CN117499199A - VAE-based information enhanced decoupling network fault diagnosis method and system - Google Patents
VAE-based information enhanced decoupling network fault diagnosis method and system Download PDFInfo
- Publication number
- CN117499199A CN117499199A CN202311108081.2A CN202311108081A CN117499199A CN 117499199 A CN117499199 A CN 117499199A CN 202311108081 A CN202311108081 A CN 202311108081A CN 117499199 A CN117499199 A CN 117499199A
- Authority
- CN
- China
- Prior art keywords
- fault
- vae
- sample
- input
- normal
- 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 98
- 238000003745 diagnosis Methods 0.000 title claims abstract description 47
- 238000002955 isolation Methods 0.000 claims abstract description 45
- 238000001514 detection method Methods 0.000 claims abstract description 36
- 238000013507 mapping Methods 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 12
- 125000004122 cyclic group Chemical group 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 239000000654 additive Substances 0.000 description 6
- 230000000996 additive effect Effects 0.000 description 6
- 238000009776 industrial production Methods 0.000 description 5
- 230000006399 behavior Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 239000002826 coolant Substances 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 206010000117 Abnormal behaviour Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000001311 chemical methods and process Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 235000020281 long black Nutrition 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 239000000376 reactant Substances 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0654—Management of faults, events, alarms or notifications using network fault recovery
- H04L41/0659—Management of faults, events, alarms or notifications using network fault recovery by isolating or reconfiguring faulty entities
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0677—Localisation of faults
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention belongs to the technical field of industrial fault diagnosis, and particularly relates to a VAE-based information enhancement decoupling network fault diagnosis method and system, which utilize statistics to convert residual signals into scalar quantities, calculate a threshold value of normal sample statistics, and realize fault detection by comparing an online sample statistics value with the threshold value; and performing a new pre-mapping operation on the input of the VAE and performing a new post-mapping operation on the output of the VAE, so as to realize decoupling of network input and output, help find the specific position of a system fault signal and realize isolation of the fault signal. The invention calculates the statistics of residual signals and the input fault samples x by fixing the trained IEDN-VAE parameters f Is used for the gradient of (a),updating x by back propagation f Until the sample returns to the normal domain, noted asFor prototypes of fault samples in the normal domain, by comparing x f And (3) withAn estimate of the fault is calculated.
Description
Technical Field
The invention belongs to the technical field of industrial fault diagnosis, and particularly relates to a VAE-based information-enhanced decoupling network fault diagnosis method and system.
Background
Modern industrial systems are increasingly large in scale and increasingly complex in composition, and once the system fails, serious threats are caused to the economic benefit of industrial production and the life safety of operators. In addition, if the minor faults of the system cannot be detected and removed in time, the minor faults gradually accumulate and evolve into major system faults as the system flows spread to other parts. Therefore, accurate, timely and efficient fault diagnosis is particularly important to improve the safety and stability of modern industrial production processes.
Fault diagnosis generally includes three sub-tasks: fault detection, fault isolation, and fault estimation. The fault detection determines whether a fault has occurred in the system. The fault isolation can quickly and effectively locate the fault reason, and greatly reduce the fault recovery time. The fault estimation may further determine the fault type and the size of the fault and reconstruct the fault signal in the observed data. In general, the relationship between fault detection, fault isolation, and fault estimation is layer-by-layer, closely related. Fault detection is the first step in fault diagnosis, and fault isolation and fault estimation are subsequent extended tasks.
In the fault detection task, the residual generator has been widely used. The residual signal generation method may be classified into a model-based method, a shallow learning-based method, and a deep learning-based method. Model-based methods require the establishment of an accurate mathematical or physical model of the system under study in advance, but this requirement is often difficult to meet in today's large and complex industrial processes. Meanwhile, a great deal of expert experience is required to be applied when a fault detection scheme is designed, so that the method has the defect that the method is difficult to overcome in practical application and is greatly bound; shallow learning-based methods extract compressed features by projecting data into a specific low-dimensional latent variable space, the extracted features having a well-defined mathematical physical meaning. However, it is still difficult to process large-scale data and fit complex system behaviors; the residual error generating method based on deep learning has a complex multi-layer nonlinear structure, can fit a sufficiently complex process system model, has the advantages of extracting complex nonlinear dynamic characteristics, freely identifying system parameters and the like, is a typical representative of a variational self-encoder, and is focused in the field of industrial fault diagnosis.
Further fault isolation is required after a system fault is detected. The existing fault isolation method is mainly based on two ideas of model decoupling and contribution evaluation. Model-based fault isolation methods typically construct a decoupling transfer matrix that structurally separates the effects of different types of faults on the residual signal; the contribution evaluation-based method is to realize fault isolation by analyzing the contribution of each input quantity to the generated statistic. Although some fault isolation methods have been proposed, there is a lack of deep network-based fault decoupling structure. Because of the complex multi-layer nonlinear structure of the depth network and the characteristics of opacity and difficult interpretation, the method faces a great challenge in implementing fault isolation tasks, and is easily influenced by smearing effect and cannot analyze the contribution of residual errors to test statistics although the method based on the interpretable artificial intelligence has some success in the fault isolation tasks. Therefore, applying the decoupling concept to the fault isolation method of deep learning can solve the above-mentioned dilemma.
Fault estimation is the ultimate goal of the fault diagnosis task. The variables can not interfere with each other after fault isolation to estimate fault signals, so that correct directions can be provided for the variables, reconstruction noise can be reduced, and the accuracy of fault estimation is improved. The most common fault estimation methods (such as sliding mode observers) based on models still exist at present, and the methods have less research on fault estimators with decoupling capability, so that the precision of fault estimation is difficult to guarantee. Furthermore, deep network based methods are still very deficient for fault estimation of complex nonlinear systems.
Through the above analysis, the problems and defects existing in the prior art are as follows: in the aspect of fault detection, in the current huge and complex industrial process, the model-based fault detection method is generally difficult to establish an accurate mathematical physical model, and cannot meet the actual demands; the fault detection method based on shallow learning has weak capability of processing large-scale data, and complex system behaviors are difficult to fit. In terms of fault isolation, a fault decoupling structure based on a depth network is lacking, and the existing fault isolation method is easily affected by a smearing effect, so that the contribution of residual errors to test statistics cannot be analyzed. In the aspect of fault estimation, less research on fault estimation with decoupling capability leads to difficulty in guaranteeing fault estimation precision.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a VAE-based information enhanced decoupling network fault diagnosis method and system.
The invention is realized in such a way that the information enhancement decoupling network fault diagnosis method based on the VAE comprises the following steps:
s1: inputting the data x in normal state into IEDN-VAE to obtain outputDecoupling reconstruction of input is realized;
s2: constructing a residual error generator r related to the input and output observed values;
s3: t is adopted for residual signals 2 The statistic is converted into scalar signal and the threshold J of normal sample is determined th ；
S4: calculating T of on-line sample residual signal 2 The statistics are tested, the state of the online sample is judged, and the fault detection task is completed;
s5: the residual signal is further used for positioning a cause variable which causes faults in the input sample, and a fault isolation task is completed;
s6: and constructing a fault estimator based on a gradient descent strategy according to the fault isolation result to complete a fault estimation task.
Further, the information enhancement decoupling network in S1 includes:
s101, input data: the input is denoted as x, x= (x) 1 ,…,x m ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein m represents the number of sensors for collecting input data; the input data of the off-line process are all in a normal state, i.e
Wherein,representing the whole observation space>And->Respectively representNormal domain space and fault domain space with dimension m;
diagonalizing the x:
Diag{x}＝[χ 1 ,…,χ m ] T ＝X；
each matrix row vectorAs a new sample, +_s> The kth element of (c) satisfies:
s102, front mapping P pre ：X pre ＝XΓ＝[χ 1 ,…,χ m ] T Γ. Wherein Γ is a cyclic matrix that can guarantee X pre Fair to each variable;
the circulant matrix may be determined by its first row, e.g. the circulant matrix of the first row [1,2,3,4] may be constructed as follows
Selecting a cyclic matrixNamely: x is X pre ＝[χ 1 ,…,χ m ] T Γ pre ；
Wherein 1 is<n<Too small values of m and n affect X pre The inheritance capability of the effective information of the input X, and the generalization capability of the model can be negatively influenced if the value of n is too large;
s103, VAE reconstruction: the VAE network belongs to a typical unsupervised neural network, and comprises an encoder and a decoder, wherein the interior of the VAE network is composed of a full connection layer; any two adjacent layers (l-1, l) can be written as:
x (l) ＝σ (l) (w (l-1,l) x (l-1) +b (l) )；
wherein x is (l-1) And x (l) The values of the first layer-1 and the first layer are respectively represented, w represents a weight coefficient, b represents a bias, and sigma represents an activation function;
X pre as input to the VAE, the resulting output is noted as
Mapping function, θ, representing a VAE network VAE Representing network parameters;
s104, post-mapping P post ：
And Γ ik Respectively indicate->And Γ ith row and kth column elements, concatenating the function mappings defined previously, the entire IDN is described as follows:
wherein,namely, decoupling reconstruction of the input data x;
s105, training IEDN-VAE includes internal loss l in And external loss l out Two parts. l (L) in Representing the reconstruction loss of the internal VAE, l out Representing the decoupling reconstruction loss of the IEDN-VAE:
after sufficient training with normal data samples, l in And l out The following are satisfied:
further, the construction method of the residual error generator in S2 includes:
wherein,representing optimal model parameters after IEDN-VAE training;
the residual generator R serves to quantify the reconstructed bias of the IEDN-VAE, where R is a bias vector:
further, a normal sample threshold J is determined in S3 th The method of (1) comprises:
first, use is made of jojobaTerlin statistics (T) 2 ) Converting the S2 generated residual signal into a scalar:
wherein r is n (t) represents the sampled t normal sample residual signal, N represents the training sample size of the offline process;
normal sample T 2 Can be used as a threshold valueTo describe, determine +.The method of nuclear density estimation is used>Is a value of (2); in the nuclear density estimation method, < >>The probability density at that point is expressed as follows:
wherein k (·) represents a kernel function and ρ represents a pairBandwidth with significant impact, m in ρ T2 ＝1。The cumulative density function at this point is calculated as follows:
description of the aboveDoes not exceed a specific boundary (threshold->) Delta represents the confidence level, typically taking a value of 99.5%.
Further, the method for judging the state of the online sample in S4 includes:
after the offline training is completed, learnThe value will be used as a logical decision condition in the online process to determine if the system is in a normal state, after setting the confidence level +.>Can be obtained by the calculation of the inverse function of the above formula;
in the online process, the state of an input sample of the IEDN-VAE is unknown, and the test statistic T of a residual signal of the sample in the unknown state needs to be calculated first 2 (r); after setting the confidence level, the status of the online sample is determined by the following rules:
further, the method for applying the residual signal to the fault isolation task in S5 further includes:
according to S1, the output of IEDN-VAEAnd input x only i Related, independent of other input variables, the residual generator defined by S2 is thus furtherWhen the steps are applied to fault isolation tasks:
wherein,representing the cause variable causing the fault, < >>R represents f I-th variable of>Representation->Is the i-th diagonal element of (a);I is an m matrix of units.
Further, the method for completing the fault estimation task in S6 includes:
the offline process well trains the fault detection model, and knowledge of a normal state sample is fully learned; the task of the fault estimation is to reconstruct the addition to the normal sample x n Fault signal f, and x f Can be observed, and accurately reconstruct the normal sample before the fault signal is added (recorded as) Is particularly important for estimating the value of f; therefore, a rational reconstruction is designed>Is a loss function of (2):
IEDN-VAE parameters obtained in S1Fixed, minimize +.>According to the chain law, < > on the kth update input>The gradient to the input is expressed as:
wherein:
the (k+1) th input sample is
The input samples are iterated continuously untilStopping, wherein the input sample is in normal state, and the obtained input sample can be regarded as original fault input sample x f Corresponding normal sample->
Finally, the addition to the normal sample x can be calculated n A fault signal f of (a):
at this time:
another object of the present invention is to provide a VAE-based information enhancement decoupling network fault diagnosis system to which the VAE-based information enhancement decoupling network fault diagnosis method is applied, the VAE-based information enhancement decoupling network fault diagnosis system comprising:
and (3) a reconstruction module: for inputting data x in normal state into IEDN-VAE to obtain outputDecoupling reconstruction of input is realized;
the residual error generator building module: a residual error generator r for constructing a residual error related to the input and output observed values;
a threshold value determining module: t is adopted for residual signals 2 The statistic is converted into scalar signal and the threshold J of normal sample is determined th ；
And a fault detection module: t for calculating on-line sample residual signal 2 The statistics are tested, the state of the online sample is judged, and the fault detection task is completed;
fault isolation module: the residual signal is further used for positioning a cause variable causing faults in the input sample, so as to complete a fault isolation task;
a fault estimation module: and the fault estimator is used for constructing a gradient descent strategy based fault estimator to complete the fault estimation task according to the fault isolation result.
It is a further object of the present invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the VAE based information enhanced decoupling network fault diagnosis method.
Another object of the present invention is to provide an information data processing terminal, where the information data processing terminal is configured to implement the VAE-based information enhanced decoupling network fault diagnosis system.
In combination with the technical scheme and the technical problems to be solved, the technical scheme to be protected has the following advantages and positive effects:
firstly, the fault diagnosis method based on the VAE is an intelligent method based on data driving, and an observer is designed to simulate the behavior of a system in a normal state through reconstruction deviation, and once a fault signal occurs to the system, the observer is sensitive to intolerable reconstruction deviation; the decoupling design can help us find a specific position where intolerable deviation behavior occurs in the reconstruction vector, namely, isolation of fault signals is realized; and then, through parameters of a fixed depth network, carrying out iterative updating on the input samples by using a gradient descent algorithm to minimize a reconstruction deviation value, finally finding out a corresponding value of the tested fault sample in a normal domain, and calculating an estimated value of the fault signal. Summarizing, the decoupling network based on the VAE can distinguish the influence of each fault variable while carrying out fault detection so as to realize fault isolation, and then the value of the corresponding normal domain is found by updating the fault input sample so as to realize fault estimation.
Secondly, the VAE-based fault diagnosis method provides a new front mapping mode and a new rear mapping mode for fault isolation tasks, and can help to construct a fault decoupling structure based on a depth network, so that decoupling reconstruction of input data is better realized.
The invention provides a fault diagnosis method based on a VAE, which provides an interpretable fault estimation method aiming at a fault estimation task, calculates the gradient of a reconstruction loss and an input sample by fixing trained depth network parameters, updates the input according to a chain rule by back propagation, pulls sample data in a fault domain back to a normal domain, and finally compares an initial fault sample with a normal domain sample to calculate an estimated value of the fault.
Thirdly, whether the technical scheme of the invention overcomes the technical bias: the fault diagnosis technology realizes the monitoring of the existence of fault signals, the positioning of the root cause and the reconstruction of the size through the detection, isolation and identification of system faults. The deep learning has the capability of simulating complex process characteristics due to the multi-layer nonlinear structure, and can greatly improve the diagnosis effect when being used for fault diagnosis tasks. However, the multi-layer nonlinear structure of the depth network is commonly regarded as a "black box model" which cannot provide corresponding evidence and support for its output or decision process, so that one has doubt attitude on its output results, and considers that the results of the depth network lack interpretability. According to the invention, the system normal domain knowledge learned by the deep network model is fully utilized, the input fault sample is updated based on the gradient descent algorithm, and the fault domain sample is migrated back to the normal domain. The input found by the method can lead the trained depth network to realize the best reconstruction performance, the final input is reasonably considered as a prototype of a fault sample in a normal domain, the prototype of specific knowledge extracted by the depth network in the learning process can be interpreted, and the understanding of the multi-layer nonlinear structure learning representation is enhanced.
Fourth, this VAE-based information enhanced decoupling network (IEDN-VAE) fault diagnosis method involves multiple steps, each of which has its own unique technological improvements.
S1: decoupling reconstruction: by using the IEDN-VAE model, the method can realize decoupled reconstruction of normal state data. This is a significant technological advance, as it enables us to extract valuable information from complex data, thereby better understanding the normal operating state of the system.
S2: residual error generator: by constructing a residual error generator related to the input and output observed values, the method can accurately identify the abnormal behavior of the system. This is an important technical advance as it provides an effective way to monitor the health of the system.
S3: threshold determination: this is a significant technical improvement by converting the residual signal into a scalar signal by taking statistics and determining the threshold for normal samples, as it provides a reliable criterion for the system to effectively identify faults.
S4: and (3) fault detection: and calculating the test statistic of the residual signal of the online sample, judging the state of the online sample, and completing the fault detection task. This is an important technical advance as it allows fault detection to be performed in real-time or near real-time conditions to discover and handle faults as early as possible.
S5: fault isolation: and further using the residual signal to locate a cause variable causing the fault in the input sample, and completing the fault isolation task. This is a significant technical advance because it provides a systematic way to determine the source of the fault and thus to conduct the repair in a targeted manner.
S6: fault estimation: and constructing a fault estimator based on a gradient descent strategy according to the fault isolation result to complete a fault estimation task. This is an important technical advance as it provides an efficient way to estimate the extent of impact of a fault and thus formulate a more efficient maintenance strategy.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for fault diagnosis of a VAE-based information enhanced decoupling network according to an embodiment of the present invention;
fig. 2 is a graph of IEDN versus additive fault detection provided by an embodiment of the present invention;
fig. 3 is a schematic diagram of an exemplary additive fault estimation result provided by an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Aiming at the problems in the prior art, the invention provides a fault diagnosis method and a fault diagnosis system for an information enhancement decoupling network based on a VAE, and the invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the fault diagnosis method for the information enhancement decoupling network based on VAE provided by the embodiment of the present invention includes:
a) Offline process
S1: inputting data x in normal state into IEDN-VAE (information enhanced decoupling network) to obtain outputAnd realizing decoupling reconstruction of input.
S2: a residual generator r is constructed which is related to the input and output observations.
S3: t is adopted for residual signals 2 The statistic is converted into scalar signal and the threshold J of normal sample is determined th 。
b) Online process
S4: calculating T of on-line sample residual signal 2 And testing statistics, judging the state of the online sample, and completing the fault detection task.
S5: and further using the residual signal to locate a cause variable causing the fault in the input sample, and completing the fault isolation task.
S6: and constructing a fault estimator based on a gradient descent strategy according to the fault isolation result to complete a fault estimation task.
The information-enhanced decoupling network of S1 includes:
s101, input data: the input is denoted as x, x= (x) 1 ,…,x m ). Wherein m represents the number of sensors for collecting input data;
it should be noted that the input data of the offline process are all in a normal state, i.e
In the above-mentioned method, the step of,representing the whole observation space>And->Respectively representing a normal domain space and a fault domain space with the dimension of m;
diagonalizing the x:
Diag{x}＝[χ 1 ,…,χ m ] T ＝X
each matrix row vectorAs a new sample, +_s>The kth element of (c) satisfies:
s102, front mapping P pre ：X pre ＝XΓ＝[χ 1 ,…,χ m ] T Γ. Wherein Γ is a cyclic matrix that can guarantee X pre Fair to each variable;
the circulant matrix may be determined by its first row, e.g. the circulant matrix of the first row [1,2,3,4] may be constructed as follows
The invention selects the cyclic matrixNamely: x is X pre ＝[χ 1 ,…,χ m ] T Γ pre ；
Wherein 1 is<n<Too small values of m and n affect X pre The inheritance capability of the effective information of the input X, and the generalization capability of the model can be negatively influenced if the value of n is too large;
s103, VAE reconstruction: the VAE network belongs to a typical unsupervised neural network, and the interior of the VAE network comprises an encoder and a decoder, which are both composed of full connection layers. Can be written between any two adjacent layers (l-1, l)
x (l) ＝σ (l) (w (l-1,l) x (l-1) +b (l) )
Wherein x is (l-1) And x (l) The values of the first layer-1 and the first layer are respectively represented, w represents a weight coefficient, b represents a bias, and sigma represents an activation function;
X pre as input to the VAE, the resulting output is noted as
Mapping function, θ, representing a VAE network VAE Representing network parameters;
s104, post-mapping P post ：
And Γ ik Respectively indicate->And an element of the kth column of the Γ ith row. Concatenating the previously defined function maps, the entire IDN is described as follows:
wherein,namely, decoupling reconstruction of the input data x;
s105, training IEDN-VAE includes internal loss l in And external loss l out Two parts. l (L) in Representing the reconstruction loss of the internal VAE, l out Representing the decoupling reconstruction loss of the IEDN-VAE:
after sufficient training with normal data samples, l in And l out The following are satisfied:
the construction method of the residual error generator in the S2 comprises the following steps:
wherein,representing optimal model parameters after IEDN-VAE training;
the residual generator R serves to quantify the reconstructed bias of the IEDN-VAE, where R is a bias vector:
s3 determining the normal sample threshold J th The method of (1) comprises:
first, the Hotelling statistics (T 2 ) Converting the S2 generated residual signal into a scalar:
wherein r is n (t) represents the sampled t normal sample residual signal, N represents the training sample size of the offline process;
normal sample T 2 Can be used as a threshold valueTo describe, the present invention uses a nuclear density estimation method to determine +.>Is a value of (2); in the nuclear density estimation method, < >>The probability density at that point is expressed as follows:
wherein k (·) represents a kernel function and ρ represents a pairBandwidth with significant impact, m in ρ T2 ＝1。The cumulative density function at this point is calculated as follows:
description of the aboveDoes not exceed a specific boundary (threshold->) Delta represents the confidence level, typically taking a value of 99.5%.
The method for judging the on-line sample state in S4 includes:
after the offline training is completed, learnThe value will be used as a logical decision condition in the online process to determine if the system is in a normal state, after setting the confidence level +.>Can be obtained by the calculation of the inverse function of the above formula;
in the online process, the state of an input sample of the IEDN-VAE is unknown, and the test statistic T of a residual signal of the sample in the unknown state needs to be calculated first 2 (r); after setting the confidence level, the status of the online sample is determined by the following rules:
the method for applying the residual signal to the fault isolation task further includes:
according to S1, the output of IEDN-VAEAnd input x only i In relation to other input variables, the residual generator defined by S2 is thus further applied to the fault isolation task:
wherein,representing the cause variable causing the fault, r i f R represents f I-th variable of>Representation->Is the i-th diagonal element of (c). It should be noted that->I is an m matrix of units.
The method for completing the fault estimation task in S6 includes:
the offline process well trains the fault detection model and is already filledKnowledge of the normal state samples is learned. The task of the fault estimation is to reconstruct the addition to the normal sample x n Fault signal f, and x f Can be observed, and accurately reconstruct the normal sample before the fault signal is added (recorded as) Is particularly important for estimating the value of f; therefore, a rational reconstruction is designed>Is a loss function of (2):
IEDN-VAE parameters obtained in S1Fixed, minimize +.>According to the chain law, < > on the kth update input>The gradient to the input is expressed as: />
Wherein:
the (k+1) th input sample is
The input samples are iterated continuously untilStopping, wherein the input sample is in normal state, and the obtained input sample can be regarded as original fault input sample x f Corresponding normal sample->
Finally, the addition to the normal sample x can be calculated n A fault signal f of (a):
at this time:
in order to prove the inventive and technical value of the technical solution of the present invention, this section is an application example on specific products or related technologies of the claim technical solution.
An application embodiment of the present invention provides a computer device including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of a VAE-based information enhanced decoupling network fault diagnosis method.
The embodiment of the application of the invention provides an information data processing terminal which is used for realizing a VAE-based information enhanced decoupling network fault diagnosis system.
The embodiment of the invention defines the performance evaluation index of fault detection and estimation:
false positive rate (FAR), false negative rate (MDR) are suitable for evaluating Fault Detection (FD) performance of an observer, defined as follows:
wherein x is o Representing observations of unknown status in the online test.
Suppose N is collected online c Class data sets, each data set representing a class of faults. The ith data set has N i And each sample consists of a plurality of normal data and fault data. The total number of samples collected online can then be expressed asFrom this, the average FAR and MDR of class i can be expressed as:
wherein FA, TA, MD, RA respectively represents false alarm, accurate alarm, missing detection and correct detection.
Further, the performance of the fault estimation may be measured by Root Mean Square Error (RMSE) and arme, which may be expressed as:
wherein,and f represents an estimated fault signal and an actual fault signal added to the normal samples, respectively.
The present examples were simulated using a Continuous Stirred Tank Reactor (CSTR). CSTR is a very common device in chemical processes. The process belongs to a complex and highly nonlinear chemical reaction process, and is widely applied to actual industrial production. The materials within the apparatus can be continuously reacted in the CSTR by action of the stirrer in the apparatus, which promotes the equilibrium of temperature and concentration in the reaction system, and the model allows the present invention to simulate the operation of the CSTR by controlling feed temperature and concentration, coolant inlet temperature, and other sensor measurements. The manipulated variables of the CSTR system thus include the feed concentration C i And temperature T i And a coolant inlet temperature T ci The present invention can represent the manipulated variable u and the response variable y as:
u＝[C i T i T ci ] T
wherein the superscript s represents the variable measured by the sensor, and C and T represent the concentration and temperature of the reactant, respectively; t (T) c And Q c Respectively the temperature and flow rate of the coolant.
In the CSTR model, 4 different types of additive faults are introduced, a fourth type adding multiple fault signals. The details are shown in table 1 below, wherein subscript "0" indicates the variable value prior to adding the fault signal:
TABLE 1 introduction of faults in CSTR simulations
During the process of collecting samples, the CSTR was simulated ten times in a normal state, 1201 samples were collected each time, and the CSTR was additionally run 4 times to collect the fault data set mentioned in table 1 for online testing. Faults are introduced after the 200 th sampling interval in the fault set. In this way, 12010 and 4804 samples were collected for training and testing, respectively, and the sample compositions are shown in table 2:
TABLE 2 introduction of faults in CSTR simulations
The CSTR system was run for 20 hours per simulation run, with samples collected every minute, and the system was run 10 times in total. According to the model of the invention, super-parameter selection: the number of iterations nepach=20; batch size nb=16; reject ratio dropout=0.2; learning rate η=0.0001; reconstruction loss factor λ=10; allowable error epsilon=5×10 -3 The method comprises the steps of carrying out a first treatment on the surface of the Confidence 1- δ=99.5%; estimating the expectations of FAR as
Experimental comparison the present invention constructs a residual-based depth-from-encoder (DAE) and IEDN comparison. The parameter settings of the DAE are then the same as the IEDN.
Table 3 lists the mean and standard deviation of AFAR and AMDR over 10 independent replicates, where AFAR should be no higher than 0.5% and AMDR should be as small as possible. Although DAE has a lower AMDR, AFAR exceeds 0.5% and does not reach the expected value. Compared with the DAE model, the proposed IEDN can achieve the best FD performance, and the average value of FD results is shown in table 4 below:
TABLE 3 mean value (+ -standard deviation) of FD results in ten independent replicates
Accordingly, table 4 below shows the failure detection results of IEDN, DAE for four different types of failures, and it can be seen that both models exhibit higher FD performance for additive failures.
TABLE 4 mean value (. + -. Standard deviation) of FD results in ten independent replicates
Accordingly, fig. 2 illustrates a detection curve of an IEDN for an additive fault; in fig. 2, the present invention analyzes a multivariate fault (fault 4). Wherein the straight line represents a normal sample, the dotted line from sample 201 represents a failure sample, the long black dotted line represents a learned threshold, the upper range thereof represents the failure state of the sample predicted by the IEDN, and the lower range thereof represents the normal region of the predicted sample. As can be seen from fig. 1, the predicted sample results meet the failure detection index well.
In addition to satisfying the fault detection task, the IEDN may also be used for the fault estimation task, and fig. 3 intuitively shows the estimation result of a typical additive fault. In fig. 3, the present invention analyzes a multivariate fault (fault 4) so that there are 2 actual fault signals. The real fault signal is represented by a straight line, the corresponding predicted fault signal is represented by a broken line, and as apparent from the result of fig. 3, the predicted fault signal can well track the real fault signal, and the predicted result can meet the fault estimation index.
Embodiment one: fault diagnosis of wind generating set
In a wind generating set, the IEDN-VAE fault diagnosis method can be applied to monitor and diagnose the running state of a fan. The specific implementation scheme is as follows:
s1: and collecting data of the normal running state of the fan, including wind speed, generated energy, temperature and the like, and inputting the data into an IEDN-VAE model for decoupling reconstruction.
S2: a residual generator is constructed for generating residuals of the input and output observations.
S3: the residual signal is converted into a scalar signal and the threshold value of the normal sample is determined by statistical analysis.
S4: in the running process of the fan, data are collected in real time, residual signal test statistics of an online sample are calculated, the running state of the fan is judged, and fault detection is achieved.
S5: if a fault is detected, further analyzing residual signals, positioning cause variables causing the fault, such as abnormal wind speed, overhigh temperature and the like, and realizing fault isolation.
S6: and estimating the severity of the fault through a gradient descent strategy according to the fault isolation result, completing fault estimation, and providing a basis for maintenance decision.
Embodiment two: fault diagnosis of industrial production line
In industrial production lines, IEDN-VAE fault diagnosis methods may be used to monitor and diagnose the operating state of equipment. The specific implementation scheme is as follows:
s1: and collecting data of the normal running state of the equipment, such as working speed, temperature, pressure and the like, and inputting the data into an IEDN-VAE model for decoupling reconstruction.
S2: a residual generator is built that is associated with the input and output observations.
S3: the residual signal is converted into a scalar signal and the threshold value of the normal sample is determined by statistical analysis.
S4: in the running process of the equipment, data are collected in real time, residual signal test statistics of an online sample are calculated, the running state of the equipment is judged, and fault detection is achieved.
S5: if a fault is detected, further analyzing residual signals, positioning cause variables causing the fault, such as equipment overheat, pressure abnormality and the like, and realizing fault isolation.
S6: according to the fault isolation result, the influence degree of the fault is estimated through a gradient descent strategy, so that fault estimation is completed, and a basis is provided for subsequent maintenance or replacement decision.
It should be noted that the embodiments of the present invention can be realized in hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or special purpose design hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such as provided on a carrier medium such as a magnetic disk, CD or DVD-ROM, a programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device of the present invention and its modules may be implemented by hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., as well as software executed by various types of processors, or by a combination of the above hardware circuitry and software, such as firmware.
The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.
Claims (10)
1. A VAE-based information-enhanced decoupling network fault diagnosis method, comprising:
s1: inputting the data x in normal state into IEDN-VAE to obtain outputDecoupling reconstruction of input is realized;
s2: constructing a residual error generator r related to the input and output observed values;
s3: t is adopted for residual signals 2 The statistic is converted into scalar signal and the threshold J of normal sample is determined th ；
S4: calculating T of on-line sample residual signal 2 The statistics are tested, the state of the online sample is judged, and the fault detection task is completed;
s5: the residual signal is further used for positioning a cause variable which causes faults in the input sample, and a fault isolation task is completed;
s6: and constructing a fault estimator based on a gradient descent strategy according to the fault isolation result to complete a fault estimation task.
2. The VAE based information enhanced decoupling network fault diagnosis method of claim 1, wherein the information enhanced decoupling network in S1 includes:
s101, input data: the input is denoted as x, x= (x) 1 ,…,x m ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein m represents the number of sensors for collecting input data; the input data of the off-line process are all in a normal state, i.e
Wherein,representing the whole observation space>And->Respectively representing a normal domain space and a fault domain space with the dimension of m;
diagonalizing X:
X＝Diag{x}＝[χ 1 ,…,χ m ] T ；
each matrix row vectorAs a new sample, +_s>i＝1,2,…,m，The kth element of (c) satisfies:
s102, front mapping P pre ：X pre ＝XΓ＝[χ 1 ,…,χ m ] T Γ, where Γ is a cyclic matrix that may guarantee X pre Fair to each variable;
the circulant matrix may be determined by its first row, e.g. the circulant matrix of the first row [1,2,3,4] may be constructed as follows
Selecting a cyclic matrixNamely: x is X pre ＝[χ 1 ,…,χ m ] T Γ pre The method comprises the steps of carrying out a first treatment on the surface of the Wherein 1 is<n<Too small values of m and n affect X pre The inheritance capability of the effective information of the input X, and the generalization capability of the model can be negatively influenced if the value of n is too large;
s103, VAE reconstruction: the VAE network belongs to a typical unsupervised neural network, and comprises an encoder and a decoder, wherein the interior of the VAE network is composed of a full connection layer; any two adjacent layers (l-1, l) can be written as:
x (l) ＝σ (l) (w (l-1,l) x (l-1) +b (l) )；
wherein x is (l-1) And x (l) The values of the first layer-1 and the first layer are respectively represented, w represents a weight coefficient, b represents a bias, and sigma represents an activation function;
X pre as input to the VAE, the resulting output is noted as
F VAE (. Cndot.) represents the mapping function of the VAE network, θ VAE Representing network parameters;
s104, post-mapping P post ：
And Γ ik Respectively indicate->And Γ ith row and kth column elements, concatenating the function mappings defined previously, the entire IDN is described as follows:
wherein,namely, decoupling reconstruction of the input data x;
s105, training IEDN-VAE includes internal loss l in And external loss l out Two parts, l in Representing the reconstruction loss of the internal VAE, l out Representing the decoupling reconstruction loss of the IEDN-VAE:
after sufficient training with normal data samples, l in And l out The following are satisfied:
3. the VAE-based information enhanced decoupling network fault diagnosis method of claim 1, wherein the constructing method of the residual generator in S2 includes:
wherein,representing optimal model parameters after IEDN-VAE training;
the residual generator R serves to quantify the reconstructed bias of the IEDN-VAE, where R is a bias vector:
4. the VAE-based information enhanced decoupling network fault diagnosis method of claim 1, wherein a positive determination is made in S3Constant sample threshold J th The method of (1) comprises:
first, the Hotelling statistics (T 2 ) Converting the S2 generated residual signal into a scalar:
wherein r is n (t) represents the sampled t normal sample residual signal, N represents the training sample size of the offline process;
normal sample T 2 Can be used as a threshold valueTo describe, determine +.The method of nuclear density estimation is used>Is a value of (2); in the nuclear density estimation method, < >>The probability density at that point is expressed as follows:
wherein k (·) represents a kernel function and ρ represents a pairThe bandwidth of the signal is of significant impact,ρ +.> The cumulative density function at this point is calculated as follows:
description of the aboveDoes not exceed a specific boundary (threshold->) Delta represents the confidence level, typically taking a value of 99.5%.
5. The VAE-based information enhanced decoupling network fault diagnosis method of claim 1, wherein the method of judging the on-line sample state in S4 includes:
after the offline training is completed, learnThe value will be used as a logical decision condition in the online process to determine if the system is in a normal state, after setting the confidence level +.>Can be obtained by the calculation of the inverse function of the above formula;
in the online process, the state of an input sample of the IEDN-VAE is unknown, and the test statistic of a residual signal of the sample in the unknown state needs to be calculated firstAt the set confidenceAfter the level, the status of the online sample is determined by the following rules:
6. the VAE based information enhanced decoupling network fault diagnosis method of claim 1, wherein the method of further applying the residual signal to the fault isolation task in S5 includes:
according to S1, the output of IEDN-VAEAnd input x only i In relation to other input variables, the residual generator defined by S2 is thus further applied to the fault isolation task:
wherein,representing the cause variable causing the fault, < >>R represents f I-th variable of>Representation->Is the i-th diagonal element of (a);I is an m matrix of units.
7. The VAE-based information enhanced decoupling network fault diagnosis method of claim 1, wherein the method of completing the fault estimation task in S6 includes:
the offline process well trains the fault detection model, and knowledge of a normal state sample is fully learned; the task of the fault estimation is to reconstruct the addition to the normal sample x n Fault signal f, and x f Can be observed, and accurately reconstruct the normal sample before the fault signal is added (recorded as) Is particularly important for estimating the value of f; therefore, a rational reconstruction is designed>Is a loss function of (2):
IEDN-VAE parameters obtained in S1Fixed, minimize +.>According to the chain law, < > on the kth update input>The gradient to the input is expressed as:
wherein:
the (k+1) th input sample is
The input samples are iterated continuously untilStopping, wherein the input sample is in normal state, and the obtained input sample can be regarded as original fault input sample x f Corresponding normal sample->
Finally, the addition to the normal sample x can be calculated n A fault signal f of (a):
at this time:
8. a VAE-based information enhanced decoupled network fault diagnosis system applying the VAE-based information enhanced decoupled network fault diagnosis method according to any one of claims 1 to 7, characterized in that the VAE-based information enhanced decoupled network fault diagnosis system comprises:
and (3) a reconstruction module: for inputting data x in normal state into IEDN-VAE to obtain outputDecoupling reconstruction of input is realized;
the residual error generator building module: a residual error generator r for constructing a residual error related to the input and output observed values;
a threshold value determining module: t is adopted for residual signals 2 The statistic is converted into scalar signal and the threshold J of normal sample is determined th ；
And a fault detection module: t for calculating on-line sample residual signal 2 The statistics are tested, the state of the online sample is judged, and the fault detection task is completed;
fault isolation module: the residual signal is further used for positioning a cause variable causing faults in the input sample, so as to complete a fault isolation task;
a fault estimation module: and the fault estimator is used for constructing a gradient descent strategy based fault estimator to complete the fault estimation task according to the fault isolation result.
9. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the VAE-based information enhanced decoupling network fault diagnosis method of any one of claims 1 to 7.
10. An information data processing terminal for implementing the VAE-based information enhanced decoupling network fault diagnosis system of claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311108081.2A CN117499199B (en) | 2023-08-30 | 2023-08-30 | VAE-based information enhanced decoupling network fault diagnosis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311108081.2A CN117499199B (en) | 2023-08-30 | 2023-08-30 | VAE-based information enhanced decoupling network fault diagnosis method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117499199A true CN117499199A (en) | 2024-02-02 |
| CN117499199B CN117499199B (en) | 2024-06-14 |
Family
ID=89678810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311108081.2A Active CN117499199B (en) | 2023-08-30 | 2023-08-30 | VAE-based information enhanced decoupling network fault diagnosis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117499199B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| CN111474919A (en) * | 2020-04-27 | 2020-07-31 | 西北工业大学 | Aeroengine control system sensor fault diagnosis method based on AANN network group |
| US20200386811A1 (en) * | 2019-06-05 | 2020-12-10 | Wuhan University | Power electronic circuit fault diagnosis method based on extremely randomized trees and stacked sparse auto-encoder algorithm |
| CN116049719A (en) * | 2023-01-09 | 2023-05-02 | 长沙理工大学 | Fault detection and estimation method, system and medium based on migration type generation model |
| CN116108755A (en) * | 2023-03-09 | 2023-05-12 | 合肥工业大学 | Anti-fact confidence data generation method based on fault dictionary |
| CN116431966A (en) * | 2023-03-16 | 2023-07-14 | 浙江大学 | Reactor core temperature anomaly detection method of incremental characteristic decoupling self-encoder |
-
2023
- 2023-08-30 CN CN202311108081.2A patent/CN117499199B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| US20200386811A1 (en) * | 2019-06-05 | 2020-12-10 | Wuhan University | Power electronic circuit fault diagnosis method based on extremely randomized trees and stacked sparse auto-encoder algorithm |
| CN111474919A (en) * | 2020-04-27 | 2020-07-31 | 西北工业大学 | Aeroengine control system sensor fault diagnosis method based on AANN network group |
| CN116049719A (en) * | 2023-01-09 | 2023-05-02 | 长沙理工大学 | Fault detection and estimation method, system and medium based on migration type generation model |
| CN116108755A (en) * | 2023-03-09 | 2023-05-12 | 合肥工业大学 | Anti-fact confidence data generation method based on fault dictionary |
| CN116431966A (en) * | 2023-03-16 | 2023-07-14 | 浙江大学 | Reactor core temperature anomaly detection method of incremental characteristic decoupling self-encoder |
Non-Patent Citations (2)
| Title |
|---|
| 宋华, 张洪钺: "基于模糊奇偶方程的非线性系统传感器故障诊断", 航空学报, no. 01, 25 January 2003 (2003-01-25) * |
| 臧荣春;崔平远;: "基于神经网络的假设检验故障诊断算法研究", 系统仿真学报, no. 16, 20 August 2007 (2007-08-20) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117499199B (en) | 2024-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112202736B (en) | Communication network anomaly classification method based on statistical learning and deep learning | |
| JP6740247B2 (en) | Anomaly detection system, anomaly detection method, anomaly detection program and learned model generation method | |
| CN105700518B (en) | A kind of industrial process method for diagnosing faults | |
| CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
| CN107725283A (en) | A kind of fan trouble detection method based on depth belief network model | |
| CN111780800B (en) | Method and system for monitoring, isolating and reconstructing sensor fault | |
| CN111709447A (en) | Power grid abnormality detection method and device, computer equipment and storage medium | |
| CN113642754A (en) | Complex industrial process fault prediction method based on RF noise reduction self-coding information reconstruction and time convolution network | |
| CN108921230A (en) | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network | |
| WO2021114320A1 (en) | Wastewater treatment process fault monitoring method using oica-rnn fusion model | |
| CN111860839A (en) | Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm | |
| CN111079348B (en) | Method and device for detecting slowly-varying signal | |
| CN111241629B (en) | Intelligent prediction method for performance change trend of hydraulic pump of airplane based on data driving | |
| CN116292367A (en) | Power plant fan system abnormal condition detection method based on one-dimensional convolution | |
| Lee et al. | In-line predictive monitoring framework | |
| Pérez-Pérez et al. | Robust fault diagnosis of wind turbines based on MANFIS and zonotopic observers | |
| CN117575027B (en) | Sensor fault diagnosis method for enhancing space-time causality by using knowledge | |
| CN117056678B (en) | Machine pump equipment operation fault diagnosis method and device based on small sample | |
| Xu et al. | Anomaly detection with gru based bi-autoencoder for industrial multimode process | |
| CN118188342A (en) | Fan-oriented fault early warning and life prediction method | |
| CN117499199B (en) | VAE-based information enhanced decoupling network fault diagnosis method and system | |
| CN116907772A (en) | Self-diagnosis and fault source identification method and system of bridge structure monitoring sensor | |
| CN105068529B (en) | A kind of fault separating method and system compared based on weighted residual component | |
| Zhi-hong et al. | Sensor fault diagnosis based on wavelet analysis and LSTM neural network | |
| Hayder et al. | Applications of artificial neural networks with input and output degradation data for renewable energy systems fault prognosis |
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 |