Industrial robot motor fault early warning method and system

Abstract

The invention relates to the technical field of industrial motor faults, in particular to an industrial robot motor fault early warning method and system. The method comprises the following steps: acquiring a motor control instruction of the industrial robot in real time, and then judging an instruction value change event to obtain a motor load change event; monitoring transient working condition characteristic data of the motor in real time according to the motor load change event; flexible gear jump response abnormity evaluation is carried out based on the motor transient state working condition characteristic data, and a flexible gear comprehensive response abnormity score is generated; electromagnetic fault feature analysis is carried out according to the transient working condition feature data of the motor to obtain electromagnetic fault feature coupling data; and according to the electromagnetic fault feature coupling data and the flexible gear comprehensive response abnormal score, performing motor flexible gear health state analysis so as to realize industrial robot motor fault early warning. According to the invention, through working condition jump point detection and motor flexible gear health state analysis, the motor fault can be accurately identified, and multi-stage fault early warning is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial motor faults, and particularly to a method and system for early warning of industrial robot motor faults. Background Art

[0002] Due to its advantages such as high transmission accuracy, small size, light weight, and large transmission ratio, the harmonic reducer is widely used in the joint drive system of industrial robots. Its core component, the flexspline, realizes power transmission through flexible deformation, and its working state can directly reflect the load condition of the motor. Motor faults, especially common winding turn-to-turn short circuit faults, will cause fluctuations in the motor output torque, which in turn affects the deformation law of the flexspline. Therefore, by monitoring the deformation state of the flexspline, the health status of the motor can be reflected. However, traditional methods for early warning of industrial robot motor faults mainly focus on analyzing signals such as current, voltage, vibration, and temperature of the motor itself. These methods can achieve good results under constant working conditions, but under complex working conditions such as variable speed and variable load, signal fluctuations caused by load changes often mask fault characteristics, resulting in false alarms or missed alarms. For example, a winding turn-to-turn short circuit fault in the motor will cause changes in the motor current and vibration signals, but under sudden load changes, the current and vibration fluctuations caused by load changes are similar to fault characteristics and are difficult to distinguish. This makes it difficult to set early warning thresholds for traditional fault early warning methods when dealing with variable working conditions, with poor adaptability and unable to meet the requirements of high-reliability operation of industrial robots. Summary of the Invention

[0003] Based on this, the present invention provides a method and system for early warning of industrial robot motor faults to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for early warning of industrial robot motor faults includes the following steps:

[0005] Step S1: Obtain the industrial robot motor control instruction in real time, and then judge the instruction value change event to obtain the motor load change event; monitor the motor transient working condition characteristic data in real time according to the motor load change event;

[0006] Step S2: Perform working condition jump flexspline deformation analysis based on the motor transient working condition characteristic data to generate working condition jump flexspline deformation data; perform flexspline jump response abnormality evaluation according to the working condition jump flexspline deformation data to generate a comprehensive flexspline response abnormality score;

[0007] Step S3: Perform electromagnetic fault characteristic analysis according to the motor transient working condition characteristic data to obtain the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum respectively; perform multi-channel monitoring fault synchronous coupling on the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum to obtain electromagnetic fault characteristic coupling data;

[0008] Step S4: Analyze the health status of the motor flexspline based on the electromagnetic fault feature coupling data and the abnormal score of the comprehensive response of the flexspline, and generate motor fault status data; perform multi-level fault warning processing based on the motor fault status data to achieve motor fault warning for industrial robots.

[0009] Preferably, the present invention also provides an industrial robot motor fault warning system that executes the above-mentioned industrial robot motor fault warning method. The industrial robot motor fault warning system includes:

[0010] A transient working condition monitoring module, configured to obtain the motor control instruction of the industrial robot in real time, and then judge the instruction value change event to obtain the motor load change event; monitor the motor transient working condition characteristic data in real time according to the motor load change event;

[0011] A flexspline response analysis module, configured to perform flexspline deformation analysis during working condition jump based on the motor transient working condition characteristic data, and generate flexspline deformation data during working condition jump; perform abnormal evaluation of the flexspline jump response according to the flexspline deformation data during working condition jump, and generate an abnormal score of the comprehensive response of the flexspline;

[0012] An electromagnetic feature coupling module, configured to perform electromagnetic fault feature analysis according to the motor transient working condition characteristic data to obtain a current fault feature time series spectrum and a magnetic field distortion feature time series spectrum respectively; perform multi-channel monitoring fault synchronous coupling on the current fault feature time series spectrum and the magnetic field distortion feature time series spectrum to obtain electromagnetic fault feature coupling data;

[0013] A health evaluation and warning module, configured to analyze the health status of the motor flexspline based on the electromagnetic fault feature coupling data and the abnormal score of the comprehensive response of the flexspline, and generate motor fault status data; perform multi-level fault warning processing based on the motor fault status data to achieve motor fault warning for industrial robots.

[0014] In the solution of the present invention, by accurately capturing the changes in the motor control commands, the "events" of load or speed changes are actively identified, and the analysis focus is concentrated during these "transient operating conditions" where the motor state undergoes a jump. By specifically analyzing the transient characteristic data of the motor when responding to command changes, it is possible to more clearly identify the abnormal dynamic responses caused by internal faults (such as torque pulsations caused by inter-turn short circuits), distinguish them from the dynamic responses caused by normal load changes, and fundamentally solve the problems of difficult setting of early warning thresholds and frequent false alarms and missed alarms under variable operating conditions, significantly improving the reliability of fault early warning. The deformation of the flexspline directly reflects the motor load, and motor faults (especially inter-turn short circuits) will cause fluctuations in the output torque, which will inevitably change the deformation law of the flexspline. Based on this, the deformation data of the flexspline during the jump of the motor operating conditions is deeply analyzed to evaluate whether its dynamic response is "abnormal". This method has unique advantages: it directly focuses on the final link where the fault affects power transmission, that is, the actual deformation behavior of the flexspline. Compared with only analyzing the electrical signals or vibration signals of the motor body, monitoring the transient response of the flexspline can more intuitively and sensitively capture the small torque anomalies that are sufficient to affect the motion accuracy of the robot joint. Even if the manifestation of the fault in the motor electrical signals is not obvious or is submerged by noise, the disturbance of the torque fluctuation caused by it to the precise deformation mode of the flexspline can be effectively identified, thus improving the detection sensitivity to early and hidden faults. Through the synchronous coupling analysis of multi-dimensional electromagnetic characteristics, the robustness and specificity of fault feature extraction are enhanced. At the same time, two types of electromagnetic signals closely related to the state of the motor winding, namely the current and magnetic field distortion, are monitored, and their respective time series spectra are generated for feature analysis. This coupling analysis can effectively filter out random noise interference and form faults that are more unique and more difficult to be imitated by normal operating condition fluctuations. For example, an inter-turn short circuit will not only cause current harmonics at specific frequencies but also lead to specific distortions in the local magnetic field distribution. The synchronous abnormal patterns of these two in the transient response further reduce the possibility of misjudgment and improve the accuracy of fault identification. By integrating mechanical responses and electromagnetic characteristics for comprehensive health assessment and multi-level early warning, comprehensive and reliable motor state monitoring is achieved. Only when the suspected fault characteristics detected at the electromagnetic level also trigger corresponding abnormal responses at the mechanical transmission level (flexspline deformation) will the system give a highly confident fault judgment. This double verification mechanism greatly improves the confidence level of the final diagnosis result. The motor fault state data generated based on this comprehensive assessment can more accurately reflect the true health level of the motor. Furthermore, the implemented multi-level fault early warning processing can provide users with early warning information at different levels from "attention" to "emergency shutdown" according to the severity and development trend of the fault, support more refined predictive maintenance strategies, avoid the occurrence of catastrophic faults, ensure the high-reliability operation and long-term stability of industrial robots in complex operating environments, and ultimately improve production efficiency and safety.Therefore, a method for early warning of industrial robot motor faults according to the present invention accurately captures motor load change events by obtaining industrial robot motor control instructions in real time and judging instruction value change events, deeply analyzes the collected transient working condition data, including the analysis of the flexible gear deformation during the working condition jump based on ellipse fitting, the extraction and coupling of multi-channel electromagnetic fault characteristics, and the comprehensive health assessment combining mechanical and electromagnetic characteristics, realizes multi-dimensional and high-precision identification of fault characteristics, and effectively improves the accuracy and reliability of fault early warning. Brief Description of the Drawings

[0015] Figure 1 It is a schematic flow chart of the steps of a method for early warning of industrial robot motor faults according to the present invention;

[0016] Figure 2 is Figure 1 It is a schematic detailed implementation step flow chart of the analysis of the flexible gear deformation during the working condition jump based on the motor transient working condition characteristic data in step S2 in

[0017] The realization, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. Detailed Embodiments

[0018] The technical method of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0019] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0020] It should be understood that although terms such as "first" and "second" may be used here to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0021] To achieve the above object, please refer to Figures 1 to 2 , the present invention provides an industrial robot motor fault early warning method, including the following steps:

[0022] Step S1: Obtain the industrial robot motor control instruction in real time, and then judge the instruction value change event to obtain the motor load change event; Monitor the motor transient condition characteristic data in real time according to the motor load change event;

[0023] Step S2: Conduct a working condition jump flexible gear deformation analysis based on the motor transient condition characteristic data to generate working condition jump flexible gear deformation data; Evaluate the abnormal flexible gear jump response according to the working condition jump flexible gear deformation data to generate a comprehensive flexible gear response abnormal score;

[0024] Step S3: Conduct an electromagnetic fault characteristic analysis according to the motor transient condition characteristic data to obtain the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum respectively; Synchronously couple the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum for multi-channel monitoring faults to obtain electromagnetic fault characteristic coupling data;

[0025] Step S4: Analyze the health state of the motor flexible gear according to the electromagnetic fault characteristic coupling data and the comprehensive flexible gear response abnormal score to generate motor fault state data; Perform multi-level fault early warning processing according to the motor fault state data to achieve industrial robot motor fault early warning.

[0026] In the embodiment of the present invention, the industrial robot motor fault early warning method includes the following steps:

[0027] Step S1: Obtain the industrial robot motor control instruction in real time, and then judge the instruction value change event to obtain the motor load change event; Monitor the motor transient condition characteristic data in real time according to the motor load change event;

[0028] In an embodiment of the present invention, for example, for the third axis (shoulder pitch axis) of an ABB IRB 6700 industrial robot, this axis uses a servo motor controlled by an EtherCAT bus and a supporting harmonic reducer (assumed model CSG-40-120). First, through the EtherCAT master protocol stack deployed on an independent IPC, the target speed command PDO (Process Data Object, such as object dictionary 60FFh) broadcast from the servo driver of this axis (such as ABB ACSM1 driver) is subscribed and parsed in real time with a 2-millisecond cycle. When the robot performs an action of grasping a heavy workpiece (for example, the weight exceeds the calibrated normal load by 20%), and the controller issues a rapid acceleration command to instantly increase the motor speed from 500 revolutions per minute to 2500 revolutions per minute, the monitoring program processes the received original speed command sequence on the IPC. The latest sampling point (1678886400.134 seconds, 2500 revolutions per minute) and its previous point (1678886400.132 seconds, 2400 revolutions per minute) are extracted, and the sampling time interval is calculated to be 10 milliseconds (meeting the requirement of 8 - 12 ms), and the command value difference is 100 revolutions per minute. Based on the rated motor speed of 3000 revolutions per minute, the absolute value of the instantaneous change rate of the command is calculated as |(100 / 3000) × 100 / (10 ms / 1000 ms / s)| = 333.3% / s. This value is much greater than the preset determination threshold of 10% per second for rapid command change, so it is determined that a "motor load change event" has occurred (the T0 moment is 1678886400.134 seconds). This event immediately triggers the IPC to send a 5V, 100-microsecond synchronous acquisition trigger pulse signal through the digital output port of the NIPXIe-6363 data acquisition card. This pulse signal synchronously starts the parallel multi-channel monitoring of the installed sensors: three-phase current collected by a through-hole current sensor (50kHz sampling rate), magnetic field intensity at three points near the stator detected by a Hall array (50kHz sampling rate), displacement measured by four eddy current displacement sensors pasted on the inner wall of the flexspline (50kHz sampling rate), and the rotation angle obtained through the motor shaft encoder (synchronously read). The data acquisition system performs pre-trigger acquisition, automatically saves the data 50 milliseconds before the trigger, and continuously acquires the data 200 milliseconds after the trigger. After splicing these two segments of data, it is packaged together with the event trigger information (timestamp T0, change rate 333.3% / s, event description "grasping an overweight workpiece") to generate a complete record of "motor transient operating condition characteristic data".

[0029] Step S2: Based on the motor transient operating condition characteristic data, perform an analysis of the flexspline deformation during the operating condition jump to generate flexspline deformation data during the operating condition jump; according to the flexspline deformation data during the operating condition jump, perform an abnormal evaluation of the flexspline jump response to generate a comprehensive flexspline response abnormal score;

[0030] In the embodiments of the present invention, by using the data of four flexspline displacement sensors {di(t)} and the motor rotation angle data θ(t) included therein, at each sampling moment t (from T0 - 50 ms to T0 + 200 ms), the major axis a(t) and minor axis b(t) of the ellipse of the flexspline cross-section are calculated in real time through the direct least squares ellipse fitting algorithm. Thus, "flexspline ellipse parameter time series data" is generated. Then, based on this data and the steady-state baseline in the 50 ms before the event (for example, the calculated mean value of the major axis baseline μ_a = 19.95 mm), the major and minor axis deviation sequences {Δa(t)}, {Δb(t)} within the entire transient window are calculated. The maximum values of the absolute deviations are extracted, for example, max|Δa| = 0.18 mm, max|Δb| = 0.15 mm. Calculate the "flexspline comprehensive peak deviation amount" = sqrt(0.18^2 + 0.15^2) = 0.234 mm. Then, analyze the data after the event to determine the time when the flexspline shape returns to stability. By setting a stable threshold band (for example, the calculated major axis deviation stable band based on the last segment of data is [-0.01 mm, +0.01 mm]), starting from T0, it is found that the major and minor axis deviation sequences first enter and remain within their respective threshold bands at t_stable_start = T0 + 110 ms. Therefore, the "flexspline deformation stable duration" is 110 ms. In the interval [T0, T0 + 110 ms], peak detection is performed on the major axis deviation sequence {Δa(t)}, oscillation characteristic points are identified, and the average time interval between adjacent peaks (or valleys) is calculated to obtain the "average oscillation period" of 18.2 ms. Taking the reciprocal gives the "main oscillation frequency" of 55 Hz. These three indicators (comprehensive peak deviation amount 0.234 mm, stable duration 110 ms, main oscillation frequency 55 Hz) are packaged into "flexspline deformation data under working condition jump". Subsequently, an anomaly assessment is carried out: comparing 0.234 mm with the severity grading standard (assuming 0.15 - 0.25 mm is grade 3 "severe"), the peak deformation severity is obtained as grade 3. Comparing the stable duration of 110 ms with the healthy baseline under this working condition (assuming the mean value is 70 ms and the standard deviation is 5 ms), calculate the Z-score = (110 - 70) / 5 = 8.0, and the abnormal index of the deformation recovery time is 8.0 (significantly extended). Comparing the main oscillation frequency of 55 Hz with the fault frequency library, it is found that it falls within the "flexspline early crack characteristic frequency range" (45 - 55 Hz), and the risk mark is obtained as "suspected flexspline crack risk (high)". Finally, the severity (mapped to 8 points), the time anomaly index (mapped to 10 points), and the frequency risk (mapped to 9 points) are weighted and summed according to the weights (0.4, 0.3, 0.3), and the "flexspline comprehensive response abnormal score" = (0.4×8)+(0.3×10)+(0.3×9) = 3.2 + 3.0 + 2.7 = 8.9 points (out of 10 points).

[0031] Step S3: Conduct electromagnetic fault feature analysis based on the motor transient condition characteristic data to respectively obtain the current fault feature time-sequence spectrum and the magnetic field distortion feature time-sequence spectrum; perform multi-channel monitoring fault synchronous coupling on the current fault feature time-sequence spectrum and the magnetic field distortion feature time-sequence spectrum to obtain electromagnetic fault feature coupling data;

[0032] In the embodiment of the present invention, three-phase currents and three-point magnetic field intensity signals after the event (from T0 to T0 + 200 ms) are extracted. Using the three-phase currents and the motor electrical angle obtained synchronously (calculated from the motor rotation angle and the number of pole pairs), applying the Clarke-Park transformation, the direct-axis current and the quadrature-axis current are calculated. A fourth-order Butterworth high-pass filter with a cut-off frequency of 1 kHz is respectively applied to the direct-axis current and the quadrature-axis current to extract the high-frequency ripple signals {Id_ripple(t)} and {Iq_ripple(t)}. Then, perform short-time Fourier transform (STFT) on these two ripple signals during the main stage of power impact (e.g., from T0 to T0 + 100 ms), with the parameters set as a 1024-point Hanning window and 50% overlap. In the generated time-frequency spectrogram, check the frequency band from 2 kHz to 5 kHz. It is found that within the time period from T0 + 25 ms to T0 + 40 ms, the STFT result of Id_ripple(t) has a frequency component with an amplitude reaching 0.7 A at approximately 3.5 kHz, exceeding the threshold of 0.5 A (and the spectral resolution has ensured within 0.2 A). Therefore, this time period is marked as the "current fault feature region", and this marked time-frequency spectrogram is saved as the "current fault feature time-sequence spectrum". At the same time, perform sliding window (50 ms Hanning window, 50% overlap) FFT analysis on one of the magnetic field signals (e.g., Bx(t)). Track the change of the spectral peak within the frequency band from 100 Hz to 1 kHz. It is found that during the period from T0 + 20 ms to T0 + 70 ms, the amplitude of a spectral peak located at approximately 600 Hz changes by 7 mT compared to the previous window, exceeding the threshold of 5 mT. Mark the time window information of these significantly distorted areas on the time-frequency diagram to generate the "magnetic field distortion feature time-sequence spectrum". Finally, package the generated "current fault feature time-sequence spectrum" (including abnormal region markings), the "magnetic field distortion feature time-sequence spectrum" (including distortion window markings), and their common time reference information into a data structure to form the "electromagnetic fault feature coupling data".

[0033] Step S4: Conduct motor flexspline health state analysis based on the electromagnetic fault feature coupling data and the flexspline comprehensive response anomaly score to generate motor fault state data; perform multi-level fault warning processing based on the motor fault state data to achieve motor fault warning for industrial robots.

[0034] In the embodiment of the present invention, the electromagnetic data is evaluated: the current spectrum is checked and a marked area is found, so the "electromagnetic anomaly flag of the current channel" = 1. The magnetic field spectrum is checked and a distortion window is found, so the "electromagnetic anomaly flag of the magnetic field channel" = 1. The sub-channel flags {current: 1, magnetic field: 1} are obtained. Then, based on this flag and the flexspline fraction 8.9, according to the preset rule (for example: if the flexspline fraction >= 7 and (current = 1 or magnetic field = 1), then it is a high confidence level), it is determined as "high confidence level electromechanical combined anomaly", which is mapped to the "electromechanical combined anomaly confidence level" 0.9. Then, the time series correlation evaluation is carried out: the overlap degree between the flexspline peak response period (assuming from T0 to T0 + 20 ms) and the current anomaly period (T0 + 25 ms to T0 + 40 ms) is low, but it is close to the starting period of the magnetic field distortion (T0 + 20 ms). The comprehensive judgment is medium correlation. The confidence level is corrected (assuming the medium correlation correction factor is 1.0): the corrected single-event coupling anomaly degree = 0.9 × 1.0 = 0.9. Next, the risk level analysis is carried out: the anomaly degree 0.9 is compared with the risk threshold (assuming >= 0.9 is level 3 "handle immediately"), and the risk level is determined as "level 3". Finally, a "motor fault status data" report is generated, which includes: event timestamp, axis number 3, event type "grasping overweight workpiece", corrected coupling anomaly degree 0.9, risk level "level 3 - handle immediately", main basis "the flexspline deformation seriously exceeds the standard, the recovery time is extremely long, the main vibration frequency is suspected of crack characteristics, accompanied by significant magnetic field distortion", and the recommended operation "immediately stop the robot operation, arrange an emergency inspection of the harmonic reducer of the third axis, and focus on checking whether there are cracks in the flexspline". Based on the motor fault status data of this highest risk level (level 3), the system executes the highest level of early warning processing: sends an "emergency stop" instruction to the robot controller through the industrial network, and at the same time pops up a red-highlighted alarm window on the central monitoring system (SCADA) interface, details the fault location (third axis motor / reducer) and the risk assessment result, and automatically pushes a high-priority maintenance work order to the maintenance management system (CMMS), realizing accurate and timely early warning of industrial robot motor faults.

[0035] Preferably, the real-time acquisition of the industrial robot motor control instruction in step S1 and then the determination of the instruction value change event include:

[0036] Carry out control instruction digital sampling on the speed or load instruction signal output by the industrial robot motor control system in real time to obtain the original instruction sequence value;

[0037] Extract the instruction value of the current latest sampling point according to the original instruction sequence value, and obtain the corresponding sampling time point to obtain the current industrial robot motor control data;

[0038] Extract the instruction value and its timestamp of the previous sampling period adjacent to the current industrial robot motor control data by backtracking to obtain the data of the previous instruction time point;

[0039] Calculate the difference between the timestamps of the two based on the current industrial robot motor control data and the data of the previous instruction time point to obtain the sampling time interval; calculate the difference between the instruction values of the two based on the current industrial robot motor control data and the data of the previous instruction time point to obtain the instruction value difference; among them, the sampling time interval should be set within the range of 8 - 12 ms;

[0040] Calculate the instantaneous change rate of the instruction according to the sampling time interval and the instruction value difference, and perform absolute value processing to generate the absolute value of the instruction change rate;

[0041] Judge the instruction value change event for the absolute value of the instruction change rate through a preset instruction rapid change judgment threshold to obtain the motor load change event; among them, the preset instruction rapid change judgment threshold is set to 10% per second, and when the absolute value of the instruction change rate is greater than 10% per second, it is determined as the motor load change event.

[0042] In an embodiment of the present invention, for an industrial robot system equipped with six-axis joints and adopting the EtherCAT bus communication protocol, the motor speed control command of its third axis (usually responsible for the swing of the large arm and with significant load changes) is obtained in real time. The monitoring system is implemented by an independently deployed industrial control computer (IPC), which is installed with an EtherCAT master protocol stack. First, before the robot normally runs its predetermined pick-and-place operation program, configure the EtherCAT master on the IPC and connect it to the EtherCAT fieldbus of the robot as a monitoring node in the network. Subsequently, through the master configuration tool, accurately identify and subscribe to the process data object (PDO) broadcast by the servo drive unit that controls the third-axis motor. Specifically, subscribe to the PDO mapping entry containing the "TargetVelocity" command value. For example, the parameter with an address index of 60FFh (assuming the standard CiA402 drive profile) in the servo drive object dictionary. After starting the robot operation program, the master protocol stack on the IPC strictly synchronously receives the real-time status PDO message sent from the servo drive according to the preset EtherCAT communication cycle (for example, every 2 milliseconds), and parses out the 32-bit integer value representing the motor target speed from it. This value is the original speed command sent by the controller to the motor at the current moment. This process is carried out continuously to ensure continuous capture of the instruction stream. Use a high-precision system clock (for example, synchronized with the EtherCAT distributed clock or using the microsecond-level timer provided by the operating system) to timestamp each received speed command value. The specific operation is as follows: Whenever the EtherCAT master protocol stack successfully receives and parses the target speed value from the third-axis servo drive (for example, the value is 1500, unit: revolutions per minute), immediately call the system timing function to obtain the current timestamp. Store the speed value (1500 revolutions per minute) and the corresponding timestamp as a data pair. This data pair is sequentially added to a first-in, first-out (FIFO) memory buffer. The buffer is set to a fixed length. For example, store the most recent 1000 sampling points. As new data pairs are continuously generated and enter the buffer, the oldest data pair will be removed, thus forming a sliding window-style record of the original instruction sequence values. When it is necessary to judge change events (usually executed periodically or triggered by the arrival of new data), access the end of the FIFO buffer (the position of the latest added element). The specific operation is as follows: directly read the data pair contained in the last element of the buffer. For example, the current last data pair in the buffer. Extract this speed command value, that is, 1550 revolutions per minute, as the "instruction value of the current latest sampling point". At the same time, extract its corresponding timestamp, that is, 1678886400.133456789 seconds, as the "corresponding sampling time point".Look back to find the data immediately preceding the previous sample. The operation method is to access the penultimate element of the said FIFO buffer (i.e., the element preceding the tail element of the queue). Provided that there are at least two data points in the buffer, read the data pair at this position. For example, the penultimate data pair is (1678886400.123456789 seconds, 1500 revolutions per minute). Extract this speed command value, i.e., 1500 revolutions per minute, as the "command value of the previous sampling period". At the same time, extract its corresponding timestamp, i.e., 1678886400.123456789 seconds, as the "timestamp of the previous sampling period". These two values constitute the "data of the previous command time point". After obtaining the current data and the previous data, perform the following calculations: First, calculate the difference in timestamps: Sampling time interval ΔT = T_current - T_previous = 0.010000000 seconds. Convert this result to 10.0 milliseconds. Next, check to confirm whether this sampling time interval (10.0 milliseconds) falls within the preset valid range of 8 to 12 milliseconds. In this example, 10.0 milliseconds meets the condition (8 ≤ 10.0 ≤ 12), so the calculation is valid. Then, calculate the difference in command values: Command value difference ΔV = V_current - V_previous = 50 revolutions per minute. These two calculation results - the sampling time interval (10.0 milliseconds) and the command value difference (50 revolutions per minute) - will be used for the next calculation of the instantaneous change rate. If the time interval is not within this range, this calculation is marked as invalid or special processing is performed to avoid misjudgment caused by abnormal data acquisition timing. Convert the time interval to seconds: ΔT_s = ΔT_ms / 1000.0 = 0.01 seconds. Assume that the rated or maximum operating speed of this third-axis motor is V_rated = 3000 revolutions per minute. Instruction instantaneous change rate Rate = (ΔV / V_rated) × 100 / ΔT_s. Substitute the values: Rate = 166.67% / s. This calculation indicates that the speed command value has changed by 1.6667% relative to the rated speed within 0.01 seconds, and the change rate per second after conversion is 166.67% / s. Finally, take the absolute value of this instantaneous change rate to obtain the absolute value of the instruction change rate |Rate| = |166.67% / s| = 166.67% / s. Compare it with the preset "threshold for determining rapid change of instruction". This threshold is preset to 10% per second. The basis for setting this threshold is: Through the statistical analysis of a large amount of normal operation data of this type of industrial robot in its typical operating scenarios, it is found that during stable operation, the absolute value of the change rate of the motor speed command is usually much lower than 10% per second; while in the case of rapid acceleration and deceleration, sudden load changes (such as when grasping or releasing heavy objects instantaneously) that cause a large impact on the flexspline of the harmonic reducer, this change rate will increase significantly. Therefore, 10% per second is selected as the boundary for distinguishing normal fluctuations from potential high-load impact events.The comparison process is as follows: Determine whether |Rate| is greater than 10% per second. In this example, 166.67% / s > 10% per second. Since the absolute value of the calculated instruction change rate exceeds the preset threshold, the system determines that a "motor load change event" has occurred.

[0043] Preferably, the real-time monitoring of the motor transient operating condition characteristic data according to the motor load change event in step S1 includes:

[0044] Sending a synchronous trigger signal based on the motor load change event to obtain a synchronous acquisition trigger pulse signal;

[0045] Performing parallel multi-channel monitoring according to the synchronous acquisition trigger pulse signal to generate multi-channel original sampling point data; wherein, the multi-channel original sampling point data includes original electromagnetic time series signals and original flexspline displacement and rotation angle data;

[0046] Performing parallel original digital stream segment truncation on the multi-channel original sampling point data through a preset pre-event trace duration threshold to obtain pre-event pre-sampled data segments;

[0047] Taking the time point of the motor load change event as the event start time node, continuously extracting parallel original digital streams from the multi-channel original sampling point data until the total acquisition duration that covers the transient response and extends to the new steady state is reached to obtain post-event acquisition data segments;

[0048] Splicing the pre-event pre-sampled data segments and the post-event acquisition data segments in chronological order to obtain complete window original data;

[0049] Performing event trigger attribute processing on the motor load change event and the complete window original data, and performing transient operating condition characteristic integration to generate motor transient operating condition characteristic data.

[0050] In the embodiments of the present invention, when it is determined in the previous step that a "motor load change event" occurs (i.e., the absolute value of the instruction change rate exceeds 10% per second. For example, the change rate is detected as 166.67% / s at time T0), the monitoring main program deployed on the industrial control computer (IPC) immediately executes a preset trigger action. Specifically, this action is: through the digital output (DO) channel (e.g., Port0 / Line0) of a data acquisition (DAQ) card (e.g., a multifunctional IO card of model NIPXIe-6363) connected to the IPC, a digitally pulsed signal with precise time synchronization is generated. This pulsed signal is set to a high level (e.g., 5 V TTL level) and has a duration of 100 microseconds. The rising edge of this pulse strictly corresponds to the moment T0 when the load change event is detected (or within a very small, determined delay, such as within 10 microseconds). This 100-microsecond high-level pulse with a clear time reference, actively sent by the monitoring program to an external hardware device after the internal logic judgment conditions are met, is the "synchronous acquisition trigger pulsed signal". The output voltage signal is connected to three analog input (AI) channels (AI0, AI1, AI2) of the PXIe-6363, and the sampling rate is set to 50 kilohertz (kHz) with a resolution of 16 bits. At the same time, the PXIe-4331 module is used to collect the state data of the flexspline. A number of micro strain gauges are pasted at specific positions on the outer wall of the flexspline of the harmonic reducer (e.g., forming a full-bridge to measure torsional strain), and the strain gauge bridge is connected to the channel (e.g., CH0) of the PXIe-4331. This module performs high-precision strain acquisition at a sampling rate of 25.6 kHz. In addition, a high-resolution absolute encoder (e.g., with a 24-bit resolution) is installed at each of the motor shaft and the output end of the reducer, and its signal is connected to the encoder interface module in the chassis (e.g., connected to the counter / timer resources of the PXIe-6363 for high-speed position tracking) to calculate the real-time relative rotation angle of the flexspline (the output rotation angle minus the input rotation angle converted according to the reduction ratio). When the rising edge of the trigger pulsed signal is detected at the PFI0 terminal of the PXIe-6363, the system immediately starts the data acquisition of all configured channels simultaneously at the preset sampling rates (50 kHz for current, 25.6 kHz for strain, and the counter records the position value at the corresponding moment), and stores the sampled digital values (current value, strain value, encoder count value) and their precise timestamps (provided by the synchronous clock of the PXIe chassis) into the on-board buffer (FIFO) of their respective channels, forming a parallel, time-synchronized "multi-channel raw sampling point data" stream. A "pre-event retrospective duration threshold" is set, and its value is determined in advance based on the analysis of the transient response characteristics of the robot axis under typical working conditions, aiming to capture the stable or quasi-stable state before the event occurs as a baseline. For example, this threshold is preset to 50 milliseconds (ms).The specific implementation is as follows: Before receiving the trigger instruction, the data acquisition system continuously collects data from all channels and temporarily stores it in a ring buffer. The size of this buffer can accommodate at least 50 ms of data volume (for example, for a current channel with a sampling rate of 50 kHz, 50 kHz × 0.050 s = 2500 sampling points need to be stored; for a strain channel with 25.6 kHz, 25.6 kHz × 0.050 s = 1280 sampling points). When the synchronous acquisition trigger pulse signal (occurring at time T0) is confirmed by the DAQ system, the system automatically extracts and saves the sampling data of all channels during the period from T0 - 50 ms to T0. The exact time point T0 (i.e., the effective edge moment of the synchronous trigger pulse) when the motor load change event is detected is used as the "event start time node". After time T0, the data acquisition system continues to collect data from all channels at the set sampling rates (50 kHz for current, 25.6 kHz for strain, encoder position tracking) without interruption. A "total acquisition duration" needs to be preset in advance, and this duration should be sufficient to cover the complete dynamic response process (including impact, oscillation, and stabilization) of the motor and the flexspline caused by the load change and extend to reach the new steady-state condition. This value is determined based on the study of the robot's dynamic characteristics and typical operation cycles. For example, it is set to 200 milliseconds (ms). Therefore, the data acquisition system will start collecting data from time T0 and continue until time T0 + 200 ms. The parallel raw digital streams of all channels during the period from T0 to T0 + 200 ms form the "post-event acquisition data segment". The specific operation is as follows: For each monitoring channel (for example, the U-phase current channel), the sample sequence of its pre-event pre-acquisition data segment (for example, 2500 points) is directly concatenated with the sample sequence of the post-event acquisition data segment (for example, 10000 points) in chronological order. Ensure that the data at the connection point (time T0) is continuous (i.e., the last point of the previous segment is adjacent to the first point of the next segment in time). The same operation is performed on all channels (U / V / W-phase current, flexspline strain, motor / output shaft angle). The final obtained data is a continuous, time-synchronized multi-channel data record that includes all monitoring channels and has a time span from T0 - 50 ms to T0 + 200 ms (total duration 250 ms). Perform "event trigger attribute processing": Attach the relevant information of the "motor load change event" that triggered this data acquisition (for example, the exact timestamp T0 of the event occurrence, the absolute value of the instruction change rate that caused the trigger 166.67% / s, the robot axis number where the event occurred: the third axis, the current task phase of the robot operation: for example, "instantly grasping the workpiece") as metadata to this data window.Then, perform "instantaneous operating condition feature integration": Create a structured data record (for example, create a new record in a database, or generate a file in a specific format such as an HDF5 or MAT file). This record consists of two parts: one part is the complete, multi-channel "raw data of the complete window" itself (stored, for example, as a multi-dimensional array or multiple parallel time series); the other part is additional metadata that details the attributes of the triggering event. This structured data packet that integrates the original time-series data and the context information of the triggering event is the finally generated "motor instantaneous operating condition feature data".

[0051] Preferably, the parallel multi-channel monitoring according to the synchronous acquisition trigger pulse signal includes:

[0052] Precisely install the current sensor, the magnetic sensor array, and the non-contact displacement sensor to the corresponding positions of the motor input line, the predetermined monitoring points of the stator winding, and the inner wall of the flexspline, respectively, and physically connect the signal output terminals of each sensor to the corresponding input ports of the data acquisition device to obtain a connected sensor port group;

[0053] Based on the synchronous acquisition trigger pulse signal, perform high-frequency synchronous sampling of the current and voltage of the three-phase input line of the industrial robot motor through the connected sensor port group, and perform multi-point synchronous detection of the magnetic field intensity near the motor housing or the motor air gap to obtain the original electromagnetic time-series signal;

[0054] Based on the synchronous acquisition trigger pulse signal, perform non-contact multi-point position displacement synchronous measurement on the circumferential direction of the outer wall of the flexspline in the industrial robot motor through the connected sensor port group, and synchronously obtain the motor rotation angle signal to obtain the original flexspline displacement and rotation angle data.

[0055] In the embodiments of the present invention, three LEMLA55-P Hall effect current sensors are selected and adopted in a through-hole structure. They are respectively and precisely clamped and fixed on the three-phase (U, V, W) power input cables of the motor driving the third axis (assumed to be the monitored target axis), ensuring that the cables pass completely through the center of the sensor magnetic core window. Secondly, an array module containing three Melexis MLX91208 programmable linear Hall sensors is selected. Through a custom-designed high-temperature-resistant non-metallic fixing bracket, it is precisely installed at a predetermined monitoring point near the end winding of the motor stator. For example, it is evenly distributed at 120-degree intervals in the circumferential direction and 3 millimeters away from the winding surface in the radial direction, for detecting the internal leakage magnetic field of the motor. Furthermore, four Micro-Epsilon eddyNCDT3005 eddy current displacement sensors are selected. Using a specially designed mounting base that can withstand the internal environment of the reducer, the probes are precisely fixed on the stationary parts of the harmonic reducer (such as the rigid gear or the housing), so that the measuring surfaces of the probes face specific circumferential positions on the inner wall of the flexspline (for example, evenly distributed at the 0-degree, 90-degree, 180-degree, and 270-degree azimuths where the major and minor axes of the deformed ellipse of the flexspline are expected to pass), and the initial gap between the probes and the inner wall of the flexspline is adjusted to 1 millimeter. After all the sensor bodies are installed, the signal output lines of each sensor (the current sensor outputs ±10V voltage signals, the Hall sensor array outputs 0-5V voltage signals, and the eddy current displacement sensor outputs 0-10V voltage signals) are physically connected to the designated analog input (AI) ports of the NI PXIe-6363 multifunctional data acquisition card using double-twisted cables with high shielding performance through BNC or connectors of the same quality. After receiving the synchronous acquisition trigger pulse signal (a 5V TTL level pulse lasting 100 microseconds), this signal is input to the PFI0 (programmable function interface) trigger input terminal of the NI PXIe-6363 data acquisition card. The data acquisition system (controlled by the LabVIEW program running on the PXI controller) is configured to immediately start a multi-channel synchronous analog acquisition task when a rising edge appears at the PFI0 port. This task operates on the channels assigned to the electromagnetic signals in the "connected sensor port group": through the AI0, AI1, and AI2 ports, the voltage signals output by the three-phase input current sensors are synchronously converted from analog to digital at a sampling rate of 50,000 times per second (50kHz) and a resolution of 16 bits. At the same time, if voltage monitoring is included in the system configuration, through an additional configured voltage transformer or high-voltage differential probe (for example, connected to AI10, AI11, AI12), the three-phase input line voltages are also synchronously acquired at a rate of 50kHz and a resolution of 16 bits. And through the AI3, AI4, and AI5 ports, the voltage signals (representing the local magnetic field intensity) output by the three Hall sensors installed near the stator winding are synchronously sampled at the same rate of 50kHz and a resolution of 16 bits.The sampling operations of all these channels are strictly synchronously driven by the same high-precision on-board clock (such as a 100 MHz time base) on the PXIe-6363 card, ensuring that at each sampling time point, all electromagnetic-related physical quantities are measured simultaneously. The acquired continuous and time-precisely aligned digital sequences (including the U-phase current value sequence, V-phase current value sequence, W-phase current value sequence, U-phase voltage value sequence, V-phase voltage value sequence, W-phase voltage value sequence, and the magnetic field intensity value sequences at three measurement points) together constitute the "original electromagnetic time-series signal". Start the data acquisition task for the flexible gear status monitoring channel. The specific operation is as follows: Through the AI6, AI7, AI8, and AI9 ports in the "connected sensor port group", synchronously perform analog-to-digital conversion on the voltage signals output by four eddy current displacement sensors installed inside the harmonic reducer at a sampling rate of 50,000 times per second (50 kHz) and a 16-bit resolution. These signals directly reflect the real-time distance (displacement) of the inner wall of the flexible gear relative to the sensors at four fixed circumferential positions. At the same time, the system synchronously acquires the motor rotation angle signal. This operation is achieved by connecting a high-resolution absolute encoder (such as a BiSS-C interface encoder with a 24-bit resolution) to the motor shaft end, and the encoder signal is connected to a dedicated encoder interface module configured inside the PXI chassis (or utilize the high-speed counter / timer resources on the PXIe-6363 card). At the moment when the trigger signal arrives and at each subsequent sampling clock cycle, the system precisely records the current absolute angle value reported by the encoder. The synchronous sampling mechanism of the PXIe-6363 card ensures that the sampling of the four displacement channels is precisely aligned in time with the reading of the encoder angle value (with an error at the nanosecond level). The acquired continuous and time-synchronized digital sequences (including the flexible gear displacement value sequences at four measurement points and the motor rotation angle value sequence) together constitute the "original flexible gear displacement and rotation angle data".

[0056] As an example of the present invention, refer to Figure 2 shown in Figure 1 is a schematic diagram of the detailed implementation step process for analyzing the flexible gear deformation during the working condition jump based on the motor transient working condition characteristic data described in step S2. In this example, the analysis of the flexible gear deformation during the working condition jump based on the motor transient working condition characteristic data described in step S2 includes:

[0057] S21: Based on the motor transient working condition characteristic data, perform elliptical geometric fitting on the cross-sectional profile of the flexible gear, and then extract the flexible gear elliptical parameter time series to generate flexible gear elliptical parameter time series data;

[0058] In the embodiments of the present invention, at each synchronous sampling time point t, the motor rotation angle θ(t) at this moment and the displacement values d1(t), d2(t), d3(t), d4(t) of the inner wall of the flexspline measured by four eddy current displacement sensors are extracted. The fixed installation angles of the four sensors are known to be 0 degrees, 90 degrees, 180 degrees, and 270 degrees (relative to a certain reference coordinate system). Combining the motor rotation angle θ(t), the instantaneous angles of these four measurement points on the flexspline relative to the main axis direction of the wave generator at this moment can be calculated. Based on these four spatial position points (the angles are known, and the radial positions are calculated from the displacement sensor readings d_i(t) combined with the sensor installation positions and the nominal radius of the flexspline), the "direct least squares ellipse fitting" algorithm is used. This algorithm solves a constrained optimization problem to find an optimal ellipse equation Ax^2 + Bxy + Cy^2 + Dx + Ey + F = 0 to fit these four (or more, if there are more sensors) data points. This fitting process is repeated for each sampling time point in the "complete window raw data" (time span from T0 - 50 ms to T0 + 200 ms). From the coefficients (A, B, C, D, E, F) of each fitted ellipse equation, the geometric parameters of the flexspline cross-section at this moment are calculated: the ellipse center coordinates (xc(t), yc(t)), the major axis length a(t), the minor axis length b(t), and the direction angle φ(t) of the ellipse major axis relative to the reference coordinate system. These parameters (a(t), b(t), xc(t), yc(t), φ(t)) that vary with time are collected to form a time series, which is the "time series data of flexspline ellipse parameters".

[0059] S22: Perform pre-event steady-state baseline statistics based on the motor transient condition characteristic data to generate pre-event steady-state baseline characteristic data;

[0060] In the embodiments of the present invention, for the multi-channel raw data during this time period and the flexspline elliptical parameters (e.g., the major semi-axis length a(t)) during the corresponding time period, statistical calculations are performed to establish a steady-state baseline. The specific operation is as follows: Select all N sampling values of the flexspline major semi-axis length a(t) within the time window from T0 - 50 ms to T0 (e.g., for a sampling rate of 50 kHz and a 50 ms window, N = 2500 points). Calculate its arithmetic mean μ_a = (1 / N) × Σa(t) and standard deviation σ_a = sqrt[(1 / (N - 1)) × Σ(a(t) - μ_a)^2]. Similarly, for other key signals such as the effective values of the three-phase currents of the motor (e.g., the effective value of the U-phase current I_U_rms) and the motor speed (which can be obtained by differentiating the angular displacement data) during this time period, their respective means and standard deviations are also calculated. For example, calculate the mean μ_I_U_rms and standard deviation σ_I_U_rms of the effective value of the U-phase current within this window. These calculated statistics (such as μ_a, σ_a, μ_I_U_rms, σ_I_U_rms, etc.) are grouped together to form the "steady-state baseline characteristic data before the event". This data set quantifies the average level and fluctuation range of various indicators of the robot motor and flexspline when operating in a relatively stable state before the load change event occurs.

[0061] S23: Perform post-event dynamic response analysis based on the motor transient operating condition characteristic data to generate post-event dynamic response characteristic data;

[0062] In the embodiments of the present invention, dynamic characteristic analysis is performed on the multi-channel raw data and the corresponding time-series data of the flexspline elliptical parameters (e.g., the major axis length a(t)) during this time period. The specific operations are as follows: For the sequence of the major axis length a(t) of the flexspline, first determine its maximum peak value a_peak and the time t_peak when it appears after the moment T0. Then, determine the new steady-state value a_final that the signal enters after the transient process ends (which can be obtained by calculating the average value within the time period from T0 + 150 ms to T0 + 200 ms). Calculate the overshoot Overshoot = (a_peak - a_final) / a_final × 100%. Define a stability criterion, for example, the signal enters and remains within the error band of a_final ± 2%, and find the time point t_settling when the signal last enters this error band. Then the response settling time (SettlingTime) is Ts = t_settling - T0. Apply the fast Fourier transform (FFT) to the data of a(t) from T0 to t_settling, and identify the frequency component with the largest amplitude as the main oscillation frequency f_osc. Perform similar analysis on other key signals such as the motor current and speed, and extract their dynamic characteristics such as peak value, settling time, and oscillation frequency. Integrate all these dynamic indexes (such as a_peak, Ts_a, f_osc_a, I_peak, Ts_I, f_osc_I, etc.) extracted from the post-event data segment to generate the "post-event dynamic response characteristic data".

[0063] S24: Perform flexspline deformation analysis during the working condition jump on the time-series data of the flexspline elliptical parameters based on the pre-event steady-state baseline characteristic data and the post-event dynamic response characteristic data, and generate the flexspline deformation data during the working condition jump, where the flexspline deformation data during the working condition jump includes the comprehensive peak deviation amount of the flexspline, the flexspline deformation stable duration, and the main oscillation frequency.

[0064] In an embodiment of the present invention, an index comprehensively reflecting the severity of deformation is defined. For example, calculate the major axis peak deviation Δa_peak = a_peak - μ_a and the minor axis peak deviation Δb_peak = b_peak - μ_b (note that the minor axis peak can be the minimum value, and in this case, the deviation is μ_b - b_min). A comprehensive deviation amount can be Dev_peak_comprehensive = sqrt((Δa_peak)^2+(Δb_peak)^2), or directly use the major axis peak deviation Dev_peak_comprehensive = |Δa_peak| that can better reflect stress concentration. Take Dev_peak_comprehensive = |a_peak - μ_a| as an example, and record this value (for example, calculated to be 0.05 mm) as the "flexspline comprehensive peak deviation amount". The adjustment time Ts_a of the flexspline geometric parameter (such as the major axis a(t)) obtained by using the post-event dynamic response analysis. For example, if Ts_a = 85 ms is calculated, then record this value as the "flexspline deformation stabilization duration". This represents the time required for the flexspline shape to recover from the impact to a new steady state (or close to the steady state). The main oscillation frequency f_osc_a obtained by performing FFT analysis on the flexspline geometric parameter (such as the major axis a(t)) in the post-event dynamic response analysis. For example, if the FFT analysis shows that the main oscillation frequency of a(t) during the transient process is 120 Hz, then record this value as the "main oscillation frequency". Package these three specifically calculated values (flexspline comprehensive peak deviation amount = 0.05 mm, flexspline deformation stabilization duration = 85 ms, main oscillation frequency = 120 Hz) to form a record of "flexspline deformation data under working condition jump". This data quantifies the dynamic deformation characteristics of the flexspline under a specific load change event.

[0065] Preferably, the analysis of flexspline deformation under working condition jump for the flexspline elliptical parameter time series data based on the pre-event steady-state baseline characteristic data and the post-event dynamic response characteristic data includes:

[0066] Based on the flexspline elliptical parameter time series data and the pre-event steady-state baseline characteristic data, calculate the deviations of the major axis and minor axis from their respective baseline values at each sampling moment within a preset transient window, respectively, to obtain the flexspline major axis deviation sequence and the flexspline minor axis deviation sequence;

[0067] Take the absolute values of the flexspline major axis deviation sequence and the flexspline minor axis deviation sequence respectively, and extract their respective maximum values within the preset transient window to obtain the absolute value of the maximum major and minor axis deviation;

[0068] Perform flexspline comprehensive peak processing according to the absolute value of the maximum major and minor axis deviation to obtain the flexspline comprehensive peak deviation amount;

[0069] Extract the data of the last segment in the preset transient window according to the timing data of the flexspline elliptical parameters, and perform the analysis of the steady-state deviation determination threshold for the major axis deviation sequence and the minor axis deviation sequence of the flexspline to generate a deviation stability determination threshold band;

[0070] Extract the event trigger time point according to the motor transient condition characteristic data;

[0071] Based on the deviation stability determination threshold band, search backward from the event trigger time point to obtain the starting time point when both the major axis deviation sequence and the minor axis deviation sequence first enter and remain within their respective threshold bands, and obtain the starting moment of deformation stability;

[0072] Analyze the flexspline deformation stability duration according to the starting moment of deformation stability;

[0073] Based on the flexspline deformation stability duration, identify the oscillation behavior of the dynamic response characteristic data after the event to generate a deviation oscillation characteristic point sequence;

[0074] Calculate the average oscillation period according to the deviation oscillation characteristic point sequence;

[0075] Take the reciprocal of the average oscillation period to calculate and obtain the main oscillation frequency.

[0076] In the embodiments of the present invention, "flexure ring elliptical parameter time-series data" (including the major axis sequence a(t) and the minor axis sequence b(t), with a time range from T0 - 50 ms to T0 + 200 ms) and "pre-event steady-state baseline feature data" (including the major axis baseline average value μ_a and the minor axis baseline average value μ_b) are utilized. For each sampling time point t within the "preset transient window" (i.e., the complete data window from T0 - 50 ms to T0 + 200 ms), deviation calculations are performed. The specific operation is as follows: Subtract its corresponding steady-state baseline value μ_a from the major axis sequence a(t) to obtain the major axis deviation Δa(t) = a(t) - μ_a at this moment. Similarly, subtract its steady-state baseline value μ_b from the minor axis sequence b(t) to obtain the minor axis deviation Δb(t) = b(t) - μ_b at this moment. This calculation is repeated for all time points within the window to generate two new time series: one is the "flexure ring major axis deviation sequence" {Δa(t)}, and the other is the "flexure ring minor axis deviation sequence" {Δb(t)}. Take the absolute value of each value in these two deviation sequences to generate the "major axis deviation absolute value sequence" {|Δa(t)|} and the "minor axis deviation absolute value sequence" {|Δb(t)|}. Subsequently, within the "preset transient window" (T0 - 50 ms to T0 + 200 ms), search for the maximum values of these two absolute value sequences respectively. The specific operation is as follows: Traverse all the values in the {|Δa(t)|} sequence to find the maximum value, denoted as max|Δa|. Similarly, traverse all the values in the {|Δb(t)|} sequence to find the maximum value, denoted as max|Δb|. For example, if it is calculated that max|Δa| = 0.08 mm and max|Δb| = 0.06 mm. These two values respectively represent the maximum absolute amplitudes by which the major axis and minor axis of the flexure ring deviate from their respective steady-state baselines during the entire transient process, and they are recorded as the "maximum major and minor axis deviation absolute values". The Euclidean norm is used as the comprehensive method: The calculation formula is the flexure ring comprehensive peak deviation amount = sqrt((max|Δa|)^2 + (max|Δb|)^2), which is the "flexure ring comprehensive peak deviation amount" of this transient event, and it comprehensively considers the maximum deformation degrees in both the major axis and minor axis directions. For example, select the data in the time period from T0 + 150 ms to T0 + 200 ms, and consider this section to represent the system entering a new steady state or approaching a steady state. At the same time, use the corresponding "flexure ring major axis deviation sequence" Δa(t) and "flexure ring minor axis deviation sequence" Δb(t) in this section. First, calculate the average value μ_Δa_final and the standard deviation σ_Δa_final of Δa(t) in this final section (T0 + 150 ms to T0 + 200 ms). Similarly, calculate the average value μ_Δb_final and the standard deviation σ_Δb_final of Δb(t) in this final section. Then, define the stable threshold band based on these statistics.Set a determination coefficient k (for example, k = 3, which is predetermined based on an assessment of the noise level and the required stable confidence level). Then, the stable threshold band for the major axis deviation is defined as [μ_Δa_final - k×σ_Δa_final, μ_Δa_final + k×σ_Δa_final]. Similarly, the stable threshold band for the minor axis deviation is [μ_Δb_final - k×σ_Δb_final, μ_Δb_final + k×σ_Δb_final]. These two interval ranges are the "deviation stable determination threshold band". Locate the exact timestamp when the recorded "motor load change event" occurs. When the absolute value of the "instruction change rate" first exceeds the preset threshold (10% per second), the real-time monitoring system precisely records it and uses it as the time point for generating the synchronous acquisition trigger pulse. Check the deviation values Δa(t) and Δb(t) of each sampling point in chronological order. The goal of the search is to find the first time point t_stable_start that satisfies the following conditions: for all time points t' from t_stable_start to the end of the transient window (T0 + 200 ms), it must be simultaneously satisfied that Δa(t') falls within the major axis threshold band [0.004, 0.016] and Δb(t') falls within the minor axis threshold band [-0.008, -0.002]. Specifically, when implementing, it can start traversing from T0. Once a point t that satisfies both conditions is found, tentatively set t as t_stable_start, and then continue to check backward until the end of the window. If any point in the middle does not satisfy the conditions, abandon the current t and continue to search for a new candidate t_stable_start backward. The first time point that passes the verification of all subsequent points is the one sought. For example, through search, it is found that t = T0 + 85 milliseconds is the first moment that satisfies this continuous stability condition. Then, t_stable_start = T0 + 85 ms, which is the "start moment of deformation stability". Based on the "start moment of deformation stability" t_stable_start and the "event trigger time point" T0. Calculate the time difference between the two to obtain the duration from when the flexspline is impacted to when its shape returns to stability, which is the quantified "flexspline deformation stability duration". For example, select the "flexspline major axis deviation sequence" {Δa(t)} with usually larger deviation amplitudes. The analysis time range is from the "event trigger time point" T0 to the moment t_stable_start corresponding to the "flexspline deformation stability duration" (i.e., the interval [T0, t_stable_start]). Within this interval, apply a peak detection algorithm (for example, based on the zero point of the first derivative and the sign judgment of the second derivative, or a simple neighborhood comparison method) to identify all local maximum points (peaks) and local minimum points (valleys) in the {Δa(t)} sequence. Record the timestamps corresponding to each found peak and valley and arrange them in chronological order.For example, within this 85 ms interval, it is recognized that the wave peaks occur at T0 + 15 ms, T0 + 35 ms, T0 + 55 ms, T0 + 75 ms, and the wave troughs occur at T0 + 25 ms, T0 + 45 ms, T0 + 65 ms. Using the wave peak time points in the above example: The first period P1 = (T0 + 35 ms) - (T0 + 15 ms) = 20 ms; the second period P2 = (T0 + 55 ms) - (T0 + 35 ms) = 20 ms; the third period P3 = (T0 + 75 ms) - (T0 + 55 ms) = 20 ms. Then calculate the arithmetic mean of these periods. In this example, the average oscillation period = (P1 + P2 + P3) / 3 = (20 ms + 20 ms + 20 ms) / 3 = 20 milliseconds. If using the wave trough time points: P1' = (T0 + 45 ms) - (T0 + 25 ms) = 20 ms; P2' = (T0 + 65 ms) - (T0 + 45 ms) = 20 ms. The average period (P1' + P2') / 2 = 20 ms. Therefore, the calculated "average oscillation period" is 20 milliseconds. Taking the reciprocal of the average oscillation period (which needs to be converted to seconds), the main oscillation frequency of the flexspline during the transient response process can be obtained. Using the results in the above example, the average oscillation period = 20 milliseconds = 0.020 seconds. Calculate its reciprocal: The main oscillation frequency = 1 / average oscillation period = 1 / 0.020 s = 50 Hertz (Hz).

[0077] Particularly important is that the end - segment data in the preset transient window is extracted according to the flexspline elliptical parameter time - series data, and the steady - state deviation determination threshold analysis of the major and minor axes is carried out according to the major - axis deviation sequence and minor - axis deviation sequence of the flexspline. Specifically:

[0078] Based on the timestamp information of the flexspline elliptical parameter time - series data, the end - time period of the preset transient window is determined, and based on the major - axis deviation sequence and minor - axis deviation sequence of the flexspline, the data points of these two sequences within this end - time period are respectively extracted, obtaining the major - axis deviation end - segment sequence and the minor - axis deviation end - segment sequence;

[0079] According to the major - axis deviation end - segment sequence and minor - axis deviation end - segment sequence, the arithmetic mean of the data points within each sequence is calculated respectively to estimate the average deviation level after reaching a new steady state, obtaining the major - and minor - axis new - steady - state center deviation estimated values;

[0080] Through a 5% small tolerance value of the peak deviation, a tolerance range analysis for determining the stable state is carried out on the major - and minor - axis new - steady - state deviation means and the flexspline comprehensive peak deviation amount, generating the major - and minor - axis deviation stable boundaries;

[0081] Combining the estimated value of the new steady-state center deviation of the major and minor axes and the stable boundary of the major and minor axis deviation to obtain a stable deviation determination threshold band.

[0082] In an embodiment of the present invention, according to the total duration of the preset transient window (250 ms) and the set length of the end time period (for example, set to the last 50 milliseconds), the specific time range of this end time period is determined to be from T0 + 150 ms to T0 + 200 ms. Then, traverse the timestamp sequence of the flexspline elliptical parameter time series through a programming script (for example, using the Python Pandas library to process time series data), and filter out all the time point indices that fall within the interval [T0 + 150 ms, T0 + 200 ms]. Using these indices, extract the corresponding data point subset from the complete {Δa(t)} sequence to form the "final segment sequence of major axis deviation". Similarly, use the same indices to extract the corresponding data points from the {Δb(t)} sequence to form the "final segment sequence of minor axis deviation". The specific operation is as follows: sum all 2500 data points in the "final segment sequence of major axis deviation", and then divide the sum by the number of data points (2500). The calculation formula is μ_Δa_final = (1 / 2500) × ΣΔa_end(t), where the summation traverses all time points t in the final segment sequence. For example, μ_Δa_final = 0.01 mm is calculated. Similarly, perform the same calculation on the "final segment sequence of minor axis deviation": μ_Δb_final = (1 / 2500) × ΣΔb_end(t). For example, μ_Δb_final = -0.005 mm is calculated. These two calculated average values (0.01 mm and -0.005 mm) respectively represent the central deviation levels at which the major axis and minor axis tend to be stable after the event, that is, the "estimated values of the new steady-state central deviations of the major and minor axes". The "rule of 5% of the peak deviation" is used to define the tolerance. First, calculate this tolerance value: Tolerance = 0.05 × the comprehensive peak deviation of the flexspline = 0.05 × 0.1 mm = 0.005 mm. Then, based on this tolerance value, expand the new steady-state central deviation of each axis up and down to generate the stability boundaries. The upper boundary of the major axis Upper_Bound_a = μ_Δa_final + Tolerance. The lower boundary of the major axis Lower_Bound_a = μ_Δa_final - Tolerance. Similarly, the upper boundary of the minor axis Upper_Bound_b = μ_Δb_final + Tolerance. The lower boundary of the minor axis Lower_Bound_b = μ_Δb_final - Tolerance. These four calculated values (0.015 mm, 0.005 mm, 0 mm, -0.010 mm) are the "stability boundaries of the major and minor axis deviations". The specific operation is: combine the lower boundary (0.005 mm) and the upper boundary (0.015 mm) of the major axis to define the stability determination threshold band of the major axis as a closed interval: Threshold_Band_a = [Lower_Bound_a, Upper_Bound_a].Similarly, the lower boundary (-0.010 mm) and the upper boundary (0 mm) of the minor axis are combined to define the stable determination threshold band of the minor axis: Threshold_Band_b = [Lower_Bound_b, Upper_Bound_b]. These two clearly defined range intervals together constitute the "deviation stable determination threshold band" of the final output.

[0083] Preferably, the evaluation of the abnormal response of the flexspline jump according to the flexspline deformation data during the working condition jump in step S2 includes:

[0084] Evaluating the severity of the peak deformation according to the comprehensive peak deviation of the flexspline;

[0085] Calculating the degree of abnormal extension or shortening of the stable duration of the flexspline deformation through the preset statistical duration baseline under the same working condition jump to obtain the abnormal index of the deformation recovery time;

[0086] Matching the failure mode according to the comprehensive peak deviation of the flexspline to obtain the risk mark of the flexspline vibration frequency;

[0087] Based on the severity of the peak deformation, the abnormal index of the deformation recovery time, and the risk mark of the flexspline vibration frequency, perform weighted evaluation of the abnormal response of the flexspline jump to generate the comprehensive response abnormal score of the flexspline.

[0088] In the embodiments of the present invention, for a specific model of harmonic reducer used (e.g., CSG-25-100), it is formulated by combining the relationship between the stress level predicted by finite element analysis (FEA) and the deformation amount, and the statistical distribution of peak deviations measured when a large number of harmonic reducers of this model experience various typical working condition jumps in a healthy state (such as rapid start-stop under different loads, collision recovery, etc.) (e.g., the upper limit of the 99% confidence interval). This standard divides the peak deviation amount into several levels. For example: Level 0 (normal): deviation amount < 0.05 mm; Level 1 (slight): 0.05 mm ≤ deviation amount < 0.09 mm; Level 2 (moderate): 0.09 mm ≤ deviation amount < 0.15 mm; Level 3 (severe): 0.15 mm ≤ deviation amount < 0.25 mm; Level 4 (critical): deviation amount ≥ 0.25 mm. Compare the comprehensive peak deviation amount of the input flexspline (0.1 mm) with this grading standard to determine the interval it falls into. In this example, 0.1 mm is in the "moderate" (Level 2) interval. Therefore, the evaluation result is "peak deformation severity" = Level 2. Identify the specific type of "working condition jump" that triggered this data collection (this information is stored as metadata in the "motor transient working condition characteristic data", such as "load suddenly increases from no-load to 5 kg"). Then, look up the entry that exactly matches this type in a pre-set "baseline database of working condition jump duration for the same condition". This database stores the statistical baselines of the stable flexspline deformation duration calibrated in a large number of healthy operation experiments for each defined typical working condition jump type, including the mean value (μ_Ts_base) and the standard deviation (σ_Ts_base). For example, the database shows that for the jump of "load suddenly increases from no-load to 5 kg", the healthy baseline is μ_Ts_base = 70 ms and σ_Ts_base = 5 ms. Next, calculate the standardized deviation (Z-score) of the currently measured duration (85 ms) relative to the baseline mean value. Check whether the peak deformation severity exceeds a warning threshold (e.g., set to the upper limit of Level 1 "slight", which is 0.09 mm). Since 0.1 mm > 0.09 mm, it indicates that the deformation amplitude is worthy of attention and further analysis of the vibration frequency is required. Next, compare the measured main oscillation frequency (50 Hz) with a pre-built "fault characteristic frequency library of harmonic reducer". This library is established based on fault injection experiments (such as simulating flexspline cracks, bearing wear, lubrication failure, etc.) and dynamic simulation analysis of this model of harmonic reducer, and contains the typical ranges of characteristic vibration frequencies under different fault modes. For example, the library records: the characteristic frequency range of early flexspline cracks: 45 - 55 Hz; the fault frequency of the outer ring of the input bearing (BPFO): about 180 Hz; the fault frequency of the inner ring of the output bearing (BPFI): about 125 Hz. Match the measured 50 Hz with the entries in the library. In this example, 50 Hz falls into the "characteristic frequency range of early flexspline cracks" (45 - 55 Hz).Therefore, the system generates "flexspline vibration frequency risk label" = "suspected flexspline crack risk (high)". If the frequency does not match any known failure modes, it is labeled as "does not match known failure frequency (low)". These different types of metrics are quantified and normalized, and mapped to a unified scoring range (e.g., 0 to 10 points). Severity level 2 is mapped to a score of 6 points; Z-score 3.0 (indicating significant deviation from normal) is mapped to a score of 8 points; risk label "suspected flexspline crack risk (high)" is mapped to a risk score of 9 points. Then, preset weights (W_severity, W_time, W_frequency) are assigned to these three quantified metrics. These weights are determined based on expert experience, historical failure data analysis, or the importance of each factor on the overall health in a specific application scenario, and the sum is 1. For example, set W_severity = 0.4, W_time = 0.3, W_frequency = 0.3. Finally, calculate the weighted average as the comprehensive anomaly score: flexspline comprehensive response anomaly score = (W_severity × Score_severity) + (W_time × Score_time) + (W_frequency × Score_frequency). This final score (7.5 points, out of 10) is the "flexspline comprehensive response anomaly score".

[0089] Preferably, step S3 includes the following steps:

[0090] Extract the post-event instantaneous electromagnetic signals from the motor transient condition characteristic data to obtain the motor transient current-voltage data and the motor magnetic field data respectively;

[0091] Calculate the instantaneous input active power of the motor based on the motor transient current-voltage data;

[0092] Calculate the transient power fluctuation intensity based on the instantaneous input active power of the motor to obtain the power impact response index;

[0093] Perform motor direct-quadrature axis current conversion according to the motor transient current-voltage data to obtain the motor direct-quadrature axis current component data;

[0094] Extract the motor direct-quadrature axis high-frequency ripple signals from the motor direct-quadrature axis current component data;

[0095] Perform current fault feature analysis on the motor direct-quadrature axis high-frequency ripple signals based on the power impact response index to obtain the current fault feature time-frequency spectrum; where if the amplitude of any frequency component in the range of 2 kHz to 5 kHz exceeds 0.5 A, mark this time period as the current fault feature area, and the spectral resolution is controlled within 0.2 A;

[0096] Perform a local distortion assessment on the motor magnetic field data, perform a fast Fourier transform using a 50 ms window, detect the magnetic field amplitude fluctuation in the frequency band from 100 Hz to 1 kHz, and generate a time series spectrum of the magnetic field distortion characteristics when the amplitude change of any spectral peak in this frequency band exceeds 5 mT.

[0097] In an embodiment of the present invention, a signal sequence related to the electromagnetic behavior of the motor is located and extracted from the data record, and only the part starting from the "event trigger time point" T0 (for example, 1678886400.150 seconds) and ending at the end of the "data segment collected after the event" (for example, T0 + 200 milliseconds = 1678886400.350 seconds) is selected. The specific signals extracted include: a three-phase current signal sequence (I_u(t), I_v(t), I_w(t), sampling rate 50 kHz) collected by a previously installed Hall current sensor and by an NI PXIe-6363 card; if the system configuration includes voltage acquisition, a three-phase voltage signal sequence (U_u(t), U_v(t), U_w(t), sampling rate 50 kHz) collected synchronously is extracted. These two sets of signals together constitute the "motor transient current-voltage data". At the same time, a three-channel magnetic field intensity signal sequence (for example, B_x(t), B_y(t), B_z(t), sampling rate 50 kHz) output by a Melexis MLX91208 magnetic sensor array installed near the stator winding and collected synchronously is extracted, and this set of signals constitutes the "motor magnetic field data". For each synchronous sampling time point t within this data segment (time range T0 to T0 + 200 ms), the standard calculation formula for three-phase instantaneous active power is applied: P(t) = U_u(t) × I_u(t) + U_v(t) × I_v(t) + U_w(t) × I_w(t). The specific calculation process is executed on a computer running a data analysis script (for example, using the NumPy library in Python). The script reads the three voltage values and three current values (both floating-point numbers, with units of volts and amperes respectively) at the corresponding moment point by point, performs the above multiplication and addition operations, and obtains the instantaneous input active power value P(t) (unit: watt) at that moment. This calculation is repeated for all N sampling points within the data segment (for example, for 200 ms of data and a sampling rate of 50 kHz, N = 10000 points), and finally a "motor instantaneous input active power" time series {P(t)} with a length of N is generated. To quantify the severity of the power impact during the motor's operating condition jump (within a specific period starting from T0, for example, selecting the main stage of the transient response T0 to T0 + 100 ms), the standard deviation of the power sequence during this period is calculated. The specific operation is as follows: select all power values in the {P(t)} with time stamps within the interval [T0, T0 + 100 ms]. Suppose this subsequence contains M data points (for example, M = 5000 points). First, calculate the average power μ_P = (1 / M) × ΣP(t) of this subsequence. Then calculate the standard deviation: σ_P = sqrt[(1 / (M - 1)) × Σ(P(t) - μ_P)^2]. The obtained standard deviation value σ_P (for example, the calculation result is 650 watts) is defined as the "power impact response index".This single value reflects the overall fluctuation amplitude of the input power of the motor when dealing with sudden load changes. The larger the value, the stronger the power impact. The three-phase current sequences (I_u(t), I_v(t), I_w(t)) in the "motor transient current-voltage data" extracted and the synchronous motor angular position signal θ_m(t) obtained from the "original flexspline displacement and angular position data". First, the number of pole pairs p of the target motor needs to be known (for example, p = 4). Calculate the electrical angle θ_e(t) of the motor at each sampling time t: θ_e(t) = p×θ_m(t). Then, apply the standard Park transformation (Clarke-Park transformation) to convert the currents (abc) in the three-phase stationary coordinate system to the direct-axis (d) and quadrature-axis (q) currents in the synchronous rotating coordinate system. The specific calculation is implemented in the data analysis script and usually consists of two steps: 1. Clarke transformation: Convert I_u(t), I_v(t), I_w(t) to I_α(t) and I_β(t) in the two-phase stationary coordinate system. For example, I_α(t) = sqrt(2 / 3)×(I_u(t) - 0.5×I_v(t) - 0.5×I_w(t)), I_β(t) = sqrt(2 / 3)×(sqrt(3) / 2×I_v(t) - sqrt(3) / 2×I_w(t)). 2. Park transformation: Rotate I_α(t), I_β(t) to the dq coordinate system using the calculated electrical angle θ_e(t): I_d(t) = I_α(t)×cos(θ_e(t)) + I_β(t)×sin(θ_e(t)); I_q(t) = -I_α(t)×sin(θ_e(t)) + I_β(t)×cos(θ_e(t)). Perform this transformation on all sampling points in the entire post-event time period (from T0 to T0 + 200 ms) to generate two new time series: "motor direct-axis current component data" {I_d(t)} and "motor quadrature-axis current component data" {I_q(t)}. Based on the "motor direct-quadrature axis current component data" {I_d(t)} and {I_q(t)}. To isolate the high-frequency components usually caused by inverter switching, dead-time effects, or potential electrical faults, a digital high-pass filter is applied to these two sequences respectively. A fourth-order Butterworth high-pass filter is selected, and its cut-off frequency is set to 1 kilohertz (kHz). The choice of this cut-off frequency is based on: it is higher than the fundamental frequency of the motor and its lower harmonics (usually below a few hundred hertz), but lower than the fault characteristic frequency range of interest (2 kHz to 5 kHz). The filtering operation is implemented by using signal processing library functions in the data analysis script (such as signal.butter and signal.filtfilt in the SciPy library in Python). Input {I_d(t)} into this filter to obtain the output sequence {I_d_ripple(t)}, which is the "motor direct-axis high-frequency ripple signal".Similarly, input {I_q(t)} into the same filter to obtain the output sequence {I_q_ripple(t)}, which is the "motor quadrature-axis high-frequency ripple signal". Perform time-frequency analysis on the ripple signal using the Short-Time Fourier Transform (STFT). Set the STFT parameters: the window function is selected as the Hann window, the window length is 1024 sampling points (corresponding to a duration of approximately 20.5 milliseconds at a sampling rate of 50 kHz, providing a frequency resolution of approximately 48.8 Hz), and the window overlap rate is 50% (i.e., sliding 512 points each time). Perform STFT on {I_d_ripple(t)} and {I_q_ripple(t)} within the time period [T0, T0 + 100 ms] (i.e., the interval used for calculating the power impact response index) to obtain their respective time-frequency spectrograms (time - frequency - amplitude). Next, check all the time-frequency units in the frequency range from 2 kHz to 5 kHz in these two time-frequency spectrograms. At each time slice (corresponding to the central moment of an STFT window), traverse all the frequency points from 2 kHz to 5 kHz and record their amplitudes (in amperes). Set the amplitude threshold to 0.5 amperes. If within this frequency range, it is found that the amplitude of any frequency point exceeds 0.5 amperes, then mark this time slice as the "current fault characteristic region". To ensure that the spectral resolution (amplitude accuracy) is controlled within 0.2 amperes, it is necessary to ensure the accuracy of the STFT calculation and amplitude estimation method, for example, by selecting an appropriate number of FFT points and amplitude correction factor. Finally, save the time-frequency spectrogram with the marked region (or only containing the information of the marked region) (usually two, one for the d-axis and one for the q-axis, or synthesized into one) as the "current fault characteristic time series spectrum". Set the window length to 50 milliseconds (corresponding to 2500 sampling points), the window uses the Hann window, and the overlap rate is 50% (sliding 25 milliseconds). Perform this sliding window FFT on the B_x(t) sequence in the time range from T0 to T0 + 200 ms. In the spectrum calculated for each window, focus on the frequency band from 100 Hz to 1 kHz (kHz). For each significant spectral peak (e.g., a peak whose amplitude exceeds a certain threshold of the background noise level) within this frequency band, record its frequency and amplitude. Then, compare the spectral peak amplitude of the current window with the spectral peak amplitude of the previous window (adjacent in time) at the same (or very close) frequency. Calculate the absolute value of the difference between the two amplitudes. Set a change threshold to 5 mT (millitesla). If within the 100 Hz to 1 kHz frequency band, it is detected that the change in the amplitude of any spectral peak compared to the previous window exceeds 5 mT, then it is considered that a significant magnetic field distortion has occurred within this time window. Record the time window and time-frequency information where the significant distortion occurs. Finally, generate a "magnetic field distortion characteristic time series spectrum" annotated with these distortion events.

[0098] Preferably, the analysis of the health state of the motor flexible gear according to the coupled data of electromagnetic fault characteristics and the abnormal score of the comprehensive response of the flexible gear in step S4 includes:

[0099] Conduct electromagnetic fault assessment for each channel according to the coupled data of electromagnetic fault characteristics to obtain the electromagnetic anomaly flag for each channel;

[0100] Based on the electromagnetic anomaly flag for each channel and the abnormal score of the comprehensive response of the flexible gear, conduct combined electro-mechanical anomaly determination to generate the combined electro-mechanical anomaly confidence level;

[0101] Conduct an assessment of the temporal correlation of abnormal responses during the jump event according to the coupled data of electromagnetic fault characteristics and the flexible gear deformation data during the working condition jump, and then correct the combined electro-mechanical anomaly confidence level for the single-event anomaly degree to obtain the corrected single-event coupled anomaly degree;

[0102] Conduct an analysis of the risk level of the flexible gear for the corrected single-event coupled anomaly degree, and then conduct processing on the health state of the motor to generate motor fault state data.

[0103] In an embodiment of the present invention, by scanning the time-series spectrum data structure, it is checked whether any time slice is marked as the "current fault feature region" (i.e., within the frequency band of 2 kHz to 5 kHz, there is a frequency component with an amplitude exceeding 0.5 amperes). If at least one such marked region is detected, the "electromagnetic anomaly flag of the current channel" is set to 1 (indicating an anomaly); if there is no such mark in the entire time-series spectrum, it is set to 0 (indicating normal). Secondly, the time-series spectrum of the magnetic field distortion feature is processed independently: it is checked whether any time window is marked as having significant magnetic field distortion (i.e., within the frequency band of 100 Hz to 1 kHz, the amplitude change of any spectral peak exceeds 5 mT). If at least one such distorted window is detected, the "electromagnetic anomaly flag of the magnetic field channel" is set to 1; otherwise, it is set to 0. First, a rule set is defined, which is preset according to historical data and expert knowledge and is used to combine the anomaly information in both mechanical (flexible gear) and electrical (electromagnetic) aspects. For example, the rules are set as follows: Rule 1: If the flexible gear score >= 7 and (current flag = 1 or magnetic field flag = 1), it is determined as "high-confidence electromechanical combined anomaly". Rule 2: If 5 <= flexible gear score < 7 and (current flag = 1 or magnetic field flag = 1), it is determined as "medium-confidence electromechanical combined anomaly". Rule 3: If the flexible gear score >= 7 and current flag = 0 and magnetic field flag = 0, it is determined as "medium-confidence flexible gear anomaly dominant". Rule 4: If the flexible gear score < 5 and (current flag = 1 or magnetic field flag = 1), it is determined as "medium-confidence electromagnetic anomaly dominant". Rule 5: In other cases, it is determined as "low-confidence anomaly". Substitute the input values (flexible gear score 7.5, current flag 1, magnetic field flag 0) into the rule set for matching. According to Rule 1 (7.5 >= 7 and (1 = 1 or 0 = 1)), it is determined as "high-confidence electromechanical combined anomaly". Then this determination result is mapped to a numerical "electromechanical combined anomaly confidence" between 0 and 1. For example, "high confidence" is mapped to 0.9. Therefore, the "electromechanical combined anomaly confidence" output in this embodiment is 0.9. First, time-series correlation evaluation is performed: find the time period when the flexible gear response is the most intense, such as from the event trigger point T0 to the moment T_peak when the flexible gear peak deviation occurs (which can be extracted from the time-series data of the flexible gear elliptical parameters, assumed to be T0 + 20 ms). At the same time, find the main time interval when electromagnetic anomalies (current fault feature region or magnetic field distortion window) occur (for example, the current fault feature region is concentrated in the period from T0 + 15 ms to T0 + 30 ms). Calculate the overlap degree of these two time intervals, such as using the Jaccard Index or a simple time overlap ratio. If the overlap degree is high (for example, Jaccard Index > 0.6), it is considered that there is a strong time-series correlation. In this example, there is a significant overlap between the flexible gear peak response period (assumed to be from T0 to T0 + 20 ms) and the current anomaly period (from T0 + 15 ms to T0 + 30 ms). Next, the confidence is corrected according to the correlation strength.Preset correction rules: When the correlation is strong, the confidence level is multiplied by the correction factor 1.1; when the correlation is medium, it is multiplied by 1.0; when the correlation is weak or non-existent, it is multiplied by 0.9. Due to the strong correlation, calculate the corrected anomaly: Corrected single-event coupling anomaly = Electromechanical combined anomaly confidence level × Correction factor = 0.9 × 1.1 = 0.99. This 0.99 is the output "corrected single-event coupling anomaly". Perform risk level analysis: Compare this anomaly with the pre-set risk classification threshold. This threshold system is established based on the failure mode, effects, and consequences analysis (FMEA) of this type of harmonic reducer and the statistics of historical maintenance data. For example, set the thresholds: Level 0 (healthy): Anomaly < 0.4; Level 1 (observe): 0.4 ≤ Anomaly < 0.7; Level 2 (recommend inspection): 0.7 ≤ Anomaly < 0.9; Level 3 (handle immediately): Anomaly ≥ 0.9. Compare the input anomaly of 0.99 with the threshold. Since it is ≥ 0.9, the risk level is determined to be "Level 3 (handle immediately)". Next, perform motor health status processing: Based on the determined risk level, generate a structured "motor fault status data" report. This report includes: timestamp of the event occurrence, robot axis number, triggered event type, calculated corrected single-event coupling anomaly (0.99), determined risk level (Level 3), text description of the risk level ("handle immediately"), summary of the main factors leading to high risk (e.g., "Severe peak deformation of the flexspline, long recovery time, abnormal current ripple and strongly correlated with flexspline response"), and recommended actions generated by the system according to the preset maintenance strategy (e.g., "Trigger an emergency stop signal, notify the maintenance personnel to immediately conduct a detailed endoscopic inspection of the harmonic reducer of the third axis, and prepare to replace the flexspline spare parts").

[0104] Particularly important is that the electromagnetic fault assessment of each channel according to the electromagnetic fault characteristic coupling data is specifically as follows:

[0105] Extract the time points of the current fault characteristic region and the maximum amplitude of the corresponding over-limit frequency components from the current fault characteristic time series spectrum to obtain the current characteristic over-limit amplitude sequence;

[0106] Calculate the cumulative occurrence times or total duration of over-limit events based on the current characteristic over-limit amplitude sequence, and conduct a monitoring channel status determination to obtain preliminary channel status determination data;

[0107] Extract the time points where the detected spectral peak amplitude change exceeds 5 mT and the corresponding maximum amplitude change amount from the magnetic field distortion characteristic time series spectrum to obtain the magnetic field distortion peak change sequence;

[0108] Conduct a magnetic field distortion anomaly judgment based on the magnetic field distortion peak change sequence, and correct the monitoring channel status of the preliminary channel status determination data to generate a sub-channel electromagnetic anomaly flag.

[0109] In the embodiments of the present invention, by parsing the time series spectrum data structure, all time slices marked as "current fault feature regions" are identified (i.e., STFT windows with frequency components having amplitudes exceeding 0.5 A in the frequency band from 2 kHz to 5 kHz). For each marked time slice, its central timestamp t is recorded, and the actual maximum amplitude Amax(t) (which must be greater than 0.5 A) is further searched within the 2 kHz to 5 kHz frequency range of this time slice. For example, when analyzing data of the event of grasping an overweight workpiece, it is found that the STFT window with a timestamp of T0 + 27 ms is marked, and the maximum amplitude of 0.7 A is detected at the frequency point of 3.5 kHz in this window; the window with a timestamp of T0 + 35 ms is also marked, and its maximum amplitude at 4.1 kHz is 0.6 A. All such information is collected to form a sequence containing timestamps and corresponding maximum over-limit amplitudes. The "cumulative occurrence times" is used as a quantization index. The total number of data points contained in this sequence is calculated. In this example, the sequence contains 3 data points, so the cumulative occurrence times is 3 times. Next, this number of times is compared with a preset "current channel status determination threshold". This threshold (for example, set to 2 times) is determined based on the analysis of a large amount of historical data to distinguish sporadic over-limit caused by noise or accidental interference in the healthy state from continuous or frequent over-limit caused by potential faults. Comparison result: 3 times > 2 times. Based on this comparison, "preliminary channel status determination data" is generated, and its content is: {monitoring channel: "current", preliminary status: "abnormal", determination basis: "the cumulative over-limit times of 3 times exceeds the threshold of 2 times"}. If the calculated number of times does not exceed the threshold, the preliminary status is determined to be "normal". The operation procedure scans this time series spectrum to locate all time windows marked as having significant magnetic field distortion (i.e., windows with a change in the spectral peak amplitude exceeding 5 mT compared to the previous window within the frequency band from 100 Hz to 1 kHz). For each marked time window, the program records its central timestamp t', and extracts the maximum spectral peak amplitude change amount |ΔBmax(t')| (which must be greater than 5 mT) observed within the 100 Hz to 1 kHz frequency band of this window. For example, the program identifies that the window with a central timestamp of T0 + 50 ms is marked, and the maximum amplitude change amount of the 600 Hz spectral peak inside it is 7 mT; the window with a timestamp of T0 + 75 ms is also marked, and the maximum amplitude change amount of the 450 Hz spectral peak inside it is 5.5 mT, forming a sequence containing timestamps and corresponding maximum amplitude change amounts. An abnormality judgment is made on the magnetic field channel: check whether the "magnetic field distortion peak change sequence" is empty. Since this sequence is not empty (containing two data points), it indicates that there is an abnormality in the magnetic field channel. Therefore, the abnormality flag of the magnetic field channel is set to 1. Next, the preliminary status of the current channel is corrected according to the determination result of the magnetic field channel (this correction logic aims to improve the reliability of the determination by using multi-source information).Preset correction rules: Rule 1: If the initial current state is "abnormal" and the magnetic field state is also determined to be "abnormal", then confirm that the current channel state is "abnormal" (flag = 1). Rule 2: If the initial current state is "normal" but the magnetic field state is determined to be "abnormal", then correct the current channel state to "suspected abnormal" or keep it "normal" according to the preset strategy (here it is assumed to keep "normal", flag = 0, unless there is stronger evidence). Rule 3: If the magnetic field state is determined to be "normal", then the initial state of the current channel is the final state. In this example, the initial current state is "abnormal" and the magnetic field state is also determined to be "abnormal", which conforms to Rule 1. Therefore, confirm that the current channel state is "abnormal" and its abnormal flag is 1. Finally, integrate the determination results of the two channels to generate the "electromagnetic anomaly flag for sub-channels" data.

[0110] Preferably, the present invention further provides an industrial robot motor fault warning system that executes the industrial robot motor fault warning method described above. The industrial robot motor fault warning system includes:

[0111] A transient condition monitoring module, which is used to obtain the industrial robot motor control instruction in real time, and then judge the instruction value change event to obtain the motor load change event; monitor the motor transient condition characteristic data in real time according to the motor load change event;

[0112] A flexure response analysis module, which is used to perform flexure deformation analysis during working condition jump based on the motor transient condition characteristic data to generate flexure deformation data during working condition jump; evaluate the abnormal flexure jump response according to the flexure deformation data during working condition jump to generate a comprehensive flexure response abnormal score;

[0113] An electromagnetic characteristic coupling module, which is used to perform electromagnetic fault characteristic analysis according to the motor transient condition characteristic data to obtain the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum respectively; perform multi-channel monitoring fault synchronous coupling on the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum to obtain electromagnetic fault characteristic coupling data;

[0114] A health assessment and warning module, which is used to analyze the health state of the motor flexure according to the electromagnetic fault characteristic coupling data and the comprehensive flexure response abnormal score to generate motor fault state data; perform multi-level fault warning processing according to the motor fault state data to achieve industrial robot motor fault warning.

[0115] By monitoring the changes in the motor control commands of industrial robots in real time, the present invention actively identifies the "events" of load or speed changes, and focuses the analysis on the "transient operating conditions" during which the motor states undergo jumps. This event-triggered monitoring method can more accurately capture the transient characteristic data of the motor when responding to command changes, thereby more clearly identifying the abnormal dynamic responses caused by internal faults (such as torque pulsations caused by inter-turn short circuits) and distinguishing them from the dynamic responses caused by normal load changes. This fundamentally solves the problems of difficult setting of early warning thresholds under variable operating conditions and frequent false alarms and missed alarms, and significantly improves the reliability of fault early warning.

[0116] Therefore, in any regard, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0117] The above description is only a specific implementation manner of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An industrial robot motor fault early warning method, characterized in that: The following steps are involved: Step S1: Acquire the industrial robot motor control command in real time, and then judge the command value change event to obtain the motor load change event; Real-time monitoring of the motor transient operating characteristic data based on the motor load change event; Step S2: Based on the transient working condition characteristic data of the motor, the deformation analysis of the flexible pulley under working condition jump is performed to generate the deformation data of the flexible pulley under working condition jump; based on the deformation data of the flexible pulley under working condition jump, the abnormal response of the flexible pulley under working condition jump is evaluated to generate the abnormal score of the comprehensive response of the flexible pulley; Step S3: performing electromagnetic fault characteristic analysis according to the transient operating condition characteristic data of the motor, and obtaining a current fault characteristic time series spectrum and a magnetic field distortion characteristic time series spectrum respectively; Perform multi-channel monitoring fault synchronous coupling on the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum to obtain electromagnetic fault characteristic coupling data; Step S4: analyzing the health status of the motor flexspline according to the electromagnetic fault characteristic coupling data and the flexspline comprehensive response abnormality score to generate motor fault status data; Multi-level fault warning processing is performed according to the motor fault status data to achieve industrial robot motor fault warning.

2. The industrial robot motor failure early warning method according to claim 1 is characterized in that: The step S1 of acquiring the industrial robot motor control instruction in real time and then judging the instruction value change event includes: Perform control instruction digital sampling on the speed or load instruction signal output in real time by the industrial robot motor control system to obtain the original instruction sequence value; Extract the instruction value of the latest sampling point according to the original instruction sequence value, and obtain its corresponding sampling time point to obtain the current industrial robot motor control data; According to the current industrial robot motor control data, the command value and its timestamp of the previous sampling cycle are extracted retrospectively to obtain the previous command time point data; Based on the current industrial robot motor control data and the previous instruction time point data, the difference between the timestamps is calculated to obtain the sampling time interval; based on the current industrial robot motor control data and the previous instruction time point data, the difference between the instruction values ​​is calculated to obtain the instruction value difference; wherein, the sampling time interval should be set within the range of 8 to 12 ms; Calculate the instantaneous change rate of the instruction according to the sampling time interval and the instruction value difference, and perform absolute value processing to generate the absolute value of the instruction change rate; The command value change event is judged by the absolute value of the command change rate through the preset command rapid change judgment threshold, and the motor load change event is obtained; wherein, the preset command rapid change judgment threshold is set to 10% per second, when the absolute value of the command change rate is greater than 10% per second, it is judged as a motor load change event.

3. The industrial robot motor fault early warning method according to claim 1 is characterized in that: The real-time monitoring of the transient operating condition characteristic data of the motor according to the motor load change event in step S1 includes: Send a synchronous trigger signal based on the motor load change event to obtain a synchronous acquisition trigger pulse signal; Perform parallel multi-channel monitoring according to the synchronous acquisition trigger pulse signal to generate multi-channel original sampling point data; wherein the multi-channel original sampling point data includes the original electromagnetic timing signal and the original flexible wheel displacement and rotation angle data; By using a preset pre-event tracing time threshold, the original digital stream segments of the multi-channel original sampling point data are intercepted in parallel to obtain the pre-event pre-collected data segments; The time point of the motor load change event is taken as the event start time node, and the parallel raw digital streams are continuously extracted from the multi-channel raw sampling point data until the preset total acquisition time covering the transient response and extending to the new steady state is reached, and the post-event acquisition data segment is obtained; The pre-event data segment and the post-event data segment are spliced ​​in chronological order to obtain the complete window original data; The motor load change event and the complete window raw data are processed with event trigger attributes, and the transient operating condition characteristics are integrated to generate the motor transient operating condition characteristic data.

4. The industrial robot motor fault early warning method according to claim 3 is characterized in that: The parallel multi-channel monitoring according to the synchronous acquisition trigger pulse signal comprises: The current sensor, magnetic sensor array and non-contact displacement sensor are accurately installed on the motor input line, the predetermined monitoring point of the stator winding and the corresponding position of the inner wall of the flexible wheel, and the signal output end of each sensor is physically connected to the corresponding input port of the data acquisition device to obtain a connected sensor port group; Based on the synchronous acquisition trigger pulse signal, the current and voltage of the three-phase input line of the industrial robot motor are sampled synchronously at high frequency through the connected sensor port group, and the magnetic field strength is synchronously detected at multiple points near the motor housing or the motor air gap to obtain the original electromagnetic timing signal; Based on the synchronous acquisition of trigger pulse signals, non-contact multi-point position displacement synchronous measurement is performed on the circumference of the outer wall of the harmonic reducer flexspline in the industrial robot motor through the connected sensor port group, and the motor angle signal is obtained synchronously to obtain the original flexspline displacement and angle data.

5. The industrial robot motor fault early warning method according to claim 1, characterized in that: The step S2 of performing the deformation analysis of the flexible pulley under the condition jump based on the transient condition characteristic data of the motor includes: Based on the transient working condition characteristic data of the motor, the elliptical geometric fitting of the cross-sectional profile of the flexible pulley is performed, and then the flexible pulley elliptical parameter time series data is extracted to generate the flexible pulley elliptical parameter time series data; Perform pre-event steady-state baseline statistics based on the transient operating condition characteristic data of the motor to generate pre-event steady-state baseline characteristic data; Perform post-event dynamic response analysis based on the transient operating condition characteristic data of the motor to generate post-event dynamic response characteristic data; Based on the steady-state baseline characteristic data before the event and the dynamic response characteristic data after the event, the flexible pulley deformation analysis of the working condition jump is carried out on the flexible pulley elliptical parameter time series data to generate the working condition jump flexible pulley deformation data, where the working condition jump flexible pulley deformation data includes the flexible pulley comprehensive peak deviation, the flexible pulley deformation stable time and the main oscillation frequency.

6. The industrial robot motor fault early warning method according to claim 5, characterized in that: The method of performing working condition jump flexible wheel deformation analysis on the flexible wheel elliptical parameter time series data based on the steady-state baseline characteristic data before the event and the dynamic response characteristic data after the event includes: Based on the time series data of the flexible pulley elliptical parameter and the steady-state baseline characteristic data before the event, the deviations of the major axis and the minor axis at each sampling moment in the preset transient window relative to their respective baseline values ​​are calculated to obtain the flexible pulley major axis deviation sequence and the flexible pulley minor axis deviation sequence respectively; The absolute values ​​of the long axis deviation sequence and the short axis deviation sequence of the flexible pulley are taken respectively, and their respective maximum values ​​are extracted within the preset transient window to obtain the maximum absolute value of the long and short axis deviation; Perform comprehensive peak value processing of the flexible pulley according to the absolute value of the maximum major and minor axis deviation to obtain the comprehensive peak deviation of the flexible pulley; Extract the last segment data in the preset transient window according to the time series data of the flexible pulley ellipse parameters, and perform the long and short axis steady-state deviation judgment threshold analysis according to the flexible pulley major axis deviation sequence and the flexible pulley minor axis deviation sequence to generate the deviation stability judgment threshold band; Extracting event triggering time points based on transient operating condition characteristic data of the motor; Based on the deviation stability judgment threshold band, search backward from the event triggering time point for the starting time point when both the major and minor axis deviation sequences simultaneously enter and remain within their respective threshold bands for the first time to obtain the deformation stability starting time; Analyze the deformation stability time of the flexible wheel according to the deformation stability starting time; Based on the stable deformation time of the flexible wheel, the oscillation behavior of the dynamic response characteristic data after the event is identified to generate a sequence of deviation oscillation characteristic points; Calculate the average oscillation period based on the deviation oscillation characteristic point sequence; The main oscillation frequency is obtained by taking the inverse of the average oscillation period.

7. The industrial robot motor fault early warning method according to claim 1, characterized in that: The abnormal evaluation of the flexspline jump response according to the working condition jump flexspline deformation data in step S2 includes: The severity of peak deformation is evaluated based on the comprehensive peak deviation of the flexspline; The abnormal extension or shortening degree of the flexible wheel deformation stability time is calculated by using the preset statistical time baseline under the same working condition jump, and the deformation recovery time abnormality index is obtained; Perform fault mode matching based on the comprehensive peak deviation of the flexible pulley to obtain the flexible pulley vibration frequency risk mark; Based on the peak deformation severity, deformation recovery time abnormality index and flexspline vibration frequency risk marker, a weighted assessment of the flexspline jump response abnormality is performed to generate a flexspline comprehensive response abnormality score.

8. The industrial robot motor fault early warning method according to claim 1, characterized in that: Step S3 includes the following steps: Extract the transient electromagnetic signal after the event from the transient working condition characteristic data of the motor, and obtain the transient current-voltage data and the magnetic field data of the motor respectively; Calculate the instantaneous input active power of the motor based on the instantaneous current-voltage data of the motor; The transient power fluctuation intensity is calculated based on the instantaneous input active power of the motor to obtain the power impact response index; Performing motor direct-to-quadrature axis current conversion according to the motor transient current-voltage data to obtain motor direct-to-quadrature axis current component data; Extracting the high-frequency ripple signal of the motor's direct-to-quadrature axis from the motor's direct-to-quadrature axis current component data; Based on the power impulse response index, the current fault characteristics of the high-frequency ripple signal of the motor's orthogonal axis are analyzed to obtain the current fault characteristic time series spectrum; if the amplitude of any frequency component in the range of 2kHz to 5kHz exceeds 0.5A, the time period is marked as the current fault characteristic area, and the spectrum resolution is controlled within 0.2A; The local distortion of the motor magnetic field data is evaluated by fast Fourier transform using a 50ms window, and the magnetic field amplitude fluctuation is detected in the frequency band of 100Hz to 1kHz. When the amplitude change of any spectrum peak in this frequency band exceeds 5mT, a characteristic time series spectrum of the magnetic field distortion is generated.

9. The industrial robot motor fault early warning method according to claim 1, characterized in that: The step S4 of analyzing the health status of the motor flexible pulley according to the electromagnetic fault characteristic coupling data and the flexible pulley comprehensive response abnormality score includes: According to the electromagnetic fault characteristic coupling data, the electromagnetic fault of each channel is evaluated to obtain the electromagnetic abnormality mark of each channel; Based on the electromagnetic anomaly signs of each channel and the abnormal score of the flexible pulley comprehensive response, the electromechanical joint anomaly is judged and the electromechanical joint anomaly confidence is generated; According to the electromagnetic fault characteristic coupling data and the working condition jump flexible wheel deformation data, the abnormal response time sequence correlation during the jump event is evaluated, and then the single event abnormality degree is corrected for the electromechanical joint abnormality confidence to obtain the corrected single event coupling abnormality degree; The flexible wheel risk level analysis is performed on the modified single event coupling abnormality, and then the motor health status processing is performed to generate the motor fault status data.

10. An industrial robot motor fault warning system, characterized in that: Used to execute the industrial robot motor failure early warning method as claimed in claim 1, the industrial robot motor failure early warning system comprises: The transient working condition monitoring module is used to obtain the industrial robot motor control instructions in real time, and then judge the instruction value change event to obtain the motor load change event; according to the motor load change event, the motor transient working condition characteristic data is monitored in real time; The flexible pulley response analysis module is used to analyze the flexible pulley deformation under working condition jump based on the transient working condition characteristic data of the motor, and generate the flexible pulley deformation data under working condition jump; perform the flexible pulley jump response abnormality evaluation based on the flexible pulley deformation data under working condition jump, and generate the flexible pulley comprehensive response abnormality score; The electromagnetic characteristic coupling module is used to perform electromagnetic fault characteristic analysis based on the transient operating condition characteristic data of the motor, and obtain the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum respectively; the current fault characteristic time series spectrum and the magnetic field distortion characteristic time series spectrum are subjected to multi-channel monitoring fault synchronous coupling to obtain electromagnetic fault characteristic coupling data; The health assessment and early warning module is used to analyze the health status of the motor flexible pulley based on the electromagnetic fault characteristic coupling data and the flexible pulley comprehensive response abnormality score, and generate motor fault status data; multi-level fault early warning processing is performed based on the motor fault status data to realize industrial robot motor fault early warning.

Images