CN109297713B - Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection - Google Patents
Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection Download PDFInfo
- Publication number
- CN109297713B CN109297713B CN201810892236.9A CN201810892236A CN109297713B CN 109297713 B CN109297713 B CN 109297713B CN 201810892236 A CN201810892236 A CN 201810892236A CN 109297713 B CN109297713 B CN 109297713B
- Authority
- CN
- China
- Prior art keywords
- signal
- stationary
- steam turbine
- original
- peak
- 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.)
- Active
Links
Classifications
-
- 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
-
- 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
- G06F2218/08—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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Control Of Turbines (AREA)
Abstract
The invention discloses a steam turbine main engine fault diagnosis method based on steady and non-steady vibration signal characteristic selection. The invention aims at a steam turbine in a thermal power steam turbine unit, combines and applies an integrated empirical mode decomposition (EEMD) and a recursive feature elimination method, and particularly decomposes and extracts key features of steady and non-steady signals in detail, so that the method is used for fault diagnosis of a steam turbine vibration signal. The method fully considers the characteristics of non-stationary vibration signals of the steam turbine, a large amount of noise and the like, fully excavates potential information contained in fault data, respectively extracts characteristics aiming at stationary and non-stationary data, and solves the problem that the characteristics of the non-stationary data are easily covered. Meanwhile, key features are extracted, the dimension of the feature vector is reduced, data redundancy is reduced, the accuracy of steam turbine vibration signal fault diagnosis is improved, and a field engineer can be helped to accurately repair the fault, so that safe and reliable operation of the generator and steam turbine equipment is ensured, and production benefits are improved.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of vibration signals, and particularly relates to a fault diagnosis method for a steam turbine main engine of a thermal power generating unit based on stable and non-stable vibration signal characteristic selection.
Background
With the progress of society and the development of technology, people have more and more demand for electricity. Among them, coal-fired power generation is one of the main power generation modes in China. In recent years, with the adjustment of the structure of the power generation industry, large units with large capacity, high parameters and low energy consumption gradually replace small units with high energy consumption, and the industrial process becomes increasingly complex. As the main equipment of coal-fired power generation, whether the main machine of the turbine can safely operate affects the operation condition of the whole coal-fired power generation unit, once the turbine breaks down, the whole power generation unit is probably shut down on line, and great inconvenience is brought to industrial production and daily life. The steam turbine main unit equipment structure is complicated, and numerous and the mutual influence of parameter is to the effective monitoring of steam turbine state the focus that people paid attention to always. Because of the sensitivity of the vibration to the state of the steam turbine set, when the steam turbine set is abnormal or has a fault, the vibration of the set usually reacts immediately, so the analysis of the vibration signal is an important means for monitoring the running state of the steam turbine, which is helpful for carrying out early analysis and diagnosis on the fault and taking corresponding measures in time. However, the working environment of the steam turbine is complex and severe, and the vibration on the main engine of the steam turbine set is a result of the combined action of multiple excitation sources, so that the vibration signal contains a large amount of noise, has the characteristics of strong nonlinearity and non-stationarity, and is difficult to directly utilize the vibration signal to extract the characteristics and model.
Therefore, for fault diagnosis of vibration signals, the signals need to be preprocessed first, and the signals are decomposed into a plurality of simple sub-signals, and the sub-signals contain single components, so that key fault features are easy to extract. And then selecting key characteristics of the signals, and selecting the key characteristics capable of reflecting fault information. Previous studies have been made on fault diagnosis based on vibration signals. Among them, various time-frequency signal processing methods such as Empirical Mode Decomposition (EMD), wavelet decomposition (WT), fourier transform, etc. have been widely used for preprocessing vibration signals and decomposing signals. In the aspect of feature extraction, energy entropy used for measuring energy distribution of signals on different scales is widely applied. However, only a single feature is extracted, and real fault information cannot be comprehensively reflected. However, if a plurality of features are calculated for all the sub-signals, redundancy of the features is caused, many features do not contain critical fault information, and even real fault information is hidden due to excessive irrelevant information. Therefore, a multi-feature selection model needs to be established, so that not only can key fault information be extracted, but also the redundancy of the information can be reduced.
The invention provides a fault diagnosis method based on steady and non-steady vibration signal characteristic selection aiming at a vibration signal of a steam turbine main engine of a thermal generator unit. The method comprises the steps of firstly preprocessing a vibration signal of a steam turbine shafting by utilizing integrated empirical mode decomposition, decomposing an original signal into a plurality of IMF components, then calculating a plurality of statistical characteristics of all the IMF components, selecting key characteristics by utilizing a recursive characteristic elimination algorithm, eliminating characteristic redundancy, and establishing a fault diagnosis model by utilizing the key characteristics, thereby greatly improving the performance of online fault diagnosis in the steam turbine operation process. The research report related to the invention is not seen yet.
Disclosure of Invention
The invention aims to provide a method for diagnosing the fault of a main steam turbine engine based on stable and non-stable vibration signal characteristic selection, which aims at the main steam turbine engine in a large coal-fired power generator unit.
The purpose of the invention is realized by the following technical scheme: a fault diagnosis method for a steam turbine host based on stable and non-stable vibration signal feature selection specifically comprises the following steps:
(1) the method comprises the following steps of collecting vibration signals of a steam turbine host under a normal state and different fault states, and performing model training as original vibration signals:
(1.1) carrying out integrated empirical mode decomposition on the original vibration signal, and decomposing the original vibration signal into n IMF components.
And (1.2) carrying out stationarity judgment on the IMF component of n, and dividing the IMF component into p stationary IMF components and q non-stationary IMF components.
(1.3) calculating the original statistical characteristics of the stationary IMF component and the non-stationary IMF component respectively. The original statistical characteristics include: RMS, Kurt, skewness Skaw, Peak-Peak value Peak-Peak, coefficient of Peak value Cf, slow characteristic Slowness and correlation coefficient Corcoef, wherein the slow characteristic Slowness is as follows:
n is the number of sampling points, Δ xiIs the difference in signal x.
The number of the original statistical features of the stationary IMF components is 7 × p, and the number of the original statistical features of the non-stationary IMF components is 7 × q.
(1.4) aiming at the original statistical characteristics of each state (the state comprises a normal state and different fault states), respectively applying a recursive characteristic elimination algorithm to the stationary signal xsAnd non-stationary signal xtSelecting k by selecting characteristicssKey characteristic sum k of stationary signaltA key feature of non-stationary signals.
(1.5) establishing a failureAnd (3) diagnosis model: k selected from (1.4)sKey characteristic sum k of stationary signaltAnd inputting the key characteristics of the non-stationary signals into a gradient lifting tree classifier for training to obtain a fault diagnosis model.
(2) Collecting vibration signals of a main machine of the steam turbine, and selecting k with highest feature importance according to the steps 1.1-1.4sKey characteristic sum k of stationary signaltAnd inputting the key characteristics of the non-stationary signals into a fault diagnosis model for fault diagnosis.
Wherein, the step of integrating empirical mode decomposition in the step (1.1) is as follows:
(1.1.1) adding a normally distributed white noise sequence to the original signal. And decomposing the signal added with the white noise into n IMF components through empirical mode decomposition.
(1.1.2) repeating the step (1.1.1) K-1 times, adding a new white noise sequence each time, and obtaining n IMF components:wherein i ∈ 1,2, … n, denotes the ith component, and j denotes the jth execution of modal decomposition;
(1.1.3) the IMF component finally obtained is I1,I2,I3,…Ii,…InWhereinK represents the number of execution times of modal decomposition.
Wherein the empirical mode decomposition is performed as follows
(a) Finding out all maximum value points and minimum value points of original vibration signal x (t), connecting the maximum value points and minimum value points with cubic spline curves to form an upper envelope line and a lower envelope line, wherein original data is contained between the upper envelope line and the lower envelope line, and calculating a mean value line m of the upper envelope line and the lower envelope line1Calculating a sequence of difference values of the signals as h1=x(t)-m1。
(b) Judgment h1Whether two conditions for the eigenmode components are met: (1) number of extreme points and zero crossing pointsThe numbers are the same or differ by at most one. (2) The mean of the two envelopes of the local maxima and local minima of the difference sequence is zero at any point. If h is1If the two conditions are not satisfied, repeating the step (a) and the step (h)1As the original sequence until h satisfying the eigenmode component is obtained1kUntil now.
(c) The first eigenmode component is denoted c1=h1kThe remainder r obtained1=x(t)-c1R is to1Repeating the steps (a) to (b) as new original data until the nth remainder rnIs a monotonic function, and ends up when the IMF component satisfying both conditions in (b) cannot be extracted. Finally obtainingThe original signal is decomposed into the sum of the residual and the n eigenmode components.
Further, in step 1.2, stationarity judgment is performed on the IMF components, and the specific method is to perform stationarity judgment on all IMF components by applying augmenteddickeyfuller (adf) test, so that the IMF components are divided into two parts, namely p stationary IMF components and q non-stationary IMF components, and n IMF components are obtained.
Further, in step (1.3), RMS is a root mean square value, and represents the average power of the signal; kurt is kurtosis and reflects the sharpness of the peak of the signal; skaw is skewness and reflects the skewness direction and degree of the signal; Peak-Peak is a Peak-Peak value and reflects the size of a signal fluctuation range; cf is the peak coefficient, which is the ratio of the peak value to the effective value; slowness reflects the speed of signal change; corcoef is the correlation coefficient. The specific formula is calculated as follows:
Peak_peak=xmax-xmin (4)
Corcoef=cor(x,xi) (7)
wherein, x in the formula (1)iFor the amplitude of the ith vibration signal in signal x, x ═ x1,x2,…,xN}; μ in equation (2) is the mean value of the signal x,σ is the standard deviation of the signal x; e represents a desired function; x in formula (4)maxIs the maximum value of the vibration signal, xminIs the minimum value of the vibration signal; rms in the formula (5) is the root mean square value calculated in the formula (1); cor in formula (7) is a formula for calculating the Pearson correlation coefficient, x is the original signal, xiIs the current IMF component.
Further, in step 1.4, the recursive feature elimination algorithm specifically includes the steps of: each operation is to abandon one feature F in the l features, then compare the classification effect of the l-1 features after the features are abandoned with the original classification effect, calculate the feature importance of the feature F, and the feature with the largest change of the classification effect is the feature with the highest feature importance. After l calculations are performed, the k features of highest importance are selected.
The invention has the beneficial effects that: the method starts from the characteristics of the vibration signals of the steam turbine equipment, and firstly decomposes the original vibration signals by adopting the integrated empirical mode, thereby overcoming the mode aliasing problem brought by the traditional empirical mode decomposition in the process of obtaining IMF components. Secondly, the method considers the non-stationarity characteristic of the IMF component, divides the IMF component into a stationary part and a non-stationary part, and then respectively selects the characteristics of the two parts, thereby more fully selecting the key characteristics related to the fault. And finally, establishing a fault diagnosis model based on the key characteristics, avoiding the interference of irrelevant information, improving the performance of online fault diagnosis, being beneficial to accurately positioning and repairing faults by maintenance personnel of a power plant, ensuring the safe and reliable operation of steam turbine equipment and even power plant production and indicating a new direction for the research of a fault diagnosis method based on vibration signals in a large coal-fired unit.
Drawings
FIG. 1 is a flow chart of a method for diagnosing a fault of a steam turbine main unit based on smooth and non-smooth vibration signal feature selection according to the present invention.
FIG. 2 is a graph comparing normal signals with air flow excitation and dynamic and static rub faults in the examples.
FIG. 3 is a flow diagram of an integrated empirical mode decomposition.
Fig. 4 is a graph of the effect of the vibration signal integration after empirical mode decomposition in an example.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific examples.
Vibration signals of a steam turbine set host in thermal power generation have the characteristics of nonlinearity and non-stationarity, and an original signal contains a large amount of noise, so that it is difficult to directly extract information from the original signal. The invention takes two typical faults of air flow excitation fault and dynamic and static rub fault generated by a steam turbine set host of a certain thermal power plant as examples, and as shown in figure 1, the method of the invention is explained in detail. As shown in fig. 2, the airflow excitation fault is embodied as that the vibration of the shaft system is increased, and the low-frequency component in the signal is increased; the dynamic and static rub faults are embodied as disappearance of the 'peak' of the vibration waveform and obvious clipping phenomenon. In addition, the sampling interval of the signal in the example is 0.15 ms. The invention relates to a steam turbine set fault diagnosis based on steady and non-steady vibration signal feature extraction, which comprises the following specific steps:
step 1: the method comprises the following steps of collecting vibration signals of a steam turbine host under a normal state and different fault states, and performing model training as original vibration signals:
step 1.1, performing integrated empirical mode decomposition on the original vibration signal to decompose the original vibration signal into n IMF components. The flowchart of the integrated empirical mode decomposition is shown in fig. 3, and the specific operation steps are as follows:
(1.1.1) adding a white noise sequence in normal distribution into the original signal, and decomposing the signal added with the white noise into n IMF components through empirical mode decomposition.
(1.1.2) repeating step (1.1.1) K-1 times, each time adding a new white noise sequence. And n IMF components are obtained:where i ∈ 1,2, … n, denotes the ith component, and j denotes the jth execution of modal decomposition.
(1.1.3) the IMF component finally obtained is I1,I2,I3,…Ii,…InWhereinK represents the number of execution times of modal decomposition.
Wherein the empirical mode decomposition is performed as follows
(a) Finding out all maximum value points and minimum value points of original vibration signal x (t), connecting the maximum value points and minimum value points with cubic spline curves to form an upper envelope line and a lower envelope line, wherein original data is contained between the upper envelope line and the lower envelope line, and calculating a mean value line m of the upper envelope line and the lower envelope line1Calculating a sequence of difference values of the signals as h1=x(t)-m1。
(b) Judgment h1Whether two conditions for the eigenmode components are met: (1) the number of the extreme points is the same as that of the zero-crossing points or only differs by one at most. (2) The mean of the two envelopes of the local maxima and local minima of the difference sequence is zero at any point. If h is1If the two conditions are not satisfied, repeating the step (a) and the step (h)1As the original sequence until h satisfying the eigenmode component is obtained1kUntil now.
(c) The first eigenmode component is denoted c1=h1kThe remainder r obtained1=x(t)-c1R is to1Repeating the steps (a) to (b) as new original data until the nth remainder rnIs a monotonic function, and ends up when the IMF component satisfying both conditions in (b) cannot be extracted. Finally obtainingThe original signal is decomposed into the sum of the residual error and n eigenmode components, resulting in n IMF components.
When the empirical mode decomposition is integrated, the variance of white noise and the number of the empirical mode decomposition are required to be set. The variance of the white noise is generally selected from 0.2-0.6, and the frequency of empirical mode decomposition by adding the white noise is generally 50-200. The integrated empirical mode decomposition in this example results in 9 IMF components.
And 1.2, carrying out stationarity judgment on the IMF component of n, and dividing the IMF component into p stationary IMF components and q non-stationary IMF components.
The method specifically comprises the step of carrying out stationarity judgment on all IMF components by applying AugmentedDicockeyFuller (ADF) test, so that the IMF components are divided into two parts, namely p stationary IMF components and q non-stationary IMF components. The stationary signal is defined as a signal whose distribution parameter or distribution rule does not change with time, that is, a signal whose statistical characteristic does not change with time; a non-stationary signal is defined as a signal whose distribution parameter or distribution law changes with time, i.e. a signal whose statistical properties change with time. As shown in fig. 4, in the example, 9 IMF components obtained by decomposing the original signal are subjected to stationarity discrimination, the first 4 IMF components are stationary, and the last 5 IMF components are non-stationary.
And 1.3, respectively calculating the original statistical characteristics of the stable IMF component and the non-stable IMF component. The original statistical characteristics include: RMS, Kurt, skewness Skaw, Peak-Peak value Peak-Peak, coefficient of Peak value Cf, slow characteristic Slowness and correlation coefficient Corcoef, wherein the slow characteristic Slowness is as follows:
n is the number of sampling points, Δ xiAs in signal xThe difference of (2).
In addition, RMS is a root mean square value, representing the average power of a signal; kurt is kurtosis and reflects the sharpness of the peak of the signal; skaw is skewness and reflects the skewness direction and degree of the signal; Peak-Peak is a Peak-Peak value and reflects the size of a signal fluctuation range; cf is the peak coefficient, which is the ratio of the peak value to the effective value; slowness reflects the speed of signal change; corcoef is the correlation coefficient. The specific formula is calculated as follows:
Peak_peak=xmax-xmin (4)
Corcoef=cor(x,xi) (7)
wherein, x in the formula (1)iFor the amplitude of the ith vibration signal in signal x, x ═ x1,x2,…,xN}; μ in equation (2) is the mean value of the signal x,σ is the standard deviation of the signal x; e represents a desired function; x in formula (4)maxIs the maximum value of the vibration signal, xminIs the minimum value of the vibration signal; rms in the formula (5) is the root mean square value calculated in the formula (1); cor in formula (7) is a formula for calculating the Pearson correlation coefficient, x is the original signal, xiIs the current IMF component.
The examples have 4 stationary IMF components and 5 non-stationary IMF components. Therefore, 28 original statistical features of the stationary IMF component and 35 original statistical features of the non-stationary IMF component are obtained.
Step 1.4, aiming at the original statistical characteristics of each state (the state comprises a normal state and different fault states), respectively applying a recursive characteristic elimination algorithm to the stationary signal xsAnd non-stationary signal xtSelecting k by selecting characteristicssKey characteristic sum k of stationary signaltA key feature of non-stationary signals. The recursive feature elimination algorithm comprises the following specific steps: each operation is to abandon one feature F in the l features, then compare the classification effect of the l-1 features after the features are abandoned with the original classification effect, calculate the feature importance of the feature F, and the feature with the largest change of the classification effect is the feature with the highest feature importance. After l calculations are performed, the k features of highest importance are selected.
The number of stationary part key features obtained in this example is 10, and the number of non-stationary part key features is 8.
Table 1 example stationary part key feature selection.
Table 2 example non-stationary part key feature selection.
Through the recursive feature elimination algorithm, the key features of the stationary part and the non-stationary part are different. Key features of the stationary part are RMS, Slowness, and Corcoef; the key feature of the non-stationary part is Peak-Peak, RMS. In addition, the distribution of the key features of the stationary part is uniform, each IMF component can extract the key features, and the IMF8 component and the IMF9 component in the non-stationary part contain residual signals of the integrated empirical mode decomposition and basically have no fault key information.
Step 1.5, establishing a fault diagnosis model: k selected from (1.4)sKey characteristic sum k of stationary signaltAnd inputting the key characteristics of the non-stationary signals into a gradient lifting tree classifier for training to obtain a fault diagnosis model.
Step 2, collecting vibration signals of a main machine of the steam turbine, and selecting k with highest feature importance according to the steps 1.1-1.4sKey characteristic sum k of stationary signaltAnd inputting the key characteristics of the non-stationary signals into a fault diagnosis model for fault diagnosis.
Table 3 comparison of the effect of the present invention with the previous algorithm.
Compared with the prior method, the method improves the accuracy and reliability of online fault diagnosis. By carrying out integrated empirical mode decomposition on the signal to obtain IMF components and calculating all characteristics of all the IMF components, compared with the classification effect of carrying out empirical mode decomposition on the signal and extracting energy entropy characteristics of the IMF components, the following characteristics can be seen: firstly, the integrated empirical mode decomposition reduces aliasing of the modes compared with the empirical mode decomposition; and secondly, a plurality of statistical characteristics are calculated for the IMF components, so that the fault information can be extracted more fully. By directly carrying out recursive feature elimination on the features of all IMF components without distinguishing stable and non-stable parts and extracting key features, and comparing with the retained classification effect of all features, the following steps can be seen: the recursive feature elimination algorithm reduces the feature dimension, simultaneously retains the key features, removes the interference of redundant data and further improves the accuracy of fault diagnosis. By respectively extracting key features aiming at the stationary part and the non-stationary part and comparing the key features with the feature extraction of all features without distinguishing stationarity, it can be seen that the signals are distinguished in stationarity, and more comprehensive fault information can be extracted. Non-stationary parts are often ignored because their contribution to the discrimination of faults is always less large than that of stationary parts. But the ignored parts often contain some critical information. Therefore, IMF components are subjected to stationarity distinguishing, and key information which is easy to ignore is extracted. Generally, the method adopts an integrated empirical mode to decompose the original vibration signal aiming at the main equipment of the steam turbine main engine for coal-fired power generation, thereby solving the problem that the effective characteristics of the original vibration signal are difficult to extract; secondly, the non-stationarity characteristic of the IMF component is considered, the IMF component is creatively divided into a stationary part and a non-stationary part, then the extraction of the characteristics and the selection of key characteristics are carried out, the fault information can be fully extracted, the key information can be kept, and the redundant information is eliminated, so that the fault diagnosis accuracy is remarkably improved, and the fault diagnosis can be accurately and rapidly carried out by power plant maintenance personnel.
Claims (5)
1. A steam turbine main machine fault diagnosis method based on steady and non-steady vibration signal feature selection is characterized by comprising the following steps:
(1) the method comprises the following steps of collecting vibration signals of a steam turbine host under a normal state and different fault states, and performing model training as original vibration signals:
(1.1) carrying out integrated empirical mode decomposition on the original vibration signal, and decomposing the original vibration signal into n IMF components;
(1.2) carrying out stationarity judgment on the IMF component of n, and dividing the IMF component into two parts, namely p stationary IMF components and q non-stationary IMF components;
(1.3) respectively calculating original statistical characteristics of the stable IMF component and the non-stable IMF component; the original statistical characteristics include: RMS, Kurt, skewness Skaw, Peak-Peak value Peak-Peak, coefficient of Peak value Cf, slow characteristic Slowness and correlation coefficient Corcoef, wherein the slow characteristic Slowness is as follows:
n is the number of sampling points, Δ xiAs in signal xA difference of (a);
the number of the stable IMF component original statistical features is 7 multiplied by p, and the number of the non-stable IMF component original statistical features is 7 multiplied by q;
(1.4) applying recursive feature elimination algorithm to the stationary signal x respectively according to the original statistical features in each statesAnd non-stationary signal xtSelecting k by selecting characteristicssKey characteristic sum k of stationary signaltKey features of individual non-stationary signals;
(1.5) establishing a fault diagnosis model: k selected from (1.4)sKey characteristic sum k of stationary signaltInputting key characteristics of the non-stationary signals into a gradient lifting tree classifier for training to obtain a fault diagnosis model;
(2) collecting vibration signals of a main machine of the steam turbine, and selecting k with highest feature importance according to the steps (1.1) - (1.4)sKey characteristic sum k of stationary signaltAnd inputting the key characteristics of the non-stationary signals into a fault diagnosis model for fault diagnosis.
2. The method for diagnosing the fault of the main engine of the steam turbine based on the selection of the characteristics of the steady and non-steady vibration signals as claimed in claim 1, wherein in the step (1.1), the specific method for integrating the empirical mode decomposition is as follows:
(1.1.1) adding a white noise sequence in normal distribution into an original signal, and decomposing the signal added with the white noise into n IMF components through empirical mode decomposition;
(1.1.2) repeating the step (1.1.1) K-1 times, adding a new white noise sequence each time, and obtaining n IMF components:wherein i ∈ 1,2, … n, denotes the ith component, and j denotes the jth execution of modal decomposition;
(1.1.3) the IMF component finally obtained is I1,I2,I3,…Ii,…InWherein, in the step (A),k represents the execution times of modal decomposition;
the empirical mode decomposition comprises the following steps:
(a) finding out all maximum value points and minimum value points of original vibration signal x (t), connecting the maximum value points and minimum value points with cubic spline curves to form an upper envelope line and a lower envelope line, wherein original data is contained between the upper envelope line and the lower envelope line, and calculating a mean value line m of the upper envelope line and the lower envelope line1Calculating a sequence of difference values of the signals as h1=x(t)-m1;
(b) Judgment h1Whether two conditions for the eigenmode components are met: (1) the number of the extreme points is the same as that of the zero crossing points or only differs by one at most; (2) the mean value of two envelope lines of the local maximum value and the local minimum value of the difference value sequence is zero at any point; if h is1If the two conditions are not satisfied, repeating the step (a) and the step (h)1As the original sequence until h satisfying the eigenmode component is obtained1kUntil the end;
(c) the first eigenmode component is denoted c1=h1kThe remainder r obtained1=x(t)-c1R is to1Repeating the steps (a) to (b) as new original data until the nth remainder rnThe IMF component is a monotonous function, namely the IMF component meeting the two conditions in the step (b) can not be extracted; finally obtainingThe original signal is decomposed into the sum of the residual error and n eigenmode components, resulting in n IMF components.
3. The method for diagnosing the main machine fault of the steam turbine based on the selection of the stationary and non-stationary vibration signal characteristics as claimed in claim 1, wherein in the step (1.2), the stationarity of all the IMF components is determined by applying augmented dickeyfuller (adf) test to determine the stationarity of all the IMF components, so as to divide the IMF components into p stationary IMF components and q non-stationary IMF components.
4. The method for diagnosing the main machine fault of the steam turbine based on the selection of the characteristics of the steady and non-steady vibration signals as claimed in claim 1, wherein in the step (1.3), RMS is a root mean square value and represents the average power of the signals; kurt is kurtosis and reflects the sharpness of the peak of the signal; skaw is skewness and reflects the skewness direction and degree of the signal; Peak-Peak is a Peak-Peak value and reflects the size of a signal fluctuation range; cf is the peak coefficient, which is the ratio of the peak value to the effective value; slowness reflects the speed of signal change; corcoef is a correlation coefficient; the specific formula is calculated as follows:
Peak_peak=xmax-xmin (4)
Corcoef=cor(x,xi) (6)
wherein, x in the formula (1)iFor the amplitude of the ith vibration signal in signal x, x ═ x1,x2,…,xN}; μ in equation (2) is the mean value of the signal x,σ is the standard deviation of the signal x; e represents a desired function; x in formula (4)maxIs the maximum value of the vibration signal, xminIs the minimum value of the vibration signal;rms in the formula (5) is the root mean square value calculated in the formula (1); cor in formula (6) is a formula for calculating the Pearson correlation coefficient, x is the original signal, xiIs the current IMF component.
5. The method for diagnosing the fault of the main engine of the steam turbine based on the selection of the characteristics of the steady and non-steady vibration signals as claimed in claim 1, wherein in the step (1.4), the recursive characteristic elimination algorithm comprises the following specific steps: each operation is to abandon one feature F in the l features, then compare the classification effect of the l-1 features after abandoning the features with the original classification effect, calculate the feature importance of the feature F, the feature with the largest change of the classification effect is the feature with the highest feature importance; after l calculations are performed, the k features of highest importance are selected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810892236.9A CN109297713B (en) | 2018-08-07 | 2018-08-07 | Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810892236.9A CN109297713B (en) | 2018-08-07 | 2018-08-07 | Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109297713A CN109297713A (en) | 2019-02-01 |
CN109297713B true CN109297713B (en) | 2019-12-31 |
Family
ID=65168040
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810892236.9A Active CN109297713B (en) | 2018-08-07 | 2018-08-07 | Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109297713B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110702394B (en) * | 2019-10-18 | 2021-06-08 | 西安热工研究院有限公司 | Vibration change characteristic-based vibration fault diagnosis method for steam turbine generator unit |
CN110817636B (en) * | 2019-11-20 | 2021-09-21 | 上海电气集团股份有限公司 | Elevator door system fault diagnosis method, device, medium and equipment |
CN110928237B (en) * | 2019-12-20 | 2020-12-29 | 华中科技大学 | Vibration signal-based numerical control machining center flutter online identification method |
CN111784068A (en) * | 2020-07-09 | 2020-10-16 | 北京理工大学 | EEMD-based power load combined prediction method and device |
CN111929044B (en) * | 2020-07-15 | 2023-08-08 | 西门子工厂自动化工程有限公司 | Method, apparatus, computing device and storage medium for monitoring device status |
CN112183439A (en) * | 2020-10-13 | 2021-01-05 | 上海明略人工智能(集团)有限公司 | Signal feature extraction method, system, computer-readable storage medium and electronic device |
CN112327701B (en) * | 2020-11-09 | 2021-11-02 | 浙江大学 | Slow characteristic network monitoring method for nonlinear dynamic industrial process |
WO2022261805A1 (en) * | 2021-06-15 | 2022-12-22 | 大连理工大学 | Diesel engine gearbox fault diagnosis method |
CN113592379B (en) * | 2021-06-25 | 2024-05-14 | 南京财经大学 | Key feature identification method for detecting anomaly of bulk grain container logistics transportation environment |
US20230093741A1 (en) * | 2021-09-17 | 2023-03-23 | Triad National Security, Llc | Signal processing methods and apparatus |
CN114548151B (en) * | 2022-01-12 | 2024-04-16 | 广东海洋大学 | Method, device, medium and equipment for improving fault characteristic signals |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7464006B1 (en) * | 2003-10-09 | 2008-12-09 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Analyzing nonstationary financial time series via hilbert-huang transform (HHT) |
CN106779148B (en) * | 2016-11-14 | 2017-10-13 | 中南大学 | A kind of method for forecasting wind speed of high speed railway line of multi-model multiple features fusion |
-
2018
- 2018-08-07 CN CN201810892236.9A patent/CN109297713B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN109297713A (en) | 2019-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109297713B (en) | Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection | |
Huo et al. | Incipient fault diagnosis of roller bearing using optimized wavelet transform based multi-speed vibration signatures | |
Wang et al. | Fault diagnosis of a rolling bearing using wavelet packet denoising and random forests | |
Jia et al. | GTFE-Net: A gramian time frequency enhancement CNN for bearing fault diagnosis | |
CN109241915B (en) | Intelligent power plant pump fault diagnosis method based on vibration signal stability and non-stationarity judgment and feature discrimination | |
CN109241849B (en) | Feature decomposition selection and fault diagnosis method for main engine of intelligent power plant steam turbine | |
Liu et al. | Wind turbine blade bearing fault diagnosis under fluctuating speed operations via Bayesian augmented Lagrangian analysis | |
Zhao et al. | Data augmentation via randomized wavelet expansion and its application in few-shot fault diagnosis of aviation hydraulic pumps | |
CN111238843B (en) | Fan health evaluation method based on rapid spectrum kurtosis analysis | |
CN112229633A (en) | Fan bearing fault diagnosis method based on multivariate feature fusion | |
CN106198020A (en) | Wind turbines bearing failure diagnosis method based on subspace and fuzzy C-means clustering | |
CN110346130B (en) | Boring flutter detection method based on empirical mode decomposition and time-frequency multi-feature | |
CN112163472A (en) | Rolling bearing diagnosis method based on multi-view feature fusion | |
CN109655266A (en) | A kind of Wind turbines Method for Bearing Fault Diagnosis based on AVMD and spectral coherence analysis | |
Zhu et al. | Complex disturbances identification: A novel PQDs decomposition and modeling method | |
Chen et al. | A visualized classification method via t-distributed stochastic neighbor embedding and various diagnostic parameters for planetary gearbox fault identification from raw mechanical data | |
CN113390631A (en) | Fault diagnosis method for gearbox of diesel engine | |
CN113255458A (en) | Bearing fault diagnosis method based on multi-view associated feature learning | |
CN111639852B (en) | Real-time evaluation method and system for vibration state of hydroelectric generating set based on wavelet singular value | |
CN110701087A (en) | Axial flow compressor pneumatic instability detection method based on single-classification overrun learning machine | |
Lu et al. | Bearing fault diagnosis based on clustering and sparse representation in frequency domain | |
CN109297735B (en) | Vibration signal fault diagnosis method for intelligent power plant coal mill | |
CN112734001A (en) | Wind power transmission chain intelligent fault diagnosis method based on order spectrum migration | |
CN116335892A (en) | Wind turbine generator blade abnormality detection method and system | |
CN113822565B (en) | Method for graded and refined analysis of time-frequency characteristics of fan monitoring data |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |