AC asynchronous motor fault diagnosis method and device and storage medium thereof

Abstract

The invention discloses an alternating current asynchronous motor fault diagnosis method and device and a storage medium thereof, and relates to the technical field of artificial intelligence, and the method comprises the steps: employing a non-intrusive installation mode, arranging a multi-source sensor outside an alternating current asynchronous motor, and collecting the multi-source monitoring data in the operation process of the motor; preprocessing the multi-source monitoring data; based on the electrical and mechanical coupling mechanism of the AC asynchronous motor, constructing a mechanism model reflecting the internal relation between different parameters and the electrical system state of the motor in normal operation and typical fault states; performing fusion processing on the preprocessed multi-source monitoring data by using an information fusion algorithm to obtain a fused sensing data set; a machine learning algorithm is adopted to carry out optimization parameter adjustment on the mechanism model based on the fusion sensing data set, a motor fault diagnosis model is obtained, and the motor fault diagnosis model outputs a motor fault diagnosis result based on multi-source monitoring data obtained in real time; the method has the effect of improving the fault monitoring efficiency of the alternating current asynchronous motor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an alternating asynchronous motor fault diagnosis method and device and a storage medium thereof. BACKGROUND

[0002] The alternating asynchronous motor is widely used as a power equipment in industrial production, and the stability of its running state directly affects the production efficiency and safety in production. It is of great significance to diagnose and warn the faults of the alternating asynchronous motor in time and accurately.

[0003] The existing fault diagnosis technology for the alternating asynchronous motor mainly includes the following means: one is an invasive detection mode, which needs to access the electrical circuit to detect electrical parameters or punch the mechanical structure to install sensors; two is a single parameter diagnosis method, which only relies on vibration or temperature signals to judge faults; three is a traditional state sensing technology, which judges or statistically analyzes faults through a simple threshold. Among these methods, the invasive detection can obtain direct electrical parameters, but it is easy to cause secondary fault risk; the single parameter diagnosis method has a single information, and the diagnosis accuracy is usually lower than 80%; the traditional state sensing technology has a monitoring error of more than 5% for dynamic parameters, which is difficult to meet the early fault warning demand. Therefore, there is a defect that the fault monitoring efficiency of the alternating asynchronous motor is low, and improvement is needed. SUMMARY

[0004] In order to improve the fault monitoring efficiency of the alternating asynchronous motor, the present application provides an alternating asynchronous motor fault diagnosis method, device and storage medium thereof.

[0005] In the first aspect, the application aims to achieve the following technical solutions: An alternating asynchronous motor fault diagnosis method, comprising: A non-invasive installation mode is adopted, and a multi-source sensor is arranged outside the alternating asynchronous motor to collect multi-source monitoring data in the running process of the motor; The multi-source monitoring data is preprocessed; Based on the electrical and mechanical coupling mechanism of the alternating asynchronous motor, a mechanism model reflecting the internal relationship between different parameters and the state of the electrical system of the motor under normal operation and typical fault states is constructed; The preprocessed multi-source monitoring data is fused by using an information fusion algorithm to obtain a fused sensor data set; A machine learning algorithm is used to optimize and parameterize the mechanism model based on the fused sensor data set to obtain a motor fault diagnosis model, and the motor fault diagnosis model outputs a motor fault diagnosis result based on real-time acquired multi-source monitoring data.

[0006] By adopting the technical scheme, the multi-source sensor includes a multi-axis vibration sensor, a temperature sensor, a noise sensor, and a magnetic flux sensor. The non-invasive installation mode is to avoid accessing the electrical system of the alternating current asynchronous motor and damaging the mechanical part of the motor, including avoiding punching installation on the mechanical structure, ensuring safety and comprehensiveness, integrating multi-source sensor data through an information fusion algorithm, eliminating redundancy, and improving data consistency; the mechanism model is a model reflecting the internal relationship between parameters based on the electrical and mechanical coupling mechanism of the alternating current asynchronous motor, the fused sensor data and the mechanism model are combined, a diagnosis model is trained through a machine learning algorithm (such as a support vector machine) to output a fault feature mode that can learn the alternating current asynchronous motor, and a motor fault diagnosis model with high precision early warning is realized to output a motor fault diagnosis result with specific fault information in time when the alternating current asynchronous motor outputs a fault or an abnormal feature, meet the early fault warning demand, and improve the fault monitoring efficiency of the alternating current asynchronous motor.

[0007] In a preferred example of the present application: the motor fault diagnosis model outputs a motor fault diagnosis result based on real-time multi-source monitoring data, including: segmenting and dividing the operating state of the alternating current asynchronous motor based on the multi-source monitoring data to form a plurality of operating state analysis units; taking each operating state analysis unit as a basic unit, performing fault tendency analysis according to a preset first feature parameter set to obtain a corresponding motor fault occurrence probability index, and calculating a motor fault risk level according to the motor fault occurrence probability index; taking each operating state analysis unit as a basic unit, evaluating the motor performance degradation degree according to a preset second feature parameter set, and calculating a motor performance loss index; based on the motor performance loss index and the motor fault risk level corresponding to each operating state analysis unit, generating a motor fault diagnosis result by fusion.

[0008] By adopting the above technical scheme, to realize fine segmentation evaluation of the dynamic running process, the state confusion problem caused by overall evaluation is avoided, the motor operating state is divided into a plurality of operating state analysis units, each operating state analysis unit is taken as a basic unit, fault tendency analysis and performance degradation degree evaluation are performed respectively, the diagnosis result has time traceability and process continuity, not only the fault type is identified, but also the occurrence threat degree is quantified, the performance loss index is calculated, the actual influence of the fault on the operating state of the alternating current asynchronous motor system is reflected, the comprehensive judgment of "fault possibility" and "consequence severity" is realized, and the engineering practicability of the diagnosis decision and the fault diagnosis accuracy are improved.

[0009] In a preferred example of the present application: the multi-source monitoring data includes multi-axis vibration sensing signals, shell temperature signals, magnetic flux signals, noise signals; the operating state of the asynchronous motor is segmented and divided based on the multi-source monitoring data, forming a plurality of operating state analysis units, including: The real-time collected multi-source monitoring data set is denoised and filtered and time-synchronized to obtain a preprocessed standardized data sequence; The standardized data sequence is sliced at equal intervals in a sliding time window manner, and the time window length is determined according to the rated speed of the motor and the typical fault characteristic period; each time window corresponds to an operating state analysis unit; The signals in each operating state analysis unit are feature extracted to form a unit feature vector containing multi-dimensional features.

[0010] By using the above technical solution, data preprocessing is used to eliminate noise interference and time misalignment of multi-source signals in the collection process, improve data processing quality, and set the time window length according to the rated speed of the motor and the typical fault characteristic period, so that it can completely cover the key fault periods such as bearing fault impact, rotor broken bar modulation, etc., improve the ability to capture motor fault features or abnormal features, and form a unit feature vector containing electrical, vibration, thermal and other multi-physical field information.

[0011] In a preferred example of the present application: taking each operating state analysis unit as a basic unit, a fault tendency analysis is performed according to a preset first feature parameter set to obtain a corresponding motor fault occurrence probability index, specifically including: The first feature parameter set includes three magnetic flux period change information, amplitude of bearing fault characteristic frequency in vibration signal envelope spectrum, Park vector modulus fluctuation coefficient, slip fluctuation degree, and temperature rise rate; The first feature parameters are extracted for each operating state analysis unit to form a feature input vector; The feature input vector is input into a classification sub-model of the motor fault diagnosis model, and the occurrence probability of each typical fault category is output, and the maximum probability value in the occurrence probability is selected as the motor fault occurrence probability index.

[0012] By using the above technical solution, the feature parameters in the first feature parameter set have strong relevance with the typical faults of the motor, and can effectively represent the early signs of faults such as stator short circuit, rotor broken bar, and bearing damage; by extracting these key features for each operating state analysis unit and forming an input vector, the fault features are accurately focused. Inputting the feature input vector into the classification sub-model outputs the occurrence probability of each fault type, which not only identifies the most likely fault type, but also reflects the confidence level, to realize efficient mapping from feature extraction to fault tendency diagnosis and improve the sensitivity of fault diagnosis.

[0013] In a preferred embodiment of this application, the step of calculating the motor failure risk level based on the motor failure occurrence probability index includes: Acquire historical operating status data, and based on the historical operating status data, statistically analyze the actual occurrence frequency of different fault types under different operating conditions to construct a motor fault condition failure rate database. For the current operating status analysis unit, query the conditional failure rate of relevant fault types under the corresponding operating conditions, and use it as a dynamic correction factor; The motor failure probability index is weighted and fused with the corresponding conditional failure rate to calculate the comprehensive failure risk score. Based on the preset risk grading threshold, the comprehensive fault risk score is mapped to low risk, medium risk or high risk level, and the corresponding motor fault risk level is generated.

[0014] By adopting the above technical solution, the actual frequency of occurrence of different faults under different loads, temperature, and humidity conditions is quantified as the conditional failure rate, which serves as a dynamic correction factor, effectively mitigating the risk of misjudgment caused by relying solely on current monitoring data. By weighting and integrating the fault occurrence probability index with the conditional failure rate, both the degree of anomaly in the current state and historical fault patterns are considered, thus improving the objectivity and reliability of risk assessment.

[0015] In a preferred embodiment of this application: the step of evaluating the degree of motor performance degradation based on a preset second feature parameter set, using each of the operating state analysis units as a basic unit, and calculating the motor performance loss index includes: The second set of characteristic parameters includes the rate of decrease in motor efficiency, the rate of change in magnetic flux, the growth factor of the RMS value of vibration acceleration, and the rate of temperature rise; The running period of the motor is divided into multiple running state analysis units. All sampling points involved in a single running state analysis unit are determined. Each sampling point in the single running state analysis unit is quantized and assigned a value according to the second feature parameter set to obtain the second feature score. The arithmetic mean of the second feature scores of all sampling points within a single operating status analysis unit is calculated to obtain the mean of performance degradation features. The mean of performance degradation features is then used in conjunction with the weighted comprehensive index method to calculate the motor performance loss index.

[0016] By adopting the above technical solution, the second set of characteristic parameters directly reflects the physical manifestations of motor performance degradation, and can quantify the impact of faults on system operating efficiency, stability, and safety. By assigning feature values ​​to all sampling points within each operating state analysis unit and calculating the arithmetic mean, the mean value of performance degradation characteristics is obtained, effectively suppressing interference from instantaneous fluctuations and improving evaluation stability. The weighted comprehensive index method further integrates multiple parameters, taking into account the differences in importance of each indicator, and the final generated motor performance loss index can comprehensively and objectively reflect the deterioration trend of the motor's health status.

[0017] In a preferred embodiment of this application: the step of fusing and generating motor fault diagnosis results based on the motor performance loss index and motor fault risk level corresponding to each of the operating status analysis units includes: Using each of the aforementioned operating status analysis units as basic units, the corresponding motor fault risk level is used as the row vector input and the corresponding motor performance loss index is used as the column vector input, which are mapped to a preset fault risk loss discrimination matrix. The fault risk loss discrimination matrix presets diagnostic conclusion labels based on the combination of different motor fault risk levels and motor performance loss indices. The severity level of motor faults for each operating status analysis unit is obtained by matching. The severity levels of motor faults include warning, attention, alarm, and shutdown. Based on the motor fault severity level corresponding to all operating status analysis units, a trend chart of the motor's full-cycle operating status and a comprehensive fault diagnosis report are generated as the output of the comprehensive fault diagnosis result of the motor.

[0018] By adopting the above technical solution, the motor fault risk level is input as a row vector and the performance loss index as a column vector into the fault risk loss discrimination matrix, achieving cross-discrimination of both risk and loss dimensions, thus overcoming the limitations of single-indicator decision-making. The fault risk loss discrimination matrix presets diagnostic conclusion labels according to different combinations, giving the diagnostic results clear handling guidance. By matching, the fault severity level of each operating status analysis unit is obtained, realizing the hierarchical and standardized nature of the diagnostic results.

[0019] Secondly, the objective of this invention is achieved through the following technical solution: An AC asynchronous motor fault diagnosis device is used to perform an AC asynchronous motor fault diagnosis method as described above. The device includes: The multi-source sensing module is used to non-intrusively install multi-source sensors on the outside of an AC asynchronous motor to collect multi-source monitoring data during motor operation. The data preprocessing module is used to preprocess the multi-source monitoring data; The mechanism modeling module is used to construct a mechanism model based on the electrical and mechanical coupling mechanism of AC asynchronous motors, reflecting the intrinsic relationship between different parameters of the motor and the state of the electrical system under normal operation and typical fault conditions. The information fusion module is used to fuse pre-processed multi-source monitoring data using information fusion algorithms to obtain a fused sensor dataset. The model training and diagnosis module is used to optimize and tune the mechanism model based on the fused sensor dataset using machine learning algorithms to obtain a motor fault diagnosis model. The motor fault diagnosis model outputs motor fault diagnosis results based on real-time acquired multi-source monitoring data.

[0020] Thirdly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned method for diagnosing faults in an AC asynchronous motor.

[0021] In summary, this application includes at least one of the following beneficial technical effects: 1. The use of non-invasive multi-source sensors on the exterior of the motor avoids modifications to the motor itself and requires downtime for disassembly and reassembly, significantly improving the deployment convenience and applicability of the monitoring system. By constructing a mechanistic model based on the coupling mechanism of electrical and mechanical components, the physical essence of motor faults is introduced, enhancing the interpretability of the diagnostic model. Combining information fusion and machine learning to optimize and tune the mechanistic model achieves a fusion modeling approach of "mechanism-guided + data-driven," overcoming the poor generalization ability of purely data-driven models and improving the adaptability of traditional mechanistic models to complex operating conditions. 2. By dividing the motor's operating state into multiple operating state analysis units, a refined segmented evaluation of the dynamic operation process is achieved, avoiding the state confusion problem caused by overall evaluation. Based on each unit, fault tendency analysis and performance degradation assessment are performed separately, making the diagnostic results time-traceable and process continuous. By introducing a dual criterion of fault occurrence probability index and risk level, not only are fault types identified, but their threat level is also quantified; at the same time, a performance loss index is calculated to reflect the actual impact of faults on system operation. Attached Figure Description

[0022] Figure 1 This is a flowchart of an AC asynchronous motor fault diagnosis method according to an embodiment of this application; Figure 2 This is a flowchart of step S5 in an AC asynchronous motor fault diagnosis method according to an embodiment of this application. Detailed Implementation

[0023] The present application will be further described in detail below with reference to the accompanying drawings.

[0024] In one embodiment, such as Figure 1 As shown, this application discloses a fault diagnosis method for an AC asynchronous motor, which specifically includes the following steps: S1: Using a non-intrusive installation method, multi-source sensors are placed outside the AC asynchronous motor to collect multi-source monitoring data during motor operation.

[0025] In this embodiment, the AC asynchronous motor is an induction-type three-phase AC asynchronous motor, suitable for squirrel-cage three-phase AC asynchronous motors. Non-invasive installation refers to deploying sensors externally via clamping / adsorption without damaging the original motor structure. The multi-source sensor includes an integrated sensor array, using a non-invasive installation method, with multi-axis vibration acceleration sensors, temperature sensors, noise sensors, and magnetic flux sensors arranged externally on the AC asynchronous motor. The multi-axis vibration sensor collects vibration characteristics during motor operation, including vibration acceleration and velocity in different directions; the temperature sensor monitors changes in the ambient temperature around the motor; the noise sensor captures changes in ambient noise generated during motor operation; and the Hall magnetic flux sensor acquires changes in the magnetic flux around the motor. Based on the Hall magnetic flux, speed and winding monitoring are performed, and current strength detection is also possible.

[0026] Specifically, the Hall magnetic flux sensor acquires information on the periodic changes of the three magnetic flux cycles. Based on this information, the rotational speed and current changes are calculated. The specific manifestation is as follows: Unsaturated operating condition (light load / no load): The magnetic flux is linearly related to the excitation current. The total stator current consists of the excitation component and the small load component, and the value is relatively small.

[0027] Saturation condition (rated load): The magnetic flux is close to the rated value, the excitation current increases slightly due to magnetic circuit saturation, and the stator current is mainly dominated by the load component, which is close to the rated current.

[0028] Starting condition: The magnetic flux decreases, the rotor current is extremely high, and the stator starting current is 5-6 times the rated value. The core is to counteract the strong demagnetizing magnetomotive force of the rotor.

[0029] Based on the electrical principle of a three-phase AC asynchronous motor, a corresponding mechanism model is established, which can calculate the theoretical motor speed, mechanical characteristic curves, etc. The mechanism model includes two parts: the mechanism of the three-phase AC asynchronous motor and the mechanism of the bearing. It is equivalent to using historical data of the four sensing parameters of the aforementioned multi-axis vibration acceleration sensor, temperature sensor, noise sensor and magnetic flux sensor, combined with the above mechanism, the mechanism model can learn the reasonable data range of the three-phase AC asynchronous motor under normal conditions. Based on the reasonable range of the four sensing parameters, if the corresponding normal range is exceeded, a fault warning will be triggered.

[0030] Specifically, the sampling frequencies are set as follows: 20kHz for vibration signal, 1Hz for temperature signal, and 1kHz for noise signal.

[0031] S2: Preprocess the multi-source monitoring data.

[0032] In this embodiment, data preprocessing includes noise reduction, time synchronization, and normalization. Time synchronization is based on the rising edge of the rotation speed signal, interpolated to the timestamps of other signals. During normalization, the current signal is divided by the rated current value, and the vibration signal is divided by the sensor range.

[0033] S3: Based on the electrical and mechanical coupling mechanism of AC asynchronous motors, a mechanism model is constructed to reflect the intrinsic relationship between different parameters of the motor and the state of the electrical system under normal operation and typical fault conditions.

[0034] In this embodiment, a mechanistic model of the motor is established based on its working principle and structural characteristics. This mechanistic model reflects the intrinsic relationship between parameters such as vibration, temperature, noise, and magnetic flux of the motor under normal operation and different fault conditions, and factors such as motor speed, load, and electrical system status. The electrical-mechanical coupling mechanism refers to the physical correlation between changes in the motor's electromagnetic field and mechanical vibration and temperature rise.

[0035] Specifically, the mechanism model includes an overall mechanism model of the motor and a mechanism model of the motor bearings. Information fusion algorithms are used to fuse pre-processed multi-sensor data, and the fused dataset is combined with the motor mechanism model. A big data machine learning algorithm is then used to train the model. The machine learning model is trained using a large amount of historical fault data and normal operation data, enabling it to learn the characteristic patterns of the motor under different fault states. Then, the real-time collected and processed sensor data is input into the trained machine learning-based mechanism model, allowing it to judge the motor's operating state based on the learned characteristic patterns, and to provide early warnings for various faults such as loose installation, bearing abnormalities, misalignment, imbalance, three-phase imbalance, phase loss, and mechanical friction. When the trained mechanism model determines that a fault exists in the motor, it promptly outputs information such as the fault type, fault location, and fault severity.

[0036] For example, the mechanism model of the motor bearing uses the stator inter-turn short-circuit current equation when monitoring stator faults, and the specific calculation is as follows: in, This is the short-circuit coefficient, used to quantify the severity of a short circuit. A larger value indicates a larger short-circuit circulating current and more significant fault characteristics. The value ranges from 0.1 to 0.5. This is the rated current of the motor, which serves as a reference value for the fault current, reflecting the current amplitude under normal operating conditions. is the decay time constant, used to control the decay rate of the short-circuit current. The slower the current decays, the longer the fault duration. The typical value range is 0.01~0.1 seconds. f is the power supply fundamental frequency, which determines the current oscillation frequency. The value is 50Hz or 60Hz.

[0037] When monitoring bearing faults, the mechanism model of motor bearings uses vibration characteristic frequencies for calculation. in, The number of balls in an AC asynchronous motor determines the number of harmonics at the fault frequency. As the number of balls increases, the peak value of the spectrum increases. Small and medium-sized bearings typically have 8 to 12 balls. For shaft rotation frequency, ; This refers to the ball diameter, which is positively correlated with the bearing size. The bearing pitch diameter refers to the diameter of the circle in which the balls are distributed. The contact angle affects load distribution. The larger the value, the higher the axial load capacity.

[0038] When performing thermal coupling monitoring, the mechanism model uses the following winding temperature rise differential equation: Where I is the operating current, which is the source of Joule heating. The main temperature rise is represented by R, which is the winding resistance and the core parameter of heat generation power, which increases with increasing temperature; h is the heat dissipation coefficient, which characterizes the heat dissipation capacity. For natural cooling, h = 5~25, and for forced air cooling, h = 50~100, with units of W / (m²·K); A is the heat dissipation area, which is positively correlated with the surface area of ​​the motor. is the ambient temperature, used as the baseline value for temperature rise; m is the winding mass, which is the main body of heat capacity; c is the specific heat capacity, which is an inherent property of the material, with a value range of 380~400.

[0039] S4: Use information fusion algorithms to fuse the preprocessed multi-source monitoring data to obtain a fused sensor dataset.

[0040] In this embodiment, the information fusion algorithm is a federated Kalman filter.

[0041] Design a local filter that includes a current subsystem and an oscillation subsystem, where the state variables of the current subsystem are [ , The three-phase current observations are as follows: , , ] is the original signal acquired by the sensor, and the triaxial acceleration state variables of the vibration subsystem are [ , , The observed value is [acceleration amplitude]. After demodulation of the acceleration amplitude correlation envelope, the fault characteristic frequency amplitude (such as BPFO) is extracted, and the sensitivity is 20dB higher than that of the time domain signal. The branch shaft current is the current component after Park transformation, used to control the excitation magnetic field strength of the motor and reflect the stator magnetic field orientation accuracy. The quadrature-axis current determines the motor's output torque; abnormal fluctuations indicate a broken rotor bar or a sudden change in load. = A global fusion method based on the master filter fusion rule is adopted, wherein the master filter fusion rule is as follows: in, The covariance is estimated locally to characterize the estimation accuracy of the subsystem; This is a local state estimate. It represents the subsystem-independent filtering results. The information matrix and the inverse covariance matrix are used for weighted fusion. It is the globally optimal estimate, and it represents the fused state variables. During parameter fusion, the weights of the current subsystem and the vibration subsystem can be set to 0.7 and 0.3, respectively.

[0042] Specifically, the feature vector output after fusing the sensor dataset is in the format of: "[current d-axis component, current q-axis component, vibration envelope energy, temperature gradient]".

[0043] S5: The mechanism model is optimized and its parameters are tuned using machine learning algorithms based on the fusion sensor dataset to obtain the motor fault diagnosis model. The motor fault diagnosis model outputs the motor fault diagnosis results based on the real-time acquired multi-source monitoring data.

[0044] In this embodiment, parameter tuning optimization refers to adjusting the weight parameters of the mechanistic model through Bayesian optimization.

[0045] Specifically, the training data for the motor fault diagnosis model includes positive samples of typical fault data (e.g., bearing wear, rotor bar breakage, stator short circuit) and negative samples of normal operation data. The optimization objective function for the model (using least squares estimation with L2 regularization) is as follows: in, =[Electromagnetic parameters, mechanical damping coefficient, thermal capacity coefficient]; The optimal parameter estimate is the final solution for parameter identification, achieved by fusing multi-source observation data. and prior models In reverse, we can deduce the combination of electromagnetic, mechanical, and thermal parameters that best fits the actual system. For the actual observed outputs of K samples, such as vibration acceleration, terminal voltage, and temperature rise rate; The input features for the Kth sample are, for example, current, rotational speed, and temperature; N is the number of observed samples.

[0046] In one embodiment, such as Figure 2 As shown, the motor fault diagnosis model outputs motor fault diagnosis results based on real-time acquired multi-source monitoring data, including: S51: Based on multi-source monitoring data, the operating status of the AC asynchronous motor is segmented to form several operating status analysis units.

[0047] In this embodiment, to solve the problem of mixed continuous operating condition data and avoid cross-cycle feature interference, the real-time acquired multi-source monitoring dataset is divided into multiple operating status analysis units. Each operating status analysis unit refers to an independent analysis period segmented by time window slicing, and each unit contains complete electrical and mechanical coupling features.

[0048] Specifically, the window length of the time window = 60 / (motor rated speed × typical fault characteristic frequency); the unit feature extraction includes time domain features, frequency domain features and time-frequency features.

[0049] In this embodiment, step S51 includes: S511: Perform noise reduction filtering and time synchronization processing on the real-time acquired multi-source monitoring dataset to obtain a preprocessed standardized data sequence.

[0050] In this embodiment, a 50Hz notch filter is used to eliminate power frequency interference in the current signal; wavelet threshold denoising uses a db4 wavelet basis with 5-level decomposition. The vibration signal is bandpass filtered to retain the bearing or gear fault characteristic frequency band and then subjected to adaptive white noise filtering.

[0051] S512: The standardized data sequence is sliced ​​at equal intervals using a sliding time window method. The length of the time window is determined based on the rated speed of the motor and the typical fault characteristic cycle. Each time window corresponds to an operating status analysis unit.

[0052] Specifically, the length of the time window Where k takes values ​​of 5 to 10 cycles, bearing failure = For example, a bearing failure in a 1500rpm motor. =82Hz. During a rotor bar breakage fault, =2 × slip ratio × Each operational status analysis unit contains a triplet of timestamp, speed range, and load rate.

[0053] S513: Extract features from the signals within each operating status analysis unit to form a unit feature vector containing multi-dimensional features.

[0054] Specifically, the multi-dimensional features include time-domain features, frequency-domain features, and time-frequency features; the time-domain features include kurtosis and waveform factors, where the kurtosis value... When K is greater than 3, an impact fault is indicated; N is the number of sampling points, which is the total amount of data within an analysis window; These are vibration acceleration sample values; The signal mean is given; the vibration energy intensity is calculated as follows: The vibration energy intensity is taken as the square root of the average of the squared values ​​of the signal; the larger the value, the greater the cumulative mechanical damage. Frequency domain feature extraction uses envelope spectrum analysis, which will not be elaborated here.

[0055] Time-frequency feature extraction uses wavelet packet energy entropy E calculation method: in, The j-th inherent normalized energy percentage reflects the degree of energy dominance in a specific frequency band and is used to locate faulty frequency bands; Subband energy refers to the signal energy of the j-th subband in wavelet packet decomposition; m is the total number of subbands, determined by the number of tree levels in the self-broadcast decomposition, for example, m=16 for a resolution of 3kHz. When the motor is in a healthy state, the spectral energy is concentrated in... , Approximately equal to 0.8; E value ranges from 0.5 to 1.0. When the motor is in a fault state due to bearing spalling, the energy of the spectral characteristics diffuses to ±n , The distribution is uniform, and the E value ranges from 2.5 to 3.5.

[0056] Furthermore, when the E value is less than 1.2, it is in a healthy state; when the E value is greater than 1.5 but less than 2.0, it is in an early initiation corrosion state, triggering a warning action; when the E value is greater than 2.5, it is in a severe peeling fault state, requiring shutdown and maintenance.

[0057] S52: Using each operating status analysis unit as the basic unit, perform fault tendency analysis based on the preset first feature parameter set to obtain the corresponding motor fault occurrence probability index, and calculate the motor fault risk level based on the motor fault occurrence probability index.

[0058] In this embodiment, the first set of characteristic parameters includes the periodic variation information of the three magnetic fluxes, the amplitude of the bearing fault characteristic frequency in the vibration signal envelope spectrum, the fluctuation coefficient of the Park vector magnitude, the degree of slip fluctuation, and the temperature rise rate; wherein the periodic variation information of the three magnetic fluxes is the three-phase magnetic flux ( , , The variation characteristics within a power cycle include periodic fluctuations in amplitude, phase, and harmonic components. This parameter quantifies magnetic field asymmetry and is directly related to magnetic field distortion caused by electromagnetic faults such as rotor bar breakage and stator inter-turn short circuits. In a healthy state, the three-phase magnetic flux amplitudes are balanced, the phase difference is 120°, and the harmonic components are below 5%. A rotor bar breakage fault leads to magnetic field modulation, causing magnetic flux amplitude fluctuations (>10%) and subharmonic growth; a stator short circuit causes phase shift and a prominent third harmonic (3f0 amplitude increase >20%). The amplitude of the bearing fault characteristic frequency in the vibration signal envelope spectrum is correlated with the bearing wear degree for quantification; the Park vector magnitude fluctuation coefficient... The ratio of the mean to the standard deviation of the Park vector magnitude is used to correlate stator winding short-circuit sensitivity to faults. When the fluctuation coefficient of the Park vector magnitude is greater than 0.25, the inter-turn short-circuit sensitivity index is triggered. The motor fault risk level is a dynamic probability correction result based on historical failure rates.

[0059] The slip fluctuation Δs reflects speed instability caused by sudden load changes or rotor imbalance. A slip fluctuation greater than 0.1 triggers a rotor dynamic imbalance warning. The formula for calculating the slip fluctuation Δs is: , ,in Instantaneous slip; This is the actual electrical frequency of the rotor; This is the base frequency of the power supply.

[0060] rate of temperature rise The formula is used to quantify sudden temperature rises caused by heat dissipation system failure or overload. If 10 seconds is taken as the quantization time interval for the temperature rise rate, then... If the value is 0.3~0.5℃ per second, it indicates a risk warning; reduce the load by 15%. If the temperature exceeds 0.8℃ per second, Pan Wei is classified as a high-risk individual, triggering and executing an immediate shutdown protection command.

[0061] Specifically, the bearing fault envelope spectrum amplitude is The impact energy intensity caused by the spalling of the bearing outer ring is characterized by vibration signal extraction, Hilbert transform, envelope demodulation, and FFT analysis. The amplitude is set, with a Hilbert transform window function of 6 and an envelope analysis bandwidth of 1-5kHz, covering bearing fault characteristics.

[0062] The envelope spectrum analysis process is as follows: the vibration signal is subjected to Hilbert transform, envelope demodulation is performed, and FFT spectrum analysis is conducted to extract the bearing fault frequency amplitude.

[0063] Specifically, step S52 includes: S521: Extract the first feature parameter for each operating status analysis unit to form a feature input vector.

[0064] In this embodiment, all parameter timestamps are aligned based on the rising edge of the rotational speed signal, and the time synchronization compensation formula is as follows: ,in The timestamp after compensation serves as a unified time reference for multiple source signals relative to subsequent timestamps. This is the original timestamp. It is the system clock record when the sensor acquired data; The phase difference refers to the phase offset between the target current signal and the reference signal; the characteristic normalization formula is: , where x is the original feature value; These are normalized eigenvalues, representing the degree of deviation from the health baseline; , Taken from health status, they are the health baseline mean and health baseline standard deviation, respectively; The mean feature value was obtained by running the new machine under no-load for 10 hours. It characterizes the degree of dispersion of eigenvalues ​​under the same working conditions.

[0065] Specifically, the feature input vector format is a 5-dimensional vector [Rh, , ,Δs, ] S522: Input the feature input vector into the classification sub-model of the motor fault diagnosis model, output the occurrence probability of each typical fault category, and select the maximum probability value among the occurrence probabilities as the motor fault occurrence probability index.

[0066] In this embodiment, the classification sub-model uses a Softmax regression model, employing a linear classifier with probability normalization. Different fault categories are associated with corresponding category indices, including: healthy state, rotor bar breakage, bearing outer ring spalling, stator inter-turn short circuit, and heat dissipation system failure, corresponding to type indices of 0, 1, 2, 3, and 4, respectively. The feature sensitivity parameters corresponding to healthy state, rotor bar breakage, bearing outer ring spalling, stator inter-turn short circuit, and heat dissipation system failure are: all parameters within the threshold, Rh and Δs, ... , and .

[0067] Specifically, the output formula for the motor failure probability index is: Where W is a weight matrix with a data dimension of K×d, K is the number of fault categories (taken as 5 in this embodiment), X is the input feature vector, and is the first feature parameter set; row vector W K表示The "feature-sensitive pattern" for the k-th type of fault; b is the bias vector with dimension K×1, used to refine the baseline probability of each type of fault; the maximum value is selected as the fault occurrence probability exponent. The linear transformation is Z=WX+b, where Z is the normalized score, used to quantify the matching degree between features and fault categories. Then, Softmax probability normalization is performed to output the probability distribution. During the optimization training of the classification model, the loss function adopts cross-entropy loss + L2 regularization. The optimizer is stochastic gradient descent (SGD) with a learning rate of 0.01 (exponential decay).

[0068] For example, the fault diagnosis record of a water pump motor is shown in the table below: Furthermore, the motor failure risk level is calculated based on the motor failure probability index, including: S523: Obtain historical operating status data, and based on the historical operating status data, statistically analyze the actual occurrence frequency of different fault types under different operating conditions to construct a motor fault condition failure rate database.

[0069] In this embodiment, the motor failure rate database is a quantitative mapping relationship library between operating conditions and failures. By statistically analyzing historical data on the occurrence patterns of various failures under different combinations of operating conditions, a dynamic failure rate benchmark is generated.

[0070] Specifically, attribution analysis is performed on historical failure times, and the ratio of the occurrence of each type of failure to the total operating time under specific operating condition combinations is statistically analyzed to determine the corresponding failure rate per unit time, so as to construct a motor failure condition failure rate database.

[0071] In this embodiment, the formula for calculating the failure rate per unit time is: Where K represents the fault type. This is used to encode operating conditions, such as "70_1500_40 represents 70% load + 1500 rpm + 40℃". Specifically, the database structure of the constructed motor fault condition failure rate database is shown in the table below: S524: For the current operating status analysis unit, query the conditional failure rate of relevant fault types under the corresponding operating conditions, and use it as a dynamic correction factor.

[0072] In this embodiment, the dynamic correction factor is used to obtain the failure rate correction coefficient under the current state based on the real-time operating condition matching historical database, thereby quantifying the impact of environmental factors on risk.

[0073] Specifically, given the current device state s = [load rate, speed, temperature], search the historical database for the most similar historical record (calculated using the weighted Euclidean distance formula). After finding the most recent record, assume that the data contains the... The status of the record is The corresponding preset weight vectors are 0.6, 0.3, and 0.1. The weighted Euclidean distance is... Find the nearest neighbor record. The corresponding condition failure rate is The final output dynamic correction factor , It is a global reference value, from the historical records. The pre-designed nominal value. Under healthy conditions, the value is 1.0, and under heavy load and high temperature conditions, the value can be 2.5.

[0074] S525: The probability index of motor failure is weighted and fused with the corresponding conditional failure rate to calculate the comprehensive failure risk score.

[0075] For example, the severity weights associated with each fault category are... The weighting coefficients for rotor bar breakage, bearing spalling, stator short circuit, and heat dissipation failure are 0, 1.2, 1.5, 2.0, and 1.0, respectively.

[0076] Specifically, the weighted fusion calculation formula used for the comprehensive fault risk score is as follows: ,in This represents the probability of a fault occurring as output by the S522, with a value ranging from 0 to 1. The dynamic correction factor output in step S524; The above refers to the severity weights of the faults.

[0077] S526: Based on the preset risk classification threshold, the comprehensive fault risk score is mapped to low risk, medium risk or high risk level, and the corresponding motor fault risk level is generated.

[0078] In this embodiment, a comprehensive fault risk score of less than 1.0 indicates low risk, a score of 1.0 to 2.0 indicates medium risk, and a score greater than 2.0 indicates high risk.

[0079] Specifically, the risk classification threshold can be set based on dynamic changes in maintenance costs. The changed threshold = threshold baseline value × (average maintenance cost / downtime loss) S53: Using each operating status analysis unit as the basic unit, evaluate the degree of motor performance degradation based on the preset second characteristic parameter set, and calculate the motor performance loss index.

[0080] In this embodiment, the second set of characteristic parameters includes the rate of decrease in motor efficiency, the rate of change of magnetic flux, the growth factor of the RMS value of vibration acceleration, and the rate of temperature rise; the rate of decrease in motor efficiency... The calculation is as follows ,in This represents the current operating rate of the motor. The initial efficiency of the motor; the calculation method is as follows: in, This refers to the output power. Input power. Motor efficiency reduction rate. The motor's energy conversion efficiency has decreased due to aging. When it exceeds 5%, a mild degradation warning is triggered. When the value exceeds 15%, an alarm indicating the need for maintenance is triggered. The rate of change of magnetic flux refers to the rate of change of the air gap magnetic flux in the motor per unit time. It is used to characterize the stability of the electromagnetic field and can effectively identify magnetic field distortion caused by stator winding short circuits, rotor bar breakage, or unbalanced power supply voltage. Specifically, the air gap magnetic flux density is measured using a non-invasive magnetic flux sensor with a sampling frequency of not less than 1kHz. The calculation formula is as follows: ,in This represents the magnetic flux value at the current moment. The sampling interval is typically 1 second. The assignment rule is as follows: based on the standard deviation of magnetic flux fluctuation under healthy conditions, a value of 5 is assigned for every instance of the rate of change of magnetic flux exceeding twice the standard deviation of magnetic flux fluctuation. When the value exceeds 10, an abnormal magnetic field is indicated. (Vibration acceleration RMS value growth factor) , The core indicator used to quantify the degree of mechanical deterioration is that the larger the value, the more severe the failure. To ensure that the vibration energy intensity monitored in real time reflects the instantaneous state of mechanical components, acceleration sensors are installed at the same position and in the same direction during measurement. The root mean square value of the reference vibration is the benchmark for vibration intensity under healthy conditions. The measurement method is to run the device at its rated speed under no-load conditions and collect vibration data for at least 10 cycles and calculate the average value (sampling frequency ≥ 5 kHz). When the value is between 1.0 and 1.5, the machine is in a healthy mechanical state, and the vibration energy does not change significantly. When the value is between 1.5 and 2.0, the machine is in an early stage of deterioration and may experience minor bearing wear or rotor imbalance. It is necessary to increase the frequency of motor testing. When the value is 2.0~3.0, the machine is in a moderate mechanical failure state, and raceway peeling / gear tooth breakage may occur. In this case, the machine needs to be shut down for maintenance within two weeks. When the temperature exceeds 3.0, the machine is in a severely failed mechanical state, and bearing breakage or rotor rubbing failure may occur, requiring immediate shutdown and component replacement. The temperature rise rate is the rate of temperature change per unit time, i.e., the first derivative of the temperature curve.

[0081] The value acquisition process is as follows: Start the AC asynchronous motor under no-load, accelerate to the rated speed and run stably for 5 minutes, collect the vibration signals of the XTZ three axes, calculate the RMS value of each axis, and take the maximum value as the value. .

[0082] Specifically, step S53 includes: S531: Divide the motor's operating period into multiple operating state analysis units, determine all sampling points involved in a single operating state analysis unit, and quantize and assign values ​​to each sampling point within a single operating state analysis unit according to the second feature parameter set to obtain the second feature score.

[0083] In this embodiment, the motor efficiency reduction rate Quantitative score 10× Each 1% efficiency loss is scored out of 10 points; a score >50 triggers an alert; the quantitative score for the increase factor of vibration acceleration RMS value. 20× ( -1), 2 points for every 0.1x increase, warning for >10 points; power factor change rate Quantitative score 12× Each change of 0.05 is scored as 6 points, and a score greater than 9.6 points triggers a warning; the quantitative score for the rate of temperature rise is also calculated. 15× Each 0.1℃ / s is counted as 1.5 minutes, and >7.5 minutes triggers air cooling.

[0084] S532: Calculate the arithmetic mean of the second feature scores of all sampling points within a single operating status analysis unit to obtain the mean of performance degradation features. Calculate the motor performance loss index using the mean of performance degradation features and the weighted comprehensive index method.

[0085] In this embodiment, the formula for calculating the mean of performance degradation characteristics is: Where N1 takes the value 5.

[0086] Specifically, when calculating the motor performance loss index using the weighted composite index method, the weighting coefficients for the motor efficiency decline rate, the increase factor of the vibration acceleration RMS value, the temperature rise rate, and the magnetic flux change rate are 0.30, 0.25, 0.30, and 0.15, respectively.

[0087] The formula for calculating the motor performance loss index is as follows: in, is the weighting coefficient, and k is the variable index of the second feature score corresponding to the four second feature parameters.

[0088] S54: Based on the motor performance loss index and motor fault risk level corresponding to each operating status analysis unit, the motor fault diagnosis results are generated by fusion.

[0089] Specifically, step S54 includes: S541: Taking each operating status analysis unit as the basic unit, the corresponding motor fault risk level is used as the row vector input and the corresponding motor performance loss index is used as the column vector input, which are mapped to the preset fault risk loss discrimination matrix. The fault risk loss discrimination matrix presets diagnostic conclusion labels according to the combination of different motor fault risk levels and motor performance loss indices.

[0090] In this embodiment, the row vector inputs the risk level, including low risk, medium risk, and high risk; the column vector inputs the motor performance loss index (PLI), 0-2, 2-4, 4-6, and greater than 6, forming the matrix cells of the fault risk loss discrimination matrix. Diagnostic labels are also associated with this matrix.

[0091] For example, the diagnostic label configuration of the fault risk loss discrimination matrix is ​​as follows: S542: Obtain the motor fault severity level corresponding to each operating status analysis unit through matching. The motor fault severity level includes warning, attention, alarm and shutdown.

[0092] For example, the matching and response actions for motor fault severity levels are shown in the table below: S543: Based on the motor fault severity level corresponding to all operating status analysis units, generate a trend chart of the motor's full-cycle operating status and a comprehensive fault diagnosis report, which will be output as the comprehensive fault diagnosis result of the motor.

[0093] In this embodiment, the horizontal axis of the motor's full-cycle operating status trend chart represents timestamps, which are timestamps based on the operating status analysis sequence; the primary coordinate of the vertical axis is the PLI index, and the secondary coordinate is the risk level. The comprehensive fault diagnosis report includes a pie chart of historical severity level distribution, analysis of the top 3 fault types, maintenance recommendations, equipment ID, and original data index, among other information.

[0094] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0095] In one embodiment, an AC asynchronous motor fault diagnosis device is provided, which corresponds to an AC asynchronous motor fault diagnosis method in the above embodiment.

[0096] An AC asynchronous motor fault diagnosis device includes a multi-source sensing module, a data preprocessing module, a mechanism modeling module, an information fusion module, and a model training and diagnosis module. Detailed descriptions of each functional module are as follows: The multi-source sensing module is used to non-intrusively install multi-source sensors on the outside of an AC asynchronous motor to collect multi-source monitoring data during motor operation. The data preprocessing module is used to preprocess multi-source monitoring data; The mechanism modeling module is used to construct a mechanism model based on the electrical and mechanical coupling mechanism of AC asynchronous motors, reflecting the intrinsic relationship between different parameters of the motor and the state of the electrical system under normal operation and typical fault conditions. The information fusion module is used to fuse pre-processed multi-source monitoring data using information fusion algorithms to obtain a fused sensor dataset. The model training and diagnosis module is used to optimize and tune the mechanism model based on the fusion sensor dataset using machine learning algorithms to obtain the motor fault diagnosis model. The motor fault diagnosis model outputs the motor fault diagnosis results based on the real-time acquired multi-source monitoring data.

[0097] Optionally, the model training and diagnostic module includes: The operation status segmentation unit is used to divide the continuous operation process of the motor into segments based on real-time multi-source monitoring data and to form multiple operation status analysis units using a sliding time window method. The fault tendency analysis unit is used to extract the feature vector corresponding to the preset first feature parameter set for each operating state analysis unit, and input it into the trained classification sub-model to output the motor fault occurrence probability index corresponding to the unit. The risk level assessment unit is used to combine the fault occurrence probability index with the pre-built motor fault condition failure rate database, calculate the comprehensive fault risk score through weighted fusion, and map it to three levels of motor fault risk: low, medium and high, according to a preset threshold. The performance degradation assessment unit is used to extract the quantified values ​​of the preset second feature parameter set for each operating status analysis unit, calculate the mean value of the performance degradation features, and generate the corresponding motor performance loss index through the weighted comprehensive index method. The diagnostic result generation unit is used to input the motor fault risk level and motor performance loss index corresponding to each operating status analysis unit into the preset fault risk-loss discrimination matrix, match and output the fault severity level of the unit, and generate a full-cycle operating trend chart and a comprehensive fault diagnosis report based on the severity levels of all units, as the final diagnostic result output.

[0098] For specific limitations regarding an AC asynchronous motor fault diagnosis device, please refer to the limitations of an AC asynchronous motor fault diagnosis method mentioned above, which will not be repeated here. Each module in the aforementioned AC asynchronous motor fault diagnosis device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in the processor of a computer device in hardware form or independent of it, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0099] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of an AC asynchronous motor fault diagnosis method.

[0100] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.

[0101] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0102] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for diagnosing faults in an AC asynchronous motor, characterized in that, include: A non-intrusive installation method is used to place multi-source sensors on the outside of the AC asynchronous motor to collect multi-source monitoring data during motor operation; The multi-source monitoring data is preprocessed; Based on the electrical and mechanical coupling mechanism of AC asynchronous motors, a mechanism model is constructed to reflect the intrinsic relationship between different parameters of the motor and the state of the electrical system under normal operation and typical fault conditions. The preprocessed multi-source monitoring data are fused using an information fusion algorithm to obtain a fused sensor dataset. The mechanism model is optimized and its parameters are tuned using a machine learning algorithm based on the fused sensor dataset to obtain a motor fault diagnosis model. The motor fault diagnosis model outputs motor fault diagnosis results based on real-time acquired multi-source monitoring data.

2. The method for diagnosing faults in an AC asynchronous motor according to claim 1, characterized in that, The motor fault diagnosis model outputs motor fault diagnosis results based on real-time acquired multi-source monitoring data, including: Based on the multi-source monitoring data, the operating status of the AC asynchronous motor is segmented to form several operating status analysis units. Using each of the aforementioned operating status analysis units as a basic unit, a fault tendency analysis is performed based on a preset first feature parameter set to obtain the corresponding motor fault occurrence probability index, and the motor fault risk level is calculated based on the motor fault occurrence probability index. Using each of the aforementioned operating status analysis units as a basic unit, the degree of motor performance degradation is evaluated based on a preset second feature parameter set, and the motor performance loss index is calculated. Based on the motor performance loss index and motor fault risk level corresponding to each of the aforementioned operating status analysis units, a motor fault diagnosis result is generated by fusion.

3. The method for diagnosing faults in an AC asynchronous motor according to claim 2, characterized in that, The multi-source monitoring data includes multi-axis vibration sensing signals, shell temperature signals, magnetic flux signals, and noise signals; Based on the multi-source monitoring data, the operating status of the AC asynchronous motor is segmented to form several operating status analysis units, including: The real-time multi-source monitoring dataset is subjected to noise reduction filtering and time synchronization processing to obtain a preprocessed standardized data sequence. The standardized data sequence is sliced ​​at equal intervals using a sliding time window method. The length of the time window is determined based on the rated speed of the motor and the typical fault characteristic cycle. Each time window corresponds to an operating status analysis unit. Feature extraction is performed on the signals within each operational status analysis unit to form a unit feature vector containing multi-dimensional features.

4. The method for diagnosing faults in an AC asynchronous motor according to claim 2, characterized in that, The step of using each of the operating state analysis units as basic units, and performing fault tendency analysis based on a preset first feature parameter set to obtain the corresponding motor fault occurrence probability index, specifically includes: The first set of characteristic parameters includes information on the periodic variation of the three magnetic fluxes, the amplitude of the bearing fault characteristic frequency in the vibration signal envelope spectrum, the fluctuation coefficient of the Park vector magnitude, the degree of slip fluctuation, and the rate of temperature rise. Extract the first feature parameter from each operating status analysis unit to form a feature input vector; The feature input vector is input into the classification sub-model of the motor fault diagnosis model, and the occurrence probability of each typical fault category is output. The maximum probability value among the occurrence probabilities is selected as the motor fault occurrence probability index.

5. The method for diagnosing faults in an AC asynchronous motor according to claim 4, characterized in that, The step of calculating the motor failure risk level based on the motor failure probability index includes: Acquire historical operating status data, and based on the historical operating status data, statistically analyze the actual occurrence frequency of different fault types under different operating conditions to construct a motor fault condition failure rate database. For the current operating status analysis unit, query the conditional failure rate of relevant fault types under the corresponding operating conditions, and use it as a dynamic correction factor; The motor failure probability index is weighted and fused with the corresponding conditional failure rate to calculate the comprehensive failure risk score. Based on the preset risk grading threshold, the comprehensive fault risk score is mapped to low risk, medium risk or high risk level, and the corresponding motor fault risk level is generated.

6. The method for diagnosing faults in an AC asynchronous motor according to claim 2, characterized in that, The process of using each of the aforementioned operating status analysis units as basic units, evaluating the degree of motor performance degradation based on a preset second feature parameter set, and calculating the motor performance loss index includes: The second set of characteristic parameters includes the rate of decrease in motor efficiency, the rate of change in magnetic flux, the growth factor of the RMS value of vibration acceleration, and the rate of temperature rise; The running period of the motor is divided into multiple running state analysis units. All sampling points involved in a single running state analysis unit are determined. Each sampling point in the single running state analysis unit is quantized and assigned a value according to the second feature parameter set to obtain the second feature score. The arithmetic mean of the second feature scores of all sampling points within a single operating status analysis unit is calculated to obtain the mean of performance degradation features. The mean of performance degradation features is then used in conjunction with the weighted comprehensive index method to calculate the motor performance loss index.

7. The method for diagnosing faults in an AC asynchronous motor according to claim 2, characterized in that, The motor fault diagnosis result is generated by fusing the motor performance loss index and motor fault risk level corresponding to each of the operating status analysis units, including: Using each of the aforementioned operating status analysis units as basic units, the corresponding motor fault risk level is used as the row vector input and the corresponding motor performance loss index is used as the column vector input, which are mapped to a preset fault risk loss discrimination matrix. The fault risk loss discrimination matrix presets diagnostic conclusion labels based on the combination of different motor fault risk levels and motor performance loss indices. The severity level of motor faults for each operating status analysis unit is obtained by matching. The severity levels of motor faults include warning, attention, alarm, and shutdown. Based on the motor fault severity level corresponding to all operating status analysis units, a trend chart of the motor's full-cycle operating status and a comprehensive fault diagnosis report are generated as the output of the comprehensive fault diagnosis result of the motor.

8. A fault diagnosis device for an AC asynchronous motor, characterized in that, The apparatus for performing an AC asynchronous motor fault diagnosis method as described in any one of claims 1 to 7 comprises: The multi-source sensing module is used to non-intrusively install multi-source sensors on the outside of an AC asynchronous motor to collect multi-source monitoring data during motor operation. The data preprocessing module is used to preprocess the multi-source monitoring data; The mechanism modeling module is used to construct a mechanism model based on the electrical and mechanical coupling mechanism of AC asynchronous motors, reflecting the intrinsic relationship between different parameters of the motor and the state of the electrical system under normal operation and typical fault conditions. The information fusion module is used to fuse pre-processed multi-source monitoring data using information fusion algorithms to obtain a fused sensor dataset. The model training and diagnosis module is used to optimize and tune the mechanism model based on the fused sensor dataset using machine learning algorithms to obtain a motor fault diagnosis model. The motor fault diagnosis model outputs motor fault diagnosis results based on real-time acquired multi-source monitoring data.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the AC asynchronous motor fault diagnosis method as described in any one of claims 1 to 7.

Images