CN111260024A - Fault detection method and system based on combination of long-term and short-term memory and typical correlation - Google Patents
Fault detection method and system based on combination of long-term and short-term memory and typical correlation Download PDFInfo
- Publication number
- CN111260024A CN111260024A CN202010017077.5A CN202010017077A CN111260024A CN 111260024 A CN111260024 A CN 111260024A CN 202010017077 A CN202010017077 A CN 202010017077A CN 111260024 A CN111260024 A CN 111260024A
- Authority
- CN
- China
- Prior art keywords
- long
- term memory
- short term
- neural network
- memory neural
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 74
- 230000006403 short-term memory Effects 0.000 title claims abstract description 35
- 230000007787 long-term memory Effects 0.000 title claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 claims abstract description 100
- 230000015654 memory Effects 0.000 claims abstract description 95
- 238000000034 method Methods 0.000 claims abstract description 25
- 238000013507 mapping Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 29
- 239000011159 matrix material Substances 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000010219 correlation analysis Methods 0.000 claims description 6
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 230000007774 longterm Effects 0.000 claims description 2
- 238000000844 transformation Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000000491 multivariate analysis Methods 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2263—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing using neural 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Debugging And Monitoring (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to the technical field of fault detection, and discloses a fault detection method and a system based on long-short term memory and typical correlation combination, which are used for fully analyzing dynamic and nonlinear characteristics of generated faults so as to realize fault detection; the method comprises the steps of obtaining a first long short term memory neural network for constructing a corresponding input set and a second long short term memory neural network for constructing a corresponding output set; analyzing a linear mapping relation between the first long-short term memory neural network and the second long-short term memory neural network by adopting a typical correlation method, and optimizing the first long-short term memory neural network and the second long-short term memory neural network; analyzing a residual vector between the first long-term memory neural network output layer and the second long-term memory neural network output layer by using the typical correlation of the first long-term memory neural network output layer and the second long-term memory neural network output layer, and setting a detection threshold; and acquiring real-time operation data of an object to be analyzed, inputting the real-time operation data into the first long-short term memory neural network and the second long-short term memory neural network to calculate detection statistics of the real-time data, and comparing the detection statistics of the real-time data with a detection threshold value to realize fault detection.
Description
Technical Field
The invention relates to the technical field of fault detection, in particular to a fault detection method and system based on long-short term memory and typical correlation combination.
Background
As automation technology and system security requirements increase, fault detection and performance monitoring of systems also become increasingly important. Model-based fault detection methods have gained wide acceptance over the past few decades. The performance of such methods depends on the accuracy of the model. Meanwhile, due to the development of sensor technology and information technology, data acquisition becomes easier and easier, and abundant system operation data are accumulated in the acquisition process. Therefore, the process monitoring and fault detection technology based on data driving becomes a research hotspot in the field of fault detection.
At present, fault detection methods based on multivariate analysis, such as principal component analysis, partial least squares and typical correlation analysis, are mostly used. However, the analysis principle of these methods is to find the linear transformation of the data set, and the constructed features cannot reflect the dynamic and nonlinear relations in the complex system.
Disclosure of Invention
The invention aims to provide a fault detection method and system based on long-short term memory and typical correlation combination, so as to fully analyze dynamic and nonlinear characteristics of generated faults and further realize fault detection.
To achieve the above object, the present invention provides a fault detection method based on a combination of long-short term memory and canonical correlation, comprising:
s1: acquiring historical normal operation data of an object to be analyzed as an input set, and acquiring corresponding output of each group of historical normal operation data as an output set;
s2: constructing a first long-short term memory neural network corresponding to the input set and a second long-short term memory neural network corresponding to the output set;
s3: analyzing the mapping relation between the first long-term memory neural network and the second short-term memory neural network by adopting a typical correlation method, and calculating a correlation coefficient according to the mapping relation;
s4: optimizing the first long-short term memory neural network and the second long-short term memory neural network according to the correlation linear coefficient, and training the first long-short term memory neural network and the second long-short term memory neural network by using a back propagation algorithm until the first long-short term memory neural network and the second long-short term memory neural network which accord with set convergence are obtained;
s5: inputting the first time sequence training set into the first long-short term memory neural network which accords with the set convergence, inputting the second time sequence training set into the second long-short term memory neural network which accords with the set convergence, analyzing residual vectors between the first long-short term memory neural network and the second long-short term memory neural network by using the typical correlation of the output layers of the first long-short term memory neural network and the second long-short term memory neural network, constructing a detection statistic according to the residual vectors, and setting a detection threshold according to the detection statistic;
s6: and acquiring real-time operation data of an object to be analyzed, inputting the real-time operation data into the first long-term and second short-term memory neural networks which accord with the convergence to calculate to obtain detection statistics of the real-time data, comparing the detection statistics of the real-time data with the detection threshold, and if the detection statistics of the real-time data exceeds the detection threshold, determining that a fault occurs.
Preferably, the S2 specifically includes:
establishing a first time sequence training set of the input set and a second time sequence training set of the output set according to a preset time span;
and constructing the first long short-term memory neural network according to the first time sequence training set, and constructing the second long short-term memory neural network according to the second time sequence training set.
Preferably, the output layers of the first long-short term memory neural network and the second long-short term memory neural network output the same data dimension.
Preferably, the first long-short term memory neural network or the second long-short term memory neural network includes an LSTM layer, a maximum pooling layer, a Dropout layer, and a full connectivity layer.
Preferably, the S3 includes:
s31: outputting the characteristic data of the first long-short term memory network
And the feature data output by the second long-short term memory network
Normalization processing is carried out to obtain normalized characteristic data
S32: computing
Of the covariance matrix sigma1、Σ2Sum-sigma12Finding a linear transformation using a canonical correlation analysis such that
And
with the greatest correlation.
Preferably, the S4 includes:
s41: constructing a matrix M and carrying out singular value decomposition on the matrix M as follows:
in the formula (I), the compound is shown in the specification,
the method comprises the following steps of providing a diagonal matrix, wherein the sum of diagonal elements of the diagonal matrix is corr, the corr represents the similarity degree of two groups of data after linear transformation, R is a left matrix, V is a right singular matrix, and T represents matrix transposition;
s42: the first long-short term memory network and the second long-short term memory network adopt a back propagation algorithm to jointly train network parameters by taking the maximum correlation coefficient as an optimization target.
Preferably, the S5 includes:
s51 estimating T by using a kernel density estimation algorithm2The probability distribution of the statistics, using the radial basis function, is as follows:
wherein K (g) represents a radial basis function, and g represents an argument of the radial basis function;
set confidence α, T2Statistical threshold UCLs pass formula
Is obtained in which
Wherein xkWhere k is 1,2, …, M is the sample value of x, h is the bandwidth value of the kernel function, b is an argument, and M represents the number of samples of x.
As a general technical idea, the present invention further provides a fault detection system based on a combination of long and short term memory and typical correlation, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.
The invention has the following beneficial effects:
the invention provides a fault detection method and a system based on long-short term memory and typical correlation combination, wherein the method comprises the steps of obtaining historical normal operation data of an object to be analyzed as an input set, and obtaining corresponding output of each group of historical normal operation data as an output set; constructing a first long-short term memory neural network corresponding to the input set and a second long-short term memory neural network corresponding to the output set; analyzing the mapping relation between the first long-term memory neural network and the second short-term memory neural network by adopting a typical correlation method, and calculating a correlation coefficient according to the linear mapping relation; optimizing the first long-short term memory neural network and the second long-short term memory neural network according to the correlation coefficient, and training the first long-short term memory neural network and the second long-short term memory neural network by utilizing a back propagation algorithm until the first long-short term memory neural network and the second long-short term memory neural network which accord with the set convergence are obtained; inputting a first time sequence training set into a first long and short term memory neural network which accords with set convergence, inputting a second time sequence training set into a second long and short term memory neural network which accords with the set convergence, analyzing a residual vector between the first long and short term memory neural network and the second long and short term memory neural network by using the typical correlation of the output layers of the first long and short term memory neural network and the second long and short term memory neural network, constructing a detection statistic according to the residual vector, and setting a detection threshold according to the detection statistic; the real-time operation data of an object to be analyzed is acquired and input into the first long-term and short-term memory neural networks which accord with the convergence to calculate the detection statistic of the real-time data, the detection statistic of the real-time data is compared with the detection threshold, if the detection statistic of the real-time data exceeds the detection threshold, the fault is considered to occur, the dynamic and nonlinear characteristics of the fault can be fully analyzed, and therefore fault detection is achieved.
The present invention will be described in further detail below with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of a fault detection method based on a combination of long-short term memory and typical correlations according to a preferred embodiment of the present invention;
FIG. 2 is a schematic diagram of a closed loop continuous stirred tank reactor according to a preferred embodiment of the present invention;
fig. 3 is a diagram of the effect of the failure 6 of the preferred embodiment of the present invention detected by the CCA method;
fig. 4 is a diagram of the effect of the failure 6 of the preferred embodiment of the present invention detected by the LSTM-CCA method.
Detailed Description
The embodiments of the invention will be described in detail below with reference to the drawings, but the invention can be implemented in many different ways as defined and covered by the claims.
Example 1
As shown in fig. 1, the present embodiment provides a fault detection method based on a combination of long-short term memory and typical correlation, which includes:
s1: acquiring historical normal operation data of an object to be analyzed as an input set, and acquiring corresponding output of each group of historical normal operation data as an output set;
s2: constructing a first long-short term memory neural network corresponding to the input set and a second long-short term memory neural network corresponding to the output set;
s3: analyzing the mapping relation between the first long-term memory neural network and the second short-term memory neural network by adopting a typical correlation method, and calculating a correlation coefficient according to the mapping relation;
s4: optimizing the first long-short term memory neural network and the second long-short term memory neural network according to the correlation coefficient, and training the first long-short term memory neural network and the second long-short term memory neural network by utilizing a back propagation algorithm until the first long-short term memory neural network and the second long-short term memory neural network which accord with the set convergence are obtained;
s5: inputting a first time sequence training set into a first long and short term memory neural network which accords with set convergence, inputting a second time sequence training set into a second long and short term memory neural network which accords with the set convergence, analyzing a residual vector between the first long and short term memory neural network and the second long and short term memory neural network by using the typical correlation of the output layers of the first long and short term memory neural network and the second long and short term memory neural network, constructing a detection statistic according to the residual vector, and setting a detection threshold according to the detection statistic;
s6: and acquiring real-time operation data of an object to be analyzed, inputting the real-time operation data into the first long-term and short-term memory neural networks which accord with convergence to calculate to obtain detection statistics of the real-time data, comparing the detection statistics of the real-time data with a detection threshold, and if the detection statistics of the real-time data exceeds the detection threshold, determining that a fault occurs.
The fault detection method based on the combination of the long-short term memory and the typical correlation combines the typical correlation analysis with the neural network, learns the nonlinear transformation by utilizing the parameter training of the network, and can fully analyze the dynamic and nonlinear characteristics of the generated fault so as to realize the fault detection. Can be applied to complex and dynamically-changing nonlinear systems.
Specifically, a fault-free historical operation training set is selected, wherein each sample comprises system input data U and output data Y:
in the formula, U represents an input set, Um TThe elements in the input set are represented by,
wherein Y represents an output set, Ym TThe elements in the output set are represented as,
scaling U, Y to [0,1, respectively]In between, training is conveniently carried out after the neural network is input, and the scaled system input data is obtained
Outputting the data
Selecting proper time span k according to certain rule of system input and output, and utilizing
And
construction of a time series training set Us、YsThe time sequence input to the network at the t-th moment is:
as a preferred implementation manner of this embodiment, the S2 specifically includes:
establishing a first time sequence training set of the input set and a second time sequence training set of the output set according to a preset time span;
and constructing the first long short-term memory neural network according to the first time sequence training set, and constructing the second long short-term memory neural network according to the second time sequence training set.
As a preferred implementation manner of this embodiment, the output layers of the first long-short term memory neural network and the second long-short term memory neural network output the same data dimension.
As a preferred embodiment of this embodiment, the first long-short term memory neural network or the second long-short term memory neural network includes an LSTM layer, a maximum pooling layer, a Dropout layer, and a full-link layer. The specific network structure is as follows: LSTM (n1 units) - > Dropout- > LSTM (n2 units) - > Dropout () - > LSTM (n3 units) - > MaxPooling- > all-connected layer (n4 nodes) - > output layer (n5 nodes).
As a preferred embodiment of this embodiment, the S3 includes:
s31: outputting the characteristic data of the first long-short term memory network
And the feature data output by the second long-short term memory network
Normalization processing is carried out to obtain normalized characteristic data
S32: computing
Of the covariance matrix sigma1、Σ2Sum-sigma12Finding a linear transformation using a canonical correlation analysis such that
And
with the greatest correlation.
As a preferred embodiment of this embodiment, the S4 includes:
s41: constructing a matrix M and carrying out singular value decomposition on the matrix M as follows:
in the formula (I), the compound is shown in the specification,
the method comprises the following steps of providing a diagonal matrix, wherein the sum of diagonal elements of the diagonal matrix is corr, the corr represents the similarity degree of two groups of data after linear transformation, R is a left matrix, V is a right singular matrix, and T represents matrix transposition;
s42: the first long-short term memory network and the second long-short term memory network adopt a back propagation algorithm to jointly train network parameters by taking the maximum correlation coefficient as an optimization target.
As a preferred embodiment of this embodiment, the S5 includes:
s51 estimation of nuclear densityMethod of estimating T by calculation2The probability distribution of the statistics, using the radial basis function, is as follows:
in the formula, k (g) represents a radial basis function, and g represents an argument of the radial basis function.
Set confidence α, T2Statistical threshold UCLs pass formula
Is obtained in which
Wherein xkWhere k is 1,2, …, M is the sample value of x, h is the bandwidth value of the kernel function, b is an argument, and M represents the number of samples of x.
The fault detection method based on long-short term memory and typical correlation combination provided by the invention is applicable to all reactors driven by data, and in the embodiment, the method is further explained and verified by taking a closed loop Continuous Stirred Tank Reactor (CSTR) as shown in FIG. 2 as an example.
The schematic diagram of a CSTR is shown in fig. 2, and the system reaction model can be abstracted as:
wherein the system input is u ═ CiTiTci]TThe system output is y ═ cT TcQc]T，νiK is a constant for process noise
The remaining parameters are shown in Table 1.
TABLE 1 values of constants in CSTR model
This example considers 6 minor fault scenarios as shown in table 2:
table 2 CSTR model 6 minor faults
The example utilizes the system input and output data sets obtained by operating the CSTR model to carry out experiments to verify the feasibility and the effectiveness of the invention.
The CSTR model is operated to simulate the reaction to occur for 20 hours to generate a normal data set and a fault data set, all system variables are sampled every 1min, certain random interference is given to input variables to simulate the real situation as much as possible, and the dynamic and nonlinear properties of the whole system are increased. The failure data set results from the introduction of a corresponding failure after remaining fault-free for 200 min. For each fault, a fault detection method (LSTM-CCA) and a CCA method based on long short term memory network (LSTM) feature extraction and typical correlation analysis, which are used for fault detection, are respectively established by using 1200 fault-free samples, and finally, a fault data set is used for verification, and performance evaluation of fault detection is measured by using index fault False Alarm Rate (FAR) and fault Missed Detection Rate (MDR), as shown in table 3:
table 3 6 minor faults under CSTR model
The calculation formulas of the false Fault Alarm Rate (FAR) and the failure Missing Detection Rate (MDR) are as follows:
wherein, the UCLs represent threshold values calculated in an off-line training stage. The effect of the fault 6 detected by the CCA method and the LSTM-CCA method is shown in fig. 3 and 4.
Example 2
In correspondence with the above method embodiment 1, this embodiment provides a fault detection system based on a combination of long and short term memory and typical correlation, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A method of fault detection based on a combination of long and short term memory and canonical correlation, comprising:
s1: acquiring historical normal operation data of an object to be analyzed as an input set, and acquiring corresponding output of each group of historical normal operation data as an output set;
s2: constructing a first long-short term memory neural network corresponding to the input set and a second long-short term memory neural network corresponding to the output set;
s3: analyzing the mapping relation between the first long-term memory neural network and the second short-term memory neural network by adopting a typical correlation method, and calculating a correlation coefficient according to the mapping relation;
s4: optimizing the first long-short term memory neural network and the second long-short term memory neural network according to the correlation coefficient, and training the first long-short term memory neural network and the second long-short term memory neural network by using a back propagation algorithm until the first long-short term memory neural network and the second long-short term memory neural network which accord with set convergence are obtained;
s5: inputting the first time sequence training set into the first long-short term memory neural network which accords with the set convergence, inputting the second time sequence training set into the second long-short term memory neural network which accords with the set convergence, analyzing residual vectors between the first long-short term memory neural network and the second long-short term memory neural network by using the typical correlation of the output layers of the first long-short term memory neural network and the second long-short term memory neural network, constructing a detection statistic according to the residual vectors, and setting a detection threshold according to the detection statistic;
s6: and acquiring real-time operation data of an object to be analyzed, inputting the real-time operation data into the first long-term and second short-term memory neural networks which accord with the convergence to calculate to obtain detection statistics of the real-time data, comparing the detection statistics of the real-time data with the detection threshold, and if the detection statistics of the real-time data exceeds the detection threshold, determining that a fault occurs.
2. The fault detection method based on the combination of long-short term memory and canonical correlation according to claim 1, wherein the S2 specifically includes:
establishing a first time sequence training set of the input set and a second time sequence training set of the output set according to a preset time span;
and constructing the first long short-term memory neural network according to the first time sequence training set, and constructing the second long short-term memory neural network according to the second time sequence training set.
3. The fault detection method based on a combination of long-short term memory and canonical correlation according to claim 1 or 2, characterized in that the output layers of the first long-short term memory neural network and the second long-short term memory neural network output the same data dimension.
4. The long-short term memory and canonical correlation combination based fault detection method of claim 1, wherein the first long-short term memory neural network or the second long-short term memory neural network comprises an LSTM layer, a maximum pooling layer, a Dropout layer, and a fully connected layer.
5. The fault detection method based on the combination of long-short term memory and canonical correlation according to claim 1, wherein the S3 includes:
s31: outputting the characteristic data of the first long-short term memory network
And the feature data output by the second long-short term memory network
Normalization processing is carried out to obtain normalized characteristic data
And calculates the covariance matrix sigma1、Σ2Sum-sigma12：
Σ1＝E{(io1-μ1)(io1-μ1)T}
Σ2＝E{(io2-μ2)(io2-μ2)T}
Σ12＝E{(io1-μ1)(io2-μ2)T}
In the formula, mu1And mu2Are respectively as
The mean value of (a);
s32: using canonical correlation analysis to find linear transformations such that
And
with the greatest correlation.
6. The fault detection method based on the combination of long-short term memory and canonical correlation according to claim 1, wherein the S4 includes:
s41: constructing a matrix M and carrying out singular value decomposition on the matrix M as follows:
in the formula (I), the compound is shown in the specification,
the matrix is a diagonal matrix, the sum of diagonal elements of the matrix is corr, the corr represents the similarity of two groups of data after linear transformation, R is a left matrix, V is a right singular matrix, and T represents matrix transposition;
s42: the first long-short term memory network and the second long-short term memory network adopt a back propagation algorithm to jointly train network parameters by taking the maximum correlation coefficient as an optimization target.
7. The fault detection method based on the combination of long-short term memory and canonical correlation according to claim 1, wherein the S5 includes:
s51 estimating T by using a kernel density estimation algorithm2The probability distribution of the statistics, using the radial basis function, is as follows:
wherein K (g) represents a radial basis function, and g represents an argument of the radial basis function;
set confidence α, T2Statistical threshold UCLs pass formula
Is obtained in which
Wherein xkWhere k is 1,2, …, M is the sample value of x, h is the bandwidth value of the kernel function, b is an argument, and M represents the number of samples of x.
8. A fault detection system based on a combination of long and short term memory and canonical correlation, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the steps of the method of any of the preceding claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010017077.5A CN111260024B (en) | 2020-01-08 | 2020-01-08 | Fault detection method and system based on combination of long-term memory and typical correlation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010017077.5A CN111260024B (en) | 2020-01-08 | 2020-01-08 | Fault detection method and system based on combination of long-term memory and typical correlation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111260024A true CN111260024A (en) | 2020-06-09 |
| CN111260024B CN111260024B (en) | 2023-11-24 |
Family
ID=70954127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010017077.5A Active CN111260024B (en) | 2020-01-08 | 2020-01-08 | Fault detection method and system based on combination of long-term memory and typical correlation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111260024B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113255219A (en) * | 2021-05-28 | 2021-08-13 | 浙江浙能技术研究院有限公司 | Generator bar vibration optimization method based on correction long-time and short-time memory neural network |
| CN113985173A (en) * | 2021-10-27 | 2022-01-28 | 哈尔滨工业大学 | PFC fault detection method based on statistical characteristic typical correlation analysis |
| CN114578011A (en) * | 2022-03-07 | 2022-06-03 | 上海蓝长科技集团有限公司 | Water quality monitoring method based on multi-sensor multi-source data fusion |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070288409A1 (en) * | 2005-05-31 | 2007-12-13 | Honeywell International, Inc. | Nonlinear neural network fault detection system and method |
| US20170150235A1 (en) * | 2015-11-20 | 2017-05-25 | Microsoft Technology Licensing, Llc | Jointly Modeling Embedding and Translation to Bridge Video and Language |
| US20180032871A1 (en) * | 2016-07-29 | 2018-02-01 | Google Inc. | Systems and Methods to Perform Machine Learning with Feedback Consistency |
| WO2019024083A1 (en) * | 2017-08-04 | 2019-02-07 | Nokia Technologies Oy | Artificial neural network |
| US20190096525A1 (en) * | 2017-09-26 | 2019-03-28 | Owned Outcomes Inc. | Patient data management system |
| CN109753896A (en) * | 2018-12-20 | 2019-05-14 | 西安电子科技大学 | A kind of unsupervised heterologous method for detecting change of remote sensing image based on general character autocoder |
| US20190147300A1 (en) * | 2017-11-16 | 2019-05-16 | International Business Machines Corporation | Anomaly detection in multidimensional time series data |
| US20190312898A1 (en) * | 2018-04-10 | 2019-10-10 | Cisco Technology, Inc. | SPATIO-TEMPORAL ANOMALY DETECTION IN COMPUTER NETWORKS USING GRAPH CONVOLUTIONAL RECURRENT NEURAL NETWORKS (GCRNNs) |
-
2020
- 2020-01-08 CN CN202010017077.5A patent/CN111260024B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070288409A1 (en) * | 2005-05-31 | 2007-12-13 | Honeywell International, Inc. | Nonlinear neural network fault detection system and method |
| US20170150235A1 (en) * | 2015-11-20 | 2017-05-25 | Microsoft Technology Licensing, Llc | Jointly Modeling Embedding and Translation to Bridge Video and Language |
| US20180032871A1 (en) * | 2016-07-29 | 2018-02-01 | Google Inc. | Systems and Methods to Perform Machine Learning with Feedback Consistency |
| WO2019024083A1 (en) * | 2017-08-04 | 2019-02-07 | Nokia Technologies Oy | Artificial neural network |
| US20190096525A1 (en) * | 2017-09-26 | 2019-03-28 | Owned Outcomes Inc. | Patient data management system |
| US20190147300A1 (en) * | 2017-11-16 | 2019-05-16 | International Business Machines Corporation | Anomaly detection in multidimensional time series data |
| US20190312898A1 (en) * | 2018-04-10 | 2019-10-10 | Cisco Technology, Inc. | SPATIO-TEMPORAL ANOMALY DETECTION IN COMPUTER NETWORKS USING GRAPH CONVOLUTIONAL RECURRENT NEURAL NETWORKS (GCRNNs) |
| CN109753896A (en) * | 2018-12-20 | 2019-05-14 | 西安电子科技大学 | A kind of unsupervised heterologous method for detecting change of remote sensing image based on general character autocoder |
Non-Patent Citations (3)
| Title |
|---|
| MALLINAR, NEIL等: "Deep Canonically Correlated LSTMs" * |
| 董洁;孙瑞琪;彭开香;唐鹏: "自动编码器与典型相关分析方法联合驱动的工业过程质量监测" * |
| 赵航;: "典型相关分析和神经网络相结合的早期火灾识别" * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113255219A (en) * | 2021-05-28 | 2021-08-13 | 浙江浙能技术研究院有限公司 | Generator bar vibration optimization method based on correction long-time and short-time memory neural network |
| CN113985173A (en) * | 2021-10-27 | 2022-01-28 | 哈尔滨工业大学 | PFC fault detection method based on statistical characteristic typical correlation analysis |
| CN113985173B (en) * | 2021-10-27 | 2024-03-15 | 哈尔滨工业大学 | PFC fault detection method based on statistical characteristic typical correlation analysis |
| CN114578011A (en) * | 2022-03-07 | 2022-06-03 | 上海蓝长科技集团有限公司 | Water quality monitoring method based on multi-sensor multi-source data fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111260024B (en) | 2023-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110287983B (en) | Single-classifier anomaly detection method based on maximum correlation entropy deep neural network | |
| CN111914873A (en) | Two-stage cloud server unsupervised anomaly prediction method | |
| CN111260024B (en) | Fault detection method and system based on combination of long-term memory and typical correlation | |
| CN109145516B (en) | Analog circuit fault identification method based on improved extreme learning machine | |
| CN114120041B (en) | Small sample classification method based on double-countermeasure variable self-encoder | |
| CN111652290A (en) | Detection method and device for confrontation sample | |
| CN110837872B (en) | Industrial control network intrusion detection method and system | |
| CN111680725A (en) | Gas sensor array multi-fault isolation algorithm based on reconstruction contribution | |
| CN114363195A (en) | Network flow prediction early warning method for time and spectrum residual convolution network | |
| CN115496144A (en) | Power distribution network operation scene determining method and device, computer equipment and storage medium | |
| CN112947649B (en) | Multivariate process monitoring method based on mutual information matrix projection | |
| CN116627116B (en) | Process industry fault positioning method and system and electronic equipment | |
| CN116827685A (en) | Dynamic defense strategy method of micro-service system based on deep reinforcement learning | |
| CN110244690B (en) | Multivariable industrial process fault identification method and system | |
| Cui et al. | Using EBGAN for anomaly intrusion detection | |
| CN116610973A (en) | Sensor fault monitoring and failure information reconstruction method and system | |
| CN111031042A (en) | Network anomaly detection method based on improved D-S evidence theory | |
| CN115659323A (en) | Intrusion detection method based on information entropy theory and convolution neural network | |
| CN111797732B (en) | Video motion identification anti-attack method insensitive to sampling | |
| CN114383648A (en) | Temperature instrument fault diagnosis method and device | |
| CN113642029A (en) | Method and system for measuring correlation between data sample and model decision boundary | |
| CN112884069A (en) | Method for detecting confrontation network sample | |
| CN113255927A (en) | Logistic regression model training method and device, computer equipment and storage medium | |
| CN114167729B (en) | Micro fault detection method and system based on distributed two-section self-adaptive weighting stacking automatic encoder | |
| CN116915477B (en) | Error data injection attack detection method and device based on typical variable analysis, electronic equipment and computer readable 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 |