Motor fault diagnosis method and system based on linkage of voiceprint features and frequency conversion parameters

Abstract

The invention discloses a motor fault diagnosis method and system based on linkage of voiceprint features and frequency conversion parameters, and belongs to the technical field of mechanical vibration and acoustic measurement, and the method comprises the steps: obtaining the operation frequency range of a motor, dividing a feature frequency band, collecting the historical voiceprint signals of motor operation, carrying out the feature extraction, and generating a dynamic baseline threshold library; selecting a dynamic baseline threshold value corresponding to the characteristic frequency band to carry out voiceprint characteristic comparison, outputting a primary fault diagnosis result, controlling the frequency converter to execute micro-amplitude frequency sweeping in a preset frequency amplitude range, collecting a frequency sweeping voiceprint signal, and judging whether a fault continuously exists or not; and if yes, performing energy decoupling on the sweep frequency voiceprint signal, calculating to obtain real fault intensity, and generating a frequency-fault intensity mapping graph and a fault classification result. According to the method, the dynamic baseline threshold library is established, micro-amplitude frequency sweeping is executed to verify the fault continuity, frequency conversion electromagnetic interference is decoupled, the classification result is finally generated, and high-precision and anti-interference motor fault quantitative diagnosis can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical vibration and acoustic measurement, and in particular to a motor fault diagnosis method and system based on the linkage between soundprint features and frequency conversion parameters. Background Art

[0002] In the field of mechanical vibration measurement, motor fault diagnosis is often achieved by analyzing acoustic signatures during operation. These signatures contain acoustic wave information generated by the vibrations of the motor's mechanical structure. Their frequency and time domain characteristics can reveal fault conditions such as bearing damage and rotor imbalance. For variable-frequency motors in particular, the dynamic changes in operating frequency create a strong correlation between acoustic signatures and the frequency conversion parameters.

[0003] Existing technologies typically use a fixed threshold method for fault diagnosis. A single sensor collects soundprint signals within a specific frequency band, extracts energy amplitude or harmonic components as features, and compares them with a preset fixed threshold. Static models are built based on vibration sensor data, ignoring the interference of electromagnetic noise on soundprint signals during frequency conversion. Some solutions incorporate frequency segmentation but fail to dynamically divide frequency bands based on the motor's resonance characteristics.

[0004] However, fixed thresholds cannot adapt to motor load fluctuations and changes in operating frequency, resulting in an increased false alarm rate; the contamination of the inverter's electromagnetic harmonics on the soundprint characteristics is not taken into account, and electromagnetic interference is easily misjudged as a mechanical failure; there is a lack of a verification mechanism for fault persistence, making it difficult to distinguish between transient anomalies and real damage; in addition, existing methods are unable to quantify fault intensity and accurately locate characteristic frequencies, which restricts the accuracy of fault classification. Summary of the Invention

[0005] To solve the above problems, the present invention provides a motor fault diagnosis method and system based on the linkage of voiceprint features and frequency conversion parameters. By establishing a dynamic baseline threshold library, performing micro-frequency sweeps to verify fault persistence, and decoupling frequency conversion electromagnetic interference, it finally generates a frequency-fault intensity mapping diagram and classification results, which can achieve high-precision, interference-resistant quantitative diagnosis of motor faults.

[0006] The above objectives can be achieved through the following solutions: A motor fault diagnosis method based on the linkage of soundprint features and frequency conversion parameters includes obtaining the operating frequency range of the motor, dividing the operating frequency range into multiple characteristic frequency bands according to the resonance point of the motor; collecting historical soundprint signals of the motor operation in each characteristic frequency band and performing feature extraction to generate a dynamic baseline threshold library; real-time monitoring of the actual operating frequency and actual soundprint signal of the motor, selecting the corresponding dynamic baseline threshold in the dynamic baseline threshold library for soundprint feature comparison according to the characteristic frequency band where the actual operating frequency is located, and outputting a primary fault diagnosis result; according to the primary fault diagnosis result, controlling the inverter to perform a micro-frequency sweep within a preset frequency amplitude range, collecting the frequency sweep soundprint signal during the frequency sweep process, and judging whether the fault persists based on the frequency sweep soundprint signal; if the fault persists, performing energy decoupling on the frequency sweep soundprint signal to calculate the true fault intensity; based on the true fault intensity, generating a frequency-fault intensity mapping diagram and a fault classification result.

[0007] Optionally, obtaining the operating frequency range of the motor and dividing the operating frequency range into multiple characteristic frequency bands according to the resonance point of the motor includes: obtaining historical vibration spectrum data of the motor, identifying the points where vibration energy suddenly increases as resonance points; and dividing the operating frequency range into high frequency bands, medium frequency bands and low frequency bands with adjacent resonance points as boundaries.

[0008] Optionally, the collecting of historical soundprint signals of the motor operation in each characteristic frequency band and performing feature extraction to generate a dynamic baseline threshold library includes: collecting historical soundprint signals of the motor operation in each characteristic frequency band, and extracting the preset fault frequency band energy, full-band energy, sound wave signal amplitude, fundamental sound energy, and electromagnetic harmonic sound energy of the corresponding characteristic frequency band; calculating the ratio of the preset fault frequency band energy to the full-band energy as the frequency domain energy focusing value, and calculating the benchmark energy focusing value based on the frequency domain energy focusing value; counting the number of times the sound wave signal amplitude exceeds the preset impact threshold per unit time as the time domain pulse density, and calculating the benchmark pulse density based on the time domain pulse density; measuring the ratio of the fundamental sound energy to the electromagnetic harmonic sound energy as the harmonic distortion rate, and calculating the benchmark harmonic distortion rate based on the harmonic distortion rate; integrating the benchmark energy focusing value, benchmark pulse density and benchmark harmonic distortion rate of each characteristic frequency band to obtain a dynamic baseline threshold library.

[0009] Optionally, the output of the primary fault diagnosis result includes: real-time monitoring of the actual operating frequency and actual soundprint signal of the motor; based on the actual operating frequency, selecting the baseline energy focusing value, baseline pulse density and baseline harmonic distortion rate of the corresponding characteristic frequency band from the dynamic baseline threshold library to obtain a baseline threshold set; performing feature extraction on the actual soundprint signal to obtain the actual energy focusing value, actual pulse density and actual harmonic distortion rate, and the actual feature set; calculating the deviation between the actual feature set and the baseline threshold set, and generating the primary fault diagnosis result according to the deviation.

[0010] Optionally, the calculation of the deviation between the actual feature set and the benchmark threshold set, and generating a primary fault diagnosis result based on the deviation includes: calculating the actual pulse density deviation based on the actual pulse density and the corresponding benchmark pulse density; calculating the actual energy focusing deviation based on the actual energy focusing value and the corresponding benchmark energy focusing value; calculating the actual distortion deviation based on the actual harmonic distortion rate and the corresponding benchmark harmonic distortion rate; judging whether the actual pulse density deviation is greater than a preset pulse density deviation threshold, or the actual energy focusing deviation is greater than a preset energy focusing deviation threshold, or the actual distortion deviation is greater than a preset distortion deviation threshold; if so, determining that a fault exists; if not, determining that a fault does not exist.

[0011] Optionally, judging whether the fault persists based on the frequency sweeping soundprint signal includes: performing feature extraction on the frequency sweeping soundprint signal to determine the fault feature strength at the start of the frequency sweep; obtaining the fault feature strength at the end of the frequency sweep based on the fault feature strength at the start of the frequency sweep, and calculating the attenuation rate of the fault feature strength; when the attenuation rate of the fault feature strength is less than a preset attenuation threshold, determining that the fault persists.

[0012] Optionally, the feature extraction of the swept frequency soundprint signal and determination of the fault feature strength at the start of the sweep includes: feature extraction of the swept frequency soundprint signal at the start of the sweep to obtain the swept frequency energy focusing value, the swept frequency pulse density and the swept frequency harmonic distortion rate, and the swept frequency feature set; based on the swept frequency feature set and the reference threshold set, calculating the swept frequency pulse density deviation, the swept frequency energy focusing deviation and the swept frequency distortion deviation; sorting the swept frequency pulse density deviation, the swept frequency energy focusing deviation and the swept frequency distortion deviation from large to small, and selecting the value ranked first as the fault feature strength at the start of the sweep.

[0013] Optionally, if the fault persists, the sweeping soundprint signal is energy decoupled, and the true fault intensity is calculated, including: if the fault persists, the percentage value of the real-time output power of the motor and the rated power during the sweeping process is collected to obtain the motor load rate; based on the sweeping soundprint signal, the sweeping harmonic distortion rate during the sweeping process is extracted; based on the motor load rate and the sweeping harmonic distortion rate, the variable frequency interference factor is calculated; and the true fault intensity is calculated using the variable frequency interference factor and the fault feature intensity during the sweeping process.

[0014] Optionally, generating a frequency-fault intensity mapping diagram and a fault classification result based on the real fault intensity includes: generating a frequency-fault intensity mapping diagram using the sampling frequency points within the frequency sweep range and the real fault intensity corresponding to the sampling frequency points; obtaining a real fault intensity peak value based on the frequency-fault intensity mapping diagram; when the real fault intensity peak value is less than a preset first threshold value, outputting a normal classification; when the real fault intensity peak value is greater than or equal to the first threshold value and less than a preset second threshold value, outputting a minor fault classification; when the real fault intensity peak value is greater than or equal to the second threshold value, outputting a serious fault classification; when the minor fault classification or the serious fault classification is output, obtaining a fault characteristic frequency according to the real fault intensity peak value and the frequency-fault intensity mapping diagram; and obtaining a fault classification result by matching the fault characteristic frequency with a preset fault characteristic frequency library.

[0015] Based on the same inventive concept, the present invention also provides a motor fault diagnosis system based on the linkage of soundprint features and frequency conversion parameters, the system comprising: a frequency band division module for obtaining the operating frequency range of the motor and dividing the operating frequency range into multiple characteristic frequency bands according to the resonance point of the motor; a dynamic baseline generation module for collecting historical soundprint signals of the motor operation in each characteristic frequency band and performing feature extraction to generate a dynamic baseline threshold library; a forward diagnosis module for real-time monitoring the actual operating frequency and actual soundprint signal of the motor, selecting the corresponding dynamic baseline threshold in the dynamic baseline threshold library according to the characteristic frequency band in which the actual operating frequency is located for soundprint feature comparison, and outputting a primary fault diagnosis result; a reverse diagnosis module for controlling the inverter to perform a micro-frequency sweep within a preset frequency amplitude range according to the primary fault diagnosis result, collecting the swept soundprint signal during the sweep process, and judging whether the fault persists based on the swept soundprint signal; a fault decoupling module for performing energy decoupling on the swept soundprint signal if the fault persists, and calculating the true fault intensity; and a fault determination module for generating a frequency-fault intensity mapping diagram and a fault classification result based on the true fault intensity.

[0016] Compared with the prior art, the present invention has the following advantages: 1. This invention significantly improves the robustness of fault diagnosis and early warning capabilities through an adaptive matching mechanism of a dynamic baseline threshold library. By dividing characteristic frequency bands based on motor resonance characteristics and establishing a multi-dimensional voiceprint feature reference library, it can adaptively match normal operating state characteristics under different working conditions, effectively reducing the risk of misjudgment caused by environmental noise and load fluctuations. 2. This invention uses micro-sweep frequency to stimulate fault response characteristics and combines them with attenuation rate analysis to accurately verify the persistence of the fault. This method uses the frequency converter to actively control the frequency of micro-variations, capturing the dynamic attenuation characteristics of the fault characteristic intensity, and effectively distinguishing between true mechanical faults and transient electromagnetic interference or random noise. 3. This invention uses the energy decoupling mechanism of motor load rate and harmonic distortion rate to eliminate the influence of inverter electromagnetic interference on voiceprint characteristics. By calculating the variable frequency interference factor and decoupling the electromagnetic noise component in the swept frequency voiceprint signal, the true fault intensity reflecting only mechanical damage is extracted, thereby improving the accuracy of fault quantitative assessment. 4. The present invention achieves collaborative determination of fault type and severity through a frequency-fault intensity mapping diagram; combined with the hierarchical comparison of the actual fault intensity peak and the preset threshold, and the matching of the fault characteristic frequency with the typical fault library, the fault classification result and severity level are synchronously output, providing a multi-dimensional and accurate basis for maintenance decision-making.

[0017] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 It is a flow chart of a motor fault diagnosis method based on the linkage between voiceprint features and frequency conversion parameters according to an embodiment of the present invention.

[0020] Figure 2 Schematic diagram of frequency band division according to an embodiment of the present invention.

[0021] Figure 3 It is a structural diagram of a motor fault diagnosis system based on the linkage between voiceprint features and frequency conversion parameters according to an embodiment of the present invention.

[0022] Figure 4This is a frequency-fault intensity mapping diagram according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0024] Reference Figure 1 One embodiment of the present invention proposes a motor fault diagnosis method based on the linkage between voiceprint features and variable frequency parameters. By establishing a dynamic baseline threshold library, performing micro-frequency sweeps to verify fault persistence, and decoupling variable frequency electromagnetic interference, it finally generates a frequency-fault intensity mapping diagram and classification results, which can achieve high-precision, interference-resistant quantitative diagnosis of motor faults.

[0025] The method of this embodiment specifically includes: Obtaining an operating frequency range of the motor, and dividing the operating frequency range into a plurality of characteristic frequency bands according to a resonance point of the motor; Collect historical soundprint signals of motor operation in each characteristic frequency band and perform feature extraction to generate a dynamic baseline threshold library; Monitor the actual operating frequency and actual soundprint signal of the motor in real time, select the corresponding dynamic baseline threshold in the dynamic baseline threshold library according to the characteristic frequency band of the actual operating frequency, perform soundprint feature comparison, and output a primary fault diagnosis result; According to the primary fault diagnosis result, controlling the frequency converter to perform a slight frequency sweep within a preset frequency amplitude range, collecting a frequency sweep soundprint signal during the frequency sweep process, and judging whether the fault persists based on the frequency sweep soundprint signal; If the fault persists, energy decoupling is performed on the frequency sweep voiceprint signal to calculate the true fault intensity; Based on the actual fault intensity, a frequency-fault intensity mapping diagram and a fault classification result are generated.

[0026] Specifically, a dynamic coordination mechanism between voiceprint features and frequency conversion parameters is established. First, the operating frequency range is divided according to the resonance characteristics of the motor to generate characteristic frequency bands with clear physical meanings. By collecting historical voiceprint signals, multi-dimensional features such as frequency domain energy focusing value, time domain pulse density and harmonic distortion rate are extracted to build an adaptive dynamic baseline threshold library. During real-time monitoring, the voiceprint feature deviation is calculated based on the benchmark threshold of the corresponding frequency band matched with the actual operating frequency to achieve primary fault diagnosis. If an abnormality is detected, the frequency converter is triggered to perform a micro-frequency sweep within the preset range, and the fault persistence is confirmed by analyzing the fault feature intensity attenuation rate of the swept frequency voiceprint signal. For persistent faults, the frequency conversion interference factor is calculated by combining the motor load rate and the swept frequency harmonic distortion rate, decoupling the influence of electromagnetic noise on the voiceprint features, and finally generating a frequency-fault intensity mapping diagram and classification results based on the actual fault intensity. This method significantly improves the reliability and accuracy of fault diagnosis. Adaptive matching of the dynamic baseline threshold library enhances the detection sensitivity of early-stage weak faults. Micro-amplitude frequency sweeps are used to stimulate fault response characteristics, combined with an energy decoupling mechanism, to effectively distinguish between true mechanical faults and transient electromagnetic interference. Frequency-fault intensity mapping enables quantitative assessment of fault type and severity, providing a clear basis for maintenance decisions. This also reduces the false positive rate caused by environmental noise and operating condition fluctuations, optimizing the robustness of the diagnostic process.

[0027] Optionally, obtaining the operating frequency range of the motor and dividing the operating frequency range into a plurality of characteristic frequency bands according to the resonance point of the motor includes: Obtain historical vibration spectrum data of the motor and identify the point where vibration energy suddenly increases as the resonance point; The operating frequency range is divided into high frequency band, medium frequency band and low frequency band with adjacent resonance points as boundaries.

[0028] Specifically, the historical vibration spectrum data of the motor is obtained. The data is collected by a vibration sensor installed on the motor housing, and the sampling frequency is not less than 2 times the maximum operating frequency of the motor. The method for identifying the vibration energy surge point is: perform peak detection on the vibration spectrum data, and when the difference between the vibration energy amplitude corresponding to a certain frequency point and the energy amplitude of the adjacent frequency points on its left and right exceeds the preset energy mutation threshold, the point is determined to be a vibration energy surge point, that is, a resonance point. The operating frequency range of the motor is divided into continuous characteristic frequency bands with the frequency interval between all adjacent resonance points as the boundary, specifically divided into a high frequency band (from the highest resonance point to the maximum allowable frequency of the motor), a medium frequency band (the main working frequency band between two adjacent resonance points) and a low frequency band (from the lowest allowable frequency of the motor to the lowest resonance point). For example, Figure 2As shown in the figure, when the resonance points f1 = 25Hz, f2 = 50Hz, and f3 = 75Hz are identified, the low frequency band is divided into 0-25Hz, the mid-frequency band is divided into 25-50Hz and 50-75Hz, and the high frequency band is divided into 75-100Hz. By identifying the resonance points, the area with differentiated motor frequency response characteristics is accurately segmented, which improves the physical significance of the characteristic frequency band division, makes the subsequent soundprint feature comparison more consistent with the actual dynamic behavior of the motor, and avoids misjudgment caused by unreasonable frequency band division.

[0029] Optionally, collecting historical soundprint signals of motor operation in each characteristic frequency band and performing feature extraction to generate a dynamic baseline threshold library includes: Collect historical soundprint signals of the motor running in each characteristic frequency band, and extract the preset fault frequency band energy, full frequency band energy, sound wave signal amplitude, fundamental wave sound energy, and electromagnetic harmonic sound energy of the corresponding characteristic frequency band; Calculating a ratio of a preset fault frequency band energy to a full frequency band energy as a frequency domain energy focusing value, and calculating a reference energy focusing value based on the frequency domain energy focusing value; Counting the number of times the amplitude of the acoustic wave signal exceeds a preset impact threshold per unit time as a time-domain pulse density, and calculating a reference pulse density based on the time-domain pulse density; Measuring a ratio of fundamental wave acoustic energy to electromagnetic harmonic acoustic energy as harmonic distortion rate, and calculating a reference harmonic distortion rate based on the harmonic distortion rate; The benchmark energy focusing value, benchmark pulse density and benchmark harmonic distortion rate of each characteristic frequency band are integrated to obtain the dynamic baseline threshold library.

[0030] Specifically, for each characteristic frequency band, multiple soundprint signal samples of historical operating cycles are collected, and each sample is time series data. First, perform fast Fourier transform (FFT) on the historical soundprint signal to obtain a frequency domain representation; extract the preset fault frequency band energy, which is obtained by integrating and calculating the sum of the squares of the signal energy amplitudes within the preset fault frequency band (for example, the common frequency band of motor bearing faults is 2000-5000Hz); extract the full-band energy, which is obtained by integrating and calculating the sum of the squares of the signal energy amplitudes within the entire frequency range (for example, 0-10000Hz); extract the acoustic wave signal amplitude, and directly read the peak value or root mean square value from the time domain signal; extract the fundamental sound energy, and extract the energy amplitude at the current operating fundamental frequency of the motor through FFT; extract the electromagnetic harmonic sound energy, and extract the sum of the harmonic component energy amplitudes at integer multiples of the fundamental frequency through FFT. Calculate the frequency domain energy focus value : , in is the preset fault band energy, The energy of the full frequency band is obtained by integrating the energy through FFT. The dimensions are the same as the energy unit, and the ratio is dimensionless. Based on the frequency domain energy focusing value, the benchmark energy focusing value is calculated, which is obtained by taking the arithmetic average of the frequency domain energy focusing values ​​of multiple historical samples. The time domain pulse density is counted by setting a unit time (for example, 1 second) and a preset impact threshold (based on the historical signal amplitude statistics), and counting the number of times the acoustic wave signal amplitude exceeds the preset impact threshold within the unit time. : , in is the number of times exceeding the threshold, The unit time length is times / time. Based on the time domain pulse density, the benchmark pulse density is calculated by taking the arithmetic average of the time domain pulse density of multiple historical samples. : , in is the electromagnetic harmonic sound energy, is the fundamental sound energy, both of which are obtained through FFT energy extraction, with the same dimension as energy units, and the ratio is dimensionless. Based on the harmonic distortion rate, the benchmark harmonic distortion rate is calculated, which is obtained by statistically averaging the harmonic distortion rates of multiple historical samples. The benchmark energy focusing value, benchmark pulse density and benchmark harmonic distortion rate of each characteristic frequency band are integrated to form a dynamic baseline threshold library, which stores three benchmark value data sets corresponding to each characteristic frequency band. The benchmark threshold is established through statistical analysis of the characteristics of historical soundprint signals to capture the frequency domain energy distribution, time domain pulse characteristics and harmonic component ratio under normal operating conditions of the motor, thereby providing a reliable reference for real-time fault diagnosis. Enhance the robustness and adaptability of fault feature extraction, reduce the risk of misjudgment caused by environmental noise interference, improve the sensitivity to early weak faults, and optimize diagnostic accuracy through multi-dimensional feature integration.

[0031] Optionally, outputting the primary fault diagnosis result includes: Real-time monitoring of the actual operating frequency and actual soundprint signal of the motor; Based on the actual operating frequency, selecting a baseline energy focusing value, a baseline pulse density, and a baseline harmonic distortion rate corresponding to a characteristic frequency band from the dynamic baseline threshold library to obtain a baseline threshold set; Performing feature extraction on the actual voiceprint signal to obtain an actual energy focus value, an actual pulse density, an actual harmonic distortion rate, and an actual feature set; The deviation between the actual feature set and the reference threshold set is calculated, and a primary fault diagnosis result is generated according to the deviation.

[0032] Optionally, calculating the deviation between the actual feature set and the reference threshold set, and generating a primary fault diagnosis result according to the deviation includes: Calculating an actual pulse density deviation based on the actual pulse density and the corresponding reference pulse density; Calculating an actual energy focusing deviation according to the actual energy focusing value and the corresponding reference energy focusing value; Calculating an actual distortion deviation based on the actual harmonic distortion rate and the corresponding reference harmonic distortion rate; Determining whether the actual pulse density deviation is greater than a preset pulse density deviation threshold, or the actual energy focus deviation is greater than a preset energy focus deviation threshold, or the actual distortion deviation is greater than a preset distortion deviation threshold; If so, it is determined that a fault exists; If not, it is determined that no fault exists.

[0033] Specifically, according to the above method, the actual pulse density obtained by extracting the actual voiceprint signal characteristics and the reference pulse density of the corresponding characteristic frequency band in the dynamic baseline threshold library are used to calculate the actual pulse density deviation. : , in The number of times the amplitude of the acoustic wave signal exceeds the preset impact threshold per unit time is counted in real time, with the dimension being times per time. The actual energy focus deviation is calculated based on the actual energy focus value extracted from the actual voiceprint signal feature and the benchmark energy focus value in the benchmark threshold set. : , in It is the ratio of the preset fault frequency band energy to the full frequency band energy calculated in real time, dimensionless; The actual distortion deviation is calculated based on the actual harmonic distortion rate obtained by extracting the actual voiceprint signal features and the benchmark harmonic distortion rate in the benchmark threshold set. : , in It is the ratio of fundamental wave sound energy to electromagnetic harmonic sound energy measured in real time, dimensionless; is a historical benchmark value and is dimensionless. It is determined whether the actual pulse density deviation is greater than the preset pulse density deviation threshold, whether the actual energy focus deviation is greater than the preset energy focus deviation threshold, and whether the actual distortion deviation is greater than the preset distortion deviation threshold. The preset threshold is determined based on the historical operating data statistics of the motor model. If any deviation exceeds its corresponding threshold, it is determined that a fault exists; if all deviations do not exceed the corresponding threshold, it is determined that no fault exists. The primary fault diagnosis result is output in the form of a Boolean value. Through multi-dimensional comparison of real-time voiceprint features with historical benchmark thresholds, feature deviations are detected to identify anomalies, and the frequency domain and time domain features are complementary to enhance diagnostic reliability. This significantly improves the accuracy and robustness of fault diagnosis, reduces the false alarm rate caused by environmental noise interference, and improves the detection sensitivity of weak early faults.

[0034] Optionally, judging whether the fault persists based on the frequency sweep voiceprint signal includes: Extracting features from the frequency sweep voiceprint signal to determine the strength of the fault feature at the start of the frequency sweep; Based on the fault characteristic strength at the start of the frequency sweep, the fault characteristic strength at the end of the frequency sweep is obtained, and the attenuation rate of the fault characteristic strength is calculated; When the decay rate of the fault characteristic intensity is less than a preset decay threshold, it is determined that the fault persists.

[0035] Optionally, the extracting features from the frequency sweep voiceprint signal to determine the fault feature strength at the start of the frequency sweep includes: Extract features of the sweeping voiceprint signal at the start of the sweeping frequency to obtain a sweeping frequency energy focus value, a sweeping frequency pulse density, a sweeping frequency harmonic distortion rate, and a sweeping frequency feature set; Based on the frequency sweep feature set and the reference threshold set, a frequency sweep pulse density deviation, a frequency sweep energy focus deviation, and a frequency sweep distortion deviation are calculated; The frequency sweep pulse density deviation, the frequency sweep energy focus deviation, and the frequency sweep distortion deviation are sorted from large to small, and the value ranked first is selected as the fault feature intensity at the start time of the frequency sweep.

[0036] Specifically, based on the reference threshold set used in the judgment of the primary fault diagnosis result, according to the above method, the sweep pulse density deviation, sweep energy focus deviation and sweep distortion deviation are calculated, and the sweep pulse density deviation, sweep energy focus deviation and sweep distortion deviation are sorted from large to small. The value with the highest ranking is selected as the fault feature strength at the start of the sweep, and the attenuation rate of the fault feature strength is calculated. : , in is the fault feature strength at the start of the frequency sweep, obtained by extracting the signal features at the start point of the frequency sweep and calculating the deviation; is the fault feature strength at the end of the sweep, which is obtained by extracting the signal features at the end of the sweep and calculating the deviation. If the fault feature strength at the start of the sweep is the sweep energy focus deviation, then the fault feature strength at the end of the sweep is also the sweep energy focus deviation. The frequency sweep interval is obtained from the inverter control parameters and is measured in seconds. The decay rate is measured in seconds. The decay rate is determined to be less than a preset decay threshold, which is set based on historical motor normal operation data and is measured in seconds. If the decay rate is less than the preset decay threshold, the fault is considered persistent; otherwise, the fault is considered non-persistent. A small frequency sweep is used to stimulate the motor's dynamic response and analyze the rate of change in the fault signature intensity. Slow decay indicates that the fault signature is stable and persistent, rather than a transient disturbance, thereby enhancing diagnostic reliability. This effectively distinguishes transient anomalies such as electromagnetic noise from true mechanical faults, reducing false positives and improving the accuracy of confirming persistent faults.

[0037] Optionally, if the fault persists, performing energy decoupling on the frequency sweep voiceprint signal to calculate the true fault intensity includes: If the fault persists, the percentage of the motor's real-time output power to the rated power during the frequency sweep process is collected to obtain the motor load rate; Extracting a frequency sweep harmonic distortion rate during the frequency sweep process based on the frequency sweep voiceprint signal; Calculating a variable frequency interference factor based on the motor load rate and the swept frequency harmonic distortion rate; The actual fault intensity is calculated using the variable frequency interference factor and the fault characteristic intensity during the frequency sweep process.

[0038] Specifically, when the fault is determined to be persistent, the real-time output power of the motor during the frequency sweeping process is first collected by the power sensor, and the rated power value of the motor is obtained, and the motor load rate is calculated to be equal to the real-time output power divided by the rated power; wherein the real-time output power is directly measured by the power sensor, and the rated power is the motor nameplate parameter or preset value. Then, based on the frequency sweeping soundprint signal collected during the frequency sweeping process, a fast Fourier transform analysis is performed to extract the sound energy of the frequency sweeping soundprint signal at the current operating fundamental frequency of the motor as the fundamental sound energy, and the sum of the harmonic component sound energy at the integer multiple frequency of the fundamental frequency is extracted as the electromagnetic harmonic sound energy, and the frequency sweeping harmonic distortion rate is calculated, which is obtained by the above method. Then, based on the motor load rate and the frequency sweeping harmonic distortion rate, the variable frequency interference factor is calculated. : , in is the motor load factor, is the harmonic distortion rate of the frequency sweep. Then, the fault feature strength is obtained. This value is the maximum deviation value in the frequency sweep feature set. The maximum value is determined by sorting the frequency sweep pulse density deviation, frequency sweep energy focus deviation and frequency sweep distortion deviation using the aforementioned method. The value in the frequency sweep process is obtained based on the determined deviation to obtain the fault feature strength in the frequency sweep process. That is, if the fault feature strength at the start of the frequency sweep is the frequency sweep distortion deviation, then the fault feature strength in the frequency sweep process is the corresponding frequency sweep distortion deviation in the frequency sweep process. Finally, the actual fault strength is calculated using the variable frequency interference factor and the fault feature strength. : , in, The fault characteristic intensity during the frequency sweep process is used to determine the actual fault intensity change during the frequency sweep process. By analyzing the voiceprint characteristics in conjunction with the frequency conversion parameters, the characteristic frequency bands are first divided according to the motor resonance point and a dynamic baseline threshold library is established to achieve primary fault diagnosis. The voiceprint signal is then collected during the micro-frequency sweep process to determine the fault persistence. Finally, energy decoupling is performed by combining the motor load factor and the frequency sweep harmonic distortion rate to eliminate frequency conversion interference, calculate the actual fault intensity, and generate a fault classification. This effectively distinguishes between true mechanical faults and electromagnetic noise interference, improves the reliability and accuracy of fault diagnosis, enhances the detection capability of weak early-stage faults, reduces the misjudgment rate, and optimizes the quantitative assessment of fault severity, providing more accurate support for motor maintenance decisions.

[0039] Optionally, generating a frequency-fault intensity mapping diagram and a fault classification result based on the actual fault intensity includes: Generate a frequency-fault intensity mapping diagram using sampling frequency points within the frequency sweep range and the actual fault intensity corresponding to the sampling frequency points; Obtaining a true fault intensity peak value based on the frequency-fault intensity mapping diagram; When the true fault intensity peak value is less than a preset first threshold, outputting a normal classification; When the true fault intensity peak is greater than or equal to the first threshold and less than a preset second threshold, outputting a minor fault classification; When the true fault intensity peak value is greater than or equal to the second threshold, outputting a serious fault classification; When outputting the minor fault classification or the severe fault classification, obtaining a fault characteristic frequency according to the true fault intensity peak and the frequency-fault intensity mapping diagram; The fault characteristic frequency is matched with a preset fault characteristic frequency library to obtain a fault classification result.

[0040] Specifically, first obtain the sequence of sampling frequency points recorded during the frequency sweep process and the true fault intensity value corresponding to each sampling frequency point; the sampling frequency point sequence is extracted from the inverter control log and is measured in Hertz; the true fault intensity value is calculated using the aforementioned method. A scatter plot is plotted with frequency as the horizontal axis and true fault intensity as the vertical axis, and a continuous curve is generated using cubic spline interpolation to form a frequency-fault intensity mapping diagram. The true fault intensity peak value is obtained from the frequency-fault intensity mapping diagram. The true fault intensity peak value is then read and compared with a preset first and second thresholds; the first and second thresholds are statistically obtained from a historical fault database based on the motor model, and the second threshold is greater than the first threshold. If the true fault intensity peak value is less than the first threshold, a normal classification result is output. If the true fault intensity peak value is greater than or equal to the first threshold and less than the second threshold, a minor fault classification result is output. If the true fault intensity peak value is greater than or equal to the second threshold, a major fault classification result is output. If a minor or major fault classification result is output, the frequency point corresponding to the true fault intensity peak value is located in the frequency-fault intensity mapping diagram and used as the fault characteristic frequency. Finally, the fault characteristic frequency is matched against a preset fault characteristic frequency library. This library, which stores characteristic frequency ranges corresponding to typical fault types (such as bearing outer race fault characteristic frequency ranges and rotor bar broken characteristic frequency ranges), is precalculated using motor model parameters and physical fault models. A nearest neighbor matching algorithm is used to calculate the absolute difference between the fault characteristic frequency and the median of each fault characteristic frequency range in the library. The fault type with the smallest difference is selected as the fault classification output. A frequency sweep process maps the frequency to the actual fault intensity. Combined with a multi-level threshold determination and fault characteristic frequency matching mechanism, this method achieves quantitative grading of motor fault severity and accurate identification of fault types, effectively improving the interpretability and maintenance guidance value of fault diagnosis results while avoiding misclassification issues caused by variable frequency interference.

[0041] Based on the same inventive concept, Figure 3 As shown, the present invention also provides a motor fault diagnosis system based on the linkage between voiceprint features and frequency conversion parameters, the system comprising: A frequency band division module, used to obtain the operating frequency range of the motor and divide the operating frequency range into multiple characteristic frequency bands according to the resonance point of the motor; The dynamic baseline generation module is used to collect historical soundprint signals of motor operation in each characteristic frequency band and perform feature extraction to generate a dynamic baseline threshold library; A forward diagnosis module is used to monitor the actual operating frequency and actual soundprint signal of the motor in real time, select the corresponding dynamic baseline threshold in the dynamic baseline threshold library according to the characteristic frequency band of the actual operating frequency, perform soundprint feature comparison, and output a primary fault diagnosis result; a reverse diagnosis module, configured to control the frequency converter to perform a slight frequency sweep within a preset frequency amplitude range based on the primary fault diagnosis result, collect a frequency sweep soundprint signal during the frequency sweep, and determine whether the fault persists based on the frequency sweep soundprint signal; A fault decoupling module is used to perform energy decoupling on the frequency sweep voiceprint signal if the fault persists, and calculate the true fault intensity; The fault determination module is used to generate a frequency-fault intensity mapping diagram and a fault classification result based on the actual fault intensity.

[0042] Example 1 To demonstrate the feasibility and advanced nature of this invention, it was applied to the logistics sorting system of a large automated warehouse. This system utilizes hundreds of inverter-controlled motors, driving high-speed conveyor belts. The stable operation of these motors is crucial to ensuring the efficiency of the entire warehouse. Traditional regular maintenance methods are costly and difficult to detect early failures. Sudden motor failures (such as bearing wear and rotor imbalance) often cause the sorting line to shut down, resulting in significant economic losses.

[0043] In this embodiment, the fault diagnosis system of the present invention was deployed in the automated warehousing center to continuously monitor and analyze data for 50 SEW-Eurodrive DFV100M4 variable frequency motors on one of the key sorting lines for 6 months.

[0044] The system first acquired historical vibration spectrum data for this motor model under normal operation. Using a peak detection algorithm, it identified significant structural resonance points near 25Hz and 60Hz. Therefore, the system divided the motor's rated operating frequency range (0-100Hz) into three characteristic frequency bands: low frequency (0-25Hz), mid-frequency (25-60Hz), and high frequency (60-100Hz). The system then collected historical soundprint signals from healthy motors operating for over 1000 hours within each characteristic frequency band. By extracting features from these signals, dynamic baseline thresholds were established for each characteristic frequency band. For the mid-frequency band (25-60Hz), for example, the baseline thresholds were determined as: a baseline energy focus value of 0.12, a baseline pulse density of 8 beats / second, and a baseline harmonic distortion rate of 0.05. This data constitutes a dynamic baseline threshold library, providing an adaptive reference standard for subsequent real-time diagnosis.

[0045] During the six-month monitoring period, the system continuously analyzed the motor's real-time operating frequency and acoustic signature. At 10:30 AM on November 15, 2024, the system detected an abnormal acoustic signature from motor M-07 on conveyor belt 7, operating at 48 Hz (a mid-frequency band). The system extracted its signature and compared it with the mid-frequency benchmark in the dynamic baseline library. The system found an actual pulse density of 32 pulses per second. The calculated deviation of the actual pulse density was (32-8) / 8 = 3.0, or 300%, far exceeding the preset deviation threshold of 100%. The system therefore issued a "primary fault" alarm. To verify the persistence of the fault, the system immediately controlled the motor's inverter and performed a small frequency sweep of ±2 Hz (from 46 Hz to 50 Hz) around the current operating frequency of 48 Hz for four seconds. The system collected the acoustic signature during the sweep and analyzed the pulse density, a key deviation indicator. At the start of the frequency sweep (46 Hz), the pulse density deviation was 2.9; at the end of the sweep (50 Hz), the deviation was 2.85. The calculated decay rate of the fault signature intensity was only (2.9 - 2.85) / 4 = 0.0125 / s. This value is far less than the decay threshold of 0.5 / s used to distinguish transient interference, so the system determines that the fault is a true, persistent mechanical fault.

[0046] After confirming the fault's persistence, the system entered the fault decoupling and quantification phase. The system measured the real-time load factor of motor M-07 during the frequency sweep as 0.75, and its harmonic distortion rate during the frequency sweep as 0.20. Based on this, the system calculated the frequency conversion interference factor as 0.75 × 0.20 = 0.15. Using this interference factor, the system decoupled the fault characteristic intensity (i.e., pulse density deviation) measured during the frequency sweep to eliminate the influence of electromagnetic noise. The calculated true fault intensity was 2.9 × (1 - 0.15) ≈ 2.47. This value more accurately reflects the degree of damage to the mechanical structure itself.

[0047] The system uses the actual fault intensity corresponding to each frequency point during the frequency sweep process to generate a "frequency-fault intensity mapping diagram". Figure 4 As shown in the figure, at 47.5Hz, the true fault intensity reaches a peak of 2.55. This peak is between the preset "minor fault" threshold (1.5) and "serious fault" threshold (4.0), so the system determines the severity level of the fault as "minor fault".

[0048] The system also matched the fault characteristic frequency of 47.5Hz, corresponding to the peak, with a pre-set fault characteristic frequency library. This frequency closely matched the characteristic frequency range of "early bearing outer race wear" in the library (45Hz-50Hz, coupled with the rotational frequency and harmonics). Ultimately, the system output a diagnostic result: "Motor M-07 has a minor bearing outer race wear fault. Inspection is recommended during the next scheduled maintenance window." Based on this accurate diagnosis, the maintenance team prepared spare parts in advance and replaced the motor's bearings over the weekend, avoiding an unplanned downtime and saving an estimated eight hours of production delay.

[0049] Table 1 Comparison of motor fault diagnosis accuracy and false alarm rate Table 2 Typical fault diagnosis event record It can be seen from the above data and tables that the method of the present invention has shown excellent performance in practical applications. Table 1 clearly shows the huge advantages of the present invention in early warning accuracy and false alarm rate compared with traditional methods, and proves the effectiveness of the dynamic baseline and micro-sweep verification mechanism. Table 2 records several typical diagnostic cases, which not only successfully identified real mechanical failures, but also accurately distinguished transient electromagnetic interference, avoided unnecessary maintenance, and fully demonstrated the system's anti-interference ability and high reliability. By quantifying the intensity of faults and accurately classifying them, the system provides the maintenance team with a clear and executable decision-making basis, and ultimately realizes the transformation from "passive maintenance" to "predictive maintenance", significantly improving the operational stability and economic benefits of the entire warehousing system.

[0050] It should be noted that the formulas appearing above can translate physical quantities of different properties into unitless standard values ​​or superimposable parameters of the same dimension through the principle of dimensional consistency and mathematical standardization means (such as normalization, dimensionless parameter conversion or unit system unification), thereby eliminating the interference of different dimensions on the operation logic, so that the formulas have mathematical operation rationality and objective law adaptability while retaining the distribution characteristics of the original data. It is a conventional technical means and will not be elaborated here. The electrical connection between the above-mentioned units does not necessarily mean a direct connection of the circuit. The indirect connection method can be applied to the embodiments of the present invention as long as the purpose of the present invention is achieved. The above is only an exemplary embodiment of the present invention and the scope of the present invention cannot be limited thereto.

[0051] That is, any equivalent changes and modifications made according to the teachings of the present invention are still within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art not described herein.

Claims

1. A motor fault diagnosis method based on the linkage between voiceprint features and frequency conversion parameters, characterized in that: The method comprises: Obtaining an operating frequency range of the motor, and dividing the operating frequency range into a plurality of characteristic frequency bands according to a resonance point of the motor; Collect historical soundprint signals of motor operation in each characteristic frequency band and perform feature extraction to generate a dynamic baseline threshold library; Monitor the actual operating frequency and actual soundprint signal of the motor in real time, select the corresponding dynamic baseline threshold in the dynamic baseline threshold library according to the characteristic frequency band of the actual operating frequency, perform soundprint feature comparison, and output a primary fault diagnosis result; According to the primary fault diagnosis result, controlling the frequency converter to perform a slight frequency sweep within a preset frequency amplitude range, collecting a frequency sweep soundprint signal during the frequency sweep process, and judging whether the fault persists based on the frequency sweep soundprint signal; If the fault persists, energy decoupling is performed on the frequency sweep voiceprint signal to calculate the true fault intensity; Based on the actual fault intensity, a frequency-fault intensity mapping diagram and a fault classification result are generated.

2. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 1 is characterized in that: The obtaining of the operating frequency range of the motor and dividing the operating frequency range into a plurality of characteristic frequency bands according to the resonance point of the motor includes: Obtain historical vibration spectrum data of the motor and identify the point where vibration energy suddenly increases as the resonance point; The operating frequency range is divided into high frequency band, medium frequency band and low frequency band with adjacent resonance points as boundaries.

3. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 1 is characterized in that: The collecting of historical soundprint signals of the motor running in each characteristic frequency band and feature extraction to generate a dynamic baseline threshold library includes: Collect historical soundprint signals of the motor running in each characteristic frequency band, and extract the preset fault frequency band energy, full frequency band energy, sound wave signal amplitude, fundamental wave sound energy, and electromagnetic harmonic sound energy of the corresponding characteristic frequency band; Calculating a ratio of a preset fault frequency band energy to a full frequency band energy as a frequency domain energy focusing value, and calculating a reference energy focusing value based on the frequency domain energy focusing value; Counting the number of times the amplitude of the acoustic wave signal exceeds a preset impact threshold per unit time as a time-domain pulse density, and calculating a reference pulse density based on the time-domain pulse density; Measuring a ratio of fundamental wave acoustic energy to electromagnetic harmonic acoustic energy as harmonic distortion rate, and calculating a reference harmonic distortion rate based on the harmonic distortion rate; The benchmark energy focusing value, benchmark pulse density and benchmark harmonic distortion rate of each characteristic frequency band are integrated to obtain the dynamic baseline threshold library.

4. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 3 is characterized in that: Outputting the primary fault diagnosis result includes: Real-time monitoring of the actual operating frequency and actual soundprint signal of the motor; Based on the actual operating frequency, selecting a baseline energy focusing value, a baseline pulse density, and a baseline harmonic distortion rate corresponding to a characteristic frequency band from the dynamic baseline threshold library to obtain a baseline threshold set; Performing feature extraction on the actual voiceprint signal to obtain an actual energy focus value, an actual pulse density, an actual harmonic distortion rate, and an actual feature set; The deviation between the actual feature set and the reference threshold set is calculated, and a primary fault diagnosis result is generated according to the deviation.

5. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 4 is characterized in that: The calculating the deviation between the actual feature set and the reference threshold set, and generating a primary fault diagnosis result according to the deviation includes: Calculating an actual pulse density deviation based on the actual pulse density and the corresponding reference pulse density; Calculating an actual energy focusing deviation according to the actual energy focusing value and the corresponding reference energy focusing value; Calculating an actual distortion deviation based on the actual harmonic distortion rate and the corresponding reference harmonic distortion rate; Determining whether the actual pulse density deviation is greater than a preset pulse density deviation threshold, or the actual energy focus deviation is greater than a preset energy focus deviation threshold, or the actual distortion deviation is greater than a preset distortion deviation threshold; If so, it is determined that a fault exists; If not, it is determined that no fault exists.

6. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 4 is characterized in that: The determining whether the fault persists based on the frequency sweep voiceprint signal includes: Extracting features from the frequency sweep voiceprint signal to determine the strength of the fault feature at the start of the frequency sweep; Based on the fault characteristic strength at the start of the frequency sweep, the fault characteristic strength at the end of the frequency sweep is obtained, and the attenuation rate of the fault characteristic strength is calculated; When the decay rate of the fault characteristic intensity is less than a preset decay threshold, it is determined that the fault persists.

7. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 6 is characterized in that: Extracting features from the frequency sweep voiceprint signal to determine the fault feature strength at the start of the frequency sweep includes: Extract features of the sweeping voiceprint signal at the start of the sweeping frequency to obtain a sweeping frequency energy focus value, a sweeping frequency pulse density, a sweeping frequency harmonic distortion rate, and a sweeping frequency feature set; Based on the frequency sweep feature set and the reference threshold set, a frequency sweep pulse density deviation, a frequency sweep energy focus deviation, and a frequency sweep distortion deviation are calculated; The frequency sweep pulse density deviation, the frequency sweep energy focus deviation, and the frequency sweep distortion deviation are sorted from large to small, and the value ranked first is selected as the fault feature intensity at the start time of the frequency sweep.

8. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 6 is characterized in that: If the fault persists, energy decoupling is performed on the swept frequency voiceprint signal to calculate the true fault intensity, including: If the fault persists, the percentage of the motor's real-time output power to the rated power during the frequency sweep process is collected to obtain the motor load rate; Extracting a frequency sweep harmonic distortion rate during the frequency sweep process based on the frequency sweep voiceprint signal; Calculating a variable frequency interference factor based on the motor load rate and the swept frequency harmonic distortion rate; The actual fault intensity is calculated using the variable frequency interference factor and the fault characteristic intensity during the frequency sweep process.

9. The motor fault diagnosis method based on voiceprint characteristics and frequency conversion parameters linkage according to claim 1 is characterized in that: The generating of a frequency-fault intensity map and a fault classification result based on the actual fault intensity includes: Generate a frequency-fault intensity mapping diagram using sampling frequency points within the frequency sweep range and the actual fault intensity corresponding to the sampling frequency points; Obtaining a true fault intensity peak value based on the frequency-fault intensity mapping diagram; When the true fault intensity peak value is less than a preset first threshold, outputting a normal classification; When the true fault intensity peak is greater than or equal to the first threshold and less than a preset second threshold, outputting a minor fault classification; When the true fault intensity peak value is greater than or equal to the second threshold, outputting a serious fault classification; When outputting the minor fault classification or the severe fault classification, obtaining a fault characteristic frequency according to the true fault intensity peak and the frequency-fault intensity mapping diagram; The fault characteristic frequency is matched with a preset fault characteristic frequency library to obtain a fault classification result.

10. A motor fault diagnosis system based on the linkage between voiceprint features and frequency conversion parameters, applied to the motor fault diagnosis method based on the linkage between voiceprint features and frequency conversion parameters as described in any one of claims 1 to 9, characterized in that: The system comprises: A frequency band division module, used to obtain the operating frequency range of the motor and divide the operating frequency range into multiple characteristic frequency bands according to the resonance point of the motor; The dynamic baseline generation module is used to collect historical soundprint signals of motor operation in each characteristic frequency band and perform feature extraction to generate a dynamic baseline threshold library; A forward diagnosis module is used to monitor the actual operating frequency and actual soundprint signal of the motor in real time, select the corresponding dynamic baseline threshold in the dynamic baseline threshold library according to the characteristic frequency band of the actual operating frequency, perform soundprint feature comparison, and output a primary fault diagnosis result; a reverse diagnosis module, configured to control the frequency converter to perform a slight frequency sweep within a preset frequency amplitude range based on the primary fault diagnosis result, collect a frequency sweep soundprint signal during the frequency sweep, and determine whether the fault persists based on the frequency sweep soundprint signal; A fault decoupling module is used to perform energy decoupling on the frequency sweep voiceprint signal if the fault persists, and calculate the true fault intensity; The fault determination module is used to generate a frequency-fault intensity mapping diagram and a fault classification result based on the actual fault intensity.

Images