A method and system for detecting a moving object drive failure

Abstract

This application provides a method and system for detecting faults in a moving object drive system. The method acquires the phase current signals of a three-phase motor in the moving object drive system, along with the corresponding operating parameters. Then, it constructs a current feature vector characterizing the motor winding state based on the phase current signals, and constructs a nominal current feature vector under normal operating conditions based on the operating parameters. Finally, it generates a residual signal based on the difference between the current feature vector and the nominal current feature vector, and determines whether an inter-turn short-circuit fault exists in the motor windings based on the statistical characteristics of the residual signal. This achieves early and effective detection of inter-turn short-circuit faults in the motor windings without adding additional sensors.

Description

Technical Field

[0001] This application relates to data processing technology, and more particularly to a method and system for detecting faults driven by moving objects. Background Technology

[0002] With the widespread application of industrial robots and collaborative robots in high-precision assembly, flexible manufacturing and other scenarios, robot joint drive systems typically adopt the form of servo drives and three-phase permanent magnet synchronous motors or asynchronous motors.

[0003] Inter-turn short-circuit faults in motor stator windings are relatively common in actual operation. Their causes include insulation aging, overheating, overload impact, electromagnetic stress, and environmental corrosion. Current technologies typically rely on simple current amplitude or protection thresholds to identify inter-turn short-circuit faults in motor windings. However, because early inter-turn short circuits only affect a small number of turns, the change in phase current at the port is small, making it difficult to detect these early, minor inter-turn short-circuit faults.

[0004] Therefore, there is an urgent need for a method that can detect short circuit faults between turns in motor windings at an early stage without adding extra sensors and by reusing existing servo controllers and motor feedback signals as much as possible. Summary of the Invention

[0005] This application provides a method and system for detecting faults in moving objects, which enables early detection of short-circuit faults between turns of a motor winding without adding additional sensors.

[0006] In a first aspect, this application provides a method for detecting faults in motion object driving processes, including:

[0007] Acquire the phase current signal of the three-phase motor in the motion object drive system and the corresponding operating condition parameters of the three-phase motor;

[0008] Based on the phase current signal, a current feature vector characterizing the motor winding state is constructed, and a nominal current feature vector under normal motor operating conditions is constructed according to the operating condition parameters.

[0009] A residual signal is generated based on the difference between the current feature vector and the nominal current feature vector, and the presence of an inter-turn short circuit fault in the motor winding is determined based on the residual signal in the motion object drive system.

[0010] Secondly, this application provides a motion object drive fault detection system, characterized in that it includes:

[0011] The acquisition module is used to acquire the phase current signal of the three-phase motor in the motion object driving system and the corresponding operating condition parameters of the three-phase motor;

[0012] The processing module is used to construct a current feature vector characterizing the motor winding state based on the phase current signal, and to construct a nominal current model of the motor under normal operating conditions based on the operating condition parameters.

[0013] The detection module is used to generate a residual signal based on the difference between the current feature vector and the nominal current model, and to determine whether there is an inter-turn short circuit fault in the motor winding in the moving object drive system based on the statistical characteristics of the residual signal.

[0014] Thirdly, this application provides an electronic device, comprising:

[0015] Processor; and,

[0016] Memory for storing the executable instructions of the processor;

[0017] The processor is configured to perform any of the possible methods described in the first aspect by executing the executable instructions.

[0018] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement any of the possible methods described in the first aspect.

[0019] The moving object drive fault detection method and system provided in this application acquire the phase current signal of the three-phase motor in the moving object drive system and the corresponding operating condition parameters of the three-phase motor. Then, based on the phase current signal, a current feature vector characterizing the state of the motor winding is constructed, and a nominal current feature vector under normal operating conditions of the motor is constructed according to the operating condition parameters. Then, a residual signal is generated according to the difference between the current feature vector and the nominal current feature vector. Based on the statistical characteristics of the residual signal, it is determined whether there is an inter-turn short circuit fault in the motor winding in the moving object drive system, thereby realizing early and effective detection of inter-turn short circuit faults in the motor winding without adding additional sensors. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0021] Figure 1 This is a schematic diagram illustrating an application scenario of a motion object-driven fault detection method according to an example embodiment of this application;

[0022] Figure 2 This is a schematic flowchart illustrating a motion object driving fault detection method according to an example embodiment of this application;

[0023] Figure 3 This is a flowchart illustrating an implementation of S150 according to an example embodiment of this application;

[0024] Figure 4 This is a schematic diagram of the structure of a motion object driven fault detection system according to an example embodiment of this application;

[0025] Figure 5 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0026] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0028] Figure 1 This is a schematic diagram illustrating an application scenario of a motion object-driven fault detection method according to an example embodiment of this application. Figure 2 This is a schematic flowchart illustrating a motion object drive fault detection method according to an example embodiment of this application. Figures 1-2 As shown, the method provided in this embodiment includes:

[0029] S110: Obtain the phase current signal of the three-phase motor in the motion object drive system and the corresponding operating condition parameters of the three-phase motor.

[0030] In this step, the phase current signal of the three-phase motor in the motion object drive system and the corresponding operating condition parameters of the three-phase motor are obtained. The operating condition parameters include at least one of the target torque command and speed command of the servo controller.

[0031] S120. Construct a current feature vector characterizing the state of the motor windings based on the phase current signal.

[0032] In this step, coordinate transformation and sequence component decomposition can be performed based on the phase current signal to construct a current feature vector characterizing the state of the motor winding.

[0033] Specifically, the phase current signal can first undergo Clarke and Park transforms to obtain the d-axis and q-axis currents. Then, symmetrical component decomposition can be performed on the phase current signal to obtain the zero-sequence and negative-sequence current components. Finally, within a preset time window, a current feature vector can be constructed based on the statistics and spectral quantities of the d-axis and q-axis currents, as well as the amplitude characteristics of the zero-sequence and negative-sequence current components.

[0034] The Clarke transform on the three-phase currents projects the currents from the three-phase stationary coordinate system (abc) to the two-phase orthogonal stationary αβ coordinate system, achieving a redundancy-free representation of the three-phase current information. Then, the Park transform rotates the currents from the αβ coordinate system to the dq coordinate system, which rotates synchronously with the motor rotor flux linkage. This ensures that the d-axis current primarily represents the excitation component, and the q-axis current primarily represents the torque component. Local flux imbalances caused by inter-turn short circuits are directly reflected in the amplitude and waveform distortion of the d-axis and q-axis currents, which can then be quantified using statistical and spectral quantities such as their mean, variance, and harmonic amplitude.

[0035] At the same time, the three-phase current is decomposed into positive sequence, negative sequence and zero sequence components. Under normal symmetrical operation, the negative sequence and zero sequence components are theoretically close to zero. However, the short circuit between winding turns destroys the phase current symmetry and will produce observable amplitudes in the negative sequence current and zero sequence current.

[0036] Extracting features such as the amplitude, root mean square (RMS), or dominant frequency amplitude of zero-sequence and negative-sequence currents within a preset time window can stably reflect the degree of phase current imbalance. By unifying the features extracted from the dq coordinate system and the sequence component space to form a current feature vector, joint monitoring of the three electromagnetic dimensions—winding flux excitation, torque generation, and three-phase symmetry—can be achieved.

[0037] Optionally, the length of the aforementioned preset time window can be dynamically adjusted according to the operating condition parameters to improve the detection sensitivity and real-time performance of inter-turn short-circuit faults in low-speed heavy-load conditions and high-speed light-load conditions, respectively.

[0038] Specifically, the operating condition intensity index can be calculated based on the target torque command and / or speed command from the servo controller to characterize the current operating condition. Optionally, the target torque command can be normalized to obtain a normalized torque command value. The speed command can also be normalized to obtain a normalized speed command value. Then, the normalized torque command value and the normalized speed command value are weighted according to a preset weighting coefficient to obtain a weighted operating condition intensity index. And / or, the product of the normalized torque command value and the normalized speed command value can be used as a joint operating condition intensity index. Finally, the weighted operating condition intensity index and / or the joint operating condition intensity index are compared with a preset operating condition threshold to indicate whether the current operating condition is a low-speed heavy-load condition or a high-speed light-load condition.

[0039] When the working condition strength index meets the low-speed heavy-load working condition judgment conditions, the length of the preset time window is increased to the target window length within the upper limit range of the first window length, so as to improve the statistical significance of the small current characteristics of the winding inter-turn short circuit fault, thereby improving the detection sensitivity.

[0040] When the operating condition strength index meets the judgment conditions for high-speed light-load operating conditions, the length of the preset time window is reduced to the target window length within the lower limit range of the second window length, so as to shorten the update cycle of the current feature vector and thus improve the real-time performance of fault detection.

[0041] It's worth noting that if a robot's servo drive system uses a three-phase star (Y) connection topology without a neutral line, then according to Kirchhoff's current law, the sum of the three-phase currents is physically forced to be zero in this topology. This means that even if an early inter-turn short circuit fault occurs inside the motor, generating an asymmetrical induced electromotive force, the fault characteristics cannot be detected by conventional current sensors in the form of zero-sequence current due to the lack of a zero-sequence current flow path. When dealing with such motors without a neutral line, the lack of zero-sequence characteristics often leads to a significant decrease in detection sensitivity, or even the inability to identify the fault.

[0042] To address this, when the three-phase motor uses a star connection and has no neutral wire, the zero-sequence current component can be obtained by first acquiring the three-phase voltage commands output from the servo controller's current loop, calculating the algebraic sum of the three-phase voltage commands, and obtaining the commanded zero-sequence voltage. Then, using the nominal current model, based on the phase current signals and operating parameters, the theoretical phase voltages of the three-phase motor under fault-free conditions are reconstructed in reverse, and the algebraic sum of the theoretical phase voltages is calculated to obtain the model zero-sequence voltage. Next, the difference between the commanded zero-sequence voltage and the model zero-sequence voltage is calculated, and the target subharmonic component in the difference is extracted as the zero-sequence voltage residual. Finally, a preset virtual admittance coefficient is used to map the zero-sequence voltage residual to a virtual zero-sequence current, and this virtual zero-sequence current is used as the zero-sequence current component to construct the current feature vector.

[0043] The determination of the aforementioned preset virtual admittance coefficient can be achieved by first obtaining the real-time electrical angular velocity of the three-phase motor and the zero-sequence inductance parameters in the nominal model. Then, based on the physical characteristic that the back EMF distortion caused by the inter-turn short-circuit fault in the motor winding is mainly concentrated in the third harmonic frequency band, the inductive reactance value corresponding to the third harmonic of the real-time electrical angular velocity is calculated, and the reciprocal of the inductive reactance value is determined as the virtual admittance coefficient. Furthermore, the virtual admittance coefficient can be dynamically updated according to the change in real-time electrical angular velocity to eliminate the influence of speed change on the zero-sequence voltage residual amplitude, so that the mapped virtual zero-sequence current reflects the severity of the fault itself.

[0044] Furthermore, it's worth explaining that when a motor experiences an inter-turn short circuit, the demagnetizing magnetic field generated by the short-circuit loop leads to an asymmetry in the three-phase back electromotive force (EMF). In a neutral-less system, this asymmetry cannot generate a zero-sequence current; instead, it transforms into a potential drift of the motor's neutral point relative to the DC bus midpoint (i.e., zero-sequence voltage). Since the robot servo system employs high-performance field-oriented control, its current loop closed-loop controller automatically adjusts the output d / q-axis voltage commands to maintain the target current (i.e., resist back EMF distortion). This transfers the neutral point potential drift information caused by the fault and embeds it in the algebraic sum of the three-phase voltage commands. The above steps first calculate the command zero-sequence voltage output by the servo controller, and then use the nominal current model and real-time operating parameters to inversely deduce the model zero-sequence voltage under fault-free conditions. By calculating the difference between the two, the voltage distortion residual caused solely by the fault is extracted from the complex command signal. Finally, using the circuit duality principle, a preset virtual admittance coefficient is introduced to map this voltage residual value into an equivalent virtual zero-sequence current.

[0045] This enables the virtual reconstruction of zero-sequence fault characteristics with clear physical meaning without adding additional voltage sensor hardware. Specifically, by introducing virtual admittance to map the voltage domain residual back to the current domain, the reconstructed virtual zero-sequence current can be seamlessly integrated into existing statistical detection algorithms based on current feature vectors. This significantly improves the robot's detection sensitivity and coverage for early, weak inter-turn short-circuit faults in a neutral-line-less configuration, while effectively eliminating conventional common-mode voltage interference injected by space vector pulse width modulation and reducing the false alarm rate.

[0046] S130. Construct the nominal current feature vector of the motor under normal operating conditions based on the operating condition parameters.

[0047] In this step, a nominal current model of a three-phase motor under fault-free conditions can be established based on the motor's equivalent circuit parameters, rated operating parameters, and the target torque and speed commands of the servo controller. The nominal current feature vector is then calculated from the nominal current model according to the operating condition parameters. The nominal current feature vector corresponds to the current feature vector in terms of feature dimension.

[0048] Specifically, the equivalent circuit equations of a three-phase motor in the synchronous rotating coordinate system can be established based on one or more equivalent circuit parameters such as stator resistance, stator inductance, rotor inductance, mutual inductance, and number of pole pairs, as well as one or more rated operating parameters such as rated voltage, rated current, and rated speed. These equations include, but are not limited to, voltage equations, voltage balance equations, flux linkage equations, torque equations, and motion equations.

[0049] More specifically, one could obtain the equivalent circuit parameters and rated operating parameters of the three-phase motor, and express the stator voltage equations of the three-phase motor as d-axis voltage equations and q-axis voltage equations in a synchronous rotating coordinate system. Under steady-state operating conditions, the current-time derivative terms in the d-axis and q-axis voltage equations are approximated to zero to obtain the steady-state d-axis and q-axis voltage balance equations. Optionally, one could also combine the number of motor pole pairs and the rated speed or corresponding synchronous electric angular velocity parameters to determine the corresponding reactance parameters and voltage drop terms in the steady-state d-axis and q-axis voltage balance equations to obtain the steady-state equivalent circuit of the three-phase motor under fault-free conditions.

[0050] Then, based on the target torque and speed commands of the servo controller, the steady-state d-axis current and steady-state q-axis current that satisfy the target torque and speed commands are solved using the electromagnetic torque balance relationship and voltage balance relationship, thus obtaining the nominal d-axis current and nominal q-axis current.

[0051] Next, based on the three-phase symmetry condition and the neutral point potential constraint, the corresponding nominal zero-sequence current component and nominal negative-sequence current component are calculated. Then, the statistics and spectral quantities of the nominal d-axis current and nominal q-axis current, as well as the amplitude characteristics of the nominal zero-sequence current component and nominal negative-sequence current component, are combined on the same feature dimension as the current feature vector to construct the nominal current feature vector.

[0052] It is worth noting that the three-phase inverter in a robot servo drive system typically consists of six power semiconductor switching devices. Although the ideal design assumes that the physical characteristics of the three-phase bridge arms are perfectly symmetrical, in actual industrial production, due to the limitations of manufacturing process discreteness, non-ideal characteristics such as inconsistent saturation conduction voltage drops, differences in switching delay times, and dead time jitter inevitably exist between the switching transistors of different phases.

[0053] According to the symmetrical component method theory, this three-phase asymmetrical voltage source characteristic on the inverter output side will directly induce a negative sequence current component in the motor stator windings. In the above fault detection scheme, since it is impossible to distinguish whether this negative sequence current is generated by an inter-turn short circuit in the motor windings (fault source) or by the inherent hardware asymmetry of the inverter (non-fault source), the system may misjudge the inverter's imbalance as a motor fault, or be forced to raise the detection threshold to avoid false alarms, thereby losing the ability to detect early, weak inter-turn short circuit faults.

[0054] To address this, determining the aforementioned nominal current model can begin by constructing a voltage error compensation model that characterizes the three-phase output imbalance of the inverter in the robot drive system. The process of constructing this voltage error compensation model is essentially an online identification and feature extraction process. First, the d-axis and q-axis voltage command values ​​of the servo controller's current loop output are obtained. Simultaneously, using pre-acquired motor resistance, inductance, and permanent magnet flux linkage parameters, combined with real-time sampled phase current and electric angular velocity, the theoretical terminal voltage required to maintain the current current state is calculated based on the motor's fundamental voltage balance equation.

[0055] Subsequently, the instantaneous error vector between the voltage command value and the theoretical terminal voltage is calculated. This error vector contains voltage distortion information caused by the inverter's nonlinear characteristics. Using an adaptive bandpass filter or software phase-locked loop with a bandwidth center frequency set to a certain value, the second harmonic voltage component is extracted from the instantaneous error vector in real time. This second harmonic voltage component constitutes the voltage error compensation model, which quantitatively characterizes the degree of imbalance in the inverter's three-phase output under the current operating conditions. It is then input as a feedforward compensation term into the subsequent nominal current model to correct the nominal current prediction benchmark.

[0056] Then, in the synchronous rotating coordinate system, the deviation vector between the voltage command of the servo controller and the voltage value calculated based on the motor fundamental equation is calculated, and the second harmonic voltage component twice the fundamental frequency is extracted from the deviation vector as the negative sequence voltage characteristic caused by inverter hardware asymmetry.

[0057] Finally, the second harmonic voltage component is injected into the nominal current model as a feedforward compensation quantity, so that the nominal current characteristic vector output by the nominal current model includes the inherent negative sequence current component caused by the inverter hardware asymmetry, so as to cancel the interference of non-fault sources when generating residual signals.

[0058] The voltage imbalance component, which appears as a negative-sequence fundamental wave in the three-phase stationary coordinate system, transforms into a second-harmonic AC component with a frequency twice the fundamental frequency after being projected onto the synchronous rotating coordinate system (dq axis) via Park transformation. Motor parameter errors typically manifest as DC deviations. The above steps utilize the deviation between the voltage command from the robot servo controller and the ideal back EMF calculated based on the motor's fundamental equation. A frequency domain extraction algorithm (such as a notch filter or phase-locked loop) is used to separate this second-harmonic voltage vector, identifying it as an inherent imbalance characteristic of the inverter. Subsequently, this inherent imbalance characteristic is injected as a feedforward into the nominal current model, forcing the theoretical current output by the model to also contain the same distortion. Finally, in the residual calculation stage, the distortion predicted by the model cancels out the actual measured distortion caused by the inverter, thus isolating the fault characteristics.

[0059] As can be seen, when the actual current is differentially divided with the corrected nominal current to generate a residual signal, the common error component caused by inverter asymmetry is canceled out, and the residual signal retains only the true fault characteristics caused by inter-turn short circuits in the motor windings. This not only significantly reduces the false alarm rate caused by differences in driver hardware, but also allows the system to set a lower alarm threshold, thereby achieving very early warning of inter-turn short circuit faults and preventing motor burnout.

[0060] S140. Generate a residual signal based on the difference between the current characteristic vector and the nominal current characteristic vector.

[0061] In this step, the current characteristic vector is differiating from the nominal current characteristic vector to obtain the current residual vector. Then, at least one of the Mahalanobis distance or weighted Euclidean distance is calculated based on the current residual vector as a residual signal characterizing the degree of inter-turn short-circuit fault in the winding.

[0062] Specifically, within the same preset time window used for constructing the current feature vector, each feature component of the current feature vector is subtracted one-to-one from the corresponding nominal feature component in the nominal current feature vector along its feature dimension, resulting in a sequence of differences for each feature component. This sequence of differences is then updated within the preset time window to obtain a time-varying sequence of current residual vectors. The difference sequence at the current moment is then arranged sequentially according to the feature sorting rules used when constructing the current feature vector, serving as the current residual vector for that moment, characterizing the deviation between the actual motor winding state and the nominal normal operating condition.

[0063] Next, under normal operating conditions of the three-phase motor, the covariance matrix of the current residual vector and / or the variance and correlation coefficient of each feature component are estimated based on historical current residual vector samples to obtain statistical weight parameters for distance measurement. Optionally, the correlation coefficients mentioned above include Pearson correlation coefficients between each pair of feature components.

[0064] During robot operation, for the current residual vector at each moment, the distance between the current residual vector and the zero vector is used as the residual signal. The Mahalanobis distance is obtained by left-multiplying the current residual vector by the inverse of the covariance matrix and right-multiplying it by the transpose of the current residual vector. The weighted Euclidean distance is obtained by weighting and summing the squares of each characteristic component of the current residual vector according to its corresponding statistical weights. The sequence of Mahalanobis distance and / or weighted Euclidean distance over time is used as the residual signal time series to continuously characterize the changing trend of the inter-turn short-circuit fault degree of the motor winding during the operation of the robot drive system.

[0065] S150. Based on the statistical characteristics of the residual signal, determine whether there is a short circuit fault between turns of the motor winding in the motion object drive system.

[0066] Figure 3 This is a schematic flowchart illustrating an implementation of S150 according to an example embodiment of this application. Figure 3 As shown, the above-mentioned S150 includes:

[0067] S151. Collect historical residual signals under normal operating conditions of a three-phase motor, estimate the statistical baseline parameters of the residual signals, and obtain the residual baseline model.

[0068] Specifically, during the normal operation phase of the robot when there is no short circuit between turns in the motor windings, residual signal sequences are continuously calculated within multiple preset time windows, forming a historical residual signal sample set. Then, statistical analysis is performed on this historical residual signal sample set to calculate at least one of the following: sample mean, sample variance, and / or higher-order statistics. These serve as statistical baseline parameters characterizing the distribution characteristics of the residual signals under normal operating conditions. The statistical baseline parameters, along with the number of historical residual signal samples used to generate them, the collection time period, and the corresponding operating condition range, are stored as a residual baseline model corresponding to different operating conditions. This model is used to standardize real-time residual signals and adaptively adjust fault detection thresholds during robot operation.

[0069] S152. During robot operation, the real-time residual signal is standardized based on the residual baseline model.

[0070] Specifically, during the actual operation of the robot, the residual signal at the current moment can be obtained through real-time calculation. Based on the current operating condition parameters, statistical baseline parameters that match the current operating condition can be selected from the residual baseline model. Alternatively, the target statistical baseline parameters under the current operating condition can be obtained by interpolating the statistical baseline parameters corresponding to multiple adjacent operating conditions.

[0071] Then, the current residual signal is standardized using the target statistical baseline parameters. When the residual signal is considered a scalar, the standardized residual signal is obtained by subtracting the sample mean from the residual baseline model and dividing by the square root of the corresponding sample variance. When the residual signal is considered a vector, the current residual vector is subjected to a linear transformation using the inverse of the covariance matrix as the transformation matrix, and / or by subtracting the mean dimensionally and normalizing according to the standard deviation. This yields standardized residual signals with comparable scales for each feature component under different operating conditions. The standardized residual signals are then used as input to subsequent sequence detection algorithms to improve the robot's robustness in detecting inter-turn short-circuit faults in motor windings under different operating conditions and its ability to suppress false alarms.

[0072] S153. Perform sequence detection on the standardized residual signal. When the detection statistic exceeds the preset threshold, it is determined that there is a short circuit fault between turns of the motor winding.

[0073] The standardized residual signal can be sequence-detected using at least one of the Cumulative Sum Control Chart (CUSUM) algorithm or the Generalized Likelihood Ratio (GLRT) algorithm. When the detection statistic exceeds a preset threshold, an inter-turn short-circuit fault in the motor winding is determined. Specifically, when using the CUSUM algorithm, the upper and lower biased cumulative sum statistics are initialized to zero at the start of robot operation. At each sampling time, the upper and lower biased cumulative sum statistics are recursively updated based on the deviation between the current standardized residual signal and the expected residual level under normal operating conditions. A truncation constraint of not less than zero is applied to the cumulative sum result at each step.

[0074] The updated upper bias cumulative sum statistic and / or lower bias cumulative sum statistic are compared with the corresponding CUSUM control threshold. When either cumulative sum statistic exceeds the corresponding threshold for the first time, an inter-turn short circuit fault in the motor winding is immediately determined, and the time of the first exceedance is recorded as the fault detection time.

[0075] When using the GLRT algorithm, during robot operation, standardized residual signal samples within a certain length of a sliding observation window are used to construct corresponding probabilistic statistical models under both the fault-free assumption and the assumption of an inter-turn short-circuit fault. The logarithmic statistic of the generalized likelihood ratio is calculated by comparing the maximum likelihood function values ​​under the two assumptions. The logarithmic statistic of the generalized likelihood ratio is compared with a GLRT detection threshold set based on the target false alarm rate and false alarm rate. When the logarithmic statistic exceeds the GLRT detection threshold, it is determined that an inter-turn short-circuit fault exists in the motor winding within the current sliding observation window.

[0076] Next, by introducing the statistical baseline parameters of the residual baseline model into the threshold settings of the CUSUM algorithm and / or GLRT algorithm, a balance is achieved between detection sensitivity and false alarm rate, so as to ensure that early, stable and reliable online detection of short circuit faults between turns of motor windings can be achieved during long-term operation of the robot.

[0077] Furthermore, the progression characteristics of inter-turn short-circuit faults in the winding can be determined based on the temporal variation trend of the residual signal. These progression characteristics can characterize the rate of change of the residual signal. When the progression characteristic reaches a warning threshold, a warning message is output to the robot's host computer or safety control module. When the progression exceeds a severe threshold, an emergency stop control of the robot's drive system is triggered.

[0078] In this embodiment, the phase current signals of the three-phase motor in the robot drive system and the corresponding operating condition parameters of the three-phase motor are acquired. Then, a current feature vector characterizing the state of the motor winding is constructed based on the phase current signal, and a nominal current feature vector under normal operating conditions is constructed based on the operating condition parameters. A residual signal is then generated based on the difference between the current feature vector and the nominal current feature vector. Based on the statistical characteristics of the residual signal, it is determined whether there is a short circuit fault between the turns of the motor winding in the robot drive system. This enables early and effective detection of short circuit faults between the turns of the motor winding without adding additional sensors.

[0079] Specifically, the above embodiment utilizes the existing three-phase current sampling signals of the servo driver and the torque and speed commands output by the controller. Without adding additional sensors, it constructs a multi-dimensional current feature vector by performing coordinate transformation and symmetrical component decomposition on the three-phase current, including d / q axis components, zero-sequence components, negative-sequence components and their statistics and spectral quantities. Combined with the operating condition parameters, it generates the corresponding nominal current feature vector based on the equivalent circuit model of the motor.

[0080] By calculating the residual vector between actual and nominal features, and using Mahalanobis distance or weighted Euclidean distance to form a scalar residual signal, combined with the residual baseline model established in the health phase, a sequence detection method is used to accumulate, amplify, and determine the long-term small offset of the residual online. This enables early identification of current asymmetry and harmonic distortion caused by slight inter-turn short circuits under varying operating conditions. Based on the residual trend, the degree of fault progression is assessed, and warning or shutdown commands are output, thereby improving the robot drive system's sensitivity to early fault detection of motor windings and operational safety.

[0081] To facilitate understanding, the above embodiments achieve specific technical effects by no longer solely focusing on whether the current suddenly becomes very large. Instead, they first transform the representation of the three-phase current into a way that more easily reveals subtle anomalies. It's like not focusing on the current of the three motor phases themselves, but rather on whether the three phases are still sufficiently coordinated and symmetrical, without any minor inconsistencies or distortions. Through mathematical transformation, these aspects of coordination and abnormal components are extracted. Then, combined with the controller's known current task requirements (such as the desired torque and rotation speed), the characteristics that the current should exhibit under perfectly healthy conditions are calculated. Finally, a detailed comparison is made with the actual measured characteristics to see if there is a persistent, even very slight, difference between the current performance and the expected performance.

[0082] Finally, these small, persistent deviations are accumulated for analysis: occasional deviations are ignored, while persistent deviations are considered a potential early sign of inter-turn short circuits within the motor, triggering warnings or shutdowns based on the degree of deviation. This allows for early detection of problems before they escalate into serious burnout, without the need for additional sensors, thus improving the robot's reliability and safety.

[0083] Figure 4 This is a schematic diagram of the structure of a moving object driven fault detection system according to an example embodiment of this application. For example... Figure 4 As shown, the motion object drive fault detection system 300 provided in this embodiment includes:

[0084] The acquisition module 310 is used to acquire the phase current signal of the three-phase motor in the motion object driving system and the operating condition parameters corresponding to the three-phase motor.

[0085] The processing module 320 is used to construct a current feature vector characterizing the motor winding state based on the phase current signal, and to construct a nominal current model of the motor under normal operating conditions based on the operating condition parameters.

[0086] The detection module 330 is used to generate a residual signal based on the difference between the current feature vector and the nominal current model, and to determine whether there is a short circuit fault between the motor windings in the moving object drive system based on the residual signal.

[0087] Figure 5 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 5 As shown, the electronic device 400 provided in this embodiment includes: a processor 401 and a memory 402; wherein:

[0088] Memory 402 is used to store computer programs, and the memory may also be flash memory.

[0089] Processor 401 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0090] Alternatively, the memory 402 can be either standalone or integrated with the processor 401.

[0091] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include:

[0092] Bus 403 is used to connect the memory 402 and the processor 401.

[0093] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0094] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0095] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0096] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for detecting faults in motion-driven objects, characterized in that, include: Acquire the phase current signal of the three-phase motor in the motion object drive system and the corresponding operating condition parameters of the three-phase motor; Based on the phase current signal, a current feature vector characterizing the motor winding state is constructed, and a nominal current feature vector under normal motor operating conditions is constructed according to the operating condition parameters. A residual signal is generated based on the difference between the current feature vector and the nominal current feature vector, and the presence of an inter-turn short circuit fault in the motor winding is determined based on the residual signal in the motion object drive system.

2. The method for detecting motion object drive faults according to claim 1, characterized in that, The construction of the current feature vector characterizing the motor winding state based on the phase current signal includes: The phase current signal is subjected to coordinate transformation to obtain the d-axis current and q-axis current; Symmetric component decomposition is performed on the phase current signal to obtain the zero-sequence current component and the negative-sequence current component. Within a preset time window, the current feature vector is constructed based on the statistics and spectral quantities of the d-axis current and the q-axis current, as well as the amplitude characteristics of the zero-sequence current component and the negative-sequence current component.

3. The method for detecting motion object drive faults according to claim 2, characterized in that, The length of the preset time window is dynamically adjusted according to the operating condition parameters.

4. The method for detecting motion object drive faults according to claim 1, characterized in that, The step of constructing the nominal current model of the motor under normal operating conditions based on the operating condition parameters includes: The nominal current feature vector is determined based on the motor's equivalent circuit parameters, rated operating parameters, and the target torque and speed commands of the servo controller. The nominal current feature vector corresponds to the current feature vector in terms of feature dimension.

5. The method for detecting motion object drive faults according to claim 1, characterized in that, The step of generating a residual signal based on the difference between the current feature vector and the nominal current feature vector includes: The current residual vector is obtained by differencing the current feature vector with the nominal current feature vector. The residual signal is calculated based on the current residual vector, using either the Mahalanobis distance or the weighted Euclidean distance, as a residual signal characterizing the degree of inter-turn short-circuit fault in the winding.

6. The method for detecting motion object drive faults according to claim 1, characterized in that, The step of determining whether there is an inter-turn short circuit fault in the motor winding of the moving object drive system based on the residual signal includes: Historical residual signals are collected under normal operating conditions of the three-phase motor, and the statistical baseline parameters of the residual signals are estimated to obtain the residual baseline model. During robot operation, the real-time residual signal is standardized based on the residual baseline model. The standardized residual signal is subjected to sequence detection. When the detection statistic exceeds the preset threshold, it is determined that there is a short circuit fault between turns of the motor winding.

7. The method for detecting motion object drive faults according to claim 1, characterized in that, After generating the residual signal based on the difference between the current feature vector and the nominal current model, the method further includes: The progression characteristics of the inter-turn short-circuit fault in the winding are determined based on the time variation trend of the residual signal; When the progress characteristic reaches the warning threshold, a warning message is output to the robot's host computer or safety control module. When the progress exceeds a critical threshold, the emergency stop control of the motion object driving system is triggered.

8. A fault detection system for a moving object, characterized in that, include: The acquisition module is used to acquire the phase current signal of the three-phase motor in the motion object driving system and the corresponding operating condition parameters of the three-phase motor; The processing module is used to construct a current feature vector characterizing the motor winding state based on the phase current signal, and to construct a nominal current model of the motor under normal operating conditions based on the operating condition parameters. The detection module is used to generate a residual signal based on the difference between the current feature vector and the nominal current model, and to determine whether there is an inter-turn short circuit fault in the motor winding in the moving object drive system based on the residual signal.

9. An electronic device, characterized in that, include: processor; as well as, Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 7.

Images