CN115017978A - Fault classification method based on weighted probability neural network - Google Patents
Fault classification method based on weighted probability neural network Download PDFInfo
- Publication number
- CN115017978A CN115017978A CN202210543721.1A CN202210543721A CN115017978A CN 115017978 A CN115017978 A CN 115017978A CN 202210543721 A CN202210543721 A CN 202210543721A CN 115017978 A CN115017978 A CN 115017978A
- Authority
- CN
- China
- Prior art keywords
- fault
- neural network
- matrix
- sample
- weighted
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000011159 matrix material Substances 0.000 claims description 41
- 238000012549 training Methods 0.000 claims description 22
- 239000013598 vector Substances 0.000 claims description 22
- 238000012360 testing method Methods 0.000 claims description 15
- 210000002569 neuron Anatomy 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000003062 neural network model Methods 0.000 claims description 8
- 230000035945 sensitivity Effects 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 7
- 238000012847 principal component analysis method Methods 0.000 claims description 5
- 239000000126 substance Substances 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 2
- 210000004027 cell Anatomy 0.000 claims 1
- 238000010801 machine learning Methods 0.000 abstract description 3
- 239000000498 cooling water Substances 0.000 description 8
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 238000013145 classification model Methods 0.000 description 3
- 238000003066 decision tree Methods 0.000 description 3
- 238000007637 random forest analysis Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 206010000372 Accident at work Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010206 sensitivity analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000004781 supercooling Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a fault classification method based on a weighted probability neural network, which relates to the technical field of machine learning and solves the technical problem that the conventional power station equipment cannot accurately classify faults. Known faults and novel faults can be effectively processed, and fault classification of the power station equipment under the full-working-condition operation is achieved. Meanwhile, aiming at novel faults, new fault samples are easily added into a trained network, only corresponding hidden layer units are needed to be added, retraining is not needed, and the economical efficiency and safety of operation of a power plant are improved.
Description
Technical Field
The application relates to the technical field of machine learning, in particular to a fault classification method based on a weighted probability neural network.
Background
In the "industrial 4.0" context, power plants are equipped with a large number of instruments to capture the values of hundreds of variables at each moment, constituting a huge data set. When equipment breaks down, due to the fact that information in data is complex and diverse, abnormal states cannot be rapidly and accurately processed only through expert experience. If the fault is not processed timely, further development of the fault state can cause industrial accidents, further casualties and serious economic loss, and even unit halt. Therefore, an intelligent fault classification tool is needed to help determine fault information and take corrective measures, so that the safety and the economy of the unit are ensured.
The fault classification is based on fault detection of equipment, and according to characteristics and changes of operating parameters when the equipment fails, the fault is analyzed from data by combining the working principle and the thermal performance of the equipment, fault variable information is provided, fault categories are distinguished quickly and accurately, and a fault solution is convenient to determine. Along with the power station equipment maximization, the degree of complication constantly promotes, different trouble often couples a plurality of same fault characteristics, and the classification of trouble can't be accurately discerned to traditional processing mode, therefore accurate fault classification has important meaning to solving the trouble problem fast, the operation of guarantee unit safety and stability.
Disclosure of Invention
The application provides a fault classification method based on a weighted probability neural network, which aims to accurately classify faults and realize real-time classification of the faults of power station equipment under the full-working-condition operation, thereby solving the problem of the faults quickly.
The technical purpose of the application is realized by the following technical scheme:
a fault classification method based on a weighted probability neural network comprises the following steps:
s1: acquiring a training sample set and a testing sample set of faults based on historical fault data of the power station equipment, forming a sample matrix by the training sample set, and performing dimensionality reduction on the sample matrix to obtain a score matrix and a load matrix;
s2: extracting fault characteristics in the score matrix and the load matrix by adopting a principal component analysis method based on reconstruction, and taking the extracted fault characteristics as input samples of a weighted probability neural network;
s3: training the weighted probability neural network through the fault characteristics to obtain a weighted probability neural network model;
s4: and inputting the test sample set into the trained weighted probability neural network model to obtain a fault classification result.
The beneficial effect of this application lies in:
(1) the system and the method provided by the application have high learning speed; the learning rule of the selected method in the training process is simple, the calculation speed is high, and the training process is completed at one time.
(2) On the basis of a traditional Bayesian decision-based probabilistic neural network, weighting factors are added in a mode layer and a summation layer of the probabilistic neural network to measure the class separability of samples, so that novel fault information is further provided for fault classification decision.
(3) The weighted probability neural network classification model provided by the application can effectively process known faults and novel faults, and can realize fault classification of power station equipment under full-working-condition operation. Meanwhile, aiming at novel faults, new fault samples are easily added into a trained network, only corresponding hidden layer units are needed to be added, retraining is not needed, and the economical efficiency and safety of operation of a power plant are improved.
(4) The fault classification module classifies samples based on the Bayes minimum risk criterion, prior knowledge can be utilized to the greatest extent, and no matter how complex the classification problem is, the optimal solution under the Bayes criterion can be guaranteed to be obtained as long as sufficient training samples exist.
Drawings
FIG. 1 is a flow chart of a method described herein;
FIG. 2 is a diagram of a weighted probabilistic neural network architecture;
FIG. 3 is a graph illustrating the weight analysis results of known faults;
fig. 4 is a diagram illustrating the weight analysis result of the new fault.
Detailed Description
The technical solution of the present application will be described in detail below with reference to the accompanying drawings.
The condenser is used as important equipment of a steam turbine unit of a power plant, and has extremely important influence on the vacuum index of the unit. This application uses certain power plant's condenser system as the research object, carries out the analysis to its trouble under different operating modes. Firstly, collecting data of a condenser in an SIS (information system) of the power plant, acquiring operation data and offline historical fault data of the condenser under different working conditions, and marking the fault data correspondingly to serve as training data of a weighted probability neural network model.
TABLE 1 condenser System Fault characterization parameters
| Numbering | Name (R) | Numbering | Name (R) |
| F 1 | Temperature of condenser | F 14 | Power of circulating water pump |
| F 2 | Condenser pressure | F 15 | Mass flow at the air extractor inlet |
| F 3 | Liquid level of condenser | F 16 | Mixed pressure at air extractor outlet |
| F 4 | Steam turbine exhaust mass flow | F 17 | Ejector outlet mix temperature |
| F 5 | End difference | F 18 | Mixed pressure drop of condenser system |
| F 6 | Degree of supercooling | F 19 | Mixed temperature rise of condenser system |
| F 7 | Mass flow of circulating cooling water | F 20 | Additional heat source |
| F 8 | Inlet temperature of circulating cooling water | F 21 | Air leakage rate |
| F 9 | Outlet temperature of circulating cooling water | F 22 | Valve position of air extractor |
| F 10 | Inlet pressure of circulating cooling water | F 23 | Control valve position for circulating cooling water |
| F 11 | Outlet pressure of circulating cooling water | F 24 | Relative roughness in pipe |
| F 12 | Temperature rise of circulating cooling water | F 25 | Condensate pump outlet pressure |
| F 13 | Pressure drop of circulating cooling water | F 26 | Outlet pressure of circulating water pump |
When a condenser system breaks down, various operation parameters of the system are influenced by different degrees, and the parameters comprise real-time operation data obtained by using a sensor on site and performance indexes obtained based on thermodynamic calculation. The characteristic parameters for condenser system fault classification are shown in table 1 and are used for fault characteristic extraction and identification of fault information.
As shown in fig. 1, the fault classification method using the weighted probabilistic neural network includes the following steps:
s1: the method comprises the steps of obtaining a training sample set and a testing sample set of faults based on historical fault data of power station equipment, forming a sample matrix by the training sample set, and performing dimensionality reduction on the sample matrix to obtain a score matrix and a load matrix.
Specifically, the fault data of the power station equipment is acquired by collecting the fault data of the field equipment through a sensor, combining the fault case library data, extracting the operation experience and the fault simulation rule, the fault data of the power station equipment is used for constructing a training sample set of a classifier to obtain an original observation matrix X,
wherein N represents the total number of samples, and each sample includes m observation variables, then the original observation matrix X is represented as:
carrying out dimensionality reduction processing on the original observation matrix X, wherein the dimensionality reduction processing comprises the following steps:
(1) performing zero mean and unit variance processing on each column of the original observation matrix X to obtain a covariance matrix represented as:
through eigen decomposition, the covariance matrix can be further decomposed into:
(2) then one sample vector x can be projected into the principal component subspace and the residual subspace, respectively, as:
(3) decomposing the original observation matrix X into a score matrix T and a load matrix P according to the projection of the sample vector X; wherein, in the original observation matrix X, each row represents a sample, and the observation variable X i ＝[x i (1),x i (2),...,x i (N)]；
l represents the number of pivot elements; 1, 2.
S2: and extracting fault features in the score matrix and the load matrix by adopting a principal component analysis method based on reconstruction, and taking the extracted fault features as input samples of the weighted probability neural network.
In order to successfully identify fault characteristics, index SPE reflecting residual error space is adopted, and observation variable x is calculated i Is reconstructed to the contribution value
As a fault signature, the reconstructed contribution value based on the respective sample variables
And forming a fault characteristic data set.
The method can indicate real fault variables, and the fault feature extraction basis is as follows: aiming at the fault variable, the maximum amplitude value is reduced after the SPE index of the fault variable is reconstructed. And according to the response degree of each reconstructed variable, taking the variable with the maximum response as a main feature, and taking the feature and the reconstructed data group as extracted features for identifying the type of the fault.
In contribution analysis, the variable x is observed i Contribution to the indicator SPE
Expressed as:
wherein, the first and the second end of the pipe are connected with each other,
represents the ith column, direction and x in the unit matrix i The same;
representing a projection matrix;
representing a projection matrix
The ith diagonal element of (1).
S3: and training the weighted probability neural network through the fault characteristics to obtain a weighted probability neural network model.
And introducing a weighted probability neural network of a machine learning algorithm, carrying out detailed modeling on the equipment by combining with prior knowledge, describing the relation between the characteristics and the fault data type, and carrying out classification and identification on the fault. Class separability represents the ability to distinguish between various classes in a feature vector, usually as one of the performance criteria in a classification algorithm, i.e., a set of feature vectors can be passed through the classifier to achieve maximum class separability. However, the conventional probabilistic neural network assigns the same weight to all the mode layers, and cannot satisfy the maximization of class separability of different mode layers. The weighted probabilistic neural network is based on perfect statistical principles, derived from Bayesian decisions and nonparametric kernel density estimators, rather than heuristic methods. Due to the existence of the weight factors, the nodes of the summation layer are related to each node of the mode layer, the classification result can be ensured to approach the Bayesian optimal decision by selecting proper smooth factor values, and the classification precision of the classifier can be improved by introducing and updating the weight factors. The network structure of the weighted probabilistic neural network is shown in fig. 2.
The weighted probability neural network comprises a sample input layer, a mode layer, a summation layer and a decision output layer which are sequentially connected.
Sample input layer: inputting data to be classified, wherein the number of neurons of the data is the dimensionality of sample data after dimensionality reduction; the sample layer is associated with the mode layer by a gaussian function.
Mode layer: i.e., the radial basis layer, matches the feature vectors of the input data with the respective patterns, and calculates the distance between the input sample vector and the center vector of each neuron. The relationship between the sample vector x and the jth center vector of class i is expressed as:
where σ represents a smoothing factor; upsilon is ij Representing a neuron vector x ij The weighting factor takes into account the class separability of the different modes, and for high class separability (meaning better class discrimination), upsilon ij The ratio is large; for low class separability, upsilon ij The ratio is small. x ═ x 1 ,x 2 ,...,x m )；N i Represents class C i Total number of samples in (1).
And a summation layer: the number of summation nodes is equal to the class of the training sample, and the neuron calculates the output sum of the mode nodes of the same class and then performs weighted average.
A decision output layer: and based on Bayesian classification decision, deciding the class with the maximum posterior probability and outputting the class of the fault.
The application derives an analytic formula of a weighting factor of a weighted probability neural network based on sensitivity analysis, the weighting factor is used for reflecting the importance of a qth mode to a jth category between a mode layer and a summation layer, and the calculation process of the weighting factor comprises the following steps:
the sensitivity coefficient S of the neuron in all modes is calculated and expressed as:
wherein the content of the first and second substances,
representing the vector parameter of the jth class, the sensitivity parameter S of the jth class is represented as:
wherein S is j The elements in column r represent: and for a specific input mode q, calculating a sensitivity parameter of a kernel density estimation function of an r neuron corresponding to a j category in the weighted probability neural network. Due to S j Is a vector and therefore the gradient needs to be determined as follows:
aggregate all patterns Q and obtain a sensitivity vector, Q1 j Then the sensitive vector a j Expressed as:
a is to j The normalization of the elements in (a) results in the weight ω of each element in the jth class j Expressed as:
wherein, when n ═ infinity,
then w j Is normalized to [0,1 ]]The interval, at this time, the summation layer of the weighted probabilistic neural network is represented as:
wherein r represents a j Middle (r) th element, n 1 ,n 2 Not less than 1; when calculating weighted probabilistic neural network weight factors, n 1 ＝∞，n 2 ＝2
S4: and inputting the test sample set into the trained weighted probability neural network model to obtain a fault classification result.
Specifically, real-time fault data are input into a trained weighted probabilistic neural network model, fault categories are output accordingly, fault sources are determined, and classification results are transmitted into a DCS.
As a specific embodiment, the application records 70% of the data set used for constructing the training sample as Tr 1 The 30% data set used to construct the test samples is recorded as Te 1 . And (3) performing feature extraction by using a principal component analysis method based on reconstruction, acquiring a fault feature sample as a training sample of the classifier, inputting the fault feature sample into the weighted probability neural network classifier, and outputting a classification result.
To weight the necessity and effectiveness of probabilistic neural network methods to test set Te 1 One of the observations is an example. Firstly, obtaining the sample fault characteristics based on a reconstructed principal component analysis method, then inputting the sample fault characteristics into a weighted probability neural network classifier, outputting probability values belonging to all classes, and outputting the maximum probability class as a predicted fault class.
The results show that the method is able to correctly identify the fault condition and assign the highest probability value (1.0) to fault 3, which corresponds to an aspirator functional anomaly, as shown in table 2. By this factIn the barrier extraction process, F is obtained according to the response degree of the variable contribution value 16 Is the primary characteristic, the root cause of the fault can be determined. Based on the above information of the failure, a solution for resolving the failure is provided.
TABLE 2 probability calculation based on PNN classes
| Class of |
0 | 1 | 2 | 3 | 4 | 5 |
| |
0 | 0 | 0 | 1 | 0 | 0 |
To verify the accuracy of weighted probabilistic neural network based classification (WPNN), it was compared to the traditional Probabilistic Neural Network (PNN) and other five common classification models, including: support Vector Machines (SVM), Random Forests (RF), Decision Trees (DT), extreme random trees (ET), K Nearest Neighbor (KNN). All classification models are countedData set Tr 1 Training was performed and data set Te was used 1 The test was performed and the accuracy of the classification results is shown in table 3.
TABLE 3 Classification accuracy of six Classification methods
| WPNN | PNN | SVM | RF | DT | ET | KNN | |
| Accuracy of classification | 99.8％ | 98.93％ | 95.48％ | 90.38％ | 89.61％ | 92.45％ | 96.98％ |
As can be seen from Table 3, the comparison results of the 7 classes of classifiers show that the traditional probabilistic neural network method and the weighted probabilistic neural network method have good classification performance in the fault classification module and are superior to other types of artificial neural networks and classifiers. The weighted probabilistic neural network method can train an optimal classifier, because the weighted probabilistic neural network method basically has good performance of the traditional probabilistic neural network and can approximate the probability density function of a multi-element Gaussian function. In addition, the weighted probabilistic neural network includes weighting factors characterizing class separability, accessible to each of the pattern neurons' contribution to the network outcome.
Novel fault classification: in order to verify the effect of the weighted probabilistic neural network method on the classification of novel faults, a fifth type of fault sample is selected from an original training set Tr 1 Removing, constructing a new training data set Tr 2 Thus, fault 5 may be a new type of fault that does not exist in the training set. In test set Te 1 And taking a sample of the fifth type fault as a test sample.
TABLE 4 WPNN-based Category probability calculation
As can be seen from table 4, the probability assigned to the failure 3 is only 0.163, and cannot be directly used as a basis for determining the failure type, and therefore, it is necessary to determine the failure type based on the weight factor calculation result. The weighting factors are calculated through the weighted probability neural network, the proportion of the weighting factors of the 3 rd type fault sample is obtained and is shown in figure 3, and the proportion of the weighting factors of the third type fault of the fault test sample is shown in figure 4. The normal third type failure samples were weighted at 89.45%, while the failure test samples were weighted only 34.04% in the third type. The test sample weight calculation results do not exhibit significant class separability, and therefore the test sample is considered to be a new type of failure.
In the face of novel faults, the novel faults need to be extracted by combining fault characteristicsInformation on in-process fault variable isolation, i.e. F 3 Condenser water level corresponding to variable is main fault characteristic, F 25 、F 13 The condenser full water fault is considered as a secondary fault characteristic. However, if the condenser is full of water, the pressure variable of the outlet of the condenser does not have a corresponding large amplitude in reconstruction analysis, so that the full water of the condenser can only be considered as a fault symptom, but not a fault main characteristic. Selecting F with the largest secondary response amplitude 25 The variables serve as the failure master. The novel fault can be determined to be the fault of the condensate pump by combining the water level full sign of the condenser and the pressure change of the condenser system.
The foregoing is an exemplary embodiment of the present application, and the scope of the present application is defined by the claims and their equivalents.
Claims (5)
1. A fault classification method based on a weighted probability neural network is characterized by comprising the following steps:
s1: acquiring a training sample set and a testing sample set of faults based on historical fault data of the power station equipment, forming a sample matrix by the training sample set, and performing dimensionality reduction on the sample matrix to obtain a score matrix and a load matrix;
s2: extracting fault characteristics in the score matrix and the load matrix by adopting a principal component analysis method based on reconstruction, and taking the extracted fault characteristics as input samples of a weighted probability neural network;
s3: training the weighted probability neural network through the fault characteristics to obtain a weighted probability neural network model;
s4: and inputting the test sample set into the trained weighted probability neural network model to obtain a fault classification result.
2. The method for classifying a fault according to claim 1, wherein in step S1, a training sample set of a classifier is constructed, an original observation matrix X is obtained,
wherein, N is shown inIndicating the total number of samples, and each sample includes m observation variables, the original observation matrix X is represented as:
carrying out dimensionality reduction on the original observation matrix X, and decomposing the original observation matrix X into a score matrix T and a load matrix P; wherein, in the original observation matrix X, each row represents a sample, and the variable X is observed i ＝[x i (1),x i (2),...,x i (N)]；
l represents the number of pivot elements; 1,2, m.
3. The fault classification method according to claim 2, characterized in that in step S2, the observation variable x is calculated i Is reconstructed to the contribution value
Obtaining fault characteristics, and then reconstructing contribution values based on various sample variables
Forming a fault signature data set comprising: by observing the variable x i Contribution to the indicator SPE
Obtaining a reconstruction contribution value
Expressed as:
wherein the content of the first and second substances,
represents the ith column, direction and x in the unit matrix i The same;
representing a projection matrix;
representing a projection matrix
The ith diagonal element of (1).
4. The fault classification method according to claim 3, wherein in the step S2, the weighted probabilistic neural network includes a sample input layer, a mode layer, a summation layer, and a decision output layer which are connected in sequence;
the pattern layer is configured to match feature vectors of input data with patterns, and calculate a distance between an input sample and a center vector of each neuron, where a relationship between a sample vector x and an ith class jth center vector is expressed as:
where σ represents a smoothing factor; v is a cell ij Representing a neuron vector x ij The ratio of the "between class variance" and the "within class variance" of (c); x ═ x 1 ,x 2 ,...,x m )；N i Represents class C i Total number of samples in (1).
5. The method of fault classification according to claim 4, characterized in that a weighting factor of said weighted probabilistic neural network is calculated, the weighting factor being used to reflect the importance of the qth pattern to the jth class between the pattern layer and the summation layer, the calculation of the weighting factor comprising:
the sensitivity coefficient S of the neuron in all modes is calculated and expressed as:
wherein the content of the first and second substances,
representing the vector parameter of the jth class, the sensitivity parameter S of the jth class is represented as:
wherein S is j The elements in column r represent: for an input mode q, calculating a sensitivity parameter of a kernel density estimation function of an r-th neuron corresponding to a j-th category in a weighted probability neural network;
aggregate all patterns Q and obtain a sensitivity vector, Q1 j Then the sensitive vector a j Expressed as:
a is to be j The element in (1) is normalized to obtainWeight ω of each element in jth class j Expressed as:
wherein, when n ═ infinity,
then w j Is normalized to [0,1 ]]The interval, at this time, the summation layer of the weighted probabilistic neural network is represented as:
wherein r represents a j Middle (r) th element, n 1 ,n 2 Not less than 1; when calculating weighted probabilistic neural network weight factors, n 1 ＝∞，n 2 ＝2。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210543721.1A CN115017978A (en) | 2022-05-18 | 2022-05-18 | Fault classification method based on weighted probability neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210543721.1A CN115017978A (en) | 2022-05-18 | 2022-05-18 | Fault classification method based on weighted probability neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115017978A true CN115017978A (en) | 2022-09-06 |
Family
ID=83069512
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210543721.1A Pending CN115017978A (en) | 2022-05-18 | 2022-05-18 | Fault classification method based on weighted probability neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115017978A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819345A (en) * | 2023-08-29 | 2023-09-29 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Battery system fault identification method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6594620B1 (en) * | 1998-08-17 | 2003-07-15 | Aspen Technology, Inc. | Sensor validation apparatus and method |
| CN104142254A (en) * | 2014-08-06 | 2014-11-12 | 浙江大学 | Fault diagnosis method for flow passage of steam turbine |
| CN112199890A (en) * | 2020-10-11 | 2021-01-08 | 哈尔滨工程大学 | System-level fault diagnosis method for integrated nuclear power device |
-
2022
- 2022-05-18 CN CN202210543721.1A patent/CN115017978A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6594620B1 (en) * | 1998-08-17 | 2003-07-15 | Aspen Technology, Inc. | Sensor validation apparatus and method |
| CN104142254A (en) * | 2014-08-06 | 2014-11-12 | 浙江大学 | Fault diagnosis method for flow passage of steam turbine |
| CN112199890A (en) * | 2020-10-11 | 2021-01-08 | 哈尔滨工程大学 | System-level fault diagnosis method for integrated nuclear power device |
Non-Patent Citations (1)
| Title |
|---|
| YUCHEN DAI等: "A Novel Fault Diagnosis Method Integrating Reconstruction-based Principal Component Analysis with Probabilistic Neural Network on Condenser", 《2022 5TH INTERNATIONAL CONFERENCE ON ENERGY, ELECTRICAL AND POWER ENGINEERING (CEEPE)》, 24 April 2022 (2022-04-24), pages 496 - 501, XP034130120, DOI: 10.1109/CEEPE55110.2022.9783345 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819345A (en) * | 2023-08-29 | 2023-09-29 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Battery system fault identification method and device, electronic equipment and storage medium |
| CN116819345B (en) * | 2023-08-29 | 2023-11-21 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Battery system fault identification method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106355030B (en) | A kind of fault detection method based on analytic hierarchy process (AHP) and Nearest Neighbor with Weighted Voting Decision fusion | |
| CN110132598B (en) | Fault noise diagnosis algorithm for rolling bearing of rotating equipment | |
| CN106371427B (en) | Industrial process Fault Classification based on analytic hierarchy process (AHP) and fuzzy Fusion | |
| Eslamloueyan | Designing a hierarchical neural network based on fuzzy clustering for fault diagnosis of the Tennessee–Eastman process | |
| CN103914064B (en) | Based on the commercial run method for diagnosing faults that multi-categorizer and D-S evidence merge | |
| CN109086793A (en) | A kind of abnormality recognition method of wind-driven generator | |
| Zhang et al. | Fault detection and diagnosis of chemical process using enhanced KECA | |
| CN112200104B (en) | Chemical engineering fault diagnosis method based on novel Bayesian framework for enhanced principal component analysis | |
| CN111598150A (en) | Transformer fault diagnosis method considering operation state grade | |
| CN111580506A (en) | Industrial process fault diagnosis method based on information fusion | |
| Du et al. | Minimal Euclidean distance chart based on support vector regression for monitoring mean shifts of auto-correlated processes | |
| Tian et al. | Identification of abnormal conditions in high-dimensional chemical process based on feature selection and deep learning | |
| CN109409425B (en) | Fault type identification method based on neighbor component analysis | |
| CN112904810B (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
| CN109710661A (en) | Based on Global Genetic Simulated Annealing Algorithm to the method for high-pressure heater state analysis | |
| CN109240276B (en) | Multi-block PCA fault monitoring method based on fault sensitive principal component selection | |
| Jyoti et al. | Data clustering approach to industrial process monitoring, fault detection and isolation | |
| CN114757269A (en) | Complex process refined fault detection method based on local subspace-neighborhood preserving embedding | |
| CN115021679A (en) | Photovoltaic equipment fault detection method based on multi-dimensional outlier detection | |
| CN110427019B (en) | Industrial process fault classification method and control device based on multivariate discriminant analysis | |
| CN115017978A (en) | Fault classification method based on weighted probability neural network | |
| CN111242204A (en) | Operation and maintenance management and control platform fault feature extraction method | |
| CN114116832A (en) | Power distribution network abnormity identification method based on data driving | |
| CN107728476B (en) | SVM-forest based method for extracting sensitive data from unbalanced data | |
| CN117112992A (en) | Fault diagnosis method for polyester esterification stage |
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 |