Motor fault detection method and system based on harmonic cycle kurtosis

Abstract

The invention discloses a motor fault detection method and system based on harmonic cycle kurtosis, and the method comprises the steps: extracting an amplitude spectrum based on an original vibration signal, and carrying out the spectrum amplitude modulation of the amplitude spectrum through different weights, and obtaining a plurality of enhanced amplitude spectrums; performing normalization processing on the enhanced amplitude spectrum and the amplitude spectrum, and calculating a difference value between the normalized amplitude spectrum and the enhanced amplitude spectrum to obtain difference spectrums corresponding to different weights; reconstructing all the difference spectrums to obtain a plurality of reconstructed signals, and selecting an optimal reconstructed signal based on a harmonic cycle kurtosis index; and carrying out envelope spectrum analysis on the optimal reconstruction signal, and extracting a fault characteristic frequency to carry out motor fault detection. According to the method, the difference spectrum is constructed and the reconstructed signal is optimized, so that fault features can be highlighted under strong noise interference; through the harmonic cycle kurtosis index, the optimal reconstruction signal is automatically selected, the fault type can be effectively and accurately identified, and the accuracy and efficiency of washing machine motor fault detection in a real scene are improved.

Description

Technical Field

[0001] This invention relates to the field of motor fault detection technology, and in particular to a motor fault detection method and system based on harmonic cyclic kurtosis. Background Technology

[0002] As a household electric appliance, the washing machine relies on a motor power system to drive the inner drum and pulsator to rotate alternately in opposite directions, cleaning clothes through water rinsing, tumbling, and detergent breakdown. Due to its advantages such as high washing efficiency, precise energy consumption control, convenient operation, and diverse functions, it is widely used in household washing, rinsing, and spin-drying of everyday clothes, and has also expanded into areas such as centralized laundry systems in hotels, disinfection and cleaning of bedding in medical institutions, and cleaning of industrial protective clothing. As the key power source for the washing machine's washing function, the motor power system operates under complex conditions such as load changes, frequent start-stop cycles, high torque impacts, and humid and hot environments, making it prone to failure and directly affecting the overall stability of the machine. Common motor power system failure modes include those involving motor bearings, rotors, stators, and gears. Under complex operating conditions for extended periods, components such as bearings, rotors, stators, and gears in the motor power system often experience wear, cracks, and even breakage. If these faults are not detected and addressed promptly, they can lead to washing and spin-drying failures or motor burnout, threatening equipment safety and impacting user experience. In the trend of intelligent manufacturing, real-time health monitoring and stable operation assurance capabilities have become core competitive elements for washing machines. Therefore, fault detection of the power system components of the washing machine motor becomes particularly important.

[0003] The Chinese patent document "A Bearing Fault Monitoring Method and Monitoring Device Using the Same, and a Washing Machine," publication number CN111189640A, published on May 22, 2020, describes a method for monitoring bearings. This method involves acquiring the vibration acceleration signal of a monitored bearing, including bearings in the washing machine motor and bearings connecting the inner and outer tubs. The method amplifies and denoises the vibration acceleration signal. Based on the denoised vibration acceleration signal, it extracts the bearing fault characteristic frequency and calculates the theoretical value of the fault frequency. The actual extracted value of the bearing fault characteristic frequency and the theoretical value of the fault frequency are compared to determine if the monitored bearing has failed. If a fault occurs, the location of the fault is determined, and the result is displayed to the user. This method is simple, highly operable, and can monitor the bearing status and fault location in real time, promptly notifying the user to repair or replace the bearing. However, in the operating environment of a washing machine, background noise dominates the vibration signal. This broadband noise interference not only masks the fault resonance band but also prevents traditional narrowband demodulation methods from effectively separating and extracting the fault characteristic frequency, resulting in inaccurate fault detection results. Summary of the Invention

[0004] The present invention aims to overcome the problem in the prior art that background noise dominates the vibration signal of washing machines in complex operating environments, and that its broadband noise interference not only masks the fault resonance band, but also makes it impossible for traditional narrowband demodulation methods to effectively separate characteristic frequencies. The invention provides a motor fault detection method and system based on harmonic cyclic kurtosis.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A motor fault detection method based on harmonic cyclic kurtosis includes: The amplitude spectrum is extracted based on the original vibration signal. The amplitude spectrum is modulated by different weights to obtain several enhanced amplitude spectra. The enhanced amplitude spectrum and the amplitude spectrum are normalized, and the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum is calculated to obtain the difference spectrum corresponding to different weights. All differential spectra are reconstructed to obtain several reconstructed signals, and the optimal reconstructed signal is selected based on the harmonic cyclic kurtosis index. Envelope spectrum analysis is performed on the optimal reconstructed signal to extract fault characteristic frequencies for motor fault detection.

[0006] This invention first modulates and normalizes the vibration signal to obtain a normalized amplitude spectrum and a normalized enhanced amplitude spectrum. Then, it calculates the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum under different weights to obtain the difference spectrum corresponding to different weights. Multiple reconstructed signals are obtained from these differences. Next, the optimal reconstructed signal is selected using a harmonic cyclic kurtosis index. Finally, envelope spectrum analysis is performed on the optimal reconstructed signal to detect the fault type of the motor power system. On one hand, by subtracting the normalized amplitude spectrum from the normalized enhanced amplitude spectrum, the difference spectrum is obtained, thereby enhancing fault characteristics under strong noise interference. On the other hand, a harmonic cyclic kurtosis index is proposed to automatically select the optimal reconstructed signal. This effectively and accurately identifies the fault type of components in the washing machine motor power system, highlights fault characteristics under strong noise interference, and improves the accuracy and efficiency of fault detection in real-world washing machine motor power systems.

[0007] Preferably, the reconstruction of all difference spectra to obtain several reconstructed signals includes: The difference spectrum corresponding to different weights is reconstructed into a time-domain signal by using inverse Fourier transform, thus obtaining the reconstructed signal under different weights.

[0008] Preferably, the selection of the optimal reconstructed signal based on the harmonic cyclic kurtosis index includes: Calculate the harmonic cyclic kurtosis index of all reconstructed signals, and select the reconstructed signal with the largest index result as the optimal reconstructed signal.

[0009] Preferably, the calculation of the harmonic cyclic kurtosis index includes: Calculate the corresponding harmonic amplitude product based on the envelope spectrum of the reconstructed signal, and calculate the signal period T using the fundamental frequency corresponding to the maximum value of the harmonic amplitude product; calculate the square of the signal envelope SE of the reconstructed signal. The ratio of the total autocorrelation characteristics of the square of the reconstructed signal envelope under different signal period delays to the autocorrelation value of the reconstructed signal at zero delay is used as the harmonic cyclic kurtosis index.

[0010] Preferably, the total amount of autocorrelation features is: The autocorrelation function Rse(hT) of the square of the signal envelope SE, with the signal period T as the delay interval and the delay coefficient h starting from 1 up to the preset value, is summed. The autocorrelation function Rse(hT) represents the degree of similarity between SE and itself after a delay of hT.

[0011] Preferably, the calculation of the harmonic amplitude product includes: For any fundamental frequency fi, the product of the envelope spectrum of the fundamental frequency and the envelope spectra of the harmonic frequencies from 2 times to num times the fundamental frequency is taken as the harmonic amplitude product.

[0012] Preferably, the method of using different weights to modulate the amplitude spectrum to obtain several enhanced amplitude spectra includes: presetting the weight value range and the interval step size, selecting different weights to perform exponentiation on the amplitude spectrum of the original vibration signal to obtain several enhanced amplitude spectra.

[0013] Preferably, obtaining the difference spectrum corresponding to different weights includes: Calculate the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum corresponding to any weight, retain the part of the difference with amplitude greater than or equal to 0, and set the part less than 0 to 0, to obtain the difference spectrum corresponding to that weight.

[0014] Preferably, the step of extracting fault feature frequencies for motor fault detection includes: A fault characteristic frequency library containing fault types and corresponding frequencies is pre-set. The extracted fault characteristic frequencies are compared with the fault characteristic frequency library to determine the fault type of the motor.

[0015] A motor fault detection system based on harmonic cyclic kurtosis includes: The feature extraction module constructs a difference spectrum by processing the acquired raw vibration signal through spectral amplitude modulation and normalization. The signal reconstruction and optimization module uses inverse Fourier transform to reconstruct the difference spectrum into a time-domain signal, and selects the optimal reconstructed signal through the harmonic cyclic kurtosis index. Fault Analysis and Detection Module: Performs envelope spectrum analysis on the optimal reconstructed signal, extracts fault characteristic frequencies, and performs fault detection.

[0016] The present invention has the following beneficial effects: Based on the difference spectrum amplitude modulation guided by harmonic cyclic kurtosis, it can accurately identify key fault features containing obvious periodicity from vibration signals, and achieve accurate classification of fault types through envelope spectrum analysis, thereby realizing high-precision detection of faults in the power system of washing machine motors; by constructing the difference spectrum and optimizing the reconstructed signal, it can highlight fault features under strong noise interference, thereby improving the overall robustness of fault detection. Attached Figure Description

[0017] Figure 1 This is a flowchart of a motor fault detection method based on harmonic cyclic kurtosis in this invention.

[0018] Figure 2 This is a schematic diagram of the bearing fault signal processing result of the motor in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the electronic rotor fault signal processing results in an embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of the motor stator fault signal processing results in an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0022] Current fault feature extraction methods for motor power systems are mostly developed based on envelope spectrum analysis. The core of these methods is to construct an optimal narrowband filter to capture the modulation sidebands near the natural frequency, and then demodulate the fault characteristic frequency. The effectiveness of this method depends on the fault resonance band formed by the natural frequency excited by the fault impact. When the constructed optimal narrowband filter can filter out the fault resonance band, the fault characteristic frequency can be demodulated through envelope spectrum analysis. However, in the operating environment of a washing machine, background noise dominates the vibration signal. Its broadband noise interference not only masks the fault resonance band but also prevents traditional narrowband demodulation methods from effectively separating the characteristic frequency.

[0023] To solve the above-mentioned technical problems, the present invention provides, as follows: Figure 1 The method for motor fault detection based on harmonic cyclic kurtosis, as shown, includes: The amplitude spectrum is extracted based on the original vibration signal. The amplitude spectrum is modulated by different weights to obtain several enhanced amplitude spectra. The enhanced amplitude spectrum and the amplitude spectrum are normalized, and the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum is calculated to obtain the difference spectrum corresponding to different weights. All differential spectra are reconstructed to obtain several reconstructed signals, and the optimal reconstructed signal is selected based on the harmonic cyclic kurtosis index. Envelope spectrum analysis is performed on the optimal reconstructed signal to extract fault characteristic frequencies for motor fault detection.

[0024] It should be noted that this invention first modulates and normalizes the vibration signal to obtain a normalized amplitude spectrum and a normalized enhanced amplitude spectrum. Next, it calculates the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum under different weights to obtain the difference spectrum corresponding to different weights. This difference spectrum is then reconstructed to obtain multiple reconstructed signals. The optimal reconstructed signal is then selected using a harmonic cyclic kurtosis index. Finally, envelope spectrum analysis is performed on the optimal reconstructed signal to detect the fault type of the motor power system. On one hand, by subtracting the normalized amplitude spectrum from the normalized enhanced amplitude spectrum, the difference spectrum is obtained, thereby enhancing fault characteristics under strong noise interference. On the other hand, a harmonic cyclic kurtosis index is proposed to automatically select the optimal reconstructed signal. This effectively and accurately identifies the fault type of components in the washing machine motor power system, highlighting fault characteristics under strong noise interference, and improving the accuracy and efficiency of fault detection in real-world scenarios. It not only helps to detect potential faults early, preventing equipment damage and safety accidents, but also significantly improves detection efficiency and accuracy, and reduces maintenance costs.

[0025] It is worth noting that the method of this invention not only improves the accuracy of fault detection but also optimizes the detection process and reduces complexity. This method has broad application prospects, applicable not only to industrial washing machines but also to other fields requiring similar fault detection; it is based on signal processing technology, particularly frequency domain analysis and time domain reconstruction. Its core idea is to identify fault characteristics by analyzing the vibration signals generated during motor operation. Specifically, the method includes the following steps.

[0026] Vibration Signal Acquisition: Motors generate different vibration modes under normal operation and fault conditions. Acquiring these vibration signals provides information reflecting the motor's operating status. Frequency Domain Analysis: Fourier transform converts the acquired time-domain vibration signals into frequency-domain signals, allowing analysis of the signal's frequency components and identification of specific frequencies associated with the fault. Difference Spectrum Construction: Difference spectra are constructed through spectral amplitude modulation and normalization to highlight fault characteristic frequencies and reduce noise interference. Time-Domain Reconstruction: Inverse Fourier transform reconstructs the difference spectrum into a time-domain signal, providing a foundation for further fault feature analysis. Fault Feature Extraction: Envelope spectrum analysis extracts fault characteristic frequencies from the reconstructed time-domain signal; these frequencies are associated with specific fault types in the motor. Fault Diagnosis: The extracted fault characteristic frequencies are compared with a predefined fault characteristic frequency library to determine the specific fault type of the motor.

[0027] As a specific implementation, the motor fault detection method requires obtaining the original vibration signal from the washing machine motor power system; performing a Fourier transform on the obtained original vibration signal to obtain the amplitude spectrum Amp(f) of the original vibration signal.

[0028] After calculating the amplitude spectrum Amp(f) of the original vibration signal, several enhanced amplitude spectra are obtained by spectral amplitude modulation using different weights. The weight values ​​are preset to a range and an interval step size. Different weights are selected to perform exponentiation on the amplitude spectrum of the original vibration signal, resulting in several enhanced amplitude spectra. Specifically, the enhanced amplitude spectrum Ampe(f) is the MO power of the amplitude spectrum Amp(f) as the corresponding weight; MO represents the specific weight, and its value range can be selected from -0.5 to 1.5. The interval compensation can be selected as 0.1, thus obtaining enhanced amplitude spectra corresponding to multiple weights.

[0029] Normalize the amplitude spectrum and all enhanced amplitude spectra separately. Specifically, you can choose Z-score normalization, normalizing based on the mean and variance of the amplitude spectrum, and normalizing based on the mean and variance of the enhanced amplitude spectrum. Calculate the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum corresponding to any weight. Retain the portion of the difference where the amplitude is greater than or equal to 0 (the frequency of positive amplitude in the difference is the potential fault frequency), and set the portion where the amplitude is less than 0 to 0, thus obtaining the difference spectrum corresponding to that weight. Iterate through all enhanced amplitude spectra to obtain the difference spectrum corresponding to each weight.

[0030] The difference spectrum corresponding to different weights is reconstructed into a time-domain signal using inverse Fourier transform, resulting in reconstructed signals under different weights. The harmonic cyclic kurtosis index of all reconstructed signals is then calculated, and the reconstructed signal with the largest index result is selected as the optimal reconstructed signal.

[0031] The core idea of ​​the harmonic cyclic kurtosis index is to first estimate the fundamental harmonic frequency based on the envelope spectrum. This index can adaptively select the optimal reconstructed signal, which is then used as prior knowledge and input into the cyclic kurtosis to measure the fault pulse intensity. The specific calculation process of the harmonic cyclic kurtosis index is as follows.

[0032] First, the envelope spectrum of the reconstructed signal is calculated. As one optional method for calculating the envelope spectrum, a Hilbert transform is performed on the reconstructed signal to obtain the corresponding analytic signal. Then, the magnitude of the analytic signal is calculated to obtain the corresponding envelope signal, which is then subjected to a Fourier transform to obtain the final envelope spectrum. Alternatively, other existing envelope spectrum calculation methods can be used to obtain the envelope spectrum ES(f) of the reconstructed signal, where f represents the corresponding frequency.

[0033] Then, the harmonic amplitude product is calculated. For any fundamental frequency fi, the product of the envelope spectrum of the fundamental frequency and the envelope spectra of the harmonic frequencies corresponding to multiples of the fundamental frequency from 2 to num times is taken as the harmonic amplitude product HAP(fi). The signal period T is then calculated using the fundamental frequency f0 corresponding to the maximum value of the harmonic amplitude product. The calculation method is that the signal period T is equal to the experimental sampling frequency Fs divided by the fundamental frequency f0, where the experimental sampling frequency Fs is the sampling frequency during data acquisition.

[0034] After calculating the signal period T, the square of the signal envelope SE of the reconstructed signal is calculated. Specifically, it is the square of the modulus of the analytic signal obtained by performing a Hilbert transform on the reconstructed signal. Then, the ratio of the total autocorrelation characteristics of the square of the signal envelope SE of the reconstructed signal under different signal period delays to the autocorrelation value of the reconstructed signal at zero delay is used as the harmonic cyclic kurtosis index HCK.

[0035] Specifically, the total autocorrelation feature is: the summation of the autocorrelation function Rse(hT) of the square of the signal envelope SE with a delay interval of signal period T and a delay coefficient h from 1 to a preset value (usually chosen as 3); the autocorrelation function Rse(hT) represents the degree of similarity between SE and itself after a delay of hT.

[0036] After selecting the optimal reconstructed signal, envelope analysis is performed on the obtained optimal reconstructed signal to extract the fault characteristic frequency of the signal. The extracted fault characteristic frequency is then compared with a pre-set fault characteristic frequency library to determine the fault type of the motor.

[0037] When determining the type of motor fault, comparing the extracted fault characteristic frequencies with the theoretical fault characteristic frequencies in a fault characteristic frequency library is the primary diagnostic method. This method is direct and effective because it relies on the degree of matching between the fault characteristic frequencies and known theoretical values. Besides directly comparing frequency values, the error between the extracted fault characteristic frequencies and theoretical frequencies can also be calculated. For example, absolute error: calculates the absolute difference between the extracted and theoretical frequencies. Relative error: calculates the ratio of the absolute error to the theoretical frequency, expressed as a percentage. Mean square error (MSE): calculates the average of the sum of squares of all frequency errors. Root mean square error (RMSE): calculates the square root of the mean square error, providing a standard measure of error. Mean absolute error (MAE): calculates the average of the absolute values ​​of all frequency errors. Maximum error: identifies the maximum value among all frequency errors, used to assess the error in the worst-case scenario. These error metrics are applicable not only to comparisons of single frequencies but also to comparisons of multiple frequencies or harmonics, providing richer evaluation tools for fault detection. These metrics allow for a more accurate assessment of the degree of matching between the fault characteristic frequencies and theoretical values, thereby improving the accuracy and reliability of fault detection.

[0038] As a specific example, a washing machine motor power system typically consists of key components such as a motor, reducer, drive belt, clutch, and brake. During operation, these components may experience wear, cracks, or breakage due to complex operating conditions such as load variations, frequent starts and stops, high torque impacts, and humid and hot environments. These faults cause abnormal vibrations in the motor power system, and these vibration signals carry characteristic information about the faults. For example, bearing damage will produce vibrations at a specific frequency, while rotor imbalance will cause periodic vibrations. Therefore, by analyzing these vibration signals, potential faults in the motor power system can be identified.

[0039] This embodiment provides a fault detection method based on such fault mechanism analysis. It analyzes and diagnoses faults in the motor power system by accurately extracting characteristic frequencies from the vibration signal. First, vibration signals from the motor during operation are acquired. Then, Fourier transform is used to convert the time-domain signal into a frequency-domain signal to analyze its frequency components. Next, difference spectra are constructed through spectral amplitude modulation and Z-score normalization to highlight fault characteristic frequencies and reduce noise interference. Then, inverse Fourier transform is used to reconstruct the difference spectrum into a time-domain signal, and envelope spectrum analysis is performed to extract fault characteristic frequencies. Finally, by comparing the extracted fault characteristic frequencies with a predefined fault characteristic frequency library, the specific fault type of the motor is determined.

[0040] It is worth noting that the motor fault detection method of the present invention is not only applicable to fault detection of motor power systems in industrial washing machines, but also widely applicable to motor power systems of other types of washing machines, including household washing machines, commercial washing machines, dryers, washer-extractors, and other laundry equipment. Although these devices differ in structure and application scenarios, the basic working principle and fault modes of their motor power systems are similar to those of industrial washing machines.

[0041] In fault detection of washing machine motor power systems, vibration signal acquisition is a crucial step. Motors generate different vibration modes under normal operation and fault conditions, and these vibration signals carry rich information reflecting the motor's operating status. High-precision vibration sensors are deployed in key parts of the motor, such as bearing locations, gearboxes, and the motor housing, to capture the vibration signals generated during operation. These sensors can monitor the motor's vibration modes in real time, providing fundamental data for fault detection.

[0042] The data sources for this invention are diverse. Signals can be collected from various key components of the washing machine motor power system, such as the motor, reducer, drive belt, clutch, and brake. The vibration, temperature, and sound signals generated by these components during operation are important data sources for fault detection. In addition to data directly collected from the motor power system, data can also be collected through sensors placed in different parts of the system, such as vibration sensors, temperature sensors, and sound sensors. These sensors can monitor the operating status of the motor power system in real time, providing rich data support for fault detection. Furthermore, other fault detection-related data can be considered, such as electrical parameters of the motor, such as current, voltage, power, and frequency, as well as operating parameters such as motor speed and load changes. This data can be acquired through the motor control system or measured by other external sensors (this collected data, like vibration signals, contains fault characteristic information; therefore, the data processing method and harmonic kurtosis index of this invention can also be used for feature extraction and fault detection). By comprehensively analyzing this data, the operating status of the motor power system can be more comprehensively evaluated, improving the accuracy and reliability of fault detection.

[0043] As an optional embodiment, a data acquisition step for a household washing machine motor power system under specific operating conditions is as follows: First, acquire the operating parameters and environmental parameters of the household washing machine motor power system. Operating parameters include the target speed and load type, while environmental parameters include indoor temperature and humidity. Specifically, the operating parameters input by the user via buttons or touchscreen are collected through the control panel. These parameters include the target speed setting, with a range of 50 RPM to 1000 RPM, and load types categorized as light load, medium load, and heavy load. Simultaneously, environmental parameters, including indoor temperature and relative humidity, are collected by temperature and humidity sensors installed indoors. The temperature and humidity sensors collect data at preset intervals (e.g., 30 seconds) and transmit the data to the control unit.

[0044] Secondly, the current operating mode of the washing machine motor power system is determined based on operating and environmental parameters. Specifically, after receiving the operating and environmental parameters, the control unit first determines the difference between the target speed and the current speed. When the difference is greater than a first preset threshold (e.g., 100 RPM), it automatically switches to high-speed operation mode to quickly reach the target speed. When the difference is within a preset range (e.g., from a second preset threshold of 50 RPM to a first preset threshold of 100 RPM), it selects an intelligent operation mode based on the load type setting. In this mode, the control unit dynamically adjusts the motor speed according to the load type and the changing trend of indoor temperature. When the difference is less than the second preset threshold (e.g., 50 RPM), it switches to energy-saving operation mode. In this mode, the motor operates at a preset low speed to maintain energy-saving effects. This automatic selection scheme of operating mode based on operating and environmental parameters can not only adapt to different user needs and environmental conditions, but also improve operating efficiency and energy utilization through intelligent mode switching.

[0045] The corresponding data sampling frequency and sampling duration are determined based on the operating mode. Specifically, the control unit automatically adjusts the data sampling strategy according to the determined operating mode. In high-speed operating mode, because the motor operates under continuous high load, more frequent monitoring of the operating status is required. Therefore, the data sampling frequency is set to a higher level (e.g., once every 10 seconds), and the sampling duration is set to continuous sampling. In intelligent operating mode, considering the dynamic changes in motor load, the data sampling frequency is set to a medium level (e.g., once every 30 seconds), and the sampling duration is set to the operating cycle. In energy-saving operating mode, because the operating load is relatively stable, the data sampling frequency is set to a lower level (e.g., once every 2 minutes), and the sampling duration is set to the operating segment.

[0046] Multi-dimensional indicator data is collected according to the data sampling frequency and duration. Specifically, the actual motor speed data is collected through a motor speed sensor; the motor load data is collected through a load sensor; and the motor operating temperature data is collected through a temperature sensor, according to the data sampling frequency and duration. The collected data undergoes preprocessing by a data processing module, including outlier removal, data smoothing, and standardization. The preprocessed data is stored in a data cache according to timestamps and is updated periodically. This adaptive sampling scheme based on operating modes ensures both the timeliness and completeness of data collection, while avoiding the storage and processing burden caused by redundant data collection. At the same time, the differentiated sampling strategy under different modes can better reflect the operating characteristics of the motor under various operating conditions.

[0047] In addition to a motor fault detection method based on harmonic cyclic kurtosis, this invention also provides a motor fault detection system based on harmonic cyclic kurtosis, comprising: Data acquisition module: Responsible for collecting signals such as vibration, temperature, current, and voltage from key components of the washing machine's motor power system. Alternative solutions include using different types of sensors or wireless sensor networks to adapt to varying monitoring needs and environmental conditions.

[0048] Data processing module: This module utilizes a large-capacity storage device to store the collected data and employs a high-performance processor to execute data processing algorithms, such as feature extraction, noise reduction, and model training. Alternative solutions may include using more advanced data processing techniques, such as machine learning algorithms, to improve the accuracy and efficiency of fault detection. In the embodiments of this invention, the main function of the data processing module is to preprocess the collected data, including removing outliers, smoothing the data, and standardizing it. The preprocessed data is then stored in a data cache according to its timestamp and the cache is updated periodically.

[0049] The feature extraction module performs a Fourier transform on the acquired time-domain vibration signal, converting it from the time domain to the frequency domain to obtain the signal's amplitude spectrum. Power operations are then performed on the amplitude spectrum using different weights to obtain several enhanced amplitude spectra. The vibration signal is processed through spectral amplitude modulation and Z-score normalization to construct a difference spectrum to enhance fault characteristics. An inverse Fourier transform is then used to reconstruct the difference spectrum into a time-domain signal, providing a foundation for further analysis of fault characteristics.

[0050] The signal reconstruction and optimization module uses inverse Fourier transform to reconstruct the difference spectrum into a time-domain signal, and selects the optimal reconstructed signal through the harmonic cyclic kurtosis index.

[0051] The fault analysis and detection module performs envelope spectrum analysis on the optimal reconstructed signal, extracts fault characteristic frequencies, and compares them with a predefined fault characteristic frequency library to identify the specific fault type of the motor. To improve the accuracy and reliability of fault detection, multiple error indices are used to evaluate the degree of matching between the extracted fault characteristic frequencies and theoretical values, such as absolute error, relative error, mean square error, root mean square error, mean absolute error, or maximum error.

[0052] It should be noted that in the system of this invention, the data acquisition module is responsible for collecting vibration signals from key components of the washing machine motor's power system. These signals are the basis for fault detection; potential faults are identified by analyzing the spectral characteristics of the vibration signals. In addition to vibration signals, the following types of data can also be collected to provide more comprehensive information on the motor's operating status: the motor's current and voltage signals can reflect its operating status, including problems such as overload, underload, or imbalance. By analyzing these signals, electrical faults in the motor can be further identified. Temperature changes in the motor and other key components can reveal faults caused by overheating. Temperature sensors can be deployed in key areas such as the motor housing, gearbox, and drive belt to monitor temperature changes in real time. Sound analysis can monitor gear wear and meshing in the reducer, as well as the operational reliability and response time of the clutch. Sound sensors can be installed near the motor's junction box or control panel to capture sound signals during operation.

[0053] The system primarily receives vibration signals collected from the washing machine motor power system, which may include other operating parameters such as current, voltage, and temperature. The system outputs fault detection results, including an assessment of the fault type and severity. This invention determines the specific fault type in the motor power system by comparing extracted fault characteristic frequencies with a predefined fault characteristic frequency library. Although this invention primarily relies on traditional signal processing techniques, alternative approaches can be considered, such as using different types of sensors or sensor deployment schemes, like wireless sensor networks, to adapt to different monitoring needs and environmental conditions. In addition to vibration signals, more operating parameters such as motor speed and load changes can be collected to provide more comprehensive information on the motor's operating status. The system also includes sufficient communication interfaces to transmit data from the sensors to the data processing center via wired or wireless communication.

[0054] It is worth noting that the fault detection system of this invention includes several key components, which may be located in the same or different positions within the washing machine motor power system. The core of the system is a fault detection computer, which receives sensor data from various key components of the washing machine motor power system, including but not limited to vibration sensors, temperature sensors, and current sensors. These sensors are responsible for collecting key parameters such as vibration, temperature, current, and voltage during motor operation and transmitting this data to the fault detection computer for processing.

[0055] The fault detection computer can be connected to an output device, such as a computer monitor and / or a display or touchscreen device, to provide fault diagnosis results to the user. In some embodiments, the fault detection computer can be integrated with the washing machine motor power system, for example, it can be built into the motor's control panel. In other embodiments, the fault detection computer can be located remotely, receiving sensor data and performing fault diagnosis via wired or wireless communication.

[0056] Furthermore, fault detection systems can also be deployed in cloud environments, utilizing cloud services for data processing and analysis. For example, Microsoft Azure, Amazon Web Services, or Google Cloud Services can be used to achieve remote fault monitoring and diagnosis. In this scenario, the fault detection computer can receive data from sensors and transmit it to a quality monitoring server in the cloud for processing. This cloud deployment approach offers greater flexibility and scalability, while also facilitating system maintenance and upgrades.

[0057] This invention relates to a fault detection system specifically designed for washing machine motor power systems. It employs a series of signal processing techniques to achieve fault detection. The core of the system lies in its ability to process and analyze vibration signals collected from the motor power system, as well as other relevant operating parameters such as current, voltage, and temperature. By collecting the motor's vibration signals and other operating parameters, and through steps such as preprocessing, Fourier transform, difference spectrum construction, signal reconstruction, and envelope spectrum analysis, fault characteristic frequencies are extracted and compared with a predefined fault characteristic frequency library to identify potential motor faults. In the process of converting time-domain signals to frequency-domain signals, for non-stationary signals, wavelet transform or short-time Fourier transform can be optionally used, as these methods provide better time-frequency analysis capabilities. In fault feature extraction, other signal analysis techniques such as Hilbert-Huang transform or empirical mode decomposition can optionally be used, which can further reveal the intrinsic characteristics of the signal and enhance the accuracy of fault diagnosis.

[0058] Example 1: Washing Machine Motor Bearing Fault Detection. In this example, the goal is to identify faults in the washing machine motor bearing. Vibration signals from the motor bearing were collected, and it was noted that the theoretical fault characteristic frequency of the bearing is 45.73Hz. Preliminary frequency domain analysis indicates that identifying fault characteristics directly from the raw data is difficult.

[0059] The fault detection method of this invention performs in-depth analysis of vibration signals. This includes obtaining the amplitude spectrum using Fourier transform, and then constructing a difference spectrum through spectral amplitude modulation and Z-score normalization. The difference spectrum is reconstructed into a time-domain signal through inverse Fourier transform, and the optimal reconstructed signal is selected using the harmonic cyclic kurtosis index. Through envelope spectrum analysis, the bearing fault characteristic frequency of 45Hz and its harmonics were clearly demodulated, which matches the theoretical characteristic frequency value of the bearing fault, successfully detecting the fault type as a washing machine motor bearing fault.

[0060] like Figure 2 The results of the vibration signal analysis are shown below. Figure 2 (a) is the time-domain spectrum of the original signal. Figure 2 (b) is the frequency domain spectrum of the original signal after Fourier transform. Figure 2 (c) shows the envelope spectrum of the original signal after envelope analysis. Analysis of the original signal reveals that it is difficult to detect bearing fault characteristics from it. Furthermore, the method of this invention is used to analyze the signal. Figure 2 (d) is the time-domain spectrum of the optimal reconstructed signal selected in this invention. Figure 2 (e) The frequency domain spectrum of the optimal reconstructed signal after Fourier transform is selected for this invention. Figure 2 (f) is the envelope spectrum of the optimal reconstructed signal selected in this invention after envelope analysis.

[0061] Example 2: Detection of Misalignment in Washing Machine Motor Rotors. In this embodiment, vibration signals are collected using vibration sensors installed on the motor rotor, based on a motor rotation frequency of 60Hz. These signals are first converted to generate a time-domain spectrum, and then the frequency-domain spectrum is obtained through Fourier transform.

[0062] Next, feature extraction and optimization steps were performed on these frequency domain data. This included constructing a difference spectrum using spectral amplitude modulation and Z-score normalization, and reconstructing the difference spectrum into a time-domain signal using inverse Fourier transform. Then, the harmonic cyclic kurtosis index was used to select the optimal reconstructed signal. In the fault feature analysis and detection stage, envelope spectrum analysis was performed on the optimal reconstructed signal to extract fault-related characteristic frequencies. These characteristic frequencies were compared with a predefined fault characteristic frequency library to determine the specific fault type of the motor. Ultimately, a rotor misalignment fault was successfully identified, with its characteristic frequency of 120.04 Hz and its harmonics matching theoretical values. Figure 3 The results of the vibration signal analysis are shown below. Figure 3 (a) is the time-domain spectrum of the original signal. Figure 3 (b) The envelope spectrum of the optimal reconstructed signal selected in this invention after envelope analysis.

[0063] Example 3: Washing Machine Motor Stator Fault Detection. In this embodiment, a fault in the washing machine motor stator was detected. Vibration signals from the motor stator were collected from the system, and the theoretical fault characteristic frequency of the motor stator was recorded as 100Hz. The time-domain signal was converted to the frequency domain using Fourier transform, and its frequency spectrum was analyzed. Preliminary analysis showed that directly detecting fault characteristics from the raw signal was challenging. After applying the method of this invention, the signal was further processed, including spectral amplitude modulation, Z-score normalization, and a difference spectrum was constructed. The difference spectrum was reconstructed into a time-domain signal using inverse Fourier transform, and the optimal reconstructed signal was selected using the harmonic cyclic kurtosis index. Finally, the fault characteristic frequency of 100.15Hz and its harmonics were extracted from the optimal reconstructed signal, which perfectly matched the theoretical characteristic frequency value of the stator fault, thus confirming the fault type as a washing machine motor stator fault. Figure 4 The results of the vibration signal analysis are shown below. Figure 4 (a) is the time-domain spectrum of the original signal. Figure 4 (b) The envelope spectrum of the optimal reconstructed signal selected in this invention after envelope analysis.

[0064] The above embodiments are further elaborations and descriptions of the present invention to facilitate understanding, and are not intended to limit the present invention in any way. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A motor fault detection method based on harmonic cyclic kurtosis, characterized in that, include: Amplitude spectra are extracted from the original vibration signals, and several enhanced amplitude spectra are obtained by using different weights to modulate the amplitude spectra. The enhanced amplitude spectrum and the amplitude spectrum are normalized, and the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum is calculated to obtain the difference spectrum corresponding to different weights. All difference spectra are reconstructed to obtain several reconstructed signals, and the optimal reconstructed signal is selected based on the harmonic cyclic kurtosis index. Envelope spectrum analysis is performed on the optimal reconstructed signal to extract fault characteristic frequencies for motor fault detection.

2. The motor fault detection method based on harmonic cyclic kurtosis according to claim 1, characterized in that, The reconstruction of all difference spectra yields several reconstructed signals, including: The difference spectrum corresponding to different weights is reconstructed into a time-domain signal by using inverse Fourier transform, thus obtaining the reconstructed signal under different weights.

3. A motor fault detection method based on harmonic cyclic kurtosis according to claim 1 or 2, characterized in that, The selection of the optimal reconstructed signal based on the harmonic cyclic kurtosis index includes: Calculate the harmonic cyclic kurtosis index of all reconstructed signals, and select the reconstructed signal with the largest index result as the optimal reconstructed signal.

4. A motor fault detection method based on harmonic cyclic kurtosis according to claim 1 or 2, characterized in that, The calculation of the harmonic cyclic kurtosis index includes: Calculate the corresponding harmonic amplitude product based on the envelope spectrum of the reconstructed signal, and calculate the signal period T using the fundamental frequency corresponding to the maximum value of the harmonic amplitude product; calculate the square of the signal envelope SE of the reconstructed signal. The ratio of the total autocorrelation characteristics of the square of the reconstructed signal envelope under different signal period delays to the autocorrelation value of the reconstructed signal at zero delay is used as the harmonic cyclic kurtosis index.

5. The motor fault detection method based on harmonic cyclic kurtosis according to claim 4, characterized in that, The total amount of the autocorrelation features is: The autocorrelation function Rse(hT) of the square of the signal envelope SE, with the signal period T as the delay interval and the delay coefficient h starting from 1 up to the preset value, is summed. The autocorrelation function Rse(hT) represents the degree of similarity between SE and itself after a delay of hT.

6. The motor fault detection method based on harmonic cyclic kurtosis according to claim 4, characterized in that, The calculation of the harmonic amplitude product includes: For any fundamental frequency fi, the product of the envelope spectrum of the fundamental frequency and the envelope spectra of the harmonic frequencies from 2 times to num times the fundamental frequency is taken as the harmonic amplitude product.

7. A motor fault detection method based on harmonic cyclic kurtosis according to claim 1, 2, 5, or 6, characterized in that, The method of using different weights to modulate the amplitude spectrum to obtain several enhanced amplitude spectra includes: By setting the weight value range and interval step size, different weights are selected to perform exponentiation on the amplitude spectrum of the original vibration signal to obtain several enhanced amplitude spectra.

8. The motor fault detection method based on harmonic cyclic kurtosis according to claim 7, characterized in that, The difference spectrum corresponding to different weights is obtained as follows: Calculate the difference between the normalized amplitude spectrum and the enhanced amplitude spectrum corresponding to any weight, retain the part of the difference with amplitude greater than or equal to 0, and set the part with amplitude less than 0 to 0, to obtain the difference spectrum corresponding to that weight.

9. A motor fault detection method based on harmonic cyclic kurtosis according to claim 1, 2, 5, 6, or 8, characterized in that, The step of extracting fault feature frequencies for motor fault detection includes: A fault characteristic frequency library containing fault types and corresponding frequencies is pre-set. The extracted fault characteristic frequencies are compared with the fault characteristic frequency library to determine the fault type of the motor.

10. A motor fault detection system based on harmonic cyclic kurtosis, applicable to the motor fault detection method as described in any one of claims 1-9, characterized in that, include: The feature extraction module constructs a difference spectrum by processing the acquired raw vibration signal through spectral amplitude modulation and normalization. The signal reconstruction and optimization module uses inverse Fourier transform to reconstruct the difference spectrum into a time-domain signal, and selects the optimal reconstructed signal through the harmonic cyclic kurtosis index. Fault Analysis and Detection Module: Performs envelope spectrum analysis on the optimal reconstructed signal, extracts fault characteristic frequencies, and performs fault detection.

Images