A robot joint stiffness-damping integrated dynamics parameter calibration method

Abstract

This invention relates to the field of robot joint dynamic parameter testing and calibration technology, specifically a method for calibrating robot joint stiffness and damping integrated dynamic parameters. The method includes: mounting the robot joint module under test on a floating adaptive mounting support; adaptively adjusting the mounting posture and coaxially calibrating based on feedback from a six-dimensional force sensor; controlling a servo loading component to apply a static load according to a preset torque gradient, constructing a torque-angular displacement relationship curve and fitting it to obtain stiffness parameters; controlling a transient excitation component to apply a controllable impact, collecting vibration decay time-series data, and extracting damping parameters using a combination of the logarithmic decay rate method and Hilbert transform; combining temperature data, coaxiality deviation data, and sensor zero-drift data for comprehensive error compensation, outputting the compensated stiffness matrix and damping matrix, and generating a test report. This invention improves the accuracy, stability, and traceability of stiffness and damping parameter calibration.

Description

Technical Field

[0001] This invention relates to the field of robot joint dynamics parameter testing and calibration technology, specifically a method for calibrating robot joint stiffness and damping integrated dynamic parameters. Background Technology

[0002] As a key actuator in industrial robots, collaborative robots, and special equipment, the stiffness and damping of robot joint modules directly affect motion accuracy, vibration suppression capability, and overall control stability. To ensure parameter consistency of robot joint modules during assembly, control, and factory testing, it is usually necessary to calibrate the stiffness and damping of the joint modules and obtain calibration results that reflect their mechanical response characteristics to meet the application requirements of subsequent control compensation, performance evaluation, and quality traceability.

[0003] However, when calibrating the dynamic parameters of robot joint modules, stiffness and damping tests are often conducted separately. The testing process is easily affected by factors such as installation eccentricity, coaxiality deviation, temperature drift, and sensor zero drift, resulting in error coupling between output torque data, angular displacement data, and vibration attenuation data. Especially in test scenarios where static loading and transient excitation coexist, if stiffness or damping parameters are directly calculated based on a single test data, or if simple methods are used to handle inconsistent sampling timing, missing data, and missing compensation terms, the calibration results are likely to deviate from the actual working conditions, thereby affecting the accuracy and stability of the joint module dynamic parameter calibration. Summary of the Invention

[0004] The purpose of this invention is to provide a method for calibrating the dynamic parameters of robot joint stiffness and damping in an integrated manner. This method avoids the problems of error coupling and deviation from actual working conditions caused by factors such as installation eccentricity, coaxiality deviation, temperature drift, sensor zero drift, and inconsistent sampling timing in the separate stiffness and damping test. It also achieves more accurate and stable comprehensive error compensation, thereby obtaining high-precision stiffness matrix and damping matrix calibration results.

[0005] The objective of this invention can be achieved through the following technical solution: a method for calibrating the integrated dynamic parameters of robot joint stiffness and damping, based on a test platform. The platform includes a floating adaptive mounting bracket, a six-dimensional force sensor, a precision encoder assembly, a servo loading component, a transient excitation component, and a temperature sensor. The method includes: acquiring basic information of the robot joint module under test and installing it on the floating adaptive mounting bracket; based on feedback data from the six-dimensional force sensor, adaptively adjusting the installation posture and performing coaxial calibration on the robot joint module under test; controlling the servo loading component to apply a static load to the robot joint module under test according to a preset torque gradient, and synchronously acquiring output torque data and angular displacement data using the six-dimensional force sensor and the precision encoder assembly; based on the input... Torque and angular displacement data are used to construct a torque-angular displacement relationship curve, which is then fitted to obtain the stiffness parameters of the robot joint module under test. A transient excitation component is controlled to apply impact to the robot joint module under load, and vibration decay time-series data is collected using a precision encoder group. Based on the vibration decay time-series data, the damping parameters of the robot joint module are extracted using a combination of the logarithmic decay rate method and the Hilbert transform. Temperature data is obtained through a temperature sensor, and combined with installation coaxiality deviation data and sensor zero drift data, comprehensive error compensation is performed on the stiffness and damping parameters. The stiffness and damping parameters are then constructed into stiffness and damping matrices, respectively, and the compensated stiffness and damping matrices are output as calibration results to generate a test report.

[0006] Preferably, based on the feedback data from the six-dimensional force sensor, the installation posture of the robot joint module under test is adaptively adjusted and coaxially calibrated, including: identifying the installation eccentricity state according to the force information output by the six-dimensional force sensor; performing multi-degree-of-freedom fine-tuning of the installation position of the robot joint module under test through the floating adaptive installation support to perform posture compensation; and locking the installation state after adjusting the output axis of the robot joint module under test and the loading axis of the servo loading component to a preset coaxial range.

[0007] Preferably, controlling the servo loading component to apply a static load to the joint module of the robot under test according to a preset torque gradient includes: controlling the joint module of the robot under test to maintain a horizontal posture; driving its loading axis to apply a static load step by step according to a preset torque gradient through the servo loading component, wherein a preset torque step is used between adjacent load levels.

[0008] Preferably, obtaining the stiffness parameters of the joint module of the robot under test includes: performing least-squares fitting on the torque-angular displacement relationship curve to obtain the fitting function, and calculating the torsional stiffness parameter based on the slope of the fitting function; acquiring radial off-center load or bending load data collected by a six-dimensional force sensor, and calculating the bending stiffness parameter based on the torque change and angular displacement change in the loaded direction; and performing temperature drift correction on the torsional stiffness parameter and bending stiffness parameter to obtain the corrected stiffness parameter.

[0009] Preferably, the platform further includes a replaceable load linkage assembly. Before applying the impact, the method further includes: adjusting the joint module of the robot under test, equipped with the replaceable load linkage assembly, to a vertically loaded state; extracting the damping parameters of the joint module of the robot under test using the logarithmic decay rate method combined with Hilbert transform, including: extracting adjacent vibration peaks based on vibration decay time series data and calculating the logarithmic decay rate; performing a Hilbert transform on the vibration decay time series data to obtain instantaneous envelope information; calculating the equivalent damping coefficient based on the logarithmic decay rate and instantaneous envelope information, and correcting the equivalent damping coefficient by combining the installation coaxiality deviation data to obtain the damping parameters.

[0010] Preferably, comprehensive error compensation is performed on the stiffness and damping parameters, including: establishing a temperature drift compensation model; establishing an installation coaxiality error compensation model; establishing a sensor zero drift compensation model; and fusing the parameters corrected by the temperature drift compensation model, the installation coaxiality error compensation model, and the sensor zero drift compensation model to obtain the compensated stiffness matrix and damping matrix.

[0011] Preferably, the compensated stiffness matrix and damping matrix are expressed as follows: the compensated stiffness matrix is ​​equal to the initial stiffness matrix superimposed with temperature drift compensation, installation coaxiality error compensation, and sensor zero drift compensation; the compensated damping matrix is ​​equal to the initial damping matrix superimposed with temperature drift compensation, installation coaxiality error compensation, and sensor zero drift compensation; wherein, the compensated stiffness matrix and damping matrix correspond to the final calibration result after comprehensive compensation.

[0012] Preferably, the output of the compensated stiffness matrix and damping matrix is ​​used as the calibration result and a test report is generated, including: binding the compensated stiffness matrix and damping matrix with the basic information of the joint module of the robot under test; generating a test report containing test condition information, output torque data, angular displacement data, vibration decay time series data, error compensation results and calibration conclusions.

[0013] The beneficial effects of this invention are as follows: By integrating stiffness testing and damping testing onto the same testing platform, and combining a floating adaptive mounting bracket, a six-dimensional force sensor, a precision encoder assembly, a servo loading component, and a transient excitation component, this invention achieves integrated calibration of joint module installation posture adaptive adjustment, coaxial calibration, static loading testing, and impact attenuation testing. This effectively reduces the impact of installation eccentricity, coaxiality deviation, temperature drift, and sensor zero drift on the test results. At the same time, error compensation and matrix output improve the accuracy, stability, and traceability of stiffness and damping parameter calibration. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0015] Figure 1 This is a flowchart of the robot joint stiffness and damping integrated dynamic parameter calibration method according to an embodiment. Detailed Implementation

[0016] The technical solutions of 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.

[0017] Please see Figure 1 A method for calibrating the dynamic parameters of robot joint stiffness and damping is disclosed. This method relies on a test platform, which includes a floating adaptive mounting bracket, a six-dimensional force sensor, a precision encoder assembly, a servo loading component, a transient excitation component, and a temperature sensor. The method includes: acquiring basic information of the robot joint module under test and mounting it on the floating adaptive mounting bracket; based on feedback data from the six-dimensional force sensor, adaptively adjusting the installation posture and performing coaxial calibration on the robot joint module under test; controlling the servo loading component to apply a static load to the robot joint module under test according to a preset torque gradient, and simultaneously acquiring output torque data and angular displacement data using the six-dimensional force sensor and the precision encoder assembly; and based on the output torque data and angular displacement data... The data is used to construct a torque-angular displacement relationship curve, which is then fitted to obtain the stiffness parameters of the robot joint module under test. A transient excitation component is controlled to apply impact to the robot joint module under load, and vibration decay time-series data is collected using a precision encoder group. Based on the vibration decay time-series data, the damping parameters of the robot joint module are extracted using a combination of the logarithmic decay rate method and the Hilbert transform. Temperature data is obtained through a temperature sensor, and combined with installation coaxiality deviation data and sensor zero drift data, comprehensive error compensation is performed on the stiffness and damping parameters. The stiffness and damping parameters are then constructed into stiffness and damping matrices, respectively, and the compensated stiffness and damping matrices are output as calibration results to generate a test report.

[0018] Based on feedback data from a six-dimensional force sensor, the installation posture of the robot joint module under test is adaptively adjusted and coaxially calibrated. This includes: identifying the installation eccentricity state based on the force information output by the six-dimensional force sensor; performing multi-degree-of-freedom fine-tuning of the installation position of the robot joint module under test using a floating adaptive mounting bracket to perform posture compensation; adjusting the output axis of the robot joint module under test and the loading axis of the servo loading component to a preset coaxial range and then locking the installation state; controlling the servo loading component to apply a static load to the robot joint module under test according to a preset torque gradient, including: controlling the robot joint module under test to maintain a horizontal posture; driving its loading axis to apply a static load stepwise according to a preset torque gradient through the servo loading component, wherein a preset torque step is used between adjacent load levels.

[0019] After the joint module of the robot under test is installed on the test platform, the system first reads the basic information of the joint module to match the corresponding loading range and sampling configuration. This basic information is not directly written once and then not used again after entering the test process. Instead, it is first broken down by the control unit into four types of control variables: loading limit, target sampling frequency, attitude constraint, and safe stopping boundary. Among them, the loading limit and safe stopping boundary remain in a read-only state throughout the entire test cycle, while the sampling frequency and attitude constraint are allowed to switch between the stiffness test phase and the damping test phase. Since the joint module may have positional offset and axis deviation after assembly, the test platform relies on a floating adaptive mounting support to support the module, and the force status is continuously provided by a six-dimensional force sensor. The system monitors the radial force and bending moment output by the six-dimensional force sensor in real time. If its absolute value is greater than or equal to the preset force threshold, it is identified as having an installation eccentricity state. This threshold can be determined based on three times the nominal accuracy of the zero position of the six-dimensional force sensor.

[0020] Based on this, the control unit performs corresponding compensation action mapping according to the polarity of the non-torsional forces such as X-axis radial force, Y-axis radial force, bending moment about the X-axis, and bending moment about the Y-axis. The polarity of the non-torsional forces is extracted through a sign function, and the X-axis radial force... Y-axis radial force The deviation polarity mapping is the position correction of the support in the X and Y translational degrees of freedom, according to the control law formula: ; ;in, , Here, sgn is the position correction value, and sgn is the sign function. The system presets a linear translation fine-tuning step size; it maps the deviation polarity of the bending moment around the X and Y axes to the correction direction of the support's tilt rotation degree of freedom around the corresponding axis.

[0021] The actuators in each direction of the drive support perform multi-degree-of-freedom fine-tuning step by step according to the preset linear translation or angular fine-tuning step, and perform attitude compensation step by step in the opposite direction of the above mapping until the absolute value of the force response in the above four non-torsional directions is less than the preset force threshold. That is, it is determined that the output axis and the loading axis have entered the preset coaxial range, and the installation state is locked to avoid new offsets during subsequent testing. After locking the installation state, it means that the installation attitude adaptive adjustment and coaxial calibration of the joint module of the robot under test have been completed, so that it is in a preset attitude that meets the requirements of subsequent testing.

[0022] During the support adjustment process, the system continuously collects the force results for no less than two sampling windows after each fine-tuning. If the sum of the absolute values ​​of the four non-torsional directions in the subsequent window does not decrease compared to the previous window, the fine-tuning is marked as invalid, and the step size of the corresponding degree of freedom is halved before trying again. If the same degree of freedom is invalid three times in a row, compensation for that degree of freedom is stopped, and only the remaining degrees of freedom are retained for adjustment, and a restricted mark is retained in the installation record. After entering the stiffness test stage, the system keeps the joint module in a horizontal position, and the servo loading component applies static load step by step according to the preset torque step. Output torque data and angular displacement data need to be collected for each load level. Due to the difference in sampling frequency and communication lag between the six-dimensional force sensor and the precision encoder, the system adds absolute timestamps to the two sensor signals based on hardware synchronous triggers.

[0023] Using the data axis corresponding to the benchmark sampling frequency as the benchmark time axis, linear interpolation or nearest neighbor matching resampling is performed on the target high sampling rate data at the timestamp position to eliminate the timing misalignment caused by communication delay and physical hysteresis; the two types of data that have been strictly time aligned are written into the same test cycle record for subsequent fitting calls.

[0024] The system sets four logical states for the acquisition link: pending alignment, aligned, interpolated, and eliminated. Within the same load level, when the timestamp difference between two data streams does not exceed the preset alignment tolerance, it directly switches to the aligned state.

[0025] When the difference exceeds the alignment tolerance but is still within a single sampling step, it enters the interpolation state; when the difference exceeds a single sampling step, or when any data stream is interrupted for more than two sampling cycles, the batch of samples enters the rejection state and is no longer included in the current load level average calculation.

[0026] Once the target torque is reached, the servo loading component enters a holding state. Only when the torque fluctuation and angular displacement fluctuation are both lower than or equal to their respective stability thresholds within multiple consecutive sampling windows is the load level recorded as stable and allowed to be written into the fitted sequence.

[0027] If the system remains unstable for more than the preset waiting time within the holding state, it will automatically revert to a preset torque step and reapproach until the stability condition is met or the abandonment flag for that level is triggered. The relationship curve formed between torque and angular displacement is used to calculate stiffness parameters. The fitting process uses conventional least squares processing, and the slope of the curve is converted into torsional stiffness.

[0028] When the six-dimensional force sensor records radial off-center load or bending load, the bending stiffness is obtained by combining the change in torque and the change in angular displacement in the corresponding direction.

[0029] For the temperature rise that occurs during the test, the system records the temperature through the temperature sensor and performs drift correction after parameter calculation to ensure that the stiffness result is consistent with the working conditions during the test.

[0030] After entering the damping test phase, the joint module is subjected to a controllable impact applied by the transient excitation component under load, and the precision encoder group continuously collects vibration decay time sequence data.

[0031] The system first extracts adjacent vibration peaks from the time-series data, and then calculates the logarithmic decay rate by calculating the natural logarithm of the ratio of the amplitudes of adjacent peaks. Convert it into a preliminary damping ratio And through a nonlinear transformation formula: Then, perform a Hilbert transform on the same set of data to obtain instantaneous envelope information and derive the envelope damping ratio based on this instantaneous envelope information. Finally, calculate the equivalent damping coefficient by combining the initial damping ratio and the envelope damping ratio.

[0032] If there is a record of installation coaxiality deviation during the test, the damping calculation result will be further corrected in conjunction with the deviation. For sensor zero drift that occurs in the test link, the system will use the zero drift record as a compensation input, and together with the temperature data and coaxiality deviation data, it will be used to correct the stiffness parameters and damping parameters to form a compensated stiffness matrix and damping matrix, and then write them into the test report after binding them with the basic information.

[0033] When temperature data, encoder data, or six-dimensional force sensor data cannot be fully acquired within a certain sampling period, the system does not directly interrupt the current test process. Since directly copying the record of the previous period will cause a smooth step in the angular displacement and torque curves, resulting in an error in the stiffness fitting slope, the system adopts a differentiated compensation strategy: for low-frequency environmental variables, such as temperature data, the stable record of the previous valid period is directly carried over.

[0034] For dynamically loaded parameters, such as encoder data and six-dimensional force sensor data, a control hysteresis of a single sampling period is introduced based on the data cache stack. Linear interpolation is performed based on the effective data nodes adjacent to the missing point to supplement the parameter calculation basis for that delay period. An interpolation supplementation mark is set for the data of that period so that the data source can be identified in subsequent reports.

[0035] If three or more consecutive sampling periods are missing within the same load level, the system will no longer continue interpolation and expansion. Instead, it will terminate the data acquisition for that load level and mark the entire level as invalid. Different degradation paths will be executed according to the current test stage: stiffness test will jump to the next load level, and damping test will retain the acquired attenuation segment and wait for the next excitation to be retested. This can avoid the calculation distortion or failure to generate the entire calibration result due to the loss of single-point data.

[0036] In this embodiment, the stiffness parameters of the joint module of the robot under test are obtained by: performing least squares fitting on the torque-angular displacement relationship curve to obtain the fitting function, and calculating the torsional stiffness parameter based on the slope of the fitting function; acquiring radial off-center load or bending load data collected by a six-dimensional force sensor, and calculating the bending stiffness parameter based on the torque change and angular displacement change in the load direction; and performing temperature drift correction on the torsional stiffness parameter and bending stiffness parameter to obtain the corrected stiffness parameter.

[0037] The platform also includes a replaceable load linkage assembly. Before performing the damping test, the method further includes: adjusting the robot joint module under test, equipped with the replaceable load linkage assembly, to a vertically loaded state; extracting the damping parameters of the robot joint module under test using the logarithmic decay rate method combined with Hilbert transform, including: extracting adjacent vibration peaks based on vibration decay time series data and calculating the logarithmic decay rate; performing a Hilbert transform on the vibration decay time series data to obtain instantaneous envelope information; calculating the equivalent damping coefficient based on the logarithmic decay rate and instantaneous envelope information, and correcting the equivalent damping coefficient by combining the installation coaxiality deviation data to obtain the damping parameters.

[0038] When the joint module is subjected to radial off-center load or bending load, the system synchronously reads the radial off-center load and bending load data output by the six-dimensional force sensor, and calculates the bending stiffness by dividing the change in torque in the load direction by the corresponding change in angular displacement. After the torsional stiffness and bending stiffness are obtained by fitting, temperature drift correction is uniformly performed to form the corrected stiffness parameters. After this processing, the result does not depend on a single working condition, but corresponds to the form of loading.

[0039] Before the damping test, the platform is configured with a replaceable load linkage group and the loaded module is adjusted to a vertical state to ensure that the test conditions are consistent with the load relationship related to the direction of gravity. Before executing transient excitation, the system sets a pre-check sequence: first, verify whether the vertical attitude is within the allowable deviation range, then verify whether the locking state of the load linkage group and the loaded mass match the current joint basic information, and confirm that the encoder zero position has been refreshed and the transient excitation component is in the ready-to-fire state.

[0040] Impact triggering is only permitted if all the above checks pass; otherwise, the system remains in the ready state and damping sampling is prohibited. After transient excitation, the vibration decay time series data collected by the encoder is first split into adjacent peak sequences to calculate the logarithmic decay rate to obtain the system damping ratio. The same set of time series data is then subjected to Hilbert transform to obtain instantaneous envelope information.

[0041] The damping ratio calculation process specifically includes: subtracting the preliminary damping ratio obtained by the logarithmic decay rate method from the envelope damping ratio derived based on instantaneous envelope information. The derivation of the envelope damping ratio can be performed according to the following data flow: the system extracts the time points corresponding to adjacent peak values ​​on the instantaneous envelope curve. and and the corresponding amplitude and Calculate the logarithmic ratio Divide by the time difference To output the decay time constant .

[0042] The decay time constant Divide by the preset undamped natural circular frequency in the system's basic information The envelope damping ratio is calculated; if the absolute value of the difference between the two is less than a preset threshold, the arithmetic mean of the two is taken as the overall system damping ratio. If the difference is greater than or equal to the preset threshold, the envelope damping ratio is directly used as the overall system damping ratio. .

[0043] Based on the dynamic mapping relationship, the system will integrate the damping ratio. The moment of inertia J and undamped natural circular frequency of the joint module obtained from the basic information. The product of these factors is then multiplied by a constant 2 to quantitatively calculate the equivalent damping coefficient. : The system first calculates the root mean square amplitude of the encoder angular displacement noise during the static period before excitation, and uses its preset multiple as the lower limit of peak recognition. The attenuation sequence is scanned in chronological order, and only the points that simultaneously satisfy the condition that the time interval between the local maximum and the previous peak is greater than or equal to the minimum vibration half-cycle, and whose peak amplitude is higher than or equal to the lower limit of recognition, are recorded as effective peaks.

[0044] If the number of effective peaks after the scan is insufficient to form two sets of adjacent peaks, the system will not proceed to the logarithmic attenuation rate calculation, but will directly switch to the single-branch damping estimation based on the instantaneous envelope, and mark the periodic damping result as obtained by degradation. If there is still an installation coaxiality deviation under vertical load, the system will extract the specific coaxiality deviation value, such as the radial eccentricity distance in millimeters.

[0045] The system multiplies the coaxiality deviation value by a preset eccentric coupling coefficient, which can be established through a pre-existing eccentric tension-compression calibration experiment. Its dimension is defined as the ratio of damping to length, i.e. (N⋅m⋅s / rad) / mm. The dimensionally consistent damping correction compensation amount is then calculated. This correction compensation amount is subtracted from the equivalent damping coefficient to eliminate parasitic eccentric friction interference, and finally the damping parameters are obtained.

[0046] If noise interference causes vibration peaks to be incompletely identified within a certain sampling period, the system extracts the instantaneous envelope curve obtained by Hilbert transforming the time series data. At the corresponding time node where the peak is missing, the amplitude of the envelope curve is directly read as the estimated peak value to complete the sequence and calculate the logarithmic decay rate normally, thereby obtaining the equivalent damping coefficient. If the coaxiality deviation data is not updated in this period, the most recent valid record is used for correction, and the corresponding mark is retained in the test report.

[0047] If the peak value after envelope calculation still does not satisfy the monotonic decay relationship with the two effective peak values ​​before and after it, the system cancels the supplemented peak value and terminates the calculation of the current excitation segment, and then re-executes the transient excitation to avoid introducing non-physical rebound peak values ​​during the supplementation process.

[0048] In this embodiment, comprehensive error compensation is performed on the stiffness and damping parameters, including: establishing a temperature drift compensation model; establishing an installation coaxiality error compensation model; establishing a sensor zero drift compensation model; and fusing the parameters corrected by the temperature drift compensation model, the installation coaxiality error compensation model, and the sensor zero drift compensation model to obtain the compensated stiffness matrix and damping matrix.

[0049] The compensated stiffness matrix can be expressed as: K= The 6×6 dimensional stiffness matrix obtained from the initial fitting is denoted as... The compensation amounts introduced by temperature drift, installation coaxiality error, and sensor zero drift on the stiffness matrix are respectively denoted as: , and The compensated damping matrix can be expressed as: C = The 6×6-dimensional damping matrix initially calculated is denoted as... The compensation amounts introduced by temperature drift, installation coaxiality error, and sensor zero drift on the damping matrix are denoted as follows: , and .

[0050] The compensated stiffness matrix and damping matrix are output as the calibration results of the joint module of the robot under test, and a test report is generated, including: binding the compensated stiffness matrix and damping matrix with the basic information of the joint module of the robot under test; generating a test report containing test condition information, output torque data, angular displacement data, vibration decay time series data, error compensation results and calibration conclusions.

[0051] After initial fitting, the stiffness and damping parameters are compensated for errors by combining the temperature rise, support eccentricity, and sensor baseline drift data obtained during the test. Since the torsional stiffness, bending stiffness, and equivalent damping coefficient obtained from the initial fitting are all one-dimensional scalar parameters, in order to avoid data structure dimensionality discontinuity in subsequent calculations, the system explicitly uses the Z-axis as the loading coordinate system for torsion output and executes the following matrix filling rule: the calculated torsional stiffness value is placed into the 6th element of the main diagonal of the 6×6 diagonal matrix, which corresponds to the position of the degree of freedom of rotation around the Z-axis.

[0052] The bending stiffness values ​​are placed into the 4th and 5th elements of the main diagonal, corresponding to the degrees of freedom of rotation about the X and Y axes, respectively; and the remaining elements of the matrix are set to zero, thus constructing an initial stiffness matrix with defined dimensions. and initial damping matrix .

[0053] The system establishes temperature drift compensation model, installation coaxiality error compensation model and sensor zero drift compensation model respectively, and maps various scalar compensation quantities into a 6×6 deviation matrix to act on the stiffness matrix and damping matrix.

[0054] 6×6 dimensional temperature sensitivity matrix Both the 6×2 order parasitic coupling stiffness coefficient matrix and the system's pre-defined values ​​were established by the system through pre-executed offline calibration experiments.

[0055] Specifically, in the compensation model calculation, the temperature drift compensation model uses the scalar temperature rise value obtained in real time. As an input, it is compared with the preset 6×6-dimensional temperature sensitivity coefficient matrix within the system. Perform scalar-matrix multiplication and output the compensation amount as a 6×6 deviation matrix. .

[0056] The coaxiality error compensation model takes the two-dimensional residual eccentricity vector containing the radial offset as input and performs matrix multiplication with the system's predefined 6×2 parasitic coupling stiffness coefficient matrix. The result is a 6-dimensional column vector containing the stiffness deviation values ​​for each degree of freedom. The elements of this column vector are then sequentially placed into positions 1 to 6 on the main diagonal of a new 6×6 matrix according to their corresponding degrees of freedom, constructing a full-dimensional diagonal compensation matrix. .

[0057] In cases where sensor zero drift directly affects the data reference position, since the original average output values ​​of the sensor, such as torque and angular displacement, have different dimensions from the stiffness and damping matrix, they cannot be directly added as deviations. Therefore, when calculating the sensor zero drift compensation, [further details are needed]. First, extract the baseline drift mean of the six-dimensional force sensor torque output sequence and the encoder angular displacement output sequence within a preset time period under no-load, static, and constant-temperature conditions.

[0058] Before performing least-squares fitting, the mean torque baseline drift is subtracted point by point from the output torque data sequence to complete the reference position correction, and the stiffness compensation is... The entire matrix is ​​always set to all zeros.

[0059] Compensation amount corresponding to the damping matrix , The compensation amount is derived by multiplying the sensitivity coefficients and the coupling matrix, following a similar logic to that used for stiffness. Then, by dividing the mean of the torque baseline drift by the maximum angular velocity change to obtain the equivalent damping deviation scalar, a 6×6 dimensional compensation matrix is ​​constructed by filling it in according to the same rules.

[0060] The dimensional conflicts caused by the superposition of different physical quantities are eliminated by constructing a compensation matrix; the compensation results are no longer saved as scattered parameters, but are uniformly formed into a matrix form, which is convenient for direct use in joint module control or factory records.

[0061] In the matrix representation, the stiffness matrix and damping matrix obtained from the initial fitting are used as base values. The compensation values ​​corresponding to temperature drift, coaxiality error and zero drift are respectively superimposed on the base values ​​to obtain the final calibration result.

[0062] The compensated stiffness matrix is ​​equal to the initial stiffness matrix superimposed with the temperature drift compensation, the installation coaxiality error compensation, and the sensor zero drift compensation. Similarly, the compensated damping matrix is ​​equal to the initial damping matrix superimposed with the temperature drift compensation, the installation coaxiality error compensation, and the sensor zero drift compensation.

[0063] The system sets a fixed order for calculating and superimposing compensation amounts: first, it generates compensation amounts based on the original test sequence. and Then add the temperature term, the coaxiality term, and the zero drift term.

[0064] After each superposition step is completed, a dimensional consistency and numerical boundary check is performed. If a main diagonal element has a sign inversion, exceeds the upper limit of the allowable parameter for that type of joint, or falls into the non-positive range, the compensation amount just added in that step is canceled, the compensation amount is set to zero as a whole, and subsequent calculations are continued. At the same time, it is recorded in the report as out of bounds and not used.

[0065] In the event of a conflict between the three types of compensation rules, the system will process the conflict in a cascading priority order: safety boundary first, measured geometric error second, and environmental drift third. That is, zero drift compensation will not cover the safety limit result already triggered by coaxiality error, and temperature compensation will not break through the aforementioned limit in the opposite direction. After the result is generated, it will be bound to the basic information of the joint module of the robot under test to avoid object mismatch during subsequent retrieval.

[0066] The test report simultaneously saves test condition information, output torque data, angular displacement data, vibration decay time series data, and error compensation results, and provides calibration conclusions, forming a complete archival record.

[0067] When specific sensor data is missing within a certain period, resulting in the inability to obtain complete compensation, the system sets the parameter corresponding to the missing compensation item to a zero matrix, making the compensation item a zero matrix, that is, making the compensation item zero, to ensure that the superposition of the basic matrix can be performed normally, thereby obtaining the current stiffness matrix and damping matrix.

[0068] Meanwhile, the system does not forcibly write unfounded calculated values ​​into the matrix. Instead, when outputting the calibration results, it adds missing item labels to the test report for items that were not corrected.

[0069] If the same compensation item is missing for multiple consecutive cycles, the system retains the aforementioned zeroing strategy, but stops writing the result of that cycle into the final factory ledger, and only allows it to be displayed as a process monitoring result.

[0070] This ensures both the robustness of matrix operations and the clarity of the compensation source in the final report, avoiding the difficulty in tracing data after mixing from different sources.

Claims

1. A method for calibrating the integrated dynamic parameters of robot joint stiffness and damping, characterized in that... This is achieved using a testing platform, which includes a floating adaptive mounting bracket, a six-dimensional force sensor, a precision encoder assembly, a servo loading component, a transient excitation component, and a temperature sensor. The methods include: Acquire the basic information of the joint module of the robot under test and install it on the floating adaptive mounting support; Based on the feedback data from the six-dimensional force sensor, the installation posture of the joint module of the robot under test is adaptively adjusted and coaxially calibrated. The servo loading component is controlled to apply a static load to the joint module of the robot under test according to a preset torque gradient, and the six-dimensional force sensor and the precision encoder group are used to synchronously collect and output torque data and angular displacement data. Based on the output torque data and angular displacement data, a torque-angular displacement relationship curve is constructed, and the stiffness parameters of the joint module of the robot under test are obtained by fitting. The transient excitation component is controlled to apply an impact to the joint module of the robot under test under load, and vibration decay time series data is collected by the precision encoder group; based on the vibration decay time series data, the damping parameters of the joint module of the robot under test are extracted by combining the logarithmic decay rate method with the Hilbert transform. Temperature data is acquired by a temperature sensor, and combined with installation coaxiality deviation data and sensor zero drift data, comprehensive error compensation is performed on the stiffness and damping parameters. The stiffness and damping parameters are constructed into stiffness matrices and damping matrices, respectively. The compensated stiffness and damping matrices are output as calibration results and a test report is generated.

2. The method for calibrating integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... Based on the feedback data from the six-dimensional force sensor, the installation posture of the joint module of the robot under test is adaptively adjusted and coaxially calibrated, including: The installation eccentricity is identified based on the force information output by the six-dimensional force sensor. The installation position of the joint module of the robot under test is finely adjusted in multiple degrees of freedom by the floating adaptive mounting bracket, and attitude compensation is performed. After adjusting the output axis of the robot joint module under test and the loading axis of the servo loading component to a preset coaxial range, lock the installation state.

3. The method for calibrating the integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... Controlling the servo loading component to apply a static load to the joint module of the robot under test according to a preset torque gradient includes: Control the joint module of the robot under test to maintain a horizontal posture; The servo loading component drives its loading axis to apply static load step by step according to a preset torque gradient, wherein a preset torque step is used between adjacent load levels.

4. The method for calibrating the integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... The stiffness parameters of the joint module of the robot under test are obtained, including: The torque-angular displacement relationship curve is fitted with least squares to obtain the fitting function, and the torsional stiffness parameter is calculated based on the slope of the fitting function. The radial off-center load or bending load data collected by the six-dimensional force sensor are obtained, and the bending stiffness parameters are calculated based on the change in torque and the change in angular displacement in the load direction. Temperature drift correction is applied to the torsional stiffness parameters and bending stiffness parameters to obtain the corrected stiffness parameters.

5. The method for calibrating the integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... The platform also includes a replaceable load linkage assembly, and the method further includes, before applying the impact: Adjust the joint module of the robot under test, which is equipped with the replaceable load linkage, to a vertical load-bearing state. The damping parameters of the joint module of the robot under test are extracted by combining the logarithmic decay rate method with the Hilbert transform, including: Based on the vibration attenuation time series data, adjacent vibration peak values ​​are extracted and the logarithmic attenuation rate is calculated. Perform Hilbert transform on the vibration decay time series data to obtain instantaneous envelope information; The equivalent damping coefficient is calculated based on the logarithmic decay rate and the instantaneous envelope information, and then corrected by combining the installation coaxiality deviation data to obtain the damping parameter.

6. The method for calibrating the integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... Comprehensive error compensation is performed on the stiffness and damping parameters, including: Establish a temperature drift compensation model; Establish an installation coaxiality error compensation model; Establish a sensor zero-drift compensation model; The parameters corrected by the temperature drift compensation model, the installation coaxiality error compensation model, and the sensor zero drift compensation model are fused to obtain the compensated stiffness matrix and damping matrix.

7. The method for calibrating integrated dynamic parameters of robot joint stiffness and damping according to claim 6, characterized in that, The compensated stiffness matrix can be expressed as: K= ； The 6×6 dimensional stiffness matrix obtained from the initial fitting is denoted as . The compensation amounts introduced by temperature drift, installation coaxiality error, and sensor zero drift on the stiffness matrix are respectively denoted as: , and The compensated damping matrix can be expressed as: C= ； The 6×6-dimensional damping matrix obtained from the initial calculation is denoted as . The compensation amounts introduced by temperature drift, installation coaxiality error, and sensor zero drift on the damping matrix are denoted as follows: , and .

8. The method for calibrating the integrated dynamic parameters of robot joint stiffness and damping according to claim 1, characterized in that... The output of the compensated stiffness matrix and damping matrix serves as the calibration results, and a test report is generated, including: The compensated stiffness matrix and damping matrix are bound to the basic information of the joint module of the robot under test; Generate a test report that includes test condition information, output torque data, angular displacement data, vibration decay time series data, error compensation results, and calibration conclusions.

Images