CN113051822A - Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model - Google Patents
Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model Download PDFInfo
- Publication number
- CN113051822A CN113051822A CN202110319202.2A CN202110319202A CN113051822A CN 113051822 A CN113051822 A CN 113051822A CN 202110319202 A CN202110319202 A CN 202110319202A CN 113051822 A CN113051822 A CN 113051822A
- Authority
- CN
- China
- Prior art keywords
- sample
- model
- industrial system
- lstm
- anomaly detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 46
- 230000005856 abnormality Effects 0.000 claims abstract description 24
- 230000002159 abnormal effect Effects 0.000 claims abstract description 15
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 238000000638 solvent extraction Methods 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 27
- 239000013598 vector Substances 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 15
- 230000003993 interaction Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 6
- 238000012806 monitoring device Methods 0.000 claims description 6
- 238000012512 characterization method Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000002372 labelling Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
An industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model comprises the following steps: 1) sample partitioning and normalization: dividing original industrial system data into samples by adopting a sliding window; 2) constructing an abnormality detection model: adopting an image attention network and an LSTM automatic coding machine to construct an anomaly detection model; 3) real-time anomaly detection: an abnormality degree score is calculated based on the reconstruction error, and an abnormal state is determined based on the calculation. The invention adopts an automatic coding machine to train an abnormality detection model in an unsupervised mode without providing an abnormality labeling sample; the graph attention network is adopted to mine the association among different dimensions of the industrial system, and the anomaly detection accuracy in the complex industrial system is improved.
Description
Technical Field
The invention relates to a machine learning technology, in particular to an industrial system anomaly detection method.
Background
With the rapid development of the industrial internet, more acute and efficient automation control and resource allocation of the industrial system are realized. However, as the industrial internet breaks the boundaries of the network world and the physical world, industrial manufacturing systems are more vulnerable to external malicious activities. In addition, there are inevitable production problems in industrial manufacturing systems, such as equipment failure, performance degradation, quality defects, and the like. If the abnormal conditions such as intrusion, failure and the like in the industrial production cannot be detected in time, the serious loss can be brought to the whole manufacturing system. Therefore, the anomaly detection is a basic requirement of the industrial Internet and has very important significance for intelligent manufacturing enterprises.
The modern industrial manufacturing system realizes the sensing and recording of the production running state, environment and process through monitoring equipment such as a sensor, a controller, an intelligent instrument and the like, and accumulates a large amount of industrial data. Therefore, the data-driven model is the mainstream means of current anomaly detection. The current mainstream data-driven anomaly detection model generally has two characteristics: 1) training is performed in an unsupervised manner. The operation process and monitoring data of the industrial system are very complex, so that deep domain knowledge is needed for understanding the abnormal state of the industrial system, and the marking work cost of the abnormal sample is huge, so that enough abnormal samples cannot be obtained from the real industrial system for training. 2) Deep learning is used as the bottom layer model. This is because the industrial system data is usually collected continuously by a large number of monitoring devices, and is a high-dimensional time series data, which makes it difficult to rely on manual feature engineering.
Most of the existing unsupervised anomaly detection models based on deep learning are based on a recurrent neural network and an automatic coding machine: the method comprises the steps of firstly mapping original industrial system data to a low-dimensional feature space based on an encoder (the encoder is usually designed based on a recurrent neural network), regarding the features of the original data in the space as potential modes of the original data, then reconstructing the original data from the low-dimensional feature space based on a decoder, and if the difference between the reconstructed data and the original data is large, regarding that the original data does not conform to the potential modes, namely, an exception occurs.
However, the existing anomaly detection model based on the recurrent neural network + the automatic coding machine still has the following defects: first, the existing method considers multidimensional time series industrial system data as an integral input anomaly detection model, but actually different monitoring devices of the industrial system are not completely independent, so that potential correlation exists between different dimensions of the industrial system data. Secondly, the existing method treats the data of the multidimensional time sequence industrial system equally in different dimensions (usually different monitoring devices), but actually, the influence degrees of the different dimensions on the abnormal state under different scenes are inconsistent.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model, an automatic coding machine is adopted to train an anomaly detection model in an unsupervised mode, and an anomaly labeling sample is not required to be provided; the graph attention network is adopted to mine the association among different dimensions of the industrial system, and the anomaly detection accuracy in the complex industrial system is improved.
The technical scheme adopted by the invention for solving the technical problems is as follows:
an industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model comprises the following steps:
1) sample partitioning and normalization: dividing original industrial system data into samples by adopting a sliding window;
2) constructing an abnormality detection model: adopting an image attention network and an LSTM automatic coding machine to construct an anomaly detection model;
3) real-time anomaly detection: an abnormality degree score is calculated based on the reconstruction error, and an abnormal state is determined based on the calculation.
Further, in the step 1), the original multi-dimensional time sequence of the industrial system data is given
Wherein T is data capacity, F is data dimension, and the steps of sample division and normalization are as follows:
(1-1) sample division: dividing X into a sample set XS consisting of a plurality of samples based on a sliding window with the width W and the step size S, wherein the number of the divided samples is N, and each sample
(1-2) normalization: carrying out standardization operation on the data based on a Z-Score method, so that the mean value of the data in each dimension in each window is 0, and the standard deviation of the data in each dimension is 1;
in the step 2), the step of constructing the anomaly detection model is as follows:
(2-1) association graph construction: converting each sample x into a form G of a correlation graphx(V, E, a). V, E, A is a node set, an edge set, and an attribute set, respectively, and is described as follows: first, each node V in ViRepresenting one dimension of sample data, wherein one dimension corresponds to one monitoring device in the industrial system; next, each edge E in EijRepresentative node viAnd vjThere is an association between them, and an edge is set for each pair of nodes, i.e. GxIs a full connection diagram; again, each element a in AiIs v isiSample data vector in the indicated dimension, representing vi(ii) an attribute of (d);
(2-2) model construction: the anomaly detection model is a deep neural network and comprises an interaction layer, a coding layer, a decoding layer and a reconstruction layer;
(2-3) model training: in order to realize unsupervised model training, the mean square error between a reconstructed sample y and an original sample x generated by a decoder is used as a loss function of the model, and the model is optimally trained in a gradient descent mode on the basis.
The interaction layer uses the graph attention network to process the input of the model, and the input of the interaction layer is the sample G in the form of the association graph obtained in the step (2-1)xThe processing steps are as follows: step1 for GxAny one side eijSetting a learnable weight wijThe calculation method is shown as formula (1), wherein q is a learnable parameter vector, σ () is a nonlinear activation function, ^ is used for splicing a plurality of vectors, and L is viThe number of the neighbor nodes; step2 for each node viAnd calculating the weighted average vector of the neighbor nodes asviIs characterized by a vector giAs shown in formula (3); step3 splices the characterization vectors of all nodes into a matrix
As the output of the interaction layer, wherein the row of z is the dimension of the node representation vector, and the column is the number of the nodes;
the coding layer processes the interaction layer output z of consecutive N samples using the LSTM network1、z2、…、zNThe processing procedure is shown as formula (4), i.e. the current hidden state matrix htFrom its previous hidden state matrix ht-1And a current input state matrix ztCo-generating;
ht＝LSTM(ht-1,zt) (4)
the decoding layer uses LSTM network to process the output h of the coding layer1、h2、…、hNThe processing strategy is consistent with the coding layer, and a characteristic matrix sequence r is output1、r2、…、rN；
The reconstruction layer processes the output r of the decoding layer using a full-link layer with a neuron number of F1、r2、…、rNObtaining the original sample
Of the reconstructed samples
Further, in the step (3), the real-time abnormality detection step includes:
(3-1) scoring the degree of abnormality: giving a real-time sample x, and firstly inputting the real-time sample x into a trained anomaly detection model to obtain a reconstructed sample y; then, the mean square error of x and y is calculated as the abnormal degree fraction p of xx；
(3-2) abnormality determination: if p isxIf the value is larger than the specified threshold value, x is judged to be abnormal.
The invention has the following beneficial effects: 1. an automatic coding machine is adopted to train an abnormality detection model in an unsupervised mode, and an abnormality labeling sample is not required to be provided; 2. the graph attention network is adopted to mine the association among different dimensions of the industrial system, and the anomaly detection accuracy in the complex industrial system is improved.
Drawings
FIG. 1 is a flow chart of an industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model;
FIG. 2 is a network architecture diagram of an anomaly detection model;
FIG. 3 is a schematic diagram of a dimension characterization method based on a graph attention network;
fig. 4 is a diagram of an encoding-decoding network structure.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
Referring to fig. 1 to 4, an industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model includes the following steps:
1) sample partitioning and normalization: dividing original industrial system data into samples by adopting a sliding window;
2) constructing an abnormality detection model: adopting an image attention network and an LSTM automatic coding machine to construct an anomaly detection model;
3) real-time anomaly detection: an abnormality degree score is calculated based on the reconstruction error, and an abnormal state is determined based on the calculation.
Further, in the step 1),industrial system data given an original multi-dimensional time sequence
Wherein T is data capacity, F is data dimension, and the steps of sample division and normalization are as follows:
(1-1) sample division: dividing X into a sample set XS consisting of a plurality of samples based on a sliding window with the width W and the step size S, wherein the number of the divided samples is N, and each sample
(1-2) normalization: carrying out standardization operation on the data based on a Z-Score method, so that the mean value of the data in each dimension in each window is 0, and the standard deviation of the data in each dimension is 1;
in the step 2), the step of constructing the anomaly detection model is as follows:
(2-1) association graph construction: converting each sample x into a form G of a correlation graphx(V, E, a). V, E, A is a node set, an edge set, and an attribute set, respectively, and is described as follows: first, each node V in ViRepresenting one dimension of sample data, wherein one dimension corresponds to one monitoring device in the industrial system; next, each edge E in EijRepresentative node viAnd vjThere is an association between; since it is generally not known in advance which dimensions have associations between them, an edge is set for each pair of nodes, i.e. GxIs a full connection diagram; again, each element a in AiIs v isiSample data vector in the indicated dimension, representing vi(ii) an attribute of (d);
(2-2) model construction: referring to fig. 2, the anomaly detection model is a deep neural network including an interaction layer, a coding layer, a decoding layer, and a reconstruction layer;
(2-3) model training: in order to realize unsupervised model training, the mean square error between a reconstructed sample y and an original sample x generated by a decoder is used as a loss function of the model, and the model is optimally trained in a gradient descent mode on the basis.
Referring to FIG. 3, the interaction layer uses the graph attention network to process the input of the model, which is the sample G in the form of the association graph obtained in step (2-1)xThe processing steps are as follows: step1 for GxAny one side eijSetting a learnable weight wijThe calculation method is shown as formula (1), wherein q is a learnable parameter vector, σ () is a nonlinear activation function, ^ is used for splicing a plurality of vectors, and L is viThe number of the neighbor nodes; step2 for each node viCalculating its weighted average vector with all neighboring nodes as viIs characterized by a vector giAs shown in formula (3); step3 splices the characterization vectors of all nodes into a matrix
As the output of the interaction layer, wherein the row of z is the dimension of the node representation vector, and the column is the number of the nodes;
referring to fig. 4, the coding layer processes an interaction layer output z of consecutive N samples using an LSTM network1、z2、…、zNThe processing procedure is shown as formula (4), i.e. the current hidden state matrix htFrom its previous hidden state matrix ht-1And a current input state matrix ztCo-generating;
ht＝LSTM(ht-1,zt) (4)
the decoding layer uses LSTM network to process the output h of the coding layer1、h2、…、hNThe processing strategy is consistent with the coding layer, and a characteristic matrix sequence r is output1、r2、…、rN；
The reconstruction layer processes the output r of the decoding layer using a full-link layer with a neuron number of F1、r2、…、rNObtaining the original sample
Of the reconstructed samples
Further, in the step (3), the real-time abnormality detection step includes:
(3-1) scoring the degree of abnormality: giving a real-time sample x, and firstly inputting the real-time sample x into a trained anomaly detection model to obtain a reconstructed sample y; then, the mean square error of x and y is calculated as the abnormal degree fraction p of xx；
(3-2) abnormality determination: if p isxIf the value is larger than the specified threshold value, x is judged to be abnormal.
The embodiments described in this specification are merely illustrative of implementations of the inventive concepts, which are intended for purposes of illustration only. The scope of the present invention should not be construed as being limited to the particular forms set forth in the examples, but rather as being defined by the claims and the equivalents thereof which can occur to those skilled in the art upon consideration of the present inventive concept.
Claims (5)
1. An industrial system anomaly detection method based on a graph attention network and an LSTM automatic coding model is characterized by comprising the following steps:
1) sample partitioning and normalization: dividing original industrial system data into samples by adopting a sliding window;
2) constructing an abnormality detection model: adopting an image attention network and an LSTM automatic coding machine to construct an anomaly detection model;
3) real-time anomaly detection: an abnormality degree score is calculated based on the reconstruction error, and an abnormal state is determined based on the calculation.
2. The method for detecting the abnormality of the industrial system based on the graph attention network and the LSTM automatic coding model as claimed in claim 1, wherein in the step 1), the original multi-dimensional time sequence of the industrial system data is given
Wherein T is data capacity, F is data dimension, and the steps of sample division and normalization are as follows:
(1-1) sample division: dividing X into a sample set XS consisting of a plurality of samples based on a sliding window with the width W and the step size S, wherein the number of the divided samples is N, and each sample
(1-2) normalization: the data were normalized based on the Z-Score method such that the mean and standard deviation of the data in each dimension in each window was 0 and 1.
3. The method for detecting the abnormality of the industrial system based on the graph attention network and the LSTM automatic coding model as claimed in claim 1 or 2, wherein in the step 2), the step of constructing the abnormality detection model is as follows:
(2-1) association graph construction: converting each sample x into a form G of a correlation graphxWhere V, E, A is node set, edge set, and attribute set, respectively, as follows: first, each node V in ViRepresenting one dimension of sample data, wherein one dimension corresponds to one monitoring device in the industrial system; next, each edge E in EijRepresentative node viAnd vjThere is an association between them, and an edge is set for each pair of nodes, i.e. GxIs a full connection diagram; again, each element a in AiIs v isiSample data vector in the indicated dimension, representing vi(ii) an attribute of (d);
(2-2) model construction: the anomaly detection model is a deep neural network and comprises an interaction layer, a coding layer, a decoding layer and a reconstruction layer;
(2-3) model training: in order to realize unsupervised model training, the mean square error between a reconstructed sample y and an original sample x generated by a decoder is used as a loss function of the model, and the model is optimally trained in a gradient descent mode on the basis.
4. The method for detecting the abnormality of the industrial system based on the graph attention network and the LSTM automatic coding model as claimed in claim 3, wherein the interaction layer uses the graph attention network to process the input of the model, and the input of the interaction layer is the sample G obtained in the step (2-1) in the form of the associated graphxThe processing steps are as follows: step1 for GxAny one side eijSetting a learnable weight wijThe calculation method is shown as formula (1), wherein q is a learnable parameter vector, σ () is a nonlinear activation function, ^ is used for splicing a plurality of vectors, and L is viThe number of the neighbor nodes; step2 for each node viCalculating its weighted average vector with all neighboring nodes as viIs characterized by a vector giAs shown in formula (3); step3 splices the characterization vectors of all nodes into a matrix
As the output of the interaction layer, wherein the row of z is the dimension of the node representation vector, and the column is the number of the nodes;
the coding layer processes the interaction layer output z of consecutive N samples using the LSTM network1、z2、…、zNThe processing procedure is shown as formula (4), i.e. the current hidden state matrix htFrom its previous hidden state matrix ht-1And a current input state matrix ztCo-generating;
ht＝LSTM(ht-1,zt) (4)
the decoding layer uses LSTM network to process the output h of the coding layer1、h2、…、hNThe processing strategy is consistent with the coding layer, and a characteristic matrix sequence r is output1、r2、…、rN；
The reconstruction layer processes the output r of the decoding layer using a full-link layer with a neuron number of F1、r2、…、rNObtaining the original sample
Of the reconstructed samples
5. The method for detecting the anomaly of the industrial system based on the graph attention network and the LSTM automatic coding model as claimed in claim 1 or 2, wherein in the step (3), the real-time anomaly detection step is as follows:
(3-1) scoring the degree of abnormality: giving a real-time sample x, and firstly inputting the real-time sample x into a trained anomaly detection model to obtain a reconstructed sample y; then, the mean square error of x and y is calculated as the abnormal degree fraction p of xx；
(3-2) abnormality determination: if p isxIf the value is larger than the specified threshold value, x is judged to be abnormal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110319202.2A CN113051822A (en) | 2021-03-25 | 2021-03-25 | Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110319202.2A CN113051822A (en) | 2021-03-25 | 2021-03-25 | Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113051822A true CN113051822A (en) | 2021-06-29 |
Family
ID=76515731
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110319202.2A Pending CN113051822A (en) | 2021-03-25 | 2021-03-25 | Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113051822A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113811009A (en) * | 2021-09-24 | 2021-12-17 | 之江实验室 | Multi-base-station cooperative wireless network resource allocation method based on space-time feature extraction reinforcement learning |
| CN113962273A (en) * | 2021-09-22 | 2022-01-21 | 北京必示科技有限公司 | Multi-index-based time series anomaly detection method and system and storage medium |
| CN114265882A (en) * | 2021-12-24 | 2022-04-01 | 中冶赛迪重庆信息技术有限公司 | Method, system, device and medium for detecting time sequence signal point abnormity |
| CN115018012A (en) * | 2022-07-21 | 2022-09-06 | 北京航空航天大学 | Internet of things time sequence anomaly detection method and system under high-dimensional characteristic |
| CN115049472A (en) * | 2022-05-12 | 2022-09-13 | 之江实验室 | Unsupervised credit card anomaly detection method based on multi-dimensional feature tensor |
| CN115098563A (en) * | 2022-07-14 | 2022-09-23 | 中国海洋大学 | Time sequence abnormity detection method and system based on GCN and attention VAE |
| CN115983087A (en) * | 2022-09-16 | 2023-04-18 | 山东财经大学 | Method for detecting time sequence data abnormity by combining attention mechanism and LSTM and terminal |
| CN116664000A (en) * | 2023-06-13 | 2023-08-29 | 无锡物联网创新中心有限公司 | Industrial equipment abnormality detection method and related device based on long-short-term memory network |
| WO2023197617A1 (en) * | 2022-04-11 | 2023-10-19 | 浙江工业大学 | Method for detecting and diagnosing production abnormality of industrial system on basis of multi-dimensional sensing data |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112217674A (en) * | 2020-10-12 | 2021-01-12 | 西安交通大学 | Alarm root cause identification method based on causal network mining and graph attention network |
-
2021
- 2021-03-25 CN CN202110319202.2A patent/CN113051822A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112217674A (en) * | 2020-10-12 | 2021-01-12 | 西安交通大学 | Alarm root cause identification method based on causal network mining and graph attention network |
Non-Patent Citations (1)
| Title |
|---|
| 沈潇军: "一种基于LSTM 自动编码机的工业***异常检测方法", 电信科学, 20 July 2020 (2020-07-20) * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113962273A (en) * | 2021-09-22 | 2022-01-21 | 北京必示科技有限公司 | Multi-index-based time series anomaly detection method and system and storage medium |
| CN113811009A (en) * | 2021-09-24 | 2021-12-17 | 之江实验室 | Multi-base-station cooperative wireless network resource allocation method based on space-time feature extraction reinforcement learning |
| CN114265882A (en) * | 2021-12-24 | 2022-04-01 | 中冶赛迪重庆信息技术有限公司 | Method, system, device and medium for detecting time sequence signal point abnormity |
| WO2023197617A1 (en) * | 2022-04-11 | 2023-10-19 | 浙江工业大学 | Method for detecting and diagnosing production abnormality of industrial system on basis of multi-dimensional sensing data |
| CN115049472A (en) * | 2022-05-12 | 2022-09-13 | 之江实验室 | Unsupervised credit card anomaly detection method based on multi-dimensional feature tensor |
| CN115049472B (en) * | 2022-05-12 | 2024-01-26 | 之江实验室 | Unsupervised credit card anomaly detection method based on multidimensional feature tensor |
| CN115098563A (en) * | 2022-07-14 | 2022-09-23 | 中国海洋大学 | Time sequence abnormity detection method and system based on GCN and attention VAE |
| CN115018012A (en) * | 2022-07-21 | 2022-09-06 | 北京航空航天大学 | Internet of things time sequence anomaly detection method and system under high-dimensional characteristic |
| CN115983087A (en) * | 2022-09-16 | 2023-04-18 | 山东财经大学 | Method for detecting time sequence data abnormity by combining attention mechanism and LSTM and terminal |
| CN115983087B (en) * | 2022-09-16 | 2023-10-13 | 山东财经大学 | Method for detecting time sequence data abnormality by combining attention mechanism with LSTM (link state machine) and terminal |
| CN116664000A (en) * | 2023-06-13 | 2023-08-29 | 无锡物联网创新中心有限公司 | Industrial equipment abnormality detection method and related device based on long-short-term memory network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113051822A (en) | Industrial system anomaly detection method based on graph attention network and LSTM automatic coding model | |
| CN110162018B (en) | Incremental equipment fault diagnosis method based on knowledge distillation and hidden layer sharing | |
| CN111504676B (en) | Equipment fault diagnosis method, device and system based on multi-source monitoring data fusion | |
| CN112461537B (en) | Wind power gear box state monitoring method based on long-time and short-time neural network and automatic coding machine | |
| CN111798051A (en) | Air quality space-time prediction method based on long-short term memory neural network | |
| CN105891422B (en) | The electronic nose Gas Distinguishing Method that the limit learns drift compensation is migrated based on source domain | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| Khediri et al. | Variable window adaptive kernel principal component analysis for nonlinear nonstationary process monitoring | |
| CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
| CN109635763B (en) | Crowd density estimation method | |
| CN111046961B (en) | Fault classification method based on bidirectional long-time and short-time memory unit and capsule network | |
| CN113723010A (en) | Bridge damage early warning method based on LSTM temperature-displacement correlation model | |
| WO2023197617A1 (en) | Method for detecting and diagnosing production abnormality of industrial system on basis of multi-dimensional sensing data | |
| CN110991471B (en) | Fault diagnosis method for high-speed train traction system | |
| CN114970715A (en) | Variable working condition fault diagnosis method and system under small sample and unbalanced data constraint | |
| CN115358259A (en) | Self-learning-based unsupervised cross-working-condition bearing fault diagnosis method | |
| CN112738014A (en) | Industrial control flow abnormity detection method and system based on convolution time sequence network | |
| CN115205689A (en) | Improved unsupervised remote sensing image anomaly detection method | |
| CN112784920A (en) | Cloud-side-end-coordinated dual-anti-domain self-adaptive fault diagnosis method for rotating part | |
| Fu et al. | MCA-DTCN: A novel dual-task temporal convolutional network with multi-channel attention for first prediction time detection and remaining useful life prediction | |
| JP7384218B2 (en) | Anomaly detection device, anomaly detection method and program | |
| CN116306289A (en) | Multi-source domain self-adaption-based electromechanical device cross-domain residual life prediction method | |
| CN116259172A (en) | Urban road speed prediction method considering space-time characteristics of traffic network | |
| CN114048546A (en) | Graph convolution network and unsupervised domain self-adaptive prediction method for residual service life of aircraft engine | |
| CN112381213A (en) | Industrial equipment residual life prediction method based on bidirectional long-term and short-term memory network |
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 |