CN111340072B - Industrial process fault detection method based on data overall information and neighborhood structure - Google Patents
Industrial process fault detection method based on data overall information and neighborhood structure Download PDFInfo
- Publication number
- CN111340072B CN111340072B CN202010088167.3A CN202010088167A CN111340072B CN 111340072 B CN111340072 B CN 111340072B CN 202010088167 A CN202010088167 A CN 202010088167A CN 111340072 B CN111340072 B CN 111340072B
- Authority
- CN
- China
- Prior art keywords
- statistics
- data
- spe
- industrial process
- points
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Entrepreneurship & Innovation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Computing Systems (AREA)
- Educational Administration (AREA)
- Genetics & Genomics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Physiology (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention provides an industrial process fault detection method based on data overall information and a neighborhood structure, and belongs to the field of process monitoring. Includes pre-processing data of training set under normal working condition, calculating overall information, optimal projection matrix, calculating latent variable space and residual error, and calculating T associated with data set 2 Statistics and SPE (Q) statistics, and estimating T by non-parametric test 2 Upper control limits for statistics and SPE (Q) statistics; and with T of the monitoring dataset 2 The statistics are compared with SPE (Q) statistics to determine if the control limit is exceeded. The invention utilizes the idea of digging global and local characteristics of data to obtain better performance, effectively avoids the problem of information loss caused by data dimension reduction and expansion, and greatly improves the effect of fault detection in process monitoring.
Description
Technical Field
The invention relates to an industrial process fault detection method based on data overall information and a neighborhood structure, and belongs to the field of process monitoring.
Background
The production safety problem is one of the problems of great concern in industrial processes in the new rapid development stage of industrialization, informatization, township and agricultural modernization. Industrial processes refer to industries such as petrochemical, electrical, metallurgical, papermaking, pharmaceutical, food, and the like; they are characterized by continuity; the raw materials and the products are mostly homogeneous phase (solid, liquid or gas) materials. The quality of the product is characterized by purity and various physical and chemical properties. The industries of chemical industry, oil refining, metallurgy, light industry, building materials, pharmacy and the like all belong to the process industry.
The process monitoring aims to prevent faults from occurring by monitoring the running state of the production process and continuously detecting the change and state information of the process, so that the system runs stably under a given performance index. Therefore, a high-efficiency, accurate and real-time fault detection and diagnosis system is established in the production process, faults are accurately and timely detected, diagnosed and eliminated, and the method has important social benefit and great economic significance for guaranteeing the personal safety of operators, ensuring the smooth implementation of chemical processes, improving the product quality, reducing the product cost and protecting the society and natural environment. The fault detection is the basis for developing subsequent researches and is one of research hotspots in the field of process monitoring.
Traditional process monitoring methods are divided into three main categories: quantitative mechanism model based, knowledge based and data driven based methods. With the development of information technology and the coming of big data age, the computer performance and the data storage capacity are greatly improved, and the acquisition and measurement of mass data become more convenient and faster. The data provides more comprehensive, richer and more detailed process states for objective phenomena, can help people to better know morphological characteristics and objective rules of things, and also provides a useful information basis for process modeling, monitoring and control. In recent years, the rapid development of technologies such as data mining, machine learning and the like lays a strong theoretical foundation for data-driven process monitoring.
Principal component analysis is widely applied and achieves good results in industrial production processes as a data-driven process control method. However, the principal component analysis method can only capture the global structure of the process data, only embody the whole information among the data, neglect the neighborhood structure among the data, and have strong correlation among the industrial data, so that a great number of false positives and false negatives occur in the process monitoring, and the effect is greatly reduced. In recent years, in order to solve the problem, a plurality of manifold learning methods are presented, analysis is performed from the view of the neighborhood structure among data, the local relation among data is fully excavated, but the global structure among data is ignored, the existing fault detection method only starts from the view of the whole information of the data or the neighborhood structure relation among data, and has the problems of information loss and the like caused by the fact that the data is reduced from high dimension to low dimension, so that the maximization of the whole variance of the data can be realized, and the method for ensuring that the neighborhood structure among the data is unchanged is deficient in fault detection.
Disclosure of Invention
The invention provides a method for generating data-based overall information and neighbor informationA domain structure industrial process fault detection method, the method comprising two parts of off-line modeling and on-line monitoring, the off-line modeling comprising: collecting industrial process data of the industrial process when the system is normal and the industrial process data of the industrial process when the system is in a fault state as original process information of the system; preprocessing industrial process data by adopting methods such as removing single attribute variables, filling missing data, removing abnormal data and the like, and carrying out standardized processing to obtain an industrial process data set with zero mean and unit variance; deriving T associated with an industrial process data set 2 Statistics and SPE (Q) statistics; acquisition of T 2 Upper control limits for statistics and SPE (Q) statistics;
the online monitoring includes: obtaining an industrial process test data set and its associated T using the same method as described above 2 new Statistics and SPE (Q) new The method comprises the steps of carrying out a first treatment on the surface of the According to T 2 new Statistics and SPE (Q) new And whether the control upper limit is exceeded at the same time, thereby realizing fault detection.
In one embodiment of the invention, the acquiring T of the industrial process dataset 2 The statistics, upper control limit, SPE (Q) statistics, upper control limit method includes:
step one: obtaining an average of an industrial dataset X
And covariance matrix C,>
R D ×n n vector sets with dimension D;
step two: acquiring a neighborhood of a data set X; firstly, calculating Euclidean distances among all sample points, setting proper neighbor numbers according to the number of the samples, and then selecting neighbor points of each sample point through a cluster learning algorithm;
step three: calculating the distance from the sample point in each neighborhood to the neighboring center point; setting W as a sparse symmetric matrix; if there is no point connection between the two points, W ij =0; if the two points are adjacent points, W is calculated by a thermonuclear function;
step four: solving an optimal projection matrix G by using an intelligent learning algorithm;
step five: establishing a monitoring model; comprising the following steps: a latent variable feature space model Y and a residual space model E;
step six: computing T associated with data set X 2 Statistics and SPE (Q) statistics;
step seven: estimating T by adopting a nonparametric test method 2 Upper control limits for statistics and SPE (Q) statistics.
In one embodiment of the present invention, the cluster learning algorithm is a k-nearest neighbor algorithm, including the following steps: and calculating Euclidean distance between all sample points, setting proper neighbor number k according to the number of the samples, selecting neighbor points of each sample point through a k neighbor algorithm, and if one point is in the neighborhood of the other point between the two points, connecting edges between the two points, otherwise, not connecting edges.
In one embodiment of the present invention, the thermonuclear function is:
wherein t is a thermonuclear function parameter, t ε R, and R is a real set.
In one embodiment of the present invention, the optimal projection matrix G is obtained by using a genetic algorithm, and is obtained according to the following formula:
M＝(1-W T )(1-W)
where λ is the smoothing parameter.
In one embodiment of the present invention, the latent variable feature space Y is: y=gx.
In one embodiment of the present invention, the residual spatial model E is: e=x-YG T 。
In one embodiment of the present invention, the AND dataSet X associated T 2 The statistics are:
wherein x is i For the ith vector, y in dataset X i Is the ith vector in latent variable feature space Y.
In one embodiment of the invention, the SPE (Q) statistic associated with data set X is:
wherein I is an identity matrix.
In one embodiment of the invention, the method of non-parametric testing is a method of nuclear density estimation.
The industrial process fault detection method based on the data integral information and the neighborhood structure is used for fault detection of crude oil desalting and dewatering processes.
The beneficial effects are that:
the invention not only realizes the great reservation of the whole information of the data in the fault detection, but also takes account of the maintenance of the neighborhood structure among the data, and can change the maintenance proportion of the whole information and the neighborhood structure by adjusting the smooth parameter, thereby overcoming the problems of the existing fault detection method such as information loss generated by the fact that the data is only reduced from high dimension to low dimension from the global angle or the local relation of the data, and the like, and improving the capability of extracting nonlinear fault characteristics; the method utilizes the idea that the global and local characteristics of the data are mined to obtain better performance, effectively avoids the problem of information loss caused by data dimension reduction and expansion, and greatly improves the effect of fault detection in process monitoring.
Drawings
FIG. 1 is a flow diagram of an implementation of an industrial process fault detection means based on data integrity information and neighborhood structure.
FIG. 2 is a characteristic space distribution diagram of the latent variable in the desalting and dewatering process of crude oil in example 1.
FIG. 3 is a T-shape of crude oil desalting and dewatering Process of example 1 2 Statistics monitor graph.
FIG. 4 is a graph of SPE (Q) statistic monitoring of crude oil desalting and dewatering in example 1.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples.
Example 1
The process of the present invention will be specifically described below in connection with a crude oil desalting and dewatering process. The crude oil desalting and dewatering is to the untreated light petroleum, namely crude oil, by injecting water and demulsifier, mixing to make the oil and water fully contact, the impurities which can be dissolved in the water in the crude oil enter the water, and breaking the emulsion by an electric desalting device under the action of an electric field and temperature to realize oil-water separation and sedimentation, thereby completing the electrochemical desalting and dewatering process.
As shown in fig. 1, an industrial process fault detection method based on data overall information and a neighborhood structure is divided into two parts of offline modeling and online monitoring, wherein the offline modeling comprises:
s101: collecting a process variable sample in a crude oil desalting and dewatering process as system original process information, wherein the process variable comprises the following steps: the total number of the collected process variable samples is 1680, 960 samples in the total number are collected when the system is normal, and 720 samples are collected when the system is in a fault state; preprocessing the training data set by removing single attribute variables, filling missing data, removing abnormal data and the like, carrying out standardization processing to obtain a data set X with zero mean and unit variance,
s102: acquiring global information of the data set X, including a mean value
And covariance matrix C; the calculation is as follows:
s103: acquiring a neighborhood of a data set X; firstly, calculating Euclidean distance between all sample points, setting proper neighbor number according to the number of the samples, and then selecting neighbor points of each sample point through a cluster learning algorithm, wherein the cluster learning algorithm selects a K-NN (K neighbor) algorithm which is more mature in the prior art, and the method comprises the following steps: the euclidean distance between all the sample points is calculated, then the proper number K of neighbors is set according to the number of samples, in this embodiment, k=10, then the neighboring point of each sample point is selected through the K-NN algorithm, if one point is in the neighborhood of the other point, then the two points are connected by edges, otherwise, the two points are not connected by edges.
S104: calculating the distance from the sample point in each neighborhood to the neighboring center point; setting W as a sparse symmetric matrix; if there is no point connection between the two points, W ij =0; if the two points are adjacent points, then W is calculated from the thermonuclear function, i.e
Wherein t is R.
S105: solving an optimal projection matrix G by using an intelligent learning algorithm, such as a genetic algorithm, a neural network and the like; in the embodiment, a genetic algorithm is adopted, and an optimal projection matrix G is obtained according to the following solution;
M＝(1-W T )(1-W) (0.4)
where λ is the smoothing parameter, λ=0.5 is chosen.
S106: establishing a monitoring model; comprising the following steps: a latent variable feature space model Y and a residual space model E;
calculating latent variable feature space according to equation (0.6)
Y＝GX (0.6)
The latent variable feature space diagram is depicted in fig. 2, and it is clear that normal data points and fault data points are separated in the feature space.
Residual E is calculated according to equation (0.7),
E＝X-YG T (0.7)
s107: computing T associated with data set X 2 Statistics and SPE (Q) (square forecast error) statistics;
s108: estimating T by adopting a nonparametric test method 2 The upper control limit of statistic and SPE (Q) statistic is obtained by adopting a nuclear density estimation method and Matlab simulation calculation 2 The upper control limit for the statistic is 1.39 and the upper control limit for the SPE (Q) statistic is 111.64.
The online monitoring includes:
s201: the method comprises the steps of selecting relevant process variables such as demulsifier content, temperature, pressure, cl content, water density and the like in the crude oil desalting and dewatering process as original process information of a system, collecting 600 samples in total, wherein the first 400 samples are collected when the system is normal, the last 200 samples are collected when the system is in a fault state, and performing standardized processing identical to S101 to obtain a test data set X new ；
S202: obtaining the latent variable space Y of the monitoring data according to formulas (0.6) and (0.7) new And residual E new ；
S203: the monitoring data X are obtained by using the formulas (0.8) and (0.9) new Associated T 2 new Statistics and SPE (Q) new Statistics;
s204: according to T 2 The statistics and SPE (Q) statistics exceed their upper control limits at the same time to achieve fault detection. T (T) 2 The statistics and SPE (Q) statistics simultaneously exceed the control limit and the system is in failure. As shown in FIGS. 3 and 4, the dotted line is the upper control limit, and T of the 200 data can be seen 2 The statistics and SPE (Q) statistics are beyond the control limits, and the crude oil desalting and dewatering process is in a fault mode at the stage. The Matlab simulation calculation shows that the false alarm rate is 1.00% and the false alarm rate is 0.00%.
The scope of the present invention is not limited to the above-described embodiments, but is intended to be limited to the appended claims, any modifications, equivalents, improvements and alternatives falling within the spirit and principle of the inventive concept, which can be made by those skilled in the art.
Claims (2)
1. An industrial process fault detection method based on data overall information and a neighborhood structure, the method comprising: two parts, off-line modeling and on-line monitoring, the off-line modeling comprising: collecting industrial process data of the industrial process when the system is normal and the industrial process data of the industrial process when the system is in a fault state as original process information of the system; preprocessing industrial process data by adopting methods such as removing single attribute variables, filling missing data, removing abnormal data and the like, and carrying out standardized processing to obtain an industrial process data set with zero mean and unit variance; deriving T associated with an industrial process data set 2 Statistics and SPE (Q) statistics; acquisition of T 2 Upper control limits for statistics and SPE (Q) statistics;
the online monitoring includes: obtaining an industrial process test data set and its associated T using the same method as described above 2 new Statistics and SPE (Q) new The method comprises the steps of carrying out a first treatment on the surface of the According to T 2 new Statistics and SPE (Q) new Whether or not to simultaneously exceed controlUpper limit, thus realizing fault detection;
t of the acquisition of an industrial process dataset 2 The statistics, upper control limit, SPE (Q) statistics, upper control limit method includes:
step one: obtaining an average of an industrial dataset X
And covariance matrix C,>
R D×n n vector sets with dimension D;
step two: : acquiring a neighborhood of a data set X; firstly, calculating Euclidean distances among all sample points, setting proper neighbor numbers according to the number of the samples, and then selecting neighbor points of each sample point through a cluster learning algorithm;
step three: calculating the distance from the sample point in each neighborhood to the neighboring center point; setting W as a sparse symmetric matrix; if there is no point connection between the two points, W ij =0; if the two points are adjacent points, W is calculated by a thermonuclear function;
step four: solving an optimal projection matrix G by using an intelligent learning algorithm;
step five: establishing a monitoring model; comprising the following steps: a latent variable feature space model Y and a residual space model E;
step six: computing T associated with data set X 2 Statistics and SPE (Q) statistics;
step seven: estimating T by adopting a nonparametric test method 2 Upper control limits for statistics and SPE (Q) statistics;
the cluster learning algorithm is a k-nearest neighbor algorithm, and comprises the following steps: calculating Euclidean distance between all sample points, setting proper neighbor number k according to the number of the samples, selecting neighbor points of each sample point through a k neighbor algorithm, and if one point is in the neighborhood of the other point between the two points, connecting edges between the two points, otherwise, not connecting edges;
the thermonuclear function is:
wherein t is a thermonuclear function parameter, t epsilon R, R is a real number set;
the optimal projection matrix G is obtained by adopting a genetic algorithm and is obtained according to the following formula:
M＝(1-W T )(1-W)
wherein λ is a smoothing parameter;
the latent variable feature space Y is: y=gx;
the residual space model E is: e=x-YG T ；
The T associated with data set X 2 The statistics are:
wherein x is i For the ith vector, y in dataset X i Is the ith vector in the latent variable feature space Y;
the SPE (Q) statistic associated with data set X is:
wherein I is an identity matrix.
2. The industrial process fault detection method based on the data overall information and the neighborhood structure as claimed in claim 1 is used for industrial processes such as chemical industry, oil refining, light industry and the like; process information in industrial processes consists of parameters that characterize the quality of a product by various physical and chemical properties.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010088167.3A CN111340072B (en) | 2020-02-12 | 2020-02-12 | Industrial process fault detection method based on data overall information and neighborhood structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010088167.3A CN111340072B (en) | 2020-02-12 | 2020-02-12 | Industrial process fault detection method based on data overall information and neighborhood structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111340072A CN111340072A (en) | 2020-06-26 |
| CN111340072B true CN111340072B (en) | 2023-06-20 |
Family
ID=71185275
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010088167.3A Active CN111340072B (en) | 2020-02-12 | 2020-02-12 | Industrial process fault detection method based on data overall information and neighborhood structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111340072B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111967489B (en) * | 2020-06-28 | 2023-06-06 | 北京理工大学 | Manufacturing process abnormity monitoring method based on quality data manifold characteristics |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106338977A (en) * | 2016-10-17 | 2017-01-18 | 沈阳化工大学 | Nonlinear process fault detection method based on differential locality preserving projection (DLPP) |
| CN109144039A (en) * | 2018-11-04 | 2019-01-04 | 兰州理工大学 | A kind of batch process fault detection method keeping extreme learning machine based on timing extension and neighborhood |
| CN109522948A (en) * | 2018-11-06 | 2019-03-26 | 山东科技大学 | A kind of fault detection method based on orthogonal locality preserving projections |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101827108B1 (en) * | 2016-05-04 | 2018-02-07 | 두산중공업 주식회사 | Plant fault detection learning method and system |
-
2020
- 2020-02-12 CN CN202010088167.3A patent/CN111340072B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106338977A (en) * | 2016-10-17 | 2017-01-18 | 沈阳化工大学 | Nonlinear process fault detection method based on differential locality preserving projection (DLPP) |
| CN109144039A (en) * | 2018-11-04 | 2019-01-04 | 兰州理工大学 | A kind of batch process fault detection method keeping extreme learning machine based on timing extension and neighborhood |
| CN109522948A (en) * | 2018-11-06 | 2019-03-26 | 山东科技大学 | A kind of fault detection method based on orthogonal locality preserving projections |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111340072A (en) | 2020-06-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112527788B (en) | Method and device for detecting and cleaning abnormal value of transformer monitoring data | |
| CN112036089A (en) | Coal mill fault early warning method based on DPC-MND and multivariate state estimation | |
| CN111080074B (en) | System service security situation element obtaining method based on network multi-feature association | |
| CN111931849B (en) | Hydropower unit operation data trend early warning method | |
| CN110009126B (en) | Online alarm analysis method based on fusion of PLS model and PCA contribution degree | |
| CN107831300A (en) | A kind of transformer insulation oil based on three-dimensional trapezoidal Probabilistic Fuzzy collection deteriorates appraisal procedure | |
| Zhang et al. | A novel fault diagnosis method based on stacked LSTM | |
| CN101738998A (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
| CN111340072B (en) | Industrial process fault detection method based on data overall information and neighborhood structure | |
| CN110879586A (en) | Phase modulator fault diagnosis and state monitoring method and system | |
| Yang et al. | An intelligent singular value diagnostic method for concrete dam deformation monitoring | |
| Tang et al. | Prediction of bearing performance degradation with bottleneck feature based on LSTM network | |
| CN111881594A (en) | Non-stationary signal state monitoring method and system for nuclear power equipment | |
| Zhong et al. | Bayesian inference based reorganized multiple characteristics subspaces fusion strategy for dynamic process monitoring | |
| Liu et al. | Toward robust process monitoring of complex process industries based on denoising sparse auto-encoder | |
| CN105718733B (en) | Fault prediction method based on fuzzy nearness and particle filter | |
| Yao et al. | A new condition-monitoring method based on multi-variable correlation learning network for wind turbine fault detection | |
| CN111934903A (en) | Docker container fault intelligent prediction method based on time sequence evolution genes | |
| Yajie et al. | Complex data study on mechanical fault diagnosing | |
| CN114638039A (en) | Structural health monitoring characteristic data interpretation method based on low-rank matrix recovery | |
| CN114492636A (en) | Transformer winding state signal acquisition system | |
| Zhao et al. | Research of elevator fault diagnosis based on decision tree and rough set | |
| Zhao et al. | Research on Transformer Oil Multi-frequency Ultrasonic Monitoring Technology Based on Convolutional Neural Network | |
| CN113449809A (en) | Cable insulation on-line monitoring method based on KPCA-NSVDD | |
| CN118011223B (en) | Fault diagnosis method and system for electric power energy storage system |
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 |