A multi-station processing closed-loop control method for an electric sofa special-shaped hardware

Abstract

The present application relates to numerical control program control technical field, disclose a kind of multi-station processing closed-loop control method of electric sofa special-shaped hardware, including according to the power data of preceding sequence station generates micro unloading parameter and step displacement threshold value and controls servo fixture operation, in micro unloading detection interval superimposes high-frequency torque flutter signal, according to the current sequence and clamping displacement sequence of feedback calculation hysteresis loop integral area and construct space elastic internal stress feature vector, by energy and yielding flexibility matrix and kinematics jacobian matrix the feature vector is converted into expected space geometric springback deviation vector, to nominal target coordinate reverse bias generates dynamic compensation coordinates and processes, obtains the actual space error vector after processing and decomposes into translation and torsion error component, respectively updates force and error flexibility conversion matrix and weight mapping matrix.The present application realizes clamping deformation adaptive adjustment and processing trajectory dynamic compensation.

Description

Technical Field

[0001] This invention relates to the field of CNC program control technology, specifically to a closed-loop control method for multi-station machining of irregularly shaped hardware parts for electric sofas. Background Technology

[0002] The irregularly shaped hardware parts for electric sofas require multiple processing steps such as stamping and bending during production. Due to the irregular geometry of the hardware parts and the accumulation of elastic residual stress in the previous processing, they exhibit nonlinear mechanical characteristics when clamped in subsequent processing steps.

[0003] Conventional CNC machining clamping control mainly uses constant clamping force or fixed position closed-loop control, which cannot sense and adapt to changes in residual stress inside the workpiece. The dry friction between the workpiece and the fixture interface hinders the spontaneous release of elastic internal stress in the hardware during the clamping stage, causing the internal stress to be excited by external loads during the cutting process and resulting in non-uniform release, or causing springback deviation when the fixture is released after machining, thus reducing the form and position tolerance accuracy of the finished product.

[0004] Meanwhile, existing machining error compensation relies on offline measurement and fixed geometric stiffness conversion models, failing to consider the mechanical wear caused by long-term use of fixtures and the slow drift of system compliance parameters due to batch fluctuations in workpiece material. This results in a decrease in compensation accuracy as the number of production batches increases. Furthermore, during multi-axis interpolation, the control system lacks a mechanism for dynamic physical boundary linkage verification of real-time cutting load, clamping friction holding force, and actual commanded acceleration of the motion axis. When the reverse compensation exceeds physical constraints or the cutting force changes abruptly, it can easily cause geometric distortion of the interpolation trajectory or slippage of the workpiece within the fixture. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a closed-loop control method for multi-station machining of irregularly shaped hardware parts for electric sofas. This method solves the problem that, during multi-station machining of irregularly shaped hardware parts for electric sofas, residual stress release and dry friction coupling on the contact surface can easily lead to clamping deformation and machining geometric errors.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: a multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas, including calculating the work integral, sample mean and variance based on the power data of the preceding station, and generating micro-unloading parameters and step displacement thresholds using a weight mapping matrix; The servo fixture is controlled to operate according to the micro-unloading parameters and the step displacement threshold, and a high-frequency torque chatter signal is superimposed within the micro-unloading detection action range. Obtain the total output torque current sequence and clamping displacement sequence fed back by the servo fixture, calculate the integral area of ​​the hysteresis loop and construct the spatial elastic internal stress feature vector; The spatial elastic internal stress eigenvector is converted into the expected spatial geometric rebound deviation vector by using the preset energy and yielding flexibility matrix and the kinematic Jacobian matrix. Dynamic compensation coordinates are generated by reverse offsetting the nominal target coordinates according to the expected spatial geometric rebound deviation vector, and the machining is performed at the position of the dynamic compensation coordinates. Obtain the actual spatial error vector and decompose it into translational error components and torsional error components. Update the force and error compliance transformation matrix using the translational error components and reverse-correct the weight mapping matrix using the torsional error components.

[0007] In a preferred embodiment of the present invention, the step of calculating the work integral, sample mean, and variance based on the power data of the preceding workstation, and generating micro-unloading parameters and step displacement thresholds using a weight mapping matrix, includes: After deducting the no-load power baseline and correcting the drive efficiency of the power data, the instantaneous active power within the time window is accumulated using the trapezoidal integral algorithm to obtain the work integral; when the sample variance is less than or equal to a preset variance threshold, and the sample mean or the work integral of the current workpiece is less than or equal to a preset energy level threshold, the system is determined to be in a mechanical steady state, the micro-unloading parameter is assigned to zero, and the step displacement threshold is assigned a preset closed position value; When the sample variance is greater than the preset variance threshold, or the sample mean or the work integral of the current workpiece is greater than the preset energy level threshold, the system is determined to be in a mechanically unsteady state. The work integral of the current workpiece is substituted into the mapping calculation model, and the step displacement threshold and micro-unloading parameters of the current workpiece are generated by matrix multiplication.

[0008] In a preferred embodiment of the present invention, controlling the operation of the servo fixture according to the micro-unloading parameters and the step displacement threshold includes: dividing the clamping process into a rapid approach segment, a micro-unloading detection segment, and a secondary loading and locking segment. In the rapid approach phase, the servo fixture is controlled to run. When it is determined that the servo fixture has reached the position defined by the step displacement threshold, the motion state is switched to the micro-unloading detection phase. In the micro-unloading detection phase, the target endpoint of the reverse retreat is calculated. The coordinates of the target endpoint are represented by the step displacement threshold minus the micro-unloading parameter. The servo fixture is controlled to retreat in a direction away from the workpiece to the target endpoint. In the secondary loading and locking section, the servo clamp is controlled to advance from the target endpoint back towards the final fastening coordinate direction until the preset clamping torque threshold is reached; the rapid approach section, the micro-unloading detection section and the secondary loading and locking section all use third-order polynomial curves for position interpolation.

[0009] In a preferred embodiment of the present invention, superimposing a high-frequency torque chatter signal within the micro-unloading detection action range includes: generating a discrete sine wave sequence as chatter compensation torque, and superimposing the chatter compensation torque into the basic motion torque command through an addition operation to form the final current loop input value. During the flutter signal injection, the actual current feedback value and encoder angular velocity feedback value inside the servo driver are read, and the real-time equivalent friction torque of the contact surface is extracted. The real-time equivalent friction torque is compared with the preset critical static friction torque. When the real-time equivalent friction torque is less than the product of the preset critical static friction torque and the preset conversion coefficient, the dynamic Coulomb friction conversion is determined to be successful.

[0010] In a preferred embodiment of the present invention, obtaining the total output torque current sequence and clamping displacement sequence of the servo fixture feedback, and calculating the hysteresis loop integral area includes: The total output torque current sequence is subjected to band-stop filtering to remove the narrow-band current component corresponding to the high-frequency torque chatter signal. The signal is then input into a first-order low-pass filter for smoothing, and a low-frequency current waveform is output. The motion direction flag of the clamping displacement sequence is analyzed, and the low-frequency current waveform is separated to generate a forward loading sequence and a reverse unloading sequence. A cubic spline interpolation algorithm is used to resample the forward loading sequence and the reverse unloading sequence with equal step size within the overlapping displacement interval to generate forward current curves and reverse current curves. A trapezoidal numerical integration algorithm is used to calculate the cumulative integration of the current difference between the forward current curve and the reverse current curve at corresponding positions to obtain the hysteresis loop integration area.

[0011] In a preferred embodiment of the present invention, the construction of the spatial elastic internal stress feature vector includes: reading the hysteresis loop integral area of ​​each servo axis participating in the clamping, and the hardware parameters of each servo axis including the motor torque constant and the mechanical transmission coefficient. For each servo axis, the hysteresis loop integral area, the motor torque constant, the mechanical transmission coefficient, and the length unit conversion factor are multiplied to convert the drive current into linear force work applied to the workpiece by the end effector, thus obtaining the equivalent elastic dissipation work. The equivalent elastic dissipation work of each servo axis is arranged in the index order of the spatial coordinate axes to construct the spatial elastic internal stress characteristic vector in column vector form.

[0012] In a preferred embodiment of the present invention, the step of converting the spatial elastic internal stress eigenvector into a expected spatial geometric rebound deviation vector by using a preset energy and yielding compliance matrix and a kinematic Jacobian matrix includes: The initial position compensation vector is calculated by multiplying the initial energy and yielding compliance matrix with the spatial elastic internal stress eigenvector; the initial position compensation vector is then limited using the maximum allowable deformation displacement parameter and the lower dead zone threshold, and the actual position compensation vector is generated by combining them. The actual position compensation vector is multiplied by the kinematic Jacobian matrix to calculate the expected spatial geometric rebound deviation vector, which includes translation and rotation components.

[0013] In a preferred embodiment of the present invention, generating dynamic compensation coordinates by reverse offsetting the nominal target coordinates according to the expected spatial geometric rebound deviation vector includes: The underlying CNC machining program is analyzed to extract the nominal target coordinates in the trajectory flow; the expected spatial geometric springback deviation vector is decoupled and separated into translational deviation vector and rotational deviation vector, and a three-dimensional orthogonal rotation matrix is ​​constructed based on the translational deviation vector and the rotational deviation vector, and combined to generate a homogeneous deviation matrix characterizing the workpiece spatial pose deviation. The homogeneous deviation matrix is ​​inverted to generate a reverse bias transformation matrix; the nominal target coordinates are converted into homogeneous coordinates and multiplied with the reverse bias transformation matrix to obtain the corrected target coordinates; when it is confirmed that each axial component in the corrected target coordinates is within the corresponding maximum travel interval, the corrected target coordinates are used as the dynamic compensation coordinates.

[0014] In a preferred embodiment of the present invention, machining at the dynamically compensated coordinate position includes: while maintaining a forced constraint state, based on the normal clamping force applied to the workpiece by the clamping mechanism corresponding to each servo axis, the mass of the workpiece, the real-time cutting force, and the static friction coefficient between the workpiece and the contact surface of the fixture, calculating the maximum allowable acceleration and the minimum allowable acceleration of each servo axis in the current interpolation cycle, and establishing a dynamic range of allowable acceleration. Calculate the actual commanded acceleration of each motion axis in the current cycle. When the actual commanded acceleration exceeds the allowable acceleration dynamic range, calculate the deceleration factor required for the over-limit axis based on the deviation between the actual commanded acceleration of the over-limit axis and the boundary value, and extract the maximum value as the global deceleration factor. By using the global deceleration factor to proportionally attenuate the current nominal tangential velocity, the command speed and command acceleration of all participating servo axes are adjusted synchronously so that the actual command acceleration of each axis is within the allowable acceleration dynamic range.

[0015] In a preferred embodiment of the present invention, the step of obtaining the actual spatial error vector and decomposing the actual spatial error vector into translational error components and torsional error components, updating the force and error compliance transformation matrix using the translational error components, and reversibly correcting the weight mapping matrix using the torsional error components includes: The actual coordinate data of preset measurement feature points on the workpiece surface are obtained by a contact probe. The three-dimensional position deviation vector of each measurement feature point is calculated by combining the corresponding nominal coordinate data. The three-dimensional position deviation vector is then decomposed by performing a dot product operation with the corresponding surface unit normal vector. The actual spatial error vector is then constructed by combining the vectors. Based on the nominal coordinate data, actual coordinate data and corresponding surface unit normal vector of each of the measured feature points, the rigid body small pose error model is substituted, and the six-dimensional spatial residual vector is solved by the least squares algorithm. The first three elements of the six-dimensional spatial residual vector are defined as the translation error component, and the last three elements are defined as the torsional error component. Substitute the generalized force vector under the current correction period, the prediction error vector and the gain vector calculated based on the translation error components into the recursive least squares algorithm model to recursively update the force and error compliance transformation matrix; extract the input state feature vector containing the work integral, sample mean and sample variance to construct a scalar loss function, combine the input state feature vector to calculate the gradient matrix of the scalar loss function with respect to the weight mapping matrix, and perform parameter updates on the weight mapping matrix along the negative direction of the gradient matrix.

[0016] This invention provides a closed-loop control method for multi-station processing of irregularly shaped hardware parts for electric sofas. It has the following beneficial effects: 1. This invention generates micro-unloading parameters based on the integral of the power done at the preceding workstation, and superimposes a high-frequency torque chatter signal within the micro-unloading detection range. This causes the contact area between the workpiece and the fixture to transition from a static friction state to a dynamic friction state, eliminating the hindering effect of dry friction on workpiece deformation and springback. By combining the hysteresis loop integral area extracted from the driving current and clamping displacement to construct a spatial elastic internal stress feature vector, accurate quantification and smooth release of the concentrated stress inside the workpiece are achieved, improving the accuracy of predicting the geometric springback posture deviation after workpiece clamping.

[0017] 2. This invention utilizes the expected spatial geometric springback deviation vector to construct a homogeneous deviation matrix and performs an inversion operation. The reverse offset transformation matrix is ​​then multiplied by the nominal target coordinates to obtain the corrected target coordinates, achieving online correction of the machining trajectory without modifying the underlying CNC machining code. Simultaneously, the normal clamping force, real-time cutting force, and actual commanded acceleration of each motion axis are dynamically combined and verified. When the safety boundary is exceeded, a global deceleration factor is used to proportionally attenuate the current nominal tangential velocity, maintaining the geometric contour of the machining trajectory without distortion while preventing workpiece slippage exceeding tolerances caused by cutting load fluctuations.

[0018] 3. This invention establishes a dual-loop parallel adaptive calibration mechanism for calculating actual spatial errors. It utilizes the decomposed translational error components to update the force and error compliance transformation matrices online using a recursive least squares algorithm, eliminating the deviations in the stiffness model caused by mechanical wear from long-term fixture use and fluctuations in blank material. Furthermore, it leverages the torsional error components to perform inverse optimization of the weight mapping matrix using a gradient descent algorithm, correcting the mapping relationship of the micro-unloading control parameters and achieving closed-loop adaptive convergence of the form and position errors in subsequent batches of hardware parts processing. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is the sample variance solution and material steady-state discrimination diagram of the present invention; Figure 3 This is a diagram showing the separation of forward and reverse current curves and the extraction of hysteresis loop area integrals according to the present invention. Figure 4 This is a diagram showing the reverse offset calculation and trajectory correction of the nominal target coordinates of the present invention. Figure 5 This is a characteristic curve of the uniaxial loading and unloading current hysteresis loop of the present invention; Figure 6 This is a box plot of the machining form and position error distribution of the present invention; Figure 7 This is a convergence curve of the scalar loss function of the present invention as iterates with processing batches; Figure 8 This is a diagram showing the dynamic acceleration envelope and actual acceleration tracking of the present invention. Detailed Implementation

[0020] 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.

[0021] See attached document Figure 1 This invention provides a multi-station control method based on flutter hysteresis feature extraction and multidimensional residual dual adaptive closed loop, comprising the following steps: S10, when the irregular hardware part is transferred from the previous station to the current station, the energy assessment module obtains the power data output by the processing driver of the previous station, performs no-load power baseline subtraction and drive efficiency correction on the power data, calculates the work integral, sample mean and variance, generates steady-state assessment results based on the work integral level and sample variance, and outputs micro-unloading parameters and step displacement threshold according to the steady-state assessment results. S20, the clamping control module generates the position control curve of the current station servo fixture based on the micro-unloading parameters and the step displacement threshold. When the position limited by the step displacement threshold is reached, the reverse displacement command is initiated. At the same time, the signal injection module superimposes a high-frequency torque chatter signal onto the current control loop of the servo fixture driver within the micro-unloading detection action range. S30, the feature extraction module obtains the total output torque current sequence and the corresponding clamping displacement sequence fed back by the servo clamping driver, performs band-stop filtering on the total output torque current sequence to filter out the narrow-band current component corresponding to the high-frequency torque chatter signal, and separates the current curve envelopes of the positive loading stage and the reverse unloading stage based on the motion direction mark, speed symbol or trajectory stage mark of the clamping displacement sequence, calculates the hysteresis loop area formed by the two current curve envelopes and constructs the spatial elastic internal stress feature vector. S40, the deviation prediction module retrieves the system's preset energy and yielding compliance matrix and kinematic Jacobian matrix. First, it converts the spatial elastic internal stress characteristic vector into the position compensation vector of each servo axis through matrix multiplication. Then, it decouples the position compensation vector of each servo axis into the six-dimensional expected spatial geometric springback deviation vector of the workpiece in the spatial rectangular coordinate system through the kinematic Jacobian matrix. The six-dimensional expected spatial geometric springback deviation vector includes three translational deviation components and three rotational deviation components. S50, the compensation execution module uses the expected spatial geometric springback deviation vector to perform reverse offset calculation on the nominal target coordinates in the standard machining program to generate dynamic compensation coordinates. Under the state of the servo fixture maintaining clamping constraints, it guides the machining execution mechanism to move to the position of the dynamic compensation coordinates to perform machining actions. S60, after the workpiece is processed and transferred to the inspection station, the residual update module obtains the actual spatial error vector output by the measuring equipment. Based on the actual coordinates and nominal coordinates of multiple measurement feature points, a rigid body small pose error model is established. The actual spatial error vector is converted into a six-dimensional spatial residual vector through the least square pose fitting algorithm. The six-dimensional spatial residual vector is decomposed into translational error components and torsional error components. The translational error components are substituted into the recursive least square algorithm model to update the force and error compliance transformation matrix coefficients. The torsional error components are substituted into the gradient descent algorithm model to reverse correct the detection parameter weight mapping moment used to generate micro-unloading parameters and step displacement threshold.

[0022] The technical solutions for each step of this invention will be explained in detail below, taking into account the underlying algorithm logic and control mechanism.

[0023] In a multi-station machining production line, irregularly shaped metal parts are transferred from the previous machining process to the current clamping process. The physical deformation or material removal operations performed on the workpiece by the previous process will change the stress distribution inside the workpiece. To evaluate this stress state, this embodiment maps material characteristics by extracting the electrical work parameters of the machining equipment.

[0024] When processing equipment performs plastic deformation or cutting to overcome the yield resistance of the workpiece material, the effective portion of the electrical energy consumed is converted into mechanical work. The total mechanical work, after correction for no-load power, transmission loss, and drive efficiency, serves as an energy characteristic quantity characterizing the trend of residual stress changes within the material. This energy characteristic quantity is positively correlated with the degree of accumulation of residual stress within the material under the same process conditions and calibration range. The energy assessment module performs the following steps in the pre-processing energy extraction process: S101, the energy assessment module listens to the action command status words issued by the upstream workstation control system via the industrial fieldbus. When a trigger signal indicating the start of a processing action is detected, the current network clock is recorded as the processing start timestamp; When an end signal indicating the termination of a processing action is detected, the corresponding network clock is recorded as the processing end timestamp. The duration between the processing start timestamp and the processing end timestamp constitutes the time window for data acquisition.

[0025] To prevent the loss of the termination signal due to field equipment failure or communication interruption, the energy assessment module is configured with a waiting time threshold. If the termination signal is not received within the waiting time threshold since the trigger signal, the energy assessment module terminates the recording of the current time window and outputs a timeout error code.

[0026] In this implementation scenario, the trigger signal and the end signal correspond to the level change of the auxiliary function code in the main program of the CNC system, or the input / output status bit of the mold position sensor associated with the programmable logic controller of the stamping equipment.

[0027] S102, within a defined time window, the energy assessment module reads instantaneous active power data from the preceding workstation's machining driver at a set sampling period. The preceding workstation's machining driver can be a spindle servo driver controlling the cutting spindle, a feed servo driver controlling the linear motion of the feed axis, or a servo pump driver driving the hydraulic forming mechanism. The preceding workstation's machining driver collects the stator three-phase voltage and phase current within the bottom current loop, performs multiplication, and filters out reactive power components before outputting the instantaneous active power. The duration of this sampling period is determined based on the communication cycle time of the industrial fieldbus and is configured to be in the millisecond range. The energy assessment module reassembles the instantaneous active power acquired in each sampling period in chronological order to generate discrete power data.

[0028] S103, the energy assessment module performs numerical integration on the generated power data. The system uses a trapezoidal integral algorithm to accumulate the instantaneous active power within the time window, obtaining the work integral. This work integral, after deducting the idle power integral of the preceding station and multiplying by the drive chain efficiency correction coefficient, characterizes the total mechanical energy applied to the workpiece by the cutting tool or mold of the preceding station, causing the material to undergo elasto-plastic deformation. The fluctuation of this work integral value is used to characterize the objective differences in material yield strength, thickness tolerance, or cutting allowance between different batches of workpieces.

[0029] Regarding the first For each workpiece, the energy assessment module calculates the integral of its preceding work. The formula is expressed as: ; In the formula, A positive integer representing the workpiece sequence number in a continuous production batch; This indicates the total number of power data sampling points collected within the time window; Indicates the first time within the time window The instantaneous active power value extracted at each sampling time, in watts; Indicates the first The instantaneous active power values ​​extracted at each sampling time; This indicates the sampling period time interval of the fieldbus communication configuration, in seconds.

[0030] The work integral calculated in a single operation can experience data jumps due to electrical noise from the equipment or interference from localized material hard spots. To reflect the stress variation trend of material batches, the energy assessment module uses a sliding time window algorithm to statistically process the work integral data. The sliding window retains recent historical data and removes outdated data, filters out random interference, and extracts trend features characterizing changes in material stiffness or margin. The process of the energy assessment module performing sliding window sample variance calculation and material steady-state discrimination includes the following steps: S104, the energy assessment module establishes a first-in-first-out (FIFO) buffer queue in its internal memory. After completing the preliminary work integral calculation for each workpiece, the energy assessment module stores the work integral into the buffer queue. The depth of the buffer queue is set to the window length. When the amount of data in the buffer queue reaches the window length, it constitutes the sliding time window for the current batch of workpieces.

[0031] Each time new data is stored, the energy assessment module simultaneously removes the oldest data with the oldest timestamp, maintaining a constant number of data points within the sliding time window. If the amount of data in the buffer queue does not reach the set window length, the energy assessment module temporarily suspends variance calculation, outputs an initial assessment result indicating insufficient sample size, and calls preset conservative detection parameters as the micro-unloading parameters and step displacement threshold for the current workpiece, causing the current workpiece to enter feature extraction mode. Steady-state discrimination is then performed only after the amount of data in the buffer queue reaches the set window length.

[0032] S105, when the amount of data in the cache queue reaches the window length setting, for the first... For each workpiece, the energy assessment module extracts all work integral data within the current sliding time window and calculates the sample mean and sample variance of the data set.

[0033] The sample mean represents the baseline level of processing energy within a time window, used to determine whether the overall processing energy level of the current batch of workpieces exceeds a preset energy threshold. The sample variance represents the dispersion of processing energy from the baseline level, reflecting the fluctuation state of stiffness or processing allowance in consecutive batches of materials.

[0034] The energy assessment module calculates the sample variance. The formula is: ; In the formula, A positive integer representing the workpiece sequence number in a continuous production batch; A positive integer representing the window length setting; Indicates the first time within the sliding time window The integral of work done by each workpiece; Indicates the first The sample mean of the work integral done by each workpiece within the sliding time window; Indicates the first The sample variance of the work integral done by each workpiece within the sliding time window. Formula introduction. Used as the denominator for unbiased estimation calculations to correct calculation biases due to small sample sizes.

[0035] S106, the energy assessment module compares the calculated sample variance with the preset variance threshold, and compares the sample mean or the work integral of the current workpiece within the current sliding time window with the preset energy level threshold to generate the assessment result.

[0036] When the sample variance is less than or equal to the variance threshold, and the sample mean or the work integral of the current workpiece is less than or equal to the energy level threshold, the energy assessment module generates an assessment result indicating that the system is in a mechanically steady state. When the sample variance is greater than the variance threshold, or the sample mean or the work integral of the current workpiece is greater than the energy level threshold, the energy assessment module generates an assessment result indicating that the system is in a mechanically unsteady state. The energy assessment module sends the assessment result to the clamping control module. The variance threshold is calibrated by extracting the variances of multiple sets of work integral samples from qualified workpieces under constant machining conditions and averaging them.

[0037] The energy assessment module acquires the steady-state assessment results of the workpiece and adjusts the servo fixture control strategy of subsequent stations based on these results. When the work integral of the preceding process increases, the residual stress inside the workpiece increases; therefore, the reverse retraction distance is increased and the detection position is triggered earlier during the clamping process. When the material is in a mechanically steady state, the detection action is skipped. The process of the energy assessment module performing detection parameter weight mapping and adaptive strategy decision-making includes the following steps: S107, when the energy assessment module receives an assessment result indicating that the system is in a mechanically steady state, it determines that the residual stress fluctuation and the overall energy level corresponding to the residual stress of the workpiece in the current processing batch do not exceed the tolerance range, and activates the adaptive strategy decision. The energy assessment module assigns the micro-unloading parameter to zero and assigns the step displacement threshold to a preset closed position value, and sends a clamping command to the clamping control module. The clamping control module receives the clamping command, controls the servo fixture to run at a set speed to the preset mechanical locking position, performs the clamping action, and completes the processing cycle.

[0038] S108, when the energy assessment module receives the assessment result indicating that the system is in a mechanically unsteady state, it determines that the workpiece has a state variable where the residual stress fluctuation exceeds the limit or the residual stress energy level exceeds the limit, and activates the feature extraction mode.

[0039] The energy assessment module retrieves a pre-set weight mapping matrix and a basic parameter correction vector from its internal memory. The weight mapping matrix characterizes the transfer relationship between the total mechanical energy input from the front-end machining and the displacement required for servo fixture detection. The energy assessment module substitutes the work done by the workpiece into the mapping calculation model and generates the workpiece's step displacement threshold and micro-unloading parameters through matrix multiplication.

[0040] Regarding the first For each workpiece, the energy assessment module calculates the step displacement threshold. With micro-unloading parameters The formula is expressed as: ; In the formula, A positive integer representing the workpiece sequence number; Indicates the first The step displacement threshold of each workpiece, in milliseconds; Indicates the first The micro-unloading parameters of each workpiece, in milliseconds; This represents the weight mapping coefficients corresponding to the displacement threshold. This represents the weight mapping coefficients corresponding to the micro-unloading amplitude. and The weight mapping matrix is ​​constructed in the form of a two-dimensional column vector. Indicates the first The integral of work done by each workpiece; This represents the preset reference trigger displacement constant, in milliseconds (ms). This represents the preset reference micro-unloading amplitude constant, in milliseconds (ms).

[0041] Reference trigger displacement constant Compared with the reference micro-unloading amplitude constant The value is preset based on the mechanical structure hard limit of the servo fixture and the nominal geometric dimensions of the workpiece.

[0042] S109, the energy assessment module compares the calculated step displacement threshold and micro-unloading parameters with the preset effective travel limit value of the servo system. If the calculated value exceeds the preset effective travel limit value, the energy assessment module overwrites the current calculated value with the preset effective travel limit value for amplitude truncation. The preset effective travel limit value is set according to the maximum and minimum safe space displacement boundaries allowed for sliding of the servo fixture. After verification, the energy assessment module sends the step displacement threshold and micro-unloading parameters to the register of the clamping control module via the industrial fieldbus as input data for the current station servo displacement interpolation.

[0043] The clamping control module receives the step displacement threshold. With micro-unloading parameters The motion trajectory guiding the servo fixture is generated using an interpolation algorithm. Microscopic geometric errors on the workpiece surface during elastic contact can lead to localized stress accumulation. By planning the position control trajectory, the fixture retracts upon contacting the workpiece surface, releasing elastic strain space and improving clamping stress uniformity. The process of the clamping control module planning the micro-motion step position control curve trajectory includes the following steps: S201, the clamping control module parses the communication message and extracts the step displacement threshold of the current machining cycle. With micro-unloading parameters The clamping control module reads the actual position feedback value of the current servo fixture as the starting coordinate. And retrieve the preset processing task target position as the final fastening coordinate. Based on the aforementioned key location nodes, the clamping control module divides the clamping process into a rapid approach segment, a micro-unloading detection segment, and a secondary loading locking segment.

[0044] S202, in the rapid approach phase, the clamping control module generates... Run to The motion commands are then used. This stage employs a trapezoidal acceleration / deceleration planning algorithm to move the servo fixture, and the velocity and acceleration boundaries obtained from the trapezoidal acceleration / deceleration planning are used as constraints for subsequent third-order polynomial position interpolation.

[0045] The clamping control module compares the actual position value fed back by the servo driver in real time. When the conditions are met When the system determines that the servo gripper has reached the step trigger position, it switches the motion state to the micro-unloading detection segment. To prevent mechanical failure from causing the trigger condition to fail, the clamping control module synchronously starts a safety timer. If the running time exceeds the preset safety approach time and the above condition is not triggered, the system determines that a position interpolation abnormality has occurred, outputs an approach timeout alarm command, and terminates the operation of the servo motor.

[0046] S203, in the micro-unloading detection section, the clamping control module calculates the target endpoint for reverse retraction, and the coordinates of the target endpoint are represented as follows: The clamping control module generates a reverse interpolation trajectory, guiding the servo fixture from a step displacement threshold. The position loop retracts away from the workpiece. During the retraction, the position loop gain parameter switches to low gain mode to coordinate with the subsequent current loop signal injection.

[0047] Once the servo clamp reaches the target endpoint and stabilizes for the preset dwell time, the system enters the secondary loading and locking phase. The preset dwell time is configured to be between 50 and 200 milliseconds to allow for the attenuation of mechanical residual vibrations in the actuator. During the secondary loading and locking phase, the clamping control module guides the servo clamp back from the target endpoint to the final clamping coordinate. Proceed in the direction until the preset clamping torque threshold is reached.

[0048] The preset clamping torque threshold is pre-set based on the nominal clamping force required by the workpiece and the torque constant of the servo motor through a proportional conversion relationship, which is used to limit the final clamping state. All stages above use third-order polynomial curves for position interpolation. The velocity and acceleration boundaries of the rapid approach segment are determined by the trapezoidal acceleration / deceleration planning results. The micro-unloading detection segment and the secondary loading locking segment generate third-order polynomial curves according to preset endpoint velocity and endpoint acceleration constraints to ensure velocity and acceleration continuity.

[0049] S204, the clamping control module performs a logic check before outputting the trajectory. If the condition is met... This indicates that the calculated retraction position exceeds the physical starting boundary, and the clamping control module will adjust the current machining cycle. Revised to Half of the value is used to prevent logical collision errors. All coordinate parameters satisfy the following in the nominal state. The coordinates are in an increasing relationship. If an inversion is detected, the clamping control module stops trajectory calculation and issues a configuration error code. The verified position sequence is sent to the driver position loop through the servo synchronous cycle, realizing closed-loop control of the servo fixture end effector.

[0050] See attached document Figure 2 During the movement of the servo fixture in the micro-unloading detection section, the clamping control module injects a high-frequency, low-amplitude torque chatter signal through the servo driver current loop. Based on the Coulomb friction model and the Stribek friction effect, in the mechanical clamping contact area, high-frequency micro-motion disrupts the viscous state of the contact surface through periodic micro-slippage. This signal superimposes a high-frequency alternating torque on the electromagnetic torque, reducing the equivalent friction coefficient of the contact surface and causing the residual stress inside the workpiece to be released within the micro-unloading displacement. The process of the clamping control module injecting the high-frequency, low-amplitude torque chatter signal includes the following steps: S205, the clamping control module monitors the execution status of the micro-unloading probe section. When the servo fixture initiates the reverse retraction interpolation action, the chatter signal generation unit is activated synchronously. The chatter signal generation unit generates a discrete sine wave sequence as chatter compensation torque according to preset configuration parameters. The amplitude of this chatter compensation torque is set to 5% to 15% of the servo motor's rated torque to generate an alternating excitation force that overcomes static friction torque without causing workpiece position deviation. Chatter frequency parameters. The value range is configured from 50Hz to 200Hz. The value is determined based on the closed-loop response bandwidth of the servo system and the high-frequency transmission characteristics of the fixture mechanical structure to ensure that the high-frequency alternating torque is transmitted to the end effector without being attenuated by the system inertia.

[0051] S206, the clamping control module reads the basic motion torque commands output from the position loop and speed loop, and adds the chatter compensation torque to the basic motion torque commands using an additive operation to form the final current loop input value. This input value serves as the driver-side input value. The given shaft current signal acts on the electromagnetic torque output of the motor. The synthesized torque command value equals the sum of the basic motion torque command value and the chatter compensation torque, where the chatter compensation torque is based on the chatter amplitude parameter. With flutter frequency parameters A sine wave is generated.

[0052] S207, to avoid the inherent frequency of the mechanical structure, the clamping control module performs frequency verification logic before signal injection. If the flutter frequency parameter... If the frequency falls within the preset bandwidth range of the servo system's mechanical resonance frequency, the clamping control module retrieves the pre-calibrated safe frequency band parameters of the system and sets the flutter frequency parameters accordingly. The frequency is reassigned to the center frequency within the safe frequency band to prevent mechanical damage or instability of the controlled object.

[0053] S208, the clamping control module continuously monitors the displacement progress of the micro-unloading probe section. When the servo clamp reaches the retraction endpoint and the dwell time ends, the clamping control module performs a smooth attenuation operation on the chatter signal, reducing the chatter amplitude parameter within a preset attenuation period. The linear gain drops to zero. After the signal is turned off, the clamping control module restores the position loop gain to its rated value and enters the secondary loading and locking phase. The smooth turn-off mechanism suppresses the instantaneous impact of high-frequency signal abrupt changes on the actuator.

[0054] After superimposing a high-frequency chatter signal on the micro-unloading detection section, the clamping control module evaluates the friction state of the workpiece-fixture contact surface to determine whether static friction has transformed into dynamic friction. Dry friction characteristics exist between the workpiece and fixture contact surfaces. When the contact surface transitions from a relatively static viscous state to a relatively sliding Coulomb dynamic friction state, the frictional resistance decreases. The change in the contact surface constraint state is assessed by observing the dynamic torque change of the servo system. The process of the clamping control module performing dynamic Coulomb friction transformation and mechanical dry friction elimination state assessment includes the following steps: S209, during the flutter signal injection period, the clamping control module reads the actual current feedback value and encoder angular velocity feedback value from the servo driver in real time. The clamping control module smooths the feedback value using a first-order low-pass filter to remove high-frequency oscillation components. The clamping control module multiplies the filtered actual current feedback value by the motor's torque constant to calculate the actual electromagnetic torque of the servo motor. By performing a first-order discrete difference operation on the filtered angular velocity feedback value, the clamping control module synchronously obtains the actual angular acceleration of the motor shaft.

[0055] The S210 clamping control module, based on the principle of rigid body dynamics equilibrium, extracts the real-time equivalent frictional torque of the contact surface. During the operation of the mechanical transmission system, the actual electromagnetic torque is used to overcome the system's inertial torque and external frictional torque.

[0056] The value of the real-time equivalent friction torque is equal to the actual electromagnetic torque minus the product of the equivalent moment of inertia and the actual angular acceleration, and further subtracts the transmission loss compensation term, gravity compensation term and transmission efficiency compensation term obtained by no-load identification to obtain the equivalent friction torque acting on the contact surface between the fixture and the workpiece.

[0057] The equivalent moment of inertia is measured and stored through the equipment's no-load acceleration and deceleration identification program. The transmission loss compensation item, gravity compensation item, and transmission efficiency compensation item are obtained through no-load reciprocating motion test, static holding test under different postures, and calibrated load traction test, and stored in the parameter table of the servo drive or clamping control module.

[0058] S211, the clamping control module compares the real-time equivalent friction torque with the preset critical static friction torque. The critical static friction torque is pre-calibrated through a stall test of the servo mechanism in the nominal clamping state. When the alternating excitation force induces slippage, the contact surface transitions from a viscous state to the Coulomb dynamic friction region, and the system friction resistance decreases. When the real-time equivalent friction torque is less than the product of the critical static friction torque and the preset conversion coefficient, the clamping control module determines that the dynamic Coulomb friction conversion is successful, and the dry friction constraint caused by the contact surface is alleviated. The preset conversion coefficient is configured to range from 0.4 to 0.7 to avoid misjudgments caused by signal noise.

[0059] S212, the clamping control module synchronously runs the abnormal monitoring logic. If, at the end of the micro-unloading detection period, the real-time equivalent friction torque is still greater than or equal to the product of the critical static friction torque and the preset conversion coefficient, it indicates that the chatter signal has not disrupted the physical engagement of the contact surfaces. The clamping control module determines that the fixture and the workpiece have rigidly jammed, outputs a clamping interference abnormality code, and blocks the servo enable signal to prevent damage to the mechanical transmission components due to continuous overload.

[0060] After acquiring the real-time feedback current from the servo driver, the clamping control module performs frequency domain separation. Actual electromagnetic torque and current loop... The shaft current components are positively correlated. The current feedback current is a superposition of low-frequency fundamental motion components and high-frequency injected oscillation components, which will cause data bias if directly used for mechanical evaluation.

[0061] The clamping control module is equipped with a band-stop filter to filter out chatter signals and extract low-frequency current waveforms reflecting the system's stress state. The clamping control module performs current frequency domain separation and waveform reconstruction in the following steps: S301, the clamping control module synchronously samples the three-phase stator current of the servo driver through a current sensor, and obtains the result through Clark and Park coordinate transformation. The actual feedback current discrete sequence of the shaft. This current sequence contains high-frequency chatter components, fundamental motion components, and pulse width modulation switching noise.

[0062] S302, the clamping control module constructs a second-order band-stop filter. The stopband center frequency of the band-stop filter is set to the aforementioned injected flutter frequency parameter. The quality factor of the band-stop filter is configured from 10 to 50. The value of the quality factor is determined based on the bandwidth characteristics of the flutter signal, and is used to constrain the stopband width of the filter, filter out narrowband signals of the flutter frequency and its adjacent frequency bands, and retain the basic low-frequency physical characteristics.

[0063] S303, the clamping control module retrieves the sampling period parameters of the current loop of the current servo system and uses a bilinear transform algorithm to discretize the second-order band-stop filter into a difference equation. The discrete sequence of the actual feedback current of the shaft is substituted into the difference equation for digital filtering to separate the high-frequency flutter component and generate an intermediate filtering sequence. To avoid output oscillation caused by transient response during the initial calculation stage, the clamping control module performs a masking operation in the first few sampling periods after the filter starts, discarding transient data, and records the steady-state output after the filter state converges.

[0064] S304, the clamping control module inputs the intermediate filtering sequence into a first-order low-pass filter for smoothing, filtering out residual high-frequency harmonics and pulse-width modulation switching noise. The cutoff frequency of the first-order low-pass filter is pre-calibrated according to the control bandwidth of the servo driver, making it lower than the pulse-width modulation carrier frequency and higher than the dynamic response frequency, with a configuration range of 300Hz to 500Hz.

[0065] The output data after low-pass filtering is the restored low-frequency current waveform. The clamping control module stores the low-frequency current waveform in an internal data buffer register as an input sample for subsequent evaluation of the workpiece clamping mechanical state and feature extraction.

[0066] See attached document Figure 3 After acquiring the low-frequency current waveform, the clamping control module combines it with the position feedback data of the servo mechanism to construct the correspondence between current and position. During the mechanical clamping process, due to the influence of system frictional resistance and the elastic hysteresis effect inside the workpiece, the driving current required at the same position point during the forward loading and reverse micro-unloading stages does not coincide, forming a mechanical hysteresis characteristic curve.

[0067] The area of ​​the hysteresis loop is used to characterize the energy characteristics of frictional dissipation and system stress release during clamping. The clamping control module performs the separation of forward and reverse current curves and the integral calculation of the hysteresis loop area, including the following steps: S305, the clamping control module aligns the low-frequency current waveform sequence in the data buffer register with the synchronously sampled actual position feedback discrete sequence based on the system timestamp. The clamping control module parses the motion direction flag bit of the actual position feedback discrete sequence.

[0068] The clamping control module extracts the current and position data of the servo fixture when it is moving toward the workpiece and generates a forward loading sequence; it also extracts the current and position data of the servo fixture when it is performing a micro-retraction action and generates a reverse unloading sequence.

[0069] S306, the clamping control module performs spatial domain alignment on the aforementioned forward and reverse sequences. The displacement span of the forward loading sequence is greater than that of the reverse unloading sequence. Based on the start and end physical coordinates of the micro-unloading detection segment, the clamping control module extracts the displacement interval data that overlaps with the reverse unloading sequence in the forward loading sequence.

[0070] Because the acceleration and deceleration motion of the servo system results in uneven spatial distribution of the sampling points at equal time intervals, the clamping control module employs a cubic spline interpolation algorithm to resample the forward loading sequence and reverse unloading sequence within the overlapping interval at equal step sizes according to a preset spatial resolution. The resampled discrete data constitutes forward and reverse current curves with position as the x-axis and current as the y-axis.

[0071] The S307 clamping control module employs a trapezoidal numerical integration algorithm. Based on the resampled, constant-step displacement interval, it calculates and integrates the current difference between the forward and reverse current curves at corresponding points. The integration result is defined as the hysteresis loop integration area, serving as a reference indicator for evaluating the degree of residual stress release within the workpiece. The constant-step displacement interval is configured to range from 0.001 mm to 0.01 mm, determined based on the physical resolution of the servo encoder.

[0072] S308, the clamping control module performs data validity verification logic on the calculated hysteresis loop integral area. Abnormal data will be generated if the forward and reverse current values ​​are reversed due to external mechanical shock or signal acquisition interruption. The clamping control module retrieves the pre-calibrated lower limit threshold of the area under no-load operation.

[0073] If the integral area of ​​the hysteresis loop is determined to be less than the lower limit threshold, the clamping control module determines that the energy assessment data is abnormal, discards the calculation results, and drives the servo clamp to return to the starting coordinates of the current micro-unloading detection segment, and re-executes the interpolation and sampling action; if the number of consecutive retries exceeds the set 3 times, the system disables servo enable and outputs a detection failure alarm.

[0074] If the integral area of ​​the hysteresis loop is determined to be greater than or equal to the lower limit threshold, the clamping control module updates the calculated integral area of ​​the hysteresis loop to the status register and enters the subsequent adaptive adjustment phase of clamping parameters. This verification logic eliminates the influence of sensor zero drift and external interference on the closed-loop control algorithm.

[0075] In multi-axis collaborative clamping, there is an elastic coupling effect between the contact points of the workpiece, and a change in force in one dimension will cause stress redistribution in other dimensions. The integral area of ​​the hysteresis loop of a single axis only reflects the energy characteristics in a local direction, and its value is difficult to directly compare across axes due to the differences in hardware parameters of different drive axes.

[0076] The clamping control module extracts servo axis data involved in clamping, and after physical dimension conversion, constructs an equivalent spatial elastic internal stress feature vector characterizing the clamping state. This equivalent spatial elastic internal stress feature vector reflects the distribution of elastic dissipated work in each clamping direction and does not directly represent the absolute value of the actual stress field inside the workpiece material. The process of constructing the multi-axis simultaneous spatial elastic internal stress feature vector by the clamping control module includes the following steps: S309, the clamping control module traverses the control bus and synchronously reads the hysteresis loop integral area of ​​all servo axes involved in workpiece clamping. The clamping control module retrieves the hardware parameters of each servo axis stored in the driver firmware, including the motor torque constant and mechanical transmission parameters.

[0077] The S310 clamping control module converts the hysteresis loop integral area into equivalent elastic dissipation work in J for each servo axis. The equivalent elastic dissipation work is obtained by multiplying the hysteresis loop integral area, the motor torque constant, the mechanical transmission coefficient, and a length unit conversion factor. This process converts the drive current into the linear force work applied to the workpiece by the end effector. The length unit conversion factor, with a value of 10, is used to convert the displacement dimension from mm to the international standard unit m. -3 The mechanical transmission coefficient is determined based on the transmission structure of the servo axis. For a ball screw mechanism with rotary linear transmission, its value is equal to twice pi multiplied by the reduction ratio and then divided by the screw lead.

[0078] S311, the clamping control module arranges the calculated equivalent elastic dissipation work of each servo axis participating in the clamping according to the index order of the spatial coordinate axes, and constructs a spatial elastic internal stress characteristic vector in the form of a column vector.

[0079] S312, the clamping control module extracts the Euclidean norm of the spatial elastic internal stress characteristic vector to evaluate the spatial level of the overall residual stress of the workpiece. The clamping control module calculates the variance between the equivalent elastic dissipation work of each servo axis as a clamping anisotropy index.

[0080] The preset spatial stress threshold is determined by the statistical mean of historical qualified data of standard samples under calibration test conditions; the preset balance threshold is configured to be 1.2 to 1.5 times the variance of the dissipation work of the standard sample. The clamping control module compares the Euclidean norm with the preset spatial stress threshold.

[0081] If the Euclidean norm is greater than or equal to the threshold, it indicates that the overall internal stress of the workpiece has not yet dropped to the standard level. The clamping control module triggers an instruction to extend the injection cycle of the current high-frequency chatter signal until it is re-evaluated and meets the standard. If the Euclidean norm is less than the threshold, it is determined that the clamping internal stress has been released to the set range.

[0082] Based on this, if the variance exceeds the preset balance threshold, the clamping control module determines that the forces at each clamping point are uneven, marks the presence of local stress concentration, and triggers a compensation loading operation on the servo axis corresponding to the largest equivalent elastic dissipation work in subsequent processes. This process reduces the risk of control dead zones caused by a single threshold judgment under multi-axis force coupling conditions.

[0083] In a multi-axis collaborative clamping system, after assessing the spatial residual stress state of the workpiece, the system needs to release the concentrated stress through mechanical position adjustment. The clamping control module introduces an admittance control principle based on energy mapping to establish the transmission relationship between clamping state characteristics and servo position adjustment amounts. The process of the clamping control module executing the physical definition and initial establishment of the compliance transformation matrix includes the following steps: S401, the clamping control module constructs the initial energy and yield compliance matrices. In conventional impedance control, the compliance matrix describes the linear mapping between external forces and end-effector deformation. To match the spatial elastic internal stress characteristic vector extracted from the preceding process, the clamping control module defines the initial energy and yield compliance matrices as mapping operators between dissipated energy and position compensation displacement. Understandably, the number of rows and columns of this matrix is ​​equal to the total number of servo axes participating in the clamping, forming a static square matrix structure. Its input is the equivalent elastic dissipated work vector, and its output is the position compensation vector for each servo axis. The unit of the matrix elements is m / J.

[0084] S402, the clamping control module runs the offline equipment stiffness calibration program to obtain the coefficient values ​​of the initial energy and yield compliance matrix. The clamping control module controls each servo axis to perform a step torque loading action on a calibration fixture with standard stiffness, simultaneously injecting a nominal high-frequency chatter signal and calling the aforementioned frequency domain separation and integration algorithm to calculate the equivalent elastic dissipation work during the loading process, and reading the actual position deformation of the servo encoder after reaching mechanical equilibrium steady state. The clamping control module divides the position deformation of each servo axis by the corresponding equivalent elastic dissipation work to calculate the reference compliance coefficient of each single axis.

[0085] S403, the clamping control module sequentially fills the acquired reference compliance coefficients into the main diagonal positions of the initial energy and yield compliance matrices. To isolate physical coupling interference between different control axes in the initial stage, the clamping control module assigns all off-diagonal elements of the matrix to zero.

[0086] S404, the clamping control module calls the spatial elastic internal stress eigenvector generated by the pre-calculation and calculates the initial position compensation vector for stress release through matrix multiplication. The initial position compensation vector is obtained by multiplying the initial energy and yield compliance matrix with the spatial elastic internal stress eigenvector. The initial position compensation vector is a column vector, and its internal elements correspond to the required position compensation amount for each servo axis, in meters (m). The units of the main diagonal elements of the initial energy and yield compliance matrices are m / J.

[0087] S405, the clamping control module sets the position compensation limiting logic and verifies the values ​​of each element in the initial position compensation vector. The clamping control module retrieves the maximum allowable deformation displacement parameter and the servo encoder physical resolution parameter configured in the driver firmware, and configures 1.2 to 1.5 times the physical resolution as the lower dead zone threshold. If the position compensation of a certain servo axis exceeds the maximum allowable deformation displacement parameter, the clamping control module clamps its value to the upper limit to prevent the workpiece from falling off due to excessive compensation. If the position compensation of a certain servo axis is less than the lower dead zone threshold, the clamping control module assigns the compensation of that axis to zero to avoid sending invalid commands to the position loop and causing system jitter.

[0088] In S406, the clamping control module synchronously superimposes the position compensation amounts of each axis, after amplitude limiting processing, onto the reference target command of the corresponding servo driver position control loop. The servo motor drives the end-effector clamping mechanism to perform displacement retraction based on the updated position deviation. This control process makes the rigid clamping mechanism exhibit compliant characteristics under force retraction, releasing the clamping internal stress accumulated within the workpiece through deformation.

[0089] In a multi-axis clamping system, after the servo mechanism performs position retraction to release internal stress, the displacement of the clamping point causes the workpiece to shift positionally in the Cartesian coordinate system. Because the normal vectors and force arms of each servo axis in space differ, single-axis retraction synthesizes translational and rotational deviations at the workpiece reference. The clamping control module calculates this offset using a kinematic mapping model to assess the impact of stress release on the workpiece machining reference. The process of the clamping control module performing the decoupling matrix operation for the expected spatial geometric springback deviation includes the following steps: S407, the clamping control module establishes the kinematic Jacobian matrix between the multi-axis clamping system and the workpiece spatial pose. This kinematic Jacobian matrix represents the differential mapping relationship between the motion of each servo axis in joint space and the six-DOF pose change of the workpiece feature points in Cartesian space. The clamping control module retrieves the kinematic parameter library of the equipment control system, extracts the homogeneous transformation matrix of each servo axis relative to the workpiece reference coordinate system, and calculates the partial derivatives of the joint position variables in the forward kinematic equations to construct the kinematic Jacobian matrix.

[0090] S408, the clamping control module acquires the position compensation amounts of each axis after amplitude limiting processing and combines them to generate the actual position compensation vector. The clamping control module multiplies the actual position compensation vector with the kinematic Jacobian matrix to calculate the expected spatial geometric rebound deviation vector.

[0091] The expected spatial geometric rebound deviation vector is a six-dimensional column vector containing three translational components and three rotational components. The first three elements represent the translational deviation of the workpiece reference point along the spatial rectangular coordinate axis, and the last three elements represent the rotational deviation of the workpiece around the spatial rectangular coordinate axis. This kinematic Jacobian matrix is ​​only used to map the servo axis position compensation amount to the workpiece six-dimensional pose deviation, and is independent of the force and error compliance transformation matrices updated by the subsequent recursive least squares algorithm.

[0092] The elements within the actual position compensation vector are the position compensation amounts for each axis after amplitude limiting. The unit for the linear servo axis position adjustment is meters (m), and the unit for the rotary servo axis position adjustment is rad. In the expected spatial geometric rebound deviation vector, the unit for the translation component is meters (m), and the unit for the rotation component is rad.

[0093] S409, the clamping control module performs a decoupling operation on the expected spatial geometric springback deviation vector, separating it into a translational deviation vector and a rotational deviation vector along the spatial coordinate axes. The clamping control module calculates the Euclidean norm of the two sets of decoupled vectors respectively, obtaining the translational offset distance and rotational angle of the workpiece reference point in space.

[0094] S410, the clamping control module performs error verification logic on the decoupled data. The clamping control module retrieves the current process machining document issued by the computer-aided manufacturing system, and extracts the maximum permissible translation error threshold and the maximum permissible rotation error threshold according to the geometric tolerance zone required by the workpiece drawing.

[0095] The clamping control module compares the translational offset distance with the maximum permissible translational error threshold, and the rotation angle with the maximum permissible rotational error threshold. If the translational offset distance is greater than or equal to the maximum permissible translational error threshold, or the rotation angle is greater than or equal to the maximum permissible rotational error threshold, it indicates that the positional backlash caused by stress release has resulted in the workpiece exceeding the machining form and position tolerances. The clamping control module then blocks the machining start signal, outputs a clamping posture out-of-tolerance alarm, triggers the system to zero the compensation amount, and re-executes the initial workpiece edge finding and positioning centering process.

[0096] If both sets of deviation values ​​are less than the corresponding error threshold, the clamping control module confirms that the current workpiece spatial pose meets the machining requirements, outputs a clamping ready signal, and starts the cutting process. This verification logic ensures that the stress release process does not affect the machining datum.

[0097] See attached document Figure 4In multi-axis clamping operations, after the workpiece shifts position due to the release of clamping stress, its current actual pose deviates spatially from the ideal pose built into the CNC program. The aforementioned position compensation of each servo axis is used to release the accumulated stress within the workpiece, while the reverse offset of the nominal target coordinates (hereinafter referred to as the reverse offset) is used to counteract the impact of the workpiece's reference pose shift on the tool path after stress release. These two actions respectively apply to clamping stress release and machining path correction, and do not constitute redundant compensation.

[0098] The clamping control module describes the current deviation state of the workpiece through a homogeneous transformation matrix, and calculates a compensation transformation to cancel out the deviation using matrix inversion. This corrects the tool path without modifying the underlying CNC program, ensuring the tool path conforms to the workpiece's geometric contour during relative motion. The process of the clamping control module performing the reverse offset calculation of the nominal target coordinates includes the following steps: S501, the clamping control module parses the underlying CNC machining program and extracts the nominal target coordinates from the trajectory flow. The nominal target coordinates include the three-dimensional coordinates of the machining tool center point in the workpiece coordinate system.

[0099] S502, the clamping control module retrieves the translational deviation vector and rotational deviation vector of the workpiece reference point decoupled from the previous process to obtain the component values ​​along each coordinate axis in space. The clamping control module uses the translational deviation vector as a three-dimensional translation vector along the Cartesian coordinate axis in space, and constructs a three-dimensional orthogonal rotation matrix based on the component angles of each axis in the rotational deviation vector, and then combines them to generate a homogeneous deviation matrix characterizing the spatial pose deviation of the workpiece.

[0100] In step S503, the clamping control module performs an inverse matrix operation on the homogeneous deviation matrix to generate a reverse offset transformation matrix. This inversion operation establishes a reverse geometric mapping from the offset pose to the original design datum, ensuring that the subsequent target compensation direction is opposite to the workpiece displacement direction. The homogeneous deviation matrix and the reverse offset transformation matrix are both dimensionless fourth-order square matrices; the three-dimensional orthogonal rotation matrix is ​​a dimensionless third-order square matrix; and the three-dimensional translation vector is a third-order column vector with elements in units of m.

[0101] S504, the clamping control module converts the nominal target coordinates into homogeneous coordinates and multiplies them with the inverse bias transformation matrix to calculate the corrected target coordinates. Both the corrected target coordinates and the nominal target coordinates are converted into fourth-order homogeneous position column vectors, with the first three position components in meters (m).

[0102] S505, the clamping control module performs soft limit verification logic on the calculated corrected target coordinates. The clamping control module retrieves the maximum travel range of each physical motion axis of the machine tool configured in the CNC system firmware. It can be understood that the maximum travel range of each physical motion axis of the machine tool is determined based on the limit length of the CNC machine tool slide rail and the preset soft limit parameters of the system.

[0103] The clamping control module verifies whether each axial component in the target coordinate system is within its corresponding maximum travel range. If it determines that any axial component in the target coordinate system exceeds its corresponding maximum travel range, the clamping control module determines that there is an overtravel risk, blocks the enable input of the current interpolator, stops the CNC spindle from running and outputs an overtravel alarm signal, and simultaneously maintains the current position in a safe stop state to wait for manual intervention. If all axial components are determined to be within their corresponding maximum travel range, the clamping control module confirms that the current compensation trajectory can be executed normally. It then sends the corrected target coordinates to the multi-axis motion interpolator, driving the multi-axis servo mechanism to control the tool to perform cutting according to the corrected trajectory. This verification process reduces the risk of hardware collisions caused by excessive coordinate reverse compensation.

[0104] In a multi-axis clamping and CNC machining collaborative system, the workpiece is subject to mechanical constraints applied by the clamping mechanism during the machining process. This state of continuously maintaining clamping force and position locking is defined as a forced constraint state.

[0105] A multi-axis motion interpolator receives the corrected target coordinates and, under the premise of satisfying the mandatory constraints, calculates the interpolation commands for each motion axis, driving the multi-axis servo mechanism to complete the cutting. In one embodiment of the present invention, the interpolation process is performed on a dynamically compensated spatial trajectory. The corrected target coordinates that are confirmed to be executable normally are used as the dynamically compensated coordinates.

[0106] Under forced constraint, the resultant force of the inertial force generated by the servo motion of each axis and the external cutting force does not exceed the maximum static friction constraint provided by the clamping mechanism, thus preventing the workpiece from slipping inside the fixture. The process of interpolation machining under forced constraint by the clamping control module includes the following steps: S506, the multi-axis motion interpolator cyclically reads the corrected target coordinate stream from the buffer. The multi-axis motion interpolator obtains the current interpolation period, which is configured to be 0.001s. The nominal tangential velocity is calculated based on the ratio of the spatial distance between two adjacent corrected target coordinates to the interpolation period.

[0107] The S507 multi-axis motion interpolator performs dynamic verification of the output acceleration of each servo axis based on the current forced constraint state. The forced constraint state is determined by the holding force applied to the workpiece by the clamping mechanism, the workpiece mass, and the cutting force acquired in real time.

[0108] The real-time cutting force component of each servo axis in the direction of external load is estimated in real time by a cutting force observer integrated into the CNC system, or dynamically acquired by a force sensor.

[0109] To prevent workpiece slippage during machining, the multi-axis motion interpolator calculates the allowable dynamic range of acceleration for each servo axis in the current interpolation cycle.

[0110] Before calculating the allowable dynamic range of acceleration, the multi-axis motion interpolator first determines whether the friction holding force in the direction corresponding to the i-th servo axis is greater than the real-time cutting component force. If satisfied If the current forced constraint state fails, the multi-axis motion interpolator will immediately reduce the current feed speed to zero, block the interpolation output, and output an alarm signal for insufficient clamping constraint.

[0111] Maximum allowable acceleration for each servo axis With minimum allowable acceleration The calculation formulas are expressed as follows: ; ; In the formula, Indicates the first The maximum allowable acceleration of each servo axis during the current interpolation cycle, in m / s². 2 ; Indicates the first The minimum allowable acceleration of each servo axis during the current interpolation cycle, in m / s². 2 ; This represents the static friction coefficient between the workpiece and the contact surface of the fixture, and is dimensionless. Indicates the first The clamping mechanism corresponding to each servo axis applies a normal clamping force to the workpiece in N. Indicates the first Real-time cutting force component of each servo axis in the direction of external load, in N; This indicates the mass of the workpiece, expressed in kg. The static friction coefficient is also mentioned. Based on the material preset of the contact surface between the workpiece and the fixture, the value range is 0.15 to 0.35.

[0112] The S508 multi-axis motion interpolator calculates the commanded velocity and actual commanded acceleration for each motion axis in the current cycle. The actual commanded acceleration is obtained by dividing the difference between the commanded velocity in the current cycle and the actual velocity in the previous cycle by the interpolation cycle, with units of m / s. 2 .

[0113] If, within the current interpolation cycle, the calculated actual command acceleration for any axis exceeds the corresponding allowable acceleration dynamic range, the multi-axis motion interpolator executes multi-axis synchronous deceleration logic. If, after synchronous deceleration, the actual command acceleration for each axis still cannot fall within the corresponding allowable acceleration dynamic range, or if the allowable acceleration dynamic range does not have a valid feasible range, the multi-axis motion interpolator stops trajectory output for the current interpolation cycle and triggers a safety stop or clamping force compensation process.

[0114] The multi-axis motion interpolator calculates the required deceleration factor for the over-limit axis based on the deviation between the actual commanded acceleration and the boundary value, and extracts the maximum value of the deceleration factor among all axes as the global deceleration factor.

[0115] The multi-axis motion interpolator uses a global deceleration factor to proportionally attenuate the current nominal tangential velocity, thereby synchronously adjusting the command speed and command acceleration of all participating servo axes. This ensures that the actual command acceleration of each axis is within the allowable dynamic acceleration range, while maintaining the spatial geometry of the machining trajectory without distortion.

[0116] The S509 multi-axis motion interpolator converts the dynamically corrected position increments of each axis into servo drive commands. The interpolator synchronously distributes these commands to the position loop inputs of each servo drive via an industrial Ethernet bus. Each servo drive uses a distributed clock for synchronization, with a clock synchronization accuracy configured to 100ns. The servo motors respond to the position commands, driving the machine tool actuators to perform cutting operations according to the corrected spatial trajectory.

[0117] During interpolation machining, the clamping control module continuously monitors the actual constraint force value fed back by the clamping force sensor. If the actual constraint force decreases and falls below the preset safety clamping threshold, the system triggers an emergency stop, blocking the interpolation output and locking the mechanical brakes of each axis. The preset safety clamping threshold is determined by multiplying the nominal clamping force of the clamping mechanism by a safety factor, with the safety factor ranging from 0.75 to 0.85. This verification process reduces the risk of workpiece slippage caused by changes in cutting load.

[0118] After machining, the workpiece experiences form and position deviations due to the stress relief during clamping and the cutting force load during machining. The clamping control module obtains the actual spatial error of the workpiece surface through online detection and spatial geometric projection, and extracts the error components in the machining direction through vector decomposition. The process of obtaining and decomposing the actual spatial error vector by the clamping control module includes the following steps: S601, the clamping control module calls the online detection program to control the contact probe on the CNC machine tool spindle to move to the workpiece surface. The contact probe touches the preset measurement feature points on the workpiece surface to obtain the actual coordinate data of each measurement feature point in the machine tool coordinate system.

[0119] The number of preset measurement feature points is configured to be 4 to 12 based on the complexity of the workpiece surface, and they are distributed on the geometric reference surface and the profile surface of the processing area.

[0120] S602, the clamping control module retrieves the design nominal geometric model from the CNC system, extracts the nominal coordinate data corresponding to each measured feature point, and the surface unit normal vector at that point. The clamping control module uses the currently updated workpiece coordinate system transformation matrix to transform the actual coordinate data in the machine tool coordinate system to the workpiece coordinate system, and performs vector subtraction between the transformed actual coordinate data and the nominal coordinate data to obtain the three-dimensional position deviation vector of each measured feature point.

[0121] S603, the clamping control module performs a dot product operation between the three-dimensional position deviation vector and the corresponding surface unit normal vector, performs vector decomposition, extracts the actual spatial error of the normal along the normal direction of the machined surface, and combines them to construct the actual spatial error vector.

[0122] Actual spatial error of normal to measured feature points The calculation formula is expressed as: ; In the formula, Indicates the first The actual spatial error of the normal vector of each measured feature point is expressed in meters. Indicates the first position in the workpiece coordinate system The actual coordinate vector of each measured feature point, in meters; This represents the nominal coordinate vector, with units of meters (m). This represents the surface unit normal vector, which is dimensionless. The actual spatial error vector is generated by the measurements of each feature point. Composition, as 3D column vector, This represents the total number of feature points measured.

[0123] S604, the clamping control module performs out-of-tolerance verification logic on the actual spatial error vector. The clamping control module extracts the acceptable geometric tolerance threshold and the allowable repair limit threshold from the CNC system process document. The acceptable geometric tolerance threshold is configured with a value range of 1×10-1. -5 m to 5×10 -5 m, the range of values ​​for the allowed repair limit threshold is configured to be 1×10. -4 m to 5×10 -4 m.

[0124] The clamping control module extracts the maximum absolute value of each element in the actual spatial error vector as the comprehensive machining form and position error. If the comprehensive machining form and position error is determined to be less than or equal to the acceptable form and position tolerance threshold, the clamping control module outputs a machining qualified signal and ends the machining process.

[0125] If the overall machining form and position error is determined to be greater than the qualified form and position tolerance threshold but less than or equal to the allowable repair limit threshold, the clamping control module determines that the workpiece is in a repairable state, maintains the forced constraint state, and feeds back the actual spatial error vector to the front control loop, recalculates the reverse offset transformation matrix to correct the tool path and performs secondary machining.

[0126] The residual update module establishes a rigid body small pose error model based on the three-dimensional position deviation vector of multiple measured feature points. This model represents the spatial error of the workpiece relative to the nominal workpiece coordinate system as a six-dimensional spatial residual vector, which includes translational residuals along the X, Y, and Z axes, as well as torsional residuals around the X, Y, and Z axes.

[0127] The residual update module substitutes the nominal coordinates, actual coordinates, and corresponding surface unit normal vectors of each measured feature point into the rigid body small pose error model, and solves for the six-dimensional spatial residual vector using the least squares algorithm. When the aforementioned measuring device only outputs the actual spatial error of the normal direction, the residual update module constructs an observation matrix based on the normal projection relationship at each measured feature point, and uses the least squares algorithm to solve for the six-dimensional spatial residual vector that minimizes the sum of the squares of the residuals of each normal error.

[0128] The residual update module defines the first three elements of the six-dimensional residual vector as translation error components and the last three elements as torsional error components. The translation error components are used to update the force and error compliance transformation matrix in the recursive least squares algorithm, while the torsional error components are used to update the weight mapping matrix in the gradient descent algorithm.

[0129] If the overall machining form and position error is determined to be greater than the allowable repair limit threshold, the clamping control module determines that the workpiece is out of tolerance and should be scrapped, blocks subsequent machining operations, and outputs a machining out-of-tolerance alarm signal.

[0130] In a multi-axis clamping and CNC machining collaborative system, the compliance transformation matrix is ​​used to characterize the mapping relationship between the control force applied by the clamping mechanism and the spatial pose change of the workpiece. Due to mechanical wear during fixture use and batch-to-batch fluctuations in the physical properties of the workpiece material, the initially set compliance parameters deviate from the actual state.

[0131] The clamping control module uses the actual spatial error vector obtained by online detection after machining, and the translational error component calculated from the actual spatial error vector, to update the force and error compliance transformation matrix through a recursive least squares algorithm, so as to improve the accuracy of pose compensation during subsequent workpiece machining.

[0132] The force and error compliance transformation matrix is ​​used to characterize the mapping relationship between the generalized clamping force vector and the translational error components after machining, which is different from the aforementioned energy and yield compliance matrix. The process of the clamping control module performing compliance transformation matrix correction based on recursive least squares method includes the following steps: S605, the clamping control module obtains the actual spatial error vector of the current machining cycle and retrieves the current generalized force vector collected by the clamping force sensor. The actual spatial error vector is... The column vector is obtained from the aforementioned online detection and vector decomposition calculation; the generalized force vector is... A column vector whose internal elements consist of the normal clamping forces applied by each clamping mechanism in the multi-axis clamping system.

[0133] In one embodiment of the present invention, when the algorithm runs for the first time and the cycle index is corrected to 1, the clamping control module initializes the initial force and error compliance transformation matrix and the initial covariance matrix. The initial force and error compliance transformation matrix are preset based on the stiffness of the fixture structure according to the finite element analysis theory. Its input is a generalized force vector, and its output is either a translation error component or an error component projected onto the normal of a selected measurement point. The initial covariance matrix is ​​configured as a diagonal matrix, with its diagonal elements ranging from 1×10⁻⁶. 3 Up to 1×10 5 .

[0134] S606, the clamping control module combines the compliance transformation matrix of the previous correction cycle with the current generalized force vector to calculate the prediction error vector for the current correction cycle. The prediction error vector represents the residual between the theoretical form and position error estimated based on the current force and the error compliance transformation matrix and the actual form and position error obtained from actual measurement.

[0135] Prediction error vector The calculation formula is expressed as: ; In the formula, Indicates the first The prediction error vector under each correction period is: A dimensional column vector whose elements are in units of m; Indicates the first The actual spatial error vector under each correction period is: A dimensional column vector whose elements are in units of m; Indicates the first The force and error compliance transformation matrix under each correction cycle is: An n-order matrix whose internal elements are in units of m / N; Indicates the first The generalized force vector under each correction period is: A dimensional column vector whose elements are in units of N; This represents the correction cycle index, which is a positive integer.

[0136] S607, the clamping control module calculates the gain vector for the current correction cycle. The gain vector is calculated based on the covariance matrix of the previous correction cycle, the current generalized force vector, and a preset forgetting factor, and is used to determine the weighting of the current residual for updating the matrix parameters.

[0137] Gain vector The calculation formula is expressed as: ; In the formula, Indicates the first The gain vector under each correction period is: A dimensionless column vector; Indicates the first The covariance matrix under each correction period is: A square matrix of order, dimensionless; The forgetting factor is dimensionless and its value range is configured to be 0.950 to 0.995. Its specific value is preset based on the time-varying noise characteristics of the CNC machining system. This represents the matrix transpose operator.

[0138] S608, the clamping control module uses the gain vector and prediction error vector to recursively update the compliance transformation matrix and covariance matrix.

[0139] Flexibility Transformation Matrix The corrected formula is expressed as: ; covariance matrix The update formula is expressed as: ; In the formula, Indicates the first The force and error compliance transformation matrix updated after each correction cycle is: An n-order matrix whose internal elements are in units of m / N; Indicates the first The covariance matrix updated after each correction period is: A square matrix of order 1, dimensionless.

[0140] S609, the clamping control module performs boundary convergence verification on the updated force and error compliance transformation matrix. In one embodiment of the invention, the clamping control module extracts a preset compliance physical limit boundary range from the CNC system firmware. This range is preset based on the theoretical stiffness limit range of the fixture material, and the value range of its elements is configured as 1×10. -8 m / N to 5×10 -6 m / N. The clamping control module verifies whether each element in the updated compliance transformation matrix is ​​within the compliance physical limit boundary range.

[0141] If any element in the flexibility transformation matrix exceeds the physical limit boundary of flexibility, the clamping control module determines that the recursive calculation is subject to measurement noise interference and there is a risk of divergence. It resets the over-limit element to the corresponding boundary value and resets the current covariance matrix to the preset initial diagonal matrix in order to suppress the covariance saturation phenomenon and cut off the accumulation of error. If all elements are determined to be within the physical limit range of compliance, the clamping control module stores the updated compliance transformation matrix in the CNC system model database as the initial optimization matrix for performing kinematic mapping and pose compensation during the machining of the next batch of workpieces. This verification process is used to maintain the physical rationality of parameter identification and reduce the risk of algorithm dead zones and system malfunctions caused by abnormal sensor data.

[0142] In a multi-axis clamping system, the generation of micro-unloading parameters and step displacement thresholds depends on a pre-established probe parameter mapping network, the output characteristics of which are determined by a weight mapping matrix.

[0143] After machining, the clamping control module, based on the torsional error component calculated from the measured actual spatial error vector, uses the gradient descent algorithm to perform reverse optimization and update the weight mapping matrix. This corrects the mapping relationship between the micro-unloading parameters and the step displacement threshold, and reduces the form and position deviations of subsequent batches of workpieces. The process of reverse optimization of the mapping weights based on the gradient descent algorithm performed by the clamping control module includes the following steps: S610, the clamping control module extracts the actual spatial error vector and input state feature vector of the current processing batch. In one embodiment of the invention, the front-end control mapping network is configured as a single-layer linear mapping network model. The actual spatial error vector is... The column vector is obtained from the feedback of the detection and vector decomposition module, and the torsional error component is obtained by solving the actual spatial error vector through the rigid body small pose error model.

[0144] The input state feature vector is a normalized dimensionless vector. The column vector contains the integral of the preceding work, the mean of the sliding window samples, the variance of the sliding window samples, the blank machining allowance, the material hardness coefficient, the target depth of cut, and the commanded feed rate.

[0145] Before the first batch of processing, the initial weight mapping matrix is ​​initialized and assigned values ​​through offline finite element analysis. The clamping control module constructs a scalar loss function with the goal of minimizing torsional error while taking into account translational error constraints. The scalar loss function is defined as half of the square of the L2 norm of the torsional error component, or as half of the square of the weighted L2 norm of the torsional error component and the translational error component.

[0146] scalar loss function The calculation formula is expressed as: ; In the formula, Indicates the first scalar loss function for each processing batch, in units of m 2 ; Indicates the first The actual spatial error vector for each processing batch is: A dimensional column vector whose elements are in units of m; Represents the matrix transpose operator; This represents the batch index, which is a positive integer.

[0147] S611, the clamping control module calculates the gradient matrix of the scalar loss function with respect to the weight mapping matrix based on the chain rule and the input state feature vector of the current batch.

[0148] The sensitivity matrix of the detection parameters is used as the Jacobian matrix in the gradient calculation. The clamping control module obtains the sensitivity of the torsional error component relative to the step displacement threshold and the micro-unloading parameter based on historical calibration data or online perturbation identification results, and calculates the gradient matrix by combining the force and error compliance transformation matrix updated by the recursive least squares method in the previous step.

[0149] gradient matrix The calculation formula is expressed as: ; In the formula, Indicates the first The gradient matrix corresponding to the weight mapping matrix for each processing batch is: An m-order matrix whose elements are in units of m. 2 / N; Indicates the first The updated compliance transformation matrix for each processing batch is as follows: An n-order matrix whose internal elements are in units of m / N; Indicates the first The input state feature vector for each processing batch is: A dimensionless column vector; The dimension of the generalized force vector is the same as the number of clamping mechanisms.

[0150] S612, the clamping control module, combined with the preset learning rate, performs parameter updates on the weight mapping matrix along the negative direction of the gradient matrix.

[0151] Weight mapping matrix The update formula is expressed as: ; In the formula, Indicates that the updated version is used for the first... The weight mapping matrix for each processing batch is a 2×D matrix, and its output corresponds to the step displacement threshold and micro-unloading parameters. Indicates the first The weight mapping matrix used in each processing batch is a 2×D matrix; The learning rate is represented by a unit determined based on the dimensions of the input state feature vector, the torsional error component, and the weight mapping matrix, and its value range is configured to be from 0.001 to 0.050.

[0152] S613, the clamping control module performs out-of-bounds truncation processing on the updated weight mapping matrix. The clamping control module extracts the step displacement threshold boundary, micro-unloading parameter boundary, and servo clamping effective travel boundary from the CNC system safety configuration file, and compares and verifies the step displacement threshold and micro-unloading parameter obtained by applying the updated weight mapping matrix to the input state feature vector with the corresponding boundary thresholds.

[0153] If the step displacement threshold or micro-unloading parameter is determined to exceed the corresponding boundary threshold, the clamping control module assigns the parameter the corresponding boundary threshold value and sets the learning rate accordingly. The attenuation is performed according to a preset ratio, with the attenuation coefficient configured to be between 0.5 and 0.8. This operation is used to control the trigger position and retraction range of the servo fixture's detection action to be within the mechanical safety travel range, reducing the risk of workpiece instability due to excessive detection displacement or abnormal trigger position.

[0154] Specific application examples: This invention discloses a closed-loop control method for multi-station machining of irregularly shaped hardware parts for electric sofas. The method evaluates the material stress state by integrating the work done in the preceding station and adaptively plans the micro-unloading detection trajectory; during the detection period, a high-frequency flutter signal is injected to extract the clamping mechanical hysteresis characteristics and construct a spatial elastic internal stress feature vector; Subsequently, the internal stress characteristics are mapped to the expected geometric springback deviation using the compliance transformation matrix and the Jacobian matrix, and reverse offset compensation machining is performed on the CNC nominal coordinates while maintaining the forced constraint state; finally, based on the online measurement residual, the system compliance and weight mapping matrix is ​​updated in reverse by the recursive least squares method and the gradient descent algorithm.

[0155] The results show that this method effectively quantifies and releases the spatial distribution of residual stress inside the workpiece, reduces the average comprehensive machining form and position error by 74.43%, reduces the batch error sample variance by 88.16%, and effectively eliminates the clamping instability interference phenomenon during the machining process.

[0156] In a specific implementation, taking the multi-axis machining of thin-walled aerospace aluminum alloy parts (AL7075) as an example, the system is configured with three orthogonally distributed linear servo clamping mechanisms. In the preceding milling station, the energy assessment module calculates the integral of the preceding work, generating a trigger displacement threshold of 0.068m and a micro-unloading amplitude of 0.0065m. After the servo fixture moves to the trigger position, it reverses and injects a high-frequency chatter signal with a frequency of 120Hz and an amplitude of 10% of the rated torque into the current control loop.

[0157] After frequency domain separation and spatial domain resampling, the current curves of the servo axis forward loading and reverse unloading are extracted and integrated to construct a spatial elastic internal stress feature vector characterizing the equivalent elastic dissipation work. This vector is then used to decouple and output the expected spatial geometric springback deviation to generate dynamic compensation coordinates. To prevent workpiece slippage caused by the coupling between position compensation and external cutting force, the multi-axis motion interpolator calculates the allowable maximum / minimum acceleration boundaries for each servo axis in real time.

[0158] The dynamic verification formula for the maximum allowable acceleration is as follows: ;in, Indicates the first Maximum permissible acceleration (m / s2) for each servo axis; The preset static friction coefficient between the workpiece and the fixture is 0.2. The normal clamping force applied to the clamping mechanism is taken as 4000N; The real-time cutting force along the axial direction is taken as 450N; The workpiece mass is taken as 8.0 kg. The current maximum acceleration boundary is calculated to be 43.75 m / s². Based on this, the system verifies the command acceleration and introduces a global deceleration factor to ensure safe completion of trajectory correction cutting under forced anti-slip constraint.

[0159] After the compensation process is completed, the system obtains the actual measurement residual of the online probe and performs adaptive closed-loop update of the force and error compliance transformation matrix.

[0160] Flexibility transformation matrix of the current correction period The update is based on the recursive least squares algorithm, and its iterative formula is as follows: ;in, This is the updated force and error compliance transformation matrix (m / N); This is the historical compliance transformation matrix for the previous correction period; The prediction error vector (m) is calculated by combining the current actual space error vector and the generalized force vector. This is the transpose of the gain vector calculated based on the covariance matrix and the forgetting factor (default is 0.980).

[0161] Through this recursive formula, the system can dynamically adjust the compliance parameters according to the actual small form and position error residuals, effectively eliminating the nonlinear deviations introduced by mechanical component wear and material batch hardness fluctuations, and improving the convergence accuracy of subsequent batch workpiece detection mapping and reverse pose compensation.

[0162] Reference Appendix Figure 5 The horizontal axis represents the clamping displacement (m), and the vertical axis represents the low-frequency actual feedback current (A) after filtering out flutter. In the legend, the upper solid line represents the forward loading current curve, and the lower dashed line represents the reverse unloading current curve. The closed shaded area formed by the envelope of the two curves represents the elastic dissipation work equivalent to the internal stress release level. Reference Appendix Figure 6 The horizontal axis compares the traditional rigid clamping control group with the experimental group of the adaptive method of this invention, and the vertical axis represents the comprehensive machining form and position error (m). In the figure, the error box of the experimental group is compressed and the median is close to the minimum value. Reference Appendix Figure 7 The horizontal axis represents the batch index, and the vertical axis represents the scalar loss function on a logarithmic scale (m). 2 As shown in the figure, the solid broken line enters a horizontally stable convergence interval after the 5th batch, verifying the rapid response of the weight mapping optimization mechanism; Reference Appendix Figure 8 The horizontal axis represents the interpolation time (s), and the vertical axis represents the acceleration (m / s²). 2 In the illustration, the two dashed lines that fluctuate with the cutting force are the maximum and minimum allowable acceleration boundary envelopes, respectively. The solid line representing the actual commanded acceleration is strictly constrained within the dashed line envelope, which intuitively verifies the effectiveness of the forced constraint anti-slip logic.

Claims

1. A closed-loop control method for multi-station processing of irregularly shaped hardware parts for electric sofas, characterized in that, Includes the following steps: The work integral, sample mean and variance are calculated based on the power data of the preceding workstation, and the micro-unloading parameters and step displacement threshold are generated using the weight mapping matrix. The servo fixture is controlled to operate according to the micro-unloading parameters and the step displacement threshold, and a high-frequency torque chatter signal is superimposed within the micro-unloading detection action range. Obtain the total output torque current sequence and the corresponding clamping displacement sequence fed back by the servo fixture, calculate the integral area of ​​the hysteresis loop and construct the spatial elastic internal stress feature vector; By using a preset energy and yielding flexibility matrix and a kinematic Jacobian matrix, the spatial elastic internal stress characteristic vector is converted into an expected spatial geometric rebound deviation vector. Based on the expected spatial geometric rebound deviation vector, the nominal target coordinates are reverse-biased to generate dynamic compensation coordinates, and the machining actuator is controlled to process at the position of the dynamic compensation coordinates. Obtain the actual spatial error vector and decompose it into translational error components and torsional error components. Update the preset force and error compliance transformation matrix using the translational error components, and reverse correct the weight mapping matrix using the torsional error components.

2. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The steps of calculating the work integral, sample mean, and variance based on the power data of the preceding workstation, and generating the micro-unloading parameters and step displacement threshold using the weight mapping matrix specifically include: After deducting the no-load power baseline and correcting the drive efficiency of the power data, the trapezoidal integral algorithm is used to accumulate the instantaneous active power within the time window to obtain the work integral. When the sample variance is less than or equal to a preset variance threshold, and the sample mean or the work integral of the current workpiece is less than or equal to a preset energy level threshold, the system is determined to be in a mechanical steady state, the micro-unloading parameter is assigned to zero, and the step displacement threshold is assigned a preset closed position value. When the sample variance is greater than the preset variance threshold, or the sample mean or the work integral of the current workpiece is greater than the preset energy level threshold, the system is determined to be in a mechanically unsteady state. The work integral of the current workpiece is substituted into the mapping calculation model, and the step displacement threshold and micro-unloading parameters of the current workpiece are generated by matrix multiplication.

3. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The steps for controlling the operation of the servo fixture based on the micro-unloading parameters and the step displacement threshold specifically include: The clamping process is divided into a rapid approach section, a micro-unloading detection section, and a secondary loading and locking section. In the rapid approach phase, the servo gripper is controlled to run. When it is determined that the servo gripper has reached the position limited by the step displacement threshold, the motion state is switched to the micro-unloading detection phase. In the micro-unloading detection section, the target endpoint of the reverse retraction is calculated. The coordinates of the target endpoint are represented by the step displacement threshold minus the micro-unloading parameter. The servo fixture is controlled to retract to the target endpoint in a direction away from the workpiece. In the secondary loading and locking section, the servo clamp is controlled to advance from the target endpoint back towards the final fastening coordinate direction until the preset clamping torque threshold is reached; The rapid approach segment, the micro-unloading detection segment, and the secondary loading locking segment all use third-order polynomial curves for position interpolation.

4. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 3, characterized in that, The step of superimposing a high-frequency torque chatter signal within the micro-unloading detection action range specifically includes: A discrete sine wave sequence is generated as the flutter compensation torque. The flutter compensation torque is then added to the basic motion torque command through an addition operation to form the final current loop input value. During the flutter signal injection, the actual current feedback value and encoder angular velocity feedback value inside the servo driver are read to extract the real-time equivalent friction torque of the contact surface. The real-time equivalent friction torque is compared with the preset critical static friction torque. When the real-time equivalent friction torque is less than the product of the preset critical static friction torque and the preset conversion coefficient, the dynamic Coulomb friction conversion is determined to be successful.

5. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The steps of obtaining the total output torque current sequence and the corresponding clamping displacement sequence of the servo fixture feedback, and calculating the integral area of ​​the hysteresis loop, specifically include: The total output torque current sequence is subjected to band-stop filtering to filter out the narrow-band current component corresponding to the high-frequency torque flutter signal, and then input into a first-order low-pass filter for smoothing to output a low-frequency current waveform. The motion direction flag of the clamping displacement sequence is analyzed, and the low-frequency current waveform is separated to generate a forward loading sequence and a reverse unloading sequence. A cubic spline interpolation algorithm is used to resample the forward loading sequence and the reverse unloading sequence at equal step sizes within the overlapping displacement interval to generate forward current curves and reverse current curves. The trapezoidal numerical integration algorithm is used to calculate the cumulative integration of the current difference between the forward current curve and the reverse current curve at corresponding points to obtain the hysteresis loop integration area.

6. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The steps for constructing the spatial elastic internal stress feature vector specifically include: Read the hysteresis loop integral area of ​​each servo axis involved in the clamping, as well as the hardware parameters of each servo axis, including the motor torque constant and the mechanical transmission coefficient. For each servo axis, the hysteresis loop integral area, the motor torque constant, the mechanical transmission coefficient, and the length unit conversion factor are multiplied to convert the drive current into linear force work applied to the workpiece by the end effector, thus obtaining the equivalent elastic dissipation work. The equivalent elastic dissipation work of each servo axis is arranged in the index order of the spatial coordinate axes to construct the spatial elastic internal stress characteristic vector in column vector form.

7. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The step of converting the spatial elastic internal stress eigenvector into the expected spatial geometric rebound deviation vector by using a preset energy and yielding flexibility matrix and a kinematic Jacobian matrix specifically includes: The initial position compensation vector is calculated by multiplying the initial energy and yielding compliance matrix with the spatial elastic internal stress eigenvector. The initial position compensation vector is limited using the maximum allowable deformation displacement parameter and the lower limit dead zone threshold, and then combined to generate the actual position compensation vector. The actual position compensation vector is multiplied by the kinematic Jacobian matrix to calculate the expected spatial geometric rebound deviation vector, which includes translation and rotation components.

8. The multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The steps for generating dynamic compensation coordinates by reverse offsetting the nominal target coordinates based on the expected spatial geometric rebound deviation vector specifically include: Analyze the underlying CNC machining program and extract the nominal target coordinates from the trajectory flow; The expected spatial geometric springback deviation vector is decoupled and separated into a translational deviation vector and a rotational deviation vector. A three-dimensional orthogonal rotation matrix is ​​constructed based on the translational deviation vector and the rotational deviation vector, and the homogeneous deviation matrix characterizing the spatial pose deviation of the workpiece is generated by combining them. Perform the inverse matrix operation on the homogeneous bias matrix to generate the inverse bias transformation matrix; The nominal target coordinates are converted into homogeneous coordinates and multiplied with the inverse bias transformation matrix to obtain the corrected target coordinates. When it is confirmed that each axial component in the corrected target coordinates is within the corresponding maximum travel interval, the corrected target coordinates are used as the dynamic compensation coordinates.

9. A multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The specific steps of the controlled machining actuator in machining at the dynamically compensated coordinate position are as follows: Under the condition of maintaining forced constraints, based on the normal clamping force applied to the workpiece by the clamping mechanism corresponding to each servo axis, the mass of the workpiece, the real-time cutting force, and the static friction coefficient between the workpiece and the contact surface of the fixture, the maximum allowable acceleration and the minimum allowable acceleration of each servo axis under the current interpolation cycle are calculated, and the dynamic range of allowable acceleration is established. Calculate the actual commanded acceleration of each motion axis in the current cycle. When the actual commanded acceleration exceeds the allowable acceleration dynamic range, calculate the deceleration factor required for the over-limit axis based on the deviation between the actual commanded acceleration of the over-limit axis and the boundary value, and extract the maximum value as the global deceleration factor. By using the global deceleration factor to proportionally attenuate the current nominal tangential velocity, the command speed and command acceleration of all participating servo axes are adjusted synchronously so that the actual command acceleration of each axis is within the allowable acceleration dynamic range.

10. A multi-station closed-loop control method for processing irregularly shaped hardware parts for electric sofas according to claim 1, characterized in that, The step of using the torsional error component to reverse correct the weight mapping matrix specifically includes: The actual coordinate data of preset measurement feature points on the workpiece surface are obtained by a contact probe. The three-dimensional position deviation vector of each measurement feature point is calculated by combining the corresponding nominal coordinate data. The three-dimensional position deviation vector is then decomposed by performing a dot product operation with the corresponding surface unit normal vector. The actual spatial error vector is then constructed by combining the vectors. Based on the nominal coordinate data, actual coordinate data and corresponding surface unit normal vector of each of the measured feature points, the rigid body small pose error model is substituted, and the six-dimensional spatial residual vector is solved by the least squares algorithm. The first three elements of the six-dimensional spatial residual vector are defined as the translation error component, and the last three elements are defined as the torsional error component. Substitute the generalized force vector under the current correction period, the prediction error vector and the gain vector calculated based on the translation error components into the recursive least squares algorithm model to recursively update the preset force and error compliance transformation matrix. Extract the input state feature vector containing the work integral, sample mean, and sample variance to construct a scalar loss function. Combine the input state feature vector to calculate the gradient matrix of the scalar loss function with respect to the weight mapping matrix. Update the parameters of the weight mapping matrix along the negative direction of the gradient matrix.

Images