CN109100648B - CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method - Google Patents

CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method Download PDF

Info

Publication number
CN109100648B
CN109100648B CN201810466453.1A CN201810466453A CN109100648B CN 109100648 B CN109100648 B CN 109100648B CN 201810466453 A CN201810466453 A CN 201810466453A CN 109100648 B CN109100648 B CN 109100648B
Authority
CN
China
Prior art keywords
impeller
current generator
ocean current
fault
characteristic
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
Application number
CN201810466453.1A
Other languages
Chinese (zh)
Other versions
CN109100648A (en
Inventor
王天真
温平平
刘卓
郑义来
李志超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Maritime University
Original Assignee
Shanghai Maritime University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Maritime University filed Critical Shanghai Maritime University
Priority to CN201810466453.1A priority Critical patent/CN109100648B/en
Publication of CN109100648A publication Critical patent/CN109100648A/en
Application granted granted Critical
Publication of CN109100648B publication Critical patent/CN109100648B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/34—Testing dynamo-electric machines
    • G01R31/343—Testing dynamo-electric machines in operation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
    • G01R19/165—Indicating that current or voltage is either above or below a predetermined value or within or outside a predetermined range of values
    • G01R19/16528—Indicating that current or voltage is either above or below a predetermined value or within or outside a predetermined range of values using digital techniques or performing arithmetic operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches

Abstract

The invention discloses a CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method, and the specific calculation process of the diagnosis method is shown in figure 1. Under the condition of winding faults of an impeller of the ocean current generator, the diagnosis method utilizes ARMA and CNN to respectively carry out feature extraction on a stator current sample and an impeller image sample of a single period of the ocean current generator and carry out convolution fusion and fusion feature distance calculation; and inputting all the obtained related winding fault characteristics into softmax for classification diagnosis. The invention carries out fault fusion diagnosis by extracting the characteristics of the stator current of the ocean current generator and the image of the ocean current generator, increases the efficiency of the diagnosis of the winding fault of the impeller of the ocean current generator, enhances the stability of the ocean current generator system and provides ideas for other occasions of image and voltage circuit data characteristic fusion.

Description

CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method
The technical field is as follows:
the invention relates to impeller winding fault diagnosis of an ocean current generator, in particular to a CNN-ARMA-Softmax-based impeller winding fault fusion diagnosis method of an ocean current generator.
Background art:
the ocean current power generation system is equipment for converting ocean current energy into electric energy, compared with common land equipment which works in a dry state, a clean state, a controllable temperature state and convenient maintenance state, the ocean current power generation set has higher accuracy requirements on fault detection and diagnosis, and the ocean current power generation set is influenced by the ocean environment to generate attachments, biological siltation, corrosion and rusting, so that multi-fault coupling is caused. Therefore, the method for detecting and diagnosing the fault of the ocean current power generation system under the unstable working condition is particularly important to research.
The invention content is as follows:
the invention aims to solve the problem that the impeller of the ocean current generator is wound, in order to find out that the impeller is wound by benthon in time and avoid large-area biological deposition of the impeller, stator current data and ocean current generator image data are simultaneously used as diagnostic information, the characteristic information of the winding fault of the impeller of the ocean current generator is extracted through a CNN algorithm aiming at image processing and an ARMA algorithm aiming at time period data and used as a convolution calculation kernel, the convolution kernel is carried out on the convolution kernel and the distance between characteristics with other period stator current data, finally, softmax regression is adopted, the probability of each category is output, the output has probability statistical significance, and the uncertainty problem is convenient to analyze.
In order to achieve the above purpose, the technical solution proposed by the present invention includes the following:
1. device for establishing simulation of winding fault of impeller of ocean current generator
The low-speed ocean current power generation system is utilized to simulate the underwater working state of the ocean current generator, and according to the biofouling process, the process that the impeller is attached by organisms and is propagated to large-area coverage and the process that the impeller is attached by the benthos and generates fouling gradually is simulated. According to the degree of attachment and coverage of the impeller by the benthos, namely the proportion (0-100%) of the area of the impeller surface of the ocean current generator covered by the benthos to the total area of the impeller, sequentially taking stator current data as analysis data to establish a fault and classification characteristic set;
2. extracting local period information and fault image local information of stator current
Extracting parts from unlabeled stator current data and ocean current generator image data to be used as characteristic extraction samples;
3. data pre-processing
Carrying out zero-mean processing on sample image data, and reducing the correlation among image pixels by ZCA whitening; meanwhile, low-pass filtering processing is carried out on the stator current data, and partial noise interference is reduced.
4. Extracting feature information
Extracting preprocessed stator current sample data by using an FFT algorithm, and recording main frequency information as current characteristics; and processing image sample data by using the CNN network, and taking the data of the full connection layer as image characteristics.
5. Convolution process for computing feature fusion
Carrying out convolution operation on the training sample, the extracted image and the current characteristic and storing the convolution characteristic;
6. calculating a characteristic distance
In order to reduce the data volume of the convolution characteristic and facilitate the operation of a classifier, the direct distance of the convolution characteristic is used for replacing the convolution characteristic.
7. Model training phase
And training the softmax classifier according to the convolution characteristic data after the distance characteristic is calculated so as to achieve a classification effect.
8. Real-time diagnostic phase
And (3) collecting a real-time test sample, performing convolution operation on the test picture and the features extracted by the fast FFT, calculating the distance between the features by adopting the same method in the step 6, inputting the distance into a trained softmax classifier, obtaining a probability output result of the real-time sample, and taking the class with the maximum probability as a diagnosis result of sample classification.
The invention has the technical effects that: the invention adopts a CNN-ARMA-SOFTMAX ocean current generator impeller winding fault fusion diagnosis strategy. Firstly, when the impeller attachment fault is processed, the image signal and the current signal are used as fault diagnosis data, the characteristics related to the impeller winding object fault are quickly extracted, local information can act on each periodic domain through convolution, meanwhile, the dimension of characteristic convolution can be reduced by utilizing the distance calculation among the characteristics, the fault degree analysis is conveniently carried out, and the characteristics of different data types are fused. The output of the softmax classifier is probability output, has the significance of probability statistics, and is convenient for analyzing uncertainty problems. The invention is also suitable for other calculation occasions needing to process images and voltage and current data.
Description of the drawings:
FIG. 1: winding fault fusion diagnosis scheme work flow chart
FIG. 2: actual graph of imbalance fault caused by impeller winding of ocean current generator
The specific implementation mode is as follows:
in order to make the technical means, the creation features and the achievement objects of the invention clearly visible, the invention is further explained below with reference to the specific drawings.
A fault fusion diagnosis strategy for impeller winding of an ocean current generator mainly comprises four parts, wherein image and current data are collected, data are preprocessed, features are extracted, diagnosis and classification are carried out, and the specific calculation process is shown in figure 1. The invention mainly combines the four parts to find the winding fault caused by the attachment of the impeller in time. The CNN-ARMA-SOFTMAX-based marine current generator impeller winding fault fusion diagnosis strategy provided by the invention is specifically described below.
Step 1: experiment platform for constructing attachment fault of impeller of ocean current generator
The low-speed ocean current power generation system is utilized to simulate the underwater working state of the ocean current generator, and according to the biofouling process, the process that the impeller is wound by the benthos and gradually generates fouling is simulated from the attachment and the multiplication of the impeller by the organisms to the large-area coverage. According to the degree of attachment and coverage of the impeller by the benthos, namely the proportion (0% -100%) of the area covered by the benthos on the surface of the impeller of the ocean current generator in the total area of the impeller, sequentially shooting color image information of the front of the impeller through an underwater sensor to serve as analysis data, establishing a fault and classification characteristic set, and acquiring impeller image data as shown in figure 2;
step 2: extracting local information
Extracting a local picture from the image without the label as a characteristic extraction sample;
extracting local period information from the unlabeled stator current data group as a feature extraction sample;
and step 3: data pre-processing
Carrying out zero-mean processing on sample image data, and reducing the correlation among image pixels by ZCA whitening; meanwhile, low-pass filtering processing is carried out on the stator current data, and partial noise interference is reduced.
And 4, step 4: extracting fault feature information
The method for predicting the winding fault characteristics of the ocean current generator based on the ARMA and the stator current comprises the following steps:
(1) using the current time t0The preceding time length is T0Establishing a characteristic parameter time sequence x (n) with a time interval delta T by the one-dimensional characteristic parameter data;
(2) carrying out model identification and parameter calibration by using the time sequence x (n) to obtain the order, autoregressive parameters and moving average parameters of the ARMA model; white noise is considered by the ARMA model by utilizing the model order and the parameters;
(3) according to the established ARMA model, evaluating the state representation capability of the model by utilizing the kurtosis value of actually measured stator current data;
(4) based on the established ARMA model, for t0Predicting the characteristic parameters subsequent to the moment, and predicting the characteristic parameters at each tiValue u of timeiAnd with a given decision threshold A1、A2Carrying out comparison; the judgment conditions of bearing failure and failure are ui＞A1、ui＞A2Thereby obtaining the time T for reaching the fault and failure state1、T2And correspondingly, the time t when the wind fault and the failure of the ocean current generator occur1、t2(t1，2＝t0+T1,2)；
(5) And according to the confidence coefficient, the predicted and actual characteristic parameter evolution trend, carrying out weighted calculation on the fault characteristic sequence group.
The marine current generator winding fault prediction based on the CNN and the marine current generator image comprises the following steps: extracting and processing sample data by using a convolution algorithm, and extracting main characteristics of the sample data as a convolution kernel according to the calculated winding characteristic distance;
and 5: convolution calculation fusing winding fault features
Performing convolution operation on the training sample and the extracted features and storing the convolution features, wherein f is the preprocessed picture, f (tau) represents a sub-matrix of the picture, g represents a convolution kernel, and the convolution features are obtained by translating the convolution kernel and multiplying the translation kernel by corresponding elements of the sub-matrix of the corresponding picture;
Figure GDA0002460628480000041
step 6: calculating the characteristic distance of winding fault
For the classifier to work, a winding fault characteristic distance calculation method is selected, wherein Kullback-L eibler (K L) divergence is used,
Figure GDA0002460628480000042
is a uniformly distributed vector of random numbers,
Figure GDA0002460628480000043
the ith value of the random number vector is represented by rho, which is a convolution characteristic, and the rho represents the ith characteristic value;
Figure GDA0002460628480000044
and 7: training softmax classifier
Utilizing the training characteristics Y obtained in the step 6trainTraining the softmax classifier and setting class labels F ∈ {1,2, …, k }, wherein 1 is the state of no attachment to the blade, 2,3, …, k is the degree of imbalance of the impeller covered by the attachment, initializing a parameter matrix of theta ∈ k × q, q is the number of neurons in the input layer of the softmax classifier, then the output of the softmax classifier is a first order probability matrix, which indicates that the sample belongs to class "k". The class label can be from 1 to k, and the system equation is as follows:
Figure GDA0002460628480000045
each row of the matrix can be taken as a classifier parameter corresponding to one classification label, and the total number is k rows; x(i)Training features X, y representing the ith input(i)Represents X(i)P denotes the probability form of the prediction, e denotes an exponential function, hθ(x(i)) Predicted probability results for k classes；hθ(x(i)) The cost parameters are as follows:
Figure GDA0002460628480000051
1{ g } is an indicative function whose value is 1 when the value of the equation in parenthesis is true, and 0 otherwise; the second term on the right side of the equation is a weight attenuation term which can change the original cost function into a strict convex function and ensure that a unique solution is obtained; m is the number of input training samples, λ2Is the attenuation coefficient of the attenuation term, j represents the unbalance degree of the attachment predicted as the j-th class, and thetaijRepresenting a jth class weight of an ith training sample; the expression of the cost function versus the partial derivative function of the parameter θ is as follows:
Figure GDA0002460628480000052
knowing a training sample, a cost function and a partial derivative thereof, and solving a parameter theta by using a steepest descent gradient iterative algorithm;
and 8: testing phase
And inputting the test convolution characteristics into a classifier to obtain the probability output of each sample, and taking the class corresponding to the result with the maximum probability as a final classification result.

Claims (1)

1. A CNN-ARMA-Softmax-based fusion diagnosis method for the winding fault of an impeller of an ocean current generator,
the method is characterized by comprising the following steps:
step 1: experimental platform for constructing fault of impeller winding of ocean current generator
Simulating the underwater working state of the ocean current generator by using a low-speed ocean current power generation system, and simulating the process that the impeller is wound by the organisms and grows to cover a large area according to the biofouling process, wherein the process that the impeller is wound by the benthos and gradually generates fouling; according to the degree of the impeller wound and covered by the benthos, namely the proportion of the area of the impeller surface of the ocean current generator wound by the benthos to the total area of the impeller is 0-100%, sequentially shooting color image information of the front surface of the impeller through an underwater sensor to serve as analysis data, and establishing a winding fault and classification characteristic set;
step 2: extracting local information
Extracting a local picture from the image without the label as a characteristic extraction sample;
extracting local period information from the unlabeled stator current data group as a feature extraction sample;
and step 3: data pre-processing
Carrying out zero-mean processing on sample image data, and reducing the correlation among image pixels by ZCA whitening; meanwhile, low-pass filtering processing is carried out on the stator current data, so that part of noise interference is reduced;
and 4, step 4: extracting fault feature information
The method for predicting the winding fault characteristics of the ocean current generator based on the ARMA and the stator current comprises the following steps:
(1) using the current time t0The preceding time length is T0Establishing a characteristic parameter time sequence x (n) with a time interval delta T by the one-dimensional characteristic parameter data;
(2) carrying out model identification and parameter calibration by using the time sequence x (n) to obtain the order, autoregressive parameters and moving average parameters of the ARMA model; white noise is considered by the ARMA model by utilizing the model order and the parameters;
(3) according to the established ARMA model, evaluating the state representation capability of the model by utilizing the kurtosis value of actually measured stator current data;
(4) based on the established ARMA model, for t0Predicting the characteristic parameters subsequent to the moment, and predicting the characteristic parameters at each tiValue u of timeiAnd with a given decision threshold A1、A2Carrying out comparison; the judgment conditions of bearing failure and failure are ui＞A1、ui＞A2Thereby obtaining the time T for reaching the fault and failure state1、T2And correspondingly the time t when the ocean current machine is in failure or failure1、t2Wherein t is1＝t0+T1,t2＝t0+T2；
(5) According to the confidence coefficient, the predicted and actual characteristic parameter evolution trend, a fault characteristic sequence group is calculated in a weighted mode;
the marine current generator winding fault prediction based on the CNN method and the marine current generator image comprises the following steps: extracting and processing sample data by using a convolution algorithm, and extracting main characteristics of the sample data as convolution kernels according to distances among different winding characteristics;
and 5: convolution calculation fusing winding fault features
Performing convolution operation on the training sample and the extracted features and storing the convolution features, wherein f is the preprocessed picture, f (tau) represents a sub-matrix of the picture, g represents a convolution kernel, and the convolution features are obtained by translating the convolution kernel and multiplying the translation kernel by corresponding elements of the sub-matrix of the corresponding picture;
Figure FDA0002492802020000021
step 6: calculating the characteristic distance of winding fault
In order to facilitate the operation of the classifier, a winding fault characteristic distance calculation method is selected, wherein Kullback-L eibler divergence is used,
Figure FDA0002492802020000022
is a uniformly distributed vector of random numbers,
Figure FDA0002492802020000023
is the ith value of the random number vector, and rho is the convolution characteristiciRepresents the ith characteristic value;
Figure FDA0002492802020000024
and 7: training softmax classifier
Utilizing the training characteristics Y obtained in the step 6trainTo train the softmax classifier and set a class label F ∈1,2, L, k, wherein 1 is the state of no attachment of the blade, 2,3, L, k is the degree of imbalance of the impeller covered by the attachment, a parameter matrix of theta ∈ k × q is initialized, q is the number of neurons in the input layer of the softmax classifier, the output of the softmax classifier is a first-order probability matrix, the probability represents that the sample belongs to the class "k", therefore, the class label thereof can be from 1 to k, and the system equation is as follows:
Figure FDA0002492802020000025
each row of the matrix can be taken as a classifier parameter corresponding to one classification label, and the total number is k rows; x(i)Training features X, y representing the ith input(i)Represents X(i)P denotes the probability form of the prediction, e denotes an exponential function, hθ(x(i)) Predicted probability outcomes for k classes; h isθ(x(i)) The cost parameters are as follows:
Figure FDA0002492802020000031
1{ g } is an indicative function whose value is 1 when the value of the equation in parenthesis is true, and 0 otherwise; the second term on the right side of the equation is a weight attenuation term which can change the original cost function into a strict convex function and ensure that a unique solution is obtained; m is the number of input training samples, λ2Is the attenuation coefficient of the attenuation term, j represents the unbalance degree of the attachment predicted as the j-th class, and thetaijRepresenting a jth class weight of an ith training sample; the expression of the cost function versus the partial derivative function of the parameter θ is as follows:
Figure FDA0002492802020000032
knowing a training sample, a cost function and a partial derivative thereof, and solving a parameter theta by using a steepest descent gradient iterative algorithm;
and 8: testing phase
And inputting the test convolution characteristics into a classifier to obtain the probability output of each sample, and taking the class corresponding to the result with the maximum probability as a final classification result.
CN201810466453.1A 2018-05-16 2018-05-16 CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method Active CN109100648B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810466453.1A CN109100648B (en) 2018-05-16 2018-05-16 CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810466453.1A CN109100648B (en) 2018-05-16 2018-05-16 CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method

Publications (2)

Publication Number Publication Date
CN109100648A CN109100648A (en) 2018-12-28
CN109100648B true CN109100648B (en) 2020-07-24

Family

ID=64796470

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810466453.1A Active CN109100648B (en) 2018-05-16 2018-05-16 CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method

Country Status (1)

Country Link
CN (1) CN109100648B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109029989B (en) * 2018-09-28 2020-02-14 西北工业大学 Gear box fault diagnosis method based on infrared thermal imaging
CN110109015B (en) * 2019-05-31 2020-06-26 中南大学 Asynchronous motor fault monitoring and diagnosing method based on deep learning
CN110863935B (en) * 2019-11-19 2020-09-22 上海海事大学 Method for identifying attached matters of blades of ocean current machine based on VGG16-SegUnet and dropout
CN113822417A (en) * 2021-09-22 2021-12-21 中国电建集团成都勘测设计研究院有限公司 Transformer fault type diagnosis method combining two machine learning methods
CN114487826B (en) * 2022-02-14 2022-12-06 爱科赛智能科技(浙江)有限公司 Motor starting locked rotor detection method based on current kurtosis
CN116930749B (en) * 2023-06-15 2024-02-13 浙江联大科技有限公司 System and method for detecting resistance of tubular motor

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103176128A (en) * 2013-03-28 2013-06-26 华南理工大学 Method and system for forcasting state of wind generating set and diagnosing intelligent fault
CN105590102A (en) * 2015-12-30 2016-05-18 中通服公众信息产业股份有限公司 Front car face identification method based on deep learning
CN106408024A (en) * 2016-09-20 2017-02-15 四川大学 Method for extracting lung lobe contour from DR image
CN106529555A (en) * 2016-11-04 2017-03-22 四川大学 DR (Digital Radiography) slice lung contour extraction method based on fully convolutional network
CN106529446A (en) * 2016-10-27 2017-03-22 桂林电子科技大学 Vehicle type identification method and system based on multi-block deep convolutional neural network
CN107153842A (en) * 2017-04-27 2017-09-12 西安交通大学 The fan blade diaphragm damage detecting method split based on edge
CN107256546A (en) * 2017-05-23 2017-10-17 上海海事大学 Ocean current machine blade attachment method for diagnosing faults based on PCA convolution pond SOFTMAX
CN107957551A (en) * 2017-12-12 2018-04-24 南京信息工程大学 Stacking noise reduction own coding Method of Motor Fault Diagnosis based on vibration and current signal

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103176128A (en) * 2013-03-28 2013-06-26 华南理工大学 Method and system for forcasting state of wind generating set and diagnosing intelligent fault
CN105590102A (en) * 2015-12-30 2016-05-18 中通服公众信息产业股份有限公司 Front car face identification method based on deep learning
CN106408024A (en) * 2016-09-20 2017-02-15 四川大学 Method for extracting lung lobe contour from DR image
CN106529446A (en) * 2016-10-27 2017-03-22 桂林电子科技大学 Vehicle type identification method and system based on multi-block deep convolutional neural network
CN106529555A (en) * 2016-11-04 2017-03-22 四川大学 DR (Digital Radiography) slice lung contour extraction method based on fully convolutional network
CN107153842A (en) * 2017-04-27 2017-09-12 西安交通大学 The fan blade diaphragm damage detecting method split based on edge
CN107256546A (en) * 2017-05-23 2017-10-17 上海海事大学 Ocean current machine blade attachment method for diagnosing faults based on PCA convolution pond SOFTMAX
CN107957551A (en) * 2017-12-12 2018-04-24 南京信息工程大学 Stacking noise reduction own coding Method of Motor Fault Diagnosis based on vibration and current signal

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于卷积神经网络（CNN）和CUDA加速的实时视频人脸识别;孔英会等;《科学技术与工程》;20161231;第16卷(第35期);第96-100、107页 *

Also Published As

Publication number Publication date
CN109100648A (en) 2018-12-28

Similar Documents

Publication Publication Date Title
CN109100648B (en) CNN-ARMA-Softmax-based ocean current generator impeller winding fault fusion diagnosis method
CN107256546B (en) Ocean current machine blade attachment fault diagnosis method based on PCA convolution pooling SOFTMAX
CN109580215B (en) Wind power transmission system fault diagnosis method based on deep generation countermeasure network
CN110261394A (en) Online fan blade damages real-time diagnosis system and method
CN106991666B (en) A kind of disease geo-radar image recognition methods suitable for more size pictorial informations
CN103530620A (en) Method for identifying bird nest on electric transmission line tower
CN113505664B (en) Fault diagnosis method for planetary gear box of wind turbine generator
CN108919116B (en) Ocean current generator unbalanced stator current fault diagnosis method based on MCCKAF-FFT-Softmax
CN114619292A (en) Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network
CN107680097A (en) A kind of method of electric power line pole tower identification Bird's Nest
CN111860593A (en) Fan blade fault detection method based on generation countermeasure network
CN114359695A (en) Insulator breakage identification method based on uncertainty estimation
CN113780242A (en) Cross-scene underwater sound target classification method based on model transfer learning
CN114863361A (en) Distribution line bird nest detection method based on USRNet and YOLOv3
CN108985385A (en) Based on the quick Weakly supervised object detection method for generating confrontation study
McLeay et al. Deep convolutional neural networks with transfer learning for waterline detection in mussel farms
CN110863935B (en) Method for identifying attached matters of blades of ocean current machine based on VGG16-SegUnet and dropout
CN117435956A (en) Electrochemical energy storage battery data analysis method based on IGWO-ADAM-BP neural network
CN117056678A (en) Machine pump equipment operation fault diagnosis method and device based on small sample
CN116843662A (en) Non-contact fault diagnosis method based on dynamic vision and brain-like calculation
CN115754713A (en) Motor fault diagnosis method based on frequency diagram and EFA-BP deep learning
Chen et al. Attention mechanism-based CNN for surface damage detection of wind turbine blades
CN111523557A (en) Wind power intelligent fault diagnosis method based on big data
Pahmi et al. Artificial neural network based forecasting of power under real time monitoring environment
CN115130731A (en) Method, equipment and medium for detecting fault of wind driven generator

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