Quantitative test method and system for peak stall torque of a robot joint module

Abstract

The application discloses a kind of quantitative test method and system of robot joint module peak stall torque, it is related to robot joint performance evaluation technical field, comprising: extracting the torque time history curve of joint module, peak stall torque is defined as controlled low-frequency fatigue loading process, the mapping model of torque amplitude and failure cycle number is established;Test system is constructed, torque-current static calibration is carried out;Specific excitation function is embedded in servo driver current loop;Bus voltage and rotational speed change rate are monitored in real time, using back electromotive force compensation algorithm and combining energy dissipation discharge loop, the voltage and current impact of unloading moment is inhibited;Repeatedly loaded until structure is damaged, cycle number is recorded, peak torque-life curve is drawn, as whole machine life prediction basis.The application realizes the quantitative correlation of peak stall torque and fatigue life, realizes the quantitative evaluation from static index to dynamic life, provides basis for robot whole machine life prediction, improves test precision.

Description

Technical Field

[0001] This invention relates to the field of robot joint performance evaluation technology, specifically a quantitative testing method and system for the peak stall torque of a robot joint module. Background Technology

[0002] High dynamic response capability is a core technical indicator for humanoid robots to achieve complex movements such as somersaults, running and jumping. Peak stall torque, as the cornerstone for evaluating the instantaneous power output capability of joint modules, directly determines the robot's extreme motion performance.

[0003] Existing standards (such as GB / T 30549-2014) only define peak stall torque as "the maximum torque that can be output for a short time," lacking in-depth analysis of its physical failure mechanism. Moreover, in actual operating conditions, when the joint module is subjected to peak torque, its internal mechanical structure (such as reducer gears and output shaft) is subjected to cyclic limit loads. Based on the first principles of materials mechanics, this failure is essentially a low-frequency fatigue failure related to the material's SN curve (stress-life curve) and specific operating conditions.

[0004] Existing technologies mostly focus on torque control algorithms and system implementation, but there is no quantitative testing scheme for the relationship between "operating conditions-life-peak torque". This makes it impossible to accurately assess the lifespan and structural boundaries of joint modules under specific impact conditions during robot development, which poses a risk of over-design or insufficient strength.

[0005] Based on this, a quantitative testing method and system for the peak stall torque of a robot joint module is provided, which can eliminate the drawbacks of existing technical solutions. Summary of the Invention

[0006] The purpose of this invention is to provide a quantitative testing method and system for the peak stall torque of a robot joint module, in order to solve the problem that existing standards in the background art only define the peak torque without relating it to operating conditions and lifespan, and lack quantitative testing methods, which leads to design risks.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A quantitative testing method for the peak stall torque of a robot joint module, specifically including the following steps:

[0009] Step S1: By sampling and analyzing the whole machine data of the humanoid robot under typical working conditions, the torque time history curve of the joint module is extracted. Based on the material mechanics analysis, the peak stall torque is defined as a controlled low-frequency fatigue loading process, and a mapping relationship model between torque amplitude and failure cycle number is established.

[0010] Step S2: Establish a test system including mechanical stall tooling, joint module, servo driver and host computer, and perform static calibration of torque-current mapping relationship to ensure that the servo driver current loop control can reflect the output torque.

[0011] Step S3: Embed a specific excitation function in the servo driver current loop interrupt, set the total test period to T, and the specific excitation function includes a pre-load stage and a peak load stage in a single loading, and performs symmetrical reverse loading at time point T / 2.

[0012] Step S4: Monitor the bus voltage and speed change rate in real time. At the moment of torque unloading, use the reverse electromotive force compensation algorithm to actively adjust the duty cycle and open the energy dissipation circuit to offset the instantaneous back pressure and prevent the servo driver from being over-voltage protected and the hardware from being damaged.

[0013] Step S5: Repeat step S3, record the number of cycles when the joint module suffers structural damage, and plot the peak torque-life cycle curve of the joint module as input for predicting the overall lifespan of the humanoid robot.

[0014] Further, step S1 specifically includes: training the humanoid robot's agent under typical working conditions through a reinforcement learning model, and deploying it on the entire humanoid robot to collect and analyze data; storing torque time history data and generating torque time history curves based on lightweight communication middleware; defining the peak stall torque as a controlled low-frequency fatigue loading process based on material mechanics analysis; establishing a uniaxial strain amplitude-life relationship using the Manson-Coffin equation; constructing a mapping relationship model between torque amplitude and failure cycle number; with the load spectrum as the input and the mapping diagram between torque amplitude and failure cycle number as the output.

[0015] Furthermore, the reinforcement learning model is either the MIMIC model or the AMP model, and the typical working conditions include jumping and falling recovery.

[0016] Furthermore, the specific operations of static calibration in step S2 include:

[0017] The joint module to be tested is installed on the test bench of the mechanical stall fixture. The output end of the joint module to be tested is connected to the input end of the torque sensor. The output end of the torque sensor is locked. The current value of the joint module and the torque value of the torque sensor are sampled at N points from 0 to the peak torque, where N is a positive integer greater than or equal to 5.

[0018] Furthermore, the specific excitation function in step S3 is used to generate a precision torque command in the servo driver current loop to test the steady-state torque output and electrical characteristics of the joint module in a pure stall state, providing clean static load data. The specific logic of the specific excitation function includes:

[0019] Let the peak torque be The total testing period is , The time is 10 seconds, and the torque command is... ;

[0020] The torque command expression for the preloading phase is: , ,in, For the test time, The preloading phase lasts 50ms. A linear ramp of time gradually increases the torque from 0 to... It uses gentle force to mesh gears and eliminate mechanical backlash in the transmission chain;

[0021] The torque command expression for the peak loading stage is: , ,in, For 100ms, in At that moment, the instruction came from Step to The motor quickly enters and remains in a stall state for 100ms. The current and voltage data collected at this time are used to reflect the pure steady-state stall torque characteristics.

[0022] During the symmetrical reverse loading process, At that time, the above-mentioned phase process is repeated, and the torque command is... This is to obtain reverse stall data and form a complete bidirectional load cycle.

[0023] Furthermore, the back EMF compensation algorithm in step S4 is used to predict speed based on encoder changes, and to calculate and compensate for the back EMF through the back EMF coefficient. The specific operations of the back EMF compensation algorithm include:

[0024] A sampling resistor and encoder are used to monitor the bus voltage and speed change rate in real time. The encoder signal is acquired at high frequency, filtered, and then the motor's mechanical angular velocity is estimated in real time. The instantaneous back electromotive force (EMF) estimate is calculated using the motor's back EMF constant. The expression is as follows: ,in, For the instantaneous back electromotive force estimation, Let be the back electromotive force constant of the motor. Given the motor's mechanical angular velocity, the instantaneous back electromotive force estimate is directly superimposed onto the voltage command output by the current loop as a feedforward quantity. The expression is: ,in, This is the voltage command that the servo driver ultimately outputs to the motor. The back EMF compensation algorithm is used to actively counteract the back EMF disturbance, suppress the voltage and current surges during unloading, protect the servo driver, and improve dynamic stability.

[0025] Furthermore, in step S4, the energy dissipation circuit converts the potential energy generated by the instantaneous reverse voltage into heat energy by connecting the bypass resistor. The instantaneous reverse voltage is generated by the permanent magnet on the rotor cutting the magnetic field when it undergoes instantaneous reverse micro-motion.

[0026] Furthermore, the criteria for determining structural damage in step S5 are at least one of the following: broken teeth in the joint module and deformation exceeding the preset tolerance.

[0027] A quantitative testing system for the peak stall torque of a robot joint module is used to execute a quantitative testing method for the peak stall torque of a robot joint module, comprising:

[0028] A mechanical stall tooling, used as a testing device to fix the joint module and realize the stall function, is a double flange structure;

[0029] Joint module, as a device under test, is used to execute mechanisms for active degrees of freedom in a humanoid robot;

[0030] The servo driver, as the electronic control board of the joint module, takes DC power and communication commands as inputs and controls the servo movement of the joint module through the servo program in the electronic control board.

[0031] The host computer is connected to the joint module via CAN communication. It is used to store torque time history data, send control commands to the joint module, and receive current feedback signals to complete the life quantification assessment.

[0032] Furthermore, one set of flanges in the dual-flange structure is used to lock the fixed flange of the joint module, and the other set of flanges is used to lock the output flange of the joint module.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0034] 1. This invention defines the peak stall torque as a controlled low-frequency fatigue loading process and establishes a mapping model between torque amplitude and failure cycle number, thereby realizing the quantitative evaluation from static indicators to dynamic life and the quantitative correlation between peak stall torque and fatigue life, providing a scientific basis for predicting the overall life of the robot.

[0035] 2. This invention, by embedding a piecewise excitation function and a back electromotive force compensation algorithm, removes dynamic impact interference during testing, provides clean static load data, and protects the servo driver from back pressure damage. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0037] Figure 2 This is a schematic diagram of the system structure of the present invention.

[0038] Figure 3 This is a sampling analysis diagram of the overall data of the humanoid robot of the present invention.

[0039] Figure 4 This is a schematic diagram of the torque-current mapping relationship of the present invention.

[0040] Figure 5 This is a schematic diagram of the sampling data for the excitation function of the present invention.

[0041] Figure 6 This is a schematic diagram of the peak torque-life cycles curve of the present invention.

[0042] Figure labeling: Mechanical stall tooling 10, joint module 20, servo driver 30, host computer 40. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0044] Example 1

[0045] In this embodiment, wherein, as Figure 1 - Figure 6 As shown, a quantitative testing method for the peak stall torque of a robot joint module includes the following steps:

[0046] Step S1: Sampling and analysis of the humanoid robot's overall data under typical working conditions (e.g., Figure 3 As shown), the torque time history curve of joint module 20 is extracted. Based on the material mechanics analysis, the peak stall torque is defined as a controlled low-frequency fatigue loading process, and a mapping relationship model between torque amplitude and failure cycle number is established.

[0047] Step S2: Establish a test system including a mechanical stall fixture 10, a joint module 20, a servo driver 30, and a host computer 40 (e.g., Figure 2 As shown), perform torque-current (TI) mapping (as shown). Figure 4 The static calibration (as shown) ensures that the servo driver current loop control can reflect the output torque;

[0048] Step S3: Embed a specific excitation function (such as...) in the servo driver current loop interrupt. Figure 5As shown), the total test period is set to T. A single loading of a specific excitation function includes a preloading phase and a peak loading phase, and symmetrical reverse loading is performed at time point T / 2.

[0049] Step S4: Monitor the bus voltage and speed change rate in real time. At the moment of torque unloading, use the reverse electromotive force compensation algorithm to actively adjust the duty cycle and open the energy dissipation circuit to offset the instantaneous back pressure and prevent the servo driver 30 overvoltage protection and hardware damage.

[0050] Step S5: Repeat step S3. After multiple cycles, structural damage will occur in the joint module 20 (such as broken teeth or deformation exceeding tolerance). Record the number of cycles when structural damage occurs in the joint module 20, and plot the peak torque-life cycle curve of the joint module 20 (e.g., ...). Figure 6 As shown in the figure, it serves as the input for predicting the overall lifespan of the humanoid robot;

[0051] In this embodiment, the peak torque-life cycles curve (e.g.) Figure 6 The generation process (as shown) is as follows: Apply peak torques of different amplitudes to multiple joint modules 20 of the same model and perform repeated loading tests until structural failure occurs. Record the failure cycle number of each sample and obtain the logarithmic linear relationship between torque amplitude and cycle number by least squares fitting, i.e., fatigue life curve. This curve is used as input data for predicting the overall life of the machine.

[0052] In this embodiment, the present invention realizes the scientific transformation from static performance indicators to dynamic life assessment by constructing a mapping relationship model between torque amplitude and failure cycle number, filling the technical gap in life prediction of humanoid robot joint modules under extreme loads.

[0053] Among them, such as Figure 1 As shown, step S1 specifically includes: training the humanoid robot's agent under typical working conditions through a reinforcement learning model, and deploying it to the entire humanoid robot to collect and analyze data; storing torque time history data and generating torque time history curves based on lightweight communication middleware (data layer communication protocol LCM); defining the peak stall torque as a controlled low-frequency fatigue loading process based on material mechanics analysis; establishing a uniaxial strain amplitude-life relationship using the Manson-Coffin equation; constructing a mapping relationship model between torque amplitude and failure cycle number (Cycles); the input is the load spectrum, and the output is a mapping diagram of torque amplitude and failure cycle number; the Manson-Coffin equation is a classic fatigue model describing the strain-life relationship of materials under cyclic loading.

[0054] Specifically, the reinforcement learning model (RL) can be either the MIMIC model or the AMP model. Typical scenarios include jumping and fall recovery. The MIMIC model is an imitation learning model used for robot motion imitation, while the AMP model is an adversarial motion prior model used to generate natural motion trajectories.

[0055] Specifically, the static calibration operations in step S2 include:

[0056] The joint module 20 to be tested is installed on the test bench of the mechanical stall fixture 10. The output end of the joint module 20 to be tested is connected to the input end of the torque sensor. The output end of the torque sensor is locked. The current value of the joint module 20 and the torque value of the torque sensor are sampled at N points from 0 to the peak torque, where N is a positive integer greater than or equal to 5.

[0057] Specifically, the specific excitation function in step S3 is used to generate a precise torque command in the servo driver current loop to test the steady-state torque output and electrical characteristics of the joint module 20 in a pure stall state, providing clean static load data. The specific logic of the specific excitation function includes:

[0058] Let the peak torque be The total testing period is , The time is 10 seconds, and the torque command is... ;

[0059] The torque command expression for the preloading phase is: , ,in, For the test time, The preloading phase lasted 50ms. A linear ramp of time gradually increases the torque from 0 to... Gentle force is used to mesh gears, eliminate mechanical backlash in the transmission chain, and establish initial contact stress, thereby avoiding impact caused by gaps during subsequent step loading. Backlash is the gap between components such as gears or couplings in the transmission chain, which can lead to nonlinear impacts in torque transmission.

[0060] The torque command expression for the peak loading phase is: , ,in, For 100ms, in At that moment, the instruction came from Step to The system backlash has been eliminated in the first stage, and this step will not cause a significant kinetic energy impact. The motor quickly enters and remains in a stall state. The current and voltage data collected at this time are used to reflect the pure steady-state stall torque characteristics, eliminating the interference of dynamic inertia.

[0061] During symmetrical reverse loading (simulating alternating stress), At that time, the above-mentioned phase process is repeated, and the torque command is... To obtain reverse stall data and form a complete bidirectional load cycle;

[0062] The key to the above stages lies in decoupling dynamic impact and static response. The preloading stage is responsible for overcoming the nonlinear backlash of the system, bringing the mechanical system into a "tight" linear contact state. Subsequently, a step loading is applied on the basis of eliminated backlash. At this time, almost all the energy output by the motor is used to generate static torque (converted into contact stress between gears and bearings) rather than acceleration, thus ensuring... The purity of the stage data provides a reliable load input for accurate strain-life analysis;

[0063] In this embodiment, the loading period of the excitation function can be adjusted according to actual test requirements, with a typical value of 10 seconds. In actual testing, the time parameter can be appropriately adjusted according to the joint module specifications and response characteristics to ensure that overheating or overload is avoided while eliminating backlash.

[0064] Specifically, considering the problem of harmful back electromotive force (EMF) generated by sudden speed changes during unloading after maintaining peak stall torque, as discovered through actual measurements, this invention designs a back EMF compensation algorithm. This algorithm is used to predict speed based on encoder changes and calculate and compensate for the back EMF through a back EMF coefficient. The specific operations of the back EMF compensation algorithm include:

[0065] A sampling resistor and encoder are used to monitor the bus voltage and speed change rate in real time. The encoder signal is acquired at high frequency, filtered, and then the motor's mechanical angular velocity is estimated in real time. The instantaneous back electromotive force (EMF) estimate is calculated using the motor's back EMF constant. The expression is as follows: ,in, The instantaneous back electromotive force (EMF) estimate, specifically the estimated value of the back EMF generated when the motor rotor abruptly cuts the magnetic field at the instant of unloading from the joint module 20, is the core calculation target of the compensation algorithm. V is the back electromotive force constant of the motor, with units of V·s / rad or V / (rad / s). It represents the proportional relationship between the motor speed and the back electromotive force and is determined by the motor design specifications. Here, represents the motor's mechanical angular velocity, specifically the instantaneous mechanical rotational angular velocity of the motor rotor, measured in rad / s. The estimated instantaneous back electromotive force is directly superimposed onto the voltage command output from the current loop as a feedforward quantity. The expression is: ,in, The voltage command that the servo driver 30 ultimately outputs to the motor has been superimposed with a back electromotive force compensation amount to counteract the disturbance of the back electromotive force. The original voltage command generated by the servo driver 30 based on torque demand does not take into account the influence of back electromotive force. The back electromotive force compensation algorithm is used to actively cancel the disturbance of back electromotive force, suppress the voltage and current surge at the moment of unloading, protect the servo driver 30 and improve dynamic stability.

[0066] Specifically, in step S4, the energy dissipation circuit converts the potential energy generated by the instantaneous back pressure into heat energy by connecting the bypass resistor. The instantaneous back pressure is generated by the permanent magnet on the rotor cutting the magnetic field when it undergoes instantaneous reverse micro-motion. Instantaneous reverse micro-motion refers to the slight reverse rotation of the rotor caused by mechanical elastic rebound at the moment of unloading.

[0067] Specifically, the criteria for determining structural damage in step S5 are at least one of the following: broken teeth in the joint module 20 or deformation exceeding the preset tolerance. The preset tolerance refers to the allowable deformation range set in the design stage of the joint module 20, and the specific value is determined according to the application scenario, material properties and structural design parameters of the joint module.

[0068] like Figure 3 As shown in the figure, this is an overall data sampling analysis diagram of the humanoid robot joint module 20. By sampling and analyzing the humanoid robot under typical working conditions, the torque time-history curve is obtained. The horizontal axis of the figure is the sampling timestamp, and the vertical axis is the torque value (unit: Nm). The blue curve represents the preset command torque, and the red curve represents the actual torque fed back by the joint module 20. This figure intuitively presents the dynamic matching relationship between the command torque and the actual torque. By comparing the degree of fit between the two curves, the response accuracy and following stability of the joint module 20 to the torque command are verified, and it is determined whether there are problems such as torque output delay or excessive fluctuation in the joint module 20 during actual operation.

[0069] Figure 4 The torque-current mapping relationship diagram of the joint module 20 is drawn based on the static calibration results of step S2. The horizontal axis of the diagram is the current (unit: A), and the vertical axis is the torque value of the joint module output terminal sampled by the torque sensor (unit: Nm). This diagram establishes a quantitative correspondence between current and torque. By presenting the variation law of the joint module output torque under different currents, the actual output torque of the joint module 20 can be calculated by collecting current signals and combining the mapping relationship of this diagram during the test.

[0070] Figure 5This is a test sampling data graph of the piecewise waveform excitation function. The horizontal axis of the graph is the timestamp (in seconds), and the vertical axis is the torque value (in Nm). The graph shows a complete cycle of the excitation function consisting of two positive and negative square waves. The positive and negative square waves correspond to the forward and reverse "preloading-peak loading" processes, respectively. It clearly shows the complete waveform of the torque linearly increasing from 0 to 10% of the peak torque (preloading stage), stepping to the peak torque and holding it (peak loading stage), and then repeating the process in reverse. This graph verifies that the excitation function outputs accurately according to the preset logic, confirms that the preloading stage can effectively eliminate backlash, and the peak loading stage can stably maintain the torque. It ensures that the piecewise loading method can eliminate inertial impulse interference and provide clean data for subsequent fatigue analysis.

[0071] Figure 6 This is a peak torque-life cycles graph. The horizontal axis represents life cycles (Cycles) on a logarithmic scale (1000-1000000), and the vertical axis represents the peak torque value (unit: Nm). This curve serves as the input for predicting the overall lifespan of the humanoid robot, realizing a closed-loop process from data acquisition to lifespan assessment. The data points in this graph represent "the number of cycles on the horizontal axis when the joint module experiences fatigue failure after repeatedly loading the corresponding vertical axis torque value," presenting the correlation between peak torque and joint module fatigue life. This facilitates the direct determination of the sustainable working number of the joint module under different peak torques. For example, the number of cycles corresponding to a certain peak torque can serve as a scientific basis for setting the robot maintenance cycle.

[0072] Therefore, according to Figures 3 to 6 It can be seen that the present invention can support a quantitative testing process from working condition analysis to life prediction through experimental data, system calibration, excitation function execution and result presentation, and has the feasibility of implementation.

[0073] Example 2

[0074] The difference from Example 1 is that, as in Example 1, Figure 2 As shown, this invention provides a quantitative testing system for the peak stall torque of a robot joint module, used to perform a quantitative testing method for the peak stall torque of a robot joint module, including:

[0075] Mechanical stall tooling 10 serves as a testing device for fixing joint module 20 and realizing stall function. Mechanical stall tooling 10 is a double flange structure.

[0076] The joint module 20, as the device under test, is used to execute the active degree of freedom mechanism of the humanoid robot. The active degree of freedom mechanism refers to the key mechanism part of the humanoid robot that has the power drive capability and can actively realize posture adjustment or motion output. The joint module 20 provides power, rather than relying on external force or passively following the movement. The degree of freedom refers to the dimension in which the mechanism can move independently (such as rotation and translation). For example, the shoulder joint of the humanoid robot has 3 degrees of freedom (pitch, yaw, roll). Each degree of freedom corresponds to an active drive mechanism. The active degree of freedom mechanism includes, but is not limited to, the shoulder joint, elbow joint, wrist joint, hip joint, knee joint, ankle joint, waist rotation mechanism, etc.

[0077] The servo driver 30 (device under test controller) serves as the electrical control board for the joint module 20. Its inputs are DC power and communication commands. It controls the servo movement of the joint module 20 through the servo program within the electrical control board.

[0078] The host computer 40 (data processing program) is connected to the joint module 20 via CAN communication. It is used to store torque time history data, send control commands to the joint module 20 and receive current feedback signals to complete life quantification assessment.

[0079] A torque sensor, connected to the output terminal of joint module 20, is used for torque-current calibration and to collect the output torque value of joint module 20.

[0080] Sampling resistor, used for real-time monitoring of bus voltage;

[0081] The encoder is used to monitor the rate of change of motor speed in real time and participate in back electromotive force compensation.

[0082] Specifically, one set of flanges in the dual-flange structure is used to lock the fixed flange of the joint module 20, and the other set of flanges is used to lock the output flange of the joint module 20. The dual-flange structure is made of high-strength aluminum alloy material, and its flange spacing is adjustable to adapt to the installation of modules of different sizes. The mechanical stall tooling 10 is fixed to the test bench by bolts to ensure no displacement or deformation during peak loading, thus ensuring the repeatability and accuracy of the test.

[0083] In summary, this invention, through quantitative testing methods and systems, enables the scientific evaluation of peak stall torque and fatigue life of robot joint modules. It possesses high stability, high precision, and strong practicality, and is suitable for life prediction and reliability design of high-performance joint modules such as humanoid robots.

[0084] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A quantitative testing method for the peak stall torque of a robot joint module, characterized in that, Specifically, the following steps are included: Step S1: By sampling and analyzing the whole machine data of the humanoid robot under typical working conditions, the torque time history curve of the joint module is extracted. Based on the material mechanics analysis, the peak stall torque is defined as a controlled low-frequency fatigue loading process, and a mapping relationship model between torque amplitude and failure cycle number is established. Step S2: Establish a test system including mechanical stall tooling, joint module, servo driver and host computer, and perform static calibration of torque-current mapping relationship to ensure that the servo driver current loop control can reflect the output torque. Step S3: Embed a specific excitation function in the servo driver current loop interrupt, set the total test period to T, and the specific excitation function includes a pre-load stage and a peak load stage in a single loading, and performs symmetrical reverse loading at time point T / 2. Step S4: Monitor the bus voltage and speed change rate in real time. At the moment of torque unloading, use the reverse electromotive force compensation algorithm to actively adjust the duty cycle and open the energy dissipation circuit to offset the instantaneous back pressure and prevent the servo driver from being over-voltage protected and the hardware from being damaged. Step S5: Repeat step S3, record the number of cycles when the joint module suffers structural damage, and plot the peak torque-life cycle curve of the joint module as the input for predicting the overall lifespan of the humanoid robot. In step S3, the specific excitation function is used to generate a precision torque command in the servo driver current loop to test the steady-state torque output and electrical characteristics of the joint module in a pure stall state, providing clean static load data. The specific logic of the specific excitation function includes: Let the peak torque be The total testing period is , The time is 10 seconds, and the torque command is... ; The torque command expression for the preloading phase is: , ,in, For the test time, The preloading phase lasts 50ms. A linear ramp of time gradually increases the torque from 0 to... It uses gentle force to mesh gears and eliminate mechanical backlash in the transmission chain; The torque command expression for the peak loading stage is: , ,in, For 100ms, in At that moment, the instruction came from Step to The motor quickly enters and remains in a stall state for 100ms. The current and voltage data collected at this time are used to reflect the pure steady-state stall torque characteristics. During the symmetrical reverse loading process, At that time, the above-mentioned phase process is repeated, and the torque command is... This is to obtain reverse stall data and form a complete bidirectional load cycle.

2. The quantitative testing method for peak stall torque of a robot joint module according to claim 1, characterized in that, Step S1 specifically includes: training the humanoid robot's agent under typical working conditions through a reinforcement learning model, deploying it on the entire humanoid robot to collect and analyze data, storing torque time history data and generating torque time history curves based on lightweight communication middleware, defining the peak stall torque as a controlled low-frequency fatigue loading process based on material mechanics analysis, establishing a uniaxial strain amplitude-life relationship using the Manson-Coffin equation, constructing a mapping relationship model between torque amplitude and failure cycle number, with the load spectrum as the input and the mapping diagram between torque amplitude and failure cycle number as the output.

3. The quantitative testing method for peak stall torque of a robot joint module according to claim 2, characterized in that, The reinforcement learning model is either the MIMIC model or the AMP model, and the typical working conditions include jumping and falling recovery.

4. The quantitative testing method for peak stall torque of a robot joint module according to claim 1, characterized in that, The specific operations for static calibration in step S2 include: The joint module to be tested is installed on the test bench of the mechanical stall fixture. The output end of the joint module to be tested is connected to the input end of the torque sensor. The output end of the torque sensor is locked. The current value of the joint module and the torque value of the torque sensor are sampled at N points from 0 to the peak torque, where N is a positive integer greater than or equal to 5.

5. The quantitative testing method for peak stall torque of a robot joint module according to claim 1, characterized in that, The back EMF compensation algorithm in step S4 is used to predict speed based on encoder changes, and to calculate and compensate for the back EMF through the back EMF coefficient. The specific operations of the back EMF compensation algorithm include: A sampling resistor and encoder are used to monitor the bus voltage and speed change rate in real time. The encoder signal is acquired at high frequency, filtered, and then the motor's mechanical angular velocity is estimated in real time. The instantaneous back electromotive force (EMF) estimate is calculated using the motor's back EMF constant. The expression is as follows: ,in, For the instantaneous back electromotive force estimation, Let be the back electromotive force constant of the motor. Given the motor's mechanical angular velocity, the instantaneous back electromotive force estimate is directly superimposed onto the voltage command output by the current loop as a feedforward quantity. The expression is: ,in, This is the voltage command that the servo driver ultimately outputs to the motor. The back EMF compensation algorithm is used to actively counteract the back EMF disturbance, suppress the voltage and current surges during unloading, protect the servo driver, and improve dynamic stability.

6. The quantitative testing method for peak stall torque of a robot joint module according to claim 5, characterized in that, In step S4, the energy dissipation circuit converts the potential energy generated by the instantaneous reverse pressure into heat energy by connecting the bypass resistor. The instantaneous reverse pressure is generated by the permanent magnet on the rotor cutting the magnetic field when it makes a slight reverse jog.

7. The quantitative testing method for peak stall torque of a robot joint module according to claim 1, characterized in that, The criteria for determining structural damage in step S5 are at least one of the following: broken teeth in the joint module or deformation exceeding the preset tolerance.

8. A quantitative testing system for the peak stall torque of a robot joint module, used to execute the quantitative testing method for the peak stall torque of a robot joint module as described in any one of claims 1 to 7, characterized in that, include: A mechanical stall tooling, used as a testing device to fix the joint module and realize the stall function, is a double flange structure; Joint module, as a device under test, is used to execute mechanisms for active degrees of freedom in a humanoid robot; The servo driver, as the electronic control board of the joint module, takes DC power and communication commands as inputs and controls the servo movement of the joint module through the servo program in the electronic control board. The host computer is connected to the joint module via CAN communication. It is used to store torque time history data, send control commands to the joint module, and receive current feedback signals to complete the life quantification assessment.

9. A quantitative testing system for the peak stall torque of a robot joint module according to claim 8, characterized in that, One set of flanges in the dual-flange structure is used to lock the fixed flange of the joint module, and the other set of flanges is used to lock the output flange of the joint module.

Images