on the overall elastic stiffness of the currently clamped flexible polishing tool. Its specific value is usually obtained by process engineers through offline tool compression deformation calibration experiments and pre-configured in the system parameters. Through this linear compensation mechanism, the CNC system can automatically lower the tool position in phase regions with abundant material, using the tool's elastic deformation to offset the interference differences caused by the abrupt change in allowance.

[0078] In step S312, the edge computing unit synchronously calculates the polishing feed rate modulation. Besides adjusting the radial depth of cut to increase physical contact force, the material removal rate also depends on the physical residence time of the polishing tool in the processing area. For interference peak ranges with large material allowances, actively reducing the feed rate of the linear feed axis can effectively extend the polishing grinding time. To precisely control the deceleration magnitude, the edge computing unit generates a pre-compensated feed rate that is inversely modulated with the rigid load. The mathematical relationship followed by this modulation logic is:

[0079] ;

[0080] In the formula, To pre-compensate the feed rate; This is the nominal polishing feed rate; This is a phase mapping function for a comprehensive rigid load; This is the velocity suppression gain coefficient; This represents the minimum feed rate threshold. In the above formula, a maximum value function is introduced. This is to establish a physical lower limit protection mechanism at the bottom layer to prevent the compensation speed calculated at high load extreme points from dropping to zero or a negative value, thereby avoiding the phenomenon of local surface overheating or stopping in place during the interpolation process of multi-axis linkage CNC machine tools.

[0081] In step S313, the edge computing unit reconstructs and distributes the multi-axis spatial interpolation trajectory. After obtaining the radial position offset and pre-compensated feed speed sequence in the spatial phase domain, the edge computing unit synchronously superimposes the above-mentioned compensation components onto the original polishing toolpath data model, generating a multi-axis linkage drive command superimposed with the feedforward pre-compensation. This command is transmitted to the CNC system via the industrial communication bus. The underlying servo driver of the multi-axis linkage CNC machine tool receives the reconstruction command and drives the linear feed axis and rotary axis to perform a smooth polishing operation according to the dynamically changing depth and speed. For the specific underlying command parsing algorithm of multi-axis linkage spatial trajectory interpolation, those skilled in the art can use conventional non-uniform rational B-spline curve interpolation or micro-straight line segment smooth transition interpolation methods. The specific motion control implementation mechanism is a well-known technology in the field and will not be described in detail here.

[0082] During the execution of the flexible polishing tool machining trajectory by the multi-axis CNC machine tool according to the aforementioned pre-compensation instructions, the system simultaneously activates a data feedback monitoring mechanism. Since the pre-compensation trajectory is a theoretical derivation based on rigid cutting data, whether its compensation effect in actual execution meets the real requirements of the physical surface typically requires closed-loop verification through real-time tribological impedance. Therefore, the edge computing unit synchronously constructs a flexible tribological load phase mapping function characterizing the current polishing physical feedback. In this embodiment, the specific synchronous construction process includes the following steps.

[0083] Step S321: The edge computing unit performs real-time acquisition and net value extraction of servo status parameters. The edge computing unit reuses the hardware-level hard real-time alignment mechanism established during the system initialization phase. During the polishing process, the edge computing unit acquires the servo operating status parameters of the underlying servo driver in real time through the industrial communication bus. Considering the significant differences between the physical mass and air damping coefficient of flexible polishing tools and rigid chamfering tools, as a preferred approach, the system performs a no-load benchmark test specifically for the polishing process after automatic tool change. The multi-axis linkage CNC machine tool performs polishing no-load operation according to a predetermined spatial trajectory. The edge computing unit records the polishing no-load current and subtracts this benchmark value from the total data acquired in real time. Specifically, the edge computing unit subtracts the spindle no-load current from the total spindle torque current to obtain the net value of the polishing spindle torque current; simultaneously, after deducting the corresponding no-load benchmark from the thrust current of each linear feed axis participating in interpolation, the thrust current is vectorized by taking the square root of the sum of squares to obtain the net value of the composite thrust current of the polishing feed axis. In addition, based on the difference between the underlying instruction position and the actual encoder position, the position following error in the polishing stage is also synthesized by taking the square root of the sum of squares.

[0084] Step S322: The edge computing unit performs filtering and dimensionless normalization mapping on the polishing feature parameters. Compared with rigid cutting, the frictional resistance fluctuations in the flexible polishing process are much weaker, and its signal is often susceptible to high-frequency electrical noise interference from the servo drive system. In this embodiment, the edge computing unit performs low-pass filtering on the extracted polishing net parameter sequence to filter out electromagnetic interference that is not within the machine tool's mechanical frequency band. To unify the physical dimensions of the parameters, which differ greatly, the edge computing unit continues to use an extreme value normalization algorithm. By calculating the difference between the current data point and the sequence minimum value, and dividing by the range between the sequence maximum and minimum values, the filtered polishing parameters are uniformly mapped to... Within the dimensionless interval. After numerical transformation, the edge computing unit obtains a normalized polishing parameter sequence for feature synthesis. For the specific hardware implementation of discrete digital filtering, those skilled in the art can use conventional field-programmable gate arrays or digital signal processors to build it. Its internal calculation and execution logic is well-known in the field and will not be described in detail here.

[0085] In step S323, the edge computing unit performs a weighted construction of the flexible friction load phase mapping function. At the physical level, the working mechanism of the flexible polishing tool is surface contact friction rather than rigid point contact cutting, which causes a change in the energy distribution ratio of cutting resistance on the spindle and linear feed axis. To accurately reproduce the flexible machining characteristics, the CNC system is configured with flexible spindle weights, flexible feed weights, and flexible error weights independent of rigid cutting. It should be noted that the above flexible weights also follow the allocation constraint that the sum is 1, and their specific values ​​are pre-set in the CNC system by the process engineers based on the energy ratio measurement results of offline polishing friction tests. The edge computing unit calculates the weights based on the synchronous phase angle of the spindle rotation axis. The normalized polishing feature parameters of each dimension are weighted and summed according to the specific weights mentioned above. The edge computing unit thus constructs a flexible friction load phase mapping function in univariate form, denoted as... .

[0086] The aforementioned construction process is completed synchronously in a multi-threaded background during the polishing operation. The generated flexible friction load phase mapping function intuitively records the actual physical friction resistance encountered by the flexible polishing tool at each circumferential phase point. This data model is temporarily stored in the memory space of the edge computing unit, providing the necessary real-time comparison benchmark for the system to perform micro-failure diagnosis and macro-wear assessment within the subsequent processing auxiliary time window.

[0087] After completing the aforementioned multi-source data acquisition and mapping function construction steps, the multi-axis CNC machine tool enters the auxiliary processing time window for workpiece loading and unloading. At this time, the edge computing unit initiates the dual-process load feature cross-decoupling algorithm. In long-cycle polishing operations, the abrasive unit or bristle material of the flexible polishing tool gradually undergoes macroscopic volume wear. This wear manifests physically as a decrease in the overall elastic stiffness of the tool and a general attenuation of global frictional resistance. To eliminate the interference of environmental noise in a single process and accurately quantify this physical failure state, the edge computing unit uses a rigid load as a reference benchmark and executes a macroscopic wear index evaluation mechanism based on global integration. The specific processing procedure in this embodiment includes the following steps.

[0088] Step S411: The edge computing unit calculates the global integral work done by the flexible friction load. To quantify the overall frictional energy output by the flexible polishing tool within the current basic analysis cycle, the edge computing unit maps the flexible friction load phase function stored in memory. Throughout A definite integral is performed within the phase interval. The resulting integral area characterizes the actual global frictional work done in the polishing process, and is denoted as... For specific numerical solutions to discrete sequence definite integrals, those skilled in the art can implement them using the trapezoidal rule or Simpson's rule. The numerical analysis process is well-known in the field and will not be elaborated here.

[0089] In step S412, the edge computing unit synchronously calculates the global integral work done under the rigid cutting load. Since the basic cutting allowance varies between different batches of stud blanks, the absolute magnitude of the polishing resistance is not only related to tool wear but also directly affected by the volume of residual material from the previous process. Therefore, the system extracts the theoretically required baseline resistance. The edge computing unit maps the phase function of the comprehensive rigid load. In the same phase interval Perform the same definite integral operation within the inner region to obtain the rigid integral area characterizing the global material interference margin, denoted as .

[0090] Step S413: The edge computing unit cross-decouples to generate the macroscopic wear index. After obtaining the actual friction work and the theoretical material allowance, the edge computing unit constructs a ratio model between the two to eliminate the interference of blank allowance fluctuations on wear determination. Considering that in actual production, if the blank allowance is extremely small or the machine tool experiences abnormal working conditions such as idle movement, the rigid integral area of ​​the denominator may approach zero, thus causing program overflow. To avoid such algorithmic dead zones, as a preferred approach, the system introduces a very small positive bias constant in the denominator of the ratio calculation. The edge computing unit calculates the macroscopic wear index. The mathematical relationship it follows is:

[0091] ;

[0092] In the formula, The actual global friction work done; The area is a rigid integral. The reference friction conversion coefficient; This is a positive bias constant. In this embodiment, the specific value of the reference friction conversion coefficient is usually determined by the process engineer during the first few standard machining cycles when the new tool is put into use, by recording... and The average ratio is used to calibrate and is then embedded into the configuration parameters of the edge computing unit. Simultaneously, a positive bias constant is used. The specific value is usually set based on the lower limit of the floating-point arithmetic precision of the underlying processor of the edge computing unit, for example, a value of , It can provide a low-level division-to-zero protection mechanism without affecting the accuracy of normal calculations.

[0093] Step S414: The edge computing unit performs status determination and early warning output based on the wear index. As the service life of the flexible polishing tool increases, the actual frictional work it can provide with the same material allowance also increases. Gradually decreasing, leading to the calculated The wear index shows a downward trend. The edge computing unit compares the calculated macroscopic wear index with the wear judgment threshold set in the system. This judgment threshold is usually reverse-calibrated by technicians through roughness testing experiments. That is, when the surface roughness of the machined workpiece approaches the upper limit allowed by the process, the corresponding macroscopic wear index is recorded and set as a warning threshold and stored in the system in advance. When the macroscopic wear index is found to be lower than the judgment threshold for several consecutive processing cycles, the edge computing unit determines that the flexible polishing tool has reached the macroscopic wear limit. Subsequently, the edge computing unit sends a tool life end signal to the CNC system through the industrial communication bus, triggering the machine tool alarm shutdown or driving the automatic tool changer to replace the spare tool, thereby ensuring the surface polishing quality of subsequent workpieces.

[0094] In addition to performing global macroscopic wear assessment, in actual machining scenarios, flexible polishing tools are prone to localized brush filament plastic deformation or abrasive element spalling due to instantaneous overload when machining high-interference areas such as thread terminations. These microscopic failures typically do not cause a significant attenuation of global frictional energy, but directly result in incomplete removal of surface burrs at specific phases. To accurately identify these highly concealed failure modes, the edge computing unit introduces a local transfer coefficient evaluation mechanism based on peak interval correlation. The specific processing procedure in this embodiment includes the following steps.

[0095] In step S421, the edge computing unit performs spatial phase truncation of the local feature sequence. The edge computing unit receives the main peak phase boundary output by the asymmetric material interference peak interval optimization mechanism. Based on this physical location boundary, the edge computing unit extracts the comprehensive rigid load phase mapping function and the flexible friction load phase mapping function from the memory space, respectively, and obtains the local rigid cutting load sequence. and local flexible frictional load sequence ,in The two local sequences obtained by post-processing in terms of time sequence correspond precisely in terms of spatial phase to the core region of maximum material interference at the end of the physical workpiece.

[0096] Step S422: The edge computing unit calculates the morphological correlation of the local sequences. In normal force-equivalent compensation machining, the waveform fluctuations of the flexible friction load in the peak region should maintain a high degree of morphological synchronization with the waveform characteristics of the rigid cutting load. When the tool locally undergoes physical yielding, this morphological synchronization is disrupted, usually resulting in the flexible load waveform becoming abnormally flat. As a preferred approach, the edge computing unit uses the Pearson correlation coefficient algorithm to perform morphological analysis on the two local sequences. Specifically, the system calculates the covariance of the local rigid sequence and the local flexible sequence, and divides it by the product of their standard deviations to generate a dimensionless local morphological correlation index. To prevent overflow of low-level operations due to the sequence standard deviation approaching zero when there are abnormal dead zones in the signal segment, the system adds a very small tolerance constant to the standard deviation product term, such as the smallest floating-point value supported by the edge computing unit processor. For the conventional discrete numerical calculation process of sequence covariance and standard deviation, those skilled in the art can use general mathematical statistical logic for code deployment, which is a well-known technique in the field and will not be elaborated upon here.

[0097] Step S423: The edge computing unit synchronously calculates the local work transfer coefficient in the core peak region. Since a single morphological correlation cannot fully reflect the actual attenuation of contact stress, the edge computing unit calculates the local work transfer coefficient of the two intercepted local sequences. Discrete integrals are performed and summed within the interval to obtain the local flexible friction work and the local rigid cutting interference. The edge calculation unit uses the local flexible friction work as the numerator and the product of the local rigid cutting interference and a preset proportionality coefficient as the denominator, and performs a ratio calculation to obtain the local work transfer coefficient. In this embodiment, the aforementioned preset proportional coefficient is typically determined by technicians during the initial production phase of a defect-free flexible polishing tool by measuring the ratio of local work done to local interference in several standard samples and taking the arithmetic mean for benchmark calibration. Similarly, to ensure the integrity of the algorithm logic under extreme conditions, the denominator term of this ratio calculation is also protected by a preset positive bias constant for division by zero.

[0098] In step S424, the edge computing unit performs multi-dimensional fusion judgment and micro-failure diagnosis output. The edge computing unit logically compares the calculated local morphology correlation index and local work transfer coefficient with the correlation diagnosis threshold and transfer coefficient diagnosis threshold configured within the system, respectively. In this step, the two diagnostic thresholds are typically set and stored in the CNC system register by process engineers using a pre-manufactured local defect calibration tool for destructive cutting tests. The threshold values ​​at which the failure characteristics abruptly change are recorded, with a safety margin allowed. When the local morphology correlation index is detected to be lower than the correlation diagnosis threshold, and the local work transfer coefficient is simultaneously lower than the transfer coefficient diagnosis threshold, the edge computing unit determines that the flexible polishing tool may have experienced a structural micro-failure in the corresponding phase region. The system then generates a list containing the specific occurrence phase. The local defect early warning record provides accurate spatial coordinates for subsequent manual re-inspection intervention or triggering the machine tool to start the local fixed-point re-polishing process.

[0099] After confirming the existence of micro-failures in a specific phase region of the flexible polishing tool through a local transfer coefficient evaluation mechanism, the system further classifies and diagnoses the specific physical modes of these failures. In actual production environments, micro-failures of polishing tools often manifest as plastic collapse deformation of the bristles or brittle spalling of abrasive units. The tool repair strategies and the urgency of machine tool intervention differ depending on the failure mode. The edge computing unit performs micro-failure mode classification by analyzing the temporal morphological characteristics of the local flexible friction load. The specific processing procedure in this embodiment includes the following steps.

[0100] Step S431: The edge computing unit calculates the dynamic gradient variance of the local flexible friction load sequence. At the physical contact level, if the polishing tool undergoes brittle spalling, the irregular edges of the fracture surface typically induce high-frequency abrupt changes in resistance when scraping the workpiece surface; conversely, if plastic collapse occurs, the contact surface between the tool and the workpiece becomes smooth and flat, and the change in frictional resistance tends to be slower. To quantify this dynamic characteristic change, the edge computing unit calculates the dynamic gradient variance of the local flexible friction load sequence stored in memory. Perform a first-order forward difference operation to obtain the local gradient sequence. Based on this, the edge computing unit calculates the statistical variance of the local gradient sequence and uses it as the dynamic gradient variance characterizing the intensity of high-frequency fluctuations in the signal. In this embodiment, the specific numerical solution method for the discrete sequence difference and variance can be implemented by those skilled in the art using conventional statistical algorithms. The underlying calculation logic is a well-known technology in this field and will not be described in detail here.

[0101] Step S432: The edge computing unit extracts the peak energy concentration of the local load. To further distinguish the local stress patterns, the edge computing unit extracts the peak energy concentration from the local flexible friction load sequence. Extract the maximum absolute load value Simultaneously calculate the arithmetic mean load value of the local sequence within a specific interval. As a preferred method, the system divides the maximum absolute load value by the arithmetic mean load value, and uses this ratio to generate a dimensionless peak energy concentration. To prevent the local average load from approaching zero due to uninterrupted operation or signal loss, which could lead to a division-by-zero dead zone in the algorithm, the system adds a positive bias constant based on the lower limit of the processor's computational precision to the arithmetic average load value of the denominator. The aforementioned bias constants are typically set to a value of [value missing]. The fractional part is used to ensure the continuity of the underlying division operation without interfering with the normal numerical proportions. This concentration index intuitively reflects the degree of concentration of local contact stress at the main wave peak.

[0102] Step S433: The edge computing unit performs classification based on multi-dimensional feature thresholds. After acquiring the two feature indicators mentioned above, the edge computing unit logically compares them with the baseline thresholds configured in the system. Specifically, the CNC system has preset gradient mutation thresholds, gradient smoothing thresholds, and concentration lower limit thresholds. These judgment thresholds are usually set by technicians through comparative cutting tests using scrap tool samples from different historical failure modes, extracting the physical boundaries of feature parameters, and reserving a certain safety tolerance. When the edge computing unit detects the calculated dynamic gradient variance... When the gradient abrupt change threshold exceeds the threshold mentioned above, the edge computing unit classifies the failure mode of the flexible polishing tool in the corresponding phase interval as brittle spalling of the abrasive unit. Conversely, when the dynamic gradient variance is less than the gradient smoothing threshold, and the peak energy concentration is calculated synchronously... When the concentration falls below the lower threshold, it indicates that the local impedance waveform is abnormally flat and lacks necessary peak support. The edge computing unit then classifies the failure mode as plastic collapse deformation of the brush bristle material. To ensure the completeness of the judgment logic and avoid algorithm dead zones, if the calculated feature parameters fall within the boundary overlap region between the aforementioned thresholds, the edge computing unit marks it as an unclassified general micro-failure warning, and the operator performs a manual offline re-inspection.

[0103] In step S434, the edge computing unit outputs the classification and diagnostic results and triggers targeted maintenance commands. After completing the mode determination, the edge computing unit packages the identified specific micro-failure modes along with the phase coordinate boundaries to generate a diagnostic log, which is then sent to the CNC system via the industrial communication bus. The multi-axis CNC machine tool executes differentiated follow-up actions based on the received diagnostic results. For example, when brittle spalling is identified, since the falling abrasive residue may scratch the workpiece surface, the CNC system will immediately trigger a spindle emergency stop command and issue an audible and visual alarm, requiring manual intervention to clean the worktable. When plastic collapse is identified, since it mainly affects polishing efficiency rather than directly causing structural damage, the system can schedule the automatic tool changer to replace the spare flexible polishing tool after the current workpiece is processed, or temporarily increase the radial pre-compensation depth of the tool when processing similar workpieces in subsequent processes, thereby maintaining the continuity of automated production while ensuring product yield.

[0104] In deburring and polishing operations on complex contours, the cutting tool experiences localized transient cutting loads when a multi-axis CNC machine tool passes through areas with abrupt changes in material allowance, such as thread finishing. Traditional control strategies typically alleviate this load by reducing the feed rate of the linear axis. However, this approach not only sacrifices overall machining efficiency but also easily induces localized chatter at low speeds. To improve this situation, the system introduces a tool axis vector deflection and dimensionality-reduction adaptive control mechanism based on dynamic constraints. The edge computing unit disperses transient cutting forces by dynamically adjusting the tool's cutting contact posture. In this embodiment, the specific processing steps for mapping the tool axis vector deflection based on the load gradient include the following steps.

[0105] Step S511: The edge computing unit calculates the local spatial gradient of the composite rigid load. At the physical machining level, the rate of change of material allowance directly determines the transient force gradient during tool entry or exit. The edge computing unit extracts the composite rigid load phase mapping function from memory and performs a first-order backward difference operation along the spatial phase domain. Thus, the edge computing unit obtains a load gradient sequence characterizing the rate of change of allowance. For the specific numerical solution algorithm of the discrete sequence difference, those skilled in the art can use central difference or higher-order finite difference rules for code implementation; the numerical analysis process is well-known in the field and will not be elaborated here.

[0106] In step S512, the edge computing unit performs a nonlinear mapping from the load gradient to the basic tool axis deflection angle. When a positive surge in the load gradient is detected, it indicates that the tool is about to enter a high interference region. At this time, the guide tool axis vector tilts at a certain angle along the cutting feed direction, which can change the effective contact area of ​​the tool side edge, thereby reducing the physical contact stress per unit area. As a preferred approach, the edge computing unit constructs the basic deflection angle mapping logic. To prevent frequent tool axis oscillations caused by small system load fluctuations, this mapping logic integrates a dead zone filtering mechanism. Specifically, the edge computing unit compares the absolute value of the calculated load gradient at the current phase with a preset load gradient dead zone threshold. When the absolute value of the load gradient is less than or equal to the dead zone threshold, it is determined that the current cutting noise range is within the normal range, and the output base deflection angle is zero. When the load gradient is greater than the positive dead zone threshold, the system multiplies the difference between the load gradient and the dead zone threshold by a preset attitude conversion gain coefficient to obtain the positive base deflection angle. Similarly, when the load gradient is less than the negative dead zone threshold, the sum of the load gradient and the dead zone threshold is multiplied by the attitude conversion gain coefficient to calculate the negative base deflection angle. In this embodiment, the specific value of the attitude conversion gain coefficient is usually pre-calibrated by the process engineer based on the geometric parameters of the currently clamped tool and the response characteristics of the machine tool's rotary axis, and stored in the configuration register of the edge computing unit. At the same time, the setting of the load gradient dead zone threshold is usually based on the extreme value of the load noise floor collected during the machine tool's idling state to ensure that weak signal disturbances are not incorrectly converted into motion commands.

[0107] In step S513, the edge computing unit performs dynamic limiting on the basic deflection angle based on the physical limits of the machine tool. Since the rotary axes of a multi-axis CNC machine tool are limited by the rated torque output of the underlying servo driver, their maximum physical angular velocity and angular acceleration have upper limits. Directly issuing the calculated basic deflection angle could easily lead to excessive servo following error or even trigger an overload alarm on the driver. The edge computing unit performs low-pass filtering and amplitude truncation on the basic deflection angle sequence based on the pre-configured maximum allowable angular velocity and angular acceleration parameters of the rotary axes in the CNC system. The system thus generates the final smooth deflection angle, ensuring that the differential rate of change of this angle always remains within the allowable kinematic constraints of the underlying hardware.

[0108] In step S514, the edge computing unit reconstructs the spatial tool axis vector and issues pose commands. After obtaining a smooth deflection angle that satisfies the dynamic constraints, the edge computing unit performs a matrix rotation transformation with the basic tool axis vector set in the original machining program to synthesize a dynamic tool axis vector command with adaptive avoidance capability. To ensure the consistency of the deflection direction and cutting safety, the above rotation transformation uses the local machining plane formed by the feed direction vector of the current tool trajectory and the basic tool axis vector as a reference, driving the tool axis to deflect around the normal vector perpendicular to the reference plane by the smooth deflection angle. The system sends the reconstructed multi-axis linkage pose data to the CNC system in real time through the industrial communication bus. The underlying servo driver drives the linear feed axis and rotary axis to move in coordination according to the new commands, enabling the tool to complete the machining operation with a more stable posture when facing sudden changes in the allowance, thereby ensuring the physical consistency of the machined surface.

[0109] After generating the aforementioned dynamic tool axis vector command with adaptive avoidance capability, the system needs to further evaluate the actual execution capability of the underlying hardware. In multi-axis CNC machine tools, the rotary axis is limited by the physical inertia of the mechanical transmission mechanism, and its dynamic response limit is usually lower than that of the linear feed axis. If the original linear feed speed is maintained when the posture deviates significantly, the rotary axis will not be able to reach the target pose within the specified interpolation cycle, thus causing physical following error in the spatial contour. To avoid such physical overload problems, the edge computing unit executes a rotary axis kinematic time bottleneck evaluation and physical limit solution mechanism. In this embodiment, the process includes the following steps.

[0110] Step S521: The edge computing unit extracts the kinematic increment of the current spatial interpolation micro-segment. Based on the machine coordinate system data within the current interpolation cycle, the edge computing unit extracts the composite linear displacement increment of the linear feed axis and the angular displacement increment of the rotary axis. For a five-axis linkage operation with multiple rotary axes, the system extracts the angular displacement increment of each rotary axis and uses the increment with the largest absolute value as the basis parameter for subsequent evaluation to ensure coverage of the most stringent machine tool kinematic constraints.

[0111] Step S522: The edge computing unit calculates the physical limit time required for the rotary axis to reach the target position. To quantify the underlying execution capability of the rotary axis, the edge computing unit calculates the shortest time constrained by the maximum dynamic parameters under the current angular displacement increment, based on the trapezoidal velocity planning model set inside the machine tool. Specifically, the edge computing unit calculates the critical angular displacement required for the rotary axis to reach the maximum allowable angular velocity. This critical value is equal to the square of the maximum allowable angular velocity divided by the maximum allowable angular acceleration. When the absolute value of the extracted angular displacement increment is less than or equal to the critical angular displacement, it indicates that the rotary axis can complete the displacement action during the acceleration and deceleration phases. The system calculates the square root of the ratio of the absolute value of the angular displacement increment to the maximum allowable angular acceleration and multiplies it by two to obtain the physical limit time. Conversely, when the absolute value of the angular displacement increment is greater than the critical angular displacement, it indicates that the rotary axis motion process includes an intermediate uniform velocity phase. The system divides the absolute value of the angular displacement increment by the maximum allowable angular velocity, adds the ratio of the maximum allowable angular velocity to the maximum allowable angular acceleration, and uses this sum as the physical limit time. As a preferred approach, the two underlying dynamic limit parameters mentioned above are read and fixed by the system during the initialization phase from the motor nameplate parameters and transmission ratio configuration table of the underlying servo driver, ensuring that the calculation benchmark conforms to the actual mechanical work capacity of the machine tool.

[0112] Step S523: The edge computing unit calculates the theoretical feed time and performs a bottleneck comparison. The edge computing unit divides the extracted linear displacement increment by the set feed rate specified in the original machining program to obtain the theoretical feed time. To prevent the division operation from overflowing the dead zone due to the set feed rate being extremely small or zero, the system compares the set feed rate with the minimum safe feed rate configured in the machine tool's underlying system before performing the division operation. If the original set value is lower than the safe threshold, the minimum safe feed rate is forcibly used as the denominator. Subsequently, the edge computing unit performs a logical comparison between the calculated physical limit time and the theoretical feed time. Considering the discretization characteristics of the underlying clock cycle of the CNC system, direct comparison of floating-point numbers is prone to frequent misjudgments due to extremely small truncation errors. To avoid logical dead zones, the system adds a time tolerance constant to the theoretical feed time. This time tolerance constant is usually set to one or two times the duration of the current CNC system's underlying interpolation cycle. When the calculated physical limit time is greater than the theoretical feed time after superimposed tolerance, the edge computing unit determines that the rotary axis has triggered a kinematic time bottleneck in the current spatial micro-segment, indicating that the original linear feed speed has exceeded the physical following limit of the machine tool attitude deflection.

[0113] Through the aforementioned evaluation logic, the edge computing unit predicts the coordination bottlenecks of each control axis of the machine tool before the actual physical action occurs. This prediction result and its accompanying time parameters are temporarily stored in the system memory queue, providing a quantified constraint boundary for subsequent algorithms to adaptively reduce the linear axis feed rate to match the response capability of the rotary axis.

[0114] After predicting the kinematic time bottleneck of the rotary axis, the edge computing unit forcibly reduces the motion speed of the linear feed axis for the spatial interpolation micro-segment that triggers the time bottleneck. This operation causes the physical time consumption of the linear feed axis to proactively compromise with the extreme response time of the rotary axis, thereby maintaining the spatial coordination of each motion axis within the underlying servo cycle. In this embodiment, the specific processing steps for forced dimensionality reduction reconstruction of feed speed and compliant machining execution include the following steps.

[0115] In step S531, the edge computing unit performs a feed rate reduction calculation based on limit constraints. In the synchronous following control of the spatial trajectory, the system needs to reallocate speed resources according to the time constraints of the underlying transmission mechanism. Specifically, the edge computing unit uses the linear displacement increment of the linear feed axis within the current micro-segment as the numerator, and the sum of the previously solved physical limit time of the rotation axis and the time offset constant used to prevent computational overflow as the denominator, performing a ratio calculation. Through this time-for-space reduction strategy, the system calculates the reconstructed feed rate constrained by the rotation axis dynamics bottleneck. As a preferred approach, this offset constant is typically set to the minimum scale value of the underlying controller clock accuracy, for example... The time limit is on the order of seconds to ensure that the algorithm does not generate a division-to-zero dead zone when the linear displacement increment is extremely small or the time limit approaches zero.

[0116] In step S532, the edge computing unit performs time-domain smoothing filtering on the reconstructed feed speed. In discrete CNC interpolation cycles, directly issuing the forced dimensionality-reduced speed setpoint can easily generate step signals between adjacent micro-segments. This sudden speed change not only excites resonance in the machine tool's mechanical structure but also increases the transient current load on the servo driver 141. Therefore, the edge computing unit introduces a first-order inertial low-pass filtering algorithm to smooth the discrete reconstructed feed speed sequence. During the algorithm's execution, the system multiplies the reconstructed feed speed calculated for the current micro-segment by a first weighting coefficient and sums it with the smoothed feed speed output from the previous interpolation cycle multiplied by a second weighting coefficient, thereby generating a smoothly transitioning current smoothed feed speed, where the sum of the first and second weighting coefficients is always 1. In this embodiment, the specific values ​​of the two weighting coefficients are determined by a preset filtering time constant, which is typically calibrated by the debugging personnel based on the actual response bandwidth of the machine tool's underlying servo driver and stored in the CNC system. For the discrete numerical solution logic of first-order inertial filtering, those skilled in the art can use conventional digital signal processing methods to implement the code. The underlying algorithm principle is a well-known technology in this field and will not be elaborated here.

[0117] In step S533, the edge computing unit performs synchronous reassembly and distribution of multi-axis linkage commands. After acquiring the smoothed feed rate and the dynamic tool axis deflection vector generated in the previous step, the edge computing unit packages the control variables of these two dimensions into data. The system encapsulates the originally isolated feed reduction command and attitude avoidance command into a synchronous data frame that conforms to the machine tool's underlying communication protocol. Subsequently, the edge computing unit transmits the reassembled interpolation command to the underlying buffer of the CNC system in real time via the industrial communication bus.

[0118] In step S534, the CNC system drives the physical machine tool to perform adaptive compliant machining. Based on the received recombination interpolation command, the underlying servo driver synchronously outputs torque current to the drive motors of the linear feed axis and rotary axis. During this physical execution phase, when facing areas with complex contours or abrupt changes in material allowance, the multi-axis CNC machine tool can actively disperse local cutting interference stress by tilting and deflecting the tool axis vector, and simultaneously maintain motion transition at a reduced feed rate that adapts to the physical limits of the rotary axis. This multi-dimensional adaptive control mechanism effectively avoids the problems of following deviations and local overloads under traditional rigid machining strategies, ensuring the physical consistency and operational safety of the polishing process in multi-axis linkage mode.

[0119] Specific application examples:

[0120] To verify the feasibility of the technical solution of this application, the following section, combining a specific set of machining physical parameters and the aforementioned formulas, fully deduces the data processing and control logic execution process of the system during actual machining. The current machining task is set as the end chamfering and polishing of an M12 standard stud, and the system's nominal polishing feed rate is set to... Minimum feed rate threshold .

[0121] 1. System initialization and rigid cutting data extraction (in conjunction with appendix) Figure 3 Feature analysis)

[0122] In rigid cutting operations, the spindle rotates carrying a rigid chamfering tool. The spindle's rotation axis reaches the core phase angle within the thread finishing abrupt change allowance range. At that time, the edge computing unit triggers hard real-time latch, obtains the underlying net parameters, and normalizes them to obtain: normalized spindle current. Normalized feed current Normalized following error The preset machine tool rigidity weighting coefficients within the CNC system are as follows: spindle current weighting...

[0123] Feed current weighting Follow-up error weights Edge computing units invoke the weighted synthesis formula:

[0124] ;

[0125] Substitute the data to calculate the comprehensive rigid load value:

[0126] .

[0127] The system uses an optimization mechanism to identify the horizontal axis between 120 and 160 degrees. If the value is continuously greater than the dynamic threshold, the interval is established as such. Figure 3 The light gray area represents the "thread end abrupt change margin range".

[0128] 2. Pre-compensation and flexible polishing feature acquisition (in conjunction with appendix) Figure 3 Feature analysis)

[0129] The machine tool's automatic tool changer uses a flexible nylon polishing brush for the polishing process. To overcome abrupt changes in margin, the system performs bidirectional feedforward compensation for radial position and velocity.

[0130] The force-potential equivalent transformation coefficient is set as follows: Then the radial position offset of that phase point The machine tool actively presses down at this point. Set the speed suppression gain coefficient. Calculate the pre-compensated feed rate using the speed modulation formula.

[0131] ;

[0132] Substitute the data into the calculation:

[0133] The machine tool successfully and smoothly reduced the feed rate in this high interference region. .

[0134] 3. Dual-process load cross-decoupling and failure diagnosis

[0135] After processing the 200th workpiece continuously, the edge computing unit initiates diagnostic calculations for the auxiliary time window.

[0136] (1) Macroscopic wear index assessment: The system assesses the overall wear index of the entire system. arrive Integrate over the interval to obtain the global frictional work. rigid integral area Let the reference friction conversion coefficient be... Positive bias constant .

[0137] Call the macroscopic wear formula:

[0138] .

[0139] The wear threshold is set to 0.8. The system determined that the overall wear limit had not been reached.

[0140] (2) Microscopic failure diagnosis: Local calculations were performed on the peak range of 120 to 160 degrees. The local morphological correlation index was calculated to be 0.45 using the Pearson correlation coefficient formula (below the set diagnostic threshold of 0.60). The local integral was further calculated to obtain the local flexible friction work. Local rigid cutting interference Preset proportional coefficient .

[0141] Call the formula for the local work transfer coefficient:

[0142] .

[0143] Since both the local morphological correlation index (0.45 < 0.60) and the local work transfer coefficient (0.50 < the set transfer threshold of 0.65) are below the threshold, and the energy concentration of the peak in this segment is low, the system accurately verifies and determines that the polishing brush filaments plastically collapse at this specific angle.

[0144] 4. Trajectory Adaptive Reconstruction and Compliance Processing

[0145] To address the aforementioned micro-failures, when processing the 201st workpiece, the system generates an avoidance posture command with a basic deflection angle of 5 degrees based on load space gradient sequence difference calculation.

[0146] Entering the limit time evaluation: extracting the linear displacement increment of this micro-segment By combining the maximum angular velocity and angular acceleration of the rotation axis, the physical limit time to complete the 5-degree deflection can be calculated. Theoretical feed time Less than This triggers a kinematic bottleneck. The system invokes the feed rate dimensionality reduction formula (assuming a constant time bias). :

[0147] ;

[0148] Then, a first-order inertial low-pass filter is applied (with a current weight of 0.4 and a historical weight of 0.6):

[0149] Smoothed feed rate The underlying servo driver reorganizes the instructions accordingly, and while issuing a 5-degree deflection angle, adaptively reduces the instantaneous feed rate of that segment. It perfectly matches the dynamic limits of the machine tool's underlying structure, completely avoiding servo tracking over-tolerance alarms.

[0150] Experimental verification and effect comparison (with appendix) Figure 4 Performance Comparison Analysis:

[0151] To verify the actual technical effect of the present invention, a comparative experiment was conducted on 500 studs of the same model under the same working conditions, comparing the multi-axis linkage adaptive machining method of the present invention with the traditional constant speed isobaric multi-axis polishing method. The data characteristics are shown in the attached figure. Figure 4 A visually intuitive display.

[0152] Appendix Figure 4 Coordinate system explanation: The horizontal axis (X-axis) lists the four physical indicators used in the comparative evaluation. To ensure a reasonable display on the vertical axis within the same numerical order of magnitude: the units for the first item (burr residual height) and the third item (maximum following error) are micrometers (µm); the second item (surface roughness) has its original value magnified 10 times, and the unit is labeled as 1 / 10 micrometer; the fourth item (tool life) has its original lifespan reduced by a factor of 10, and the unit is labeled as 10 pieces. The vertical axis (Y-axis) represents the measured values ​​of the test indicators, indicating the specific numerical values ​​recorded for each corresponding test item, distributed within the range of 0 to 55.

[0153] Appendix Figure 4 Legend and numerical analysis: In the figure, the light gray bars represent the data results produced by the traditional constant speed, isobaric multi-axis polishing method; the dark gray bars represent the data results produced by the adaptive multi-axis linkage polishing method of this invention. The numbers above the bars represent the precise measurement readings of the corresponding experimental items.

[0154] The specific experimental comparisons and conclusions are as follows:

[0155] The residual burr height in the thread termination area: the traditional method reads 42.5 μm, while the method of this invention reads 8.2 μm (a reduction of 80.7%).

[0156] Surface roughness at the end of the workpiece: The actual value of the traditional method is 1.25 μm (reading 12.5 in the figure), while the actual value of the method of the present invention is 0.38 μm (reading 3.8 in the figure, an improvement of 69.6%).

[0157] Maximum servo following error of the rotary axis: The reading of the traditional method is 35.4, while the reading of the method of this invention is 12.1 (peak low 65.8%).

[0158] Effective lifespan of flexible polishing tools: The actual value of the traditional method is 280 pieces / tool ​​(reading 28.0 in the figure), while the actual value of the method of this invention is 450 pieces / tool ​​(reading 45.0 in the figure, an extension of 60.7%).

[0159] Comprehensive analysis conclusions: as attached Figure 4 The columnar feature differences shown intuitively reflect that traditional methods are prone to physical displacement or insufficient elastic force when facing high interference zones such as thread termination, resulting in incomplete burr removal. In contrast, this invention, through comprehensive rigid load feedforward pre-compensation, causes a sharp decrease in burr residue. Simultaneously, the microscopic failure diagnosis and dimension-reduced adaptive tool axis deflection strategy introduced in this invention successfully disperses localized concentrated frictional stress, protecting the polishing tool from the risk of localized instantaneous breakage and significantly increasing tool life. Furthermore, the inclusion of a predictive mechanism targeting kinematic time bottlenecks completely eliminates position tracking errors caused by physical overload, resulting in substantial improvements in machining performance across all measured dimensions of test indicators.

Claims

1. A multi-axis linkage machining method for chamfering and polishing the end of a stud, characterized in that, Includes the following steps: Execute system initialization and establish a real-time latching mechanism between the underlying servo driver and the edge computing unit inside the CNC system for servo operating status parameters and spindle phase angle; Rigid chamfering and feature extraction are performed, and the rigid chamfering tool is invoked to perform machining. Based on the synchronously acquired servo operating state parameters, a comprehensive rigid load phase mapping function is constructed, and the peak interval characterizing the material interference allowance distribution is extracted. The feedforward pre-compensation and flexible friction feature acquisition are performed, and a flexible polishing tool is replaced. The comprehensive rigid load phase mapping function is converted into a radial position offset and a pre-compensation feed speed to execute the pre-compensation polishing trajectory. Polishing feedback data is collected simultaneously to construct a flexible friction load phase mapping function.

2. The multi-axis linkage machining method according to claim 1, characterized in that, The steps for constructing the comprehensive rigid load phase mapping function include: The net value of spindle torque current, net value of linear feed axis composite thrust current, and spatial composite position following error of linear feed axis collected by the underlying servo driver are obtained and low-pass filtered. An extreme value normalization algorithm is used to map the filtered multidimensional bottom net parameters to a dimensionless interval of 0 to 1; Based on the spindle phase angle, and combined with the preset spindle current weight, feed current weight, and follower error weight, the normalized parameters are weighted and summed to construct the comprehensive rigid load phase mapping function.

3. The multi-axis linkage machining method according to claim 1, characterized in that, The step of extracting the peak intervals characterizing the interference margin distribution of the material includes: Calculate the arithmetic mean and standard deviation of the comprehensive rigid load phase mapping function within a preset basic analysis period, sum the arithmetic mean and the standard deviation by a preset multiple, and use the calculated sum as a dynamic peak value determination threshold. By comparing the load value of the comprehensive rigid load phase mapping function with the peak value determination threshold point by point along the main axis phase angle, continuous candidate interference intervals are extracted. The area integral of the load function sequence within each candidate interference interval is calculated, and the interval with the largest integral area is selected as the peak interval. The starting phase angle and ending phase angle corresponding to the peak interval are then output.

4. The multi-axis linkage machining method according to claim 1, characterized in that, The step of converting the comprehensive rigid load phase mapping function into radial position offset and pre-compensated feed rate includes: Extract the load value of the comprehensive rigid load phase mapping function under the target phase of the current interpolation cycle, and multiply the load value with the preset force-position equivalent conversion coefficient. Use the calculation result as the radial position offset amount to compensate to the normal direction of the basic interpolation trajectory within the preset original polishing toolpath data model. Extract the set nominal polishing feed rate, subtract the product of the load value and the preset speed suppression gain coefficient, and generate the pre-compensated feed rate by taking the maximum value function in combination with the preset minimum feed rate threshold. The radial position offset and the pre-compensated feed rate are superimposed on the original polishing toolpath data model to generate multi-axis linkage drive commands, which are then issued and executed.

5. The multi-axis linkage machining method according to claim 1, characterized in that, Also includes: A cross-decoupling calculation of the load characteristics of the two processes is performed. The phase mapping function of the comprehensive rigid load and the phase mapping function of the flexible friction load are compared and analyzed. The macroscopic wear index, local morphology correlation index and local work transfer coefficient are calculated to determine the physical failure mode of the flexible polishing tool.

6. The multi-axis linkage machining method according to claim 5, characterized in that, The step of calculating the macroscopic wear index to determine the physical failure mode of the flexible polishing tool includes: The flexible friction load phase mapping function and the comprehensive rigid load phase mapping function are respectively calculated by definite integrals in the phase interval from 0 to 2π to obtain the global friction work and the rigid integral area. The macroscopic wear index is generated by performing a ratio calculation using the global friction work as the numerator and the product of the rigid integral area and the preset reference friction conversion coefficient and the preset positive bias constant as the denominator. When the macro wear index is found to be lower than the set wear judgment threshold for multiple consecutive processing cycles, the physical failure mode of the flexible polishing tool is determined to have reached the macro wear limit.

7. The multi-axis linkage machining method according to claim 5, characterized in that, The step of calculating the local morphology correlation index and the local work transfer coefficient to determine the physical failure mode of the flexible polishing tool includes: Based on the peak interval, local rigid cutting load sequences and local flexible friction load sequences are extracted from the comprehensive rigid load phase mapping function and the flexible friction load phase mapping function, respectively. The covariance of the local rigid cutting load sequence and the local flexible friction load sequence is calculated, and the covariance is divided by the product of the standard deviations of the local rigid cutting load sequence and the local flexible friction load sequence to obtain the Pearson correlation coefficient as the local morphological correlation index. Discrete integrals are performed on the local rigid cutting load sequence and the local flexible friction load sequence respectively to obtain the local rigid cutting interference and the local flexible friction work, respectively. The local work done by the local flexible friction is divided by the product of the local rigid cutting interference and the preset proportional coefficient to generate the local work transfer coefficient. When the local morphological correlation index is lower than a preset correlation diagnosis threshold and the local work transfer coefficient is lower than a preset transfer coefficient diagnosis threshold, the physical failure mode of the flexible polishing tool in the corresponding phase region is determined to be microscopic failure.

8. The multi-axis linkage machining method according to claim 1, characterized in that, Also includes: During the execution of the pre-compensated polishing trajectory, a trajectory adaptive reconstruction based on dynamic constraints is performed. A dynamic tool axis deflection vector is generated based on the load gradient of the comprehensive rigid load phase mapping function. The physical limit time is evaluated in combination with the kinematic boundary parameters of the machine tool rotary axis. The feed rate command after forced reduction and correction is reconstructed and issued to complete the deflection avoidance machining.

9. The multi-axis linkage machining method according to claim 8, characterized in that, The step of generating the dynamic tool axis deflection vector includes: The load gradient sequence is obtained by performing a differential operation on the comprehensive rigid load phase mapping function; The absolute value of the load gradient sequence is compared with the set load gradient dead zone threshold. The difference exceeding the load gradient dead zone threshold is multiplied by the attitude conversion gain coefficient to generate the base deflection angle. Based on the maximum permissible angular velocity and maximum permissible angular acceleration of the machine tool's rotary axis, the basic deflection angle is filtered and its amplitude is truncated to generate a smooth deflection angle. The smooth deflection angle is then subjected to matrix rotation transformation with the basic tool axis vector of the current interpolation micro-segment to synthesize the dynamic tool axis deflection vector.

10. The multi-axis linkage machining method according to claim 9, characterized in that, The steps of reconstructing and issuing the forced reduction and correction of the feed rate command include: Extract the linear displacement increment of the linear feed axis within the current interpolation micro-segment, calculate the shortest time to complete the synchronous motion based on the underlying dynamic constraints of the machine tool rotary axis, and use it as the physical limit time; The linear displacement increment is used as the numerator, and the sum of the physical limit time and the preset time offset constant is used as the denominator to perform a ratio calculation to obtain the forced dimension reduction reconstruction feed rate. The forced dimensionality reduction and reconstruction feed rate are compared with the pre-compensated feed rate, and the result is smoothed in the time domain using a first-order inertial low-pass filter algorithm. The result is then encapsulated with the synthesized dynamic tool axis deflection vector into a synchronous data frame and sent to the underlying servo driver.

Images