CN114298087A - Attention mechanism-based hybrid CNN-LSTM dropper anomaly detection method - Google Patents
Attention mechanism-based hybrid CNN-LSTM dropper anomaly detection method Download PDFInfo
- Publication number
- CN114298087A CN114298087A CN202111469577.3A CN202111469577A CN114298087A CN 114298087 A CN114298087 A CN 114298087A CN 202111469577 A CN202111469577 A CN 202111469577A CN 114298087 A CN114298087 A CN 114298087A
- Authority
- CN
- China
- Prior art keywords
- dropper
- lstm
- cnn
- mixed
- attention mechanism
- 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
- 230000007246 mechanism Effects 0.000 title claims abstract description 38
- 238000001514 detection method Methods 0.000 title claims abstract description 31
- 230000001133 acceleration Effects 0.000 claims abstract description 46
- 230000005856 abnormality Effects 0.000 claims abstract description 9
- 238000013135 deep learning Methods 0.000 claims abstract description 7
- 238000000034 method Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 230000009471 action Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 238000003745 diagnosis Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 abstract description 5
- 230000007774 longterm Effects 0.000 abstract 1
- 238000012423 maintenance Methods 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241001391944 Commicarpus scandens Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003137 locomotive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a mixed CNN-LSTM dropper abnormality detection method based on an attention mechanism, belonging to the technical field of automatic detection of electrified railways. The invention provides a mixed CNN-LSTM dropper abnormality detection method based on an attention mechanism, which comprises a one-dimensional convolution part and an LSTM part, wherein the one-dimensional convolution part extracts local space characteristics of signal data, the LSTM part extracts long-term dependence characteristics of the signal data, the attention mechanism is further introduced, important channel signal information influencing abnormality detection is focused, and finally abnormality detection of a catenary dropper is realized. The invention fully considers the characteristics of the vibration acceleration signal data of the contact net, improves the deep learning network, has higher detection precision and stronger robustness on the breakage of the dropper of the contact net, and thus improves the working efficiency of dropper maintenance and repair.
Description
Technical Field
The invention belongs to the technical field of automatic detection of electrified railways, and particularly relates to a hybrid CNN-LSTM dropper abnormality detection method based on an attention mechanism.
Background
The pantograph-catenary system formed by the contact network and the pantograph is an important factor influencing the stable current collection of the electric locomotive, the contact network is laid in the open along a high-speed line, is influenced by natural factors such as external environment, climate and the like and the huge impact action of the pantograph for a long time, is a weak link of the electrified railway, and the faults of the pantograph-catenary-system account for more than 90% of the total number of the traction power supply system. The dropper is a key part of a contact net, is a bridge between a contact line and a carrier cable, is easy to break down under the action of external force and external severe environment, and seriously influences the running safety of a train.
The traditional contact net dropper abnormity detection is mainly a manual inspection method, and a technician is relied on to check on the spot or check a shot image, so that the defects of high detection cost, high false detection rate, poor timeliness and the like are overcome, and the high requirement of safe operation of a train is difficult to meet. Therefore, how to utilize machine learning and artificial intelligence technology to realize the detection of string hanging abnormity on vibration signal data acquired by the high-speed rail 6C system is a technical problem which is very important for the railway department. The early detection of the catenary dropper abnormality is mainly based on a signal processing technology, the characteristics of vibration acceleration signals collected by a high-speed rail 6C system are extracted, fault recognition is carried out after the characteristics are extracted, and the core is data characteristic extraction and pattern recognition classification. The manual feature extraction needs to analyze the operation mechanism of the system, select a proper signal analysis method, and if the selection is not proper, important features may be lost, so that the final diagnosis accuracy is low. The method overcomes the defect of manual feature extraction, and has important research value in realizing the intelligent detection of the end-to-end dropper.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a mixed CNN-LSTM string hanging abnormity detection method based on an attention mechanism. And an attention mechanism is further introduced, important channel signal information influencing the abnormity detection is paid attention to, and finally the abnormity detection of the catenary dropper is realized.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
the method comprises the following steps:
a hybrid CNN-LSTM dropper anomaly detection method based on an attention mechanism, the method comprising:
step 1: acquiring an acceleration signal, and mounting an acceleration sensor on each span of carrier cables and contact lines; acquiring horizontal and vertical acceleration signal data of each acceleration sensor under 4 working conditions that a first dropper is broken, a midspan dropper is broken, the midspan dropper and the first dropper are broken simultaneously, the first dropper is broken and the midspan dropper is not broken under the action of pulsating wind;
step 2: building a deep learning environment, and installing a Keras library based on a deep learning framework TensorFlow on a server;
and step 3: constructing a mixed CNN-LSTM model based on an attention mechanism;
and 4, step 4: training a mixed CNN-LSTM model based on an attention mechanism;
and 5: and acquiring vibration acceleration signal data of the contact network in real time, inputting the trained attention-based hybrid CNN-LSTM dropper abnormity detection model, performing fault diagnosis, and judging whether the dropper is broken.
In the scheme, the step 3: constructing a mixed CNN-LSTM model based on an attention mechanism:
step 3.1: and constructing an attention module of the mixed CNN-LSTM model based on an attention mechanism. Firstly, performing one-dimensional convolution, and extracting a time mode matrix of the variable in the convolution kernel range; performing maximum pooling and global average pooling to obtain difference feedback and average feedback of the vibration acceleration signals of the convolution kernel, and combining the two feature descriptions; finally, the attention weight coefficient of each signal is obtained by fusing the characteristics of the vibration acceleration signals through the full connection layer;
step 3.2: and constructing a 1DCNN module of the mixed CNN-LSTM model based on an attention mechanism. Firstly, setting a convolution kernel and extracting the spatial characteristics of an acceleration signal; a pooling layer is added behind the convolutional layer, and the parameters with training are reduced while the characteristics irrelevant to the position are obtained;
step 3.3: and constructing an LSTM module of the mixed CNN-LSTM model based on an attention mechanism. Setting a neural network structure in an LSTM module as an LSTM layer, and learning the time characteristics of the acceleration signal;
step 3.4: the attention module, the 1DCNN model and the LSTM module are connected through a Sequential module built in a keras, an output layer is added, and the connected model is a mixed CNN-LSTM model based on the attention system.
In the above scheme, the step 3.1 is specifically as follows:
(1) CNN detection time patterns
Applying a CNN filter to acceleration signal data S, S ∈ RH×WWith H10 and W200, using C convolution kernels Ci∈R1×WPerforming convolution operation on S, performing weighted summation on a convolution kernel and a convolved part, then moving downwards with step length of 1, and performing convolution operation on all signals in sequence, wherein a calculation formula is as follows:
convolution operation generates time pattern matrix Fc∈RH×CWherein
Representing the convolution value of the ith row vector and the jth filter;
(2) attention weighting
For time pattern matrix FCMaximum pooling (Maxpooling) and global average pooling (averageposing) were performed to obtain differential feedback of 10 acceleration signals per convolution kernel
And average feedback
Wherein
Will be provided with
And
merging, fusing features through a full connection layer, and finally obtaining different acceleration signals through a sigmoid activation functionAttention weight coefficient of number, its calculation process is as follows:
in the scheme, the step 4: training a mixed CNN-LSTM model based on an attention mechanism:
step 4.1: dividing a data set;
step 4.2: carrying out hyper-parameter tuning by using a Bayesian optimization method;
and 4.3, inputting a vibration acceleration signal data set, training a neural network, and forming a mixed CNN-LSTM dropper abnormity detection model based on an attention mechanism.
Compared with the prior art, the invention has the following beneficial effects:
the method combines the respective characteristics of the convolutional neural network and the LSTM neural network, extracts the spatial characteristics of a multi-channel signal through one-dimensional convolution, and extracts the time dependence characteristic of the signal through the LSTM.
Drawings
FIG. 1 is an overall block diagram of an implementation of the present invention;
FIG. 2 is a schematic diagram of an installation position of an acceleration sensor of a contact network;
FIG. 3 is a block diagram of an attention module;
FIG. 4 is a flow chart of a hybrid CNN-LSTM model training based on an attention mechanism.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
as shown in fig. 1, a hybrid CNN-LSTM dropper breakage detection method based on an attention mechanism specifically includes the following steps:
step 1: acquiring an acceleration signal;
as shown in fig. 2, 5 sensors are mounted on each jumper contact line and catenary:
1# acceleration sensor: 1m inside the supporting point 1 on the carrier cable;
2# acceleration sensor: the inner side of the supporting point 2 on the carrier cable is 1 m;
3# acceleration sensor: the mid-span position of the carrier cable;
4# acceleration sensor: the contact line is arranged at the middle 5 th hanging string;
5# acceleration sensor: the contact line is across the 9 th dropper.
Acquiring horizontal and vertical acceleration signal data of 5 acceleration sensors under 4 working conditions that a first dropper is broken, a midspan dropper is broken, the midspan dropper and the first dropper are broken simultaneously, the first dropper is broken and the midspan dropper is not broken under the action of pulsating wind, and transmitting the acceleration signal data acquired by the sensors to a server end through an NB-IoT network.
Step 2: and (3) building a deep learning environment, and installing a Keras library based on a deep learning framework TensorFlow on a server.
And step 3: constructing a mixed CNN-LSTM model based on an attention mechanism, which comprises the following specific steps:
step 3.1: and constructing an attention module of the mixed CNN-LSTM model based on an attention mechanism. As shown in fig. 3, first, a one-dimensional convolution is performed to extract a time pattern matrix of the variable within the convolution kernel; performing maximum pooling and global average pooling to obtain difference feedback and average feedback of the vibration acceleration signals of the convolution kernel, and combining the two feature descriptions; finally, the attention weight coefficient of each signal is obtained by fusing the characteristics of the vibration acceleration signals through the full connection layer;
the step 3.1 is specifically as follows:
(1) CNN detection time patterns
Applying CNN filters to acceleration signalsData S for input acceleration signal data S ∈ R200×10Turning S' epsilon R in time dimension and characteristic dimension10×200Using C convolution kernels Ci∈R1×200Performing one-dimensional convolution, performing weighted summation on a convolution kernel and a convolved part, then moving downwards by taking the step size as 1, and sequentially performing convolution operation on all signals, wherein the calculation formula is as follows:
convolution operation generates time pattern matrix FC∈R10×CWherein
Representing the convolution value of the ith row vector and the jth filter;
(2) attention weighting
For time pattern matrix FCMaximum pooling (Maxpooling) and global average pooling (averageposing) were performed to obtain differential feedback of 10 acceleration signals per convolution kernel
And average feedback
Wherein
In order to comprehensively extract the characteristics of the signals, the method comprises
And
merging, fusing features through a full connection layer, and finally passing a sigmoidActivating a function to obtain attention weight coefficients corresponding to 10 acceleration signals, wherein the calculation process is as follows:
step 3.2: and constructing a 1DCNN module of the mixed CNN-LSTM model based on an attention mechanism. Firstly, setting a convolution kernel and extracting the spatial characteristics of an acceleration signal; a pooling layer is added behind the convolutional layer, and the parameters with training are reduced while the characteristics irrelevant to the position are obtained;
step 3.3: and constructing an LSTM module of the mixed CNN-LSTM model based on an attention mechanism. Setting a neural network structure in an LSTM module as an LSTM layer, and learning the time characteristics of the acceleration signal;
step 3.4: the attention module, the 1DCNN model and the LSTM module are connected through a Sequential module built in a keras, an output layer is added, and the connected model is a mixed CNN-LSTM model based on the attention system.
And 4, step 4: training a mixed CNN-LSTM model based on an attention mechanism, wherein the model training process is shown in FIG. 4 and comprises the following specific steps:
step 4.1: the data set is partitioned. The sample capacity is 800, and the training set, the verification set and the test set are divided into 512 samples, 128 samples and 160 samples respectively;
step 4.2: and (4) carrying out hyper-parameter tuning by using a Bayesian optimization method. Determining the number of convolution layer units, the size of convolution kernels, the number of LSTM layer units, an activation function and the like by a Bayesian optimization method;
the Bayesian optimization algorithm is a hyper-parametric optimization algorithm. The basic framework is as shown in algorithm 1, and firstly, a probabilistic proxy model f is initialized, which is a fitting to a complex objective function, and common models are: gaussian process, random forest, TPE, etc.; initializing historical observation set D0The observation set consists of a plurality of pairs of data (x, y), wherein x represents a group of hyper-parameters, and y represents target values of the group of hyper-parameters; secondly, searching the next group of hyper-parameter evaluation points x by maximizing the acquisition function alphan+1(ii) a Then the objective function value y of the group of hyper-parameters is calculatedn+1(ii) a Finally will { xn+1,yn+1}Added to the historical observation set DnAnd updating the probability agent model f, and continuously iterating until reaching a defined threshold value.
And 4.3, inputting a vibration acceleration signal data set, training a neural network, and forming a mixed CNN-LSTM dropper abnormity detection model based on an attention mechanism.
And 5: acquiring vibration acceleration signal data of a contact network in real time, inputting a trained attention-based hybrid CNN-LSTM dropper abnormity detection model, performing fault diagnosis, and judging whether a dropper is broken, wherein the method comprises the following specific steps:
step 5.1: acquiring vertical and horizontal acceleration signal data of a catenary and a sensor on a contact line;
step 5.2: and inputting the acquired 10 vibration acceleration signal data into a trained attention-based mixed CNN-LSTM model, and judging whether the hanger is broken or not.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A mixed CNN-LSTM dropper anomaly detection method based on an attention mechanism is characterized in that,
the method comprises the following steps:
step 1: acquiring an acceleration signal, and mounting an acceleration sensor on each span of carrier cables and contact lines; acquiring horizontal and vertical acceleration signal data of each acceleration sensor under 4 working conditions that a first dropper is broken, a midspan dropper is broken, the midspan dropper and the first dropper are broken simultaneously, the first dropper is broken and the midspan dropper is not broken under the action of pulsating wind;
step 2: building a deep learning environment, and installing a Keras library based on a deep learning framework TensorFlow on a server;
and step 3: constructing a mixed CNN-LSTM model based on an attention mechanism;
and 4, step 4: training a mixed CNN-LSTM model based on an attention mechanism;
and 5: and acquiring vibration acceleration signal data of the contact network in real time, inputting the trained attention-based hybrid CNN-LSTM dropper abnormity detection model, performing fault diagnosis, and judging whether the dropper is broken.
2. The method of claim 1 for detecting abnormality of mixed CNN-LSTM dropper based on attention mechanism,
and step 3: constructing a mixed CNN-LSTM model based on an attention mechanism:
step 3.1: constructing an attention module of a mixed CNN-LSTM model based on an attention mechanism;
step 3.2: constructing a 1DCNN module of a mixed CNN-LSTM model based on an attention mechanism;
step 3.3: constructing an LSTM module of a mixed CNN-LSTM model based on an attention mechanism;
step 3.4: the attention module, the 1DCNN model and the LSTM module are connected through a Sequential module built in a keras, an output layer is added, and the connected model is a mixed CNN-LSTM model based on the attention system.
3. The method for detecting abnormality of mixed CNN-LSTM dropper based on attention mechanism as claimed in claim 2, wherein the step 3.1 is as follows:
(1) CNN detection time patterns
Applying a CNN filter to acceleration signal data S, S ∈ RH×WWith H10 and W200, using C convolution kernels Ci∈R1×WPerforming convolution operation on S to convolve the convolution kernel with the SThe weighted summation is carried out on the parts, then the weighted summation is moved downwards by the step size of 1, and the convolution operation is carried out on all the signals in turn, and the calculation formula is as follows:
convolution operation generates time pattern matrix Fc∈RH×CWherein
Representing the convolution value of the ith row vector and the jth filter;
(2) attention weighting
For time pattern matrix FCMaximum pooling (Maxpooling) and global average pooling (averageposing) were performed to obtain differential feedback of 10 acceleration signals per convolution kernel
And average feedback
Wherein
Will be provided with
And
merging, fusing features through a full connection layer, and finally obtaining attention weight coefficients of different acceleration signals through a sigmoid activation function, wherein the calculation process is as follows:
4. the method of claim 3 for detecting abnormality of mixed CNN-LSTM dropper based on attention mechanism,
and 4, step 4: training a mixed CNN-LSTM model based on an attention mechanism:
step 4.1: dividing a data set;
step 4.2: carrying out hyper-parameter tuning by using a Bayesian optimization method;
and 4.3, inputting a vibration acceleration signal data set, training a neural network, and forming a mixed CNN-LSTM dropper abnormity detection model based on an attention mechanism.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111469577.3A CN114298087B (en) | 2021-12-03 | 2021-12-03 | Mixed CNN-LSTM dropper anomaly detection method based on attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111469577.3A CN114298087B (en) | 2021-12-03 | 2021-12-03 | Mixed CNN-LSTM dropper anomaly detection method based on attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114298087A true CN114298087A (en) | 2022-04-08 |
| CN114298087B CN114298087B (en) | 2024-07-12 |
Family
ID=80965703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111469577.3A Active CN114298087B (en) | 2021-12-03 | 2021-12-03 | Mixed CNN-LSTM dropper anomaly detection method based on attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114298087B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672343A (en) * | 2019-09-29 | 2020-01-10 | 电子科技大学 | Rotary machine fault diagnosis method based on multi-attention convolutional neural network |
| CN111721324A (en) * | 2020-05-15 | 2020-09-29 | 中铁第一勘察设计院集团有限公司 | Contact net dropper breakage detection method based on acceleration signals |
| CN112200032A (en) * | 2020-09-28 | 2021-01-08 | 辽宁石油化工大学 | Attention mechanism-based high-voltage circuit breaker mechanical state online monitoring method |
| US20210201010A1 (en) * | 2019-12-31 | 2021-07-01 | Wuhan University | Pedestrian re-identification method based on spatio-temporal joint model of residual attention mechanism and device thereof |
| CN113288162A (en) * | 2021-06-03 | 2021-08-24 | 北京航空航天大学 | Short-term electrocardiosignal atrial fibrillation automatic detection system based on self-adaptive attention mechanism |
-
2021
- 2021-12-03 CN CN202111469577.3A patent/CN114298087B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672343A (en) * | 2019-09-29 | 2020-01-10 | 电子科技大学 | Rotary machine fault diagnosis method based on multi-attention convolutional neural network |
| US20210201010A1 (en) * | 2019-12-31 | 2021-07-01 | Wuhan University | Pedestrian re-identification method based on spatio-temporal joint model of residual attention mechanism and device thereof |
| CN111721324A (en) * | 2020-05-15 | 2020-09-29 | 中铁第一勘察设计院集团有限公司 | Contact net dropper breakage detection method based on acceleration signals |
| CN112200032A (en) * | 2020-09-28 | 2021-01-08 | 辽宁石油化工大学 | Attention mechanism-based high-voltage circuit breaker mechanical state online monitoring method |
| CN113288162A (en) * | 2021-06-03 | 2021-08-24 | 北京航空航天大学 | Short-term electrocardiosignal atrial fibrillation automatic detection system based on self-adaptive attention mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114298087B (en) | 2024-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106779151B (en) | A kind of line of high-speed railway wind speed multi-point multi-layer coupling prediction method | |
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| CN111898634B (en) | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption | |
| CN106372731B (en) | A kind of high wind line of high-speed railway wind speed spatial network structure forecast method | |
| CN110161343B (en) | Non-invasive real-time dynamic monitoring method for external powered device of intelligent train | |
| CN111709448A (en) | Mechanical fault diagnosis method based on migration relation network | |
| CN105203645B (en) | A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion | |
| WO2021013190A1 (en) | Meteorological parameter-based high-speed train positioning method and system in navigation blind zone | |
| CN110985310B (en) | Wind driven generator blade fault monitoring method and device based on acoustic sensor array | |
| CN112200237A (en) | Abnormal diagnosis method for time sequence monitoring data of structural health monitoring system | |
| CN113610188B (en) | Bow net contact force non-section abnormality identification method and device | |
| CN111310686A (en) | Steel wire rope broken wire damage identification method of IPSO-BP neural network | |
| CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
| CN112763215B (en) | Multi-working-condition online fault diagnosis method based on modular federal deep learning | |
| CN116052435B (en) | Urban road congestion influence range definition and road influence calculation method | |
| CN116032003A (en) | Power grid inspection method and system based on power private network data processing | |
| CN106384507A (en) | Travel time real-time estimation method based on sparse detector | |
| CN111461183A (en) | Flat wheel fault detection method for optimizing BP neural network by improving artificial fish swarm algorithm | |
| CN114048880B (en) | Particle swarm optimization-based opening valve fault diagnosis method | |
| CN113807404B (en) | Intelligent monitoring and emergency system of intelligent traffic network for multi-source information fusion | |
| CN115371724A (en) | Method and device for evaluating track state | |
| CN113593605B (en) | Industrial audio fault monitoring system and method based on deep neural network | |
| CN114298087B (en) | Mixed CNN-LSTM dropper anomaly detection method based on attention mechanism | |
| CN117312962A (en) | Long-service elevator guide rail fault diagnosis method based on transfer learning and data driving | |
| CN117454100A (en) | Missing data reconstruction method for engineering structure |
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 |