CN113029616A - Compressor surge early fault feature extraction method based on enhanced entropy weight - Google Patents
Compressor surge early fault feature extraction method based on enhanced entropy weight Download PDFInfo
- Publication number
- CN113029616A CN113029616A CN202110106238.2A CN202110106238A CN113029616A CN 113029616 A CN113029616 A CN 113029616A CN 202110106238 A CN202110106238 A CN 202110106238A CN 113029616 A CN113029616 A CN 113029616A
- Authority
- CN
- China
- Prior art keywords
- surge
- fault
- characteristic
- early
- characteristic parameters
- 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.)
- Granted
Links
Images
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
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/005—Testing of complete machines, e.g. washing-machines or mobile phones
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
Abstract
The invention discloses a method for extracting early surge fault characteristics of a compressor based on enhanced entropy weight dimensionality reduction, which belongs to the technical field of fault diagnosis of mechanical equipment and can more accurately and effectively extract early surge fault sensitive characteristics. The method comprises the following steps: and collecting m groups of normal vibration data samples of the compressor to be monitored and fault vibration data samples of early surge faults. Surge early failure refers to rotating stall failure. And respectively extracting n time domain and frequency domain characteristic parameters from the normal vibration data sample and the fault vibration data sample, and constructing a sample characteristic matrix. And carrying out normalization processing on the sample characteristic matrix, and calculating enhancement information entropy values of the n characteristic parameters. And measuring weight coefficients representing the surge early fault capability of the n characteristic parameters based on the enhanced information entropy values of the characteristic parameters. And extracting characteristic parameters of which the weight coefficients exceed a set threshold value according to the magnitude relation of the weight coefficients, and using the characteristic parameters as sensitive characteristics of the early-stage faults of the surging to be monitored.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of mechanical equipment, in particular to a compressor surge early fault feature extraction method based on an enhanced entropy weight.
Background
The compressor is widely applied to the fields of petrochemical industry, smelting, machinery, power and the like. In the actual operation process, the working condition is easy to deviate from the normal working condition due to factors such as improper regulation of the working condition, environmental change and the like, and a surge fault occurs. Once surge fault occurs, important parts such as a main shaft, an impeller, a bearing, a seal and the like of the compressor are often damaged, and the safe and stable operation of equipment is seriously threatened. Therefore, how to quickly and accurately identify the surge fault becomes a key factor for maintaining the stable operation of the unit and guaranteeing the continuous production of enterprises.
On the basis of a surge fault mechanism, field personnel often judge according to abnormal sounds emitted in the operation process of equipment and the intensity of fluctuation of characteristic parameters such as observed temperature, pressure and the like, but the empirical subjective judgment method has high false alarm rate and alarm leakage rate due to the fact that the method is difficult to quantify. Although the alarm threshold value can be learned by itself by means of machine learning methods and historical data features, surge fault identification based on a single characteristic parameter still has the following two problems: 1) when a surge fault occurs, various fault characteristics are often accompanied, and the dominant fault characteristics are different in different application occasions, so that the detection method formed based on a single characteristic parameter has poor adaptability, and the problem of reduced identification precision caused by the difference of equipment and operation conditions is easy to occur; 2) most of the characteristic parameter values are obviously changed when the surge fault develops to the middle and later stages, and are suitable for fault identification or interlocking protection at the middle and later stages, and the characteristics are weak in the initial stage of the surge fault, namely the rotating stall fault, so that the characteristics are difficult to be used for early identification of the surge fault. Therefore, it is necessary to develop multidimensional characteristic parameter fusion analysis to study the early sensitive characteristics of the compressor surge fault.
The multi-dimensional characteristic parameters have fault information overlapping, so that the provided information redundancy is caused, the fault identification accuracy cannot be improved, the algorithm complexity is increased, and the problem that the characteristics cannot be visualized is caused. The characteristic selection method commonly used at present can exert a relatively ideal effect under the condition that the characteristics of different types of data are obviously different, but the characteristics of a surge signal are weak in the early stage and are interfered by vibration noise, so that the distinguishing degree of characteristic parameter values between a normal state and a fault state is extremely low, and the characteristic selection method is difficult to accurately select sensitive characteristics and cannot be used for extracting the characteristics of the early stage of the surge fault.
Therefore, at present, the research on the early-stage fault characteristics of compressor surge still has certain defects, and the problem of how to effectively extract the fault characteristics which accurately and timely reflect the surge state of the compressor needs to be solved urgently.
Disclosure of Invention
In view of the above, the invention provides a method for extracting characteristics of early-stage faults of compressor surge based on enhanced entropy weight dimensionality reduction, which can adopt the enhanced entropy weight to perform optimization and screening on high-dimensional complete characteristics of a time domain and a frequency domain, and more accurately and effectively extract characteristics of early-stage faults of surge. .
In order to achieve the purpose, the technical scheme of the invention comprises the following steps:
and S1, collecting m groups of normal vibration data samples of the compressor to be monitored and fault vibration data samples of the early surge fault. Surge early failure refers to rotating stall failure.
And S2, extracting n time domain and frequency domain characteristic parameters from the normal vibration data sample and the fault vibration data sample respectively, and constructing a sample characteristic matrix.
And S3, carrying out normalization processing on the sample characteristic matrix, and calculating enhancement information entropy values of the n characteristic parameters.
And S4, measuring weight coefficients of the n characteristic parameters for representing the surge early fault capacity based on the enhanced information entropy values of the characteristic parameters.
And S5, extracting characteristic parameters of which the weight coefficients exceed a set threshold value according to the magnitude relation of the weight coefficients, and using the characteristic parameters as sensitive characteristics of the early-stage faults of the surge to be monitored.
Further, in S1, acquiring m groups of normal vibration data samples of the compressor to be monitored and fault vibration data samples of the surge early fault, including the following steps: acquiring original vibration signals of a compressor to be monitored in a normal working state and an early surge state by using a vibration sensor; and (4) intercepting m groups of normal vibration data and fault vibration data samples of early surge faults from the original vibration signals.
Further, the vibration sensor is used for acquiring an original vibration signal of the compressor to be monitored in a normal working state and an early surge state, and the method specifically comprises the following steps: the vibration speed sensor is arranged on a sliding bearing seat of a compressor to be monitored through a bolt, and the sampling frequency f of the vibration speed sensorsSet not lower than the rotation frequency f of the compressor rotor to be monitoredr25.6 times of the total weight of the powder.
Further, m groups of normal vibration data and fault vibration data samples of the surge early fault are intercepted from the original vibration signal, and the method specifically comprises the following steps: setting the length of a truncated sample according to the sampling frequency of the original vibration signal, wherein the length T of the sample is 0.25 times of the sampling frequency; m is not less than 20.
Further, n time domain and frequency domain characteristic parameters are extracted from the normal vibration data sample and the fault vibration data sample respectively, and a sample characteristic matrix is constructed, specifically: extracting effective value, mean value, peak-peak value, kurtosis, form factor, peak factor, margin factor, pulse factor and frequency domain energy E from the normal vibration data sample and the fault vibration data sample respectively1～EL(ii) a Wherein E1～ELThe energy of the 1 st to the L-th frequency domains is sequentially obtained. The frequency domain energy is 0-1/2.56 xfsIn the range of the rotor rotation frequency frIs the spectral energy of the sub-band. Wherein the total number L of frequency domain energies is:
the energy of the first frequency domain is
Wherein L1, 2.. said., L; x (f) is a spectral vector of the sample
f is a frequency variable; x (t) is a data vector of samples, t is a time variable; t is the sample length. Setting n to be not less than 19; and respectively extracting n characteristic parameters from the normal vibration data sample and the fault vibration data sample, and constructing a sample characteristic matrix R.
Further, in S3, the sample feature matrix is normalized, specifically: the parameter of the ith row and the jth column in the sample characteristic matrix is Rij(ii) a Wherein j is the jth characteristic parameter j ═ 1, 2.. times, n; i is the ith characteristic value, i is 1, 2. . According to the j-th characteristic parameter, the minimum value in m characteristic values is RjminAnd maximum value of RjmaxAnd carrying out normalization processing on the sample feature matrix to obtain a normalized feature matrix G. The parameter of the ith row and the jth column in the normalized feature matrix G is Gij：
Further, in S3, calculating enhancement information entropy values of the n feature parameters, specifically: based on the normalized feature matrix G, n feature parameters are calculated, and the enhanced information entropy value H is equal to { H ═ H }1,H2,…,Hn}，H1,H2,…,HnRespectively are the enhancement information entropy values of the 1 st to the nth characteristic parameters.
Wherein P isijIn the form of a probability matrix, the probability matrix,
u is an enhanced probability matrix
Further, in S4, based on the enhancement information entropy of the characteristic parameter, the weight coefficients of the n characteristic parameters, which characterize the surge early failure capability, are measured, specifically: the weighting coefficient of the n characteristic parameters for representing the surge early failure capability is W ═ W1,W2,…,Wn}；W1,W2,…,WnThe weighting coefficients of the 1 st to nth characteristic parameters are respectively;
then
Wherein C isjIs the entropy weight coefficient of the jth characteristic parameter,
further, in S5, according to the magnitude relationship of the weight coefficient, extracting a characteristic parameter of which the weight coefficient exceeds a set threshold as a sensitive characteristic of the surge early fault to be monitored, specifically: and sequencing the weights of the n characteristic parameters according to the sizes, reserving the first k characteristic parameters as sensitive characteristics of the early-stage fault of the surge to be monitored, and determining the value of k according to application requirements.
Further, the value of k is determined according to the application requirement, and specifically includes: and if the extracted characteristic parameters are used for state monitoring, k is 1, namely, one characteristic parameter which is most sensitive to surge faults is used for monitoring the working state of the compressor to be monitored in real time. And if the extracted characteristic parameters are used for identifying whether the compressor to be monitored enters a surge state, the k value is between 5 and 10.
Has the advantages that:
1. the invention provides a method for extracting characteristics of early surge faults of a compressor based on an enhanced entropy weight, which extracts high-dimensional complete characteristics in a time domain and a frequency domain from normal and surge vibration data of the compressor, considers weak characteristics of early faults and noise interference, and enhances an information entropy calculation formula to obtain the enhanced entropy weight of each characteristic. And then measuring the enhanced entropy weights of the characteristic parameters of the time domain and the frequency domain by an entropy weight method to obtain objective weights. And extracting the sensitive characteristics of the early-stage surge faults based on the weight according to the characteristic that the weight value can represent the sensitivity degree of the characteristics to the surge faults. The method enhances the early weak fault characteristics in the compressor vibration signal, inhibits noise interference, measures the sensitivity of multidimensional characteristic parameters to the surge early fault, and further guides the extraction of the early fault characteristics.
2. The method for extracting the early surge fault characteristics of the compressor based on the enhanced entropy weight can effectively extract the early surge fault sensitive characteristics, can be used for monitoring the state of the on-site compressor and diagnosing the surge fault in real time, solves the problem that the surge fault of the compressor is difficult to accurately and timely identify in engineering application, is beneficial to maintaining stable operation of a unit, and ensures continuous production of enterprises.
Drawings
FIG. 1 is a schematic diagram of a compressor surge early fault feature extraction process based on an enhanced entropy weight in the method provided by the present invention;
FIG. 2 is a graph of a speed signal for a compressor in a normal operating condition in accordance with an embodiment of the present invention;
FIG. 3 is a graph of a speed signal for a compressor in surge condition in accordance with an embodiment of the present invention;
fig. 4 is a visual illustration of the early failure feature of the compressor extracted in the embodiment of the present invention.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
The invention provides a method for extracting characteristics of early surge faults of a compressor based on an enhanced entropy weight, which comprises the following steps of:
s1, collecting m groups of normal vibration data samples of the compressor to be monitored and fault vibration data samples of the early surge fault
An early surge fault refers to a rotating stall fault, and in particular to the type of fault where the compressor does not roar as much as the typical sharp airflow without surge, but its outlet pressure fluctuates.
In the embodiment of the invention, the vibration sensor is used for acquiring the original vibration signal of the compressor to be monitored in a normal working state and an early surge state;
the vibration speed sensor is arranged on a sliding bearing seat of a compressor to be monitored through a bolt, and the sampling frequency f of the vibration speed sensorsSet not lower than the rotation frequency f of the compressor rotor to be monitoredr25.6 times of the total weight of the powder.
And (4) intercepting m groups of normal vibration data and fault vibration data samples of early surge faults from the original vibration signals. In the embodiment of the invention, the length of a truncated sample is set according to the sampling frequency of an original vibration signal, and the length T of the sample is 0.25 times of the sampling frequency; m is not less than 20.
And S2, extracting n time domain and frequency domain characteristic parameters from the normal vibration data sample and the fault vibration data sample respectively, and constructing a sample characteristic matrix.
In the embodiment of the invention, effective values, mean values, peak-to-peak values, kurtosis, form factors, peak factors, margin factors, pulse factors and frequency domain energy E are respectively extracted from normal vibration data samples and fault vibration data samples1～EL(ii) a Wherein E1～ELThe energy of the 1 st to the L-th frequency domains is sequentially obtained.
The frequency domain energy is 0-1/2.56 xfsIn the range of the rotor rotation frequency frIs the spectral energy of the sub-band.
Wherein the total number L of frequency domain energies is:
the energy of the first frequency domain is
Wherein L1, 2.. said., L; x (f) is a spectral vector of the sample
f is a frequency variable; x (t) is a data vector of samples, t is a time variable; t is the sample length.
Setting n to be not less than 19; and respectively extracting n characteristic parameters from the normal vibration data sample and the fault vibration data sample, and constructing a sample characteristic matrix R.
S3, carrying out normalization processing on the sample characteristic matrix, and calculating enhancement information entropy values of n characteristic parameters; the method comprises the following steps:
the parameter of the ith row and the jth column in the sample characteristic matrix is Rij(ii) a Wherein j is the jth characteristic parameter j ═ 1, 2.. times, n; i is the ith characteristic value, i is 1, 2. .
According to the j-th characteristic parameter, the minimum value in m characteristic values is RjminAnd maximum value of RjmaxCarrying out normalization processing on the sample feature matrix to obtain a normalized feature matrix G;
the parameter of the ith row and the jth column in the normalized feature matrix G is Gij：
Based on the normalized feature matrix G, n feature parameters are calculated, and the enhanced information entropy value H is equal to { H ═ H }1,H2,…,Hn}，H1,H2,…,HnRespectively are the enhancement information entropy values of the 1 st to the nth characteristic parameters.
Wherein P isijIn the form of a probability matrix, the probability matrix,
u is an enhanced probability matrix
S4, weighing weight coefficients representing surge early fault capacity of the n characteristic parameters based on the enhanced information entropy values of the characteristic parameters; the method comprises the following steps:
the weighting coefficient of the n characteristic parameters for representing the surge early failure capability is W ═ W1,W2,…,Wn}；W1,W2,…,WnThe weighting coefficients of the 1 st to nth characteristic parameters are respectively;
then
Wherein C isjIs the entropy weight coefficient of the jth characteristic parameter,
and S5, extracting characteristic parameters of which the weight coefficients exceed a set threshold value according to the magnitude relation of the weight coefficients, and using the characteristic parameters as sensitive characteristics of the early-stage faults of the surge to be monitored. The method comprises the following steps:
and sequencing the weights of the n characteristic parameters according to the sizes, reserving the first k characteristic parameters as sensitive characteristics of the early-stage fault of the surge to be monitored, and determining the value of k according to application requirements.
In the embodiment of the invention, if the extracted characteristic parameters are used for monitoring the state, k is 1, namely, the characteristic parameter which is most sensitive to surge faults is used for monitoring the working state of the compressor to be monitored in real time;
and if the extracted characteristic parameters are used for identifying whether the compressor to be monitored enters a surge state, k value is between 5 and 10, namely, a plurality of characteristic parameters which are sensitive to surge faults are adopted for fusion analysis, and whether the compressor to be monitored enters the surge state is judged.
The method extracts high-dimensional complete characteristics in a time domain and a frequency domain from normal and surge vibration data of the compressor, and then calculates the enhancement information entropy value of each characteristic. And measuring the enhanced entropy weights of a plurality of time domain and frequency domain characteristic parameters by improving an entropy weight method to obtain objective weights. And according to the characteristic that the weighted value can represent the sensitivity degree of the characteristics to the surge fault, extracting the sensitive characteristics of the surge early fault based on the weight.
Example (b):
this example was carried out on an actual gas turbine-compressor train. A speed sensor is adopted to measure a vibration speed signal of a sliding bearing seat of the compressor, and the sensor is fixed on the bearing seat through a bolt. During the experiment, the rotor was operated at around 3000rpm with a sampling frequency of 1024 Hz.
A flow chart of a compressor surge early fault feature extraction method based on an enhanced entropy weight is shown in figure 1, and the method specifically comprises the following steps:
s1, acquiring original vibration signals of the compressor to be monitored in a normal state and an early surge state by using a vibration sensor; installing a vibration speed sensor on a sliding bearing seat of a compressor to be monitored by using a bolt, and setting a sampling frequency fsNot less than 10 times the frequency of the conversion.
In this embodiment the rotor speed n is 3000rpm, so according to fsThe sampling frequency calculated by more than or equal to 10 Xn/60 cannot be lower than 500Hz, and 1024Hz is taken in the embodiment.
And (4) intercepting m groups of normal and surge early fault vibration data samples from the original signal. In the embodiment of the invention, the length of a truncated sample is set according to the sampling frequency of an original vibration signal, and the length of the sample is 0.5 times of the sampling length (namely the sampling frequency) in 1 s; m is not less than 20.
In this embodiment, the sampling frequency is 1024Hz, so that N is 0.5 xfsThe sample length was calculated to be 512. In this embodiment, m is 120.
S2, extracting n time domain and frequency domain characteristic parameters from normal data and fault sample data respectively, and constructing a high-dimensional sample characteristic matrix Rij. In the embodiment of the invention, the time domain characteristic parameters and the frequency domain characteristic parameters are respectively effective values, mean values, peak-to-peak values, kurtosis, wave form factors, peak value factors, margin factors, pulse factors and frequency domain energy E1～EL. The frequency domain energy is 0-1/2.56 xfsIn the range of (1) with a frequency of rotation frN is set not lower than 19 for the band energy of the sub-band.
Wherein the total number L of frequency domain energies is:
the energy of the first frequency domain is
Wherein L1, 2.. said., L; x (f) is a spectral vector of the sample
f is a frequency variable; x (t) is a data vector of samples, t is a time variable; t is the sample length;
sampling frequency f used in the present embodimentsIs 5120Hz, frequency conversion frIs 50 Hz. Therefore, if L is 40, n is set to 49 as n — L + 9.
Respectively extracting the 49 time domain and frequency domain characteristic parameters from 120 groups of normal and fault sample data, and constructing a high-dimensional sample characteristic matrix Rij，(i＝1,2,..,120；j＝1,2,…,49)。
And S3, carrying out normalization processing on the sample characteristic matrix, and calculating enhancement information entropy values of the characteristic parameters of the n time domains and the frequency domains. The normalization processing in the embodiment of the invention is realized according to the following formula.
In the embodiment of the invention, in order to enhance the early-stage weak fault characteristics of the surge of the compressor and inhibit the interference of noise and the like, the traditional entropy calculation formula is enhanced and processed based on the normalized characteristic matrix GijCalculating enhanced entropy values H ═ H for n features1,H2,…,HnThe specific expression is as follows:
in the embodiment of the present invention, the enhancement information entropy values of 49 features are shown in the following table.
TABLE 1 characteristic parameters and enhancement information entropy values
S4, based on the enhanced information entropy values of the n time domain and frequency domain characteristic parameters, weighing the weight coefficients of the n characteristic characterization surge early fault capability. In the embodiment of the invention, based on 49 characteristic enhanced entropy values, the improved entropy weight method is utilized to measure the capacity weight of each characteristic for representing surge fault, and the weight coefficient of n characteristics is obtained as W ═ W { (W)1,W2,…,Wn}. The modified entropy weight expression is as follows:
in the embodiment of the present invention, the weighting coefficients of 49 features are shown in the following table.
TABLE 2 characteristic parameters and weighting coefficients
In S5, a feature parameter with a large weight coefficient is extracted based on the magnitude relationship of the weight coefficient. In the embodiment of the invention, according to the characteristic that the characteristic with larger weight can better reflect the surge fault, the weights of n characteristics are sorted according to the size, and the first k characteristics, namely the characteristic parameters with higher fault sensitivity, are reserved. The value of k depends on the application requirements.
If the extracted features are used for state monitoring, k can be 1, namely a feature parameter which is most sensitive to surge faults is used for monitoring the working state of the compressor to be monitored in real time;
if the extracted features are used for intelligent identification, k value can be between 5 and 10, and whether the compressor to be monitored enters a surge state or not is judged by adopting fusion analysis of a plurality of feature parameters sensitive to surge faults.
In the embodiment, k is 1 and is used for monitoring the surge state of the compressor in real time. And extracting frequency domain energy which is most sensitive to surge faults and is characterized in that the frequency domain energy is in a frequency range of 350-400 Hz.
The time domain waveforms of the vibration speed signal of the compressor measured in the present embodiment correspond to the normal state and the surge state, respectively, as shown in fig. 2 and 3. The characteristic values of the most sensitive characteristics extracted in the embodiment in the normal state and the surge state are shown in fig. 4, and fig. 4 shows that the method can effectively extract the surge sensitive fault characteristics of the compressor, can be used for accurate and early identification of the surge fault of the compressor, and verifies the effectiveness of the method.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A compressor surge early fault feature extraction method based on an enhanced entropy weight is characterized by comprising the following steps:
s1, collecting normal vibration data samples of m groups of compressors to be monitored and fault vibration data samples of early surge faults;
the surge incipient fault is a rotating stall fault;
s2, extracting n time domain and frequency domain characteristic parameters from the normal vibration data sample and the fault vibration data sample respectively, and constructing a sample characteristic matrix;
s3, carrying out normalization processing on the sample characteristic matrix, and calculating enhancement information entropy values of n characteristic parameters;
s4, weighing weight coefficients representing surge early fault capacity of the n characteristic parameters based on the enhanced information entropy values of the characteristic parameters;
and S5, extracting characteristic parameters of which the weight coefficients exceed a set threshold value according to the magnitude relation of the weight coefficients, and using the characteristic parameters as sensitive characteristics of the early-stage faults of the surge to be monitored.
2. The method according to claim 1, wherein in the step S1, the method for collecting m groups of normal vibration data samples of the compressor to be monitored and fault vibration data samples of the surge early fault comprises the following steps:
acquiring original vibration signals of a compressor to be monitored in a normal working state and an early surge state by using a vibration sensor;
and intercepting m groups of normal vibration data and fault vibration data samples of early surge faults from the original vibration signal.
3. The method according to claim 2, characterized in that the vibration sensor is used to acquire the original vibration signals of the compressor to be monitored in the normal operating state and in the early surge state, specifically:
the vibration speed sensor is arranged on a sliding bearing seat of the compressor to be monitored through a bolt, and the sampling frequency f of the vibration speed sensorsIs set not lower thanThe rotational frequency f of the compressor rotor to be monitoredr25.6 times of the total weight of the powder.
4. The method according to claim 2 or 3, wherein said extracting m sets of normal vibration data and fault vibration data samples of surge early fault from said raw vibration signal comprises:
setting a truncation sample length according to the sampling frequency of the original vibration signal, wherein the sample length T is 0.25 times of the sampling frequency; m is not less than 20.
5. The method according to claim 1,2 or 3, wherein the n time-domain and frequency-domain characteristic parameters are extracted from the normal vibration data sample and the fault vibration data sample, respectively, to construct a sample characteristic matrix, specifically:
extracting effective value, mean value, peak-to-peak value, kurtosis, form factor, peak factor, margin factor, pulse factor and frequency domain energy E from the normal vibration data sample and the fault vibration data sample respectively1～EL(ii) a Wherein E1～ELSequentially 1 st to Lth frequency domain energy;
the frequency domain energy is 0-1/2.56 xfsIn the range of said rotor rotation frequency frIs the spectral energy of the sub-band;
wherein the total number L of frequency domain energies is:
the energy of the first frequency domain is
Wherein L1, 2.. said., L; x (f) is a spectral vector of the sample
f is a frequency variable; x (t) is a data vector of samples, t is a time variable; t is the sample length;
setting n to be not less than 19; and extracting the n characteristic parameters from the normal vibration data sample and the fault vibration data sample respectively to construct a sample characteristic matrix R.
6. The method according to claim 1,2, 3 or 5, wherein in S3, the sample feature matrix is normalized, specifically:
the parameter of the ith row and the jth column in the sample characteristic matrix is Rij(ii) a Wherein j is the jth characteristic parameter j ═ 1, 2.. times, n; i is the ith characteristic value, i is 1, 2. (ii) a
According to the j-th characteristic parameter, the minimum value in m characteristic values is RjminAnd maximum value of RjmaxNormalizing the sample feature matrix to obtain a normalized feature matrix G;
the parameter of the ith row and the jth column in the normalized feature matrix G is Gij：
7. The method according to claim 6, wherein in S3, the enhanced information entropy values of the n feature parameters are calculated, specifically:
calculating the n feature parameter calculation enhancement information entropy values H ═ { H } based on the normalized feature matrix G1,H2,…,Hn}，H1,H2,…,HnRespectively are the enhancement information entropy values of the 1 st to the nth characteristic parameters.
Wherein P isijIn the form of a probability matrix, the probability matrix,
u is an enhanced probability matrix
8. The method according to claim 7, wherein in S4, based on the enhanced information entropy of the characteristic parameter, the weighting coefficients characterizing the surge early failure capability of the n characteristic parameters are measured, specifically:
the weighting coefficient of the n characteristic parameters for representing the surge early failure capability is W ═ W { (W)1,W2,…,Wn}；W1,W2,…,WnThe weighting coefficients of the 1 st to nth characteristic parameters are respectively;
then
Wherein C isjIs the entropy weight coefficient of the jth characteristic parameter,
9. the method according to claim 1,2, 3, 5, 7 or 8, wherein in S5, the characteristic parameter with the weight coefficient exceeding the set threshold is extracted according to the magnitude relationship of the weight coefficient, and is used as the sensitive characteristic of the surge early fault to be monitored, specifically:
and sequencing the weights of the n characteristic parameters according to the sizes, reserving the first k characteristic parameters as sensitive characteristics of the early-stage faults of the surge to be monitored, and determining the value of k according to application requirements.
10. The method of claim 9, wherein the value of k is determined according to application requirements, and specifically comprises:
if the extracted characteristic parameters are used for state monitoring, k is 1, namely, one characteristic parameter which is most sensitive to surge faults is used for monitoring the working state of the compressor to be monitored in real time;
and if the extracted characteristic parameters are used for identifying whether the compressor to be monitored enters a surge state, the k value is between 5 and 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110106238.2A CN113029616B (en) | 2021-01-26 | 2021-01-26 | Compressor surge early fault feature extraction method based on enhanced entropy weight |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110106238.2A CN113029616B (en) | 2021-01-26 | 2021-01-26 | Compressor surge early fault feature extraction method based on enhanced entropy weight |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113029616A true CN113029616A (en) | 2021-06-25 |
| CN113029616B CN113029616B (en) | 2022-08-02 |
Family
ID=76459798
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110106238.2A Active CN113029616B (en) | 2021-01-26 | 2021-01-26 | Compressor surge early fault feature extraction method based on enhanced entropy weight |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113029616B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114894460A (en) * | 2022-05-10 | 2022-08-12 | 同济大学 | IMU-based method for monitoring and evaluating damage state of anti-seismic support and hanger |
| CN114935357A (en) * | 2022-03-14 | 2022-08-23 | 浙江倍时信息科技有限公司 | Equipment health monitoring system based on entropy value calculation |
| CN115310490A (en) * | 2022-08-17 | 2022-11-08 | 中国核动力研究设计院 | Rotating equipment fault analysis method based on multi-domain feature and sensitive feature selection |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040068387A1 (en) * | 2002-10-04 | 2004-04-08 | Pierino Bonanni | Method and system for detecting precursors to compressor stall and surge |
| CN101713395A (en) * | 2009-11-20 | 2010-05-26 | 西安交通大学 | Surge monitoring method based on incremental nonlinear manifold learning |
| CN104712542A (en) * | 2015-01-12 | 2015-06-17 | 北京博华信智科技股份有限公司 | Reciprocating compressor sensitive characteristic extracting and fault diagnosis method based on internet of things |
| CN106546918A (en) * | 2016-10-27 | 2017-03-29 | 中国大唐集团科学技术研究院有限公司西北分公司 | A kind of method for diagnosing faults of Hydropower Unit |
| CN111307206A (en) * | 2020-02-20 | 2020-06-19 | 北京天泽智云科技有限公司 | Multi-source information fusion-based compressor surge automatic identification method |
| WO2021007084A1 (en) * | 2019-07-08 | 2021-01-14 | Purdue Research Foundation | Compressor stall warning using nonlinear feature extraction algorithms |
-
2021
- 2021-01-26 CN CN202110106238.2A patent/CN113029616B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040068387A1 (en) * | 2002-10-04 | 2004-04-08 | Pierino Bonanni | Method and system for detecting precursors to compressor stall and surge |
| CN101713395A (en) * | 2009-11-20 | 2010-05-26 | 西安交通大学 | Surge monitoring method based on incremental nonlinear manifold learning |
| CN104712542A (en) * | 2015-01-12 | 2015-06-17 | 北京博华信智科技股份有限公司 | Reciprocating compressor sensitive characteristic extracting and fault diagnosis method based on internet of things |
| CN106546918A (en) * | 2016-10-27 | 2017-03-29 | 中国大唐集团科学技术研究院有限公司西北分公司 | A kind of method for diagnosing faults of Hydropower Unit |
| WO2021007084A1 (en) * | 2019-07-08 | 2021-01-14 | Purdue Research Foundation | Compressor stall warning using nonlinear feature extraction algorithms |
| CN111307206A (en) * | 2020-02-20 | 2020-06-19 | 北京天泽智云科技有限公司 | Multi-source information fusion-based compressor surge automatic identification method |
Non-Patent Citations (7)
| Title |
|---|
| LEI ZHANG等: "Simulation of Entropy Generation under Stall Conditions in a Centrifugal Fan", 《ENTROPY》 * |
| 何燕慧等: "频域分析及其在离心式压缩机故障诊断中的应用", 《甘肃科技》 * |
| 吴春志等: "基于多尺度降维的柴油机信号信息熵增强方法", 《振动与冲击》 * |
| 张明等: "基于多源信息融合的往复式压缩机故障诊断方法", 《机械工程学报》 * |
| 谢林等: "基于BP神经网络和证据理论的离心压缩机喘振诊断方法", 《机械设计与制造》 * |
| 贺雅等: "多转子燃气轮机不平衡部位辨识方法研究", 《机械设计与制造》 * |
| 马超等: "基于熵权重和灰色关联度的多级变速器故障诊断", 《组合机床与自动化加工技术》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114935357A (en) * | 2022-03-14 | 2022-08-23 | 浙江倍时信息科技有限公司 | Equipment health monitoring system based on entropy value calculation |
| CN114894460A (en) * | 2022-05-10 | 2022-08-12 | 同济大学 | IMU-based method for monitoring and evaluating damage state of anti-seismic support and hanger |
| CN114894460B (en) * | 2022-05-10 | 2023-03-28 | 同济大学 | IMU-based method for monitoring and evaluating damage state of anti-seismic support and hanger |
| CN115310490A (en) * | 2022-08-17 | 2022-11-08 | 中国核动力研究设计院 | Rotating equipment fault analysis method based on multi-domain feature and sensitive feature selection |
| CN115310490B (en) * | 2022-08-17 | 2024-03-29 | 中国核动力研究设计院 | Rotary equipment fault analysis method based on multi-domain feature and sensitive feature selection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113029616B (en) | 2022-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113029616B (en) | Compressor surge early fault feature extraction method based on enhanced entropy weight | |
| CN107003663B (en) | The monitoring of device with movable part | |
| WO2011125130A1 (en) | Plant diagnosis device, diagnosis method, and diagnosis program | |
| Soylemezoglu et al. | Mahalanobis-Taguchi system as a multi-sensor based decision making prognostics tool for centrifugal pump failures | |
| US8396689B2 (en) | Method for analysis of the operation of a gas turbine | |
| JP3321487B2 (en) | Device / equipment diagnosis method and system | |
| CN109211548A (en) | A kind of mechanical failure diagnostic method | |
| CN104712542A (en) | Reciprocating compressor sensitive characteristic extracting and fault diagnosis method based on internet of things | |
| CN104573850A (en) | Method for evaluating state of thermal power plant equipment | |
| CN106441843B (en) | A kind of rotating machinery fault method for waveform identification | |
| WO2019043600A1 (en) | Remaining useful life estimator | |
| CN114004306A (en) | Equipment fault evaluation system and method based on multi-dimensional data of Internet of things | |
| CN113383160A (en) | Method for computer-implemented monitoring of a wind turbine | |
| CN107844067A (en) | A kind of gate of hydropower station on-line condition monitoring control method and monitoring system | |
| CN102966526A (en) | Diagnosis method for low-frequency fluid vibration excitation failure of compressor based on axial vibration analysis | |
| Du et al. | Intelligent turning tool monitoring with neural network adaptive learning | |
| CN104615123B (en) | K-nearest neighbor based sensor fault isolation method | |
| CN114021620A (en) | Electrical submersible pump fault diagnosis method based on BP neural network feature extraction | |
| CN110441063B (en) | Method for monitoring and diagnosing cracks of large high-speed rotor shaft | |
| JP6685124B2 (en) | Diagnostic device and diagnostic method | |
| Oddan | Multivariate statistical condition monitoring | |
| CN110146267B (en) | Method for detecting flutter of valve plate of totally-enclosed refrigeration compressor | |
| Song et al. | Rolling Bearing Fault Diagnosis Under Different Severity Based on Statistics Detection Index and Canonical Discriminant Analysis | |
| CN111551853A (en) | Hydro-generator stator core lamination loosening fault detection method | |
| dos Santos Pedotti et al. | Failure analysis on a water pump based on a low-cost MEMS accelerometer and Machine Learning Classifiers |
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 |