CN109213121B - Method for diagnosing clamping cylinder fault of fan braking system - Google Patents
Method for diagnosing clamping cylinder fault of fan braking system Download PDFInfo
- Publication number
- CN109213121B CN109213121B CN201810888737.XA CN201810888737A CN109213121B CN 109213121 B CN109213121 B CN 109213121B CN 201810888737 A CN201810888737 A CN 201810888737A CN 109213121 B CN109213121 B CN 109213121B
- Authority
- CN
- China
- Prior art keywords
- signal
- layer
- fault
- output
- evidence
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0262—Confirmation of fault detection, e.g. extra checks to confirm that a failure has indeed occurred
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a method for diagnosing the fault of a clamping cylinder of a fan braking system, which comprises the following steps of 1, collecting brake shoe gap-time characteristic curve data; 2. preprocessing data; 3. extracting a fault characteristic vector from a brake shoe clearance-time characteristic curve by utilizing a wavelet packet; 4. determining a training parameter according to the dimension of the input signal; 5. dividing the input signal sample obtained in the step 2 into a training set and a test set; 6. performing network training according to the training parameters obtained in the step 4 to obtain a fuzzy neural network model; 7. optimizing the fuzzy neural network model, and testing the fuzzy neural network model; 8. the recognition results of the fuzzy neural networks are used as independent evidences, and a D-S evidence theory is used for fusion to obtain a comprehensive diagnosis result; 9: and outputting the diagnosis result. The invention has the advantages that: the method adopts various technical means to diagnose the clamping cylinder fault of the fan braking system, and improves the accuracy of fault diagnosis.
Description
Technical Field
The invention belongs to the technical field of wind power generation, and particularly relates to a clamping cylinder fault diagnosis method for a fan braking system.
Background
There are generally two methods for improving the reliability of a fan brake system: firstly, the reliability of a braking system is improved; and secondly, detecting the running state of the brake system so as to diagnose the fault of the brake system. In the aspect of improving reliability, the failure of a fan braking system is generally caused by the obvious reduction of the friction coefficient of a brake shoe material, and factors for reducing the friction coefficient mainly include the temperature rise, the sliding speed and the pressure change of the brake shoe. Among the three factors, the influence of temperature rise is the largest, and as the fan runs at a high speed in a load state, the temperature of a brake shoe can be increased rapidly in the braking process, so that the friction coefficient is reduced, and the change of the friction coefficient can have great influence on the performance of a braking system. At present, some documents also derive an analytic expression of the temperature rise of the brake shoe, but the analytic expression is nonlinear and quite complex in structure, and the effect is poor in practical application.
In the aspect of fault diagnosis, the fault of the fan brake system is always a difficult point of fan fault diagnosis, with the development of big data theory, signals are mined from a large amount of information in an information system, and a method of signal fusion by artificial intelligence becomes a new direction of fan brake system fault diagnosis. The current artificial intelligence fault diagnosis has the disadvantages of unsatisfactory diagnosis result, low accuracy and high false alarm rate.
Many field failure examples show that the cylinder piston of the brake system is stuck to be a common failure. Based on the method, the invention provides an artificial intelligence-based method for diagnosing the clamping cylinder fault of the braking system of the fan. The fuzzy neural network used by the invention is a product of combining the fuzzy theory and the neural network, integrates the advantages of the neural network and the fuzzy theory, and integrates learning, association, identification and information processing.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method for diagnosing the stuck cylinder fault of the fan braking system, which combines a fuzzy neural network with a D-S evidence to accurately detect the running state of a fan and improve the accuracy of diagnosing the stuck cylinder fault of the fan braking system.
The technical problem to be solved by the invention is realized by the technical scheme, which comprises the following steps: step 1, data acquisition: deeply excavating a brake shoe gap-time signal from a large amount of information about the running state of a fan braking system contained in an SCADA information system;
step 2, data preprocessing: removing singular points in signal data, denoising the signal, normalizing the data and carrying out similarity processing on output signals;
step 3, extracting fault characteristic quantity: extracting a fault characteristic vector from a brake shoe gap-time signal by utilizing a wavelet packet;
step 4, determining training parameters according to the dimensionality of the input signal; the training parameters comprise hidden layer node number, period for displaying intermediate result, maximum iteration times, learning rate and error threshold value; the dimensionality of the input signal corresponds to the number of input layer nodes;
step 5, dividing the input signal sample obtained in the step 2 into a training set and a testing set, wherein the sample contains normal data and fault data;
step 6, performing network training according to the training parameters and the training set data obtained in the step 4 to obtain a fuzzy neural network model;
step 7, optimizing the fuzzy neural network model, testing the fuzzy neural network model: comparing the output value of the fuzzy neural network model with ideal output (namely actually measured data or a value set by the user) to obtain an error, performing error back transmission, updating parameters in the model until the error meets the requirement, and obtaining an optimal fuzzy neural network model; verifying the performance of the network with the test set data;
step 8, fusing D-S evidence theories: the output values of the fuzzy neural networks are sent to a decision fusion module, the recognition results of the fuzzy neural networks are used as independent evidences, and a D-S evidence theory is used for fusion to obtain a comprehensive diagnosis result;
and step 9: and outputting a diagnosis result: and sending the diagnosis result into a diagnosis system for analysis, and presenting the final result in a human-computer interaction interface.
According to the invention, the existing data is fully utilized by utilizing the data in the SCADA information system, the big data theory is applied, the wavelet theory, the fuzzy neural network, the evidence theory and other tools are adopted, the fault diagnosis is carried out around the cylinder clamping fault of the oil cylinder of the fan braking system based on the viewpoint of information fusion, the acquired data is fully utilized through the neural network, and the fault tolerance of a certain signal is increased by the D-S evidence theory fusion method.
Compared with the prior art, the invention has the technical effects that:
1. the invention effectively realizes the extraction of the fault characteristic quantity by utilizing the wavelet theory.
2. The invention processes the data with strong correlation and accelerates the convergence speed of the neural network.
3. The data in the system is fully utilized to train the fuzzy neural network to obtain an optimal model, and then the test set data is utilized again to test the model, so that the accuracy of the model is ensured.
4. The clamping cylinder fault of the fan braking system is diagnosed by fusing a plurality of sensor signals, so that the accuracy of the clamping cylinder fault diagnosis is further improved, the feasibility is provided for intelligent health management of the wind turbine generator, and the safe, stable and efficient operation of a power grid can be realized.
Drawings
The drawings of the invention are illustrated as follows:
FIG. 1 is a flow chart of a fault method of the present invention;
FIG. 2 is an exploded view of a wavelet packet signal according to the present invention;
FIG. 3 is a diagram of a fuzzy neural network topology according to the present invention;
FIG. 4 is a flowchart of fuzzy neural network model optimization and testing in accordance with the present invention;
FIG. 5 is a framework diagram of the D-S evidence theory for fault diagnosis in the present invention.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
at present, various sensors are used for detecting the running state of a fan, two systems, namely a CMS (state monitoring system) system and an SCADA (supervisory control and data acquisition) system, are installed in a wind field, and the two systems can acquire various data of a generator set in real time. As shown in fig. 1, the present invention comprises the steps of:
step 1, data acquisition: deeply excavating a large amount of information about the running state of a fan braking system contained in an SCADA information system, and respectively acquiring signals of two eddy current sensors through an A/D acquisition board to obtain a brake shoe gap-time signal;
the brake shoe clearance-time signal is mainly used for the subsequent feature vector extraction.
Step 2, data preprocessing: removing singular points in signal data, denoising the signal, normalizing the data and performing similarity processing on output signals, wherein the similarity processing is to adjust the output data to be the same when the input is the same and the output deviation is small;
the singular point of the data signal refers to a signal which is suddenly larger or smaller than the signal at the previous moment and the signal at the later moment, and due to the harsh operating environment of the fan, the sensor may be interfered by the outside or a certain sensor fails, so that the data measured by the sensor may be wrong, the data measured by the sensor must be removed, or the final result is seriously affected.
The principle of rejecting singular points is as follows: the data with the deviation larger than 3 sigma is removed according to the statistical probability theory by calculating the standard deviation sigma of the input sample signal.
The input quantity has different physical meanings and different dimensions, and the normalized data all change between [0, 1], so that each input component has the same important position in the network training. The data must be normalized for the prediction problem. The formula of the data normalization process is as follows:
in the formula, x*For normalized values of the data, xiFor the ith data of the signal, xmax、xminThe maximum and minimum values of the sample data, respectively.
The similarity processing means that when the input is the same and the output deviation is small, the output data is adjusted to be the same;
in the output signal similarity processing, an output deviation threshold is set according to the requirement, and when the output deviation meets the requirement, the output data can be adjusted to be the same, so that the convergence speed of the operation is accelerated.
Step 3, extracting fault characteristic quantity: extracting a fault characteristic vector from a brake shoe gap-time signal by utilizing a wavelet packet;
the steps of extracting the fault characteristic quantity are as follows:
step 31, carrying out wavelet decomposition on the brake shoe gap-time signal;
the wavelet packet signal decomposition diagram is shown in fig. 2: performing three-layer wavelet packet decomposition on the signal, and respectively extracting signal characteristics of 8 frequency components from low frequency to high frequency in the third layer; in fig. 2, the (0, 0) node represents the original signal S, (1, 1) node represents the first-layer low-frequency coefficient of wavelet packet decomposition, (1, 2) node represents the first-layer high-frequency coefficient of wavelet packet decomposition, (2, 1) and (2, 2) nodes represent the second-layer low-frequency coefficient of wavelet packet decomposition, (2, 3) and (2, 4) nodes represent the second-layer high-frequency coefficient of wavelet packet decomposition, and (3, 1) … … (3, 8) node represents the coefficients of the third-layer first to eighth nodes of wavelet packet decomposition.
Step 32, reconstructing the wavelet packet decomposition coefficients, and extracting signals in each frequency band range; with S31Denotes x31(x31Is the signal of the first node of the third layer of wavelet packet decomposition), S32Denotes x32(x32Is the signal of the second node of the third layer of wavelet packet decomposition), and so on; the original signal S can be represented as: s ═ S31+S32+S33+S34+S35+S36+S37+S38
Step 33, calculating the total energy of each frequency band signal; because the input signal is a random signal, the output signal is also a random signal; setting a reconstruction signal S3j(j-1, 2, …, 8) corresponds to an energy E3j(j ═ 1,2, …, 8), then there are
In the formula: x is the number of3j(j ═ 1,2, …, 8) denotes the reconstructed signal S3jThe amplitude of the discrete points of (a).
Step 34, constructing a feature vector, wherein the feature vector is constructed as follows:
T=[E31,E32,E33,E34,E35,E36,E37,E38]
when the energy is large, E3j(j-1, 2, …, 8) is typically a large number that is used for subsequent data analysisSome inconveniences, so the feature vector T is normalized by
T′=[E31/E,E32/E,E33/E,E34/E,E35/E,E36/E,E37/E,E38/E]
The vector T' is the normalized feature vector.
Step 4, determining training parameters according to the dimensionality of the input signal; the training parameters comprise hidden layer node number, period for displaying intermediate result, maximum iteration times, learning rate and error threshold value; the dimensionality of the input signal corresponds to the number of input layer nodes;
determining formula of hidden layer node number:
in the formula: m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10.
And selecting an error threshold according to the actual precision requirement of the engineering.
And 5, dividing the input signal sample obtained in the step 2 into a training set and a testing set, wherein the input signal sample contains normal data and fault data. The input signal samples are data relating to the operational status of the fan brake system as measured by various sensors.
Step 6, performing network training according to the training parameters and the training set data obtained in the step 4 to obtain a fuzzy neural network model;
the structure of the fuzzy neural network is shown in figure 3, the fuzzy neural network has 3 layers which are an input layer, a membership function generation layer and an output layer respectively, the input layer has d neurons, the membership function generation layer has h neurons, the output layer has g neurons, X represents input data, y represents output result, the input variable membership function is selected as a Gaussian function, and the Gaussian membership function is selected as the output resultMu andsigma represents a mean value and a standard value respectively,representing the input to the a-th node of the k-th layer,representing the output of the a-th node of the k-th layer, membership valueThe connection weight between the first and second layers is constant 1, and the function is a fuzzy membership function mubc(. h), a fuzzy membership function between the b factor and the c index; w is avThe weight coefficient between the second and third layers.
The processing procedures of each layer of the fuzzy neural network are as follows:
let the f-th neuron input and output of the input layer be x respectivelyfAnd OfThe p-th neuron input and output of the membership function generation layer are respectivelyAndthe q-th neuron input and output of the output layer are respectivelyAndthen
Step 61, the output of the input layer is equal to its input value, i.e.
Of=xf,
f is the input layer neuron number, f is 1,2, …, d;
in the formula: m isrpAnd σrpMean and standard deviation of Gaussian membership functions of the p-th fuzzy set respectively representing the r-th input quantity, rrpThe membership degree of the r factor to the p index is the adjustable parameter of the network;
in the formula:to blur the output of the qth neuron of the neural network,generating the output of the pth neuron in the layer for the membership function, wpqIs the weight coefficient between the p-th neuron of the second layer and the q-th neuron of the third layer.
The fuzzy neural network has three types of adjustable parameters: one is a rule parameter, which is a weight coefficient between the second and third layers; the second type of adjustable parameter is the mean value m of the Gaussian membership functionrpAnd standard deviation σrpWhich are located in nodes of the second layer and represent parameters of the input membership function; the third type is an evaluation result output layer.
Step 7, optimizing the fuzzy neural network model, and testing the fuzzy neural network model;
comparing the output value of the fuzzy neural network model with ideal output (namely actually measured data or a value set by the user) to obtain an error, performing error back transmission, updating parameters in the model until the error meets the requirement, and obtaining an optimal fuzzy neural network model; the test set data is used to verify the performance of the network.
As shown in fig. 4, the fuzzy neural network model optimization and testing includes the following steps:
step 71, inputting a sample;
step 72, inputting parameters and determining an error threshold;
step 73, comparing the actual output of the network with the ideal output, and obtaining an output error E by using a secondary cost function;
let t samples from the training set be given, the samples are inputd is a neuron in the input layer, and the output sampleg is a neuron in the output layer, s is 1,2, …, t,the ideal output signal of the u-th neuron for the s-th sample,the network output signal of the u-th neuron of the s-th sample
Step 74, judging whether the obtained error is smaller than an error threshold value, if so, executing step 76, otherwise, executing step 75;
step 75, if the error is larger than the error threshold value, performing error reverse transmission, and performing reverse transmission layer by layer from the third layer to the first layer; method for correcting adjustable parameters w of network by fuzzy methodv,mrp,σrpThe correction values of the parameters are as follows:
the updating algorithm of each adjustable parameter is as follows:
wv(T+1)=wv(T)+η1Δwv(T)
mrp(T+1)=mrp(T)+η2Δmrp(T)
σrp(T+1)=σrp(T)+η3Δσrp(T)
in the formula, wv(T) and wv(T +1) is a parameter function before and after updating; eta1,η2,η3The learning rates of the parameters are respectively; t discrete time variable.
Under the condition that the output error is larger than the error threshold, continuing to train the parameters in the model until the error is smaller than the threshold;
step 76, obtaining an optimal fuzzy neural network model when the error is smaller than the error threshold value;
step 77, the performance of the network is verified using the test set data.
Step 8, fusing D-S evidence theories: the output values of the fuzzy neural networks are normalized and then sent to a decision fusion module, a plurality of sensors are used for testing signal data, the data measured by one sensor is used for one fuzzy neural network, then the identification results of all the fuzzy neural networks are used as independent evidences, and a D-S evidence theory is used for fusion to obtain a comprehensive diagnosis result.
And modifying the evidences by utilizing an entropy principle to reduce the conflict among the evidences. As shown in fig. 5, D-S evidence fusion includes the following steps:
step 81, modifying evidence entropy: for F evidences, N fault types, mbaAnd (3) representing a basic probability assignment function of the a-th evidence to the b faults, namely BPA values, wherein the evidence entropy is as follows:
to itThe reciprocal is normalized to obtain the weight w of each evidencea:
Respective original BPA values and Total weighted average of evidence item aThe deviation of (d) is noted as:
εa=(ε1a,ε2a,…,εba,…,εNa) Wherein a is 1,2, …, F
Note the bookIs maAnddegree of total deviation therebetween, using deltaaTo adjust the original evidence, the modified evidence is denoted as m'aAnd then:
m′a=mba-εbaδawherein a is 1,2, …, F, b is 1,2, …, N
Step 82, fusing the evidence modified in the step 81 by using a D-S evidence theory, wherein a fusion formula of G pieces of evidence is as follows:
in the formula: m (A) represents a fused basic probability assignment function of the fault A, G represents the number of sensors, namely evidence bodies, K represents a conflict factor, the larger the K is, the larger the conflict between the evidences is, and conversely, the smaller the K is, the smaller m is1、m2、…mGBasic probabilistic valuation functions, A, representing individual pieces of evidence1、A2、…AnRepresenting a fault set containing a fault A in each evidence;
step 83, making a final diagnosis decision by using a relevant decision rule;
if it isThe failure modes in three failure domains (all possible failure modes of the brake system, denoted by theta) satisfy:
m(A)≥ε1
m(A)-m(B)≥ε2
in the formula, epsilon1,ε2Is a predetermined threshold, then A is the result of the decision, empirically, where ε1=0.6,ε20.3, m (a), m (b), m (c) represent A, B, C basic probability assignment functions for three failure modes.
Step 9, outputting a diagnosis result;
and sending the diagnosis result into a diagnosis system for analysis, and presenting the final result in a human-computer interaction interface. The specific operation process of the step is as follows: and sending the prediction result into an information base for storage, and transmitting the prediction result to an inference machine, wherein the inference machine analyzes the prediction result in the information base, matches information in a knowledge base to obtain a corresponding fault reason, outputs the fault reason to an interpreter to obtain a corresponding explanation, and presents the explanation in a human-computer interaction interface, so that a user can obtain a corresponding inference process.
Claims (7)
1. A method for diagnosing a stuck cylinder fault of a fan braking system comprises the following steps:
step 1, data acquisition: deeply excavating a brake shoe gap-time signal from a large amount of information about the running state of a fan braking system contained in an SCADA information system;
step 2, data preprocessing: removing singular points in signal data, denoising the signal, normalizing the data and carrying out similarity processing on output signals;
step 3, extracting fault characteristic quantity: extracting a fault characteristic vector from a brake shoe gap-time signal by utilizing a wavelet packet;
step 4, determining training parameters according to the dimensionality of the input signal; the training parameters comprise hidden layer node number, period for displaying intermediate result, maximum iteration times, learning rate and error threshold value; the dimensionality of the input signal corresponds to the number of input layer nodes;
step 5, dividing the input signal sample obtained in the step 2 into a training set and a testing set, wherein the input signal sample contains normal data and fault data;
step 6, performing network training according to the training parameters and the training set data obtained in the step 4 to obtain a fuzzy neural network model;
step 7, optimizing the fuzzy neural network model, testing the fuzzy neural network model: comparing the output value of the fuzzy neural network model with the ideal output to obtain an error, performing error back transmission, updating parameters in the model until the error meets the requirement, and obtaining an optimal fuzzy neural network model; verifying the performance of the network with the test set data;
step 8, fusing D-S evidence theories: the output values of the fuzzy neural networks are sent to a decision fusion module, the recognition results of the fuzzy neural networks are used as independent evidences, and a D-S evidence theory is used for fusion to obtain a comprehensive diagnosis result;
the method is characterized in that the D-S evidence fusion comprises the following steps:
step 1), modifying evidence entropy: for F evidences, N fault types, mbaAnd representing the BPA value of the a th evidence to the b faults, the evidence entropy is as follows:
normalizing the reciprocal to obtain the weight w of each evidencea:
respective original BPA values and Total weighted average of evidence item aThe deviation of (d) is noted as:
εa=(ε1a,ε2a,…,εba,…,εNa)
note the bookIs maAnddegree of total deviation therebetween, using deltaaTo adjust the original evidence, the modified evidence is denoted as m'aAnd then:
m′a=mba-εbaδa
step 2), fusing the evidences modified in the step 1) by using a D-S evidence theory, wherein the fusion formula of the G evidences is as follows:
in the formula: m (A) represents the fused basic probability assignment function of the fault A, G represents the number of sensors, namely evidence bodies, K represents a conflict factor, m1、m2、…mGBasic probabilistic valuation functions, A, representing individual pieces of evidence1、A2、…AnRepresenting a fault set containing a fault A in each evidence;
step 3), making a final diagnosis decision by using a relevant decision rule;
theta indicates all possible failure modes of the brake system ifThe failure modes in the three failure domains meet the following conditions:
m(A)≥ε1
m(A)-m(B)≥ε2
in the formula, epsilon1,ε2If the value is a preset threshold value, A is the result of judgment, and m (A), m (B), m (C) represent A, B, C BPA values of three fault modes;
and step 9: and outputting a diagnosis result: and sending the diagnosis result into a diagnosis system for analysis, and presenting the final result in a human-computer interaction interface.
2. The method for diagnosing the stuck cylinder fault of the fan brake system according to claim 1, wherein in the step 2, the formula of data normalization processing is as follows:
in the formula, x*For normalized values of the data, xiFor the ith data of the signal, xmax、xminThe maximum and minimum values of the sample data, respectively.
3. The blower brake system clamping cylinder fault diagnosis method according to claim 2, wherein in the step 3, the step of extracting fault characteristic quantity comprises the following steps:
step 31, carrying out wavelet decomposition on the brake shoe gap-time signal;
performing three-layer wavelet packet decomposition on the signal, and respectively extracting signal characteristics of 8 frequency components from low frequency to high frequency in the third layer;
step 32, reconstructing the wavelet packet decomposition coefficients, and extracting signals in each frequency band range; with S31Denotes x31Reconstructed signal of x31The wavelet packet decomposes the signal of the first node of the third layer; s32Denotes x32Reconstructed signal of x32Decomposing the signal of the second node of the third layer by the wavelet packet, and so on; the original signal S can be represented as: s ═ S31+S32+S33+S34+S35+S36+S37+S38;
Step 33, calculating the total energy of each frequency band signal;
because the input signal is a random signal, the output signal is also a random signal; setting a reconstruction signal S3jCorresponding energy is E3jThen, there are:
in the formula: x is the number of3jRepresenting the reconstructed signal S3jJ ═ 1,2, …, 8;
step 34, constructing a feature vector, wherein the feature vector is constructed as follows:
T=[E31,E32,E33,E34,E35,E36,E37,E38]
when the energy is large, E3jUsually a larger value, and the feature vector T is normalized to obtain
T′=[E31/E,E32/E,E33/E,E34/E,E35/E,E36/E,E37/E,E38/E]
The vector T' is the normalized feature vector.
4. The blower brake system clamping cylinder fault diagnosis method according to claim 3, characterized in that in step 4, the determination formula of the number of hidden layer nodes is as follows:
in the formula: m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10.
5. The method for diagnosing the stuck cylinder fault of the fan brake system according to claim 4, wherein in the step 6, the processing procedures of each layer of the fuzzy neural network are as follows:
let the f-th neuron input and output of the input layer be x respectivelyfAnd OfThe p-th neuron input and output of the membership function generation layer are respectivelyAndthe q-th neuron input and output of the output layer are respectivelyAndthen
Step 61, the output of the input layer equals its input value:
Of=xf
f is the number of neurons in the input layer, f is 1,2, …, d, d is the total number of neurons in the input layer;
step 62, membership function layer:in the formula, mrpAnd σrpMean and standard deviation of Gaussian membership functions of the p-th fuzzy set respectively representing the r-th input quantity, rrpThe membership degree of the r factor to the p index is the adjustable parameter of the network;
6. The blower brake system clamping cylinder fault diagnosis method according to claim 5, wherein in step 7, the error E is calculated by the following formula:
wherein t is the total number of samples in the self-training set, s is the sample number, s is 1,2, …, t, g is the total number of neurons in the output layer, and u is the neuron number of the output layer;the ideal output signal of the u-th neuron for the s-th sample,the signal is output for the network of the u-th neuron for the s-th sample.
7. The method for diagnosing the stuck cylinder fault of the fan brake system as claimed in claim 6, wherein in the step 7, parameters in the updated model are as follows: weight coefficient w between the second and third layersvMean value m of Gaussian-type membership functionsrpAnd standard deviation σrpThe correction values of the parameters are as follows:
the updating algorithm of each parameter is as follows:
wv(T+1)=wv(T)+η1Δwv(T)
mrp(T+1)=mrp(T)+η2Δmrp(T)
σrp(T+1)=σrp(T)+η3Δσrp(T)
in the formula, wv(T) and wv(T +1) is a parameter function before and after updating; eta1,η2,η3The learning rates of the parameters are respectively;
t discrete time variable.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810888737.XA CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810888737.XA CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109213121A CN109213121A (en) | 2019-01-15 |
CN109213121B true CN109213121B (en) | 2021-05-18 |
Family
ID=64988180
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810888737.XA Active CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109213121B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111497812A (en) * | 2019-01-31 | 2020-08-07 | 天津所托瑞安汽车科技有限公司 | Vehicle braking system monitoring method based on platform |
CN111781435B (en) * | 2019-04-04 | 2023-05-23 | 中车唐山机车车辆有限公司 | Fault detection method and device for four-quadrant rectifier |
CN110826690A (en) * | 2019-10-10 | 2020-02-21 | 深圳供电局有限公司 | Equipment state identification method and system and computer readable storage medium |
CN110617960A (en) * | 2019-10-12 | 2019-12-27 | 华北电力大学 | Wind turbine generator gearbox fault diagnosis method and system |
CN113963456A (en) * | 2020-07-20 | 2022-01-21 | 核工业理化工程研究院 | Method and system for analyzing operation data of multiple high-speed rotating devices |
CN111985820B (en) * | 2020-08-24 | 2022-06-14 | 深圳市加码能源科技有限公司 | FNN and DS fusion-based fault identification method for charging operation management system |
CN112101457B (en) * | 2020-09-15 | 2023-11-17 | 湖南科技大学 | PMSM demagnetizing fault diagnosis method based on torque signal fuzzy intelligent learning |
CN112798956A (en) * | 2020-12-31 | 2021-05-14 | 江苏国科智能电气有限公司 | Wind turbine generator fault diagnosis method based on multi-resolution time sequence cyclic neural network |
CN113790890B (en) * | 2021-09-10 | 2022-08-05 | 南京航空航天大学 | Wavelet packet decomposition weight fuzzy entropy and ELM-based bearing fault classification method and device |
CN114091528A (en) * | 2021-11-11 | 2022-02-25 | 烟台杰瑞石油服务集团股份有限公司 | Fault diagnosis method, diagnosis model construction method, apparatus, device and medium |
CN114690038B (en) * | 2022-06-01 | 2022-09-20 | 华中科技大学 | Motor fault identification method and system based on neural network and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2447888A1 (en) * | 2010-10-27 | 2012-05-02 | Honeywell International Inc. | System and method for determining fault diagnosability of a health monitoring system |
CN105223809A (en) * | 2015-07-10 | 2016-01-06 | 沈阳工业大学 | The synchronous control system of the fuzzy neural network compensator of H type platform and method |
CN205240976U (en) * | 2015-10-16 | 2016-05-18 | 日立电梯(中国)有限公司 | Arresting gear fault detection system |
CN106779063A (en) * | 2016-11-15 | 2017-05-31 | 河南理工大学 | A kind of hoist braking system method for diagnosing faults based on RBF networks |
-
2018
- 2018-08-07 CN CN201810888737.XA patent/CN109213121B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2447888A1 (en) * | 2010-10-27 | 2012-05-02 | Honeywell International Inc. | System and method for determining fault diagnosability of a health monitoring system |
CN105223809A (en) * | 2015-07-10 | 2016-01-06 | 沈阳工业大学 | The synchronous control system of the fuzzy neural network compensator of H type platform and method |
CN205240976U (en) * | 2015-10-16 | 2016-05-18 | 日立电梯(中国)有限公司 | Arresting gear fault detection system |
CN106779063A (en) * | 2016-11-15 | 2017-05-31 | 河南理工大学 | A kind of hoist braking system method for diagnosing faults based on RBF networks |
Non-Patent Citations (2)
Title |
---|
基于证据理论的综合故障诊断方法研究;王致杰、等;《上海电机学院学报》;20091231;第12卷(第4期);第267-270页 * |
网络环境下矿井提升机智能故障诊断关键技术研究;王峰;《中国优秀博士学位论文全文数据库工程科技Ⅰ辑》;20140731;正文第58-95页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109213121A (en) | 2019-01-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109213121B (en) | Method for diagnosing clamping cylinder fault of fan braking system | |
CN111562358B (en) | Transformer oil gas content prediction method and system based on combined model | |
CN106338406B (en) | The on-line monitoring of train traction electric drive system and fault early warning system and method | |
CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
CN109800875A (en) | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine | |
CN108445868B (en) | Intelligent automobile fault diagnosis system and method based on modern signal processing technology | |
CN112070128B (en) | Transformer fault diagnosis method based on deep learning | |
Liang et al. | Rolling bearing fault diagnosis based on one-dimensional dilated convolution network with residual connection | |
CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
CN113505655B (en) | Intelligent bearing fault diagnosis method for digital twin system | |
CN110163075A (en) | A kind of multi-information fusion method for diagnosing faults based on Weight Training | |
CN110866448A (en) | Flutter signal analysis method based on convolutional neural network and short-time Fourier transform | |
CN113188807B (en) | Automatic abs result judging algorithm | |
Li et al. | Gear pitting fault diagnosis using raw acoustic emission signal based on deep learning | |
CN111680875A (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
CN105241665A (en) | Rolling bearing fault diagnosis method based on IRBFNN-AdaBoost classifier | |
CN112508242A (en) | Method for constructing bearing fault location and classification model of wind power generator | |
Han et al. | An intelligent fault diagnosis method of variable condition gearbox based on improved DBN combined with WPEE and MPE | |
CN116776075A (en) | Fan blade health monitoring system and monitoring method thereof | |
CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
CN113780432B (en) | Intelligent detection method for operation and maintenance abnormity of network information system based on reinforcement learning | |
CN113688885A (en) | Deep space probe autonomous fault diagnosis method based on pulse neural network | |
CN112163630A (en) | Compound fault diagnosis method and device based on unbalanced learning |
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 |