CN117011690A - Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium - Google Patents
Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium Download PDFInfo
- Publication number
- CN117011690A CN117011690A CN202311281760.XA CN202311281760A CN117011690A CN 117011690 A CN117011690 A CN 117011690A CN 202311281760 A CN202311281760 A CN 202311281760A CN 117011690 A CN117011690 A CN 117011690A
- Authority
- CN
- China
- Prior art keywords
- hidden danger
- features
- data
- submarine cable
- feature
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 239000011159 matrix material Substances 0.000 claims abstract description 112
- 230000000007 visual effect Effects 0.000 claims abstract description 61
- 238000000605 extraction Methods 0.000 claims abstract description 53
- 238000002156 mixing Methods 0.000 claims abstract description 47
- 230000009466 transformation Effects 0.000 claims abstract description 25
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 18
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000013507 mapping Methods 0.000 claims abstract description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 238000004140 cleaning Methods 0.000 claims description 7
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000000844 transformation Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 9
- 230000008569 process Effects 0.000 abstract description 6
- 238000013135 deep learning Methods 0.000 abstract description 4
- 230000007547 defect Effects 0.000 abstract description 3
- 238000003058 natural language processing Methods 0.000 abstract description 3
- 239000000203 mixture Substances 0.000 description 10
- 238000012423 maintenance Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/05—Underwater scenes
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Abstract
The application discloses a submarine cable hidden danger identification method, a submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and a submarine cable hidden danger identification medium, wherein the submarine cable hidden danger identification method comprises the following steps: the method comprises the steps of preprocessing the acquired visual data, corpus data and sensor data, adopting a corresponding feature extraction network to perform feature extraction to generate visual features, corpus features and sensor features, mapping the features to a unified feature space through a feedforward neural network, constructing an equal-length matrix, performing self-attention operation based on the equal-length matrix, inputting the equal-length matrix into a first multi-layer perceptron to perform feature transformation, determining space-time mixing features, inputting the space-time mixing features into a cascaded second multi-layer perceptron and softmax layer to perform identification classification, and outputting hidden danger detection results of submarine cables. The deep learning technology and the natural language processing technology are adopted in the whole submarine cable hidden danger identification process, the defect complementation advantages of visual data, corpus data and sensor data in various mode data are fully combined, the working efficiency of submarine cable hidden danger identification is improved, and meanwhile hidden danger identification accuracy is improved.
Description
Technical Field
The application relates to the technical field of submarine cable detection, in particular to a submarine cable hidden danger identification method.
Background
Submarine cables play a significant role in global communication and data transmission, and due to the special operating environment of submarine cables, maintenance and management of the submarine cables face a plurality of challenges, and timely discovery and treatment of hidden dangers are important to the maintenance and management of the submarine cables.
The traditional submarine cable hidden danger identification method mainly depends on manual inspection, including visual inspection, electrical performance test and the like of submarine cables, is time-consuming, labor-consuming and low in accuracy, and is difficult to meet the increasing submarine cable maintenance requirements. In order to solve the problems, people begin to explore and use an image recognition technology to detect hidden danger of the submarine cable, but the hidden danger detection of the submarine cable is often only dependent on single visual information of the submarine cable, and the accuracy rate of hidden danger recognition of the submarine cable still needs to be further improved.
Disclosure of Invention
The application provides a submarine cable hidden danger identification method, device, equipment and medium, which solve the technical problem that the hidden danger identification accuracy of submarine cables is low because the hidden danger detection of submarine cables in the prior art is often only dependent on single visual information.
The application provides a submarine cable hidden danger identification method, which comprises the following steps:
when a plurality of groups of data to be identified of the submarine cable are obtained, carrying out data preprocessing on all the data to be identified, and respectively carrying out feature extraction by adopting a preset feature extraction network to generate visual features, corpus features and sensor features;
mapping the visual features, the corpus features and the sensor features to a unified feature space by adopting a preset feedforward neural network respectively, and combining the output multiple groups of equilong features to construct an equilong matrix;
after the self-attention operation is executed based on the equal-length matrix, inputting a preset first multi-layer perceptron to perform feature transformation, and determining space-time mixing features;
and inputting the space-time mixing characteristics into a cascaded preset second multi-layer perceptron and a softmax layer to carry out identification classification, and outputting hidden danger detection results of the submarine cable.
Optionally, when multiple sets of data to be identified of the submarine cable are obtained, performing data preprocessing on all the data to be identified, and performing feature extraction by using a preset feature extraction network respectively, so as to generate visual features, corpus features and sensor features, including:
when a plurality of groups of data to be identified of the submarine cable are obtained, carrying out data cleaning, denoising and normalization on all the data to be identified, and outputting target visual data, target corpus data and target sensor data; the data to be identified comprises visual data, corpus data and sensor data;
performing feature extraction on target visual data by adopting a preset convolutional neural network to generate visual features;
inputting the target corpus data into a preset first time sequence network, inputting the target sensor data into a preset second time sequence network for feature extraction, and generating corpus features and sensor features.
Optionally, after the performing the self-attention operation based on the equal-length matrix, inputting a preset first multi-layer perceptron to perform feature transformation, and determining the space-time mixing feature, including:
performing multiple linear transformations based on the equal-length matrix to determine a query matrix, a key matrix and a value matrix;
multiplying the query matrix and the transposed matrix of the key matrix to construct an attention matrix;
after softmax operation is carried out on the attention moment array, multiplication operation is carried out on the attention moment array and the value matrix, so that a target value matrix is generated;
and after the target value matrix and the equal-length matrix are spliced line by line, inputting the target value matrix and the equal-length matrix into a preset first multi-layer perceptron to perform feature transformation, and determining space-time mixing features.
Optionally, the method further comprises:
acquiring a historical space-time mixed characteristic sequence and a historical hidden danger characteristic sequence library which are associated with the submarine cable at a prediction moment;
inputting the historical space-time mixing characteristic sequence into a preset third time sequence network to perform characteristic prediction, and outputting the predicted space-time mixing characteristic at the predicted moment;
the historical space-time mixed characteristic sequence and the predicted space-time mixed characteristic are connected in series to form a one-dimensional characteristic;
determining a condition code based on the one-dimensional feature and the historical hidden danger feature sequence library;
and carrying out identification classification by adopting the prediction space-time mixing characteristics and the condition codes through a cascade preset third multi-layer perceptron and a softmax layer, and outputting hidden danger prediction results.
Optionally, the step of determining a condition code based on the one-dimensional feature and the historical potential hazard feature sequence library includes:
extracting a plurality of history hidden trouble feature sequences from the history hidden trouble feature sequence library;
respectively carrying out linear transformation on the one-dimensional features and each history hidden trouble feature sequence to construct a prediction query matrix and a plurality of prediction key matrixes;
performing inner product operation on the prediction query matrix and each prediction key matrix respectively, and outputting a plurality of confidence coefficients;
and arranging all the confidence degrees from high to low, sequentially selecting a plurality of historical hidden danger feature sequences corresponding to the confidence degrees with a preset number of thresholds for series connection, and inputting a preset full-connection layer for channel dimension reduction to generate a condition code.
Optionally, the third multi-layer perceptron comprises a first full-connection layer module and a second full-connection layer module; the step of identifying and classifying by adopting the prediction space-time mixing characteristics and the condition codes through a cascade preset third multi-layer perceptron and a softmax layer and outputting hidden danger prediction results comprises the following steps:
carrying out feature extraction on the predicted space-time mixing features by adopting a first full-connection layer module, and outputting first features;
after the first features and the condition codes are connected in series, inputting the first features and the condition codes into a second full-connection layer module for feature extraction to generate second features;
and performing feature mapping based on the second features through the softmax layer, and outputting hidden danger prediction results.
Optionally, the method further comprises:
acquiring a historical space-time mixed feature library of a plurality of submarine cables;
extracting a plurality of historical space-time mixed features in a hidden danger critical period from each historical space-time mixed feature library to respectively form a historical hidden danger feature sequence;
and forming a history hidden danger feature sequence library by adopting all the history hidden danger feature sequences.
The application provides a submarine cable hidden danger identification device, which comprises:
the device comprises a feature extraction module, a data processing module and a data processing module, wherein the feature extraction module is used for carrying out data preprocessing on all the data to be identified and respectively adopting a preset feature extraction network to carry out feature extraction when a plurality of groups of data to be identified of the submarine cable are acquired, so as to generate visual features, corpus features and sensor features;
the equilong matrix construction module is used for mapping the visual features, the corpus features and the sensor features to a unified feature space by adopting a preset feedforward neural network respectively, and combining the output multiple groups of equilong features to construct an equilong matrix;
the space-time mixing characteristic module is used for inputting a preset first multi-layer perceptron to perform characteristic transformation after the self-attention operation is performed on the basis of the equal-length matrix, and determining space-time mixing characteristics;
and the hidden danger detection module is used for inputting the space-time mixing characteristics into a cascaded preset second multi-layer perceptron and a softmax layer to carry out identification and classification and outputting hidden danger detection results of the submarine cable.
An electronic device according to a third aspect of the present application includes a memory and a processor, where the memory stores a computer program, where the computer program, when executed by the processor, causes the processor to execute the steps of the submarine cable hidden danger identification method according to any one of the first aspect of the present application.
A fourth aspect of the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program when executed implements the submarine cable hidden danger identification method according to any one of the first aspects of the present application.
From the above technical scheme, the application has the following advantages:
according to the application, the acquired visual data, corpus data and sensor data are subjected to data preprocessing, the corresponding feature extraction network is adopted to perform feature extraction to generate visual features, corpus features and sensor features, the features are mapped to a unified feature space through the feedforward neural network, an isometric matrix is constructed, self-attention operation is performed based on the isometric matrix, a preset first multi-layer perceptron is input to perform feature transformation, space-time mixing features are determined, the space-time mixing features are input to a cascaded preset second multi-layer perceptron and softmax layer to perform identification classification, and hidden danger detection results of submarine cables are output. The deep learning technology and the natural language processing technology are adopted in the whole submarine cable hidden danger identification process, the defect complementation advantages of visual data, corpus data and sensor data in various mode data are fully combined, the working efficiency of submarine cable hidden danger identification is improved, and meanwhile hidden danger identification accuracy is improved.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the application, and that other drawings can be obtained from these drawings without inventive faculty for a person skilled in the art.
FIG. 1 is a flow chart of steps of a submarine cable hidden danger identification method provided by an embodiment of the application;
fig. 2 is a schematic diagram of a framework of a submarine cable hidden danger identification method according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a feature processing flow of a third sensor according to an embodiment of the present application;
fig. 4 is a block diagram of a submarine cable hidden danger identification device according to an embodiment of the present application.
Detailed Description
The embodiment of the application provides a submarine cable hidden danger identification method, device, equipment and medium, which are used for solving the technical problem that the hidden danger identification accuracy of submarine cables is low because the hidden danger detection of submarine cables in the prior art is often only dependent on single visual information.
In order to make the objects, features and advantages of the present application more comprehensible, the technical solutions in the embodiments of the present application are described in detail below with reference to the accompanying drawings, and it is apparent that the embodiments described below are only some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Referring to fig. 1, fig. 1 is a flowchart illustrating steps of a submarine cable hidden danger identification method according to an embodiment of the present application.
The application provides a submarine cable hidden danger identification method, which comprises the following steps:
and 101, when a plurality of groups of data to be identified of the submarine cable are obtained, carrying out data preprocessing on all the data to be identified, and respectively carrying out feature extraction by adopting a preset feature extraction network to generate visual features, corpus features and sensor features.
Optionally, step 101 comprises the sub-steps of:
when multiple groups of data to be identified of the submarine cable are obtained, carrying out data cleaning, denoising and normalization on all the data to be identified, and outputting target visual data, target corpus data and target sensor data; the data to be identified comprises visual data, corpus data and sensor data;
performing feature extraction on target visual data by adopting a preset convolutional neural network to generate visual features;
inputting the target corpus data into a preset first time sequence network, inputting the target sensor data into a preset second time sequence network for feature extraction, and generating corpus features and sensor features.
The data to be identified refers to data directly collected when hidden danger identification is carried out on the submarine cable, and the data comprise visual data, corpus data and sensor data. Visual data can be acquired by a submarine cable through a submarine camera or an optical fiber sensor, and the visual data comprises information such as the surface state of the submarine cable and the surrounding environment. The corpus data comprises information such as maintenance reports, detection records and the like of the submarine cable, wherein the maintenance reports comprise text descriptions such as maintenance history, hidden danger types, discovery and processing conditions and the like of the submarine cable. Sensor data including temperature, pressure, vibration, etc. information of the sea cable.
The data preprocessing comprises preprocessing such as data cleaning, denoising, normalization and the like.
Data cleaning refers to removing dirty data with poor quality caused by equipment failure or other reasons, such as partial lines or pixels in visual data are missing, the exposure is too low or too high, and the underwater environment is too turbid without effective information; data such as temperature, pressure, vibration and the like have interference around or data of a sensor fault period; sentences which do not accord with the specification or relate to the task in the corpus data.
Denoising refers to reducing the noise level of data mainly by using technologies such as filtering and the like, for example, reducing the noise of visual data by using methods such as bilateral filtering and the like, reducing the noise of sensor data by using bandpass/lowpass filtering or time-frequency decomposition reconstruction technology, and screening out some nonsensical characters in the language data.
Normalization, for example, unifying the resolution, size and length of the visual data through resampling, resize and other operations, scaling and translating the sensor data value range distribution of the same type to convert to the same distribution, and converting the language data into word embedding by using a word2vec model and other models or into higher-order language features by using a Bert, clip (encoder therein) and other language models.
The feature extraction network comprises a convolutional neural network and a time sequence network.
Target visual data, target corpus data and target sensor data refer to visual data, corpus data and sensor data obtained by preprocessing collected visual data, corpus data and sensor data.
Visual characteristics, corpus characteristics and sensor characteristics refer to characteristics which are output after the characteristics extraction network respectively performs characteristic extraction on visual data, corpus data and sensor data after data preprocessing.
In the embodiment of the application, when the visual data, the corpus data and the sensor data of the submarine cable to be subjected to hidden danger identification are obtained, corresponding preprocessing work such as cleaning, denoising, normalizing and the like is carried out on the visual data, the corpus data and the sensor data to obtain corresponding target visual data, target corpus data and target sensor data, a preset Convolutional Neural Network (CNN) is adopted for image feature extraction on the target visual data, visual features are output, and the target corpus data and the target sensor data are respectively input into a preset first time sequence network and a preset second time sequence network for semantic feature extraction, so that corpus features and sensor features are output.
Preferably, the feature extraction may also be performed on the target visual data using a residual network (ResNet) or a dense connectivity network (DenseNet).
And 102, mapping the visual features, the corpus features and the sensor features to a unified feature space by adopting a preset feedforward neural network respectively, and combining the output multiple groups of equilong features to construct an equilong matrix.
In embodiments of the application, visual features are presentedCorpus characteristics->And sensor feature->Respectively inputting three preset feedforward neural networks to perform characteristic processing, and outputting three groups of equal-length characteristics with equal length in the same characteristic spaceCombining three groups of equilong features to construct an equilong matrix>Let the number of equal length characteristic channels be C +.>Is a 3×c matrix.
And 103, after the self-attention operation is executed based on the equal-length matrix, inputting a preset first multi-layer perceptron to perform feature transformation, and determining the space-time mixing features.
Step 103 comprises the sub-steps of:
performing multiple linear transformations based on the equal-length matrix to determine a query matrix, a key matrix and a value matrix;
multiplying the query matrix by the transposed matrix of the key matrix to construct an attention matrix;
after softmax operation is carried out on the attention moment array, multiplication operation is carried out on the attention moment array and the value matrix, so that a target value matrix is generated;
and after the target value matrix and the equal length matrix are spliced line by line, inputting the target value matrix and the equal length matrix into a preset first multi-layer perceptron to perform characteristic transformation, and determining the space-time mixing characteristics.
The space-time hybrid features refer to features generated by fusion after feature extraction and self-attention operation of multiple groups of data to be identified of the submarine cable.
In the embodiment of the application, the equal-length matrix is respectively matched with different weight matrices、/>And->Multiplying by multiple linear transforms to determine the query matrix +.>Key matrix->Sum matrix->I.e. +.>、And->Wherein->Representing matrix multiplication +.>、/>And->Is C x k, k= {64,128,.}, or other empirical length, query matrix +.>Key matrixKSum matrixVIs a 3 xk matrix. Employing a query matrix->And key matrix->Multiplication is performed on the transposed matrix of (2) to construct an attention matrix +.>I.e. +.>. After performing a softmax operation on each row in the attention matrix, the sum value matrix +.>Multiplication is performed to generate a target value matrix +.>I.e.. Matrix the target value->Expansion and equal length matrix->Is of equal lengthAnd performing row-by-row splicing to obtain the characteristic with the length of 3× (k+C), and inputting the characteristic into a preset first multi-layer perceptron to perform characteristic transformation to generate a space-time mixing characteristic.
Preferably, the spatial-temporal hybrid features are generated by fusion based on visual data, corpus data and sensor data, and a multi-Modal Deep Learning Fusion Network (MDLFN) or a cross-modal information bottleneck (CMB) method can be adopted.
And 104, inputting the space-time mixing characteristics into a cascaded preset second multi-layer perceptron and a softmax layer for identification and classification, and outputting hidden danger detection results of the submarine cable.
The hidden danger detection result refers to the hidden danger of a submarine cable with a fault and the hidden danger of a submarine cable without a fault.
In the embodiment of the application, the current space-time hybrid characteristics of the submarine cable are input into the cascaded preset second multi-layer perceptron and softmax layer for identification and classification, so that the hidden danger detection result of the submarine cable can be obtained.
Preferably, the method further comprises:
s1, acquiring a historical space-time mixed characteristic sequence and a historical hidden danger characteristic sequence library which are associated with a submarine cable at a prediction moment;
s2, inputting the historical space-time mixed characteristic sequence into a preset third time sequence network to conduct characteristic prediction, and outputting predicted space-time mixed characteristics at a prediction moment;
s3, connecting the historical space-time mixed characteristic sequence with the predicted space-time mixed characteristic in series to form a one-dimensional characteristic;
s4, determining a condition code based on the one-dimensional characteristics and the historical hidden danger characteristic sequence library;
s5, identifying and classifying by adopting prediction space-time mixing characteristics and condition codes through a third cascade preset multi-layer perceptron and a softmax layer, and outputting hidden danger prediction results.
Optionally, the method further comprises:
acquiring a historical space-time mixed feature library of a plurality of submarine cables;
extracting a plurality of historical space-time mixed features in a hidden danger critical period from each historical space-time mixed feature library to respectively form a historical hidden danger feature sequence;
and forming a history hidden danger feature sequence library by adopting all the history hidden danger feature sequences.
Optionally, step S4 includes:
extracting a plurality of history hidden trouble feature sequences from a history hidden trouble feature sequence library;
respectively carrying out linear transformation on the one-dimensional characteristics and each history hidden trouble characteristic sequence to construct a prediction query matrix and a plurality of prediction key matrixes;
performing inner product operation on the prediction query matrix and each prediction key matrix respectively, and outputting a plurality of confidence coefficients;
and arranging all the confidence degrees from high to low, sequentially selecting a plurality of historical hidden danger feature sequences corresponding to the confidence degrees with a preset number of thresholds for series connection, and inputting a preset full-connection layer for channel dimension reduction to generate a condition code.
Optionally, the third multi-layer perceptron comprises a first fully-connected layer module and a second fully-connected layer module; the step S5 comprises the following steps:
carrying out feature extraction on the predicted space-time mixing features by adopting a first full-connection layer module, and outputting first features;
after the first feature and the condition code are connected in series, inputting the first feature and the condition code into a second full-connection layer module for feature extraction to generate a second feature;
and performing feature mapping based on the second features through the softmax layer, and outputting hidden danger prediction results.
The prediction time refers to the time at which the spatio-temporal hybrid feature prediction is performed.
The prediction of the spatio-temporal mixture features refers to the spatio-temporal mixture features predicted using a third timing network.
Historical spatio-temporal mixture feature sequences refer to sequences of historical spatio-temporal mixture feature compositions used to make spatio-temporal mixture feature predictions. A historical spatiotemporal mixture feature library refers to a database that includes all of the historical spatiotemporal mixture features of submarine cables. Historical spatiotemporal mixing features refer to spatiotemporal mixing features calculated prior to the predicted time instant.
The history hidden trouble feature sequence library refers to a database containing a plurality of history hidden trouble feature sequences. The historical hidden trouble feature sequence refers to a sequence formed by mapping historical space-time mixing features of hidden trouble of the submarine cable. The hidden danger critical time period refers to a time period when critical time of fault hidden danger of all submarine cables in existing real data is within a certain preset step range, wherein the critical time is when the submarine cable state is changed from normal to the fault hidden danger near a certain time.
Condition encoding refers to a condition constraint on the third multi-layer perceptron. As the faults occur with obvious historical repeatability in reality, in order to explicitly express the repeatability, the historical data is fully utilized, the network early warning performance is improved, and the condition constraint can be carried out on the third multi-layer perceptron through the condition coding.
The hidden danger prediction result refers to a result of predicting that a submarine cable has a hidden trouble and a submarine cable does not have a hidden trouble at the prediction moment.
In the embodiment of the application, the hidden danger state of the submarine cable can be predicted and early-warned besides the identification and judgment of the hidden danger state of the submarine cable at the current moment.
The method comprises the steps of obtaining a plurality of historical space-time mixed feature libraries of the submarine cables, respectively extracting a plurality of historical space-time mixed feature composition sequences of the submarine cables in any hidden danger critical period from each historical space-time mixed feature library, obtaining a plurality of historical hidden danger feature sequences and storing the historical hidden danger feature sequences as a historical hidden danger feature sequence library.
In the prediction process, a historical space-time mixed characteristic sequence related to the submarine cable at the prediction time is acquired and input into a preset third time sequence network to conduct characteristic prediction, the prediction space-time mixed characteristic at the prediction time is output, and the historical space-time mixed characteristic sequence and the prediction space-time mixed characteristic are connected in series to construct one-dimensional characteristics. For example, record the current time astWill betMultiple historical spatiotemporal mixing features at or before a time spanThe rolling output prediction time is regarded as the third time sequence network of the sequence inputt+1、t +2, when the prediction time ist+2, correlating the historical space-time mixed characteristic sequencesWith the predicted time oftPredicted spatiotemporal mixture characteristics at +2 +.>Serial combination of one-dimensional characteristics->。
Extracting a plurality of history hidden trouble feature sequences from a history hidden trouble feature sequence library, and respectively carrying out prediction weight matrix on one-dimensional features and each extracted history hidden trouble feature sequence、/>Multiplying to make linear transformation to construct predictive query matrix +.>And a plurality of predictive key matrices->I.e. +.>、/>Wherein->Indicate->A history feature sequence->Indicate->And predicting key matrix of the history hidden danger characteristic sequence. Performing inner product operation on the prediction query matrix and each prediction key matrix respectively to correspondingly generate a plurality of confidence degrees, namely +.>WhereinIndicate->Confidence of each historical hidden danger feature sequence. Arranging all confidence degrees from high to low, and sequentially selecting the historical hidden danger feature sequences (namely ++) in the preset quantity threshold N before>And after the serial connection is carried out according to the confidence level from high to low, inputting one or more preset full-connection layers for channel dimension reduction so as to reduce the channel number to a reasonable range, and then generating the condition code of the third multi-layer perceptron. It will be appreciated that this reasonable range is empirical, and 16, 32, …, 1024 may be provided, depending on the actual conditions.
The preset third multi-layer perceptron comprises a first full-connection layer module and a second full-connection layer module, and each full-connection layer module comprises a cascade full-connection layer, an activation layer and a dropout layer. Referring to fig. 2 and 3, when predicting a momenttAnd when +2, inputting the predicted space-time mixed characteristic as a characteristic 1 into a first full-connection layer module for characteristic extraction, outputting the first characteristic as a characteristic 2, inputting the first characteristic and the conditional code into a cascaded second full-connection layer module for characteristic processing after being connected in series, generating a second characteristic, and performing characteristic mapping based on the second characteristic through a softmax layer to generate a hidden danger prediction result. It can be understood that the number of the first full-connection layer modules and the second full-connection layer modules in the third multi-layer perceptron can be multiple, and without losing generality, a certain first full-connection layer module in the third multi-layer perceptron starts, and the subsequent features are connected in series with the condition codes and then input into the next full-connection layer module, such as feature 2 and feature 3 in fig. 3, and at this time, the feature 3 is the second feature after the condition codes are connected in series and then input into the next full-connection layer module for feature processing output.
It is understood that the first, second and third timing networks may be any timing network including a Recurrent Neural Network (RNN), a long-short-term memory network (LSTM), an attention network, and the like.
Optionally, the data preprocessing further includes preprocessing of the annotation. Labeling, namely labeling the moment or scene corresponding to the data as potential sea cable faults or potential sea cable faults, and using the potential sea cable faults or potential sea cable faults for training of the neural network as a true-true (ground-trunk) calculation loss function.
In the embodiment of the application, after the training samples are obtained and labeled, the related neural network can be trained according to the flow shown in fig. 2, and the corresponding loss function values are calculated to perform iterative updating in the direction until the loss function values are converged, so that the neural network in submarine cable hidden identification is determined.
In the embodiment of the application, the acquired visual data, corpus data and sensor data are subjected to data preprocessing, the corresponding feature extraction network is adopted to perform feature extraction to generate visual features, corpus features and sensor features, the features are mapped to a unified feature space through a feedforward neural network, an isometric matrix is constructed, self-attention operation is performed based on the isometric matrix, a preset first multi-layer perceptron is input for feature transformation, space-time mixing features are determined, the space-time mixing features are input into a cascade preset second multi-layer perceptron and softmax layer for identification and classification, and hidden danger detection results of submarine cables are output. The deep learning technology and the natural language processing technology are adopted in the whole submarine cable hidden danger identification process, the defect complementation advantages of visual data, corpus data and sensor data in various mode data are fully combined, the working efficiency of submarine cable hidden danger identification is improved, and meanwhile hidden danger identification accuracy is improved.
Referring to fig. 4, fig. 4 is a block diagram illustrating a submarine cable hidden danger recognition device according to an embodiment of the present application.
A submarine cable hidden danger identification device, comprising:
the feature extraction module 401 is configured to, when multiple sets of data to be identified of the submarine cable are obtained, perform data preprocessing on all the data to be identified, and perform feature extraction by using a preset feature extraction network respectively, so as to generate visual features, corpus features and sensor features;
the equilong matrix construction module 402 is configured to map visual features, corpus features and sensor features to a unified feature space by using a preset feedforward neural network, and combine the output multiple groups of equilong features to construct an equilong matrix;
the space-time mixing feature module 403 is configured to perform a self-attention operation based on the equal-length matrix, and then input a preset first multi-layer perceptron to perform feature transformation, so as to determine a space-time mixing feature;
the hidden danger detection module 404 is configured to input the space-time hybrid characteristic into a cascaded preset second multi-layer perceptron and softmax layer to perform identification and classification, and output a hidden danger detection result of the submarine cable.
Optionally, the feature extraction module 401 is specifically configured to:
when multiple groups of data to be identified of the submarine cable are obtained, carrying out data cleaning, denoising and normalization on all the data to be identified, and outputting target visual data, target corpus data and target sensor data; the data to be identified comprises visual data, corpus data and sensor data;
performing feature extraction on target visual data by adopting a preset convolutional neural network to generate visual features;
inputting the target corpus data into a preset first time sequence network, inputting the target sensor data into a preset second time sequence network for feature extraction, and generating corpus features and sensor features.
Optionally, the feature extraction module 403 is specifically configured to:
performing multiple linear transformations based on the equal-length matrix to determine a query matrix, a key matrix and a value matrix;
multiplying the query matrix by the transposed matrix of the key matrix to construct an attention matrix;
after softmax operation is carried out on the attention moment array, multiplication operation is carried out on the attention moment array and the value matrix, so that a target value matrix is generated;
and after the target value matrix and the equal length matrix are spliced line by line, inputting the target value matrix and the equal length matrix into a preset first multi-layer perceptron to perform characteristic transformation, and determining the space-time mixing characteristics.
Optionally, the hidden danger prediction module further includes:
the characteristic acquisition unit is used for acquiring a historical space-time mixed characteristic sequence and a historical hidden danger characteristic sequence library which are associated with the submarine cable at the prediction moment;
the time-space mixing characteristic prediction unit is used for inputting the historical time-space mixing characteristic sequence into a preset third time sequence network to perform characteristic prediction and outputting the predicted time-space mixing characteristic at the predicted moment;
the one-dimensional feature generation unit is used for connecting the historical space-time mixed feature sequence with the predicted space-time mixed feature in series to form a one-dimensional feature;
the condition code determining unit is used for determining a condition code based on the one-dimensional characteristic and the history hidden danger characteristic sequence library;
and the hidden danger prediction result output unit is used for carrying out identification and classification by adopting prediction space-time mixing characteristics and condition codes through a third cascaded multi-layer perceptron and a softmax layer and outputting hidden danger prediction results.
Optionally, the condition code determining unit is specifically configured to:
extracting a plurality of history hidden trouble feature sequences from a history hidden trouble feature sequence library;
respectively carrying out linear transformation on the one-dimensional characteristics and each history hidden trouble characteristic sequence to construct a prediction query matrix and a plurality of prediction key matrixes;
performing inner product operation on the prediction query matrix and each prediction key matrix respectively, and outputting a plurality of confidence coefficients;
and arranging all the confidence degrees from high to low, sequentially selecting a plurality of historical hidden danger feature sequences corresponding to the confidence degrees with a preset number of thresholds for series connection, and inputting a preset full-connection layer for channel dimension reduction to generate a condition code.
Optionally, the three-layer perceptron comprises a first full-connection layer module and a second full-connection layer module; the hidden danger prediction result output unit is specifically used for:
carrying out feature extraction on the predicted space-time mixing features by adopting a first full-connection layer module, and outputting first features;
after the first feature and the condition code are connected in series, inputting the first feature and the condition code into a second full-connection layer module for feature extraction to generate a second feature;
and performing feature mapping based on the second features through the softmax layer, and outputting hidden danger prediction results.
Optionally, the method further comprises a history hidden trouble feature sequence library construction module for:
acquiring a historical space-time mixed feature library of a plurality of submarine cables;
extracting a plurality of historical space-time mixed features in a hidden danger critical period from each historical space-time mixed feature library to respectively form a historical hidden danger feature sequence;
and forming a history hidden danger feature sequence library by adopting all the history hidden danger feature sequences.
The embodiment of the application also provides electronic equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the submarine cable hidden danger identification method according to any embodiment of the application.
The embodiment of the application also provides a computer readable storage medium, on which a computer program is stored, and when the computer program is executed, the submarine cable hidden danger identification method according to any embodiment of the application is realized.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus and modules described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims (10)
1. The submarine cable hidden danger identification method is characterized by comprising the following steps of:
when a plurality of groups of data to be identified of the submarine cable are obtained, carrying out data preprocessing on all the data to be identified, and respectively carrying out feature extraction by adopting a preset feature extraction network to generate visual features, corpus features and sensor features;
mapping the visual features, the corpus features and the sensor features to a unified feature space by adopting a preset feedforward neural network respectively, and combining the output multiple groups of equilong features to construct an equilong matrix;
after the self-attention operation is executed based on the equal-length matrix, inputting a preset first multi-layer perceptron to perform feature transformation, and determining space-time mixing features;
and inputting the space-time mixing characteristics into a cascaded preset second multi-layer perceptron and a softmax layer to carry out identification classification, and outputting hidden danger detection results of the submarine cable.
2. The submarine cable hidden danger identification method according to claim 1, wherein when a plurality of groups of data to be identified of the submarine cable are acquired, all the data to be identified are subjected to data preprocessing, and feature extraction is performed by using a preset feature extraction network respectively, so that visual features, corpus features and sensor features are generated, and the method comprises the following steps:
when a plurality of groups of data to be identified of the submarine cable are obtained, carrying out data cleaning, denoising and normalization on all the data to be identified, and outputting target visual data, target corpus data and target sensor data; the data to be identified comprises visual data, corpus data and sensor data;
performing feature extraction on target visual data by adopting a preset convolutional neural network to generate visual features;
inputting the target corpus data into a preset first time sequence network, inputting the target sensor data into a preset second time sequence network for feature extraction, and generating corpus features and sensor features.
3. The submarine cable hidden danger identification method according to claim 1, wherein the step of determining the space-time hybrid feature by inputting a preset first multi-layer perceptron to perform feature transformation after performing the self-attention operation based on the equal-length matrix comprises the steps of:
performing multiple linear transformations based on the equal-length matrix to determine a query matrix, a key matrix and a value matrix;
multiplying the query matrix and the transposed matrix of the key matrix to construct an attention matrix;
after softmax operation is carried out on the attention moment array, multiplication operation is carried out on the attention moment array and the value matrix, so that a target value matrix is generated;
and after the target value matrix and the equal-length matrix are spliced line by line, inputting the target value matrix and the equal-length matrix into a preset first multi-layer perceptron to perform feature transformation, and determining space-time mixing features.
4. The submarine cable hidden danger identification method according to claim 1, further comprising:
acquiring a historical space-time mixed characteristic sequence and a historical hidden danger characteristic sequence library which are associated with the submarine cable at a prediction moment;
inputting the historical space-time mixing characteristic sequence into a preset third time sequence network to perform characteristic prediction, and outputting the predicted space-time mixing characteristic at the predicted moment;
the historical space-time mixed characteristic sequence and the predicted space-time mixed characteristic are connected in series to form a one-dimensional characteristic;
determining a condition code based on the one-dimensional feature and the historical hidden danger feature sequence library;
and carrying out identification classification by adopting the prediction space-time mixing characteristics and the condition codes through a cascade preset third multi-layer perceptron and a softmax layer, and outputting hidden danger prediction results.
5. The submarine cable hidden danger identification method according to claim 4, wherein the determining a condition code based on the one-dimensional feature and the historical hidden danger feature sequence library comprises:
extracting a plurality of history hidden trouble feature sequences from the history hidden trouble feature sequence library;
respectively carrying out linear transformation on the one-dimensional features and each history hidden trouble feature sequence to construct a prediction query matrix and a plurality of prediction key matrixes;
performing inner product operation on the prediction query matrix and each prediction key matrix respectively, and outputting a plurality of confidence coefficients;
and arranging all the confidence degrees from high to low, sequentially selecting a plurality of historical hidden danger feature sequences corresponding to the confidence degrees with a preset number of thresholds for series connection, and inputting a preset full-connection layer for channel dimension reduction to generate a condition code.
6. The submarine cable hidden danger identification method according to claim 4, wherein the third multi-layer perceptron comprises a first full-connection-layer module and a second full-connection-layer module; the step of identifying and classifying by adopting the prediction space-time mixing characteristics and the condition codes through a cascade preset third multi-layer perceptron and a softmax layer and outputting hidden danger prediction results comprises the following steps:
carrying out feature extraction on the predicted space-time mixing features by adopting a first full-connection layer module, and outputting first features;
after the first features and the condition codes are connected in series, inputting the first features and the condition codes into a second full-connection layer module for feature extraction to generate second features;
and performing feature mapping based on the second features through the softmax layer, and outputting hidden danger prediction results.
7. The submarine cable hidden danger identification method according to claim 4, further comprising:
acquiring a historical space-time mixed feature library of a plurality of submarine cables;
extracting a plurality of historical space-time mixed features in a hidden danger critical period from each historical space-time mixed feature library to respectively form a historical hidden danger feature sequence;
and forming a history hidden danger feature sequence library by adopting all the history hidden danger feature sequences.
8. Submarine cable hidden danger identification device, characterized by comprising:
the device comprises a feature extraction module, a data processing module and a data processing module, wherein the feature extraction module is used for carrying out data preprocessing on all the data to be identified and respectively adopting a preset feature extraction network to carry out feature extraction when a plurality of groups of data to be identified of the submarine cable are acquired, so as to generate visual features, corpus features and sensor features;
the equilong matrix construction module is used for mapping the visual features, the corpus features and the sensor features to a unified feature space by adopting a preset feedforward neural network respectively, and combining the output multiple groups of equilong features to construct an equilong matrix;
the space-time mixing characteristic module is used for inputting a preset first multi-layer perceptron to perform characteristic transformation after the self-attention operation is performed on the basis of the equal-length matrix, and determining space-time mixing characteristics;
and the hidden danger detection module is used for inputting the space-time mixing characteristics into a cascaded preset second multi-layer perceptron and a softmax layer to carry out identification and classification and outputting hidden danger detection results of the submarine cable.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the submarine cable hidden danger identification method according to any of claims 1-7.
10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed implements the submarine cable hidden danger identification method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311281760.XA CN117011690B (en) | 2023-10-07 | 2023-10-07 | Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311281760.XA CN117011690B (en) | 2023-10-07 | 2023-10-07 | Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117011690A true CN117011690A (en) | 2023-11-07 |
| CN117011690B CN117011690B (en) | 2024-02-09 |
Family
ID=88567583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311281760.XA Active CN117011690B (en) | 2023-10-07 | 2023-10-07 | Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117011690B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021056709A1 (en) * | 2019-09-24 | 2021-04-01 | 平安科技（深圳）有限公司 | Method and apparatus for recognizing similar questions, computer device and storage medium |
| CN113298110A (en) * | 2021-03-24 | 2021-08-24 | 国网河北省电力有限公司沧州供电分公司 | Submarine cable fault diagnosis method, device and equipment |
| CN113468764A (en) * | 2021-07-22 | 2021-10-01 | 西南石油大学 | Submarine cable state prediction method based on time sequence |
| CN114856941A (en) * | 2022-05-18 | 2022-08-05 | 江苏科技大学 | Offshore wind power plant and unit fault diagnosis operation and maintenance system and diagnosis operation and maintenance method thereof |
| CN116842423A (en) * | 2023-06-16 | 2023-10-03 | 成都航天科工大数据研究院有限公司 | Aeroengine fault diagnosis method and system based on multi-mode deep learning |
-
2023
- 2023-10-07 CN CN202311281760.XA patent/CN117011690B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021056709A1 (en) * | 2019-09-24 | 2021-04-01 | 平安科技（深圳）有限公司 | Method and apparatus for recognizing similar questions, computer device and storage medium |
| CN113298110A (en) * | 2021-03-24 | 2021-08-24 | 国网河北省电力有限公司沧州供电分公司 | Submarine cable fault diagnosis method, device and equipment |
| CN113468764A (en) * | 2021-07-22 | 2021-10-01 | 西南石油大学 | Submarine cable state prediction method based on time sequence |
| CN114856941A (en) * | 2022-05-18 | 2022-08-05 | 江苏科技大学 | Offshore wind power plant and unit fault diagnosis operation and maintenance system and diagnosis operation and maintenance method thereof |
| CN116842423A (en) * | 2023-06-16 | 2023-10-03 | 成都航天科工大数据研究院有限公司 | Aeroengine fault diagnosis method and system based on multi-mode deep learning |
Non-Patent Citations (1)
| Title |
|---|
| CHUNYING XU ET AL: "A Hybrid Model Integrating CNN–BiLSTM and CBAM for Anchor Damage Events Recognition of Submarine Cables", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》, pages 1 - 11 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117011690B (en) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111079532B (en) | Video content description method based on text self-encoder | |
| CN111325323B (en) | Automatic power transmission and transformation scene description generation method integrating global information and local information | |
| CN115471216B (en) | Data management method of intelligent laboratory management platform | |
| CN112818035A (en) | Network fault prediction method, terminal equipment and storage medium | |
| CN116414990B (en) | Vehicle fault diagnosis and prevention method | |
| CN111444399B (en) | Reply content generation method, device, equipment and readable storage medium | |
| CN112257263A (en) | Equipment residual life prediction system based on self-attention mechanism | |
| CN116862847A (en) | Interactive segmentation method and system for infrared image power equipment | |
| CN116150901A (en) | Rolling bearing residual life prediction method based on attention-enhanced time-frequency converter | |
| CN115587597A (en) | Sentiment analysis method and device of aspect words based on clause-level relational graph | |
| CN117011690B (en) | Submarine cable hidden danger identification method, submarine cable hidden danger identification device, submarine cable hidden danger identification equipment and submarine cable hidden danger identification medium | |
| CN112559741A (en) | Nuclear power equipment defect recording text classification method, system, medium and electronic equipment | |
| CN112395954A (en) | Power transmission line specific fault recognition system based on combination of natural language model and target detection algorithm | |
| CN116662307A (en) | Intelligent early warning method, system and equipment based on multi-source data fusion | |
| CN113474795A (en) | Answering cognitive queries from sensor input signals | |
| CN115935969A (en) | Heterogeneous data feature extraction method based on multi-mode information fusion | |
| CN116109978A (en) | Self-constrained dynamic text feature-based unsupervised video description method | |
| CN113920497B (en) | Nameplate recognition model training method, nameplate recognition method and related devices | |
| CN115587007A (en) | Robertta-based weblog security detection method and system | |
| CN110852066A (en) | Multi-language entity relation extraction method and system based on confrontation training mechanism | |
| CN114913448A (en) | Video understanding method, device, equipment, storage medium and computer program product | |
| CN113378546A (en) | Non-autoregressive sentence sequencing method | |
| CN116051155B (en) | User identification method, apparatus, device, storage medium, and program product | |
| CN115249133B (en) | Building construction process risk classification method based on width learning network | |
| CN116757753A (en) | User identification method, device, computer equipment and storage medium |
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 |