CN116861320A - Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation - Google Patents

Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation Download PDF

Info

Publication number
CN116861320A
CN116861320A CN202310639247.7A CN202310639247A CN116861320A CN 116861320 A CN116861320 A CN 116861320A CN 202310639247 A CN202310639247 A CN 202310639247A CN 116861320 A CN116861320 A CN 116861320A
Authority
CN
China
Prior art keywords
time
signal
short
frequency
fault
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310639247.7A
Other languages
Chinese (zh)
Inventor
赵振民
李江红
崔佳航
单航
马蓉
符立梅
陈超
王彤
尚纯洁
雷庆春
蔡飞超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202310639247.7A priority Critical patent/CN116861320A/en
Publication of CN116861320A publication Critical patent/CN116861320A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00—Testing of engines
    • G01M15/14—Testing gas-turbine engines or jet-propulsion engines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G06T7/41—Analysis of texture based on statistical description of texture
    • G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The application discloses a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation, relates to the technical field of gas turbine rotor fault diagnosis, and solves the problems of high misjudgment rate of fault diagnosis by experience in the prior art; the method comprises the following steps: inputting a fault vibration signal to be judged, and carrying out data preprocessing on the fault vibration signal; wherein, data preprocessing includes: denoising the fault vibration signal to determine a denoising signal; performing short-time Fourier transform on the denoising signals to determine short-time Fourier time-frequency domain signals; performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal; extracting characteristics of the compressed time-frequency domain signals, and determining signal characteristics; inputting the signal characteristics into a trained fault diagnosis model, and outputting a diagnosis result; the method realizes the flow of fault diagnosis, improves the diagnosis effect, reduces the misjudgment rate and ensures the normal operation of the gas turbine.

Description

Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation
Technical Field
The application relates to the technical field of gas turbine rotor fault diagnosis, in particular to a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation.
Background
Gas turbines are a common generator set and large power drive system. Because of its high efficiency and high reliability, gas turbines are widely used in a variety of heavy industrial fields such as power plants, aeroengines, large vessels, and the like.
The gas turbine rotor consists of a plurality of blades and a wheel disc, and the rotor is driven to rotate by high-speed air flow generated by gas injection, so that energy is finally converted into electricity or power to be output. The working environment is more severe and changeable, the gas turbine rotor can be aged gradually under the high-temperature and high-pressure environment after long-term operation, the residual service life can be reduced gradually, the potential possibility of failure is increased gradually, and the gas turbine rotor is easy to generate typical failures such as rub-impact, unbalance, misalignment, cracks and the like due to the complexity and the requirements of the high-speed rotating parts of the rotor. Once the fault occurs, the fault not only causes huge economic loss, but also causes catastrophic casualties, and serious social influence is formed.
At present, how to diagnose the fault according to the rub-against, unbalance and misalignment of the rotor of the gas turbine is a problem to be solved.
Disclosure of Invention
The rotor fault diagnosis method based on short-time Fourier synchronous compression transformation solves the problem that fault diagnosis is carried out by experience in the prior art, and the misjudgment rate is high, achieves fault diagnosis flow, improves diagnosis effect, reduces misjudgment rate, and ensures normal operation of a gas turbine.
The embodiment of the application provides a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation, which comprises the following steps:
inputting a fault vibration signal to be judged, and carrying out data preprocessing on the fault vibration signal; wherein the data preprocessing comprises: denoising the fault vibration signal to determine a denoising signal; performing short-time Fourier transform on the denoising signals to determine short-time Fourier time-frequency domain signals; performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal; extracting the characteristics of the compressed time-frequency domain signals to determine signal characteristics;
and inputting the signal characteristics into a trained fault diagnosis model, and outputting a diagnosis result.
In one possible implementation manner, the denoising processing for the fault vibration signal, determining a denoising signal, includes:
performing wavelet decomposition on the fault vibration signal to determine a high-frequency component and a low-frequency component;
processing the high-frequency component according to a threshold value, removing the high-frequency component exceeding the threshold value, and determining the residual high-frequency component;
and carrying out wavelet reconstruction on the residual high-frequency component and the low-frequency component, and determining a denoising signal.
In one possible implementation manner, the performing short-time fourier transform on the denoising signal to determine a short-time fourier time-frequency domain signal includes:
the fault vibration signal is expressed as:
wherein t represents the center position of the window function, K represents the number of components, A k (t) represents the instantaneous amplitude value,representing the instantaneous phase;
the short-time fourier transform S (t, f) is expressed as:
where f represents frequency, τ represents a time parameter, g (τ -t) represents a window function, and f (τ -t) represents a frequency distribution function corresponding to the window function.
In one possible implementation manner, the performing synchronous compression transformation on the short-time fourier time-frequency domain signal to determine a compressed time-frequency domain signal includes:
wherein T (T, ω) represents a frequency distribution after the synchronous short-time fourier processing is performed on the denoising signal, g (0) represents a window function at time 0, δ represents a dirac function, S (T, f) represents a signal S (T) short-time fourier transform function, ω (T, f) represents a compressed instantaneous frequency, and a specific calculation formula of the compressed instantaneous frequency is:
where i represents an imaginary unit, and Re represents a real operation.
In one possible implementation manner, the extracting the characteristics of the compressed time-frequency domain signal to determine the signal characteristics includes:
extracting energy texture features and contrast features from the compressed time-frequency domain signal image;
and extracting the energy texture characteristics and the contrast characteristics of the R, G, B channel respectively from the compressed time-frequency domain signal image.
In one possible implementation manner, the specific calculation formula of the energy texture feature is:
where G (i, j) represents a gray level co-occurrence matrix of an image, i represents a value of an abscissa of an element in the gray level co-occurrence matrix, j represents a value of an ordinate of an element in the gray level co-occurrence matrix, and k represents a maximum value of an element in the gray level co-occurrence matrix.
In one possible implementation manner, the specific calculation formula of the contrast characteristic is:
wherein G (i, j) represents a gray level co-occurrence matrix of an image, i represents a value of an abscissa of an element in the gray level co-occurrence matrix, j represents a value of an ordinate of an element in the gray level co-occurrence matrix, k represents a maximum value of an element in the gray level co-occurrence matrix, and n represents a difference between the abscissa and the ordinate of an element in the gray level co-occurrence matrix.
In one possible implementation, the energy texture feature and contrast feature include: edge number, mean degree, and aggregation coefficient.
In one possible implementation, training the fault diagnosis model includes:
acquiring a plurality of sample signals, preprocessing the data of the plurality of sample signals, and determining preprocessed data;
dividing the preprocessed data into a training data set and a test data set, inputting the training data set into an initial fault diagnosis model for training, and determining a training fault diagnosis model;
and inputting the test data set into the training fault diagnosis model to obtain a fault type, evaluating the fault type, if the fault type meets the preset condition, finishing training, and otherwise, continuing training.
In one possible implementation, the preprocessed data is a signal feature.
One or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:
the embodiment of the application adopts a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation, which comprises the following steps: inputting a fault vibration signal to be judged, and carrying out data preprocessing on the fault vibration signal; wherein, data preprocessing includes: denoising the fault vibration signal to determine a denoising signal; performing short-time Fourier transform on the denoising signals to determine short-time Fourier time-frequency domain signals; the short-time Fourier transform can decompose the signal into frequency distribution diagrams in different time periods, so that the frequency distribution characteristics of the signal can be clearly observed, and the position and the amplitude of the fault frequency can be determined; performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal; the signal-to-noise ratio of the fault signal is improved, and the fault characteristics are further enhanced; extracting characteristics of the compressed time-frequency domain signals, determining signal characteristics, extracting various characteristic information, and providing classification judgment for refined information; inputting the signal characteristics into a trained fault diagnosis model, and outputting a diagnosis result; the problem that fault diagnosis is carried out by experience and the misjudgment rate is high is effectively solved, the fault diagnosis is realized, the diagnosis effect is improved, the misjudgment rate is reduced, and the normal operation of the gas turbine is ensured.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are needed in the embodiments of the present application or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation provided by an embodiment of the application;
FIG. 2 is a flowchart of a step of denoising a fault vibration signal according to an embodiment of the present application;
FIG. 3 is a time-frequency diagram of a rotor normal short-time Fourier transform provided by an embodiment of the application;
FIG. 4 is a time-frequency chart of a rotor rub-impact short-time Fourier transform provided by an embodiment of the application;
FIG. 5 is a time-frequency diagram of a rotor imbalance short-time Fourier transform provided by an embodiment of the application;
FIG. 6 is a time-frequency diagram of a rotor misalignment mid-short time Fourier transform provided by an embodiment of the application;
FIG. 7 is a time-frequency diagram of rotor normal synchronous compression conversion provided by an embodiment of the application;
FIG. 8 is a time-frequency diagram of rotor rub-impact synchronous compression conversion provided by an embodiment of the application;
FIG. 9 is a time-frequency diagram of rotor imbalance synchronous compression conversion provided by an embodiment of the present application;
FIG. 10 is a time-frequency diagram of rotor misalignment synchronous compression conversion provided by an embodiment of the present application;
FIG. 11 illustrates an energy texture feature of a synchronous compressed time-frequency RGB map provided by an embodiment of the present application;
FIG. 12 is a contrast texture feature of a synchronous compressed time-frequency RGB map provided by an embodiment of the present application;
FIG. 13 is a graph of R channel energy texture characteristics for a synchronous compressed time-frequency graph according to an embodiment of the present application;
fig. 14 is a graph of a contrast texture feature of an R channel of a synchronous compressed time-frequency diagram according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It will be apparent that the described embodiments are some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In rotor vibration fault diagnosis, fourier transform FFT is an important frequency domain analysis method, and a rotor vibration signal is usually a time-varying signal, that is, its frequency characteristics change with time. FFT frequency domain analysis is a method of converting a time domain signal into a frequency domain signal, which can decompose the signal into its frequency components, so that the frequency characteristics of the signal can be analyzed. By subjecting the vibration signal to FFT analysis, the frequency content of the signal, as well as its intensity or amplitude, can be obtained. This information can be used to diagnose the type of vibration failure of the rotor. For example, rotor imbalance may cause it to vibrate periodically during rotation, which appears as distinct frequency peaks in the FFT frequency domain analysis. In addition, rotor imbalance tends to produce 1 multiplication; misalignment of the rotor tends to produce 1-fold, 2-fold, 3-fold; rotor rub tends to produce 0.5 frequency doubling, 1 frequency doubling, 1.5 frequency doubling, 2 frequency doubling, 2.5 frequency doubling, 3 frequency doubling, 3.5 frequency doubling, 4 frequency doubling. By analyzing the characteristic features of the FFT frequency domain, the fault types can be diagnosed and judged. However, in the case of non-stationary signals, i.e. signals whose frequency characteristics vary with time, conventional fourier transforms do not solve this problem well. This is because fourier transforms transform the entire signal and therefore cannot analyze the local time domain characteristics of the signal. The short-time fourier transform is an improvement over conventional fourier transforms in that it divides the signal into a plurality of short-time windows and fourier transforms the signal within each window. This results in frequency components of the signal within each time window so that local time-frequency characteristics of the non-stationary signal can be analyzed. Therefore, the short-time Fourier transform can analyze the time-frequency characteristics of the non-stationary signals and is more sensitive to the local time domain characteristics of the signals; the trade-off of frequency and time resolution can be controlled, with the resolution being adjusted by window size and overlap; the problem that the frequency resolution and the time resolution are mutually sacrificed can be avoided, and the accuracy and the reliability of signal processing are improved. Then, the short-time Fourier transform is limited by the time-frequency measurement inaccuracy principle, and the time resolution and the frequency resolution can not be simultaneously optimized, so that the energy can not be represented in a concentrated manner. The synchronous compression transformation achieves the purpose of improving the time-frequency aggregation through the time-frequency rearrangement of the energy coefficient. The synchronous compression transformation superimposes the frequency spectrums in the pseudo frequency interval, so that the energy is concentrated on the actual instantaneous frequency, the time-frequency aggregation is improved, and the characteristics of energy, contrast and the like are more obvious. The method can help engineers to rapidly and accurately analyze the frequency characteristics of the rotor vibration signals, thereby diagnosing the type of the rotor vibration faults. Therefore, more fault characteristics in different forms can be excavated as much as possible through the transformation of the variable domain, and the accuracy and the precision of rotor fault diagnosis can be improved. Aiming at the collision friction, unbalance and misalignment faults of the rotor of the gas turbine, the rotor fault diagnosis method of the gas turbine based on multi-transform domain feature extraction is provided, and finally, the rotor fault diagnosis method based on short-time Fourier synchronous compression transformation is verified through a modeling rotor experiment.
The embodiment of the application provides a rotor fault diagnosis method based on short-time Fourier synchronous compression transformation, which comprises the following steps S101 to S102 as shown in figure 1.
S101, inputting a fault vibration signal to be judged, and carrying out data preprocessing on the fault vibration signal; wherein, data preprocessing includes: denoising the fault vibration signal to determine a denoising signal; performing short-time Fourier transform on the denoising signals to determine short-time Fourier time-frequency domain signals; performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal; and extracting the characteristics of the compressed time-frequency domain signals, and determining the characteristics of the signals.
S102, inputting the signal characteristics into the trained fault diagnosis model, and outputting a diagnosis result.
The rotor fault diagnosis method based on short-time Fourier synchronous compression transformation, which is disclosed by the application, uses the methods of short-time Fourier transformation, synchronous compression transformation and extraction of energy and contrast characteristics of channels with different colors, so that fault characteristics can be effectively extracted from vibration signals, and the accuracy and reliability of fault diagnosis can be improved, and particularly, the method has the advantages that: frequency domain characteristics of the enhancement signal: the short-time Fourier transform can decompose the signal into frequency distribution diagrams of different time periods, and the frequency distribution characteristics of the signal can be clearly observed, so that the position and the amplitude of the fault frequency can be determined. Improving the signal-to-noise ratio of the signal: the synchronous compression transformation is an improved algorithm based on short-time Fourier transformation, and can improve the signal-to-noise ratio of fault signals and further enhance fault characteristics. Providing more feature information: the extraction of the energy and contrast characteristics of different color channels can provide richer information, including the characteristics of the energy, contrast and the like of signals, and the information can help to improve the classification and discrimination capability of faults. The accuracy and the reliability of fault diagnosis are improved: the short-time Fourier transform, synchronous compression transform and different-color channel energy and contrast characteristic extraction methods are comprehensively utilized, so that the accuracy and reliability of fault diagnosis can be improved, the misdiagnosis rate and the missed diagnosis rate are reduced, and better guarantee is provided for health monitoring and maintenance of equipment. In summary, by using the methods for fault diagnosis, the diagnosis effect can be improved, the misjudgment rate can be reduced, and the normal operation of large-scale mechanical equipment such as a gas turbine and the like can be ensured.
In one possible implementation, the fault vibration signal is subjected to denoising processing, and the denoising signal is determined, as shown in fig. 2, including the following steps S201 to S203.
S201, performing wavelet decomposition on the fault vibration signal to determine a high-frequency component and a low-frequency component.
S202, processing the high-frequency components according to the threshold value, removing the high-frequency components exceeding the threshold value, and determining the residual high-frequency components.
S203, carrying out wavelet reconstruction on the residual high-frequency components and the low-frequency components, and determining a denoising signal.
In one possible implementation, performing short-time fourier transform on the denoised signal to determine a short-time fourier time-frequency domain signal includes: the method comprises the steps of carrying out data transformation on a training data set of a modeling rotor fault vibration signal of a gas turbine by adopting short-time Fourier, converting the modeling rotor vibration signal into a short-time Fourier time-frequency domain signal, and for non-stationary and wide stationary signals, knowing the frequency characteristic change with time, so that time-frequency analysis is needed. The short-time Fourier transform process is that the signals with non-stationary and wide stationary are multiplied by a time window function before transformation, the signals in the window with limited time are assumed to be stationary, the Fourier transform of the signals of each window is calculated through sliding of the window function on the signals, finally, a group of time sequence up-spectrograms are obtained, and the time sequence up-spectrograms are spliced according to time axes, so that the short-time Fourier transform time-frequency chart of the signals is obtained. The fault vibration signal is expressed as:
wherein t represents the center position of the window function, K represents the number of components, A k (t) represents the instantaneous amplitude value,representing the instantaneous phase;
the short-time fourier transform S (t, f) is expressed as:
where f represents frequency, τ represents a time parameter, g (τ -t) represents a window function, and f (τ -t) represents a frequency distribution function corresponding to the window function. And a Hamming window with a window size of 256 is adopted for processing the modeling rotor fault vibration time domain data for short-time Fourier. Fig. 3, 4, 5 and 6 are respectively time-frequency diagrams of short-time fourier transform (in turn, normal, unbalanced, rub-impact, and misalignment) of gas turbine modeling rotor failure vibration data. However, the STFT is limited by the time-frequency measurement inaccuracy principle that the time resolution and the frequency resolution cannot be simultaneously optimized, and the energy cannot be centrally represented.
In one possible implementation manner, the synchronous compression transformation is used for carrying out short-time Fourier coefficient compression transformation on the gas turbine modeling rotor fault short-time Fourier time frequency domain signal to obtain a compressed time frequency domain signal, and the synchronous compression transformation is used for achieving the purpose of improving time frequency aggregation through time frequency rearrangement of short-time Fourier energy coefficients. The rotor normal synchronous compression conversion time-frequency diagram is shown in fig. 7, the rotor rub synchronous compression conversion time-frequency diagram is shown in fig. 8, the rotor unbalanced synchronous compression conversion time-frequency diagram is shown in fig. 9, and the rotor misalignment synchronous compression conversion time-frequency diagram is shown in fig. 10.
Specifically, the method for performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal comprises the following steps:
wherein T (T, ω) represents a frequency distribution after the synchronous short-time fourier processing is performed on the denoised signal, g (0) represents a window function at time 0, δ represents a dirac function, S (T, f) represents a signal S (T) short-time fourier transform function, ω (T, f) represents a compressed instantaneous frequency, and a specific calculation formula of the compressed instantaneous frequency is:
where i represents an imaginary unit, and Re represents a real operation.
The synchronous compression transformation superimposes the frequency spectrums in the pseudo frequency interval, so that the energy is concentrated on the actual instantaneous frequency, the time-frequency aggregation is improved, and the characteristics of energy, contrast and the like are more obvious.
In one possible implementation, the compressed time-frequency domain signal is subjected to feature extraction, and signal features are determined, including the following steps S701 to S702.
And S701, extracting energy texture features and contrast features from the image of the compressed time-frequency domain signal. Energy texture features and contrast features, including: edge number, mean degree, and aggregation coefficient.
The specific calculation formula of the energy texture features is as follows:
where G (i, j) represents the gray level co-occurrence matrix of the image, i represents the value of the abscissa of the element in the gray level co-occurrence matrix, j represents the value of the ordinate of the element in the gray level co-occurrence matrix, and k represents the maximum value of the element in the gray level co-occurrence matrix.
The specific calculation formula of the contrast characteristic is as follows:
where G (i, j) represents a gray level co-occurrence matrix of the image, i represents a value of an abscissa of an element in the gray level co-occurrence matrix, j represents a value of an ordinate of an element in the gray level co-occurrence matrix, k represents a maximum value of an element in the gray level co-occurrence matrix, and n represents a difference between the abscissa and the ordinate of the element in the gray level co-occurrence matrix.
S702, extracting energy texture features and contrast features of R, G, B channels from the compressed time-frequency domain signal image respectively. And calculating according to the energy texture features and the contrast features, respectively determining the compressed images of the time-frequency domain signals, and respectively carrying out R, G, B channel energy texture features and contrast features. As shown in fig. 11, the energy texture map of the compressed time-frequency domain signal RGB is shown in fig. 12, the contrast texture map of the compressed time-frequency domain signal RGB is shown in fig. 13, the R-channel energy texture map of the compressed time-frequency domain signal RGB is shown in fig. 14, and the R-channel contrast texture map of the compressed time-frequency domain signal RGB is shown in fig. 14.
In one possible implementation, the fault diagnosis model is trained, including the following steps S701 to S703.
S701, acquiring a plurality of sample signals, performing data preprocessing on the plurality of sample signals, and determining preprocessed data. In the signal acquisition process of the gas turbine modeling rotor experiment table, the rotor speed is set to 1200r/min, the sampling frequency is set to 2048Hz, and the sampling length is set to 1s. Each 45 sets of tests were performed under different rotor conditions (normal, contact friction, imbalance and misalignment) and then a total of 180 sequential sample data were obtained for the gas turbine modeling rotor failure dataset, including 4 failure types, imbalance, rub-on, misalignment, normal, setting the failure flags to 1, 2, 3, 4 respectively and the dataset was taken as 6:4 into training set and test set, and the fault type label is shown in table 1.
TABLE 1 Fault type Label correspondence Table
Fault type Label setting
Imbalance of 1
Rubbing with a ball 2
Misalignment of 3
Normal state 0
The preprocessing of the data comprises the steps of removing noise by wavelet transformation, short-time Fourier transformation, synchronous compression transformation and feature extraction.
S702, dividing the preprocessed data into a training data set and a test data set, inputting the training data set into an initial fault diagnosis model for training, and determining a training fault diagnosis model. And inputting the characteristic signals into an SVM classification algorithm model for learning, and obtaining a gas turbine modeling rotor fault diagnosis model through iterative training learning.
S703, inputting the test data set into a training fault diagnosis model to obtain a fault type, evaluating the fault type, if the preset condition is met, finishing training, otherwise, continuing training. And verifying the rotor fault diagnosis model by using a test data set, if the verification meets the precision requirement, iterating the rotor fault diagnosis classification model to perform real-time classification diagnosis, if the verification does not meet the precision requirement, repeating the process, continuously training-verifying the iteration until the rotor fault diagnosis classification model meets the fault diagnosis model with the precision requirement, performing energy texture and contrast characteristics with the time-frequency domain signal image and the RGB color channel of the modeling rotor fault short-time Fourier synchronous compression transformation, and performing accuracy and fault false alarm rate effects by using a Kmeans clustering method to compare. The Kmeans method is used for carrying out energy texture and contrast characteristic clustering fault diagnosis on the modeling rotor fault short-time Fourier synchronous compression transformation time-frequency domain signal image and the RGB color channel, wherein the accuracy rate is 88.88%, and the false alarm rate is more than 8%. The SVM method is used for carrying out energy texture and contrast characteristic fault diagnosis on the modeling rotor fault short-time Fourier synchronous compression transformation time-frequency domain signal image and the RGB color channel, and the accuracy rate is 94.44%, and the false alarm rate is less than 5%.
When the trained fault diagnosis model is used, the fault vibration signals are subjected to the same preprocessing, and then the signal characteristics are input into the fault diagnosis model to output the diagnosis result.
The rotor fault diagnosis method based on short-time Fourier synchronous compression change achieves the purpose of improving time-frequency aggregation through time-frequency rearrangement of energy coefficients. The synchronous compression transformation superimposes the frequency spectrums in the pseudo frequency interval, so that the energy is concentrated on the actual instantaneous frequency, the time-frequency aggregation is improved, and the characteristics of the energy, the contrast ratio and the like are more obvious, so that the rotor fault can be better diagnosed. And by combining with the SVM classification algorithm, the diagnosis accuracy of the rotor faults of the gas turbine can be effectively improved, and the fault false alarm rate can be effectively reduced. The method can realize rapid and accurate rotor fault diagnosis, and verifies the correctness and feasibility of the rotor fault diagnosis method based on short-time Fourier synchronous compression transformation.
From the above description of embodiments, it will be apparent to those skilled in the art that the present application may be implemented in software plus necessary hardware. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product or may be embodied in the implementation of data migration. The computer software product may be stored on a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., comprising instructions for causing a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to perform the methods described in the various embodiments or portions of the embodiments of the application.
In this specification, each embodiment is described in a progressive manner, and the same or similar parts of each embodiment are referred to each other, and each embodiment is mainly described as a difference from other embodiments. All or portions of the present application are operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the present application; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced with equivalents; such modifications and substitutions do not depart from the spirit of the application.

Claims (10)

1. A rotor fault diagnosis method based on short-time Fourier synchronous compression transformation is characterized by comprising the following steps:
inputting a fault vibration signal to be judged, and carrying out data preprocessing on the fault vibration signal; wherein the data preprocessing comprises: denoising the fault vibration signal to determine a denoising signal; performing short-time Fourier transform on the denoising signals to determine short-time Fourier time-frequency domain signals; performing synchronous compression transformation on the short-time Fourier time-frequency domain signal to determine a compressed time-frequency domain signal; extracting the characteristics of the compressed time-frequency domain signals to determine signal characteristics;
and inputting the signal characteristics into a trained fault diagnosis model, and outputting a diagnosis result.
2. The method of claim 1, wherein denoising the fault vibration signal to determine a denoised signal comprises:
performing wavelet decomposition on the fault vibration signal to determine a high-frequency component and a low-frequency component;
processing the high-frequency component according to a threshold value, removing the high-frequency component exceeding the threshold value, and determining the residual high-frequency component;
and carrying out wavelet reconstruction on the residual high-frequency component and the low-frequency component, and determining a denoising signal.
3. The method of claim 1, wherein said performing a short-time fourier transform on the denoised signal to determine a short-time fourier time-frequency domain signal comprises:
the fault vibration signal is expressed as:
wherein t represents the center position of the window function, K represents the number of components, A k (t) represents the instantaneous amplitude value,representing the instantaneous phase;
the short-time fourier transform S (t, f) is expressed as:
where f represents frequency, τ represents a time parameter, g (τ -t) represents a window function, and f (τ -t) represents a frequency distribution function corresponding to the window function.
4. The method of claim 1, wherein said performing a synchronous compression transformation on said short-time fourier time-frequency domain signal to determine a compressed time-frequency domain signal comprises:
wherein T (T, ω) represents a frequency distribution after the synchronous short-time fourier processing is performed on the denoising signal, g (0) represents a window function at time 0, δ represents a dirac function, S (T, f) represents a signal S (T) short-time fourier transform function, ω (T, f) represents a compressed instantaneous frequency, and a specific calculation formula of the compressed instantaneous frequency is:
where i represents an imaginary unit, and Re represents a real operation.
5. The method of claim 1, wherein the performing feature extraction on the compressed time-frequency domain signal to determine signal features comprises:
extracting energy texture features and contrast features from the compressed time-frequency domain signal image;
and extracting the energy texture characteristics and the contrast characteristics of the R, G, B channel respectively from the compressed time-frequency domain signal image.
6. The method of claim 5, wherein the energy texture feature specific calculation formula is:
where G (i, j) represents a gray level co-occurrence matrix of an image, i represents a value of an abscissa of an element in the gray level co-occurrence matrix, j represents a value of an ordinate of an element in the gray level co-occurrence matrix, and k represents a maximum value of an element in the gray level co-occurrence matrix.
7. The method of claim 5, wherein the contrast characteristic specification formula is:
wherein G (i, j) represents a gray level co-occurrence matrix of an image, i represents a value of an abscissa of an element in the gray level co-occurrence matrix, j represents a value of an ordinate of an element in the gray level co-occurrence matrix, k represents a maximum value of an element in the gray level co-occurrence matrix, and n represents a difference between the abscissa and the ordinate of an element in the gray level co-occurrence matrix.
8. The method of claim 5, wherein the energy texture feature and contrast feature comprise: edge number, mean degree, and aggregation coefficient.
9. The method of claim 1, wherein training the fault diagnosis model comprises:
acquiring a plurality of sample signals, preprocessing the data of the plurality of sample signals, and determining preprocessed data;
dividing the preprocessed data into a training data set and a test data set, inputting the training data set into an initial fault diagnosis model for training, and determining a training fault diagnosis model;
and inputting the test data set into the training fault diagnosis model to obtain a fault type, evaluating the fault type, if the fault type meets the preset condition, finishing training, and otherwise, continuing training.
10. The method of claim 9, wherein the preprocessed data is signal features.
CN202310639247.7A 2023-05-31 2023-05-31 Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation Pending CN116861320A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310639247.7A CN116861320A (en) 2023-05-31 2023-05-31 Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310639247.7A CN116861320A (en) 2023-05-31 2023-05-31 Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation

Publications (1)

Publication Number Publication Date
CN116861320A true CN116861320A (en) 2023-10-10

Family

ID=88232990

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310639247.7A Pending CN116861320A (en) 2023-05-31 2023-05-31 Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation

Country Status (1)

Country Link
CN (1) CN116861320A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117668751A (en) * 2023-11-30 2024-03-08 广东一业建设股份有限公司 High-low voltage power system fault diagnosis method and device

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117668751A (en) * 2023-11-30 2024-03-08 广东一业建设股份有限公司 High-low voltage power system fault diagnosis method and device
CN117668751B (en) * 2023-11-30 2024-04-26 广东一业建设股份有限公司 High-low voltage power system fault diagnosis method and device

Similar Documents

Publication Publication Date Title
Kumar et al. Optimization of VMD using kernel-based mutual information for the extraction of weak features to detect bearing defects
Zhang et al. Kurtosis based weighted sparse model with convex optimization technique for bearing fault diagnosis
Li et al. Fault diagnosis for rolling bearing based on VMD-FRFT
Lei et al. EEMD method and WNN for fault diagnosis of locomotive roller bearings
Kang et al. Reliable fault diagnosis of multiple induction motor defects using a 2-D representation of Shannon wavelets
Feng et al. A novel order spectrum-based Vold-Kalman filter bandwidth selection scheme for fault diagnosis of gearbox in offshore wind turbines
Chen et al. A fault pulse extraction and feature enhancement method for bearing fault diagnosis
Yao et al. Fault diagnosis of planetary gearbox based on acoustic signals
CN111060337B (en) Running equipment real-time fault diagnosis method based on expert system
Wang et al. Accuracy-improved bearing fault diagnosis method based on AVMD theory and AWPSO-ELM model
Sheng et al. Applications in bearing fault diagnosis of an improved Kurtogram algorithm based on flexible frequency slice wavelet transform filter bank
Song et al. Self-adaptive multivariate variational mode decomposition and its application for bearing fault diagnosis
Lv et al. High-order synchroextracting transform for characterizing signals with strong AM-FM features and its application in mechanical fault diagnosis
Li et al. Quantitative evaluation on the performance and feature enhancement of stochastic resonance for bearing fault diagnosis
CN112461934B (en) Aero-engine blade crack source positioning method based on acoustic emission
Xiang et al. Comparison of Methods for Different Time-frequency Analysis of Vibration Signal.
Yan et al. Bearing fault diagnosis via a parameter-optimized feature mode decomposition
CN116861320A (en) Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation
Zhang et al. A local transient feature extraction method via periodic low rank dynamic mode decomposition for bearing incipient fault diagnosis
Xu et al. An adaptive spectrum segmentation method to optimize empirical wavelet transform for rolling bearings fault diagnosis
CN108036940A (en) A kind of Method for Bearing Fault Diagnosis
CN115481657A (en) Wind generating set communication slip ring fault diagnosis method based on electric signals
Liang et al. Bearing fault diagnosis based on singular value distribution of impulse response segment
Hajnayeb et al. Vibration measurement for crack and rub detection in rotors
Cui et al. A spectral coherence cyclic periodic index optimization-gram for bearing fault diagnosis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination