Multi-configuration indexing transfer dual closed-loop compensation method and system

Abstract

The application relates to the technical field of multi-station five-axis numerical control machining and automatic transfer, in particular to a double-closed-loop compensation method and system for multi-configuration indexing transfer, which is applied to a multi-station five-axis flexible parallel numerical control island, and the numerical control island comprises multiple five-axis main machines, a transfer mechanism for transferring workpieces between the five-axis main machines, unified zero-point tooling installed at each station, a partition temperature control system for respectively controlling temperatures of cutting partitions and polishing partitions, and a base vibration sensing unit, and comprises the following steps: S1, configuration identification; S2, error library calling; and S3, double-closed-loop compensation. The application establishes two independent indexing error databases for milling and polishing, synchronously incorporates temperature and vibration into transfer compensation, takes the reference error vector calibrated under the reference temperature as the basis, superimposes an environment correction vector composed of real-time temperature deviation and vibration to form feedforward pre-compensation, and combines secondary positioning feedback after positioning to form a double-closed-loop, so that the slowly-varying error of the environment is eliminated.

Description

Technical Field

[0001] This application relates to the field of multi-station five-axis CNC machining and automated transfer technology, specifically to a dual closed-loop compensation method and system for multi-configuration indexing transfer that integrates feedforward pre-compensation and feedback closed loop and is applicable to multi-configuration transfer mechanisms. Background Technology

[0002] The multi-station five-axis flexible parallel CNC island adopts a production line architecture with multiple five-axis mainframes operating in flexible parallel operation. Workpieces (such as titanium alloy aerospace parts, GPU heat dissipation modules, sapphire spherical parts, diamond tools, etc.) are sequentially transferred between the mainframes by a transfer mechanism, completing the XYZ milling and AC oscillating head polishing processes in sequence. Common flexible production line configurations for the transfer mechanism include: cam rotary table type, triangular collaborative robot type, and H-type collaborative robot type.

[0003] Existing indexing and positioning technologies have the following shortcomings: Firstly, traditional indexing programs are only compatible with single cam turntable configurations, and the indexing algorithm is fixedly coupled with the equal division mapping of the turntable. The compensation methods for the indexing mechanism rely on fixed control parameters and simple compensation algorithms, which are incompatible with robot transportation and difficult to cope with errors caused by environmental changes.

[0004] Secondly, the robot positioning error compensation technologies are all based on pure feedforward or calibration approaches, and there is no feedback mechanism after compensation, which cannot eliminate the micro gaps and random residual deviations between the transfer mechanism and the workstation; moreover, no separate databases have been built for the milling and polishing processes.

[0005] Third, the machine tool thermal error compensation technology only applies to the machine tool spindle and does not involve the transportation process, nor does it incorporate the vibration and temperature of the base into the transportation compensation.

[0006] Fourth, zero-point positioning systems (such as commercially available zero-point positioning quick-change systems) provide a unified mechanical reference, but existing technologies do not include zero-point offset in the indexing error calculation. The transfer positioning and the workstation reference belong to two different coordinate systems, and recalibration is required after the model is changed.

[0007] Therefore, the transfer drift of milled blanks and polished semi-finished products cannot be synchronously corrected under the effects of temperature (22℃ in the cutting zone and 20℃ in the polishing zone) and base vibration, resulting in excessive dimensional dispersion in batch processing; and the algorithm cannot be reused when the production line is deployed in a mixed manner (turntable and robot are mixed). Summary of the Invention

[0008] This application aims to address the problems in existing indexing and transfer compensation methods, such as single configuration, lack of positioning feedback, failure to synchronously incorporate temperature and vibration compensation, and separation of zero-point offset and indexing error calculation. It provides an indexing and transfer dual-closed-loop compensation method and system that is compatible with multiple configuration transfer mechanisms, integrates feedforward pre-compensation and feedback closed loop, and has self-learning capabilities.

[0009] On the one hand, this application provides a dual closed-loop compensation method for multi-configuration indexing and transfer, applied to a multi-station five-axis flexible parallel CNC island. The CNC island includes multiple five-axis mainframes, a transfer mechanism for transferring workpieces between the five-axis mainframes, a unified zero-point fixture installed at each station, a zoned temperature control system for separately controlling the temperature of the cutting zone and the polishing zone, and a base vibration sensing unit, including the following steps: S1. Configuration Recognition: Read the hardware interface identifier of the transfer mechanism, automatically identify whether the configuration of the transfer mechanism is a cam turntable type or a collaborative robot type, and call up the kinematic model and indexing mapping relationship corresponding to the identified configuration; wherein, the indexing mapping relationship of the cam turntable type configuration is the equal division mapping of the indexing angle and the workstation number, and the indexing mapping relationship of the collaborative robot type configuration is the kinematic mapping of the target pose of each workstation and the robot joint quantity. S2. Error Database Call: Call the pre-established milling station indexing error database and polishing station indexing error database. The two databases are independent of each other. Each indexing error database stores the error vector with the indexing position, configuration, temperature range and load level as indexes. The temperature range indexes the range of workshop ambient temperature at the time of execution. The error vector is the reference error vector calibrated at the reference temperature of the corresponding process. The temperature range is used to calibrate data classification. During online compensation, the corresponding indexing error database is called according to the processing process category of the target station. The current temperature range and load level are used as filtering conditions. Interpolation is performed between adjacent indexing positions to calculate the basic error vector of the target indexing position. S3. Dual Closed-Loop Compensation: Before the transfer is executed, the cutting zone temperature, polishing zone temperature, and base vibration data are collected synchronously. An environmental correction vector determined by the temperature deviation and vibration is superimposed on the basic error vector to generate a pre-biased target pose and send it to the transfer mechanism for execution, forming a feedforward pre-compensation loop. After the transfer is completed, the in-situ detection unit set at the target station measures the actual pose of the workpiece on the unified zero-point fixture, calculates the residual deviation between the actual pose and the theoretical pose, and when the residual deviation is greater than a set threshold, drives the fine-tuning shaft on the target station side or the end fine-motion shaft of the transfer mechanism for secondary alignment until the residual deviation is not greater than the set threshold. The residual deviation is then written back to the corresponding indexing error database to update the record, forming a feedback closed loop.

[0010] Further, in step S1, the hardware interface identifier includes the type of the transfer mechanism drive bus, the number of servo axes, and the kinematic parameter package; when it is identified as a collaborative robot configuration, its topology is further identified as a triangular collaborative robot or an H-type collaborative robot, and the corresponding DH parameters are loaded.

[0011] Further, in step S2, the error vector includes three linear axis error components and two rotation axis error components; each record also includes the number of statistical samples and the range; the interpolation uses linear interpolation between adjacent indexing positions, or trilinear interpolation or Kriging interpolation between adjacent pose mesh nodes in a collaborative robot configuration.

[0012] Furthermore, the pre-establishment of the database in step S2 includes a database establishment and calibration step: controlling the transfer mechanism to repeatedly position each graduation position no less than 7 times, the measurement unit collects the actual pose each time, takes the mean value and writes it into the corresponding record, and the range is used as the dispersion index; when the difference between the norm of the residual deviation of the feedback closed loop back writing after normalization with the zero-point tooling clamping radius as the characteristic length and the norm of the error vector of the corresponding record in the database after the same normalization exceeds the range, the recalibration of the corresponding record is triggered.

[0013] Further, in step S3, the environmental correction vector is determined according to E_env = K_T·ΔT + K_v·A_v, where K_T and K_v are the temperature influence coefficient vector and vibration influence coefficient vector determined by calibration tests, respectively; when the target station is a milling station, ΔT is the deviation between the measured temperature of the cutting zone and its set reference temperature of 22℃; when the target station is a polishing station, ΔT is the deviation between the measured temperature of the polishing zone and its set reference temperature of 20℃; A_v is the effective amplitude of the base vibration; the superposition is to add each component of the environmental correction vector to the corresponding component of the basic error vector.

[0014] Further, in step S3, the norm of the residual deviation is defined as: the Euclidean norm calculated together with the linear deviation component after converting the rotational deviation component into an equivalent linear deviation using the clamping radius of the zero-point tooling as the characteristic length; the set threshold is 0.0002mm-0.001mm for the cam turntable configuration and 0.001mm-0.005mm for the collaborative robot configuration; the number of secondary alignment iterations does not exceed 3.

[0015] Furthermore, the method also includes a zero-point reference calculation step: the zero-point offset vector of the unified zero-point fixture is incorporated as a fixed parameter into the coordinate transformation chain of each station, so that the error vector calculation in step S2 and the residual deviation calculation in step S3 both use the zero-point offset vector as a unified reference; the zero-point offset vector is obtained through calibration and includes the fixture installation error; when the workpiece or fixture is changed, the zero-point offset vector remains unchanged, and only the transformation parameters of the workpiece coordinate system relative to the zero-point offset vector are updated.

[0016] Furthermore, the CNC island also includes a CT equalization module, which synchronously adjusts the compensation execution level in step S3 according to the cycle margin of each station; when the cycle margin of the target station is lower than the set value, the compensation execution level is reduced. The adjustment elements of the execution level include the number of secondary alignment iterations, the interpolation order, and the timing of compensation execution; when the level is reduced, the calculation of part of the environmental correction amount is moved forward to the idle period of the previous station, and the level is raised again after the cycle margin is restored.

[0017] Furthermore, the calculation and output of the feedforward pre-compensation loop and the feedback closed loop are executed by a high-speed hardware real-time closed-loop controller; the milled blank and the polished semi-finished product are synchronously transferred within the same production cycle.

[0018] On the other hand, this application also provides a dual closed-loop compensation system for multi-configuration indexing transfer, comprising: The configuration recognition module is used to read the hardware interface identifier of the transfer mechanism and automatically identify its configuration; The milling station indexing error database and the polishing station indexing error database are independent of each other and store error vectors with indexing position, configuration, temperature range and load level as indexes. The error vector is a reference error vector calibrated at the reference temperature of the corresponding process. The feedforward pre-compensation module is used to obtain the basic error vector from the corresponding error database, and to superimpose the environmental correction vector on the basic error vector according to the partition temperature and base vibration data to generate the pre-biased target pose. The secondary alignment feedback module includes an in-situ detection unit and a fine-tuning shaft drive unit set at the target workstation. It is used to calculate the residual deviation and perform secondary alignment, and write the residual deviation back to the corresponding indexing error database to update the record. The zero-point reference calculation module is used to incorporate the zero-point offset vector of the unified zero-point tooling into the coordinate transformation chain of each workstation. And a CT equalization adjustment module, used to synchronously adjust the compensation execution level according to the cycle margin of each workstation.

[0019] Compared with the prior art, the beneficial effects of this application are as follows: This application enables the same compensation algorithm to be adapted to three configurations—cam turntable, triangular collaborative robot, and H-type collaborative robot—by automatically recognizing the configuration and calling the corresponding kinematic model / indexing mapping relationship, thus achieving algorithm reuse when deploying mixed configurations on the production line. This application establishes two independent indexing error databases for milling and polishing, and incorporates temperature and vibration into the transfer compensation simultaneously. Based on the reference error vector calibrated at the reference temperature, it superimposes the environmental correction vector calculated by real-time temperature deviation and vibration to form a feedforward pre-compensation. Combined with the secondary alignment feedback after positioning, it forms a double closed loop, which eliminates both the environmental gradual change error and the random positioning error, and achieves sub-micron level transfer positioning accuracy. This application incorporates the zero-point offset vector of the zero-point tooling into a unified coordinate transformation chain, so that the calculation of indexing error and residual deviation are both based on this offset as a unified reference. After the changeover, there is no need to recalibrate; only the workpiece coordinate system transformation parameters need to be updated, which significantly shortens the changeover time. This application writes the residual deviation of the feedback closed loop back to the error database, and combines it with a recalibration triggering mechanism that uses the range as the dispersion index, so that the compensation accuracy gradually converges as the system runs, forming a self-learning closed loop. This application uses a CT equalization module to dynamically adjust the compensation execution level according to the cycle margin of each workstation, downshifting when the cycle is tight and upshifting when the cycle is ample, thus achieving a dynamic balance between accuracy and production efficiency.

[0020] The beneficial effects listed above are not exhaustive of all advantages. Other potential beneficial effects and detailed technical implementation methods will be further disclosed in the embodiments or other descriptive sections of this application. Detailed Implementation Example

[0021] This embodiment uses a multi-station, five-axis flexible parallel CNC island production line as an application scenario. The production line includes three five-axis mainframes: two XYZ milling mainframes and one AC swivel polishing mainframe, arranged sequentially along the production line. The transfer mechanism employs a mixed-line deployment of a cam-rotor configuration and an H-type collaborative robot configuration: the cam-rotor has 8 stations, driven by a servo motor via a reducer, and can achieve fine-tuning pre-offset of the indexing angle under electrical control (unlike the fixed indexing angle of a purely mechanical cam-rotor), responsible for workpiece transfer between milling mainframes and cross-zone transfer from the milling area to the polishing area; the H-type collaborative robot has 6 stations, responsible for auxiliary transfer and flexible loading / unloading from the milling area to the polishing area. The workpiece is a titanium alloy aerospace structural blank, and the processing flow is: loading, first XYZ rough milling, second XYZ finish milling, AC swivel polishing, and unloading. The cutting zone and polishing zone are maintained at set ambient temperatures of 22℃ and 20℃ respectively by independent zone temperature control systems. Each workstation is equipped with a uniform zero-point positioning quick-change system as a unified zero-point tooling. This system is a commercially available product, and its structural principle and installation method are existing technologies, which will not be elaborated here.

[0022] The production line base is a single-piece cast marble bed. A three-axis accelerometer is installed in the middle of the bed to collect the effective amplitude of the X, Y, and Z axes of the base vibration. The selection and installation of this vibration sensor are common knowledge. Each workstation is equipped with a contact-type in-situ probe with a measurement accuracy of 0.1μm. It extends after being transported to its position via a pneumatic telescopic mechanism for measurement. A piezoelectric ceramic micro-adjustment axis with a stroke of ±10μm and a closed-loop resolution of 1nm is installed on the target workstation side to achieve sub-micron displacement compensation during secondary alignment. A six-dimensional micro-motion platform is installed at the end flange of the H-type collaborative robot, with a stroke and resolution comparable to the aforementioned micro-adjustment axis. The drive and signal acquisition of the probes, micro-adjustment axes, and micro-motion platforms are all connected to the production line's real-time EtherCAT protocol bus and uniformly scheduled by the central industrial control computer.

[0023] After system power-on initialization, the zero-point reference calculation is performed first. The zero-point offset vector of the unified zero-point fixture is written as a fixed parameter into the coordinate transformation chain. This zero-point offset vector is obtained through offline calibration. Specifically, in a constant temperature environment of 20℃, a laser tracker is used to measure the actual installation position of the zero-point fixture at each station relative to the global coordinate system of the production line, and compared with the theoretical installation position to obtain the installation error vector for each station. Taking milling station 1 as an example, the zero-point offset vector obtained by calibration is [X0,Y0,Z0,A0,B0]=[12.5μm,-8.3μm,5.1μm,2.0 arcseconds,-1.5 arcseconds]. This vector contains three linear deviation components and two rotational deviation components. The above method of calibrating the zero-point fixture with a laser tracker is existing technology. After the zero-point offset vector is written, all subsequent error calculations and pose measurements are based on the zero-point fixture coordinate system as a unified reference. When changing workpieces or tooling, the zero-point offset vector remains unchanged; only the transformation parameters of the workpiece coordinate system relative to the zero-point offset vector are updated.

[0024] After initialization, the system enters online operation mode. The following describes the execution process of this method in detail, taking a complete indexing process of transferring the workpiece from milling station 1 to milling station 2 as an example.

[0025] First, the configuration identification step is performed. The system reads the hardware interface identifier register of the transfer mechanism via the EtherCAT protocol bus. The identifier information includes a drive bus type field and a servo axis quantity field. The reading result is: the drive bus type field value is 0x01, and the servo axis quantity is 1. Based on this, the system automatically identifies the current transfer mechanism as a cam turntable configuration. In this embodiment, the cam turntable is a servo-driven indexing turntable, and its indexing position can be finely adjusted by the servo system by superimposing a fine offset on the equally divided reference angle, which is different from the fixed indexing angle of a purely mechanical cam turntable (the latter is not suitable for pre-offset compensation). The system loads the equally divided mapping relationship table corresponding to the cam turntable configuration from the configuration storage area: 8 positions correspond to 8 indexing angles, and position number i corresponds to the indexing angle θ_i = 45° × (i-1). At the same time, the drive parameters of the cam turntable are loaded, including the number of pulses per revolution of the servo motor, the reduction ratio, the electronic gear ratio, etc. The loading of these parameters is common knowledge. If the read result is 0x02, then the topology identifier field is further read (0x02A for a triangular collaborative robot and 0x02B for an H-type collaborative robot) and the corresponding DH parameter table is loaded. In this embodiment, the DH parameter table for the H-type collaborative robot contains six sets of parameters, as shown in Table 1. The modeling method for DH parameters and the forward / inverse kinematics calculation method are common knowledge in robotics and will not be elaborated here.

[0026] Table 1. Parameters of H-type Collaborative Robot (DH)

[0027] Next, the error database retrieval step is executed. The system has pre-established a milling station indexing error database and a polishing station indexing error database. The two databases have the same structure but the data is completely independent and are deployed in independent partitions of the industrial control computer's real-time database.

[0028] The database construction and calibration were completed offline during the system debugging phase. Taking the milling station indexing error database as an example, the database construction process is as follows: The zone temperature control system was set to a cutting zone reference temperature of 22℃ and stabilized for more than 2 hours to bring the production line to a thermal steady state. For the cam turntable configuration, each of the 8 indexing positions was calibrated. For each indexing position, calibration was performed under three load levels, which were divided according to the workpiece mass as unloaded (pallet empty, 0kg), half-loaded (with a standard counterweight of 3kg), and full-loaded (with a standard counterweight of 8kg). For each indexing position, the cam turntable was controlled to repeat positioning 10 times under each load level. After each transfer to the correct position, the in-situ probe at the target station measured the actual position and orientation of the workpiece on the zero-point fixture, and collected five error components: X-axis linear deviation (μm), Y-axis linear deviation (μm), Z-axis linear deviation (μm), rotational deviation around the X-axis (arcseconds), and rotational deviation around the Y-axis (arcseconds). To standardize the dimensions for norm calculation and threshold comparison, this embodiment uses the clamping radius L = 100 mm of the zero-point tooling as the characteristic length, converting the rotational deviation component from arcseconds to an equivalent linear deviation: 1 arcsecond = 4.848 μrad, equivalent linear deviation = L × θ_rad ≈ 0.485 μm. All error vector norms ||E|| herein are defined as the Euclidean norm normalized to this characteristic length, i.e., the rotational component is first converted to micrometer dimensions using a factor of 0.485 μm / arcsecond before calculating the sum of squares. It should be noted that the rotational deviation around the Z-axis is guaranteed by the accuracy of the cam turntable indexing mechanism itself (the angular positioning accuracy of the indexing plate already covers this direction), and in five-axis machining, the C-axis can cover the rotational degrees of freedom around the Z-axis. Therefore, the error vector does not include the rotational component around the Z-axis, only retaining the rotational components around the X and Y axes to reduce the vector dimension and avoid double counting with the inherent accuracy of the indexing mechanism. The arithmetic mean of 10 measurements is taken as the error vector for that graduation position under that load range and written into the database. The range of the normalized error vector norms (i.e., the difference between the maximum and minimum values) among the 10 measurements is taken as the dispersion index and stored together. The temperature range field is recorded as "22℃~24℃" according to the ambient temperature during calibration (the temperature range where the reference temperature of 22℃ is located). The calibration method for the collaborative robot configuration is similar, except that "graduation position" is replaced with "end-effector pose mesh node". Nodes are divided in Cartesian space according to equally spaced meshes (in this embodiment, the mesh spacing is 100mm×100mm×100mm) and calibrated one by one.

[0029] The above database calibration was completed at the reference temperature. It should be noted that all calibrations (including automatic recalibrations triggered by exceeding range limits during subsequent operation) were performed at the reference temperature for the corresponding process (22℃ for milling, 20℃ for polishing) to ensure that each error vector in the database does not contain thermal deformation components outside the reference temperature. During online compensation, temperature-induced deformation is entirely borne by the environmental correction vector. The significance of using the temperature range as a database index field is that, although the calibration execution temperature is uniformly the reference temperature, the seasonal thermal boundary conditions during calibration differ—for example, when the summer workshop ambient temperature is 28℃, the zoned temperature control system needs to maintain the 22℃ setpoint at higher power, and the overall temperature gradient distribution of the marble bed differs from that when the winter workshop ambient temperature is 18℃, resulting in different micro-deformation states of the bed. Even error vectors obtained at the same reference temperature may still exhibit slight systematic shifts. Therefore, the temperature range field records the range of workshop ambient temperature during calibration (e.g., "24℃~28℃" for summer calibration and "18℃~22℃" for winter calibration). During online retrieval, the system prioritizes matching calibration data with the same temperature range as the current workshop ambient temperature to improve the matching degree between database records and current operating conditions. This temperature range field is only used for classifying and filtering calibration data based on environmental conditions and does not participate in temperature correction calculations during online compensation.

[0030] The complete index structure of each record in the database is: [indexing position, configuration identifier, temperature range, load level], and the stored content is: {error vector [X,Y,Z,A,B], number of samples 10, range R, timestamp}. The only difference between the milling station indexing error database and the polishing station indexing error database is the reference temperature: the reference temperature for the milling process is 22℃, and the reference temperature for the polishing process is 20℃. During database construction, the polishing station indexing error database undergoes the exact same calibration process under the thermal steady-state condition of a reference temperature of 20℃ in the polishing area. All error vectors stored in the database are reference error vectors obtained at the corresponding process reference temperature. The temperature range field is used to identify the environmental conditions during calibration and for filtering and matching during online calls, and does not participate in the temperature compensation calculation during online compensation. During online compensation, all temperature-induced deformations are uniformly handled by the environmental correction vectors in subsequent steps.

[0031] During online invocation, the current workpiece is in the milling process, and its weight, measured by a weighing sensor, is 3.2 kg. The system reads the current temperature of the cutting zone as 22.8℃, determining it to be within the "22℃~24℃" temperature range; based on the workpiece weight of 3.2 kg, it is determined to be in the "half-load" setting. Given that the target station is milling station 2, the system invokes the milling station indexing error database. The query conditions are: [indexing angle corresponding to station 2, cam rotary type, 22℃~24℃, half-load]. The system filters the database for a subset of records that meet these composite indexing conditions, performs linear interpolation between adjacent indexing positions, and calculates the base error vector E_base for the target indexing position. For example, the indexing angle corresponding to station 2 is 45°, and the records in the database for adjacent indexing angles are 0° and 90°, corresponding to error vectors E0 and E_base, respectively. 90 Then, according to the linear interpolation formula E_base=E0+(E 90 The interpolation result is calculated as E_base = [3.2μm, -1.8μm, 2.5μm, 4.1 arcseconds, -2.3 arcseconds].

[0032] The system then proceeds to the feedforward pre-compensation loop section of the dual closed-loop compensation step. Before the transfer begins, the system synchronously collects the following sensor data via the EtherCAT bus: the cutting zone temperature sensor reading is 22.8℃, the polishing zone temperature sensor reading is 19.6℃, and the base three-axis accelerometer reading, after Fourier transform and bandpass filtering, yields the effective vibration amplitude A_v = 0.18μm. This signal processing method is common knowledge. It should be noted that, based on actual measurements, the effective amplitude difference of the base vibration in the X, Y, and Z directions does not exceed 5% (the marble bed is made of isotropic material and has a symmetrical overall casting structure). Therefore, the synthesized scalar A_v is used as the unified vibration input to avoid introducing the complexity of three-dimensional vibration coupling calibration. This simplification has been experimentally verified to introduce a model error of no more than 0.03μm, within the system's accuracy margin. The target station is the milling station, ΔT = 22.8℃ - 22℃ = 0.8℃.

[0033] The system calculates the environmental correction vector: E_env = K_T·ΔT + K_v·A_v. Here, K_T is the temperature influence coefficient vector, and K_v is the vibration influence coefficient vector, both determined through offline calibration experiments. The calibration method for the temperature influence coefficient vector is as follows: Under conditions where the base is stationary and vibration is negligible, the zoned temperature control system is heated from 20℃ to 26℃ in 1℃ increments. After each temperature step stabilizes for one hour, the change in the graduation error is measured, and linear fitting is performed on each component to obtain the values ​​of each K_T component. The calibration method for the vibration influence coefficient vector is as follows: Under isothermal conditions, a vibrator is used to apply single-frequency excitation of different amplitudes to the base, and the change in the graduation error at the corresponding vibration amplitude is measured. The values ​​of each K_v component are obtained through fitting. The above method of calibrating the influence coefficients through controlled variable experiments is a conventional experimental method in this field. The coefficients obtained by calibration in this embodiment are: K_T=[0.32,0.18,0.51,1.25,0.76] (units are μm / ℃ and arcsecond / ℃, respectively), K_v=[0.09,0.07,0.11,0.05,0.04] (units are μm / (μm) and arcsecond / (μm), respectively).

[0034] Substituting the values ​​into the equation: E_env=[0.32×0.8,0.18×0.8,0.51×0.8,1.25×0.8,0.76×0.8]+[0.09×0.18,0.07×0.18,0.11×0.18,0.05×0.18,0.04×0.18]=[0.256,0.144,0.408,1.000,0.608]+[0.0162,0.0126,0.0198,0.009,0.0072]=[0.272,0.157,0.428,1.009,0.615].

[0035] Add each component of the environmental correction vector directly to the corresponding component of the base error vector: E_pre=E_base+E_env=[3.2+0.272,-1.8+0.157,2.5+0.428,4.1+1.009,-2.3+0.615]=[3.472μm,-1.643μm,2.928μm,5.109 arcseconds,-1.685 arcseconds].

[0036] The pre-bias error vector E_pre is superimposed onto the theoretical target pose to generate the pre-bias target pose, which is then sent to the cam turntable servo driver via the EtherCAT bus for execution. The cam turntable rotates to its position according to the compensated indexing angle, completing the feedforward pre-compensation.

[0037] After the transfer mechanism completes its operation, it enters the feedback loop of the double-loop compensation step. The contact-type in-situ probe at the target station (milling station 2) extends via a pneumatic mechanism to measure the actual pose of the workpiece on the unified zero-point fixture. The measurement process is as follows: the probe touches the reference surface in the X, Y, and Z directions of the workpiece, making three touches in each direction and taking the average value to obtain the actual position of the workpiece coordinate system origin in the zero-point fixture coordinate system; simultaneously, it touches the two locating pin holes on the workpiece to calculate the actual rotation angles around the X and Y axes. The above-mentioned touch and pose calculation methods are common knowledge in this field. The measurement results are compared with the theoretical pose, and the residual deviation vector E_res = [1.25μm, -0.83μm, 0.95μm, 1.82 arcseconds, -0.66 arcseconds] is calculated.

[0038] Calculate the norm of the residual deviation vector: ||E_res||=√(1.25²+0.83²+0.95²+(1.82×0.485)²+(0.66×0.485)²)=√(1.5625+0.6889+0.9025+0.7792+0.1025)=√4.0356≈2.009.

[0039] The threshold for the cam-rotor configuration is set at 0.5 μm. 2.009 > 0.5, indicating an excessive residual deviation, requiring a second alignment.

[0040] The system drives the piezoelectric ceramic fine-tuning shafts on both sides of the milling station to perform the first and second alignment. The fine-tuning shafts compensate for the displacement by reversing the residual deviation vector E_res. After compensation, the in-place probe measures again, obtaining a new residual deviation vector E_res_1 = [0.18μm, -0.09μm, 0.12μm, 0.25 arcseconds, -0.08 arcseconds]. The norm is calculated as: ||E_res_1|| = √(0.0324 + 0.0081 + 0.0144 + (0.25 × 0.485)² + (0.08 × 0.485)²) = √(0.0324 + 0.0081 + 0.0144 + 0.0147 + 0.0015) = √0.0711 ≈ 0.267. Since 0.267 < 0.5, the residual deviation is within the threshold range, and the second alignment is complete.

[0041] The final residual deviation vector E_res_1 = [0.18μm, -0.09μm, 0.12μm, 0.25 arcseconds, -0.08 arcseconds] is written back to the milling station indexing error database. The update method is as follows: Locate the record corresponding to the current index condition [station 2 indexing angle, cam rotary type, 22℃~24℃, half load], and update the stored error vector and the new residual vector using a weighted average. The weighting coefficients are 0.2 (new value weight) and 0.8 (old value weight) to improve database stability and avoid excessive influence from single outliers. Before updating, dead-zone filtering is performed on each component of the residual deviation: components with absolute values ​​lower than the in-situ probe measurement uncertainty (0.1μm / 0.1 arcseconds) are not updated; the original record value remains unchanged in this direction to avoid random fluctuations below the probe noise level contaminating the database.

[0042] Simultaneously update the timestamp and sample count of the record. The updated error vector is: [3.2×0.8+0.18×0.2,-1.8×0.8+(-0.09)×0.2,2.5×0.8+0.12×0.2,4.1×0.8+0.25×0.2,-2.3×0.8+(-0.08)×0.2]=[2.596,-1.458,2.024,3.330,-1.856]. The feedback loop is complete.

[0043] The system continuously monitors the health of the database. When the difference between the residual deviation norm of a record after a feedback loop write-back and the error vector norm stored in the database for that record exceeds the range stored for that record, the system determines that the record has deviated from its original calibration state. For example, if the residual deviation norm of a write-back is 3.5, and the error vector norm of the record in the database is 2.1, the difference is 1.4. However, the range stored for that record is 1.2, exceeding the range range. The system then triggers a recalibration reminder for that calibration position and automatically executes the recalibration process during the next production downtime. The recalibration operation is the same as the database creation calibration steps and will not be repeated here.

[0044] The following is a numerical example of the complete indexing process for a collaborative robot configuration. The only difference between the triangular and H-type collaborative robots lies in the joint layout topology (the triangular configuration consists of three sets of parallel joints symmetrically distributed around the workpiece at 120°) and the corresponding DH parameter table. The indexing error calibration method, trilinear interpolation process, and feedforward-feedback dual closed-loop compensation steps are completely consistent with the H-type configuration and will not be described separately.

[0045] The above describes the complete indexing compensation process for the cam-rotor configuration. The following example, using a complete indexing process of an H-type collaborative robot transferring a workpiece from milling station 2 to polishing station 1, provides a numerical example of the collaborative robot configuration. Here, the configuration recognition step reads the hardware interface identifier 0x02B, and the system loads the DH parameters shown in Table 1 and the corresponding forward / inverse kinematics calculation modules.

[0046] Unlike the cam-rotor configuration, the indexing error database of the collaborative robot configuration uses "end-effector pose mesh nodes" instead of "indexing position" as the indexing dimension. The mesh is divided into Cartesian spaces with equal intervals of 100mm, covering the robot's reachable workspace (in this embodiment, the workspace of the H-type collaborative robot is approximately 1200mm × 800mm × 600mm). In this example, the coordinates of the target end-effector pose corresponding to polishing station 1 in the global coordinate system are P_target = (550, 320, 180) (unit: mm), and this target point falls within the cubic region enclosed by the following 8 adjacent mesh nodes: P000=(500,300,100), corresponding to error vector E000=[18.2,-12.5,8.3,22.1,-15.0] (μm / arcsecond) P001=(500,300,200), corresponding to the error vector E001=[17.8,-11.9,9.1,21.5,-14.2]. P010 = (500, 400, 100), corresponding to the error vector E010 = [19.5, -13.8, 7.6, 23.0, -16.1]. P011=(500,400,200), corresponding to the error vector E011=[19.0,-13.2,8.5,22.4,-15.3]. P100 = (600, 300, 100), corresponding to the error vector E100 = [16.5, -10.2, 7.1, 20.8, -13.5]. P101=(600,300,200), corresponding to the error vector E101=[16.1,-9.6,7.9,20.2,-12.7] P110 = (600, 400, 100), corresponding to the error vector E110 = [17.8, -11.5, 6.4, 21.7, -14.6]. P111=(600,400,200), corresponding to the error vector E111=[17.3,-10.9,7.3,21.1,-13.8] The above 8 records are all reference error vectors obtained by calibration at the reference temperature of 20℃ in the polishing process. The temperature range is "18℃-20℃" and the load setting is "half load" (workpiece mass 3.2kg).

[0047] Calculate the normalized coordinates: x'=(550-500) / (600-500)=0.5, y'=(320-300) / (400-300)=0.2, z'=(180-100) / (200-100)=0.8.

[0048] Taking the X-axis linear deviation component (E_x) as an example, the trilinear interpolation calculation process is as follows: Interpolation along the Z-axis (8 nodes to the midpoints of 4 edges): E00_z=E000_x·(1-z')+E001_x·z'=18.2×0.2+17.8×0.8=3.64+14.24=17.88 E01_z=E010_x·(1-z')+E011_x·z'=19.5×0.2+19.0×0.8=3.90+15.20=19.10 E10_z=E100_x·(1-z')+E101_x·z'=16.5×0.2+16.1×0.8=3.30+12.88=16.18 E11_z=E110_x·(1-z')+E111_x·z'=17.8×0.2+17.3×0.8=3.56+13.84=17.40 Interpolation along the Y-axis (from 4 edges to 2 in-plane points): E0_yz=E00_z·(1-y')+E01_z·y'=17.88×0.8+19.10×0.2=14.304+3.820=18.124 E1_yz=E10_z·(1-y')+E11_z·y'=16.18×0.8+17.40×0.2=12.944+3.480=16.424 Interpolation along the X-axis (from two in-plane points to the target point): E_base_x=E0_yz·(1-x')+E1_yz·x'=18.124×0.5+16.424×0.5=9.062+8.212=17.274 Similarly, perform the above trilinear interpolation operation on the remaining four components (the values ​​of each component at the 8 nodes are shown in the table above; the calculation process is omitted) to obtain the complete basic error vector: E_base_robot=[17.274μm,-11.130μm,8.216μm,21.150 arcseconds,-13.830 arcseconds].

[0049] When the spatial distribution of errors exhibits significant non-uniformity or sparse grid nodes, Kriging interpolation can be used instead of trilinear interpolation. The key points of Kriging interpolation are as follows: Select all calibrated grid nodes within a spherical neighborhood with a radius of 300 mm centered on the target pose as known sample points (in this embodiment, the 8-node grid spacing is 100 mm, and at least 8 nodes can be included within the spherical neighborhood); fit the spatial correlation of each error component using the Gaussian variogram model γ(h) = C0 + C·[1-exp(-h² / a²)] (where h is the node spacing, and C0, C, and a are model parameters determined through cross-validation; in this embodiment, C0 = 0.5, C = 12, and a = 200 mm); perform ordinary Kriging estimation independently on each of the five error components to obtain the optimal linear unbiased prediction value of each component and the corresponding Kriging variance as a quantitative indicator of interpolation uncertainty. The specific mathematical derivation and implementation of Kriging interpolation are common knowledge in geostatistics.

[0050] The target station is the polishing station, ΔT = 19.6℃ - 20℃ = -0.4℃ (the current measured temperature of the polishing area is lower than the reference temperature). The effective amplitude of the base vibration A_v = 0.18μm. The temperature influence coefficient and vibration influence coefficient corresponding to the polishing process are independently calibrated (the calibration method is the same as that of the milling process, except that it is performed at the reference temperature of 20℃ in the polishing area). In this embodiment, K_T_polish = [0.29, 0.15, 0.48, 1.18, 0.71] and K_v_polish = [0.08, 0.06, 0.10, 0.04, 0.03].

[0051] Calculate the environment correction vector: E_env_robot=K_T_polish·ΔT+K_v_polish·A_v =[0.29×(-0.4),0.15×(-0.4),0.48×(-0.4),1.18×(-0.4),0.71×(-0.4)]+[0.08×0.18,0.06×0.18,0.10×0.18,0.04×0.18,0.03×0.18] =[-0.116,-0.060,-0.192,-0.472,-0.284]+[0.0144,0.0108,0.0180,0.0072,0.0054] =[-0.1016,-0.0492,-0.1740,-0.4648,-0.2786] (Each component is rounded to four decimal places) Generate the pre-bias error vector: E_pre_robot=E_base_robot+E_env_robot=[17.274+(-0.1016),(-11.130)+(-0.0492),8.216+(-0.1740),21.150+(-0.4648),(-13.830)+(-0.2786)]=[17.172μm,-11.179μm,8.042μm,20.685 arcseconds,-14.109 arcseconds] (each component is rounded to three decimal places).

[0052] The E_pre_robot is superimposed onto the theoretical end-effector pose of polishing station 1 to generate the pre-biased target pose. This pre-biased target pose is a Cartesian space pose matrix, which needs to be calculated into six-joint target angles [θ1,θ2,θ3,θ4,θ5,θ6] via robot inverse kinematics and then sent to the collaborative robot controller for execution via the EtherCAT bus. The inverse kinematics calculation adopts a closed-form solution method (analytical solution preferred) or a numerical iterative method, which is common knowledge in robotics. The robot moves into position according to the compensated joint angles, completing the feedforward pre-compensation.

[0053] After the collaborative robot is in position, the in-situ probe of polishing station 1 extends to measure the actual pose of the workpiece on the zero-point fixture (the measurement method is the same as that of the milling station), and calculates the residual deviation vector: E_res_robot_0=[8.50μm,-5.20μm,6.80μm,12.30 arcseconds,-9.10 arcseconds].

[0054] Calculate the norm: ||E_res_robot_0||=√(8.50²+5.20²+6.80²+(12.30×0.485)²+(9.10×0.485)²)=√(72.25+27.04+46.24+35.59+19.48)=√200.60≈14.16.

[0055] The threshold for the collaborative robot configuration is set to 3 μm (see the threshold range of 0.001 mm to 0.005 mm in step S3). 14.16 > 3, so a second alignment is required.

[0056] The system drives the robot's end effector, a six-dimensional micro-motion platform, to perform the first and second alignment. The micro-motion platform calculates the displacement of each axis based on the reverse calculation of E_res_robot_0 and executes compensating motion. After compensation, the probe measures again. E_res_robot_1=[2.10μm,-1.50μm,1.80μm,3.50 arcseconds,-2.20 arcseconds].

[0057] Norm: ||E_res_robot_1||=√(4.41+2.25+3.24+(3.50×0.485)²+(2.20×0.485)²)=√(4.41+2.25+3.24+2.88+1.14)=√13.92≈3.73. 3.73>3, still exceeding the limit, perform a second round of alignment: E_res_robot_2=[0.52μm,-0.38μm,0.45μm,0.95 arcseconds,-0.61 arcseconds].

[0058] Norm: ||E_res_robot_2||=√(0.2704+0.1444+0.2025+(0.95×0.485)²+(0.61×0.485)²)=√(0.2704+0.1444+0.2025+0.2123+0.0876)=√0.9172≈0.958. 0.958<3, the second alignment is completed, and a total of 2 iterations are performed (within the upper limit of 3 specified in step S3).

[0059] Write E_res_robot_2 back to the corresponding grid node record in the polishing station indexing error database with a weighting factor of 0.2 (new value) / 0.8 (old value) to complete the data update of the feedback closed loop.

[0060] As can be seen from the two sets of numerical examples above: due to the cumulative error of the serial joints, the norm of the basic error vector of the collaborative robot (E_base_robot=[17.274,-11.130,8.216,21.150,-13.830], ||E_base_robot||=√(17.274²+11.130²+8.216²+(21.150×0.485)²+(13.830×0.485)²)=√(298.39+123.88+67.50+105.22+44.99)=√639 The absolute positioning accuracy of the two configurations is significantly greater than that of the cam turntable (E_base=[3.2,-1.8,2.5,4.1,-2.3], norm²=10.24+3.24+6.25+3.95+1.24=24.92, ||E_base||=√(3.2²+1.8²+2.5²+(4.1×0.485)²+(2.3×0.485)²)=√(10.24+3.24+6.25+3.95+1.24)=√24.92≈4.99), which reflects the essential difference in absolute positioning accuracy between the two configurations. However, through dual closed-loop compensation—feedforward pre-compensation offsetting systematic configuration and environmental errors, and feedback closed-loop eliminating residual deviations with high-resolution (1nm-level) secondary alignment of the micro-motion platform—the final positioning accuracy of the collaborative robot configuration (0.96μm in this example) can converge to the same order of magnitude as the cam-rotor configuration (0.27μm in this example), both achieving sub-micron to micron-level transport and positioning accuracy. Furthermore, the collaborative robot configuration's database uses trilinear interpolation (or Kriging interpolation) between nodes in the three-dimensional pose mesh, which, compared to the one-dimensional linear interpolation of the cam-rotor, can more precisely describe the spatial non-uniform distribution characteristics of errors within the robot's workspace.

[0061] The production line can simultaneously load multiple workpieces, enabling synchronous transfer of milled blanks and polished semi-finished products within the same production cycle. For example, workpiece A is the loading blank, and workpiece B is a semi-finished product awaiting polishing after milling. In one indexing action of the transfer mechanism, workpiece A is sent into the milling station, while workpiece B is simultaneously transferred from the milling station to the polishing station. In one indexing action, the system calls its respective error database for the milling station and the polishing station, independently calculating the basic error vector and environmental correction vector for both stations. For the cam-rotor configuration, since one indexing action serves two stations simultaneously and only one rotational degree of freedom can be superimposed with fine-tuning, the pre-offset values ​​of the two target stations are weighted and averaged before unified indexing: the weights are allocated according to the accuracy requirements of the two stations. The positioning accuracy of the milling finishing station (station 2) directly affects the final machining dimension, so a weight of 0.7 is assigned; the margin for the polishing station is relatively ample, so a weight of 0.3 is assigned. If the accuracy requirements of the two stations are the same, the weights are assigned equally. When the pre-bias vectors of two target workstations have opposite directions or their norm ratios exceed 3:1, the weighted average will significantly degrade the compensation effect. In this case, the system automatically switches to a step-by-step transfer mode (breaking down one indexing operation into two independent actions, with each workstation compensating independently). When the cycle time allows, the step-by-step mode is prioritized to ensure accuracy. This weighting strategy and step-by-step switching logic have been verified by the production line process. In synchronous transfer scenarios, the arrival accuracy of both workstations meets the process requirements of their respective processes. For collaborative robot configurations, segmented end-path planning is performed and executed sequentially, with each segment compensating independently without interference. After arrival, the in-situ probes of the two workstations measure, provide feedback, and update their respective error databases independently, without affecting each other.

[0062] In this embodiment, the calculation of the feedforward pre-compensation loop and the feedback closed loop of the dual closed-loop compensation is executed by the FPGA hardware real-time closed-loop controller. This FPGA controller is connected to the servo driver, sensor acquisition module, and I / O module via the EtherCAT protocol bus, and runs the state machine with a 1μs timer interrupt cycle to ensure that the end-to-end latency from temperature / vibration data acquisition, error vector calculation, pre-bias pose generation to servo command issuance does not exceed 50μs. The FPGA hardware architecture and the implementation of the EtherCAT communication protocol stack are existing technologies. The CT equalization module runs on the host industrial control computer and dynamically adjusts the compensation execution level by monitoring the difference between the actual cycle time and the target cycle time (cycle margin) of each station: when the cycle margin of a station is less than 3 seconds, the maximum number of iterations for secondary alignment is reduced from 3 to 2, the interpolation order is reduced from trilinear to linear interpolation, and the calculation of some environmental corrections is moved forward to the processing idle period of the previous station to be completed in advance; when the cycle margin recovers to more than 5 seconds, it automatically returns to the original level. The beat statistics and gear switching logic of the CT equalization module are conventional control methods in this field.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, equivalent substitutions or modifications can be made to the technical details of this embodiment, such as the temperature threshold, threshold accuracy, interpolation algorithm, fine-tuning mechanism type, and sensor selection, without departing from the principles and spirit of the present invention. These improvements and refinements should also be considered within the scope of protection of the present invention.

Claims

1. A dual-closed-loop compensation method for multi-configuration indexing and transfer, applied to a multi-station five-axis flexible parallel CNC island, wherein the CNC island includes multiple five-axis mainframes, a transfer mechanism for transferring workpieces between the five-axis mainframes, a unified zero-point fixture installed at each station, a zoned temperature control system for separately controlling the temperature of the cutting zone and the polishing zone, and a base vibration sensing unit, characterized in that, Includes the following steps: S1. Configuration Recognition: Read the hardware interface identifier of the transfer mechanism, automatically identify whether the configuration of the transfer mechanism is a cam turntable type or a collaborative robot type, and call up the kinematic model and indexing mapping relationship corresponding to the identified configuration; wherein, the indexing mapping relationship of the cam turntable type configuration is the equal division mapping of the indexing angle and the workstation number, and the indexing mapping relationship of the collaborative robot type configuration is the kinematic mapping of the target pose of each workstation and the robot joint quantity. S2. Error Database Call: Call the pre-established milling station indexing error database and polishing station indexing error database. The two databases are independent of each other. Each indexing error database stores the error vector with the indexing position, configuration, temperature range and load level as indexes. The temperature range indexes the range of workshop ambient temperature at the time of execution. The error vector is the reference error vector calibrated at the reference temperature of the corresponding process. The temperature range is used to calibrate data classification. During online compensation, the corresponding indexing error database is called according to the processing process category of the target station. The current temperature range and load level are used as filtering conditions. Interpolation is performed between adjacent indexing positions to calculate the basic error vector of the target indexing position. S3. Dual Closed-Loop Compensation: Before the transfer is executed, the cutting zone temperature, polishing zone temperature, and base vibration data are collected synchronously. An environmental correction vector determined by the temperature deviation and vibration is superimposed on the basic error vector to generate a pre-biased target pose and send it to the transfer mechanism for execution, forming a feedforward pre-compensation loop. After the transfer is completed, the in-situ detection unit set at the target station measures the actual pose of the workpiece on the unified zero-point fixture, calculates the residual deviation between the actual pose and the theoretical pose, and when the residual deviation is greater than a set threshold, drives the fine-tuning shaft on the target station side or the end fine-motion shaft of the transfer mechanism for secondary alignment until the residual deviation is not greater than the set threshold. The residual deviation is then written back to the corresponding indexing error database to update the record, forming a feedback closed loop.

2. The dual closed-loop compensation method for multi-configuration indexing transport according to claim 1, characterized in that: In step S1, the hardware interface identifier includes the type of the transfer mechanism drive bus, the number of servo axes, and the kinematic parameter package; when it is identified as a collaborative robot configuration, its topology is further identified as a triangular collaborative robot or an H-type collaborative robot, and the corresponding DH parameters are loaded.

3. The dual closed-loop compensation method for multi-configuration indexing transport according to claim 1, characterized in that: In step S2, the error vector includes three linear axis error components and two rotation axis error components; each record also includes the number of statistical samples and the range; the interpolation uses linear interpolation between adjacent indexing positions, or trilinear interpolation or Kriging interpolation between adjacent pose mesh nodes in a collaborative robot configuration.

4. The dual closed-loop compensation method for multi-configuration indexing transport according to claim 3, characterized in that: The pre-establishment of the database in step S2 includes the database establishment and calibration steps: controlling the transfer mechanism to repeatedly position each graduation position no less than 7 times, the measurement unit collects the actual pose each time, takes the mean value and writes it into the corresponding record, and the range is used as the dispersion index. When the difference between the norm of the residual deviation of the feedback closed-loop write-back, which is normalized with the zero-point tooling clamping radius as the characteristic length, and the norm of the error vector of the corresponding record in the database, which is normalized in the same way, exceeds the range, the recalibration of the corresponding record is triggered.

5. The dual closed-loop compensation method for multi-configuration indexing transfer according to claim 1, characterized in that: In step S3, the environmental correction vector is determined according to E_env = K_T·ΔT + K_v·A_v, where K_T and K_v are the temperature influence coefficient vector and vibration influence coefficient vector determined by calibration tests, respectively; when the target station is a milling station, ΔT is the deviation between the measured temperature of the cutting zone and its set reference temperature of 22℃; when the target station is a polishing station, ΔT is the deviation between the measured temperature of the polishing zone and its set reference temperature of 20℃; A_v is the effective amplitude of the base vibration; the superposition is to add each component of the environmental correction vector to the corresponding component of the basic error vector.

6. The dual closed-loop compensation method for multi-configuration indexing transfer according to claim 1, characterized in that: In step S3, the norm of the residual deviation is defined as: the Euclidean norm calculated together with the linear deviation component after converting the rotational deviation component into an equivalent linear deviation using the clamping radius of the zero-point tooling as the characteristic length; the set threshold is 0.0002mm-0.001mm for the cam turntable configuration and 0.001mm-0.005mm for the collaborative robot configuration; the number of secondary alignment iterations does not exceed 3.

7. The dual closed-loop compensation method for multi-configuration indexing transfer according to claim 1, characterized in that: It also includes a zero-point reference calculation step: the zero-point offset vector of the unified zero-point fixture is incorporated as a fixed parameter into the coordinate transformation chain of each station, so that the error vector calculation in step S2 and the residual deviation calculation in step S3 are both based on the zero-point offset vector as a unified reference; the zero-point offset vector is obtained through calibration and includes the fixture installation error; when the workpiece or fixture is changed, the zero-point offset vector remains unchanged, and only the transformation parameters of the workpiece coordinate system relative to the zero-point offset vector are updated.

8. The dual closed-loop compensation method for multi-configuration indexing transfer according to claim 1, characterized in that: The CNC island also includes a CT equalization module, which synchronously adjusts the compensation execution level in step S3 according to the cycle margin of each station. When the cycle margin of the target station is lower than the set value, the compensation execution level is reduced. The adjustment elements of the execution level include the number of secondary alignment iterations, the interpolation order, and the timing of compensation execution. When the level is reduced, the calculation of some environmental corrections is moved forward to the idle period of the previous station. After the cycle margin is restored, the level is raised again.

9. The dual-closed-loop compensation method for multi-configuration indexing transfer according to any one of claims 1 to 8, characterized in that: The calculation and output of the feedforward pre-compensation loop and the feedback closed loop are executed by a high-speed hardware real-time closed-loop controller; the milled blank and the polished semi-finished product are synchronously transferred within the same production cycle.

10. A dual-closed-loop compensation system for multi-configuration indexing and transfer, characterized in that, include: The configuration recognition module is used to read the hardware interface identifier of the transfer mechanism and automatically identify its configuration; The milling station indexing error database and the polishing station indexing error database are independent of each other and store error vectors with indexing position, configuration, temperature range and load level as indexes. The error vector is a reference error vector calibrated at the reference temperature of the corresponding process. The feedforward pre-compensation module is used to obtain the basic error vector from the corresponding error database, and to superimpose the environmental correction vector on the basic error vector according to the partition temperature and base vibration data to generate the pre-biased target pose. The secondary alignment feedback module includes an in-situ detection unit and a fine-tuning shaft drive unit set at the target workstation. It is used to calculate the residual deviation and perform secondary alignment, and write the residual deviation back to the corresponding indexing error database to update the record. The zero-point reference calculation module is used to incorporate the zero-point offset vector of the unified zero-point tooling into the coordinate transformation chain of each workstation. And a CT equalization adjustment module, used to synchronously adjust the compensation execution level according to the cycle margin of each workstation.

Images