CN112880726A - Sensor fault diagnosis method and device based on variational modal decomposition sample entropy - Google Patents
Sensor fault diagnosis method and device based on variational modal decomposition sample entropy Download PDFInfo
- Publication number
- CN112880726A CN112880726A CN202011635495.7A CN202011635495A CN112880726A CN 112880726 A CN112880726 A CN 112880726A CN 202011635495 A CN202011635495 A CN 202011635495A CN 112880726 A CN112880726 A CN 112880726A
- Authority
- CN
- China
- Prior art keywords
- fault
- sensor
- sample
- sample set
- training
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D18/00—Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00
Abstract
The invention discloses a sensor fault diagnosis method and device based on variational modal decomposition sample entropy. Firstly, acquiring sensor signals in a normal state and various fault categories, and decomposing the sensor signals by using a VMD (variable matrix display) to obtain n IMFs (intrinsic mode functions) with larger kurtosis values; respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm; the sample entropy value forms a characteristic vector, and the characteristic vector and the fault category code form a characteristic sample to form a characteristic sample set; training the KNN model by using the characteristic sample set; and performing sensor fault diagnosis by using the trained KNN model to realize the identification of different sensor fault types. The fault feature extraction method provided by the invention has strong separability, and can improve the time-frequency description capacity of different sensor faults, thereby improving the accuracy of sensor fault identification.
Description
Technical Field
The invention relates to the technical field of fault diagnosis, in particular to a method and a device for diagnosing faults of a sensor in a measurement and control system.
Background
With the development of sensor technology, the measurement and control system can measure physical quantities under extremely severe environmental conditions and perform operations according to the measured values. The sensor is used as the most important signal acquisition device in the measurement and control system, and has great influence on the overall performance of the measurement and control system. However, due to the influence of factors such as the working environment of the sensor and the characteristics of sensitive materials, the sensor inevitably suffers performance degradation and even failure. Once a sensor fails, its output signal changes accordingly.
The time-frequency characteristic of the output signal of the sensor is utilized to judge the fault, which is an effective way to diagnose the fault of the sensor. In order to improve the overall reliability of the measurement and control system, the reliability of the sensor needs to be monitored in real time, and the fault type of the sensor needs to be identified.
Due to the fact that the sensor fault types are various and the forming reasons are complex, signals of faults cannot be accurately described by means of a mathematical model. Therefore, the realization of sensor fault diagnosis by using a pattern recognition mode is a hot direction in the current research field. The current solution for sensor fault diagnosis is realized by combining a fault feature extraction method and a classifier. However, since the sensor signal itself has a certain time-frequency characteristic, the fault signal extraction superimposed on the sensor signal is a main cause of poor effect of extracting the fault feature in the current method. This phenomenon is particularly pronounced in weak fault situations. The fault feature extraction effect directly influences the overall performance of the fault diagnosis method, so that the fault identification accuracy of the sensor is reduced.
Disclosure of Invention
In view of this, the invention provides a sensor fault diagnosis method and device based on variational modal decomposition sample entropy, and the proposed fault feature extraction method has strong separability, and can improve the time-frequency description capability of different sensor faults, thereby improving the sensor fault identification accuracy.
In order to solve the above-mentioned technical problems, the present invention has been accomplished as described above.
A sensor fault diagnosis method based on variational modal decomposition sample entropy comprises the following steps:
step 1: collecting sensor signals under a normal state and various fault categories;
step 2: decomposing the sensor signal by using Variational Modal Decomposition (VMD) to obtain n intrinsic modal component (IMF) components with larger kurtosis value as the optimal IMF;
and step 3: respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm;
and 4, step 4: forming a characteristic vector by using the sample entropy obtained in the step 3, and carrying out fault type coding on a normal state and a fault type, wherein the characteristic vector and the fault type coding form a characteristic sample to form a characteristic sample set;
and 5: training a K Nearest Neighbor (KNN) model by using a characteristic sample set;
step 6: and performing sensor fault diagnosis by using the trained KNN model to realize the identification of different sensor fault types.
Preferably, the determination manner of n in step 2 is: for the sensor signal in the normal state, the optimal number n of decompositions of the VMD is selected using the instantaneous frequency mean.
Preferably, n is 4.
Preferably, the step 5 is: dividing the characteristic sample set into a training sample set and a testing sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
Preferably, the embedding dimension m in the sample entropy algorithm is 1 or 2, and the similarity tolerance r is 0.1-0.25 times std; wherein std is the standard deviation of n IMFs.
A sensor fault diagnosis device based on variational modal decomposition sample entropy comprises a feature vector generation module, a training module, a KNN model and a fault diagnosis module;
the characteristic vector generation module is used for decomposing the input sensor signals by using the VMD to acquire n IMF components with larger kurtosis values as the optimal IMF; respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm; forming a feature vector by using the sample entropy;
the training module is used for acquiring sensor signals in a normal state and under various fault categories; calling a feature vector of a sensor signal in a sample generated by the feature vector generation module, carrying out fault category coding on a normal state and a fault category, and forming a feature sample by the feature vector and the fault category coding to form a feature sample set; training the KNN model by using the characteristic sample set;
the fault diagnosis module is used for acquiring data of the sensor to be diagnosed, calling the feature vector generation module to generate a corresponding feature vector, inputting the trained KNN model to execute sensor fault diagnosis, and recognizing different sensor fault types.
Preferably, the determination method of n adopted by the feature vector generation module is as follows: for the sensor signal in the normal state, the optimal number n of decompositions of the VMD is selected using the instantaneous frequency mean.
Preferably, the training module divides the feature sample set into a training sample set and a testing sample set when the KNN model is trained by using the feature sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
Has the advantages that:
(1) sensor signals under different time scales can be effectively extracted by utilizing VMD decomposition, and effective extraction of fault features is facilitated.
(2) The n IMFs most likely to generate time-frequency change can be effectively selected by utilizing the kurtosis standard, the probability that the selected n IMFs contain fault features is high, the reliability of feature extraction can be improved by selecting the IMF most likely to generate a fault signal, and the fault feature description capacity is improved.
(3) The physical significance of the sample entropy (SampEn) represents the rate at which the signal generates new information, which is consistent with the generation of sensor fault signals, and is suitable for feature extraction of the decomposed IMFs components. Meanwhile, the sample entropy is suitable for deterministic process and random process analysis, and has good robustness, consistency and anti-noise capability. The fault characteristics of the sensor can be effectively described by utilizing the sample entropy calculation of the n IMFs to obtain the fault characteristics of the sensor.
Drawings
FIG. 1 is a flow chart of the sensor fault diagnosis of the present invention.
Fig. 2 is a block diagram showing the components of the sensor failure diagnosis apparatus of the present invention.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
In order to solve the problems of low feature extraction separability and low fault identification accuracy of the current sensor fault diagnosis method, the invention provides a VMD sample entropy fault feature extraction method. Firstly, decomposing the signal by using Variable Mode Decomposition (VMD), obtaining a series of intrinsic mode component (IMF) components, wherein the IMF components comprise the characteristics of the sensor signal contained under different time scales, and selecting the optimal n IMF components from the IMF components. The kurtosis value of the signal can effectively describe the pulse characteristics of the signal, the higher the kurtosis value is, the richer the pulse characteristics contained in the signal are, and the obtained IMF components with the maximum kurtosis value are the optimal n IMF components. Then, n IMF component sample entropy values are obtained and used as sensor fault characteristics. And finally, identifying the fault type by training a KNN classification model.
As shown in fig. 1, the method comprises the steps of:
step 1: and collecting sensor signals in a normal state and a fault mode, wherein the sensor signals and the classes thereof form samples to form a training set.
Step 2: for the sensor signal in the normal state, the optimal number n of decompositions for the Variational Modal Decomposition (VMD) is selected using the instantaneous frequency mean.
The present invention may use the change in the mean value of the physico instantaneous frequency disclosed in references d.xiao, j.ding, x.li, and l.huang, "Gear fault diagnosis based on kurtosis criterion VMD and SOM neural network," Applied Sciences, vol.9, No.24, p.5424,2019 to select the optimum number of decompositions for VMD. According to the descriptions of the documents K.Dragomersky and D.Zosso, "spatial mode decomposition," IEEE transactions on signal processing, vol.62, No.3, pp.531-544,2013, the penalty factor of VMD algorithm is 2000, and the discrimination precision is 10-7Then, the optimal number of decomposition n is 4 through repeated experimental analysis.
And step 3: for each sample in the sample set, the VMD is used for decomposing the sensor signal, and n IMF components with larger kurtosis values are obtained from the obtained IMF components to be the optimal IMF, wherein the IMF contains the sensor state.
The Variational Modal Decomposition (VMD) overcomes the defects of two traditional signal adaptive Decomposition methods, namely Empirical Mode Decomposition (EMD) and Local Mean Decomposition (LMD), converts the signal Decomposition into the Variational problem, and solves the signal adaptive Decomposition problem by searching the optimal solution of the problem.
The n IMFs most likely to generate time-frequency variation can be effectively selected by utilizing the kurtosis standard, the probability that the selected n IMFs contain fault features is high, the reliability of feature extraction can be improved by selecting the IMF most likely to generate a fault signal, and the fault feature description capacity is improved. Wherein, it is a conventional means in the mathematical field to find kurtosis for a sequence, and it is not described herein.
And 4, step 4: and respectively calculating sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm.
Sample entropy is a time series complexity measure method proposed by Richman, with the more complex the signal, the larger the sample entropy value. The physical meaning of the sample entropy represents the rate at which the time series produces new information, consistent with the form of generation of the sensor fault signal.
The sample entropy is calculated as follows:
step 41: each IMF is a time sequence, and the N time sequences x (t) are grouped into a group of m-dimensional vectors according to the sequence numbers,
X(i)＝[x(i),x(i+1),…,x(i+m-1)]
wherein i is 1,2, …, N-m + 1;
step 42: defining the distance between vectors x (i) and x (j) as the maximum coordinate difference:
d[X(i),X(j)]＝max[|x(i+k)-x(j+k)|]
wherein k is 1,2, …, m-1, i, j is 1,2, …, N-m + 1;
step 43: for a given similarity tolerance r, counting the number of the distances between the ith vector and the other N-m vectors which are less than r, and calculating the ratio of the distance to the number of the vectors to be recorded as the R-m
In the formula, theta is a Heaviside function,
step 44: calculate all
The average value of (A) is recorded as
Step 45: increasing the vector dimension to m +1, and repeating the above calculation steps to obtain Cm+1(r)；
Step 46: the sample entropy of the time series x (n) is expressed as
When the length N of the time series is finite, the above equation is converted into
It can be seen that the time series sample entropy value depends on the values of the embedding dimension m and the similarity tolerance r. With reference to the approximate entropy, the sample entropy can have better statistical characteristics by taking m as 1 or 2 and r as 0.1-0.25 times std (std is the standard deviation of the time series).
And 5: and (4) forming a characteristic vector by using the sample entropy values obtained in the step (4), carrying out fault type coding on the normal state and the fault type, and forming a characteristic sample by using the characteristic vector and the fault type coding to form a characteristic sample set.
Step 6: and training a K Nearest Neighbor (KNN) model by using the characteristic sample set.
The main idea of the KNN algorithm is as follows: the distance between the sample to be classified and the training sample of the known class is calculated, and K samples closest to the sample data to be classified are found. If the K nearest samples of the sample data to be classified all belong to one class, the sample to be classified also belongs to the class. Otherwise, the category with the most occurrence times is selected, and the category is the category of the data to be detected.
In the step, the characteristic sample set is divided into a training sample set and a testing sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
And 7: and performing sensor fault diagnosis by using the trained KNN model to realize the identification of different sensor fault types.
In the step, a sensor signal to be subjected to fault diagnosis is obtained, and then the sensor signal is decomposed by using the VMD to obtain n optimal IMFs; respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm; selecting sample entropy values capable of representing fault modes to form a characteristic vector of the sample, inputting the characteristic vector into the KNN model, outputting a fault type code by the KNN model, and obtaining a sensor fault type by the code corresponding to a specific fault.
This flow ends by this point.
In order to implement the method, the invention further provides a sensor fault diagnosis device based on the variational modal decomposition sample entropy and the KNN, which comprises a feature vector generation module, a training module, a KNN model and a fault diagnosis module, as shown in FIG. 2.
The characteristic vector generation module is used for decomposing the input sensor signals by using the VMD to acquire n IMF components with larger kurtosis values as the optimal IMF, wherein the IMF comprises a sensor state; respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm; and forming a feature vector by using the sample entropy.
The training module is used for acquiring sensor signals in a normal state and a fault mode, and the sensor signals and fault categories thereof form samples to form a training set; calling a feature vector of a sensor signal in a sample generated by the feature vector generation module, carrying out fault category coding on a normal state and a fault category, and forming a feature sample by the feature vector and the fault category coding to form a feature sample set; and training the KNN model by using the characteristic sample set. The training module divides the characteristic sample set into a training sample set and a testing sample set when the KNN model is trained by the characteristic sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
The fault diagnosis module is used for acquiring data of the sensor to be diagnosed, calling the feature vector generation module to generate a corresponding feature vector, inputting the trained KNN model to execute sensor fault diagnosis, and recognizing different sensor fault types.
The above embodiments only describe the design principle of the present invention, and the shapes and names of the components in the description may be different without limitation. Therefore, a person skilled in the art of the present invention can modify or substitute the technical solutions described in the foregoing embodiments; such modifications and substitutions do not depart from the spirit and scope of the present invention.
Claims (8)
1. A sensor fault diagnosis method based on variational modal decomposition sample entropy is characterized by comprising the following steps:
step 1: collecting sensor signals under a normal state and various fault categories;
step 2: decomposing the sensor signal by using Variational Modal Decomposition (VMD) to obtain n intrinsic modal component (IMF) components with larger kurtosis value as the optimal IMF;
and step 3: respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm;
and 4, step 4: forming a characteristic vector by using the sample entropy obtained in the step 3, and carrying out fault type coding on a normal state and a fault type, wherein the characteristic vector and the fault type coding form a characteristic sample to form a characteristic sample set;
and 5: training a K Nearest Neighbor (KNN) model by using a characteristic sample set;
step 6: and performing sensor fault diagnosis by using the trained KNN model to realize the identification of different sensor fault types.
2. The method of claim 1, wherein n in step 2 is determined by: for the sensor signal in the normal state, the optimal number n of decompositions of the VMD is selected using the instantaneous frequency mean.
3. The method of claim 1, wherein n-4.
4. The method of claim 1, wherein step 5 is: dividing the characteristic sample set into a training sample set and a testing sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
5. The method according to claim 1, wherein the embedding dimension m in the sample entropy algorithm takes 1 or 2, and the similarity tolerance r takes 0.1-0.25 times std; wherein std is the standard deviation of n IMFs.
6. A sensor fault diagnosis device based on variational modal decomposition sample entropy is characterized by comprising a feature vector generation module, a training module, a KNN model and a fault diagnosis module;
the characteristic vector generation module is used for decomposing the input sensor signals by using the VMD to acquire n IMF components with larger kurtosis values as the optimal IMF; respectively solving sample entropy values corresponding to the n optimal IMFs by using a sample entropy algorithm; forming a feature vector by using the sample entropy;
the training module is used for acquiring sensor signals in a normal state and under various fault categories; calling a feature vector of a sensor signal in a sample generated by the feature vector generation module, carrying out fault category coding on a normal state and a fault category, and forming a feature sample by the feature vector and the fault category coding to form a feature sample set; training the KNN model by using the characteristic sample set;
the fault diagnosis module is used for acquiring data of the sensor to be diagnosed, calling the feature vector generation module to generate a corresponding feature vector, inputting the trained KNN model to execute sensor fault diagnosis, and recognizing different sensor fault types.
7. The apparatus of claim 6, wherein the feature vector generation module determines n by: for the sensor signal in the normal state, the optimal number n of decompositions of the VMD is selected using the instantaneous frequency mean.
8. The apparatus of claim 6, wherein the training module, when training the KNN model with the feature sample set, divides the feature sample set into a training sample set and a testing sample set; and training the KNN model by using the training sample set, and determining the optimal KNN model by using the test sample set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011635495.7A CN112880726A (en) | 2020-12-31 | 2020-12-31 | Sensor fault diagnosis method and device based on variational modal decomposition sample entropy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011635495.7A CN112880726A (en) | 2020-12-31 | 2020-12-31 | Sensor fault diagnosis method and device based on variational modal decomposition sample entropy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112880726A true CN112880726A (en) | 2021-06-01 |
Family
ID=76046602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011635495.7A Pending CN112880726A (en) | 2020-12-31 | 2020-12-31 | Sensor fault diagnosis method and device based on variational modal decomposition sample entropy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112880726A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105758644A (en) * | 2016-05-16 | 2016-07-13 | 上海电力学院 | Rolling bearing fault diagnosis method based on variation mode decomposition and permutation entropy |
| CN106092574A (en) * | 2016-05-30 | 2016-11-09 | 西安工业大学 | The Method for Bearing Fault Diagnosis selected with sensitive features is decomposed based on improving EMD |
| US20180100894A1 (en) * | 2016-10-07 | 2018-04-12 | United States Of America As Represented By The Secretary Of The Navy | Automatic Generation of Test Sequences |
| CN109472317A (en) * | 2018-11-22 | 2019-03-15 | 哈尔滨工业大学 | The real-time fault detection and real time fail partition method of multidimensional signal |
| CN109489977A (en) * | 2018-12-28 | 2019-03-19 | 西安工程大学 | Method for Bearing Fault Diagnosis based on KNN-AdaBoost |
| CN111397896A (en) * | 2020-03-08 | 2020-07-10 | 华中科技大学 | Fault diagnosis method and system for rotary machine and storage medium |
| CN112098600A (en) * | 2020-09-14 | 2020-12-18 | 哈尔滨工业大学 | Fault detection and diagnosis method for chemical sensor array |
-
2020
- 2020-12-31 CN CN202011635495.7A patent/CN112880726A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105758644A (en) * | 2016-05-16 | 2016-07-13 | 上海电力学院 | Rolling bearing fault diagnosis method based on variation mode decomposition and permutation entropy |
| CN106092574A (en) * | 2016-05-30 | 2016-11-09 | 西安工业大学 | The Method for Bearing Fault Diagnosis selected with sensitive features is decomposed based on improving EMD |
| US20180100894A1 (en) * | 2016-10-07 | 2018-04-12 | United States Of America As Represented By The Secretary Of The Navy | Automatic Generation of Test Sequences |
| CN109472317A (en) * | 2018-11-22 | 2019-03-15 | 哈尔滨工业大学 | The real-time fault detection and real time fail partition method of multidimensional signal |
| CN109489977A (en) * | 2018-12-28 | 2019-03-19 | 西安工程大学 | Method for Bearing Fault Diagnosis based on KNN-AdaBoost |
| CN111397896A (en) * | 2020-03-08 | 2020-07-10 | 华中科技大学 | Fault diagnosis method and system for rotary machine and storage medium |
| CN112098600A (en) * | 2020-09-14 | 2020-12-18 | 哈尔滨工业大学 | Fault detection and diagnosis method for chemical sensor array |
Non-Patent Citations (2)
| Title |
|---|
| DONGMING XIAO ET AL.: "Gear fault diagnosis based on kurtosis criterion VMD and SOM neural network", 《APPLIED SCIENCES》 * |
| KONSTANTIN DRAGOMIRETSKIY ET AL.: "Variational mode decomposition", 《IEEE TRANSACTIONS ON SIGNAL PROCESSING》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113485302B (en) | Vehicle operation process fault diagnosis method and system based on multivariate time sequence data | |
| Feng et al. | Design and implementation of a self-validating pressure sensor | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| Neupane et al. | Bearing fault detection using scalogram and switchable normalization-based CNN (SN-CNN) | |
| CN114090396A (en) | Cloud environment multi-index unsupervised anomaly detection and root cause analysis method | |
| CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| CN111122162A (en) | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| Wen et al. | A new method for identifying the ball screw degradation level based on the multiple classifier system | |
| CN114818817A (en) | Weak fault recognition system and method for capacitive voltage transformer | |
| CN113538353B (en) | Five-phase asynchronous motor rolling bearing fault diagnosis method based on single-channel diagram data enhancement and migration training residual error network | |
| CN116520068B (en) | Diagnostic method, device, equipment and storage medium for electric power data | |
| CN112541524B (en) | BP-Adaboost multisource information motor fault diagnosis method based on attention mechanism improvement | |
| CN113203954A (en) | Battery fault diagnosis method based on time-frequency image processing | |
| CN113919389A (en) | GIS fault diagnosis method and system based on voiceprint imaging | |
| CN116400244B (en) | Abnormality detection method and device for energy storage battery | |
| CN112880726A (en) | Sensor fault diagnosis method and device based on variational modal decomposition sample entropy | |
| CN113255771B (en) | Fault diagnosis method and system based on multi-dimensional heterogeneous difference analysis | |
| CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
| CN111474476B (en) | Motor fault prediction method | |
| CN112069621B (en) | Method for predicting residual service life of rolling bearing based on linear reliability index | |
| CN112560674B (en) | Method and system for detecting sound signal quality | |
| CN117609737B (en) | Method, system, equipment and medium for predicting health state of inertial navigation system | |
| CN115480134A (en) | Fault identification method and system based on equipment partial discharge data diagram | |
| CN114021275A (en) | Rolling bearing fault diagnosis method based on deep convolution fuzzy system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210601 |
|
| RJ01 | Rejection of invention patent application after publication |