Transport robot motor torque evaluation method

Abstract

The application discloses a kind of transport robot motor torque evaluation methods, it is related to the technical field of torque evaluation, including the motor current signal of transport robot, the actual torque signal of output shaft and drive wheel vibration signal;Actual torque signal and the time-frequency matrix of drive wheel vibration signal are constructed, and fault sensitive factor is extracted by singular value decomposition;The motor current signal is analyzed by order, and net abnormal torque is obtained;When net abnormal torque is greater than the first preset threshold and fault sensitive factor is greater than the second preset threshold, it is determined that mechanical failure occurs in the motor of the transport robot.The application improves the accuracy of the evaluation of transport robot motor torque.

Description

Technical Field

[0001] This invention relates to the field of torque evaluation technology, and in particular to a method for evaluating the torque of a transport robot motor. Background Technology

[0002] Currently, torque is a key consideration when AGV (Automated Guided Vehicle) robots transport goods. Torque, or torque force, is a measure of the force required for an object to rotate around its axis. In the context of AGV robots transporting goods, the design and performance of the AGV robot's power system also affect its torque performance. For example, to handle large goods, AGV robots should have high torque and high power, while also needing good speed adjustment capabilities to adapt to different transportation needs.

[0003] In existing technologies, such as Chinese Patent Publication No. CN115056237B, a pipeline transport robot system is disclosed. This system uses a current sensor and a Hall encoder to detect the current and rotation angle of the drive motor, thereby determining the current torque and speed of the drive motor. In other words, existing torque evaluation methods rely on a single sensor (such as a current or force sensor) to evaluate the motor torque.

[0004] However, evaluating motor torque using a single sensor cannot distinguish between torque anomalies caused by load changes and mechanical faults, especially under conditions of frequent start-stop cycles in transport robots. Specifically, when the load changes, such as increased cargo weight or sudden changes in path gradient requiring greater torque, the current detected by the current sensor will increase proportionally. Conversely, in cases of mechanical faults, such as friction-related faults like bearing jamming, the transport robot needs to overcome additional frictional losses, resulting in a persistently high current. In other words, the transient current response to load changes (such as the current surge during startup) and the current characteristics of mechanical faults (such as the sustained high current caused by jamming) have highly similar time-domain waveforms, making it impossible to distinguish between the two based solely on current amplitude or trend.

[0005] Therefore, improving the accuracy of torque assessment for transport robots has become a pressing technical problem. Summary of the Invention

[0006] The technical problem solved by this invention is that the accuracy of motor torque evaluation in the prior art is not high enough.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for evaluating the motor torque of a transport robot, the evaluation method comprising: acquiring the motor current signal, the actual torque signal of the output shaft, and the vibration signal of the drive wheel of the transport robot; constructing a time-frequency matrix of the actual torque signal and the vibration signal of the drive wheel, and extracting fault-sensitive factors through singular value decomposition; performing order analysis on the motor current signal to obtain the net abnormal torque; and determining that a mechanical fault has occurred in the motor of the transport robot when the net abnormal torque is greater than a first preset threshold and the fault-sensitive factor is greater than a second preset threshold.

[0008] Preferably, the acquisition of the motor current signal, the actual torque signal of the output shaft, and the drive wheel vibration signal of the transport robot includes: continuously acquiring the real-time current value of the drive motor through a first sensor to obtain the motor current signal; continuously acquiring the actual torque of the output shaft of the reducer through a second sensor to obtain the actual torque signal, wherein the second sensor is installed on the flange coupling between the reducer and the drive wheel; continuously acquiring the drive wheel vibration acceleration of the drive wheel of the transport robot through a third sensor; and determining the drive wheel vibration signal based on the drive wheel vibration acceleration.

[0009] Preferably, the motor current signal is a time-series sequence. The step of performing order analysis on the motor current signal to obtain the net abnormal torque includes: acquiring the angular position signal of the motor rotor of the transport robot, wherein the angular position signal is a time-series sequence; determining the rate of change of the motor rotor's speed based on the angular position signal; determining start-stop periods based on periods in the angular position signal where the rate of change of speed is greater than a first preset rate of change threshold, and determining running periods based on periods in the angular position signal where the rate of change of speed is not greater than the first preset rate of change threshold; resampling the motor current signal corresponding to the start-stop periods at a first angular domain sampling interval, and resampling the motor current signal corresponding to the running periods at a second angular domain sampling interval to generate an angular domain signal, wherein the first angular domain sampling interval is less than the second angular domain sampling interval; and performing a Fourier transform on the angular domain signal to separate the net abnormal torque.

[0010] Preferably, determining the start-stop period based on the time period in the angular position signal where the rotational speed change rate is greater than a first preset change rate threshold, and determining the running period based on the time period in the angular position signal where the rotational speed change rate is not greater than the first preset change rate threshold, includes: determining that the transport robot enters a start-stop state when the rotational speed change rate is greater than the first preset change rate threshold, and determining that the transport robot exits the start-stop state when the rotational speed change rate is less than or equal to the second preset change rate threshold, wherein the second preset change rate threshold is less than the first preset change rate threshold; determining the moment when the transport robot enters the start-stop state as the initial moment of the start-stop period, and determining the moment when the transport robot exits the start-stop state as the end moment of the start-stop period; dividing the start-stop period and the running period from the time period corresponding to the angular position signal based on the initial moment and the end moment of the start-stop period.

[0011] Preferably, the step of constructing the time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting fault-sensitive factors through singular value decomposition, includes: decomposing the actual torque signal and the drive wheel vibration signal corresponding to the high-speed period into a first preset number of layers using wavelet packet decomposition, and decomposing the actual torque signal and the drive wheel vibration signal corresponding to the low-speed period into a second preset number of layers using wavelet packet decomposition, to obtain a torque vector and a vibration vector, wherein the first preset number of layers is greater than the second preset number of layers; performing an outer product operation on the torque vector and the vibration vector to obtain the time-frequency matrix; and extracting fault-sensitive factors by performing singular value decomposition on the time-frequency matrix.

[0012] Preferably, the high-speed period is the period when the speed of the drive wheel is greater than a preset speed threshold, and the low-speed period is the period when the speed of the drive wheel is less than or equal to the preset speed threshold.

[0013] Preferably, the step of extracting the fault sensitivity factor by performing singular value decomposition on the time-frequency matrix includes: extracting the first two singular values ​​by performing singular value decomposition on the time-frequency matrix; and determining the fault sensitivity factor based on the first two singular values ​​and the drive wheel speed.

[0014] Preferably, before constructing the time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting the fault sensitivity factor through singular value decomposition, the evaluation method includes: performing a moving average filter on the actual torque signal, wherein the window width of the moving average filter is synchronized with the motor rotation cycle; and performing an adaptive notch filter on the drive wheel vibration signal.

[0015] The beneficial effects of this invention are as follows: By performing order analysis on the motor current signal, the net abnormal torque is obtained. When the net abnormal torque is greater than a first preset threshold, a mechanical fault is determined in the motor. This eliminates the influence of speed fluctuations through the net abnormal torque, and its amplitude directly reflects the true degree of torque abnormality, regardless of whether the torque abnormality is caused by load changes or faults, thereby improving the accuracy of motor torque assessment. Furthermore, by constructing a time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting fault sensitive factors through singular value decomposition, a mechanical fault is determined in the motor when the fault sensitive factor is greater than a second preset threshold. This allows the fault sensitive factor to characterize the sensitive components in the time-frequency matrix that are strongly correlated with mechanical faults, thereby accurately identifying mechanical faults in the transport robot by using the fault sensitive factor while excluding torque abnormalities caused by speed fluctuations. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the basic process of a method for evaluating the motor torque of a transport robot, provided in one embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] The embodiments of this application are applicable to evaluating the motor torque of a transport robot, wherein the transport robot may be an automated guided vehicle (AGV) robot, an autonomous mobile robot (AMR), or an intelligent guided vehicle (IGV) robot, etc.

[0019] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for evaluating the motor torque of a transport robot is provided, the evaluation method comprising:

[0020] S100 collects the motor current signal, the actual torque signal of the output shaft, and the vibration signal of the drive wheel of the transport robot.

[0021] Specifically, the real-time current value of the drive motor is continuously collected by the first sensor to obtain the motor current signal, the actual torque of the output shaft of the reducer is continuously acquired by the second sensor to obtain the actual torque signal, the second sensor is installed on the flange coupling between the reducer and the drive wheel, and the vibration acceleration of the drive wheel of the transport robot is continuously acquired by the third sensor, and the vibration signal of the drive wheel is determined based on the vibration acceleration of the drive wheel.

[0022] The first sensor can be a Hall sensor, which is embedded in the magnetic core ring of the motor power supply cable. Non-contact measurement avoids electromagnetic interference. The second sensor can be a six-dimensional force sensor, which is installed on the flange coupling between the reducer and the drive wheel. The flange coupling adopts a tapered positioning installation to ensure that the coaxiality between the six-dimensional force sensor and the reducer output shaft is ≤0.05mm, thus isolating the transmission chain from flexible deformation interference (such as gearbox vibration). The third sensor can be a piezoelectric accelerometer, which is rigidly connected to the drive wheel hub by bolt fixing bracket.

[0023] When a transport robot navigates complex terrain (such as slopes or uneven surfaces) or experiences sudden load changes (such as collisions with obstacles), its drive wheels are subjected to forces and torques in multiple directions. Traditional single sensors (such as torque sensors) can only measure torque in one direction, failing to comprehensively reflect the actual force situation. A six-dimensional force sensor can simultaneously measure forces (Fx, Fy, Fz) and torques (Mx, My, Mz) in three orthogonal directions, directly acquiring complete load information for the drive wheels and providing high-precision input data for the algorithm.

[0024] Preferably, prior to S200, the evaluation method includes:

[0025] S180 performs a moving average filter on the actual torque signal, with the window width of the moving average filter synchronized with the motor rotation cycle.

[0026] Moving average filtering is a classic signal processing method widely used for noise suppression in industrial equipment. Its technical background stems from the need to suppress periodic noise (such as torque fluctuations caused by motor rotation), and it achieves phase alignment by using a synchronization window width that matches the motor rotation period, thus avoiding signal distortion after filtering.

[0027] Motor torque signals often contain periodic disturbances related to speed (such as gear meshing harmonics). Moving average filtering, through the synchronization window width, preserves the true torque trend while suppressing high-frequency noise.

[0028] The window width is synchronized with the motor rotation cycle (e.g., when the motor speed is n RPM, the window time is 60 / n seconds) to ensure that the phase of the filtered signal is consistent with the original torque fluctuation, avoiding waveform distortion caused by traditional fixed windows; by averaging the torque data of adjacent cycles, high-frequency random noise (such as electromagnetic interference) and short-term impact interference are suppressed, while retaining torque trend characteristics (such as load sudden changes).

[0029] S190 performs adaptive notch filtering on the vibration signal of the drive wheel.

[0030] Adaptive notch filtering is a cutting-edge method in the field of digital signal processing, particularly suitable for removing narrowband interference strongly correlated with rotational speed (such as the resonant frequency of drive wheel bearings). Its algorithm design references adaptive noise cancellation theory and combines real-time frequency tracking technology to dynamically adjust the notch center frequency.

[0031] The vibration signal of the drive wheel of a transport robot often contains electromagnetic noise from the motor and ground impact interference. Adaptive notch filtering can selectively filter out noise in specific frequency bands while retaining fault characteristics.

[0032] S200 constructs a time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracts fault-sensitive factors through singular value decomposition.

[0033] To address the issue of strong noise interference in the vibration signal of the drive wheel of a transport robot due to ground impact and sudden load changes, a time-frequency joint matrix of vibration and torque signals is constructed. Singular value decomposition (SVD) is then used to extract sensitive components (i.e., fault sensitivity factors) in the matrix that are strongly correlated with mechanical faults.

[0034] Mechanical faults (such as broken gear teeth) can cause nonlinear coupling between vibration and torque signals in a specific frequency band. Load changes only alter the torque amplitude and do not change the coupling characteristics. Singular Value Decomposition (SVD) obtains a fault sensitivity factor by decomposing the singular value distribution of the matrix. This fault sensitivity factor is only sensitive to mechanical faults, and load changes do not significantly alter it. In other words, the coupling strength can be quantified through the fault sensitivity factor, enabling the identification of abnormal torque caused by mechanical faults.

[0035] The following is a specific implementation of S200:

[0036] First, the signal is preprocessed. The actual torque signal is decomposed using wavelet packet decomposition (such as Daubechies wavelet basis), and high-frequency coefficients are quantized by thresholding to filter out electromagnetic interference and mechanical noise. The drive wheel vibration signal is subjected to continuous wavelet transform (CWT) to generate a time-frequency distribution matrix covering the 0.3Hz–8kHz frequency band to capture the high-frequency characteristics of faults such as gear tooth breakage and bearing jamming.

[0037] Secondly, a time-frequency matrix is ​​constructed. The denoised torque signal is converted into a time-frequency matrix through a short-time Fourier transform (STFT), with a Hanning window (1024 points) as the window function. The continuous wavelet decomposition results of the vibration signal are allocated according to low frequency (R channel), mid frequency (G channel), and high frequency (B channel) to generate an RGB time-frequency chromatogram matrix, enhancing the visual distinguishability of fault features. The torque time-frequency matrix and the vibration RGB matrix are aligned along the time axis to construct a joint time-frequency matrix (dimension: time × frequency × channel) to achieve spatial coupling of multimodal signals. The joint time-frequency matrix is ​​expanded into discrete signals according to the time series to construct a Hankel matrix (dimension: m × n), where m is the embedding dimension (taken as 3 times the main period of the signal), and n is the signal length (N) - m + 1. The value of m can also be adaptively adjusted according to the speed fluctuation rate (such as the speed change rate during start-stop phase > 50 rpm / s) to ensure that high-frequency transient features are not truncated.

[0038] Subsequently, the Hankel matrix is ​​decomposed using SVD to obtain multiple singular values. Based on the difference in singular value distributions between normal and fault states, a sensitivity coefficient βi is defined, for example... Determine the fault component corresponding to the i-th order singular value, where, These are the singular values ​​derived from the fault state / SVD decomposition. These are singular values ​​under normal conditions, for example, when When the singular values ​​are derived from SVD and βi > 0.48, the fault components corresponding to the lower i-th order singular values ​​can be determined. In addition, the first k sensitive singular values ​​(e.g., k = 5) can be selected to construct a feature vector, which can be reduced to 2 dimensions by principal component analysis (PCA) and input into an improved binary K-means classifier to distinguish fault modes such as normal, bearing wear, and gear tooth breakage.

[0039] In another implementation of S200, S200 includes sub-steps S210 to S230:

[0040] S210, the actual torque signal corresponding to the high-speed period and the drive wheel vibration signal are decomposed into a first preset number of layers through wavelet packet decomposition, and the actual torque signal corresponding to the low-speed period and the drive wheel vibration signal are decomposed into a second preset number of layers through wavelet packet decomposition to obtain the torque vector and the vibration vector, wherein the first preset number of layers is greater than the second preset number of layers.

[0041] In S210, dynamic layer wavelet packet decomposition can be performed to obtain the drive wheel speed n in real time. If n is less than or equal to a preset speed threshold, such as 50 RPM, then a 3-layer decomposition is performed on the actual torque signal T(t) and the drive wheel vibration signal V(t). If n > 50 RPM, then a 4-layer decomposition is performed on V(t) and T(t). When the speed n is in the medium-high speed range (> 50 RPM), a 4-layer decomposition is used (frequency band resolution 0-1.25 kHz, accuracy ± 150 Hz). When the speed is low (< 50 RPM), it is reduced to 3 layers (accuracy ± 300 Hz) to avoid over-decomposition leading to noise amplification and the fragmentation of fault characteristics when the transport robot is running at low speed.

[0042] The basis function for wavelet packet decomposition can be the db4 wavelet basis, whose compact support characteristics match the transient characteristics of gear impact signals. At the same time, sensitive frequency bands containing fault impact components are automatically screened through kurtosis-entropy dual index evaluation (selecting nodes with kurtosis > 3 and entropy < 1.5).

[0043] In bearing failures, the main characteristic frequencies are concentrated in the low-frequency band (such as the fundamental frequency of gear meshing). Three-layer decomposition can effectively cover the key frequency band. Higher layers (such as four layers) may introduce noise interference due to excessive subdivision of the high-frequency band, which will reduce the accuracy.

[0044] High-speed selection of 4-layer decomposition can refine high-frequency fault features (such as bearing sidebands and gear impacts), while low-speed selection of 3-layer decomposition aims to balance low-frequency resolution and computational efficiency, and avoid noise interference with weak high-frequency features.

[0045] For example, at low speeds (n=30RPM), if the traditional fixed 4-layer decomposition is used, the fault frequency band will be divided, resulting in energy dispersion; at low speeds (n=30RPM), a 3-layer decomposition is used to ensure that the fault frequency band is completely preserved and the energy is concentrated.

[0046] Furthermore, the frequency band energy in low-speed mode can be processed to avoid frequency band aliasing problems under variable speed conditions. For example, for low-speed mode (3 layers), sub-bands within 0-1250Hz can be forcibly merged to generate a wideband energy vector, merging the original 4 layers [0-625Hz, 625-1250Hz] into a single energy value. For high-speed mode (4 layers), the energy of the 625Hz resolution sub-band can be retained to avoid energy dispersion at low speeds.

[0047] S220 performs an outer product operation on the torque vector and the vibration vector to obtain the time-frequency matrix.

[0048] The vibration signal can be taken as a wideband / high-resolution energy vector to obtain a row vector, and the torque signal can be processed synchronously into a column vector. The outer product operation is performed to construct a matrix to strengthen the fault coupling relationship. Specifically, when the gear wears, the diagonal energy of the time-frequency matrix is ​​enhanced by 3 times.

[0049] S230 obtains the fault sensitivity factor by performing singular value decomposition on the time-frequency matrix.

[0050] In an alternative approach, S230 includes sub-steps S231 to S232:

[0051] S231 extracts the first two singular values ​​by performing singular value decomposition on the time-frequency matrix.

[0052] S232, determine the fault sensitivity factor based on the first two singular values ​​and the drive wheel speed.

[0053] Decompose the time-frequency matrix and take the first two singular values ​​ε1 and ε2, β=|ε1-ε2| / (ε1+ε2) to quantify the coupling strength. Furthermore, if n<50RPM, β can be multiplied by a compensation coefficient, such as 1.2, to compensate for the attenuation of low-speed signals.

[0054] S300 performs order analysis on the motor current signal to obtain the net abnormal torque.

[0055] When the rotational speed of the transport robot fluctuates, such as during start-up and shutdown, the fluctuation causes the fault characteristic frequency to overlap with the rotational speed harmonics. This can lead to misjudgments when torque anomaly assessment is based solely on fault sensitivity factors. Therefore, S300 needs to be executed to eliminate the interference of rotational speed fluctuations on torque anomaly assessment.

[0056] The core of order analysis technology is to eliminate the interference of speed fluctuations on torque anomaly assessment by synchronously resampling the speed (converting the time domain signal into an order domain related to the speed). The net abnormal torque obtained by order analysis can eliminate the influence of speed fluctuations, and its amplitude directly reflects the true degree of torque anomaly (whether caused by load or fault).

[0057] The specific implementation of S300 is as follows: The motor current signal I(t) is resampled to I(θ) by θ(t). An FFT Fourier transform is performed on I(θ) to obtain the order domain signal I(O), where order O = frequency / fundamental frequency (fundamental frequency corresponds to order 1 = 1 cycle / rev). Abnormal orders are extracted from the order domain signal I(O). Energy concentrated in the 1st order (fundamental frequency) and integer multiples of harmonics (2nd and 3rd orders) represents normal torque. Abnormal torque exhibits energy peaks at non-integer orders (e.g., 1.5th and 2.3rd orders) (gear wear characteristics). These non-integer orders with sudden energy increases (e.g., 1.5th and 2.3rd orders) are marked in the order spectrum. Only the components of the marked orders are retained. An inverse transform is performed back to the angular domain to obtain the net abnormal torque.

[0058] To address the dynamic operating conditions of the transport robot with frequent start-stop cycles, a variable-order resolution algorithm is designed (e.g., increasing the angular domain sampling density during low-speed phases) to ensure accurate separation of abnormal torque components caused by mechanical faults or load changes, even during drastic speed changes (e.g., 0 → peak → 0 RPM), thus obtaining the net abnormal torque. Specifically, the motor current signal is a time sequence, and step S300 includes sub-steps S310–S350:

[0059] S310: Acquire the angular position signal of the motor rotor of the transport robot. The angular position signal is a timing sequence.

[0060] S320 determines the rate of change of the motor rotor speed based on the angular position signal.

[0061] The rate of change of rotational speed (dθ / dt) is the angular acceleration of the motor rotor, characterizing the intensity of the sudden change in rotational speed. dθ / dt = Δθ / Δt (unit: rad / s) 2 (or RPM / s).

[0062] S330, determine the start / stop period based on the period when the speed change rate in the angular position signal is greater than the first preset change rate threshold, and determine the running period based on the period when the speed change rate in the angular position signal is not greater than the first preset change rate threshold.

[0063] When the transport robot is in the start-stop phase, for example, if |dθ / dt|>100RPM / s, it is determined to have entered the speed change interval (start-stop period), and the sampling strategy adjustment needs to be triggered. The adjustment logic is to reduce the angular domain interval and increase the sampling density when a speed change (start-stop period) is detected.

[0064] When determining the start / stop period and running segment, preferably, S330 includes sub-steps S331 to S333:

[0065] S331, when the rotational speed change rate is greater than the first preset change rate threshold, the transport robot is determined to enter the start-stop state until the rotational speed change rate is less than or equal to the second preset change rate threshold, and then the transport robot is determined to exit the start-stop state, wherein the second preset change rate threshold is less than the first preset change rate threshold; S332, the moment when the transport robot enters the start-stop state is determined as the initial moment of the start-stop period, and the moment when the transport robot exits the start-stop state is determined as the end moment of the start-stop period; S333, the start-stop period and the running period are divided from the period corresponding to the angular position signal according to the initial moment and the end moment of the start-stop period.

[0066] In S330, dual threshold hysteresis comparison is used in S331 to S333 to avoid frequent switching. For example, the condition for entering the mutation interval (the initial time of the start-stop period) is dθ / dt > 00 RPM / s, and the condition for exiting the mutation interval (the end time of the start-stop period) is dθ / dt ≤ 50 RPM / s.

[0067] S340, the motor current signal corresponding to the start-stop period is resampled in the corner domain at the first corner domain sampling interval, and the motor current signal corresponding to the running period is resampled in the corner domain at the second corner domain sampling interval to generate a corner domain signal, wherein the first corner domain sampling interval is smaller than the second corner domain sampling interval.

[0068] For example, during the runtime phase, in steady state, 360 points are sampled per revolution, which is suitable for routine analysis under stable speed; during start-stop phases and abrupt changes, 1800 points are sampled per revolution.

[0069] S350 performs a Fourier transform on the angle domain signal to separate the net abnormal torque.

[0070] The technical principle of order analysis is to resample the time-domain signal I(t) into an angular domain signal I(θ) according to the rotor angle θ(t), and then transform it to the order domain through Fourier transform (FFT) to eliminate speed fluctuation interference. If a fixed angular domain interval Δθ is used for sampling during sudden speed changes, it will lead to uneven distribution of sampling points in the angular domain signal I(θ) (sparse in the low-speed range and dense in the high-speed range), resulting in spectral leakage. In S330, this problem can be solved by dynamically adjusting the angular domain interval Δθ through S310 to S350.

[0071] S400: When the net abnormal torque is greater than the first preset threshold and the fault sensitivity factor is greater than the second preset threshold, it is determined that a mechanical fault has occurred in the motor of the transport robot.

[0072] By using a dual-threshold joint determination, the causes of torque anomalies can be accurately classified. If only the net abnormal torque ΔT(t) > the first threshold and the fault sensitivity factor β ≤ the second preset threshold, it indicates that the torque anomaly is caused by load changes (β is not exceeded). If ΔT(t) > the first threshold and β > the second threshold, it indicates that the torque anomaly is caused by a mechanical fault (the fault leads to enhanced coupling between torque and vibration). The first threshold can be determined based on the actual situation.

[0073] In S300, order analysis ensures that ΔT(t) accurately reflects the degree of anomaly. In S200, by constructing a time-frequency matrix and performing SVD decomposition on the time-frequency matrix to obtain β, it can be ensured that β only responds to mechanical faults.

[0074] This application embodiment obtains the net abnormal torque by performing order analysis on the motor current signal. When the net abnormal torque is greater than a first preset threshold, a mechanical fault is determined in the motor. This eliminates the influence of speed fluctuations by using the net abnormal torque, and its amplitude directly reflects the true degree of torque abnormality, regardless of whether the torque abnormality is caused by load changes or faults, thereby improving the accuracy of motor torque assessment. Furthermore, by constructing a time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting fault sensitive factors through singular value decomposition, a mechanical fault is determined in the motor when the fault sensitive factor is greater than a second preset threshold. This fault sensitive factor characterizes the sensitive components in the time-frequency matrix that are strongly correlated with mechanical faults, thereby accurately identifying mechanical faults in the transport robot by using the fault sensitive factor while excluding torque abnormalities caused by speed fluctuations.

[0075] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating the torque of a transport robot motor, characterized in that, The evaluation method includes: Collect the motor current signal, the actual torque signal of the output shaft, and the vibration signal of the drive wheel of the transport robot; Construct a time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extract fault-sensitive factors through singular value decomposition; The net abnormal torque is obtained by performing order analysis on the motor current signal; When the net abnormal torque is greater than the first preset threshold and the fault sensitivity factor is greater than the second preset threshold, it is determined that a mechanical fault has occurred in the motor of the transport robot. The step of constructing the time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting fault-sensitive factors through singular value decomposition, includes: The actual torque signal corresponding to the high-speed period and the drive wheel vibration signal are decomposed into a first preset number of layers using wavelet packet decomposition, and the actual torque signal corresponding to the low-speed period and the drive wheel vibration signal are decomposed into a second preset number of layers using wavelet packet decomposition to obtain a torque vector and a vibration vector, wherein the first preset number of layers is greater than the second preset number of layers. The time-frequency matrix is ​​obtained by performing an outer product operation on the torque vector and the vibration vector; The fault sensitivity factor is obtained by performing singular value decomposition on the time-frequency matrix. The high-speed period is the period when the speed of the drive wheel is greater than a preset speed threshold, and the low-speed period is the period when the speed of the drive wheel is less than or equal to the preset speed threshold. The step of extracting fault-sensitive factors by performing singular value decomposition on the time-frequency matrix includes: The first two singular values ​​are extracted by performing singular value decomposition on the time-frequency matrix. The fault sensitivity factor is determined based on the first two singular values ​​and the drive wheel speed.

2. The evaluation method according to claim 1, characterized in that, The acquisition of the motor current signal, the actual torque signal of the output shaft, and the vibration signal of the drive wheel of the transport robot includes: The motor current signal is obtained by continuously collecting the real-time current value of the drive motor through the first sensor. The actual torque signal is obtained by continuously acquiring the actual torque of the output shaft of the reducer through a second sensor, which is installed on the flange coupling between the reducer and the drive wheel. The vibration acceleration of the drive wheel of the transport robot is continuously acquired by a third sensor. The vibration signal of the drive wheel is determined based on the vibration acceleration of the drive wheel.

3. The evaluation method according to claim 1, characterized in that, The motor current signal is a time sequence. The step of performing order analysis on the motor current signal to obtain the net abnormal torque includes: The angular position signal of the motor rotor of the transport robot is obtained, and the angular position signal is a time sequence; The rate of change of the motor rotor speed is determined based on the angular position signal; The start-stop period is determined based on the period when the rate of change of rotation speed in the angular position signal is greater than the first preset rate of change threshold, and the running period is determined based on the period when the rate of change of rotation speed in the angular position signal is not greater than the first preset rate of change threshold. The motor current signal corresponding to the start-stop period is resampled in the corner domain at a first corner domain sampling interval, and the motor current signal corresponding to the running period is resampled in the corner domain at a second corner domain sampling interval to generate a corner domain signal, wherein the first corner domain sampling interval is smaller than the second corner domain sampling interval. The net abnormal torque is separated by performing a Fourier transform on the angular domain signal.

4. The evaluation method according to claim 3, characterized in that, The step of determining the start / stop period based on the time period in the angular position signal where the rate of change of rotational speed is greater than a first preset rate of change threshold, and determining the running period based on the time period in the angular position signal where the rate of change of rotational speed is not greater than the first preset rate of change threshold, includes: When the rotational speed change rate is greater than the first preset change rate threshold, the transport robot is determined to enter the start-stop state. The transport robot is determined to exit the start-stop state when the rotational speed change rate is less than or equal to the second preset change rate threshold, wherein the second preset change rate threshold is less than the first preset change rate threshold. The moment when the transport robot enters the start-stop state is determined as the initial moment of the start-stop period, and the moment when the transport robot exits the start-stop state is determined as the end moment of the start-stop period. The start-stop period and the running period are divided from the period corresponding to the angular position signal based on the initial time and the end time of the start-stop period.

5. The evaluation method according to claim 1, characterized in that, Before constructing the time-frequency matrix of the actual torque signal and the drive wheel vibration signal, and extracting fault sensitivity factors through singular value decomposition, the evaluation method includes: The actual torque signal is subjected to a moving average filter, and the window width of the moving average filter is synchronized with the motor rotation cycle; Adaptive notch filtering is applied to the vibration signal of the drive wheel.

Images