Multi-working-condition robot joint anomaly detection method and system

Abstract

The invention relates to the technical field of robot anomaly detection, in particular to a multi-working-condition robot joint anomaly detection method and system which comprises a working condition data acquisition module, a coupling matrix generation module, a feature data generation module and a joint anomaly judgment module. Through the perspective of multi-physical signal fusion and dynamic modeling, the limitation that traditional single signal detection is prone to coupling interference is overcome, the accuracy of abnormal joint recognition and classification is remarkably improved, potential faults can be warned in advance, and the risk of non-planned shutdown is reduced; and on the other hand, through decoupling analysis and a working condition self-adaptive threshold strategy, the adaptive capacity of the detection system to variable working conditions and variable loads of the robot is enhanced, and the reliability and scientificity of the detection process are guaranteed, so that the operation safety and maintenance efficiency of the robot are improved, and the service life of the robot is prolonged. And reliable data support and decision basis are provided for realizing state-based intelligent maintenance and health management.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of robot anomaly detection, and relates to a multi-working-condition robot joint anomaly detection method and system. BACKGROUND

[0002] Existing joint anomaly detection usually analyzes each joint as an independent unit, that is, a "single-joint single model" detection architecture is adopted. However, in the actual "multi-working-condition" operating environment, the robot is always in continuous motion, and its spatial pose continuously changes. For a serial industrial robot (such as a six-axis robot), the motion of a joint will be transmitted to other joints through a mechanical linkage structure, forming a cross-joint dynamic coupling. When the detection system prompts that joint J3 is abnormal, the abnormal signal may actually be caused by the vibration of joint J2 conducted through the mechanical structure, or the whole machine resonance phenomenon induced at a specific pose, rather than the fault of joint J3 itself.

[0003] There are still some places that need to be optimized in the current robot anomaly detection, which are embodied in the following aspects:

[0004] With the conversion of the robot pose from the retracted state to the extended state, the moment of inertia matrix and the structural stiffness matrix of the whole machine change significantly. This means that the natural frequency of the system drifts dynamically with the pose, making it difficult for the detection model with fixed parameters to adapt to the dynamic characteristics in the whole working condition range.

[0005] Under different working conditions, the transmission efficiency of abnormal vibration signals in the mechanical structure changes. For example, at a pose close to the kinematic singularity point, weak fault vibration generated by a joint may be amplified and conducted to the end joint by the robot structure; while in the folded pose, similar vibration signals may be greatly attenuated, causing inconsistent signal representation. SUMMARY

[0006] In view of the problems existing in the prior art, the multi-working-condition robot joint anomaly detection method and system are provided to solve the above technical problems.

[0007] In order to achieve the above purpose and other purposes, the technical scheme adopted by the present application is as follows:

[0008] In one aspect, the present application provides a multi-working-condition robot joint anomaly detection method, which comprises the following steps:

[0009] Step S1: acquiring real-time multi-dimensional sensing data of each joint of the robot, the multi-dimensional sensing data including joint motor current, joint vibration signal and joint encoder position data; performing robot forward kinematics operation according to the joint encoder position data, constructing a real-time robot space linkage pose model, and generating robot real-time working condition state data containing current spatial pose features;

[0010] Step S2: Based on the real-time working condition state data of the robot, whole-machine variable inertia and variable stiffness dynamics analysis is performed, and a space pose vibration transmission model that drifts with the pose is constructed; the vibration wave conduction efficiency between each adjacent joint of the robot is quantitatively calculated by using the space pose vibration transmission model, and a vibration modal coupling coefficient matrix between joints corresponding to the current working condition is generated;

[0011] Step S3: The vibration modal coupling coefficient matrix between joints is used to perform structure-conducted noise decoupling processing on the joint vibration signal, the coupled vibration components transmitted to the target joint through the mechanical linkage from other joints are eliminated, the intrinsic vibration feature vector representing the health state of the target joint body is extracted, and multi-dimensional decoupled joint feature data is generated in combination with the joint motor current;

[0012] Step S4: According to the real-time working condition state data of the robot, a preset dynamic working condition reference threshold library is called, and the multi-dimensional decoupled joint feature data is adaptively abnormally judged by using the dynamic working condition reference threshold library, the specific joint where the abnormality occurs is located and the abnormal type is identified, so as to generate a robot joint abnormality detection result report.

[0013] Another aspect of the present application provides a multi-working-condition robot joint abnormality detection system, which comprises:

[0014] The working condition data acquisition module: acquires real-time multi-dimensional sensing data of each joint of the robot; performs robot kinematics forward solution operation according to joint encoder position data, constructs a real-time robot space linkage pose model, and generates real-time working condition state data of the robot;

[0015] The coupling matrix generation module: based on the real-time working condition state data of the robot, whole-machine variable inertia and variable stiffness dynamics analysis is performed, and a space pose vibration transmission model that drifts with the pose is constructed; the vibration wave conduction efficiency between each adjacent joint of the robot is quantitatively calculated by using the space pose vibration transmission model, and a vibration modal coupling coefficient matrix between joints corresponding to the current working condition is generated;

[0016] The feature data generation module: the vibration modal coupling coefficient matrix between joints is used to perform structure-conducted noise decoupling processing on the joint vibration signal, and the intrinsic vibration feature vector representing the health state of the target joint body is extracted, and multi-dimensional decoupled joint feature data is generated in combination with the joint motor current;

[0017] The joint abnormality discrimination module: according to the real-time working condition state data of the robot, a preset dynamic working condition reference threshold library is called, and the multi-dimensional decoupled joint feature data is adaptively abnormally judged, the specific joint where the abnormality occurs is located and the abnormal type is identified, so as to generate a robot joint abnormality detection result report.

[0018] As described above, the multi-condition robot joint abnormality detection method and system provided by the present application has at least the following beneficial effects:

[0019] (1) The multi-condition robot joint abnormality detection method and system provided by the present application collects real-time multi-dimensional sensing information of each joint during the operation of the robot, fuses joint motor current, joint vibration signal and joint encoder position data, and solves the forward kinematics of the robot in real time according to the encoder position data to construct a spatial link posture model, and then generates real-time working condition state data reflecting the current posture characteristics, effectively realizing multi-dimensional synchronous perception and digital modeling of the robot operating state. On this basis, based on the real-time working condition state data, the whole machine variable inertia and variable stiffness dynamics analysis is carried out, the spatial posture vibration transmission model which dynamically changes with the posture is established, thereby quantitatively evaluating the vibration wave conduction efficiency between each adjacent joint, generating a joint-to-joint vibration modal coupling coefficient matrix that fits the current working condition, and revealing the vibration energy transmission rule and coupling mechanism in the dynamic operation of the robot. The joint-to-joint vibration modal coupling coefficient matrix is used to decouple the structural conduction noise of the original joint vibration signal, filter out the cross-joint coupling vibration components transmitted to the target joint through the mechanical link, extract the intrinsic vibration feature vector directly representing the health state of the joint, and combine the joint motor current data to generate a multi-dimensional decoupled joint feature data set, significantly improving the independence and explainability of the state characteristics. Further, in combination with the real-time working condition state data of the robot, the preset dynamic working condition reference threshold library is dynamically called to adaptively identify the multi-dimensional decoupled joint characteristics, accurately locate the abnormal joint and identify the abnormal type, and finally output the robot joint abnormality detection result report.

[0020] (2) On the one hand, from the perspective of multi-physical signal fusion and dynamics modeling, the present application overcomes the limitation of traditional single signal detection which is easily affected by coupling interference, significantly improves the accuracy of abnormal joint identification and classification, helps to early warning potential failure and reduces the risk of unplanned downtime; on the other hand, through decoupling analysis and working condition adaptive threshold strategy, the adaptability of the detection system to variable working conditions and variable loads is enhanced, the reliability and scientificity of the detection process are ensured, so that the robot operating safety, maintenance efficiency and service life are improved, and reliable data support and decision basis are provided for realizing state-based intelligent maintenance and health management. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used for the description of the embodiments.

[0022] Figure 1 The present application is a method for connecting the schematic diagram.

[0023] Figure 2A schematic diagram of connection of various modules of the system of the present application. DETAILED DESCRIPTION

[0024] Embodiment 1

[0025] Referring to Figure 1 The abnormality detection method for the multi-condition robot joint comprises the following steps:

[0026] Step S1: Real-time multi-dimensional sensing data of each joint of the robot is acquired, the multi-dimensional sensing data comprising joint motor current, joint vibration signal and joint encoder position data; kinematics forward solution operation is performed according to the joint encoder position data, a real-time robot space link posture model is constructed, and robot real-time working condition state data containing current space posture features is generated.

[0027] In one possible design, step S1 comprises the following steps:

[0028] Step S11: The original running data of the robot is synchronously collected through the acceleration sensor, current transformer and absolute value encoder installed at each joint of the robot;

[0029] Step S12: The original running data is time-stamped aligned and preprocessed to obtain multi-dimensional sensing data;

[0030] Step S13: Joint angle information in the multi-dimensional sensing data is extracted, a robot link coordinate system is established by using the D-H parameter method, and the spatial coordinates of the robot end and the gravity centers of each link are calculated;

[0031] Step S14: The instantaneous speed and acceleration of the robot are calculated according to the spatial coordinates and the rate of change of the joint angle, and the spatial coordinates, instantaneous speed and acceleration are combined to generate robot real-time working condition state data.

[0032] In the specific implementation process, first, the Hall current sensor, MEMS three-axis accelerometer and absolute value multi-turn encoder deployed at the driving end of each joint of the robot are used to synchronously collect multi-source signals to generate time-synchronized multi-dimensional sensing data. Then, based on the joint angle information in the multi-dimensional sensing data, the robot link space posture is recursively analyzed level by level, and the improved D-H parameter method is used for the analysis. First, the homogeneous transformation matrix of the adjacent joint coordinate system is constructed, and the calculation formula is wherein is the transformation matrix of the i-1th joint coordinate system relative to the ith joint coordinate system, is the rotation angle of the ith joint, is the link length, is the link torsion angle, is the link offset distance, The structure parameters are factory calibration parameters of the robot, and the relative geometric relationship between adjacent links is described through the matrix. Then, the absolute spatial position of the center of gravity of each link in the base coordinate system is calculated in real time by using the chain rule of homogeneous transformation matrix, to generate the spatial link posture model data, and the center of gravity coordinate calculation formula is wherein is the position vector of the center of gravity of the i th link in the base coordinate system , is the cumulative transformation matrix from the base to the i th joint, is the position vector of the center of gravity of the i th link in its own local coordinate system, in meters, which is called from the robot dynamics parameter library.

[0033] Finally, the motion state differential calculation is performed according to the time sequence variation characteristics of the joint angle, and the spatial posture characteristics and the motion state characteristics are assembled into a high-dimensional tensor to generate the robot real-time working condition state data, and the assembly formula of the data is wherein is the robot real-time working condition state data, which is in the form of a multi-dimensional matrix, N is the number of degrees of freedom of the robot, is the sampling period, the second term is the i th joint angular velocity calculated, in radians per second, and the third term is the i th joint angular acceleration, in radians per second square, is the transpose of the spatial coordinates of the center of gravity.

[0034] Step S2: Based on the robot real-time working condition state data, the whole machine variable inertia and variable stiffness dynamics analysis is performed, and the spatial posture vibration transmission model with dynamic drift is constructed; the vibration mode coupling coefficient matrix between the joints corresponding to the current working condition is generated by quantitatively calculating the vibration wave conduction efficiency between the adjacent joints of the robot using the spatial posture vibration transmission model.

[0035] In one possible design, step S2 includes the following steps:

[0036] Step S21: According to the link spatial position in the robot real-time working condition state data, the robot rigid body dynamics parameters are called to calculate the robot whole machine mass distribution matrix and moment of inertia matrix in real time;

[0037] In the specific implementation process, based on the joint angle and link spatial position information in the robot real-time working condition state data, the pre-stored robot rigid body dynamics parameter database is called, which includes the mass, mass center position vector and local inertia tensor of each link obtained by importing the offline CAD model or parameter identification experiment, and the link velocity Jacobian matrix is constructed using these parameters, and the construction formula is wherein is the Jacobian matrix of the k th link, is the unit vector of the rotation axis of the k-1th joint, is a dimensionless parameter, is the spatial position vector of the mass center of the kth link, with the unit of meter, is the spatial position vector of the origin of the k-1th joint coordinate system, with the unit of meter;

[0038] Then, based on the constructed Jacobian matrix and rigid body dynamics parameters, spatial rotation transformation of the inertia tensor and aggregation calculation of the total system kinetic energy are performed to generate the rotational inertia matrix, and the calculation formula is wherein is the generalized inertia matrix of the robot in the joint space, with the unit of kilogram-meter square, N is the total number of robot joints, is the mass of the kth link, with the unit of kilogram, and are the linear velocity part and the angular velocity part of the Jacobian matrix, respectively, with the unit of meter / radian, wherein with the unit of meter / radian; with the unit of dimensionless;, is the inertia tensor of the kth link in its local coordinate system, with the unit of kilogram-square meter, is the rotation matrix from the local coordinate system of the kth link to the base coordinate system, and is a dimensionless matrix, The item maps the local inertia to the global coordinate system through a similarity transformation.

[0039] Step S22: Based on the rotational inertia matrix and combined with the joint flexibility coefficient of the robot, system natural frequency drift analysis is performed to obtain real-time system modal frequency data varying with the pose;

[0040] In the specific implementation process, according to the rotational inertia matrix varying with the dynamic pose, combined with the torsional stiffness characteristics of the joint reduction mechanism of the robot, the elastic potential energy matrix of the system is constructed, that is, the stiffness matrix, and the construction formula of the stiffness matrix is wherein is the stiffness matrix of the robot system, with the unit of Newton-meter / radian, which is a diagonal matrix of N×N, is the equivalent torsional stiffness coefficient of the i th joint reduction mechanism, with the unit of Newton-meter / radian, which is obtained by consulting the technical specification book of the robot reduction mechanism.

[0041] Then, based on the generalized inertia matrix and the system stiffness matrix, the undamped free vibration differential equation characteristic equation of the robot multi-degree-of-freedom system is established, and the real-time solution of the modal frequency is solved by using the generalized eigenvalue solving algorithm in numerical linear algebra, and the calculation formula of the system modal frequency is wherein is the rth order natural frequency under the current spatial pose q, with the unit of radian per second, is the inverse of the generalized inertia matrix, with the unit of 1 / kg·m2, is an operation function that extracts the rth eigenvalue of the matrix, which reflects the fact that when the robot pose stretches, the moment of inertia increases, the stiffness remains unchanged, the natural frequency will occur the nonlinear drift to low frequency physical phenomenon. Finally, the calculated natural frequencies of each order are arranged in order from small to large, to generate real-time system modal frequency data containing the distribution of fundamental frequency to high order modal frequency.

[0042] Step S23: According to the real-time system modal frequency data and the geometric length of the connecting rod, the transfer function method is used to analyze the attenuation and amplification characteristics of vibration energy in the connecting rod structure, and a spatial pose vibration transfer model is constructed;

[0043] In the embodiment process, based on the real-time system modal frequency data and the geometric length data of the connecting rod in the robot body parameter library, the dynamic calculation of the vibration frequency response ratio is carried out, and the calculation formula is , wherein is the frequency tuning ratio, and is a dimensionless parameter, is the current monitored external excitation frequency, with the unit of rad / s, which is extracted by FFT transformation of the joint vibration signal; is the rth order system natural frequency with the unit of rad / s, which comes from the real-time system modal frequency data, and the ratio reflects the closeness of the current excitation frequency and the resonance point after the system "variable configuration". Then, the frequency tuning ratio is used to construct a spatial pose vibration transfer function containing dynamic amplification and geometric attenuation characteristics, according to the single degree of freedom forced vibration model in mechanical vibration theory. The calculation formula of the transfer function is , wherein is the vibration transfer rate, and is a dimensionless parameter. The front half of the formula is a dynamic amplification factor, which describes the resonance amplification effect of the structure on the vibration energy when the excitation frequency is close to the natural frequency, is the material damping ratio, and is a dimensionless parameter, usually in the range of 0.01 to 0.1, and the latter half of the formula is a geometric attenuation factor, which describes the energy dissipation of the vibration wave in the physical connecting rod, is the material loss factor, and is a dimensionless parameter, is the geometric length of the kth connecting rod, with the unit of meter, is the propagation speed of the vibration wave in the metal material of the link, in meters per second. Finally, the calculated vibration transmissibility is combined with the topology of the robot's link to construct a spatial pose vibration transmission model, which is essentially a real-time changing filter group in the frequency domain, describing how the vibration energy of a specific frequency is "amplified" or "absorbed" in the physical structure of the link at the current specific stretching or folding pose.

[0044] Step S24: Calculate the vibration response transmissibility between any two joints using the spatial pose vibration transmission model, and combine the transmissibility of all joint pairs to generate the inter-joint vibration modal coupling coefficient matrix.

[0045] In one possible design, step S24 includes the following steps:

[0046] Step S241: Select the ith joint as the vibration source joint and the jth joint as the measured joint, and extract the link conduction path parameters between the two joints under the current working condition;

[0047] Step S242: Input a unit pulse excitation to the vibration source joint position in the spatial pose vibration transmission model, and simulate the response signal amplitude at the measured joint;

[0048] Step S243: Calculate the ratio of the response signal amplitude to the unit pulse excitation to obtain the single-path vibration transmissibility factor;

[0049] Step S244: Traverse all joint combinations of the robot, calculate the corresponding single-path vibration transmissibility factor, and arrange it according to the row-column rule to generate the inter-joint vibration modal coupling coefficient matrix.

[0050] In the specific implementation process, based on the real-time working condition state data of the robot, the conduction path analysis of all joints is performed, the analysis logic is set to take the ith joint as the vibration source and the jth joint as the measured object, the link conduction path parameters under the current spatial pose are extracted, the spatial Euclidean distance and the link axis angle between the two joints are calculated to generate path geometric feature data, and then the path dynamic impedance is evaluated using system modal frequency data to generate path dynamic impedance data. Subsequently, based on the path geometric feature data and the path dynamic impedance data, the response simulation under unit pulse excitation and the transmissibility factor calculation are performed, and the calculation formula of the single-path vibration transmissibility factor is wherein is the single-path vibration transmissibility factor, is the angle between the ith joint axis and the jth joint axis under the current spatial pose, coming from the forward kinematics data of the robot, used to represent the directional efficiency of the vibration wave along the axial transmission; is the equivalent modal mass under the current pose, in kilograms, extracted from the moment of inertia matrix; is the first order natural frequency of the system under current working condition, with unit of rad / s; is the structural damping ratio, which is a dimensionless parameter; is the linear transmission distance between two joints, is the characteristic decay length, which is calculated based on the sound speed of the connecting rod material and the frequency; is a normalization constant with unit of N·s·kg / m, which is used to eliminate the dimension and ensure the transfer factor between 0 and 1. Finally, the joint-to-joint vibration modal coupling coefficient matrix is assembled according to the calculated single-path vibration transfer factor, and the row-column mapping method is used to fill in the i-th row and j-th column of the matrix, for the main diagonal elements , set it to 1, and finally generate the joint-to-joint vibration modal coupling coefficient matrix.

[0051] Step S3: using the joint-to-joint vibration modal coupling coefficient matrix to perform structural transmission noise decoupling processing on the joint vibration signal, eliminating the coupled vibration components transmitted to the target joint through the mechanical connecting rod from other joints, extracting the intrinsic vibration feature vector representing the health status of the target joint, and combining the joint motor current to generate multi-dimensional decoupled joint feature data.

[0052] In one possible design, step S3 includes the following steps:

[0053] Step S31: performing frequency domain transformation on the joint vibration signal to obtain original joint vibration spectrum data;

[0054] Step S32: based on the joint-to-joint vibration modal coupling coefficient matrix, identifying and calculating the cross-talk vibration components received by the current joint from adjacent joints and remote joints;

[0055] Step S33: subtracting the cross-talk vibration components from the original joint vibration spectrum data to perform spectrum purification reconstruction, and obtaining the decoupled intrinsic vibration spectrum of the target joint;

[0056] Step S34: performing time domain inverse transformation and feature extraction on the intrinsic vibration spectrum of the target joint, calculating the effective value, kurtosis index and energy barycenter, and generating the intrinsic vibration feature vector;

[0057] Step S35: performing feature fusion of the intrinsic vibration feature vector and the frequency domain features of the joint motor current, thereby generating multi-dimensional decoupled joint feature data.

[0058] In one possible design, step S32 includes the following steps:

[0059] Step S321: obtaining the original vibration intensity data of all joints except the target joint;

[0060] ​Step S322: Indexing the row vector corresponding to the target joint in the inter-joint vibration modal coupling coefficient matrix to obtain the input coupling coefficient set of the target joint;

[0061] Step S323: Performing weighted convolution operation on the original vibration intensity data of all other joints using the input coupling coefficient set to simulate the interference waveform caused by structural conduction;

[0062] Step S324: Linearly superimposing the simulated interference waveform to calculate the cross-crosstalk vibration component at the current joint.

[0063] In the specific implementation process, the joint vibration signal is processed by short-time Fourier transform based on the Hanning window to convert the discrete acceleration sequence in the time domain into complex spectrum in the frequency domain, thereby obtaining the original joint vibration spectrum data. The purpose of the transformation is to deconstruct the aliasing time domain waveform into frequency components. Then, the reverse analysis of multi-source interference is performed based on the inter-joint vibration modal coupling coefficient matrix. The process first locks the target joint i to be analyzed, extracts the ith row vector from the coupling matrix as the input coupling coefficient set, and the elements in the vector characterize the vibration transmission efficiency of other joints j to joint i. Subsequently, the cross-crosstalk vibration component at the current joint is calculated using the principle of frequency domain multiplication equivalent to time domain convolution. The calculation formula is , wherein is the calculated crosstalk vibration component spectrum at the frequency , the unit is meter per second square, N is the total number of robot joints, is the vibration transmission factor varying with the pose q, which is a dimensionless parameter; is the measured vibration amplitude spectrum of the interference source joint j, the unit is meter per second square.

[0064] Subsequently, the cross-crosstalk vibration component calculated above is removed from the original joint vibration spectrum data using the spectral subtraction principle to perform spectral purification reconstruction. The calculation formula of the intrinsic vibration spectrum is , wherein is the decoupled target joint intrinsic vibration amplitude, the unit is meter per second square, is the total vibration amplitude measured by the target joint, is the over-decoupling factor, which is a dimensionless parameter, usually 1.0 to 1.2, used to compensate for model errors; is the spectral floor parameter, which is a dimensionless parameter, used to prevent negative values or non-physical zero spectrum after subtraction.

[0065] Then, the intrinsic vibration spectrum after decoupling is subjected to feature dimension reduction, and the kurtosis index representing the impact feature and the energy barycenter representing the frequency band change are calculated, respectively. The calculation formula of the intrinsic kurtosis index is , wherein The time-domain intrinsic signal restored by inverse Fourier transform has a unit of meters per second square, and M is the number of sampling points; the energy gravity center calculation formula is , wherein is a frequency point, and has a unit of hertz. This index is used to capture the high-frequency resonance band shift caused by bearing wear or gear surface pitting.

[0066] Finally, the vibration characteristics described above are fused with the frequency domain characteristics of the joint motor current to construct multi-dimensional decoupled joint characteristic data, and the construction formula of the fusion vector is , wherein is the effective value of the motor current, and has a unit of ampere, is the total harmonic distortion rate of the current, and is a dimensionless parameter, , is the amplitude of the hth harmonic current, is the amplitude of the fundamental wave current.

[0067] Step S4: According to the real-time working condition state data of the robot, a preset dynamic working condition reference threshold library is called, and the multi-dimensional decoupled joint characteristic data is adaptively abnormally judged by using the dynamic working condition reference threshold library, the specific joint where the abnormality occurs is located and the abnormal type is identified, so as to generate a robot joint abnormality detection result report.

[0068] In one possible design, step S4 includes the following steps:

[0069] Step S41: According to the speed and load information in the real-time working condition state data of the robot, the running working condition of the robot is divided into intervals, and the working condition level to which the current belongs is determined;

[0070] Step S42: According to the working condition level, the corresponding dynamic working condition reference threshold library is indexed from the preset database, and the reference threshold library contains the upper and lower limits of the health characteristics in this working condition;

[0071] Step S43: The multi-dimensional decoupled joint characteristic data is input into the dynamic working condition reference threshold library for comparison and analysis, and the characteristic deviation degree is calculated;

[0072] Step S44: When the characteristic deviation degree exceeds the preset safety margin, it is determined that the state is abnormal, and the abnormal joint number is output;

[0073] Step S45: According to the spectral distribution characteristics of the characteristic deviation degree, the fault mode library is matched, and a robot joint abnormality detection result report is generated.

[0074] In one possible design, step S43 includes the following steps:

[0075] Step S431: The intrinsic vibration kurtosis value and the current harmonic distortion rate in the multi-dimensional decoupled joint characteristic data are extracted;

[0076] Step S432: Obtain the corresponding health reference mean value and standard deviation in the dynamic working condition reference threshold library;

[0077] Step S433: Calculate the distance of the intrinsic vibration kurtosis value and the current harmonic distortion rate relative to the health reference mean value using the Mahalanobis distance algorithm;

[0078] Step S434: Normalize the calculated distance to obtain the feature deviation degree.

[0079] In the specific implementation process, the grid mapping of the multi-dimensional working condition space is performed according to the real-time working condition state data of the robot, and the current working condition pressure index is calculated by using a weighted normalization algorithm, and the calculation formula is wherein is the working condition pressure index, and is a dimensionless integer used as a database index key; is the working condition resolution coefficient, and are the current joint angular velocity and joint torque, respectively, with units of radians per second and Newton-meters, and are the rated parameters, is the current robot arm extension length, with a unit of meters, is the maximum arm extension, , and are weight coefficients, and the sum of the three is 1.

[0080] Then, the calculated working condition pressure index is used to quickly index the multivariate statistical distribution parameters from the pre-set dynamic working condition reference threshold library, and the health reference mean value vector and the health reference covariance matrix under the specific working condition are extracted, with a dimension of N×N, wherein records the natural correlation between the vibration features and the current features under the health state.

[0081] Then, the generated multi-dimensional decoupled joint feature data is input into the statistical discrimination model based on the Mahalanobis distance to calculate the feature deviation degree, so as to overcome the defect of ignoring the correlation between variables in the traditional Euclidean distance, and the calculation formula of the Mahalanobis distance is wherein is the Mahalanobis distance value, which is a dimensionless parameter, is the current measured multi-dimensional decoupled joint feature data, including intrinsic kurtosis, energy barycenter, current distortion rate, etc., is the inverse matrix of the covariance matrix, and then the calculated is normalized to obtain the feature deviation degree, and the formula is wherein The characteristic deviation is a dimensionless parameter. For features with p degrees of freedom and a confidence level of p, the feature dimension is p. The chi-square distribution critical value is used to precisely quantify the degree to which the current state deviates from the "normal state under this working condition" using statistical methods.

[0082] Finally, when the feature deviation When the score is greater than 1, a joint abnormality is determined. Fault mode matching is then performed based on the spectral distribution characteristics of the feature vectors to establish a fault similarity scoring model. The calculation formula is as follows: ,in Let be the cosine similarity between the current signal and the k-th fault mode (such as bearing outer ring damage, gear tooth breakage), and be a dimensionless parameter. This is the decoupled intrinsic vibration spectrum vector, in meters per second squared. The feature frequency template vector of the k-th fault in the fault mode library is selected by... The pattern corresponding to the maximum value is used as the recognition result, and finally a robot joint anomaly detection result report containing the abnormal joint number, the degree of deviation, and the specific fault type is generated.

[0083] It should be added that the following steps are included after step S4:

[0084] The robot analyzes the joint anomaly detection report and extracts the severity level and fault type. If the severity level is minor, a deceleration command is generated to adjust the robot's motion planning trajectory to avoid the resonant frequency range. If the severity level is severe, an emergency stop command is generated and the faulty axis is locked with a brake.

[0085] Example 2

[0086] like Figure 2 As shown, a multi-condition robot joint anomaly detection system includes a condition data acquisition module, a coupling matrix generation module, a feature data generation module, and a joint anomaly discrimination module.

[0087] The above modules are connected via wired connections to enable data transmission between them;

[0088] Working condition data acquisition module: acquires real-time multi-dimensional sensing data of each joint of the robot; performs forward kinematics calculation of the robot based on the joint encoder position data, constructs the robot spatial link attitude model in real time, and generates real-time working condition status data of the robot;

[0089] The coupling matrix generation module: based on the real-time working condition state data of the robot, the whole machine variable inertia and variable stiffness dynamics analysis is carried out, and the space pose vibration transmission model with dynamic drift of pose is constructed; the vibration wave conduction efficiency between each adjacent joint of the robot is quantitatively calculated by using the space pose vibration transmission model, and the inter-joint vibration modal coupling coefficient matrix corresponding to the current working condition is generated;

[0090] The feature data generation module: the structural conduction noise decoupling processing is carried out on the joint vibration signal by using the inter-joint vibration modal coupling coefficient matrix, the intrinsic vibration feature vector representing the health state of the target joint body is extracted, and the multi-dimensional decoupled joint feature data is generated in combination with the joint motor current;

[0091] The joint abnormality discrimination module: according to the real-time working condition state data of the robot, the preset dynamic working condition reference threshold library is called, the multi-dimensional decoupled joint feature data is adaptively discriminated, the specific joint where the abnormality occurs is located and the abnormal type is identified, so as to generate the robot joint abnormality detection result report.

[0092] It should be noted that the setting of the interval and the threshold size is for the convenience of comparison, wherein the size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data, as long as it does not affect the proportional relationship of the parameters and the quantized values. And the above formula is the calculation of the dimensionless value, the formula is obtained by software simulation of a large amount of data to get the nearest real situation formula, and the preset parameters in the formula are set by the person skilled in the art according to the actual situation.

Claims

1. A method for detecting anomalies in a multi-configuration robot joint, the method comprising: The method comprises the following steps: Step S1: acquiring real-time multi-dimensional sensing data of each joint of the robot, the multi-dimensional sensing data comprising joint motor current, joint vibration signal and joint encoder position data; performing forward kinematics calculation of the robot according to the joint encoder position data, constructing a real-time spatial linkage posture model of the robot, and generating robot real-time working condition state data comprising current spatial posture features; Step S2: performing whole-machine variable inertia and variable stiffness dynamics analysis based on the robot real-time working condition state data, constructing a spatial posture vibration transmission model that dynamically drifts with the posture; using the spatial posture vibration transmission model to quantitatively calculate the vibration wave conduction efficiency between each adjacent joint of the robot, and generating a joint-to-joint vibration modal coupling coefficient matrix corresponding to the current working condition; Step S3: using the joint-to-joint vibration modal coupling coefficient matrix to perform structure conduction noise decoupling processing on the joint vibration signal, eliminating the coupled vibration components transmitted to the target joint through the mechanical linkage from other joints, extracting the intrinsic vibration feature vector representing the health status of the target joint, and combining the joint motor current to generate multi-dimensional decoupled joint feature data; Step S4: calling a preset dynamic working condition reference threshold library according to the robot real-time working condition state data, using the dynamic working condition reference threshold library to perform self-adaptive abnormality discrimination on the multi-dimensional decoupled joint feature data, locating the specific joint where the abnormality occurs and identifying the abnormality type, and thus generating a robot joint abnormality detection result report.

2. The multi-configuration robot joint anomaly detection method of claim 1, wherein, Step S1 comprises the following steps: Step S11: synchronously collecting original running data of the robot through acceleration sensors, current transformers and absolute value encoders installed at each joint of the robot; Step S12: time stamp aligning and preprocessing the original running data to obtain multi-dimensional sensing data; Step S13: extracting joint angle information in the multi-dimensional sensing data, establishing a robot linkage coordinate system by using the D-H parameter method, and calculating spatial coordinates of the robot end and each linkage center of gravity; Step S14: calculating instantaneous speed and acceleration of the robot according to the spatial coordinates and the rate of change of the joint angle, combining the spatial coordinates, the instantaneous speed and the acceleration to generate robot real-time working condition state data.

3. The multi-configuration robot joint anomaly detection method of claim 1, wherein, Step S2 comprises the following steps: Step S21: calling robot rigid body dynamics parameters according to the linkage spatial position in the robot real-time working condition state data, and calculating robot whole-machine mass distribution matrix and moment of inertia matrix in real time; Step S22: performing system natural frequency drift analysis based on the moment of inertia matrix and in combination with the robot joint flexibility coefficient, to obtain real-time system modal frequency data that changes with the posture; Step S23: using the transfer function method to analyze the attenuation and amplification characteristics of vibration energy in the linkage structure according to the real-time system modal frequency data and the linkage geometric length, and constructing a spatial posture vibration transmission model; Step S24: using the spatial posture vibration transmission model to calculate the vibration response transmission ratio between any two joints, and combining the transmission ratios of all joint pairs to generate a joint-to-joint vibration modal coupling coefficient matrix.

4. The multi-configuration robot joint anomaly detection method of claim 3, wherein, Step S24 comprises the following steps: Step S241: Select the i th joint as the vibration source joint, select the j th joint as the measured joint, and extract the connecting rod transmission path parameters between the two joints under the current working condition; Step S242: input a unit pulse excitation into the vibration source joint position in the spatial pose vibration transmission model, and simulate the response signal amplitude at the measured joint; Step S243: calculate the ratio of the response signal amplitude to the unit pulse excitation to obtain the single-path vibration transmission factor; Step S244: traverse all joint combinations of the robot, calculate the corresponding single-path vibration transmission factor, and arrange it according to the row-column rule to generate the joint-to-joint vibration modal coupling coefficient matrix.

5. The multi-configuration robot joint anomaly detection method of claim 1, wherein, Step S3 includes the following steps: Step S31: frequency domain transformation is performed on the joint vibration signal to obtain original joint vibration frequency spectrum data; Step S32: based on the joint-to-joint vibration modal coupling coefficient matrix, identify and calculate the cross-talk vibration components received by the current joint from adjacent joints and remote joints; Step S33: subtract the cross-talk vibration components from the original joint vibration frequency spectrum data, perform frequency spectrum purification reconstruction, and obtain the decoupled target joint intrinsic vibration frequency spectrum; Step S34: time domain inverse transformation and feature extraction are performed on the target joint intrinsic vibration frequency spectrum to calculate the effective value, kurtosis index and energy barycenter, and generate the intrinsic vibration feature vector; Step S35: feature fusion is performed on the intrinsic vibration feature vector and the frequency domain features of the joint motor current to generate multi-dimensional decoupled joint feature data.

6. The multi-configuration robot joint anomaly detection method of claim 5, wherein, Step S32 includes the following steps: Step S321: obtain the original vibration intensity data of all joints except the target joint; Step S322: index the row vector corresponding to the target joint in the joint-to-joint vibration modal coupling coefficient matrix to obtain the input coupling coefficient set of the target joint; Step S323: use the input coupling coefficient set to perform weighted convolution operation on the original vibration intensity data of all other joints to simulate the interference waveform caused by structural conduction; Step S324: linearly superimpose the simulated interference waveform to calculate the cross-talk vibration component at the current joint.

7. The multi-configuration robot joint anomaly detection method of claim 1, wherein, Step S4 includes the following steps: Step S41: according to the speed and load information in the real-time working condition state data of the robot, interval division is performed on the robot running working condition to determine the working condition level to which the current belongs; Step S42: index the corresponding dynamic working condition reference threshold library from the preset database according to the working condition level, and the reference threshold library includes the upper and lower limits of the health features under the working condition; Step S43: input the multi-dimensional decoupled joint feature data into the dynamic working condition reference threshold library for comparison and analysis to calculate the feature deviation; Step S44: when the feature deviation exceeds the preset safety margin, it is determined as an abnormal state, and the abnormal joint number is output; Step S45: according to the spectral distribution characteristics of the feature deviation, match the fault mode library to generate a robot joint abnormality detection result report.

8. The multi-configuration robot joint anomaly detection method of claim 7, wherein, Step S43 includes the following steps: Step S431: extract the intrinsic vibration kurtosis value and current harmonic distortion rate in the multi-dimensional decoupled joint feature data; Step S432: Obtain the corresponding health reference mean and standard deviation in the dynamic working condition reference threshold library; Step S433: Calculate the distance of the intrinsic vibration kurtosis value and the current harmonic distortion rate relative to the health reference mean using the Mahalanobis distance algorithm; Step S434: Normalize the calculated distance to obtain the feature deviation degree.

9. A multi-configuration robot joint anomaly detection system, characterized by, The system is implemented based on the multi-working-condition robot joint abnormality detection method of any one of claims 1-8, and the system comprises: A working condition data acquisition module: acquires real-time multi-dimensional sensing data of each joint of the robot; performs forward kinematics operation on the joint encoder position data to construct a real-time robot space link posture model and generate real-time working condition state data of the robot; A coupling matrix generation module: performs whole-machine variable inertia and variable stiffness dynamics analysis based on the real-time working condition state data of the robot to construct a space posture vibration transmission model that dynamically drifts with the posture; and quantitatively calculates the vibration wave conduction efficiency between each adjacent joint of the robot using the space posture vibration transmission model to generate a joint vibration modal coupling coefficient matrix corresponding to the current working condition; A feature data generation module: decouples the structural conduction noise of the joint vibration signal using the joint vibration modal coupling coefficient matrix to extract an intrinsic vibration feature vector representing the health status of the target joint, and generates multi-dimensional decoupled joint feature data in combination with the joint motor current; A joint abnormality discrimination module: calls a preset dynamic working condition reference threshold library according to the real-time working condition state data of the robot to perform adaptive abnormality discrimination on the multi-dimensional decoupled joint feature data, locates the specific joint where the abnormality occurs and identifies the abnormality type, and thus generates a robot joint abnormality detection result report.

Images