CN-122020451-A - Motor fault diagnosis method based on inherent characteristic frequency

CN122020451ACN 122020451 ACN122020451 ACN 122020451ACN-122020451-A

Abstract

The invention discloses a motor fault diagnosis method based on natural characteristic frequency, which constructs a diagnosis system based on a conventional electromechanical fault simulation platform and signal acquisition equipment. The method comprises the steps of constructing a multi-condition normal data feature library containing environment compensation factors, preprocessing a real-time collected bidirectional acceleration signal through filtering, calibration and standardization, extracting multi-dimensional features through a parallel framework, weighting, comparing and judging fault candidate data, carrying out fault type identification after confirming faults through continuity verification and abnormal elimination, ensuring scene suitability through combination of regular and instant dual-mode self-adaptive calibration, and evaluating fault severity and feeding back and optimizing. The method has the advantages of high diagnosis accuracy, good real-time performance, strong adaptability and high engineering realizability, and is suitable for operation and maintenance of the industrial motor.

Inventors

DING XINYUE
LIU JIAN
ZHENG SU
ZHANG SHUO
DENG JIANDONG
CHEN YUMING

Assignees

北京航天测控技术有限公司

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (9)

1. The motor fault diagnosis method based on the inherent characteristic frequency is characterized by being realized based on a diagnosis system constructed by an electromechanical fault simulation platform and signal acquisition equipment, and comprises the following steps of: 1) Acquiring acceleration signals in horizontal and vertical directions of a fault-free motor, extracting time domain features, frequency domain features and feature association parameters after preprocessing, respectively establishing corresponding feature threshold intervals for the three types of features, and further constructing a normal data feature library containing the three types of features and the corresponding threshold intervals; 2) Collecting bidirectional acceleration signals of a motor to be diagnosed at an adaptive sampling rate, and eliminating noise and errors through filtering, calibration and standardization processing; 3) The feature extraction and fault candidate judgment, namely segmenting the preprocessed data to be diagnosed according to a fixed period, extracting bidirectional features, comparing the bidirectional features with a feature library threshold value, and judging the data to be fault candidate data when any abnormal condition is met; 4) Performing time continuity verification and exception removal on the fault candidate data, and confirming the fault after the verification is passed; 5) Positioning fault types, namely identifying rotor unbalance, eccentricity, shaft bending and rotor bar faults of the confirmed fault data through feature comparison; 6) And (3) characteristic self-adaptive calibration, namely updating a characteristic library threshold value and a fault judging rule at regular intervals or according to requirements, and guaranteeing scene suitability.
2. The method of claim 1, wherein the time domain features include acceleration amplitude ranges and rms values, the frequency domain features include FFT spectra, fundamental frequency component amplitudes and harmonic component amplitudes, and the feature-related parameters include correlation of different directions of FFT spectra with normal FFT spectra.
3. The method according to claim 1, wherein the abnormal condition includes that the correlation degree between the FFT spectrum in the horizontal direction or the vertical direction and the FFT spectrum in the corresponding direction in the normal data feature library is significantly lower than the normal range, the acceleration amplitude in the horizontal direction or the vertical direction is not in the corresponding direction acceleration amplitude interval in the normal data feature library, and the FFT fundamental frequency component amplitude in the horizontal direction or the vertical direction is not in the corresponding direction fundamental frequency component amplitude interval in the normal data feature library.
4. The method according to claim 1, wherein the fault is confirmed by determining that the continuous plurality of sets of data are all determined as fault candidate data, and the characteristic deviation value of each set of fault candidate data reaches a set level, and confirming that the motor has a fault.
5. The method of claim 1, wherein the rule for fault type localization is as follows: a) If the amplitude of the acceleration signals in the horizontal direction and the vertical direction is obviously increased compared with the normal data and the amplitude of the fundamental frequency component in the frequency domain exceeds the upper limit of the amplitude interval of the fundamental frequency component in the normal data feature library, judging that the rotor imbalance fault exists; b) If the correlation degree of the FFT spectrum in the vertical direction and the FFT spectrum of normal data is lower than a conventional threshold value and the amplitude of a certain subharmonic component reaches more than a certain proportion of the amplitude of a fundamental frequency component, judging that the rotor is eccentric; c) The shaft bending fault is judged if the correlation degree of the FFT spectrum in the horizontal direction and the FFT spectrum of normal data is lower than a conventional threshold value and the amplitude of any one of the harmonic components is not detected to reach a certain proportion of the amplitude of the fundamental frequency component; d) And (3) judging the rotor bar fault if the amplitude of the FFT fundamental frequency component is lower than the lower limit of the fundamental frequency component amplitude interval in the normal data feature library and the rms values of the acceleration signals in the horizontal and vertical directions are lower than the lower limit of the corresponding rms value interval in the normal data feature library.
6. The method of claim 1, wherein the step of determining the position of the substrate comprises, The periodic calibration is triggered by time or diagnosis times, and the instant calibration is triggered by feature blurring or manually, and the global or local feature threshold is updated respectively.
7. The method of claim 1, wherein the step of determining the position of the substrate comprises, And the normal data feature library introduces an environment compensation factor, a mapping model is established by collecting normal motor signals under different working conditions, and real-time features are corrected according to real-time environment parameters during feature comparison.
8. The method according to any one of claims 1 to 7, wherein, The method also comprises a diagnosis result verification and feedback step, wherein the diagnosis result is compared with the actual fault state, and the calibration flow is automatically triggered when the accuracy is lower than the threshold value.
9. The method of claim 8, wherein the feature extraction is performed in parallel with feature extraction in a horizontal direction and in a vertical direction using a parallel computing architecture.

Description

Motor fault diagnosis method based on inherent characteristic frequency Technical Field The invention relates to the technical field of motor fault diagnosis, in particular to a motor fault diagnosis method based on acceleration signal inherent characteristic frequency analysis, which is suitable for a diagnosis system constructed by a conventional electromechanical fault simulation platform and signal acquisition equipment, realizes real-time identification and positioning of motor rotor eccentricity, shaft bending, rotor imbalance and rotor bar faults, and can be widely applied to scenes such as industrial motor operation and maintenance, intelligent manufacturing equipment monitoring and the like. Background The motor is used as a core power device for industrial production, and the running state of the motor directly determines the production efficiency and the safety. According to the operation and maintenance data statistics of industrial equipment, most motor faults are caused by rotor system abnormality, and if the faults are not diagnosed in time, equipment shutdown and production interruption can be caused, and even safety accidents are caused. Therefore, developing efficient and accurate motor fault diagnosis technology has important engineering value. The existing motor fault diagnosis technology mainly comprises three types, namely a diagnosis method based on vibration signals, wherein faults are identified by analyzing time domain or frequency domain characteristics of motor vibration, but the traditional method is mostly dependent on single characteristics, the diagnosis accuracy is greatly influenced by working conditions, a diagnosis method based on current signals, faults are judged by frequency spectrum analysis of motor stator current, the current signals are easily interfered by a power grid, the sensitivity to slight faults is low, and a diagnosis method based on auxiliary signals such as temperature, noise and the like can only be used as fault early warning references and cannot realize fault type positioning. In the related diagnosis method proposed by the published patent, the diagnosis accuracy is obviously reduced when the fixed threshold value of the laboratory is directly applied to the engineering site by extracting the time domain entropy value and the frequency domain peak value characteristic and combining the neural network classification, but the vibration characteristic difference of different test tables is not considered. In addition, the prior art has the core defect of poor scene adaptability that the motor installation modes, the load fluctuation range and the environmental noise level of different test tables are different, so that the vibration characteristics of the same fault in different scenes are obviously different. For example, the fundamental frequency amplitude increment of a rotor imbalance fault of a laboratory is in a specific range, but under the high-load working condition of an industrial workshop, the increment can obviously change, and the conventional fixed threshold method is prone to miss judgment. Meanwhile, the existing method has insufficient mechanism interpretation of fault characteristics, most of the existing methods depend on experimental data statistics, and theoretical support is lacked, so that the universality of the diagnosis rules is limited. Disclosure of Invention In view of the above, the invention provides a motor fault diagnosis method based on natural characteristic frequency, which obtains high-fidelity acceleration signals through conventional signal acquisition equipment, analyzes and determines the difference of each fault characteristic by combining motor fault mechanism, constructs a time domain-frequency domain multi-characteristic fusion diagnosis model, introduces a characteristic self-adaptive calibration mechanism, forms a standardized diagnosis flow, solves the problems of weak fault distinguishing capability of multiple types of rotors, poor scene adaptability of different test tables, fixed threshold, single characteristic and fuzzy mechanism in the prior art, realizes accurate positioning of rotor eccentricity, shaft bending, rotor imbalance and rotor bar faults, adapts vibration characteristics of different experimental environments, ensures repeatability and engineering realizability of the method, and has the technical advantages of short diagnosis delay and high accuracy. The specific technical scheme is as follows: The motor fault diagnosis method based on the inherent characteristic frequency is realized based on a diagnosis system constructed by an electromechanical fault simulation platform and signal acquisition equipment, and comprises the following steps: 1) Acquiring acceleration signals in horizontal and vertical directions of a fault-free motor, extracting time domain features, frequency domain features and feature association parameters after preprocessing, respectively establishing correspondin