CN107024331A - A kind of neutral net is to train motor oscillating online test method - Google Patents
A kind of neutral net is to train motor oscillating online test method Download PDFInfo
- Publication number
- CN107024331A CN107024331A CN201710208538.5A CN201710208538A CN107024331A CN 107024331 A CN107024331 A CN 107024331A CN 201710208538 A CN201710208538 A CN 201710208538A CN 107024331 A CN107024331 A CN 107024331A
- Authority
- CN
- China
- Prior art keywords
- layer
- input
- fault
- network
- determined
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
- G01M7/02—Vibration-testing by means of a shake table
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Molecular Biology (AREA)
- Neurology (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a kind of neutral net to train motor oscillating online test method, using six layers of convolutional neural networks, the sound characteristic for choosing vibration signal is used as failure symptom, weight is updated using LM algorithms and cross entropy, six layers of convolutional neural networks include input layer, hidden layer and output layer.Method of the invention based on neutral net is diagnosed to fault type, and fault type is recognized without setting up fault file database by machine oneself, is reduced sensor and is laid quantity, adds system reliability.Propulsion over time, data are continuously increased, and the fault type of machine self-identifying will more and more precisely, and test works well.
Description
Technical field
The invention belongs to nerual network technique field, and in particular to perception of the sensor to monitoring objective, based on nerve net
The failure of network recognizes mechanism and by complicated fault analysis and handling process automatically, more particularly to a kind of neutral net is to train electrical
Machine vibration online test method.
Background technology
Railway is important Strategic Foundation facility, the main artery of national economy and the popular vehicles of country, in synthesis
In key status in communications and transportation system, very important status is occupied in national economy.In recent years, national economy is fast
Fast development on railway transport proposes higher requirement.By high speed railway construction for many years and to rapid turn of existing circuit
Type is transformed, and China Express Railway operation mileage accounts for the 50% of world's high ferro operation mileage, occupies world's high ferro operation mileage
Half of the country., will be in following railway transport of passengers with 350km/h EMU under the fast-developing overall background of China railways
Special line is largely equipped, and its security directly influences the life security of passenger.It is used as the core of bullet train traction drive
Equipment, traction electric machine is when traction state is run, as " intermediate variable " that electric energy changes to mechanical energy is realized, produces train fortune
Capable power.In on-position, traction electric machine will complete EMUs regenerative braking force as generator.Traction electric machine due to
It is suspended on bogie, therefore is usually in in dust, the environment of environment temperature acute variation, and the frequent frequent hair of load
Changing, causes bullet train operating mode constantly to change.Due to severe working environment and special structure, bullet train is caused to lead
Draw that driving unit fault is occurred frequently, it runs the traffic safety that safety directly influences whole bullet train, once occur can not be pre-
The failure known will likely cause great personnel casualty accidentses, produce huge economic losses and social influence.Therefore, height is actively developed
The research of fast train detection technology and diagnostic techniques, not only contributes to the safe and reliable operation of EMUs, while will also be at a high speed
The support that provides the necessary technical is reformed in the maintenance of train.It is to work as that condition monitoring and fault diagnosis is carried out to bullet train traction electric machine
Modern railway development hot and difficult issues urgently to be resolved hurrily.Traditional equipment fault diagnosis includes setting up fault file and status information
Storehouse, its diagnosis algorithm is divided into signal detection, fault signature extraction, equipment state identification and forecast decision-making etc., wherein：
Signal detection：According to diagnostic device and target, selection is easy to the significant condition signal of monitoring and collection, thus set up
Fault status information storehouse, belongs to initial pattern.
Fault signature is extracted：By the fault status signal of collection by signal transacting, fault signature is extracted, and is Fault Identification
Lay the foundation.
Equipment state is recognized：Use database technology by theory analysis and with reference to previous failures set up fault file storehouse for
Whether reference mode, broken down by pattern relatively more to be checked with reference mode come diagnostic device.
Forecast decision-making：By diagnostic analysis, if equipment state normally repeats procedure above；Otherwise to there is malfunction
, then failure situation is found out, trend analysis is made.
Some is it can be found that traditional fault diagnosis relies on substantial amounts of different types of data more than, and according to known
Various fault types these data are classified into fault file database.This needs the sensor of a large amount of monitoring different physical quantities
With known all fault types.These conditions are only met, state recognition can be just carried out, and analyze whether equipment occurs
Failure or the fault type of generation, the decision-making of diagnosis is completed finally according to fault type.Reality is that people can not can know that
Whole fault types, which results in the appearance of some abnormal datas, and system has no idea to carry out data state recognition,
The state of None- identified certainly will influence the decision-making of diagnosis, and then report by mistake.In order to reduce rate of false alarm, it must just increase more
Sensor carrys out diversification identification failure, and more sensor collections can also increase rate of false alarm to more data.Therefore it is traditional
Fault diagnosis need substantial amounts of sensor, and big quantity sensor can reduce the reliability of total system on the contrary.Asked based on above-mentioned
Topic, bullet train is badly in need of a kind of monitoring method, the requirement to number of sensors is reduced, while needing not know about all failure classes
Type.
The content of the invention
The method based on neutral net of the invention is diagnosed to failure, by machine oneself identification fault type without
Fault file database is set up, and then greatly reduces the quantity that sensor is laid, reliability is improved, and with the propulsion of time,
Data are continuously increased, and the fault type of machine self-identifying more and more precisely will solve conventional monitoring methods rate of false alarm height, sensor
The problem of node is more, the overall maintenance cost of reduction and maintenance time, maintenance personal is helped quick and precisely to find failure parts.
The present invention is using improved neural network algorithm (Levenberg-Marquardt).The algorithm be gradient descent method and
The combination of gauss-newton method, the local convergence characteristic of existing gauss-newton method, there is the global property of gradient method again.With regard to frequency of training
For precision, the second dervative information approximate due to make use of, this algorithm is substantially better than gradient descent method standard.LM algorithms are
A kind of fast algorithm of the numerical optimization technique of utilization standard.It need not calculate Hessian matrixes as quasi-Newton method.
The technical scheme is that, a kind of neutral net is to train motor oscillating online test method, it is characterised in that
Using six layers of convolutional neural networks, the sound characteristic of vibration signal is chosen as failure symptom, LM algorithms and cross entropy, institute is used
Stating six layers of convolutional network includes input layer, hidden layer and output layer；It the described method comprises the following steps：
1) input layer (M) number is determined, makes P represent the input sample vector of network,
2) hidden layer neuron (J) number is determined, is rule of thumb chosen,
3) output layer neuron number is determined, is determined by fault type,
4) network function is determined,
5) tested with partial data, build 24-500-500-500-1024-4 convolutional neural networks,
6) input data, carries out neural network computing.
The present invention is using six layers of convolutional neural networks：The unit number of input layer is 24, and input data is 20*800 frequency squares
Battle array.The unit number of output layer is 4,4 kinds of typical faults of correspondence, i.e. normal operation, stator failure, rolling bearing fault and rotor
Failure.The unit number of hidden layer is respectively 500-500-500-1024, and using Cross-Entropy Algorithm and LM innovatory algorithms, network is entered
Row training.Global error E=10 is set in convolutional neural networks learning process-5, the initial value of neutral net set, weights are repaiied
Positive quantity is randomly generated by being uniformly distributed within the specific limits.Step 4) described in network function include neuron train function LM,
Learning function LM and cross entropy, limit the function min and max of input vector element threshold range.Step 5) described in convolution god
Include input port input through network, every time one 20*800 of input input matrix；The weights module of input layer-hidden layer
ω.Step 6) described in neural network computing update weight using LM algorithms and Cross-Entropy Algorithm.
The LM algorithms include,
1) training error permissible value, initialization weight vector W are provided(o),
2) calculating network output and error vector E (W(o)),
3) the vectorial J (W) to network weight of calculation error,
4) local minimum points W is reachedk=W (k).
The LM algorithms computational methods are as follows：
If wk∈RnRepresent the network weight vector of kth time iteration, new weight vector wk+1It can be asked according to following rule
:
wk+1=wk+Δwk
In formula, t, o is respectively the output of network output layer and desired output (this section is as follows), and E (w) is error energy letter
Number.This formula is single output network, if the network of multi output then need to only be added up to output layer square-error, that is, add up item
From m to m*n, ξ represents weight vector w the ξ element.Jacob matrix Js (w) are:
If ξ represents weight vector w the ξ element, for the ξ ξ elements of J matrixesIf this ξ
Element represents the weights v between output node i and hidden layer section jijPartial derivative is sought, is obtained
di=-(oi-ti)·ti·(1-ti)
In formula, bjRepresent the output of hidden layer corresponding node.
If the ξ element represents output node i value, partial derivative is：
If the ξ element represents the weight of input layer and hidden layer, partial derivative is：
During the above is various, VjiThe weights of output layer are represented, a represents the input of input layer corresponding node.If the ξ element
The threshold value of hidden layer is represented, then
Cross-Entropy Algorithm is as follows：
Cross entropy is relative with entropy, such as covariance and variance.
What entropy was investigated is the expectation of single information (distribution)：
What cross entropy was investigated is the expectation of two information (distribution)：
Cross entropy cost function：
xPrimary signal is represented, z represents reconstruction signal, represent that length is d in the form of vectors, can be changed easily again
Make the form for inner product of vectors.
Cross entropy cost function is introduced for neutral net, is easily occurred to make up the derivative form of sigmoid type functions
The defect of saturation.Decline this problem to solve parameter renewal efficiency, use cross entropy cost function to replace traditional square
Error function.
For the neuronal structure of multiple input single output, as shown in Figure 3：
Its loss function is defined as：
Wherein
Final derivation is obtained：
Avoid σ ' (z) and participate in the problem of parameter updates, influence updates efficiency.
The LM algorithm flow charts are as shown in Figure 1.24 groups of voice datas are inputted, feature extraction is carried out by 5*5 convolutional layers,
Line activating is entered by Relu.Simplify data volume by maximum pond layer, entering second 5*5 convolutional layer through Relu activation carries out spy
Levy extraction.Maximum pond is carried out afterwards and 1 dimension matrix is converted into, and as the input of full articulamentum grader, finally exports 4 kinds of events
Hinder the probability of type.In training, it can be contrasted according to output result with model answer, pass through LM algorithms and Cross-Entropy Algorithm
Carry out assessment of loss function LOSS, and each layer weight is updated according to result backpropagation stomogastric nerve network.Final training result such as Fig. 2
It is shown.
The present invention chooses the sound characteristic of vibration signal as failure symptom, and interception voice signal 20*800 is used as feature
Amount, its corresponding spectrum peak is used as failure symptom, and the typical fault conduct that 4 kinds of scenes more often occur after normalized
Network is exported, and constitutes one six layers of convolutional neural networks, and its input feature value is pre-processed for the sample data collected
After obtain.
The effect of the present invention, the method based on convolutional neural networks is diagnosed to fault type, is recognized by machine oneself
Fault type reduces sensor and lays quantity, add system reliability without setting up fault file database.With
The propulsion of time, data are continuously increased, and the fault type of machine self-identifying will be more and more accurate.
Brief description of the drawings
Fig. 1 is convolutional neural networks Organization Chart of the present invention.
Fig. 2 is neural network computing error result figure of the present invention.
Fig. 3 is the neuronal structure figure of multiple input single output.
Embodiment
Below in conjunction with the accompanying drawings, the present invention is described in more detail.
A kind of neutral net to train motor oscillating online test method,
The structure for defining network first is：Single input layer, single output layer, 4 hidden layers.
Concrete operations are as follows：
1) input layer (M) number is determined, makes P represent the input sample vector of network, input layer needs 24 god
Through member, i.e. M=24.
2) hidden layer neuron (J) number is determined,
Due to the method that the determination neither one of hidden layer neuron number is fixed, chosen generally according to experience, also one
The approach of kind can be used for the number for determining hidden unit.Make the number of hidden unit variable first, or be put into enough hidden units, lead to
Cross study to reject those inoperative hidden units, untill it can not reject.Equally, it can also be put into when training and starting
Fewer neuron, study is arrived after certain number of times, and the number of hidden unit is further added by if unsuccessful, compares conjunction until reaching
Untill the hidden unit number of reason.
3) output layer neuron number is determined
The neuron number of output layer is determined that the present invention devises normal operation, stator failure, rolling by fault type
Four kinds of fault types such as dynamic bearing failure, rotor fault, output layer needs 4 outputs, and the target output of network is as shown in the table：
| Fault type | Target is exported |
| Run well | 0 0 |
| Stator failure | 0 1 |
| Rolling bearing fault | 1 0 |
| Rotor fault | 1 1 |
It can thus be concluded that object vector is 4 × 2 matrix T, i.e. T=[0 0,01,10,1 1].
4) network function is determined
The hidden layer neuron training function of network is LM and cross entropy, and the speed of service of the function is than very fast, for big
Medium-sized network is relatively adapted to；Learning function takes the cross entropy of acquiescence；
5) tested with partial data, 24-500-500-500-1024-4 neutral net is built first, wherein
Input is input port, and the input matrix of one 20 × 800 is inputted every time；Step is step-length module, and ω is the weights mould of hidden layer
Block, weight is updated by LM and cross entropy come backpropagation.
Increased income deep learning framework tensorflow using Google, relevant parameter is defined according to above-mentioned neural network structure.
Input node parameter is defined first, according to the voice signal collected【20,800】It is converted into the one-dimensional matrix of dimension【1200】, so
Input node is 1200.Output node is categorized as 4 classes according to fault type, and output node is 4.Hidden according to operational capability setting
Node layer is 500.
Input_node=1200
Output_node=4
H_node=500
It is variable by tensorflow function sets weight and offset, these changes is created when training neutral net
Amount, the value of these variables is loaded in test by the model of preservation, also needs the regularization of variable in function losing addition
Into loss functions.
When setting weight, the value of associated weight is initialized, is allowed to meet god in Gauss normal distribution, reduction optimization process
The possibility of local optimum is absorbed in through network.
Def get_weight_Varibale(shape,regularizer):
Weights=tf.get_variable (' weights ', shape, initializer=tf.truncated_
Normal_initializer (stddev=0.1))
The forward direction transmittance process of neutral net is defined, hidden layer weight matrix dimension is by input node and hiding node layer
Count to determine, and regularization.Deviant is set as constant, and dimension is determined by hiding node layer, therefore is one-dimensional matrix.Afterwards,
Matrix multiplication calculating is carried out by input parameter weight, the overall value progress relu of deviant is added and enters line activating.
Weight=tf.Variable ([5,5, input_node, h_node], regularizer)
Biases=tf.Variable ([h_node])
H_layer=tf.nn.relu (tf.matmul (input_tensor, weights)+biases))
Neutral net propagated forward algorithm defined in this section of code, either training or tests, can directly adjust
With.Next the amendment of weight in back-propagation process is defined.
It is once incoming to be limited by the operational capability and storage capacity of machine, it is therefore desirable to right because data volume is larger
Whole data set is limited, and 100 groups of data are only read in every time, is read in next group again after the completion of training every time, is circulated altogether
30000 groups.
Batch_size=100
Training_step=30000
The training process of loss function, learning rate and correlation is defined, learning rate gradually drops according to the increase of training time
Low, preventing can not convergent situation generation.Loss function is calculated using cross entropy and LM algorithms, and each group of friendship is calculated afterwards
Fork entropy result is simultaneously averaged.
TrainLM=tf.nn.mul (array (LMe, LMa**k))
LMe=tf.nn.matmul (d, y, b_end)
B_end=tf.nn.mul (b (1-b))
Variable_averages=tf.train.ExponentialMovingAverage (learning_decay,
blobal_step)
Cross_entropy=tf.nn.sparse_softmax_cross_entropy_with_lo gits (y,
tf.argmax(y_,1))
Cross_entropy_mean=tf.reduce_mead (cross_entropy)
Loss=cross_entropy_mean+tf.add_n (tf.get_collection (' losses '))+
TrainLm
The majorized function of each layer of transmission is defined, optimizes whole neutral net using gradient descent method, evaluation function is for it
The loss function of preceding definition.
Train_step=tf.train.GradientdescentOptimizer (learning_rate) .minimize
(loss,global_step)。
Input data, carries out neural network computing, and error result is as shown in Figure 2.
As can be seen from Figure this algorithm in the early stage when convergence it is very fast, gradually tended towards stability after error drops to 0.1, no
Cross the error at 450-500 milliseconds and level off to 0, illustrate that the system reaches stabilization, test works well, to neutral net here
It is trained to finish, and can be with relatively good prediction fault type.
The convolutional neural networks Organization Chart is as shown in Figure 1:
24 groups of voice datas are inputted, feature extraction is carried out by 5*5 convolutional layers, line activating is entered by Relu.By maximum
Pond layer simplifies data volume, and entering second 5*5 convolutional layer through Relu activation carries out feature extraction.Maximum pondization is carried out afterwards simultaneously
1 dimension matrix is converted into, as the input of full articulamentum grader, the probability of 4 kinds of fault types is finally exported.In training, meeting
Contrasted according to output result with model answer, by LM algorithms and Cross-Entropy Algorithm come assessment of loss function LOSS, and root
Each layer weight is updated according to result backpropagation stomogastric nerve network.Final training result is as shown in Figure 2.
Technical scheme is described in detail above.It is apparent that the present invention is not limited described is interior
Hold.Based on the above in the present invention, those skilled in the art can also make a variety of changes, but any and sheet accordingly
Invention is equivalent or similar change belongs to the scope of protection of the invention.
Claims (9)
1. a kind of neutral net is to train motor oscillating online test method, it is characterised in that using six layers of convolutional neural networks,
The sound characteristic of vibration signal is chosen as failure symptom, using LM algorithms and cross entropy, six layers of convolutional network includes defeated
Enter layer, hidden layer and output layer；It the described method comprises the following steps：
1) input layer (M) number is determined, makes P represent the input sample vector of network,
2) hidden layer neuron (J) number is determined, is rule of thumb chosen,
3) output layer neuron number is determined, is determined by fault type,
4) network function is determined,
5) tested with partial data, build 24-500-500-500-1024-4 convolutional neural networks,
6) input data, carries out neural network computing.
2. method according to claim 1, it is characterised in that the input layer is individual layer, and the hidden layer is 4 layers, described
Output layer is individual layer.
3. method according to claim 1, it is characterised in that step 1) described in input layer (M) number be M=
24, output layer neuron number is 4.
4. method according to claim 1, it is characterised in that step 3) described in fault type to run well, stator therefore
Barrier, rolling bearing fault, 4 kinds of fault types of rotor fault.
5. method according to claim 1, it is characterised in that step 4) described in network function include neuron and train function
LM, learning function LM and cross entropy, limit the function min and max of input vector element threshold range.
6. method according to claim 1, it is characterised in that step 5) described in convolutional neural networks include input port
Input, every time one 20*800 of input input matrix；The weights module ω of input layer-hidden layer.
7. method according to claim 1, it is characterised in that step 6) described in neural network computing use LM algorithms and friendship
Pitch entropy algorithm and update weight.
8. method according to claim 7, it is characterised in that the LM algorithms include,
1) training error permissible value, initialization weight vector W are provided(o),
2) calculating network output and error vector E (W(o)),
3) the vectorial J (W) to network weight of calculation error,
4) local minimum points W is reachedk=W (k).
9. method according to claim 1, it is characterised in that the unit number of the input layer is 24, correspondence 20*800 sound
Frequecy characteristic value, the unit number of the output layer is 4,4 kinds of typical faults of correspondence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710208538.5A CN107024331B (en) | 2017-03-31 | 2017-03-31 | A kind of neural network is to train motor oscillating online test method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710208538.5A CN107024331B (en) | 2017-03-31 | 2017-03-31 | A kind of neural network is to train motor oscillating online test method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107024331A true CN107024331A (en) | 2017-08-08 |
| CN107024331B CN107024331B (en) | 2019-07-12 |
Family
ID=59526721
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710208538.5A Expired - Fee Related CN107024331B (en) | 2017-03-31 | 2017-03-31 | A kind of neural network is to train motor oscillating online test method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107024331B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108230121A (en) * | 2018-02-09 | 2018-06-29 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of product design method based on Recognition with Recurrent Neural Network |
| CN108280746A (en) * | 2018-02-09 | 2018-07-13 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of product design method based on bidirectional circulating neural network |
| CN108304960A (en) * | 2017-12-29 | 2018-07-20 | 中车工业研究院有限公司 | A kind of Transit Equipment method for diagnosing faults |
| CN108596470A (en) * | 2018-04-19 | 2018-09-28 | 浙江大学 | A kind of power equipments defect text handling method based on TensorFlow frames |
| CN108899048A (en) * | 2018-05-10 | 2018-11-27 | 广东省智能制造研究所 | A kind of voice data classification method based on signal Time-frequency Decomposition |
| CN109816094A (en) * | 2019-01-03 | 2019-05-28 | 山东省科学院海洋仪器仪表研究所 | Optical dissolved oxygen sensor non-linear temperature compensation method based on neural network L-M algorithm |
| CN111397700A (en) * | 2020-03-02 | 2020-07-10 | 西北工业大学 | Wall-mounted fault detection method of Coriolis mass flow meter |
| CN112464972A (en) * | 2019-08-12 | 2021-03-09 | 美光科技公司 | Predictive maintenance of automotive powertrains |
| CN112710486A (en) * | 2019-10-24 | 2021-04-27 | 广东美的白色家电技术创新中心有限公司 | Equipment fault detection method, equipment fault detection device and computer storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334936B (en) * | 2018-01-30 | 2019-12-24 | 华中科技大学 | Fault prediction method based on migration convolutional neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE4124501A1 (en) * | 1991-07-24 | 1993-01-28 | Dieter Prof Dr Ing Barschdorff | Neuronal network esp. for testing multiple-phase electric motor - has three neuronal layers for classifying input attribute vectors using class function |
| JPH11237432A (en) * | 1998-02-24 | 1999-08-31 | Fujikura Ltd | Partial discharge discrimination method |
| CN104680233A (en) * | 2014-10-28 | 2015-06-03 | 芜湖杰诺瑞汽车电器系统有限公司 | Wavelet neural network-based engine failure diagnosing method |
| CN105510038A (en) * | 2015-12-31 | 2016-04-20 | 北京金风科创风电设备有限公司 | Wind turbine generator fault monitoring method and device |
-
2017
- 2017-03-31 CN CN201710208538.5A patent/CN107024331B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE4124501A1 (en) * | 1991-07-24 | 1993-01-28 | Dieter Prof Dr Ing Barschdorff | Neuronal network esp. for testing multiple-phase electric motor - has three neuronal layers for classifying input attribute vectors using class function |
| JPH11237432A (en) * | 1998-02-24 | 1999-08-31 | Fujikura Ltd | Partial discharge discrimination method |
| CN104680233A (en) * | 2014-10-28 | 2015-06-03 | 芜湖杰诺瑞汽车电器系统有限公司 | Wavelet neural network-based engine failure diagnosing method |
| CN105510038A (en) * | 2015-12-31 | 2016-04-20 | 北京金风科创风电设备有限公司 | Wind turbine generator fault monitoring method and device |
Non-Patent Citations (2)
| Title |
|---|
| 刘亚军: "基于小波分析和神经网络的电机故障诊断研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
| 王歆峪: "基于神经网络的电机故障诊断", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304960A (en) * | 2017-12-29 | 2018-07-20 | 中车工业研究院有限公司 | A kind of Transit Equipment method for diagnosing faults |
| CN108230121A (en) * | 2018-02-09 | 2018-06-29 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of product design method based on Recognition with Recurrent Neural Network |
| CN108280746A (en) * | 2018-02-09 | 2018-07-13 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of product design method based on bidirectional circulating neural network |
| CN108230121B (en) * | 2018-02-09 | 2022-06-10 | 艾凯克斯（嘉兴）信息科技有限公司 | Product design method based on recurrent neural network |
| CN108280746B (en) * | 2018-02-09 | 2022-05-24 | 艾凯克斯（嘉兴）信息科技有限公司 | Product design method based on bidirectional cyclic neural network |
| CN108596470A (en) * | 2018-04-19 | 2018-09-28 | 浙江大学 | A kind of power equipments defect text handling method based on TensorFlow frames |
| CN108899048A (en) * | 2018-05-10 | 2018-11-27 | 广东省智能制造研究所 | A kind of voice data classification method based on signal Time-frequency Decomposition |
| CN109816094A (en) * | 2019-01-03 | 2019-05-28 | 山东省科学院海洋仪器仪表研究所 | Optical dissolved oxygen sensor non-linear temperature compensation method based on neural network L-M algorithm |
| CN112464972A (en) * | 2019-08-12 | 2021-03-09 | 美光科技公司 | Predictive maintenance of automotive powertrains |
| WO2021077567A1 (en) * | 2019-10-24 | 2021-04-29 | 广东美的白色家电技术创新中心有限公司 | Device failure detection method, device failure detection apparatus and computer storage medium |
| CN112710486B (en) * | 2019-10-24 | 2022-01-25 | 广东美的白色家电技术创新中心有限公司 | Equipment fault detection method, equipment fault detection device and computer storage medium |
| CN112710486A (en) * | 2019-10-24 | 2021-04-27 | 广东美的白色家电技术创新中心有限公司 | Equipment fault detection method, equipment fault detection device and computer storage medium |
| CN111397700A (en) * | 2020-03-02 | 2020-07-10 | 西北工业大学 | Wall-mounted fault detection method of Coriolis mass flow meter |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107024331B (en) | 2019-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107024331A (en) | A kind of neutral net is to train motor oscillating online test method | |
| CN105550700B (en) | A kind of time series data cleaning method based on association analysis and principal component analysis | |
| CN104792530B (en) | Deep-learning rolling bearing fault diagnosis method based on SDA (stacked denoising autoencoder) and Softmax regression | |
| CN109829236A (en) | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction | |
| CN106682688A (en) | Pile-up noise reduction own coding network bearing fault diagnosis method based on particle swarm optimization | |
| CN107066759A (en) | A kind of Vibration Fault Diagnosis of Turbine Rotor method and device | |
| CN109242147A (en) | Signal fused fan condition prediction technique based on Bp neural network | |
| CN106874963B (en) | A kind of Fault Diagnosis Method for Distribution Networks and system based on big data technology | |
| CN108875918A (en) | It is a kind of that diagnostic method is migrated based on the mechanical breakdown for being adapted to shared depth residual error network | |
| CN109948833A (en) | A kind of Hydropower Unit degradation trend prediction technique based on shot and long term memory network | |
| CN109597401A (en) | A kind of equipment fault diagnosis method based on data-driven | |
| CN103810328A (en) | Transformer maintenance decision method based on hybrid model | |
| CN106779063A (en) | A kind of hoist braking system method for diagnosing faults based on RBF networks | |
| CN110046379A (en) | A kind of structure entirety damnification recognition method based on space-frequency information | |
| CN112200263B (en) | Self-organizing federal clustering method applied to power distribution internet of things | |
| CN112747924A (en) | Bearing life prediction method based on attention mechanism and residual error neural network | |
| CN115392333A (en) | Equipment fault diagnosis method based on improved end-to-end ResNet-BilSTM dual-channel model | |
| Al Tobi et al. | Using MLP‐GABP and SVM with wavelet packet transform‐based feature extraction for fault diagnosis of a centrifugal pump | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| CN112148997A (en) | Multi-modal confrontation model training method and device for disaster event detection | |
| CN113887729A (en) | Fault diagnosis method for low-voltage power line carrier communication system | |
| CN115600136A (en) | High-voltage bushing fault diagnosis method, system and medium based on multiple sensors | |
| CN113221946B (en) | Method for diagnosing fault types of mechanical equipment | |
| CN105930623A (en) | Multi-level electromechanical system reliability prediction method based on fuzzy judgment | |
| CN111325233A (en) | Transformer fault detection method and device |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190712 |