A waveform shaping-based energy storage motor fault diagnosis method and system

Abstract

The present application belongs to the technical field of pulse shaping, and particularly relates to a method and system for fault diagnosis of an energy storage motor based on waveform shaping, which comprises the following steps: obtaining burr index by evaluating local asymmetry of data points, and introducing an iterative gating smoothing algorithm to selectively process current pulse waveforms to correct the interference of noise burrs; automatically positioning starting points, peak points and steady-state points on the structure diagram of the smoothed pulse waveforms to separate out key inflection points reflecting the real physical process; combining structural peak value integration and structural rise rate to construct a feature vector, establishing a health model and a fault feature library based on Mahalanobis distance, and realizing fault decision; thus providing more accurate and robust diagnostic basis for motor state monitoring, and solving the problems of feature extraction distortion and diagnostic misjudgment caused by noise interference in traditional methods.

Description

Technical Field

[0001] This invention relates to the field of pulse shaping technology. More specifically, this invention relates to a fault diagnosis method and system for energy storage motors based on waveform shaping. Background Technology

[0002] For industrial and automation systems, especially in the scenario of monitoring the operating status and diagnosing faults of energy storage motors in high-voltage circuit breakers, the motor is a core component that ensures the system can instantly release or store energy during specific processes. Therefore, real-time and accurate monitoring of the operating status of the energy storage motor and timely diagnosis of potential faults are of great significance for ensuring the safe and stable operation of the entire system.

[0003] In existing motor fault diagnosis technologies, a mainstream approach involves collecting current signals from the motor during startup and operation using current sensing devices. Technicians attempt to identify the motor's health status by analyzing these pulse waveform data. For example, observing the peak value, rise time, or steady-state characteristics of the current can help determine if the motor has typical faults such as coil aging or mechanical stall. The core of this approach relies on extracting key diagnostic features from the pulse waveforms and comparing them with known normal or fault benchmarks to achieve diagnosis.

[0004] However, the aforementioned existing technologies have significant drawbacks in practical applications: the current pulse waveforms acquired by sensors inevitably contain a large number of noise spikes, which severely affect the accuracy of the analysis; traditional diagnostic methods rely heavily on single features that are extremely sensitive to noise, which means that an isolated noise spike may be misjudged as a true peak, leading to feature extraction distortion and misdiagnosis; in order to solve the noise problem, existing technologies often use traditional filtering algorithms such as Gaussian filtering, but while these algorithms smooth out the noise, they inevitably blur the key inflection points in the pulse waveform, destroying the structure of the pulse waveform, which also leads to inaccurate feature extraction. Summary of the Invention

[0005] To address the technical problems of existing technologies, such as the presence of numerous noise spikes in the current pulse waveform and the blurring of key inflection points during denoising using traditional algorithms like Gaussian filtering, which leads to distorted and inaccurate diagnostic feature extraction and consequently misjudgments by the diagnostic system, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a fault diagnosis method for an energy storage motor based on waveform shaping, comprising: for each data point on the current pulse waveform, obtaining a glitch index of the data point by evaluating the local asymmetry between the data point and its neighboring data points; obtaining a glitch index threshold based on the degree of abrupt change and jitter of the current pulse waveform; obtaining an iterative gated smoothing algorithm based on the glitch index threshold, the glitch index, and the local mean of the glitch index; the gated smoothing algorithm smooths only data points whose glitch index is greater than the glitch index threshold, and retains data points whose glitch index is less than or equal to the glitch index threshold; The data points are iteratively smoothed using a gated smoothing algorithm. In each iteration, the glitch index is recalculated based on the current pulse waveform. The final processed data points are then combined to form a pulse waveform structure diagram. The starting point, peak point, and steady-state point are located on the pulse waveform structure diagram. Based on the starting point, peak point, and steady-state point, the structural peak integral and structural rise rate are extracted to form a feature vector. A health model and a fault feature library are established for the feature vector. The Mahalanobis distance between the feature vector of the newly acquired motor pulse waveform and the health model is calculated. Based on the comparison results of the Mahalanobis distance with the health threshold and with the fault feature library, a fault judgment is made.

[0007] This invention generates a pulse waveform structure diagram through an iterative gated smoothing algorithm, and extracts the structural peak integral and structural rise rate to form a feature vector. This avoids the feature extraction distortion caused by noise glitches or filtering blurring of key inflection points in traditional methods. Furthermore, it combines Mahalanobis distance for fault judgment, thereby solving the technical problem of misjudgment in diagnostic systems.

[0008] Preferably, obtaining the spurt index of the data point by evaluating the local asymmetry between the data point and its neighboring data points includes: the data point at the... The time point of the next iteration The burr index satisfies the expression: ;in, For the first Second iteration time point The burr index; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; These are positive numbers, empirical values. Ampere (A), the range of values ​​is Ampere Amperes are used to prevent the denominator from being zero. It is the absolute value symbol.

[0009] This invention utilizes the local asymmetry between data points and their neighboring data points to obtain the glitch index. This glitch index can amplify the evaluation value of noise glitch points while keeping the evaluation value of real structure points at a low level, providing a basis for the subsequent iterative gating smoothing algorithm to accurately distinguish between noise and real pulse waveform structures.

[0010] Preferably, the data points are smoothed using an iterative gated smoothing algorithm based on the spurious index. The iterative gated smoothing algorithm, obtained based on the spurious index threshold, the spurious index, and the local mean of the spurious index, includes: [further details on the iterative smoothing algorithm, including the process of smoothing the data points using the spurious index threshold, the spurious index, and the local mean of the spurious index]. Second iteration time point pulse waveform amplitude Updated, number The pulse waveform amplitude in the next iteration satisfies the expression: ;in, For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For smoothing coefficients; In the first iteration Local mean of a point; This is a gated function.

[0011] This invention employs an iterative gated smoothing algorithm, which can selectively process based on the glitch index and the glitch index threshold. By using a gate function, smoothing is applied only to data points that are identified as noise glitch points, while preserving the pulse waveform amplitude of the real structural points. This removes noise while completely preserving the key inflection points in the pulse waveform, solving the problem that traditional filtering algorithms blur these inflection points.

[0012] Preferably, the local mean satisfies the expression: The gate function satisfies the expression: ;in, For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The burr index; For the first Second iteration time point The threshold for the burr index.

[0013] Preferably, locating the starting point, peak point, and steady-state point on the pulse waveform structure diagram includes: calculating the second derivative of the pulse waveform structure diagram; finding the first peak point where the second derivative is positive and locating it as the starting point; finding the maximum value point of the pulse waveform structure diagram and locating it as the peak point; and locating the point where, after finding the peak point, the absolute value of the second derivative first falls back and remains near zero, as the steady-state point.

[0014] This invention utilizes the second derivative of the pulse waveform structure diagram to locate the starting point, peak point, and steady-state point. Since the pulse waveform structure diagram has eliminated the interference of noise spikes, the location results of these key inflection points are accurate, avoiding possible misjudgments on the original current pulse waveform, and providing a reliable time anchor point for subsequent extraction of structural peak integral and structural rise rate.

[0015] Preferably, the extraction of the structural peak integral includes: obtaining the pulse waveform amplitude of the pulse waveform structure diagram at all time points between the starting point and the steady-state point; obtaining the steady-state current value corresponding to the steady-state point; calculating the difference between the pulse waveform amplitude and the steady-state current value at each time point; and multiplying all differences by the sampling time interval and summing them to obtain the structural peak integral.

[0016] This invention extracts the structural peak integral by calculating the integral of the difference between the pulse waveform amplitude and the steady-state current value between the start-up point and the steady-state point. This feature represents the total charge or energy consumption during the start-up process, and therefore it is not sensitive to noise glitches, and has better robustness compared to traditional features that rely on a single data point.

[0017] Preferably, the method for obtaining the structural rise rate is as follows: obtaining the difference between the pulse waveform amplitude corresponding to the peak point and the pulse waveform amplitude corresponding to the starting point to obtain the pulse waveform amplitude difference; obtaining the difference between the time of the peak point and the time of the starting point; dividing the pulse waveform amplitude difference by the time difference to obtain the structural rise rate.

[0018] Preferably, the establishment of the health model of feature vectors includes: acquiring current pulse waveforms of multiple sets of known healthy energy storage motors; for each set of healthy pulse waveforms, performing operations such as calculating the glitch index for each data point on the current pulse waveform, smoothing the data points using an iterative gated smoothing algorithm to generate a pulse waveform structure diagram, and locating the start point, peak point, and steady-state point on the pulse waveform structure diagram and extracting feature vectors; calculating the mean and covariance matrix of the feature vectors of the current pulse waveforms of multiple sets of known healthy energy storage motors, and using the mean and covariance matrix as the health model.

[0019] Preferably, the first Second iteration time point The spurious exponent threshold satisfies the expression: In the formula, It is the first Second iteration time point The threshold for the burr index; It is the first Second iteration time point The amplitude of the pulse waveform; It is the first Second iteration time point The amplitude of the pulse waveform; For the first The average value of the pulse waveform amplitude during each iteration; This is the total time; These are positive numbers, empirical values. Ampere (A), the range of values ​​is Ampere Ampere is used to prevent the denominator from being zero.

[0020] Secondly, the present invention provides a fault diagnosis system for energy storage motors based on waveform shaping, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned fault diagnosis method for energy storage motors based on waveform shaping is implemented.

[0021] By adopting the above technical solution, a computer program for fault diagnosis of energy storage motor based on waveform shaping is generated and stored in a memory so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and the processor for convenient use.

[0022] The beneficial effects of this invention are as follows:

[0023] This invention obtains the glitch index by evaluating the local asymmetry between a data point and its neighboring data points. Based on this glitch index and a glitch index threshold, an iterative gated smoothing algorithm is used to smooth the data points. This algorithm can selectively smooth noise glitch while preserving the true pulse waveform structure, thereby generating a clear pulse waveform structure diagram. This solves the problem that traditional filtering algorithms blur the key inflection points in the pulse waveform.

[0024] This invention accurately locates the start point, peak point, and steady-state point on the pulse waveform structure diagram, and extracts the structural peak integral and structural rise rate based on these reliable structural points. These features are a comprehensive evaluation of a range of pulse waveforms. Compared with traditional features that rely on a single feature and are extremely sensitive to noise, the feature vector is not sensitive to noise spikes and has better robustness, thus solving the problem of feature extraction distortion.

[0025] This invention establishes a health model and a fault feature library based on feature vectors. For newly acquired pulse waveforms, the Mahalanobis distance between their feature vectors and the health model is calculated and compared with the health threshold and the fault feature library to make a fault judgment. This diagnostic process is based on pulse waveform shaping and robust feature extraction, thereby improving the accuracy of diagnosis and solving the technical problem of misjudgment in the diagnostic system. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a fault diagnosis method for energy storage motors based on waveform shaping according to the present invention.

[0027] Figure 2 This is a schematic diagram illustrating the pulse waveform structure after smoothing iteration in this invention;

[0028] Figure 3 This is a schematic diagram illustrating the location of key structural points on a complete pulse waveform structure diagram in this invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0031] This invention discloses a fault diagnosis method for energy storage motors based on waveform shaping, referring to... Figure 1 This includes steps S1-S4:

[0032] S1: Acquire and preprocess the current pulse waveform of the energy storage motor.

[0033] It should be noted that, in order to obtain data on the operating status of the energy storage motor, this invention utilizes a current sensing device to completely record the current signal of the motor during the entire startup and operation process.

[0034] Specifically, a current acquisition Hall sensor is connected in series in the power supply circuit of the circuit breaker energy storage motor; when the circuit breaker energy storage mechanism is started, the sensor acquires the current signal flowing through the motor in real time at a high sampling rate of, for example, 10kHz, and converts the analog signal into a digital time series, which is recorded as a current pulse waveform. ;in For discrete time points, This represents the current amplitude at that time point; the collected pulse waveform should completely cover the entire process of the motor from startup and operation to the completion of energy storage.

[0035] S2: For each data point on the current pulse waveform, obtain the glitch index of the data point; obtain the glitch index threshold; obtain an iterative gated smoothing algorithm based on the glitch index threshold, glitch index, and local mean of the glitch index; perform iterative smoothing processing on the data point based on the gated smoothing algorithm, and assemble the finally processed data points into a pulse waveform structure diagram.

[0036] It should be noted that the structure of a real pulse waveform is continuous in time, and adjacent data points exhibit strong local correlation and trend consistency; while noise spikes are isolated, instantaneous abrupt changes, showing substantial breaks or asymmetry in morphology compared to their immediate neighbors; for example, a noise spike... It will be far higher than its neighbors and ,and and The values ​​are very close; this invention utilizes this local asymmetry to locate burrs.

[0037] Specifically, to assess this asymmetry, the present invention uses current pulse waveforms... Each data point For the first In the next iteration, its spurt index is constructed:

[0038]

[0039] in, For the first Second iteration time point The burr index; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For positive numbers, such as ; It is the absolute value symbol; The value of starts from 0, when hour, Represents the current pulse waveform without iterative smoothing. .

[0040] In the formula, the denominator The calculation is Differences between neighbors; molecule The calculation is The sum of the differences between a point and its neighbors; when When the point is a noise burr, It is an isolated spike, with a value much greater than or much less than and ,and and The values ​​are very close, which causes the denominator to approach 0, while the numerator is very large, resulting in a spike index. It will become extremely large; when When the point is a real slope or inflection point, , , It shows a monotonically increasing or decreasing trend, with the numerator and denominator values ​​being nearly equal, leading to... Approaching 1; when When the point represents the true peak, the denominator is large, resulting in a high spurt index. It will remain at a relatively small value; therefore, It can amplify the values ​​of noise and glitch points while keeping the values ​​of real structure points around 1.

[0041] Furthermore, based on the glitch index, this invention designs an iterative gated smoothing algorithm for pulse waveform structure diagrams. Each data point The process is iteratively updated to generate the final pulse waveform structure diagram. :

[0042]

[0043]

[0044]

[0045]

[0046] In the formula, For the first Second iteration time point The amplitude of the pulse waveform, Pulse waveform ; For the first Second iteration time point The amplitude of the pulse waveform; The smoothing coefficient controls the smoothing rate. For the first Second iteration time point The local mean, that is, the mean of its immediate and next neighbors; In the first The next iteration is based on The calculated burr index; For the first Second iteration time point The threshold for the burr index; For gated functions;

[0047] In each iteration, based on the current pulse waveform Calculate the glitch index for all data points Check each data point Gating function ,if Less than or equal to That is, the actual structural point. When the value equals 0, the smoothing term in the updated formula becomes 0. equal The amplitude of the pulse waveform at that point is completely preserved; if Greater than This refers to the noise burr point. When the value equals 1, the value of the spike point will be pulled closer to the mean of its neighbors; this process is repeated iteratively for all data points, for example, 10 times, or until the pulse waveform no longer changes significantly; the resulting... This is the pulse waveform structure diagram. In this diagram, all glitch indices are greater than... All points are smoothed, and all spur exponents are less than or equal to 1. The true structure was completely preserved.

[0048] More specifically, the spurious index threshold satisfies the expression:

[0049]

[0050] In the formula, It is the first Second iteration time point The threshold for the burr index; It is the first Second iteration time point The amplitude of the pulse waveform; It is the first Second iteration time point The amplitude of the pulse waveform; For the first The average value of the pulse waveform amplitude during each iteration; This is the total time; It's a normal number, an empirical value. .

[0051] The spurious index threshold is from a fixed value. It departs and adjusts in real time based on a dynamically calculated adjustment coefficient. The numerator of the adjustment coefficient combines the maximum deviation of the current pulse waveform from its maximum pulse waveform amplitude and the average change between adjacent data points, used to capture isolated glitches and high-frequency jitter, respectively; the denominator is the maximum pulse waveform amplitude of the current pulse waveform. When glitches or jitter are severe in the current pulse waveform, the numerator of the adjustment coefficient increases, causing the final threshold to automatically increase, making the system more tolerant of these noise disturbances; the denominator of the adjustment coefficient uses the maximum pulse waveform amplitude for normalization: when the motor is operating at high current, the maximum pulse waveform amplitude increases, which suppresses excessive amplification of the threshold, making the judgment standard match the dynamic range of the signal; while at low current, the maximum pulse waveform amplitude is smaller, so the threshold is relatively increased to adapt to the judgment requirements under low signal strength.

[0052] The reasonable range for the smoothing coefficient is usually between (0,1). If the smoothing coefficient is 0, the algorithm will not produce a smoothing effect; if the smoothing coefficient is greater than or equal to 1 or less than 0, the updated value may cross the local mean or diverge in the wrong direction, leading to algorithm instability. The smoothing coefficient is used to control the smoothing rate. Choosing a larger value will make the noise spikes converge to their local mean faster, which is suitable for scenarios with extremely high noise amplitudes that require rapid smoothing; choosing a smaller value will result in a slower smoothing rate, which is suitable for scenarios with weak noise or where a more gradual smoothing process is desired. In this embodiment of the invention, the example value of the smoothing coefficient is 0.5. This value provides a balanced weighting between the current value of the spike and the local mean, enabling the spike to stably and effectively approach the mean of its neighbors, ensuring a smoothing effect while avoiding instability caused by excessively rapid updates.

[0053] The number of iterations should be a positive integer, and its selection is usually matched with the smoothing coefficient. If the number of iterations is set too low, noise spikes may not be sufficiently pulled closer to their local mean, resulting in incomplete smoothing; if the number of iterations is too high, although it has little impact on the final result, it will increase the unnecessary computational burden. Generally, when the smoothing coefficient is small or the noise amplitude is extremely large, a larger number of iterations is required; when the smoothing coefficient is large or the noise is weak, a smaller number of iterations is sufficient to converge. In this embodiment of the invention, the example value is 10 iterations. This value is an empirical setting that matches a smoothing coefficient of 0.5. Its purpose is to ensure that the algorithm has enough opportunities to iterate repeatedly, so that noise spikes can be sufficiently smoothed, and finally a pulse waveform structure diagram in which the pulse waveform no longer changes significantly is obtained.

[0054] For example, Figure 2 This is a diagram of the pulse waveform structure after smoothing and iteration. The blue line in the diagram represents the original pulse waveform; the red line represents the smoothed pulse waveform.

[0055] S3: Locate the starting point, peak point, and steady-state point on the pulse waveform structure diagram, and extract the structural peak integral and structural rise rate based on the starting point, peak point, and steady-state point to form a feature vector; establish a health model and fault feature library of the feature vector.

[0056] It should be noted that after obtaining a clear and glitch-free pulse waveform structure diagram, this invention needs to extract diagnostic features that can accurately reflect the true physical characteristics of the motor. Traditional methods, due to their high dependence on a single maximum peak feature, are easily distorted by glitch interference on noisy pulse waveforms, leading to evaluation failure. This invention believes that features reflecting the overall shape of the pulse waveform, such as the area integral representing the total starting energy and the average slope representing the starting acceleration, are more robust physical quantities than single data points because they are overall assessments of an interval and are not sensitive to small perturbations at individual points. To accurately calculate these integrals and slope features, this invention needs to precisely locate the inflection points of the pulse waveform, namely the starting point, peak point, and steady-state point. These inflection points cannot be reliably identified on noisy pulse waveforms. Therefore, this invention utilizes a clear pulse waveform structure diagram precisely to accurately locate these key inflection points, thereby abandoning traditional unstable peak features and extracting more stable and physically meaningful structural peak integrals and structural rise rate features.

[0057] Specifically, in the complete pulse waveform structure diagram Precisely locate key structural points; calculate Second derivative sequence Location of the starting point That is, to search The first significantly positive peak point corresponds to the inflection point where the current begins to rise rapidly; pinpoint the peak point. That is, to search The maximum value point; the steady-state point. That is, to search after, The point at which the price first falls back and remains near zero corresponds to the point where the motor enters steady-state operation.

[0058] For example, Figure 3 This is a schematic diagram of locating key structural points on a complete pulse waveform structure diagram. The blue broken line is the smoothed pulse waveform, the green dashed line is the time at the start point, and the intersection of the green dashed line and the blue broken line is the current of the smoothed pulse waveform at the start point; the red dashed line is the time at the peak point, and the intersection of the red dashed line and the blue broken line is the current of the smoothed pulse waveform at the peak point; the purple dashed line is the time at the steady-state point, and the intersection of the purple dashed line and the blue broken line is the current of the smoothed pulse waveform at the steady-state point.

[0059] Furthermore, based on the above structural points, this invention extracts a set of feature vectors; features That is, the structural peak integral, this feature The calculation is performed during the startup phase, that is, from the starting point. to steady state point Pulse waveform structure diagram within the time interval The amplitude of the pulse waveform is higher than the steady-state current value. The total area enclosed by the portion; this feature represents the total charge or energy consumption during the startup process, compared to a single maximum peak value. It is not sensitive to any residual noise glitches and has high robustness.

[0060] Furthermore, features That is, the structural rise rate is calculated from the starting point. to peak point The average slope, through the peak point current With starting point current The difference, divided by the peak time With the start point time The difference is obtained; due to and All are burr-free Precisely extracted from the above, It can accurately reflect the starting acceleration of the motor, unaffected by noise. Misjudgment or The impact of deviation.

[0061] S4: Calculate the feature vector of the newly acquired motor pulse waveform and the Mahalanobis distance of the health model, and make a fault judgment based on the comparison results of the Mahalanobis distance and the health threshold, as well as the comparison results with the fault feature library.

[0062] It should be noted that the extracted structured features This constitutes a robust description of the motor's state; the present invention aims to establish a mapping relationship between these features and the motor's healthy and faulty states, thereby realizing automatic diagnosis of energy storage motor faults.

[0063] Specifically, multiple sets of current pulse waveforms from known healthy energy storage motors are collected. For each set of pulse waveforms, the aforementioned data acquisition, pulse waveform reshaping, and feature extraction operations are performed to obtain a series of normal feature vectors. Calculate the mean of these vectors. Covariance Matrix This is the health model.

[0064] Furthermore, for known typical faults, such as coil aging or mechanical stall, their fault pulse waveforms are collected; similarly, the aforementioned data acquisition, pulse waveform reshaping, and feature extraction operations are performed to obtain fault feature clusters.

[0065] For the newly acquired motor pulse waveform, perform the aforementioned data acquisition, pulse waveform reshaping, and feature extraction operations to obtain its feature vector. Calculate the new eigenvector. With health model Mahalanobis distance between ; in response to Less than or equal to the preset health threshold If it is normal, then it is considered normal; in response to Greater than If a fault is detected, it is considered abnormal, and further compared with a fault feature database, for example, using the minimum distance method, to pinpoint the specific fault type. Once a fault is diagnosed, the system immediately issues an alarm, prompting maintenance personnel to inspect the circuit breaker, thus realizing a fault diagnosis method for energy storage motors based on waveform shaping.

[0066] Thus, a fault diagnosis method for energy storage motors based on pulse waveform shaping has been realized.

[0067] This invention also discloses a fault diagnosis system for energy storage motors based on waveform shaping, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a fault diagnosis method for energy storage motors based on waveform shaping according to this invention.

[0068] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A fault diagnosis method for energy storage motors based on waveform shaping, characterized in that, include: Acquire and preprocess the current pulse waveform of the energy storage motor; For each data point on the current pulse waveform, the glitch index of that data point is obtained by evaluating the local asymmetry between that data point and its neighboring data points. Including: data points in the 1st The time point of the next iteration The burr index satisfies the expression: ； For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform; These are positive numbers, empirical values. Ampere, the range of values ​​is Ampere Amperes are used to prevent the denominator from being zero. The absolute value sign is used; the glitch index threshold is obtained based on the degree of abrupt change and jitter of the current pulse waveform; an iterative gated smoothing algorithm is obtained based on the glitch index threshold, the glitch index, and the local mean of the glitch index, including: for the ... Second iteration time point pulse waveform amplitude Updated, number Pulse waveform amplitude in the next iteration Satisfying the expression: ； For smoothing coefficients; In the first Second iteration Local mean of a point; For gated functions; The gated function satisfies the expression: ; For the first Second iteration time point The threshold for the burr index; The gated smoothing algorithm only smooths data points whose spurt index is greater than the spurt index threshold, and retains data points whose spurt index is less than or equal to the spurt index threshold; the smoothing process is iteratively performed on the data points based on the gated smoothing algorithm, and the spurt index is recalculated according to the current pulse waveform in each iteration, and the finally processed data points are combined into a pulse waveform structure diagram; The starting point, peak point, and steady-state point are located on the pulse waveform structure diagram. Based on the starting point, peak point, and steady-state point, the structural peak integral and structural rise rate are extracted to form a feature vector. A health model and fault feature library of the feature vector are established. The Mahalanobis distance between the feature vector of the newly acquired motor pulse waveform and the health model is calculated. Based on the comparison results of the Mahalanobis distance with the health threshold and the comparison results with the fault feature library, a fault judgment is made.

2. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 1, characterized in that, The local mean satisfies the expression: ； in, For the first Second iteration time point The amplitude of the pulse waveform; For the first Second iteration time point The amplitude of the pulse waveform.

3. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 1, characterized in that, Locating the start point, peak point, and steady-state point on the pulse waveform structure diagram includes: Calculate the second derivative of the pulse waveform structure diagram; find the first positive peak point of the second derivative and locate it as the starting point; find the maximum value point of the pulse waveform structure diagram and locate it as the peak point; after finding the peak point, locate the point where the absolute value of the second derivative first falls back and remains near zero.

4. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 1, characterized in that, The extraction of the structural peak integral includes: Obtain the pulse waveform amplitude at all time points between the start point and the steady-state point; obtain the steady-state current value corresponding to the steady-state point; calculate the difference between the pulse waveform amplitude and the steady-state current value at each time point; multiply all differences by the sampling time interval and sum them up to obtain the peak integral of the structure.

5. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 1, characterized in that, The method for obtaining the structural rise rate is as follows: Obtain the difference between the pulse waveform amplitude corresponding to the peak point and the pulse waveform amplitude corresponding to the start point to obtain the pulse waveform amplitude difference; obtain the difference between the time of the peak point and the time of the start point; The rise rate of the structure is obtained by dividing the difference in amplitude of the pulse waveform by the difference in time.

6. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 1, characterized in that, The establishment of the health model of feature vectors includes: Multiple sets of current pulse waveforms from known healthy energy storage motors are collected. For each set of healthy pulse waveforms, the following operations are performed: calculating the glitch index for each data point on the current pulse waveform; smoothing the data points using an iterative gated smoothing algorithm to generate a pulse waveform structure diagram; and locating the start point, peak point, and steady-state point on the pulse waveform structure diagram and extracting feature vectors. The mean and covariance matrix of the feature vectors of the current pulse waveforms from multiple sets of known healthy energy storage motors are calculated, and the mean and covariance matrix are used as the health model.

7. The method for fault diagnosis of energy storage motor based on waveform shaping according to claim 2, characterized in that, The first Second iteration time point The spurious exponent threshold satisfies the expression: ； In the formula, It is the first Second iteration time point The threshold for the burr index; It is the first Second iteration time point The amplitude of the pulse waveform; It is the first Second iteration time point The amplitude of the pulse waveform; For the first The average value of the pulse waveform amplitude during each iteration; This is the total time; These are positive numbers, empirical values. Ampere (A), the range of values ​​is Ampere Ampere is used to prevent the denominator from being zero.

8. A fault diagnosis system for energy storage motors based on waveform shaping, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a fault diagnosis method for an energy storage motor based on waveform shaping according to any one of claims 1-7.

Images