Robotic joint stiffness identification system and method based on standardized excitation flange

Abstract

The application belongs to the technical field of robots, and provides a robot joint stiffness identification system and method based on a standardized excitation flange, which comprises a test flange installed at the end of a robot, at least one excitation guide point on the test flange, at least two three-axis acceleration sensors fixed on preset fixed points of the test flange, and a data processing unit configured to: solve the linear acceleration signals of different excitation guide points under the application of torque excitation to obtain three-axis acceleration signals of the test flange and angular acceleration signals of the axis of the test flange, calculate the measured frequency response function matrix under each test working condition based on the three-axis acceleration signals of the test flange and the angular acceleration signals of the axis of the test flange, and identify the joint stiffness values of the robot by combining the measured frequency response function matrix under each test working condition and the theoretical frequency response function matrix. The robot joint stiffness values can be quickly, accurately and repeatedly identified.

Description

Technical Field

[0001] This invention belongs to the field of robotics technology, and in particular relates to a robot joint stiffness identification system and method based on a standardized excitation flange. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The stiffness of robot joints is a key parameter determining their dynamic performance, positioning accuracy, and vibration characteristics. In high-speed, high-precision applications, accurate identification of robot joint stiffness is crucial for optimizing control algorithms and improving operational performance. Experimental Modal Analysis (EMA), particularly the impact test, is a commonly used method for obtaining structural dynamic characteristics due to its relatively simple equipment and high testing efficiency. However, applying the existing impact test method to robot joint stiffness identification suffers from several technical problems: insufficient and non-standard excitation; missing key joint information; incomplete response measurements; inability to capture rotational dynamics; the need for large-scale robot movement; complex testing processes with poor identification accuracy; and a lack of standardized testing procedures and poor repeatability. These issues necessitate improvements in both the efficiency and accuracy of robot joint stiffness identification. Summary of the Invention

[0004] To address at least one of the technical problems mentioned above, this invention provides a robot joint stiffness identification system and method based on a standardized excitation flange. It offers a complete end-to-end solution that enables rapid, accurate, and repeatable identification of the stiffness of all robot joints.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A first aspect of the present invention provides a robot joint stiffness identification system based on a standardized excitation flange, comprising:

[0007] A test flange is installed at the end of the robot. The test flange has at least one excitation guide point. The excitation guide point and the central axis of the test flange have a preset lever arm distance, which is used to apply a quantifiable torque to stimulate the rotational dynamics of the robot's wrist joint.

[0008] At least two triaxial accelerometers are fixed at preset fixed points on the test flange to acquire linear acceleration signals at different excitation guide points under applied torque excitation.

[0009] The data processing unit is configured to: solve the linear acceleration signal of the obtained different excitation guide points under the application of the moment excitation to obtain the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis, calculate the measured frequency response function matrix under each test working condition based on the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis, and identify the joint stiffness value of the robot in combination with the measured frequency response function matrix under each test working condition and the theoretical frequency response function matrix.

[0010] Further, the test flange is a cross-shaped test flange, the cross-shaped test flange includes two orthogonally extending arms, the origin of the cross-shaped test flange coordinate system is located at the intersection of the center lines of the two arms of the bottom surface of the cross-shaped test flange, the axis and the axis are parallel to the directions of the center lines of the two arms respectively, the third axis of the cross-shaped test flange coordinate system is determined according to the directions of the axes and by the right-hand rule.

[0011] Further, the arm of the cross-shaped test flange is provided with a reinforcing boss, the reinforcing boss includes two orthogonal extension arms, the center line of each extension arm is parallel to the axes and , and the geometric center point of the center lines of the two orthogonal extension arms is located on the axis , the top surface and the side surface of the extension arm are parallel and perpendicular to the bottom surface of the cross-shaped test flange respectively, and the end of each extension arm of the reinforcing boss is vertically cut at a position with a distance from the axis to form four end planes which are perpendicular to the bottom surface of the cross-shaped test flange and the center line of the extension arm of the reinforcing boss.

[0012] Further, the excitation guide points include linear force excitation guide points and moment excitation guide points; wherein the linear force excitation guide points are arranged at the geometric center point of the center lines of the two orthogonal extension arms of the reinforcing boss, and the moment excitation guide points are arranged at the ends of the extension arms of the reinforcing boss of the cross-shaped test flange.

[0013] Further, in all test working conditions, the joints of the main arm of the robot remain fixed, and only the relative geometric relationship between the joint axes of the wrist of the robot is changed.

[0014] Further, the solving of the linear acceleration signal of the obtained different excitation guide points under the application of the moment excitation includes:

[0015] preprocessing the linear acceleration signal of the obtained different excitation guide points under the application of the moment excitation;

[0016] The linear acceleration signal after pretreatment is converted to the same coordinate system to obtain a corresponding time domain signal;

[0017] Based on the rigid body kinematics principle, the time domain signal is solved to obtain a three-axis acceleration signal of the test flange and an angular acceleration signal of the test flange axis.

[0018] Further, when calculating the measured frequency response function matrix under each test condition, the multiple test results of the same test condition are averaged to obtain the final measured frequency response function matrix result.

[0019] Further, the determination method of the theoretical frequency response function matrix is: a simplified dynamic equation is established, the simplified dynamic equation is converted into a state space form, based on the state space model, for any given joint stiffness vector, the theoretical frequency response function matrix corresponding to each condition is calculated.

[0020] Further, when the measured frequency response function matrix under each test condition and the theoretical frequency response function matrix are combined to identify the joint stiffness values of the robot, a cost function is defined based on the theoretical and experimental frequency response function matrices under all conditions, and the stiffness vector that minimizes the cost function is the final identification result.

[0021] The second aspect of the present application provides a robot joint stiffness identification method based on a standardized excitation flange, based on the robot joint stiffness identification system based on the standardized excitation flange of the first application, comprising the following steps:

[0022] Obtain the linear acceleration signal of different excitation guide points under the application of torque excitation;

[0023] The obtained linear acceleration signal of different excitation guide points under the application of torque excitation is solved to obtain a three-axis acceleration signal of the test flange and an angular acceleration signal of the test flange axis.

[0024] Based on the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis, the measured frequency response function matrix under each test condition is calculated, and the measured frequency response function matrix under each test condition and the theoretical frequency response function matrix are combined to identify the joint stiffness values of the robot.

[0025] Compared with the prior art, the present application has the following advantages:

[0026] The present application aims at the problem of insufficient and non-standard excitation, and changes the experience-dependent and arbitrary hammering action in traditional tests into standardized and repeatable engineering operations, thereby ensuring the quality and consistency of excitation from the source, and ensuring that all joint dynamics are effectively excited through standard and complete excitation.

[0027] The present application is aimed at the problem of incomplete measurement, and realizes accurate measurement of four-degree-of-freedom response through double-sensor differential measurement and corresponding settlement algorithm, breaks through the bottleneck of angular acceleration measurement, and provides four-degree-of-freedom high-quality response data containing three-dimensional translation and one-dimensional rotation for the first time for joint stiffness identification, and provides essential end-axis direction angular acceleration information for accurately decoupling and identifying the stiffness of J4, J5, J6 and other wrist joints.

[0028] The present application is aimed at the problem of complex test process and poor identification accuracy, and effectively avoids this problem through the test strategy of 'fixing the main arm and traversing the wrist', and effectively avoids the nonlinear changes of joint load and stiffness caused by large-scale motion due to the minimal influence of wrist joint motion on the overall load of the robot. At the same time, through systematic testing of three groups of necessary poses and global optimization combined with multi-working-condition data, it is ensured that the joint stiffness parameters can be accurately decoupled and identified under the assumption of linear model.

[0029] The present application is aimed at the problem of lack of standard test process and poor repeatability, and integrates standardized hardware (cross-shaped flange), standardized process (excitation point and test pose definition) and standardized algorithm (response calculation and parameter identification) into one, forming a complete end-to-end solution. From sensor installation, hammer excitation to data processing, each step has clear specifications, eliminating the human uncertainty introduced by temporary construction and arbitrary point placement in traditional methods, ensuring the high consistency and reproducibility of test results, and laying a foundation for the large-scale and engineering application of robot joint stiffness.

[0030] The advantages of the additional aspects of the present application will be partially given in the following description, partially will become obvious from the following description, or will be understood through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0031] The drawings accompanying the specification of the present application form a part thereof, serve to provide further understanding of the present application, and together with the description of the exemplary embodiments of the present application and their description serve to explain the present application, and do not constitute an improper limitation of the present application.

[0032] Figure 1 is a cross-shaped test flange installation schematic diagram provided by the embodiment of the present application;

[0033] Figure 2 is a cross-shaped test flange schematic diagram provided by the embodiment of the present application;

[0034] Figure 3 is a robot joint stiffness identification method flowchart based on a standardized excitation flange provided by the embodiment of the present application;

[0035] Figure 4 are three groups of necessary test poses of the robot provided by the embodiment of the present application; wherein (a) is necessary test pose 1, (b) is necessary test pose 2, and (c) is necessary test pose 3;

[0036] Figure 5 are three groups of necessary test pose axis schematic diagrams provided by the embodiment of the present application; wherein (a) is necessary pose 1 axis schematic diagram, (b) is necessary pose 2 axis schematic diagram, and (c) is necessary pose 3 axis schematic diagram;

[0037] 1. Robot; 2. Test flange; 3. Three-axis acceleration sensor, 4. Reinforcing boss; 5. Base; 6. Center boss; 7. Center knock point; 8. Edge knock point. DETAILED DESCRIPTION

[0038] The present application will be further described below in conjunction with the accompanying drawings and embodiments.

[0039] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0040] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0041] The present application discloses a robot joint stiffness identification system and method based on a standardized excitation flange, aiming to solve the problems of insufficient excitation, complex operation, poor repeatability, and inability to effectively decouple the dynamic characteristics of each joint in the existing robot joint stiffness test. The specific performance is as follows:

[0042] 1. Insufficient and non-standard excitation, missing key joint information: conventional tests usually hammer along the axis direction near the center of the robot end flange. The excitation force line produced by this method passes through or approaches the rotation center of the robot J4, J6 shaft, which cannot produce effective excitation torque, resulting in the dynamic characteristics of these joints (especially torsional stiffness) cannot be fully excited and measured, causing incomplete identification results. At the same time, due to the lack of standardized guidance, the randomness of the hammering point and direction introduces human error that is difficult to quantify.

[0043] 2. Response measurement is not comprehensive, and rotational dynamics cannot be captured: Most test methods only directly measure the three-dimensional linear acceleration of the robot's end, and cannot directly obtain the angular acceleration information of the end. This makes it difficult to decouple the J4, J5, and J6 joint deformation dynamics of the robot wrist part from the effects of other joints, resulting in systematic deviations in the stiffness identification of the robot wrist part.

[0044] 3. Robot needs to move in a large range, test process is complex and identification accuracy is poor: In the identification of existing methods, in order to decouple the influence of each axis on the end response as much as possible, the robot needs to move in a large range in the workspace. However, the site may not allow the robot to move in a large range, and the test sensor may need to be frequently disassembled due to cable limitations, which brings difficulties to the on-site test. In addition, the large range movement of the robot will cause significant changes in joint load, introducing additional influencing factors in joint stiffness identification, resulting in reduced joint stiffness identification accuracy.

[0045] 4. Test process lacks standardization and repeatability is poor: Due to the lack of standardized test devices and processes, existing methods highly depend on the professional experience of test engineers. The temporary setup of excitation devices, the arbitrary arrangement of sensors, the selection of hammering points, etc. all have uncertainties, which makes it difficult to guarantee the repeatability and reliability of test results, hindering the engineering and large-scale application of the technology.

[0046] The present application includes a cross-shaped test flange, two three-axis acceleration sensors arranged at the preset points of the flange, and a set of matching test and stiffness calculation methods. The cross-shaped flange provides standardized linear force and torque application guide points for hammer excitation; the double-sensor arrangement scheme can calculate the three-dimensional translational and torsional dynamic response of the end flange; and the calculation method uses high-quality excitation and response data under known geometric relationships to accurately identify the stiffness of each joint of the robot. Through the collaborative design of software and hardware, the present application provides a complete end-to-end solution that can quickly, accurately, and repeatedly identify the stiffness of all joints of the robot.

[0047] As an embodiment of the present application, the embodiment provides a robot joint stiffness identification system based on a standardized excitation flange, comprising:

[0048] A test flange is installed at the end of the robot, and at least one excitation guide point is provided on the test flange, the excitation guide point and the center axis of the test flange having a preset force arm distance for applying a quantifiable torque to excite the rotational dynamics of the wrist joints of the robot;

[0049] At least two three-axis acceleration sensors are fixed on the preset fixed points of the test flange for obtaining linear acceleration signals under the excitation of different excitation guide points;

[0050] The data processing unit is configured to: solve the linear acceleration signal of the obtained different excitation guide points under the applied torque excitation to obtain the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis, calculate the measured frequency response function matrix under each test working condition based on the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis, and identify the joint stiffness value of the robot in combination with the measured frequency response function matrix under each test working condition and the theoretical frequency response function matrix.

[0051] The present application solves the problems of insufficient excitation, complex operation, poor repeatability and inability to effectively decouple the dynamic characteristics of each joint in the existing robot joint stiffness test, and provides a complete end-to-end solution that can quickly, accurately and repeatedly identify the joint stiffness of the robot.

[0052] As shown in Figure 1 , a test flange 2 is installed at the end of the robot 1, the test flange 2 is a cross-shaped test flange, and at least two three-axis acceleration sensors 3 are installed on the preset fixed points of the cross-shaped test flange in a specified coordinate alignment manner, and an excitation guide point is arranged on the cross-shaped test flange; in the present embodiment, the robot 1 is a six-axis robot;

[0053] As shown in Figure 2 , the cross-shaped test flange is a preset structure, and in the present embodiment, two three-axis acceleration sensors 3 are arranged on the cross-shaped test flange, and in the present embodiment, IEPE type three-axis acceleration sensors are used, which are respectively denoted as three-axis acceleration sensor A and three-axis acceleration sensor B;

[0054] Specifically, the three-axis acceleration sensor A is installed on the arm end mounting base along the positive direction of the axis, and the three-axis acceleration sensor B is installed on the arm end mounting base along the negative direction of the axis. The installation of the two sensors uses the reference surface on the base to align, and the relative attitude relationship of their respective local coordinate systems and with the flange coordinate system is a known fixed value, which is stored in the end response signal solving related program, wherein, , , are three orthogonal directions of the flange coordinate system .

[0055] The cross-shaped test flange is designed to be a lightweight and high-rigidity structure to avoid the influence of the "mass loading effect" on the measured robot system. Specifically, the cross-shaped test flange has two orthogonally extended arms, defining a cross flange coordinate system The origin of the coordinate system is located at the intersection of the center lines of the two arms of the cross flange bottom surface, The axis of the coordinate system is parallel to the direction of the center line of the two arms, And The third axis of the coordinate system is parallel to the direction of the center line of the two arms, The third axis of the coordinate system is parallel to the direction of the center line of the two arms, The direction of the third axis is determined by the right-hand rule according to And The direction of the third axis is determined by the right-hand rule according to

[0056] The arms of the cross-shaped test flange are provided with reinforcing bosses 4, which are cross-shaped reinforcing bosses. The reinforcing bosses 4 include two orthogonal protruding arms, and the geometric center point of the reinforcing bosses 4 is located on the axis The center line of each protruding arm is parallel to the And The top surface and side surface of the protruding arm of the reinforcing boss 4 are parallel and perpendicular to the bottom surface of the cross-shaped test flange 2, respectively. The end of each protruding arm of the reinforcing boss 4 is vertically cut at a distance of The axis to form four end planes that are perpendicular to the bottom surface of the cross-shaped test flange and the center line of the protruding arm of the reinforcing boss. Each protruding arm of the reinforcing boss 4 is provided with a base 5 for mounting an acceleration sensor, and each base 5 uses the end plane and the bottom surface of the cross-shaped test flange as two orthogonal mounting reference surfaces to achieve precise positioning of the sensor.

[0057] Of course, in actual application, corresponding three-axis acceleration sensors can be arranged at the mounting bases of any two adjacent orthogonal arms, and the mounting position and direction of the sensor can be accurately represented in the coordinate system to ensure the consistency and repeatability of the sensor position during each test.

[0058] Preferably, the cross-shaped test flange is made of 7075 aerospace aluminum alloy to ensure high rigidity and light weight, and the extension length of the cross arm is 150mm;

[0059] As a further embodiment, the excitation guide points include linear force excitation guide points and torque excitation guide points;

[0060] Specifically, the linear force excitation guide point: the center of the cross flange has a high-rigidity center boss 6 with an axis The top surface of the center boss 6 is perpendicular to the The side surfaces are parallel or perpendicular to the And In the hammering experiment, the center point of the top surface of the center boss 6 is taken as the center knocking point 7, numbered , guiding the operator to knock along , , The axis alignment center knock point 7 knocks the top surface and each side surface of the center boss, and the cross test flange is subjected to a force along the , , Pure linear force excitation of the shaft.

[0061] Moment excitation guide point: the center points of the two end planes of the cross test flange without the acceleration sensor are taken as the edge knock points 8 of the cross flange, and are numbered as , The two points are moment excitation guide points, and the operator is guided to align along the axis direction of the coordinate system , Knocking the end of the end of the reinforced boss extension arm can generate a linear force excitation, and the force arm of the end flange is , and the moment excitation with an accurately calculated size can effectively excite the rotation dynamics of the wrist joints J4 and J6 of the robot wrist part.

[0062] The existing method usually randomly hammers the center of the end flange, and the force line passes through or is close to the shaft center of the wrist joints J4 and J6, which cannot generate effective moment, so that the torsional stiffness of these key joints cannot be identified. The present application solves the problem through the innovative design of the cross test flange:

[0063] Standardization of moment excitation: the end of the flange arm is provided with special moment excitation guide points (P1, P2). The operator hammers along the guide points, and a moment with a calculable size and a determined direction can be applied to the end. This directly solves the core problem of insufficient wrist rotation joint excitation, ensures that the dynamic characteristics of the J4 and J6 joints can be fully excited and measured, and thus complete six-joint stiffness information is obtained.

[0064] Standardization of linear force excitation: the linear force excitation guide point (P0) of the center of the flange provides a clear reference for the application of pure linear force along the three axes, and the specially designed center boss vertical surfaces provide a clear landing point for the axis direction hammering, so that the hammering can approach the desired axis direction as much as possible, greatly reducing the generation of additional moment.

[0065] Through the above design, the present application changes the experience-dependent and random hammering action in the traditional test into standardized and repeatable engineering operation, which guarantees the quality and consistency of the excitation from the source.

[0066] As an embodiment, in all test conditions, the joints of the main arm of the robot remain fixed, and only the relative geometric relationship between the wrist joint axes of the robot is changed.

[0067] As an embodiment, the linear acceleration signal calculation of the different excitation guide points under the application of moment excitation includes: ​

[0068] The linear acceleration signals of the obtained different excitation guide points under the application of torque excitation are preprocessed;

[0069] The preprocessed linear acceleration signals are converted to the corresponding time domain signals in the same coordinate system;

[0070] Based on the rigid body kinematics principle, the time domain signals are solved to obtain the three-axis acceleration signals of the test flange and the angular acceleration signals of the test flange axis.

[0071] As an embodiment, when calculating the measured frequency response function matrix under each test condition, the multiple test results of the same test condition are averaged to obtain the final measured frequency response function matrix result.

[0072] As an embodiment, the determination method of the theoretical frequency response function matrix is: establishing a simplified dynamic equation, converting the simplified dynamic equation into a state space form, and based on the state space model, for any given joint stiffness vector, calculating the theoretical frequency response function matrix corresponding to each condition.

[0073] As an embodiment, when the measured frequency response function matrix and the theoretical frequency response function matrix under each test condition are combined to identify the joint stiffness values of the robot, a cost function is defined based on the theoretical and experimental frequency response function matrices under all conditions, and the stiffness vector that minimizes the cost function is the final identification result.

[0074] As shown in Figure 3 As another embodiment of the present application, a robot joint stiffness identification method based on a standardized excitation flange is provided, which specifically includes the following steps:

[0075] Step 1: Obtain the linear acceleration signals of different excitation guide points under the application of torque excitation;

[0076] In this embodiment, according to the excitation guide points provided by the cross-shaped flange, a force hammer is used to perform multi-point and multi-direction hammering in sequence to apply standardized force and torque excitation. At the same time, the force signal of the force hammer and the three-axis acceleration signals of all acceleration sensors are recorded synchronously.

[0077] Based on the set standardized test procedure, high-quality dynamic data sufficient to decouple and identify all six joint stiffnesses are collected by changing only the wrist joint posture;

[0078] In this embodiment, the base and the main arm joints (J1, J2, J3) are fixed, only the posture of the wrist joints (J4, J5, J6) is systematically changed, and sufficient and diverse end excitation information is provided by knocking different excitation points. This test method not only avoids the test complexity and nonlinear disturbance introduced by the load change caused by the large range motion of the robot, but also fully excites the dynamic characteristics of all 6 joints, improves the identification accuracy.

[0079] Specifically, the following steps are included:

[0080] Step 101, test preparation and main arm positioning;

[0081] The cross-shaped test flange and sensor system are installed on the end of the robot to be tested, the robot is controlled to move the main arm joints J1, J2, J3 to a selected test position (for example, a working pose near the center of the working space), and is kept locked during the entire test.

[0082] Specifically, the robot is controlled to move the main arm joints J1, J2, J3 to a fixed test pose (for example, ), and keep locked during the subsequent wrist posture traversal and hammering test;

[0083] Step 102, under the premise of locking the main arm joints, adjust the robot wrist joints to move to the preset pose, and execute the excitation and collection process at each group of poses;

[0084] In this embodiment, under the premise of locking the main arm joints, the robot wrist joints J4, J5, J6 are controlled in sequence to move, so that the robot end reaches the set target test pose, and the excitation and collection process is executed;

[0085] Specifically, in order to fully excite the dynamic of each axis and effectively decouple the stiffness parameters of each joint, this embodiment defines three groups of necessary wrist typical poses, as shown in Figure 4 and Figure 5 These poses are constructed by changing the relative geometric relationship of the J4, J5, J6 joint axes, ensuring that the contribution of each joint can be clearly distinguished under different excitation.

[0086] Among them, the necessary pose 1: the J4, J6 axis is parallel, and the J5 axis is perpendicular to the J4, J6 axis and parallel to the J3 axis.

[0087] The necessary pose 2: the J4, J6 axis is perpendicular, and the J5 axis is perpendicular to the J4, J6 axis and parallel to the J3 axis.

[0088] The necessary pose 3: the J4, J6 axis is perpendicular, and the J5 axis is perpendicular to the J4, J6 axis and perpendicular to the J3 axis.

[0089] The J6 joint angle of the above-mentioned three groups of necessary poses is fixed to 0 degrees to reduce unnecessary complexity.

[0090] In this embodiment, under each group of poses, the following excitation and collection process is performed, including:

[0091] The operator holds a force hammer with a built-in force sensor, and sequentially hammers the three excitation guide points (center point P0, edge points P1 and P2) on the cross flange along the three orthogonal directions of the flange coordinate system , , The hammering direction of each excitation point is repeated 5 times to ensure data quality.

[0092] A data collection system is used to synchronously record 1 force signal output by the force hammer and 6 acceleration signals output by two three-axis acceleration sensors at a sampling rate of 10240 Hz;

[0093] Integrating all the poses, excitation points and excitation directions, the end force input and end acceleration output response of groups of working conditions are tested, and not less than times of end hammering excitation experiments are performed. Taking the three groups as an example, after completing the testing of all the three groups of poses, a total of times of effective hammering excitation and response data are obtained.

[0094] In addition to the above-mentioned necessary poses, other optional poses (for example, the J4, J5 and J6 joint angles are all 45 degrees) can be added to further enhance the diversity of the data and improve the robustness of the identification model.

[0095] The existing method often requires the robot to perform large-scale motion to decouple the influence of each joint, which not only complicates the operation, but also introduces nonlinear errors due to the dramatic change of joint load, thereby reducing the identification accuracy. The present application effectively avoids this problem through the testing strategy of “fixing the main arm and traversing the wrist”:

[0096] Simplified testing process: only the J4, J5 and J6 wrist joint poses need to be changed during the entire test, and the main arm joints (J1, J2 and J3) remain fixed. This greatly simplifies the on-site operation and avoids problems such as cable entanglement and insufficient space.

[0097] Improved identification accuracy: since the wrist joint motion has little effect on the overall load of the robot, this method effectively avoids the nonlinear changes in joint load and stiffness caused by large-scale motion. At the same time, through systematic testing of the three necessary poses and global optimization of the multi-working-condition data, it is ensured that the joint stiffness parameters can be accurately decoupled and identified under the linear model assumption.

[0098] ​Step 2: Linear acceleration signals of different excitation guide points are calculated under the linear excitation torque, and three-axis acceleration signals of the test flange and angular acceleration signals of the test flange axis are obtained;

[0099] In this embodiment, the known and fixed sensor geometric position relationship is used, and the collected multiple sets of linear acceleration signals are converted into three-axis acceleration signals of the robot end flange and angular acceleration signals of the cross flange axis through a preset algorithm.

[0100] Specifically, the following steps are included:

[0101] Step 201: The original acceleration signals obtained are preprocessed;

[0102] In this embodiment, the preprocessing specifically includes filtering and windowing processing of the force signal and the 6-way acceleration signal of each set of data.

[0103] Specifically, the filtering includes applying a band-pass filter, such as a 2-2000Hz band-pass filter, to filter out high-frequency noise irrelevant to the structural modal and low-frequency drift caused by sensor zero offset.

[0104] The windowing processing includes applying a suitable window function (such as an exponential window) to the effective signal segment of each hammering to reduce the frequency spectrum leakage caused by signal truncation.

[0105] After processing, the force signal of this hammering is obtained and two sets of clear acceleration signals 、 .

[0106] Step 202: The preprocessed acceleration signals are converted into corresponding time domain signals in the same coordinate system;

[0107] In this embodiment, the pre-stored rotation matrix and are used to convert the acceleration signals of sensors A and B 、 into the flange coordinate system , obtaining and .

[0108] Since the measurement values of sensors A and B are based on their respective local coordinate systems and , in order to perform kinematic calculation, they must be unified into the common coordinate system of the cross flange .

[0109] According to the precise installation relationship defined in step 1, sensors A and B are relative to the coordinate system The attitude is determined by the rotation matrix , Precise description.

[0110] Therefore, the two sets of acceleration signals can be converted to the following formula: In coordinate system:

[0111] (1),

[0112] (2),

[0113] in, and Indicates that sensors A and B are in Acceleration time-domain signal in coordinate system;

[0114] Step 203: Based on the principle of rigid body kinematics, the time-domain signal is solved to obtain the triaxial translational acceleration and angular acceleration about the axis of the end test flange;

[0115] In this embodiment, the triaxial translational acceleration of the entire rigid body (cross flange) is solved using the known accelerations at two points. With respect to the axis angular acceleration of rotation .

[0116] Among them, the triaxial translational acceleration of the end cross flange The solution includes:

[0117] Origin of the cross flange coordinate system Translational acceleration By measuring acceleration and The averaging method is used to obtain the result, which effectively suppresses acceleration measurement noise, and is expressed as:

[0118] (3),

[0119] Flange angular acceleration about axis The solution includes:

[0120] Cross flange around axis angular acceleration The solution can be obtained by analyzing the acceleration difference vector between the two sensors. Assume the overall arrangement of the acceleration sensors is as follows: Figure 2 As shown, A is located The sensor mounting base is located in the positive axis direction, and the accelerometer B is positioned there. Sensor mounting base in the negative direction of the axis. Then around... angular acceleration of the axis It can be represented as:

[0121] (4),

[0122] wherein, , are respectively in the direction of , component, , are respectively in the direction of , component, is the distance from the sensor mounting base to the center axis of the flange.

[0123] Step 204, the calculated three-axis linear acceleration of the cross flange and the angular acceleration around the axis as the end response signal of this hammering, together with the corresponding force signal are stored to provide complete input data for the stiffness identification of step 4.

[0124] The existing method only measures three-dimensional linear acceleration and cannot capture the rotation dynamics, resulting in difficulty in decoupling wrist joint stiffness and inaccurate identification. The present application completely changes this situation through a double-sensor differential measurement scheme and a supporting algorithm:

[0125] (a) Direct calculation of angular acceleration: by arranging two three-axis acceleration sensors with known accurate relative positions on the flange, the present application uses rigid body kinematics to successfully calculate the end angular acceleration around the axis (ωz) that cannot be obtained by traditional methods by analyzing the difference between the two sets of acceleration signals. ).

[0126] (b) Complete response data: combined with the calculated angular acceleration and the average calculated three-axis linear acceleration, the present application first provides four-degree-of-freedom high-quality response data containing three-dimensional translation and one-dimensional rotation for joint stiffness identification, providing essential end axis direction angular acceleration information for accurate decoupling and identification of J4, J5, J6, etc. wrist joint stiffness.

[0127] Step 3: Calculate the measured frequency response function matrix under each test condition based on the three-axis acceleration signal of the test flange and the angular acceleration signal of the test flange axis;

[0128] In this embodiment, the measured frequency response function (FRF) matrix is calculated using the end four-degree-of-freedom response signal and the corresponding excitation force signal .

[0129] In this embodiment, the calculated excitation force and four-degree-of-freedom response (Y) , ), calculate the theoretical frequency response function matrix of each test condition (pose and excitation point combination) experimental frequency response function (FRF) matrix The results of multiple (such as 5) repeated tests for the same condition are averaged to improve the signal-to-noise ratio.

[0130] Among them, the single hammer FRF calculation: for single hammer, the excitation is a three-dimensional force vector, and the response is a four-dimensional acceleration vector. A FRF matrix can be calculated, where each column represents an excitation direction ( , , ), and each row represents a response direction ( , , , ).

[0131] Among them, data averaging and integration: the results of multiple hammer tests for the same condition (same pose, excitation point, excitation direction) are averaged to improve the signal-to-noise ratio of the FRF.

[0132] Integrate the FRF matrices of all conditions to obtain a data set containing a set of experimental FRF matrices as a reference for subsequent parameter identification, where is the input frequency value of the FRF matrix.

[0133] Step 4: Identify the joint stiffness values of the robot by combining the measured frequency response function matrices of each test condition and the theoretical frequency response function matrices.

[0134] Specifically, the following steps are included:

[0135] Step 401, simplify the dynamic equation, convert the simplified dynamic equation to state space form, and based on the state space model, for any given joint stiffness vector, calculate the theoretical frequency response function matrix corresponding to each condition;

[0136] In this embodiment, to describe the dynamic response of the system under hammer excitation, a simplified, linear robot dynamics model is established;

[0137] Specifically, it includes:

[0138] Step 4011, simplify the dynamic equation: considering that the robot joints are locked (rigid joint angular velocity and acceleration are zero) and the joint deformation is small during hammer testing, the complex nonlinear dynamic equation can be simplified to the following linear differential equation:

[0139] (5),

[0140] In the formula, Let be the rigid joint angle vector of the robot. This is the joint dynamic deformation vector. and The first Rigid joint rotation and dynamic torsional deformation of each joint For the quality matrix, Here is the joint stiffness matrix. Here is the joint deformation damping matrix. , These are the joint stiffness vector and the deformation damping vector, respectively. and The first Torsional stiffness and deformation structural damping of each joint, Represents diagonal matrix operators. From the base coordinate system to a specific excitation point The velocity Jacobian matrix, the excitation point The center tapping point can be predefined on the cross flange. Edge tapping point and , This is the end-effector vector acting on the excitation point.

[0141] Compared to the complete dynamic equation, this equation is reasonably simplified based on the specific working conditions of the hammer impact test. It ignores the second-order effects of Coriolis force, centrifugal force, and small deformations at the joints on the mass and Jacobian matrix, enabling efficient linearization analysis while maintaining accuracy. Due to the rigid joint angle locking, and All A defined constant matrix can be obtained through calculations using the robot's forward kinematics and dynamics models. Joint stiffness matrix. and joint deformation damping matrix These are the parameters to be identified.

[0142] Step 4012: Convert the simplified dynamic equations into state-space form;

[0143] To facilitate the calculation of the FRF, the simplified dynamic equations are converted into state-space form.

[0144] First, adopting the Rayleigh damping assumption, let ,in This is an empirical coefficient that can be estimated based on material properties. The system's state vector is defined as follows: The system output is the terminal four-degree-of-freedom acceleration. , operator The vertical stack of each element. The state space equation of the system can be expressed as:

[0145] (6),

[0146] (7),

[0147] The independent variables of each dynamic matrix are Omit the representation, and let and be the zero matrix and the unit matrix, respectively. The system matrix of the state space can be specifically expressed as:

[0148] (8),

[0149] where the output Jacobian matrix is an important part of the state space, which maps the deformation velocity in the joint space to the four-degree-of-freedom motion velocity of the end flange. The matrix is The upper three rows are the standard velocity Jacobian matrix, and the fourth row is the angular velocity Jacobian of the axis, defined as where and are the velocity Jacobian and angular velocity Jacobian from the robot base coordinate system to the cross flange coordinate system, is the selection matrix of the axis angular acceleration.

[0150] Step 4013, based on the state space model, for any given joint stiffness vector, calculate the theoretical frequency response function matrix corresponding to each working condition, denoted as:

[0151] (9),

[0152] where is the imaginary unit;

[0153] For the defined group of test working conditions, the corresponding robot posture and excitation point Jacobian can be substituted to calculate the group of theoretical FRF matrix, denoted as .

[0154] Step 402, define the cost function based on the theoretical and experimental frequency response function matrices under all working conditions, and find the stiffness vector that minimizes the cost function;

[0155] Specifically, the following steps are included:

[0156] ​Step 4021, the difference between the theoretical and experimental FRF matrices under all working conditions is accumulated to define a global cost function is:

[0157] (10),

[0158] wherein, is the Frobenius norm of the matrix, which integrates the errors of all frequency points, all excitation and response channels, the number of groups of test working conditions, is the measured frequency response function matrix under the i test working condition of the th group, i is the theoretical frequency response function matrix under the th test working condition of the

[0159] th group; Step 4032, a mature nonlinear optimization algorithm (such as Levenberg-Marquardt or sequential quadratic programming) is used to find the stiffness vector that minimizes the cost function

[0160] (11),

[0161] After optimization convergence, the final identified torsional stiffness values of the robot J1 to J6 joints are output .

[0162] The present application integrates standardized hardware (cross flange), standardized process (excitation point and test pose definition) and standardized algorithm (response calculation and parameter identification) into one, forming a complete end-to-end solution. From sensor installation, hammer excitation to data processing, each step has clear specifications, eliminating the human uncertainty introduced by temporary construction, arbitrary point placement, etc. in the traditional method, ensuring the high consistency and reproducibility of the test results, and laying a foundation for the large-scale and engineering application of robot joint stiffness.

[0163] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A robot joint stiffness identification system based on standardized actuation flanges, characterized by, Comprising: A test flange mounted on the end of the robot, the test flange being provided with at least one excitation guide point, the excitation guide point and the center axis of the test flange having a preset force arm distance for applying a quantifiable torque to excite the rotational dynamics of the robot wrist joint; At least two three-axis acceleration sensors are fixed on the preset fixed points of the test flange respectively, for acquiring linear acceleration signals of different excitation guide points under the excitation of the applied torque; A data processing unit configured to: solve the acquired linear acceleration signals of different excitation guide points under the excitation of the applied torque to obtain three-axis acceleration signals of the test flange and angular acceleration signals of the test flange axis, calculate the measured frequency response function matrix under each test condition based on the three-axis acceleration signals of the test flange and the angular acceleration signals of the test flange axis, and identify the stiffness values of each joint of the robot by combining the measured frequency response function matrix under each test condition and the theoretical frequency response function matrix. The test flange is a cross-shaped test flange, the cross-shaped test flange comprising two orthogonally extending arms, defining a cross-shaped test flange coordinate system with its origin at the intersection of the center lines of the two arms of the cross-shaped test flange bottom surface, with its axis and parallel to the direction of the center lines of the two arms, respectively, with its third axis pointing according to the right-hand rule from and determination. The arm of the cross-shaped test flange is provided with a reinforcing boss, the reinforcing boss comprises two orthogonal extension arms, the center line of each extension arm is parallel to the axis, the geometric center point of the center lines of the two orthogonal extension arms is located on the axis, the top surface and the side surface of the extension arm are parallel and perpendicular to the bottom surface of the cross-shaped test flange respectively, the end of each extension arm of the reinforcing boss is vertically cut at a position with a distance of 1 / 2 to 2 times of the distance from the axis to the end of the extension arm, thereby forming four end planes which are perpendicular to the bottom surface of the cross-shaped test flange and the center line of the extension arm of the reinforcing boss. and the axis, the geometric center point of the center lines of the two orthogonal extension arms is located on the axis, the top surface and the side surface of the extension arm are parallel and perpendicular to the bottom surface of the cross-shaped test flange respectively, the end of each extension arm of the reinforcing boss is vertically cut at a position with a distance of 1 / 2 to 2 times of the distance from the axis to the end of the extension arm, thereby forming four end planes which are perpendicular to the bottom surface of the cross-shaped test flange and the center line of the extension arm of the reinforcing boss. the axis, the geometric center point of the center lines of the two orthogonal extension arms is located on the axis, the top surface and the side surface of the extension arm are parallel and perpendicular to the bottom surface of the cross-shaped test flange respectively, the end of each extension arm of the reinforcing boss is vertically cut at a position with a distance​​ The excitation guide points include linear force excitation guide points and torque excitation guide points; wherein the linear force excitation guide points are arranged at the geometric center points of the two orthogonal extension arm center lines of the reinforcing boss, and the torque excitation guide points are arranged at the ends of the extension arms of the cross-shaped test flange reinforcing boss.

2. The standardized actuation flange based robot joint stiffness identification system of claim 1, wherein, In all test conditions, the joints of the main arm of the robot remain fixed, and only the relative geometric relationship between the joints of the wrist of the robot is changed.

3. The standardized actuation flange based robot joint stiffness identification system of claim 1, wherein, The solving of the acquired linear acceleration signals of different excitation guide points under the excitation of the applied torque includes: Pretreatment of the acquired linear acceleration signals of different excitation guide points under the excitation of the applied torque; Convert the pretreated linear acceleration signals to the same coordinate system to obtain corresponding time domain signals; Based on the rigid body kinematics principle, the time domain signals are solved to obtain the three-axis acceleration signals of the test flange and the angular acceleration signals of the test flange axis.

4. The standardized actuation flange based robot joint stiffness identification system of claim 1, wherein, When calculating the measured frequency response function matrix under each test condition, the test results of the same test condition are averaged to obtain the final measured frequency response function matrix result.

5. The standardized actuation flange based robotic joint stiffness identification system of claim 1, wherein, The determination method of the theoretical frequency response function matrix is: establishing a simplified dynamics equation, converting the simplified dynamics equation into a state space form, and based on the state space model, for any given joint stiffness vector, calculating the theoretical frequency response function matrix corresponding to each condition.

6. The standardized actuation flange based robotic joint stiffness identification system of claim 1, wherein, When combining the measured frequency response function matrix under each test condition and the theoretical frequency response function matrix to identify the stiffness values of each joint of the robot, a cost function is defined based on the theoretical and experimental frequency response function matrices under all conditions, and the stiffness vector that minimizes the cost function is the final identification result.

7. A method for robot joint stiffness identification based on standardized actuation flange, characterized in that, The robot joint stiffness identification system based on the standardized excitation flange according to any one of claims 1-6, comprising the following steps: Acquiring linear acceleration signals of different excitation guide points under the excitation of the applied torque; Solving the acquired linear acceleration signals of different excitation guide points under the excitation of the applied torque to obtain three-axis acceleration signals of the test flange and angular acceleration signals of the test flange axis; The measured frequency response function matrix under each test condition is calculated based on the three-axis acceleration signals of the test flange and the angular acceleration signals of the test flange axis, and the joint stiffness values of the robot are identified by combining the measured frequency response function matrix under each test condition and the theoretical frequency response function matrix.

Images