Industrial robot servo system digital-physical collaborative service reliability testing platform and method

Abstract

This invention discloses a digital-physical collaborative service reliability testing platform and method for industrial robot servo systems. It includes a digital layer prototype and a physical layer prototype of the industrial robot. The digital layer prototype is a digital twin control model of the robot constructed by collecting load data during the robot's operation, establishing a load state matrix, and associating it with the robot's body stiffness model, joint friction model, and end effector model. The physical layer prototype contains several test channels, each corresponding to a joint of the robot. The servo systems of each joint are installed within the test channels to monitor and test the performance degradation of the servo systems under service conditions. This invention establishes functional units around the digital and physical prototypes, including test control, service loading, test instruments and sensors, and reliability analysis, to realize the reliability testing and evaluation of industrial robot servo systems under service conditions.

Description

Technical Field

[0001] This invention relates to the field of industrial robot technology, specifically to a reliability testing platform and method for the collaborative service of digital and physical components in industrial robot servo systems. Background Technology

[0002] As the core power execution and control unit of an industrial robot, the performance of the servo system directly affects the movement speed, positioning accuracy, load-bearing capacity, and operational reliability of the entire robot system. In practical applications, the servo system must withstand mechanical loads and environmental stresses under complex working conditions for extended periods. Therefore, accurate assessment of its service reliability is crucial for ensuring the stable operation of the robot system and extending its service life.

[0003] Currently, most reliability testing methods for industrial robot servo systems are based on stable experimental conditions. The reliability data obtained deviates significantly from the degradation and evolution behavior in actual dynamic operating environments, failing to accurately reflect performance changes during service and thus hindering effective support for system design, lifespan prediction, and maintenance decisions. Furthermore, some methods attempt to directly install the servo system onto the robot body for in-service testing. However, the complex structure and limited internal space of industrial robots severely restrict the placement of key test sensors and measuring instruments, resulting in limited testing dimensions, high testing costs, low testing efficiency, and difficulty in achieving comprehensive degradation monitoring under multimodal service conditions. For example, patent ZL202110354614.X proposes a workload simulation testing system and method for industrial robot servo systems. Its core lies in implementing mechanical stress loading and operational status detection for the servo system. However, this solution does not cover stress loading methods in the service environment, nor does it establish a service response control model for the servo system or a reliability modeling mechanism related to system degradation, resulting in insufficient depth and breadth in the evaluation. For example, patent ZL202210239999.X discloses a reliability testing method based on the overall structure of an industrial robot. While it achieves performance evaluation under varying load conditions to some extent, it still faces challenges such as difficulty in independently monitoring the degradation of the servo system and insufficient granularity in reliability analysis, as it primarily focuses on the entire robot. Furthermore, the digital twin construction method proposed in CN202411488766.9 does not focus on the actual service degradation behavior of industrial robot servo systems. Its model is mainly based on virtual simulation and parametric estimation, lacking support from physical test data, resulting in limited accuracy in reliability assessment. Summary of the Invention

[0004] The purpose of this invention is to provide a reliability testing platform and method for the collaborative service of industrial robot servo systems, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a reliability testing platform for the digital-physical collaborative service of an industrial robot servo system, comprising:

[0006] The industrial robot digital layer prototype is a robot digital twin control model. It is constructed by collecting load data during the operation of the industrial robot, establishing a load state matrix, and associating the robot body stiffness model, joint friction model, and end effector association model.

[0007] An industrial robot physical layer prototype includes several test channels, each test channel corresponding to a joint of the industrial robot. A servo system under test is set in the test channel. The industrial robot physical layer prototype realizes the simulation of the servo system in the service motion state of the industrial robot by linking the digital and physical information with the industrial robot digital layer prototype.

[0008] Service loading module, which is used to simulate and output the mechanical and environmental conditions of industrial robot service, and realize high-fidelity simulation output of service load of industrial robot servo system;

[0009] Test instruments and sensors are installed in each test channel of the physical layer prototype of the industrial robot to collect multimodal degradation monitoring data;

[0010] The reliability analysis unit is used to screen and extract degradation features from the degradation data collected by the test instruments and sensors, establish a degradation model based on the β_Wiener process, and analyze and output the service reliability assessment results of the industrial robot servo system.

[0011] The central control unit is used to control the test platform, including user interaction, test process control, data-object information linkage, and test data acquisition for the reliability testing of the industrial robot servo system.

[0012] Preferably, the servo system under test includes:

[0013] Servo motors, specifically corresponding to robot joints, are installed in the test channel;

[0014] An industrial robot servo driver, which receives control signals from an industrial robot servo controller and drives a servo motor to operate.

[0015] An encoder is mounted on a servo motor. The encoder is used to collect angular displacement signals of the servo motor and transmit them as feedback signals to the servo controller to realize the position control of the servo motor.

[0016] Preferably, the industrial robot digital layer prototype is a robot digital twin control model, and its construction steps are as follows:

[0017] S1. Collect real-time load data of the industrial robot during operation and construct a load state matrix that changes over time. The load state matrix includes multi-dimensional load characteristic parameters.

[0018] S2. Based on the load state matrix and combined with the industrial robot structural dynamics model, construct an industrial robot joint dynamic twin control model that is adaptive to the working conditions.

[0019] S3. Based on the dynamic twin control model of industrial robot joints, the output torque parameters of the servo driver are adjusted in real time to match the actual load changes and simulate the control response state of the servo system inside the industrial robot body.

[0020] Preferably, each test channel of the industrial robot physical layer prototype is used to perform service reliability testing of a certain joint servo system of the industrial robot. The servo motor is set in the test channel and is connected to the harmonic reducer through a first drive shaft. The test instruments and sensors include a torque sensor. The harmonic reducer is connected to one end of the torque sensor through a second drive shaft and a first coupling. The other end of the torque sensor is connected to the service loading module through a third drive shaft and a second coupling.

[0021] Preferably, the service loading module includes an environmental stress loading system and a mechanical stress loading system.

[0022] The environmental stress loading system includes a temperature and humidity control box and a vibration table, which are used to simulate the ambient temperature, humidity and vibration conditions of the industrial robot servo system during operation.

[0023] The mechanical stress loading system includes:

[0024] The service load simulation calculation and output module comprises two parts: a speed and acceleration estimation program for the industrial robot servo system and a service load calculation program. The speed and acceleration estimation program uses the angular displacement signal obtained from the encoder of the servo system under test, applies a tracking differentiator to estimate its angular velocity and angular acceleration values, and obtains the joint angular displacement, joint angular velocity, and joint angular acceleration of the servo system based on the transmission coefficient of the harmonic reducer, inputting these values ​​into the service load calculation program. The service load calculation program, based on the joint angular displacement, joint angular velocity, and joint angular acceleration output by the speed and acceleration estimation program, calculates the service load of the industrial robot servo system using the robot inverse dynamics equations and transmits it to the loading system controller to realize the mechanical stress loading of the industrial robot servo system during service.

[0025] The loading mechanism is a torque servo system, including a driver and a torque motor. The driver receives the current service load result output by the service load simulation calculation and output module to generate a drive signal, and the drive motor outputs the corresponding service loading torque.

[0026] A load controller is used to generate a drive control signal based on the difference between the output torque of the torque motor and the target torque calculated by the driver receiving the service load simulation and output by the output module, so as to realize closed-loop control of the load output of the torque servo system.

[0027] Another aspect of this invention discloses a method for testing the reliability of digital-physical collaborative service of an industrial robot servo system, including:

[0028] S10. Input the structural parameters, service environment parameters, mission trajectory parameters, and reliability test control parameters of the servo system under test into the interactive interface of the central control unit.

[0029] S20. Collect real-time load data during the operation of the industrial robot, construct a load state matrix that changes over time, and establish a digital layer prototype system by combining the structural dynamics model of the industrial robot.

[0030] S30. Install the servo system under test on the physical layer prototype system and conduct service reliability testing. The servo system under test runs according to the robot's task trajectory and outputs service feedback status in conjunction with the digital layer prototype system.

[0031] S40, the test platform environmental stress loading system simulates the service environment conditions of the servo system, and the mechanical stress loading system outputs the current service load according to the robot inverse dynamics model, so that the servo system enters the service state.

[0032] S50: The central control unit collects control parameters based on the test data input by the user, controls the data collection of each test instrument and sensor during the reliability test, and transmits the data to the reliability analysis unit.

[0033] The S60 reliability analysis unit data feature processing module uses a regularized restricted Boltzmann machine algorithm to perform unsupervised feature extraction on the collected degradation data. The extracted features are then transmitted to the reliability modeling and analysis module to establish a nonlinear β_Wiener process model, analyze and output the service reliability assessment results.

[0034] Preferably, in step S20, the real-time load data includes force data of the end effector, joint torque data, speed information and acceleration information, and the accuracy of load status perception is improved by multi-source sensor fusion technology.

[0035] Preferably, in step S20, the load state matrix is ​​dynamically refreshed according to a preset time window, a sliding time window mechanism is used to obtain the continuous load change trend, and the trend is input into the dynamic torque control model in a recursive form to realize time-series feature modeling.

[0036] Preferably, in step S20, the structural dynamics model includes a robot body stiffness model, a joint friction model, and an end effector inertia model. The dynamics model is dynamically calibrated using a parameter identification method to improve its matching with actual working conditions.

[0037] Preferably, in step S60, the regularized restricted Boltzmann machine algorithm introduces a regularization term into the objective function to improve the ability to extract degenerate features; and / or

[0038] In S60, the nonlinear β_Wiener process model introduces a β-distributed random effect into the drift term to characterize the individual differences and uncertainties in the degradation process of the servo system.

[0039] As can be seen from the above technical solution, the present invention has the following beneficial effects:

[0040] This digital-physical collaborative service reliability testing platform and method for industrial robot servo systems establishes a digital-physical prototype of the industrial robot and develops a digital-physical collaborative platform to achieve high-fidelity simulation of the servo system's service state on the industrial robot. Based on the digital-physical collaborative platform, servo system reliability testing removes the limitations imposed by the industrial robot's structure on servo system degradation monitoring, enabling multi-dimensional degradation monitoring of the industrial robot servo system and more comprehensively capturing its degradation status, thus laying the foundation for effective reliability assessment. The digital-physical collaborative testing platform and method significantly reduce the cost of deploying servo systems on industrial robots for service reliability testing and improve testing efficiency, possessing significant engineering application value. Attached Figure Description

[0041] Figure 1 A schematic diagram of the framework for a reliability testing platform for the collaborative service of digital and physical components in an industrial robot servo system.

[0042] Figure 2 A schematic diagram of the physical layer prototype test channel for a reliability testing platform for the digital-physical collaborative service of an industrial robot servo system.

[0043] Figure 3 A flowchart for the reliability testing method of digital-physical collaborative service of industrial robot servo systems. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] An industrial robot servo system digital-physical collaborative service reliability testing platform is used to implement reliability testing methods for industrial robot servo systems' digital-physical collaborative service, such as... Figure 1 As shown, it includes:

[0046] Industrial robot digital layer prototype 2 is a robot digital twin control model. It is constructed by collecting load data during the operation of the industrial robot, establishing a load state matrix, and associating the robot body stiffness model, joint friction model, and end effector association model.

[0047] The industrial robot physical layer prototype 3 includes several test channels, each test channel corresponding to a joint of the industrial robot. The test channel is equipped with a servo system 3.1 under test. The industrial robot physical layer prototype 3 realizes the simulation of the motion state of the servo system 3.1 in the service of the industrial robot by linking the digital and physical information with the industrial robot digital layer prototype 2.

[0048] Service loading module 4 is used to simulate and output the mechanical and environmental conditions of the industrial robot during service, so as to realize the high-fidelity simulation output of the service load of the industrial robot servo system.

[0049] Test instruments and sensors 5 are deployed in each test channel of the physical layer prototype of the industrial robot to collect multimodal degradation monitoring data.

[0050] Among them, the testing instruments and sensors 5 include torque sensors, power meters, power meters, temperature sensors, resistance testers, high voltage testers, and turn-to-turn testers, which enable multi-dimensional degradation monitoring of industrial robot servo systems.

[0051] The reliability analysis unit 6 is used to screen and extract degradation features from the degradation data collected by the test instrument and sensor 5, establish a degradation model based on the β_Wiener process, analyze and output the service reliability assessment results of the industrial robot servo system.

[0052] The central control unit 1 is used to control the test platform, including user interaction, test process control, data-object information linkage, and test data acquisition for the reliability test of the industrial robot servo system.

[0053] The main control unit 1 mainly consists of an industrial control computer and a PLC control system, such as... Figure 1 As shown, it is connected to the industrial robot digital layer prototype 2, physical layer prototype 3, and service loading module 4 for control.

[0054] This invention establishes a digital-physical prototype of an industrial robot (including a digital layer prototype 2 and a physical layer prototype 3 and their information interaction and linkage), and develops a digital-physical collaborative platform to achieve high-fidelity simulation of the servo system's internal state within the industrial robot in service. Based on the digital-physical collaborative platform, servo system reliability testing removes the limitations imposed by the robot's structure on servo system degradation monitoring, enabling multi-dimensional degradation monitoring of the industrial robot servo system and more comprehensively capturing its degradation status, thus laying the foundation for effective reliability assessment. The digital-physical collaborative testing platform and method significantly reduce the cost of deploying servo systems on industrial robots for in-service reliability testing and improve testing efficiency, possessing significant engineering application value.

[0055] In some embodiments, the servo system under test 3.1 includes:

[0056] Servo motor 3.1.1, specifically corresponding to a robot joint, is located in the test channel;

[0057] The industrial robot servo driver 3.1.3 is used to receive control signals from the industrial robot servo controller and drive the servo motor 3.1.1 to operate.

[0058] Encoder 3.1.2 is mounted on servo motor 3.1.1. Encoder 3.1.2 is used to collect the angular displacement signal of servo motor 3.1.1 and transmit it as a feedback signal to servo driver 3.1.3 to realize position control of servo motor 3.1.1.

[0059] In some embodiments, the industrial robot digital layer prototype is a robot digital twin control model, and its construction steps are as follows:

[0060] S1. Collect real-time load data of the industrial robot during operation and construct a load state matrix that changes over time. The load state matrix includes multi-dimensional load characteristic parameters.

[0061] S2. Based on the load state matrix and combined with the industrial robot structural dynamics model, construct an industrial robot joint dynamic twin control model that is adaptive to the working conditions.

[0062] S3. Based on the dynamic twin control model of industrial robot joints, the output torque parameters of the servo driver are adjusted in real time to match the actual load changes and simulate the control response state of the servo system inside the industrial robot body.

[0063] In S1, strain gauge sensors are installed along the force transmission paths of the industrial robot's joints or actuators to collect load data in real time during its operation. This load data includes multiple dimensions such as time, joint position, applied torque, velocity, acceleration, and temperature. The collected data is recorded at fixed time intervals, and the data at each time point is combined to form a set of multi-dimensional load characteristic parameters. For example, data is recorded every 100 milliseconds within one second, resulting in 10 sets of data per second. As time progresses, all recorded data is organized in a matrix to form a load state matrix, where the rows represent the time series and the columns represent the various load characteristic parameters.

[0064] In S2, based on the load state matrix constructed above, a dynamic twin control model of the industrial robot joints for the servo driver 3.1.3 is established in conjunction with the dynamic model of the industrial robot structure. The industrial robot structure dynamic model was determined during robot design through stiffness modeling and inertial measurement, and verified through static and dynamic experiments. In the dynamic twin control model, the theoretical driving torque value that each drive joint should bear is first calculated at each time point using Newton's second law by combining the position and acceleration data in the load state matrix with the mass distribution. This theoretical torque is compared with the actual torque value measured by the torque sensor 3.6, and the error is calculated and fed back to adjust the output target of the driver 3.1.3 in the control model. The dynamic twin control model has an adaptive function, mainly reflected in the system automatically identifying the load fluctuation pattern in each time period and automatically adjusting the control parameters such as the gain coefficient. The initial values ​​of the control parameters are derived from the preset response time and target accuracy requirements during the system calibration process. For example, the response time of the servo driver 3.1.3 is set to 50 milliseconds, and the position accuracy error threshold is set to 0.01 millimeters. These values ​​are obtained through robot design standards and working condition simulation experiments.

[0065] In S3, the target output torque value calculated by the robot joint dynamic twin control model is directly input to the control module of the servo driver 3.1.3. This module adjusts the actual output current to precisely control the torque. This process is refreshed cyclically every 10 milliseconds to ensure the high-fidelity output of the control model to the servo system's response to fluctuations in the service load. For example, if the current service load of the mechanical stress loading system 4.1 increases by 5 Nm, the joint dynamic twin control model will adjust the current increase in the next cycle according to the load change, so that the driver outputs a servo system with a response torque equivalent to that of the robot body, allowing the servo system 3.1 under test on the digital-physical collaborative test platform to operate in the service state inside the robot body.

[0066] In summary, by collecting and processing load data during the actual operation of industrial robots, a load state matrix with strong time-series characteristics is constructed. Combined with dynamic analysis of the actual structural model, a digital twin control model matching the load response changes of the robot's internal servo system is formed. Through precise output control and a fast error compensation mechanism, the digital layer prototype based on the digital twin control model maintains consistency with the robot's current state throughout the entire service test process. This effectively improves the matching between the servo system's feedback on the service load response and the feedback of the servo system under actual robot operating conditions on the digital-physical collaborative testing platform, laying the foundation for accurate evaluation of the service reliability of industrial robot servo systems.

[0067] In some embodiments, each test channel of the physical layer prototype 3 is used to perform service reliability testing of a certain joint servo system of an industrial robot. The servo motor 3.1.1 is set in the test channel. The servo motor 3.1.1 is connected to the harmonic reducer 3.3 through the first drive shaft 3.2. The harmonic reducer 3.3 is connected to one end of the torque sensor 3.6 through the second drive shaft 3.4 and the first coupling 3.5. The other end of the torque sensor 3.6 is connected to the service loading module 4 through the third drive shaft 3.7 and the second coupling 3.8.

[0068] The industrial robot physical layer prototype 3 includes several test channels, and a schematic diagram of a single test channel is shown below. Figure 2 As shown, each test channel consists of: a servo system under test 3.1, a first drive shaft 3.2, a harmonic reducer 3.3, a second drive shaft 3.4 and a first coupling 3.5, a torque sensor 3.6, a second drive shaft 3.7 and a second coupling 3.8, a service mechanical stress loading system 4.1, a temperature and humidity control box 4.2.1, and a vibration table 4.2.2.

[0069] In some embodiments, the service loading module 4 includes an environmental stress loading system and a mechanical stress loading system 4.1.

[0070] The environmental stress loading system includes a temperature and humidity control box 4.2.1 and a vibration table 4.2.2, which are used to simulate the ambient temperature, humidity and vibration conditions of the industrial robot servo system during operation.

[0071] The mechanical stress loading system 4.1, as described Figure 2 As shown, it includes:

[0072] The service load simulation calculation and output module 4.1.1 is used to simulate and output the service load when a servo system placed inside an industrial robot completes a specific task trajectory. The service load simulation calculation and output module 4.1.1 includes two parts: a speed and acceleration estimation program for the industrial robot servo system and a service load calculation program.

[0073] The industrial robot servo system speed and acceleration estimation program uses the angular displacement signal obtained from the encoder 3.1.2 of the servo system under test (3.1) to estimate its angular velocity and angular acceleration values ​​using a tracking differentiator. Based on the transmission coefficient of the harmonic reducer 3.3, it obtains the servo system joint angular displacement, joint angular velocity, and joint angular acceleration, and inputs these values ​​into the service load calculation program. The joint angular displacement, joint angular velocity, and joint angular acceleration are obtained through the following transmission relationship:

[0074]

[0075] In the formula, These are the angular displacement, angular velocity, and angular acceleration of the i-th joint of the industrial robot, respectively. The reduction ratio and transmission efficiency of the i-th joint harmonic reducer are 3.3, respectively. Let represent the angular displacement, angular velocity, and angular acceleration of the servo system at the i-th joint, respectively. Data obtained from servo motor 3.1.1 encoder 3.1.2. This is obtained through the tracking differentiator in the velocity and acceleration estimation program;

[0076] The service load calculation program, based on the joint angular displacement, joint angular velocity, and joint angular acceleration output by the velocity and acceleration estimation program, calculates the service load of the industrial robot servo system using the robot inverse dynamics equations and transmits it to the loading system controller to realize the mechanical stress loading of the industrial robot servo system during service. The general formula of the robot inverse dynamics equations is as follows:

[0077]

[0078] In the formula, For the joint torque service load of industrial robots, These are the mass matrix, Coriolis force and centrifugal force matrices, and gravity term, respectively. Let be the generalized angular displacement, angular velocity, and angular acceleration vectors of the i-th joint, respectively. This indicates the relevant parameters.

[0079] Loading mechanism 4.1.3, such as Figure 2 As shown, the loading mechanism 4.1.3 is a torque servo system, including a driver and a torque motor. The driver receives the current service load result output by the service load simulation calculation and output module 4.1.1 to generate a drive signal, and the drive motor outputs the corresponding service loading torque.

[0080] Load controller 4.1.2 is used to generate drive control signals based on the difference between the output torque of the torque motor and the target torque output by the service load simulation calculation and output module 4.1.1 received by the driver, so as to realize closed-loop control of the load output of the torque servo system.

[0081] In some embodiments, the reliability analysis unit 6 includes a data feature processing module and a reliability modeling and analysis module. The data feature processing module employs a regularized restricted Boltzmann machine (RBM) algorithm to perform unsupervised feature extraction on the collected degradation monitoring data. The RBM consists of a visible layer and a hidden layer, and its energy function can be expressed as:

[0082]

[0083] In the formula, v and h are the state vectors of the visible layer and hidden layer, respectively, and n v , n h These represent the number of units in the visible and hidden layers, respectively. i h is the state value of the i-th visible cell. j Let a be the state value of the j-th hidden unit. i Let b be the Gaussian mean of the i-th visible cell. j Let σ be the Gaussian mean of the j-th hidden unit. i Let σ be the standard deviation of the i-th visible cell. j Let ω be the standard deviation of the j-th hidden unit. ij It is the connection weight between the i-th visible unit and the j-th hidden unit.

[0084] The regularized restricted Boltzmann machine algorithm adds a regularization term to the objective function of the original restricted Boltzmann machine algorithm, improving the algorithm model's ability to extract degradation trend features of industrial robot servo systems. Its objective function expression is as follows:

[0085]

[0086]

[0087]

[0088] In the formula, These are the regularization objective function, the log-likelihood function that maximizes the test dataset, and the regularization term, respectively. t is the regularization constant; N is the total number of training samples; k This indicates the k-th running cycle; This represents the average value over the total operating cycle. This represents the state value of the j-th hidden unit in the k-th running cycle; This represents the average state of the j-th hidden unit in the total running cycle.

[0089] The multidimensional degradation signal of the industrial robot servo system, collected by the testing instrument and sensor 5, is processed by normalization, missing value removal, and feature selection. It is then used as the visible layer state vector v to input the regularized restricted Boltzmann machine network. After unsupervised learning of the network, a degradation factor sequence reflecting the degradation state of the industrial robot servo system is obtained.

[0090] The reliability modeling and analysis module, based on the degradation factor sequence of the industrial robot servo system output by the regularized restricted Boltzmann machine, establishes a degradation model using a nonlinear Wiener process with β-distributed random effects to conduct reliability analysis. Specifically, let the eigenvalue of the i-th degradation factor of the servo system at time t be... Therefore, according to the nonlinear β_Wiener process model, we have:

[0091]

[0092] In the formula, Assuming the performance parameters are in their initial state, without loss of generality, we can assume... ; Let be the drift coefficient, characterizing the rate of parameter degradation; a random effect of the β distribution is introduced into the drift coefficient, therefore we have

[0093]

[0094] It is a nonlinear function with respect to time t, characterizing the nonlinear features of parameter degradation; The diffusion coefficient characterizes the influence of internal structure and external environment on the degradation of servo system parameters; It represents standard Brownian motion, characterizing the uncertainty of the parameter degradation process.

[0095] The degradation of industrial robot servo systems increases with service life, and a degradation threshold is set at... ,but Failure time exceeding the threshold Represented as:

[0096]

[0097] It follows an inverse Gaussian distribution, from which its failure probability density function can be obtained as:

[0098]

[0099] The next step is based on the failure probability density. Other reliability metrics, such as reliability, are derived. failure rate The system outputs reliability assessment results and completes the service reliability assessment of the industrial robot servo system.

[0100] In some embodiments, such as Figure 3 As shown, another aspect of the present invention provides a method for testing the reliability of digital-physical collaborative service of an industrial robot servo system, the method comprising:

[0101] S10. Input the structural parameters, service environment parameters, mission trajectory parameters, and reliability test control parameters of the servo system under test through the interactive interface of the main control unit 1.

[0102] S20. Collect real-time load data during the operation of the industrial robot, construct a load state matrix that changes over time, and establish the digital layer prototype 2 system by combining the structural dynamics model of the industrial robot.

[0103] S30. Install the servo system under test 3.1 on the physical layer prototype 3 system and carry out service reliability testing. The servo system under test runs according to the robot task trajectory and outputs the service load response status in conjunction with the digital layer prototype 2 system.

[0104] S40, the test platform environmental stress loading system simulates the service environment conditions of the servo system, and the mechanical stress loading system 4.1 outputs the current service load according to the robot inverse dynamics model, so that the servo system 3.1 enters the service state;

[0105] S50: The central control unit collects control parameters based on the test data input by the user, controls the data collection of each test instrument and sensor 5 during the reliability test, and transmits the data to the reliability analysis unit 6.

[0106] The S60 reliability analysis unit 6 data feature processing module uses a regularized restricted Boltzmann machine algorithm to perform unsupervised feature extraction on the collected degradation data. The extracted features are transmitted to the reliability modeling and analysis module to establish a nonlinear β_Wiener process model, analyze and output the service reliability assessment results.

[0107] This method, based on data-physical collaboration technology, combines the collaborative interaction of physical and digital prototypes to assess the service reliability of servo systems. First, the main control unit 1 acquires user-input structural parameters, environmental parameters, and task trajectory to ensure precise test control. Then, the servo system under test 3.1 is installed in the physical prototype test channel. During operation, the environmental stress loading system simulates the actual service environment, while the mechanical stress loading system 4.1 outputs the service load torque conforming to the task trajectory in real time, based on information returned by the servo system encoder 3.1.2. Reliability data acquisition is coordinated by the main control unit 1 with various test instruments and sensors 5. The data is transmitted to the analysis module 6, where a regularized restricted Boltzmann machine algorithm is used for feature extraction. Finally, a nonlinear β_Wiener process model is established for life prediction and reliability assessment, forming an analysis method based on the fusion of data-driven and physical mechanisms.

[0108] In S40, the mechanical stress loading system 4.1 includes a service load simulation calculation and output module 4.1.1, a loading controller 4.1.2, and a loading mechanism 4.1.3. The service load simulation calculation and output module 4.1.1 is used to calculate the current service load of the servo system, the loading mechanism 4.1.3 is used to output the corresponding loading torque, and the loading controller 4.1.2 is used to realize closed-loop torque control.

[0109] In this embodiment, to achieve dynamic load simulation of the servo system during testing, the mechanical stress loading system 4.1 is refined into three functional modules: The service load simulation calculation and output module 4.1.1 receives motion data from the servo system in real time, combines the task trajectory and robot structural parameters, derives the theoretical service load under the current state using a multi-rigid-body dynamics model or finite element method, and converts this load data into a control signal output; the loading mechanism 4.1.3, such as a servo electro-hydraulic actuator, receives the control signal and generates a corresponding torque to load the servo system onto the physical prototype, achieving dynamic loading; the loading controller 4.1.2 is responsible for closed-loop control of the loading process, comparing the applied load with the target torque in real time, adjusting the torque based on feedback errors, ensuring loading accuracy and dynamic response performance, preventing overload or loading instability, and maintaining the consistency of system reliability testing. This mechanical stress loading system 4.1, through modular design, improves the accuracy and dynamic response capability of service load simulation in servo system reliability testing. The service load simulation calculation and output module 4.1.1, combined with information from the real-time encoder 3.1.2 and structural parameters, can accurately reproduce load conditions under different operating states. The loading mechanism 4.1.3 provides sufficient actuation force and flexibility to simulate different types of torque requirements. The loading controller 4.1.2 achieves high-precision control of the loading torque through a closed-loop control algorithm, significantly improving loading consistency and system stability, thereby ensuring high fidelity in the reliability testing process. This system enhances the physical realism of the entire test platform, providing more realistic service environment support for reliability assessment.

[0110] The service load simulation calculation and output module 4.1.1 includes a servo system speed and acceleration estimation program and a service load calculation program. The speed and acceleration estimation program is based on the angular displacement signal collected by the servo system encoder 3.1.2 and uses a tracking differentiator to estimate the angular velocity and angular acceleration.

[0111] In this embodiment, to improve the real-time performance and accuracy of load calculation in the mechanical stress loading system 4.1, the service load simulation calculation and output module 4.1.1 is further subdivided into a servo system velocity and acceleration estimation program and a service load calculation program. First, during service, the servo system's encoder 3.1.3 continuously acquires angular displacement signals and transmits this displacement data to the velocity and acceleration estimation program in real time. This estimation program uses a tracking differentiator algorithm to process the angular displacement data. By constructing a dynamic tracking system, it effectively filters out measurement noise and estimates the current angular velocity and angular acceleration of the servo system. This algorithm has good dynamic response characteristics and anti-interference capabilities, and is suitable for identifying the high-frequency changing motion state of the servo system. Then, the service load calculation program takes the estimated angular velocity and angular acceleration as input, combines the robot's structural parameters, kinematic and dynamic models, and outputs the service load signal to be simulated. This signal is then provided to the loading mechanism 4.1.3 for execution.

[0112] By embedding the velocity and acceleration estimation process into the load calculation flow and introducing a high-precision tracking differentiator algorithm, the system can acquire dynamic motion states in real time, significantly improving the response speed and calculation accuracy of load calculation. This ensures that the torque output by the loading system is closer to actual working conditions, thereby enhancing the dynamic simulation capability and physical realism of the entire servo system reliability testing platform. Simultaneously, it avoids the noise amplification problem caused by direct second-order numerical differentiation, improving system stability and estimated reliability.

[0113] The service load calculation program calculates the service load of the servo system by combining the robot's inverse dynamics equations with the servo system's joint angular displacement, angular velocity, and angular acceleration, and outputs the result to the load controller 4.1.2.

[0114] In this implementation, to improve the physical accuracy and engineering applicability of the servo system service load estimation, the service load calculation program employs a modeling method based on inverse dynamics. First, the program receives joint angular displacement, angular velocity, and angular acceleration data acquired by the servo system encoder 3.1.2 and processed by the velocity and acceleration estimation program. Then, combining the industrial robot's structural parameters (including link mass, geometric dimensions, inertial parameters, and gravity effects), a standard robot inverse dynamics model is established, such as using the Lagrange method or the Newton-Euler method, to calculate the torque. The inverse dynamics model transforms the motion state input into the actual load output acting on each joint of the servo system, obtaining the dynamic service torque required under the current task trajectory. This result is the simulated service load of the servo system and is transmitted to the loading controller 4.1.2 via the control bus to drive the loading mechanism 4.1.1 to output the corresponding torque.

[0115] This method utilizes the robot's inverse dynamics equations to establish a precise load mapping relationship, ensuring that the output load closely matches the dynamic mechanical behavior during actual service, significantly improving the realism and technical representativeness of reliability tests. Simultaneously, relying on real-time acquired motion state data, the system exhibits high responsiveness, reflecting load changes in the servo system at different motion stages in real time, which is crucial for identifying load-sensitive conditions and capturing potential failure points. By integrating this program into the service load simulation calculation and output module 4.1.1, the data link and module coupling are optimized, enhancing the overall system's real-time simulation capabilities.

[0116] In S40, the environmental stress loading system includes a temperature and humidity control box 4.2.1 and a vibration table 4.2.2, which are used to simulate temperature, humidity and vibration service conditions respectively.

[0117] In this embodiment, to fully replicate the environmental stress conditions of the servo system during actual use, the environmental stress loading system is configured as a composite loading platform consisting of a temperature and humidity control box 4.2.1 and a vibration table 4.2.2. The temperature and humidity control box 4.2.1 simulates typical temperature and humidity fluctuations in the industrial environment where the servo system operates. It features adjustable heating, cooling, humidification, and dehumidification functions, and the temperature range (e.g., -40℃ to +85℃) and relative humidity range (e.g., 20% to 95%) can be set according to the test parameters. During the test, the control box remains closed to ensure the stability and uniform distribution of the test environment parameters.

[0118] The vibration table 4.2.2 is used to simulate the mechanical vibration conditions that the servo system may encounter during operation, supporting multi-axis and multi-frequency vibration excitation modes. This vibration table 4.2.2 can generate various excitation signals, such as sinusoidal vibration, random vibration, or impact vibration, based on the ground excitation spectrum of the industrial site or the vibration characteristics of the actual equipment. These signals are then applied to the physical prototype via the table surface, subjecting the servo system to realistic dynamic disturbance stress. These two subsystems are uniformly controlled by the environmental stress loading system and can operate collaboratively to achieve combined loading of various environmental stress factors, enhancing the realistic simulation capability of service scenarios.

[0119] This environmental stress loading system, by integrating temperature and humidity control and vibration excitation equipment, effectively covers the main environmental variables in typical industrial robot operating environments, providing a multi-stress coupling loading platform for servo system reliability testing. Through the coordinated control of the main control program in the central control unit 1 and the environmental stress loading system, the loading sequence, intensity, and duration of each stress factor in the testing process can be precisely set, thereby improving the accuracy of environmental adaptability assessment in service reliability testing. Simultaneously, simultaneous loading of multiple factors helps to reveal potential collaborative failure modes, enhancing the depth and systematic nature of failure mechanism research.

[0120] In S50, the test instruments and sensors include a torque sensor, a power meter, a power meter, a temperature sensor, a resistance tester, a high voltage tester, and an inter-turn tester.

[0121] In this embodiment, to achieve multi-dimensional parameter monitoring and data acquisition of the industrial robot servo system during reliability testing, the testing system is equipped with various high-precision sensors and instruments, covering multiple key test indicators such as electrical performance, thermal performance, and mechanical response. Specifically, torque sensors are used to monitor the actual torque response of the servo system's output shaft or loaded parts in real time, and are combined with a torque controller for closed-loop control and failure identification; power meters and wattmeters are used to collect voltage, current, electrical energy, and instantaneous power data during servo system operation to evaluate electrical input and energy efficiency characteristics; temperature sensors are deployed in key areas of the servo motor housing (3.1.1), servo driver unit (3.1.3), and control board to monitor heat distribution and over-temperature risks.

[0122] Meanwhile, to assess the insulation and electrical integrity of the servo system's electrical structure, a resistance tester is configured to detect changes in circuit contact resistance or winding resistance and identify potential aging trends; a high-voltage tester is used to apply high-voltage stress to the insulation system to verify its withstand voltage performance and safety margin; and an inter-turn tester is used to identify early fault modes such as inter-turn short circuits in the motor windings, obtaining coil state evolution information through high-frequency pulse excitation and voltage waveform analysis. All the above testing instruments and sensors are uniformly connected to the central control unit 1, which performs synchronous scheduling and data acquisition according to the user-defined acquisition control parameters and transmits the data in real time to the reliability analysis unit 6 for processing.

[0123] This implementation method constructs a comprehensive test data acquisition system covering mechanical, electrical, and thermal characteristics, achieving efficient capture of key performance parameters throughout the servo system's entire lifecycle and providing comprehensive and accurate data support for reliability analysis. The collaborative work of various sensors and instruments enables multi-dimensional parameter correlation analysis, identifying potential failure trends and degradation paths, thus improving the depth and interpretability of reliability assessment. Simultaneously, the system possesses excellent modular scalability and interface compatibility, facilitating adaptation to different servo system specifications and testing requirements, enhancing the versatility and adaptability of the testing platform.

[0124] In S60, the Regularized Restricted Boltzmann Machine algorithm introduces a regularization term into the objective function to improve the ability to extract degenerate features.

[0125] In this embodiment, to improve the accuracy and robustness of feature extraction from degradation data during the service life of industrial robot servo systems, the Regularized Restricted Boltzmann Machine (R-RBM) algorithm in the reliability analysis unit 6 introduces a regularization term into its optimization objective function. Based on the traditional RBM unsupervised learning model, this algorithm adds an L1 norm or L2 norm as a regularization term embedded in the weight update process, thereby effectively suppressing redundant information and overfitting during degradation feature extraction.

[0126] In practical implementation, this algorithm takes high-dimensional, multi-source temporal degradation data collected by sensors as input. Through forward propagation and backsampling mechanisms, it learns the potential degradation characteristics of the servo system at different service stages. Regularization constraints enhance the model's ability to identify key degradation factors. The introduction of regularization terms allows the model to control the weights during parameter updates, highlighting feature dimensions with significant discriminative power and improving sensitivity to subtle degradation trends.

[0127] This implementation significantly improves the stability and robustness of degradation feature extraction by optimizing the objective function structure. It helps to accurately capture key degradation indicators under noise interference or imbalanced data samples, and enhances the reliability analysis model's ability to identify early failure modes. At the same time, it effectively controls model complexity, improves algorithm convergence efficiency and generalization ability, and provides more discriminative input features for subsequent nonlinear β_Wiener process modeling, thereby improving the accuracy of the entire service reliability assessment system.

[0128] In S60, the nonlinear β_Wiener process model introduces a β-distributed random effect into the drift term to characterize the individual differences and uncertainties in the degradation process of the servo system.

[0129] In this implementation, to improve the adaptability of degradation modeling in service reliability analysis to differences between different servo system samples, the nonlinear β_Wiener process model introduces a random effect term based on the β distribution into its drift term. The core idea of ​​this model is to introduce uncertainty modeling by treating the drift term as a random variable based on the traditional Wiener process. Specifically, the drift parameter is modeled using the β distribution to reflect the individual differences in the servo system degradation path among different samples.

[0130] In practical implementation, the model uses statistical learning to estimate four parameters (β1, β2, β3, β4) of the β distribution from historical or real-time collected degradation data. This distribution's sampled values ​​are then injected into the drift term of each test sample, giving each degradation trajectory a degree of randomness and individualization. This modeling approach not only enhances the degradation model's ability to fit complex real-world degradation behaviors but also effectively addresses the differences in performance degradation paths caused by manufacturing deviations, changes in operating conditions, or different usage habits during actual service of servo systems.

[0131] This nonlinear β_Wiener process model has good flexibility and scalability, and can simultaneously capture the nonlinear trend of degradation rate and the uncertainty distribution characteristics across samples, which helps to improve the confidence interval expression ability of lifetime prediction and the accuracy of risk assessment. In addition, the β distribution has the characteristics of limited domain and flexible form, which is suitable for modeling the gradual performance degradation characteristics commonly found in degradation systems, and is especially suitable for the service modeling needs of high-reliability, slow-degradation equipment such as industrial robot servo systems.

[0132] Overall, this invention utilizes a digital-physical collaborative testing platform to recreate the service load and environment of a servo system through the linkage of digital-physical prototypes. By integrating dynamic calculations, multi-sensor acquisition, deep learning algorithms, and statistical modeling techniques, it achieves degradation identification, performance prediction, and reliability verification of industrial robot servo systems under real working conditions, possessing the advantages of high precision, low cost, and high efficiency.

[0133] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A reliability testing platform for the collaborative service of industrial robot servo systems, characterized in that, include: The industrial robot digital layer prototype is constructed by a robot digital twin control model, which is built by collecting load data during the operation of the industrial robot, establishing a load state matrix, and associating it with the robot body stiffness model, joint friction model and end effector association model. An industrial robot physical layer prototype includes several test channels, each test channel corresponding to a joint of the industrial robot. A servo system under test is set in the test channel. The industrial robot physical layer prototype realizes the simulation of the servo system in the service motion state of the industrial robot by linking the digital and physical information with the industrial robot digital layer prototype. Service loading module, which is used to simulate and output the mechanical and environmental conditions of industrial robot service, and realize high-fidelity simulation output of service load of industrial robot servo system; Test instruments and sensors are installed in each test channel of the physical layer prototype of the industrial robot to collect multimodal degradation monitoring data; The reliability analysis unit is used to screen and extract degradation features from the degradation data collected by the test instruments and sensors, establish a degradation model based on the β_Wiener process, and analyze and output the service reliability assessment results of the industrial robot servo system. The central control unit is used to control the test platform, including user interaction, test process control, data-object information linkage, and test data acquisition control for the reliability test of the industrial robot servo system. The service loading module includes an environmental stress loading system and a mechanical stress loading system. The environmental stress loading system includes a temperature and humidity control box and a vibration table, which are used to simulate the ambient temperature, humidity and vibration conditions of the industrial robot servo system during operation. The mechanical stress loading system includes: The service load simulation calculation and output module comprises two parts: a speed and acceleration estimation program for the industrial robot servo system and a service load calculation program. The speed and acceleration estimation program uses the angular displacement signal obtained from the encoder of the servo system under test, applies a tracking differentiator to estimate its angular velocity and angular acceleration values, and obtains the joint angular displacement, joint angular velocity, and joint angular acceleration of the servo system based on the transmission coefficient of the harmonic reducer, inputting these values ​​into the service load calculation program. The service load calculation program, based on the joint angular displacement, joint angular velocity, and joint angular acceleration output by the speed and acceleration estimation program, calculates the service load of the industrial robot servo system using the robot inverse dynamics equations and transmits it to the loading system controller to realize the mechanical stress loading of the industrial robot servo system during service. The loading mechanism is a torque servo system, including a driver and a torque motor. The driver receives the current service load result output by the service load simulation calculation and output module to generate a drive signal, and the drive motor outputs the corresponding service loading torque. A load controller is used to generate a drive control signal based on the difference between the output torque of the torque motor and the target torque calculated by the driver receiving the service load simulation and output by the output module, so as to realize closed-loop control of the load output of the torque servo system.

2. The industrial robot servo system digital-physical collaborative service reliability testing platform according to claim 1, characterized in that, The servo system under test includes: Servo motors, specifically corresponding to robot joints, are installed in the test channel; An industrial robot servo driver, which receives control signals from an industrial robot servo controller and drives a servo motor to operate. An encoder is mounted on a servo motor. The encoder is used to collect angular displacement signals of the servo motor and transmit them as feedback signals to the servo driver to realize the position control of the servo motor.

3. The industrial robot servo system digital-physical collaborative service reliability testing platform according to claim 1, characterized in that, The industrial robot digital layer prototype is a robot digital twin control model, and its construction steps are as follows: S1. Collect real-time load data of the industrial robot during operation and construct a load state matrix that changes over time. The load state matrix includes multi-dimensional load characteristic parameters. S2. Based on the load state matrix and combined with the industrial robot structural dynamics model, construct an industrial robot joint dynamic twin control model that is adaptive to the working conditions. S3. Based on the dynamic twin control model of industrial robot joints, the output torque parameters of the servo driver are adjusted in real time to match the actual load changes and simulate the control response state of the servo system inside the industrial robot body.

4. The industrial robot servo system digital-physical collaborative service reliability testing platform according to claim 2, characterized in that, Each test channel of the industrial robot physical layer prototype is used to perform service reliability testing of the servo system of a certain joint of the industrial robot. The servo motor is set in the test channel and is connected to the harmonic reducer through the first drive shaft. The test instruments and sensors include a torque sensor. The harmonic reducer is connected to one end of the torque sensor through the second drive shaft and the first coupling. The other end of the torque sensor is connected to the service loading module through the third drive shaft and the second coupling.

5. The test method for the reliability test platform of the industrial robot servo system digital-physical collaborative service as described in any one of claims 1-4, characterized in that, include: S10. Input the structural parameters, service environment parameters, mission trajectory parameters, and reliability test control parameters of the servo system under test through the interactive interface of the central control unit. S20. Collect real-time load data during the operation of the industrial robot, construct a load state matrix that changes over time, and establish a digital layer prototype system by combining the structural dynamics model of the industrial robot. S30. Install the servo system under test on the physical layer prototype system and conduct service reliability testing. The servo system under test runs according to the robot's task trajectory and outputs service feedback status in conjunction with the digital layer prototype system. S40, the test platform environmental stress loading system simulates the service environment conditions of the servo system, and the mechanical stress loading system outputs the current service load according to the robot inverse dynamics model, so that the servo system enters the service state; S50: The central control unit collects control parameters based on the test data input by the user, controls the data collection of each test instrument and sensor during the reliability test, and transmits the data to the reliability analysis unit. The S60 reliability analysis unit data feature processing module uses a regularized restricted Boltzmann machine algorithm to perform unsupervised feature extraction on the collected degradation data. The extracted features are then transmitted to the reliability modeling and analysis module to establish a nonlinear β_Wiener process model, analyze and output the service reliability assessment results.

6. The test method for the reliability test platform of the industrial robot servo system digital-physical collaborative service according to claim 5, characterized in that: In step S20, the real-time load data includes force data, joint torque data, speed information, and acceleration information of the end effector, and the accuracy of load status perception is improved through multi-source sensor fusion technology.

7. The test method for the industrial robot servo system digital-physical collaborative service reliability test platform according to claim 5, characterized in that: In step S20, the load state matrix is ​​dynamically refreshed according to a preset time window, and the continuous load change trend is obtained by using a sliding time window mechanism. The trend is then input into the dynamic torque control model in a recursive form to realize the time-series feature modeling.

8. The test method for the reliability test platform of the industrial robot servo system digital-physical collaborative service according to claim 5, characterized in that: In step S20, the industrial robot structural dynamics model includes a robot body stiffness model, a joint friction model, and an end effector inertia model. The industrial robot structural dynamics model is dynamically calibrated through parameter identification methods to improve the response matching degree between the servo system and the actual robot working conditions.

9. The test method for the reliability test platform of the industrial robot servo system digital-physical collaborative service according to claim 5, characterized in that: In step S60, the regularized restricted Boltzmann machine algorithm introduces a regularization term into the objective function to improve the ability to extract degenerate features; and / or In S60, the nonlinear β_Wiener process model introduces a β-distributed random effect into the drift term to characterize the individual differences and uncertainties in the degradation process of the servo system.

Images