CN111079857B - Sewage treatment process fault monitoring method based on overcomplete width learning model - Google Patents
Sewage treatment process fault monitoring method based on overcomplete width learning model Download PDFInfo
- Publication number
- CN111079857B CN111079857B CN201911402093.XA CN201911402093A CN111079857B CN 111079857 B CN111079857 B CN 111079857B CN 201911402093 A CN201911402093 A CN 201911402093A CN 111079857 B CN111079857 B CN 111079857B
- Authority
- CN
- China
- Prior art keywords
- data
- fault
- training
- train
- matrix
- 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
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000012544 monitoring process Methods 0.000 title claims abstract description 39
- 230000008569 process Effects 0.000 title claims abstract description 34
- 239000010865 sewage Substances 0.000 title claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 56
- 239000011159 matrix material Substances 0.000 claims abstract description 50
- 238000012545 processing Methods 0.000 claims abstract description 5
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 238000012360 testing method Methods 0.000 claims description 28
- 238000013507 mapping Methods 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000003014 reinforcing effect Effects 0.000 claims description 3
- 238000012880 independent component analysis Methods 0.000 abstract description 8
- 238000010606 normalization Methods 0.000 abstract 1
- 238000004088 simulation Methods 0.000 description 5
- 241000894006 Bacteria Species 0.000 description 4
- 231100000331 toxic Toxicity 0.000 description 4
- 230000002588 toxic effect Effects 0.000 description 4
- 238000004065 wastewater treatment Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- FDWIKIIKBRJSHK-UHFFFAOYSA-N 2-(2-methyl-4-oxochromen-5-yl)acetic acid Chemical compound C1=CC=C2OC(C)=CC(=O)C2=C1CC(O)=O FDWIKIIKBRJSHK-UHFFFAOYSA-N 0.000 description 2
- 238000005273 aeration Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 150000001875 compounds Chemical class 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000010992 reflux Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 238000005842 biochemical reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 208000024335 physical disease Diseases 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000004062 sedimentation Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
-
- C—CHEMISTRY; METALLURGY
- C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F9/00—Multistage treatment of water, waste water or sewage
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
-
- C—CHEMISTRY; METALLURGY
- C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F1/00—Treatment of water, waste water, or sewage
- C02F2001/007—Processes including a sedimentation step
-
- C—CHEMISTRY; METALLURGY
- C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F7/00—Aeration of stretches of water
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Environmental & Geological Engineering (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Chemical & Material Sciences (AREA)
- Organic Chemistry (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Hydrology & Water Resources (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention discloses a sewage treatment process fault monitoring method based on an overcomplete width learning model. The invention comprises two stages of off-line training and on-line monitoring. "offline training" includes: firstly, setting labels for collected normal data and partial fault data, carrying out normalization processing on training data, wherein the labels are in a normal state and the labels are in a fault state, extracting a unmixed matrix of the normal data by an overcomplete independent component analysis method, calculating an independent component matrix of the training data, and establishing an offline training model by using a width learning system. "on-line monitoring" includes: and normalizing and labeling the newly acquired data, wherein the label of 0 is in a normal state, the label of 1 is in a fault state, and online monitoring is performed by using a parameter model of an offline model. The invention does not need excessive network layers, has rapid fault monitoring and high monitoring accuracy.
Description
Technical Field
The invention relates to the technical field of fault diagnosis based on data driving, in particular to a fault diagnosis technology aiming at a sewage treatment process. The invention relates to a specific application of a data driving-based method in the aspect of fault monitoring in a sewage treatment process.
Background
In recent decades, with the continuous development of economy and society, more and more sewage is produced in industrial production and social life, if the untreated sewage directly flows into water sources and lands of people, serious environmental pollution can be caused, if people drink the water carelessly, serious physical diseases can be even caused, and life safety is threatened. The water resources in China are abundant in total, but are unevenly distributed and have less people and average possession, and the research on sewage treatment and recycling technology is widely focused due to huge social benefits and environmental benefits. The sewage treatment process has the characteristics of strong nonlinearity, dynamic property, persistence and the like, and in order to ensure the high efficiency and stability of the sewage treatment process, it is necessary to establish an effective process monitoring scheme to timely detect the abnormal state in the sewage treatment process.
Currently, multivariate statistical techniques have been widely used for process monitoring of wastewater treatment processes, such as principal component analysis (Principal Component Analysis, PCA) and independent component analysis (Independent Component Analysis, ICA) and partial least squares analysis (Partial Least Squares, PLS), among others. Meanwhile, the process data does not always follow Gaussian distribution well, the ICA algorithm can extract non-Gaussian information in the data, but independent components extracted by the ICA algorithm are unstable. With the continued development of deep learning techniques, many people have applied deep learning techniques to process monitoring of wastewater treatment processes, such as deep belief networks (Deep Belief Networks, DBN), deep boltzmann machines (Deep Boltzmann Machines, DBM), convolutional neural networks (Convolutional neural Networks, CNN), and the like. Although most deep neural networks have strong fault monitoring capability for sewage treatment processes, the deep neural networks have complex structures and involve a large number of super-parameters, and the super-parameters of the deep neural networks are complex to set and adjust. On the other hand, in order to obtain higher accuracy in the application, the depth model often needs to be increased in network layer number or parameters, which makes the training process of the depth model often consume a lot of time and a lot of computing resources.
Disclosure of Invention
In order to achieve rapid and efficient process monitoring of a wastewater treatment process, the invention provides a wastewater treatment process fault monitoring method of an overcomplete width learning model integrating an overcomplete independent component analysis (Overcomplete Independent Component Analysis, OICA) and a width learning model (Broad Learning Model, BLM). The OICA and BLM algorithms were proposed by Anastasia Podosinnikova et al at the university of ma and Chen Junlong professor et al at the university of australia, respectively. Firstly, in order to extract non-Gaussian information in data, an OICA algorithm is used for obtaining a unmixed matrix of normal data in an offline training stage, the obtained unmixed matrix is used for calculating a stable independent component matrix containing the non-Gaussian information of training data, the independent component matrix of the training data and a label corresponding to the independent component matrix are used for training a width learning model, and finally, the online data is monitored by using a parameter model obtained through offline training to judge the running state of the current online data.
The invention adopts the following technical scheme and implementation steps:
A. offline training stage:
1) Collecting normal data and fault data in sewage treatment process to form training data set X train . Wherein, the normal data is marked with a '0' label, the fault data is marked with a '1' label, and the label matrix is marked with Y train 。X train Each group of data comprises K sampling moments and M variables. For each sampling instantx i,j The value of the j-th variable representing the i-th sampling instant. The training data is marked as follows:
normal and normal-label represent normal data and labels thereof, and fault-label represent fault data and labels thereof, respectively;
2) The training data are standardized:
first, the mean and standard deviation of normal data are calculated, wherein the mean value at all times of the j-th variable isThe j-th variableStandard deviation of all moments of (2) is +.> Finally, normalized training data are obtained +.>
3) Calculating standardized training data by using OICA algorithmThe unmixed matrix D of normal data, and the independent component matrix S of training data is obtained by using the D train Will S train Data tag Y corresponding to the data tag Y train Labeling:
wherein ,snormal And normal-label represent the independent component matrix of normal data and its label, s, respectively fault And fault-label respectively represent an independent component matrix of fault data and a label thereof;
4) Independent component matrix S obtained for OICA algorithm by using width learning model train And corresponding data tag Y train Training to obtain offline weight W m :
Input matrix A of width learning model n Is formed by mapping node Z n And reinforcing node H j Two parts. Is provided with n mapping nodes and j enhancement nodes, which are formed by independent component matrix S train The mapping is carried out to obtain: z is Z i =φ(S train W ei +β ei ) Phi is a linear activation function, W ei Is a random weight, beta ei Is a random bias. Mapping to obtain a mapping node Z n =(Z 1 ,Z 2 ,…,Z n ). Similarly, enhance node H j =(H 1 ,H 2 ,…,H j ) Is formed by mapping node Z n =(Z 1 ,Z 2 ,…,Z n ) The mapping is carried out to obtain: h j =ζ(Z n W hj +β hj ) ζ is an activation function tanh, W hj Is a random weight, beta hj Is a random bias. A is that n =[Z n |H j ]Is an extended input matrix formed by combining all mapping nodes with enhancement nodes. 100 mapped nodes and 100 enhanced nodes are determined as the number of network nodes when training offline. Thus, the width learning model can be expressed as:
Y=[φ(S train W e1 +β e1 ),…,φ(S train W en +β en )|ζ(Z n W h1 +β h1 ),…,ζ(Z n W hj +β hj )]W n
=[Z 1 ,Z 2 ,…,Z n |H 1 ,…,H j ]W n
=[Z n |H j ]W n
wherein ,W n the connection weight of the width system is obtained by using a ridge regression method, lambda is a constraint parameter, I is a unit matrix, A n =[Z n |H j ]Y is the label output obtained by training;
5) According to the steps, training the width learning model by utilizing different fault type data to obtain over-complete width model subsystems of C different fault types;
B. on-line monitoring:
6) Collecting current sewage treatment process data X test As a test dataset.
7) Average and standard deviation vs. X of normal data obtained off-line test Standardized test data are obtained by standardized processing
8) Calculation ofIndependent component matrix S of (2) test :/>Wherein D is a unmixed matrix obtained in an off-line stage;
9) Matrix S of independent components test Respectively input into the C overcomplete width model subsystems. In the C overcomplete width model subsystems, test data S is obtained by utilizing network parameters obtained through offline training test Mapping to characteristic nodes and enhancement nodes, and then utilizing a network weight matrix W obtained by offline training n For online data S test Reconstructing to obtain a reconstructed data tag
10 Reconstructed data tag obtained by calculating C overcomplete width model subsystemsIf->If the value of (1) is 1, the fault is considered to occur, and an alarm is given; otherwise, it is normal.
Advantageous effects
Compared with the prior art, the method has the advantages that the over-complete width learning model is utilized to carry out offline training on the acquired normal data and fault data to obtain the weight and bias matrix, then the parameters obtained by training are used to monitor the online data, more characteristic information can be extracted compared with the traditional multivariate statistical analysis method, a large number of layers are not required to be stacked like a deep network, complex super-parameters are not required, the complexity of the model is low, the monitoring response is rapid, and the monitoring accuracy is high.
Drawings
FIG. 1 is a block diagram of a breadth-learning model network;
FIG. 2 is a monitoring graph of single variable faults under sunny conditions;
FIG. 3 is a graph of monitoring single variable faults under drought conditions;
FIG. 4 is a graph of monitoring a composite variable fault under sunny conditions;
FIG. 5 is a graph of monitoring composite univariate faults under drought conditions;
Detailed Description
The sewage treatment process not only comprises various physical changes, but also comprises complex chemical changes and biochemical reactions, which lead to complex process flows of the sewage treatment process, and simultaneously, the sewage treatment process anomaly monitoring brings great challenges. The invention adopts a 'simulation benchmark model' (Benchmark Simulation Model, BSM 1) developed by International Water Association (IWA) to simulate the actual sewage treatment process in real time. The model consists of five reaction tanks (5999 m 3) and one secondary sedimentation tank (6000 m 3), and three aeration tanks. The aeration tank has 10 layers, the depth is 4 meters, the occupied area is 1500m < 2 >, and the reaction process has internal reflux and external reflux. The average sewage treatment flow rate is 20000m3/d, and the chemical oxygen demand is 300mg/l. The experiment uses single variable faults and compound variable faults for monitoring, the single variable faults are changes of the maximum growth rate uh of the heterotrophic bacteria, the compound variable faults are values of the maximum growth rate uh of the heterotrophic bacteria and toxic impact bh, the values of the maximum growth rate uh of the heterotrophic bacteria and the toxic impact bh are changed at the same time, the values of the maximum growth rate uh of the heterotrophic bacteria and the values of the toxic impact bh are 4 under the normal operating condition uh, and the values of the toxic impact bh are 0.3 under the normal operating condition bh. The simulation platform collects 1344 sampling point data, and each sampling point contains 16 variables. The variables used for monitoring are shown in table 1 and changed to 2 to a fault setting condition.
Table 1 monitoring the variables used
Table 2 fault set-up conditions
The specific statement that the method of the invention is applied to the offline training and online monitoring of the BSM1 simulation platform is as follows:
A. offline training stage:
1) Collecting normal data and fault data in sewage treatment process to form training data set X train . Wherein, the normal data is marked with a '0' label, the fault data is marked with a '1' label, and the label matrix is marked with Y train 。X train Each group of data comprises K sampling moments and M variables. For each sampling instantx i,j The value of the j-th variable representing the i-th sampling instant. The training data is marked as follows:
normal and normal-label represent normal data and labels thereof, and fault-label represent fault data and labels thereof, respectively;
2) The training data are standardized:
first, the mean and standard deviation of normal data are calculated, wherein the mean value at all times of the j-th variable isThe standard deviation of all moments of the jth variable is +.> Finally, normalized training data are obtained +.>
3) Calculating standardized training data by using OICA algorithmThe unmixed matrix D of normal data, and the independent component matrix S of training data is obtained by using the D train Will S train Data tag Y corresponding to the data tag Y train Labeling:
wherein ,snormal And normal-label represent the independent component matrix of normal data and its label, s, respectively fault And fault-label respectively represent an independent component matrix of fault data and a label thereof;
4) Independent component matrix S obtained for OICA algorithm by using width learning model train And corresponding data tag Y train Training to obtain offline weight W m :
Input matrix A of width learning model n Is formed by mapping node Z n And reinforcing node H j Two parts. Is provided withHaving n mapping nodes and j enhancement nodes, which are formed by independent component matrix S train The mapping is carried out to obtain: z is Z i =φ(S train W ei +β ei ) Phi is a linear activation function, W ei Is a random weight, beta ei Is a random bias. Mapping to obtain a mapping node Z n =(Z 1 ,Z 2 ,…,Z n ). Similarly, enhance node H j =(H 1 ,H 2 ,…,H j ) Is formed by mapping node Z n =(Z 1 ,Z 2 ,…,Z n ) The mapping is carried out to obtain: h j =ζ(Z n W hj +β hj ) ζ is an activation function tanh, W hj Is a random weight, beta hj Is a random bias. A is that n =[Z n |H j ]Is an extended input matrix formed by combining all mapping nodes with enhancement nodes. 100 mapped nodes and 100 enhanced nodes are determined as the number of network nodes when training offline. Thus, the width learning model can be expressed as:
Y=[φ(S train W e1 +β e1 ),…,φ(S train W en +β en )|ζ(Z n W h1 +β h1 ),…,ζ(Z n W hj +β hj )]W n
=[Z 1 ,Z 2 ,…,Z n |H 1 ,…,H j ]W n
=[Z n |H j ]W n
wherein ,W n the connection weight of the width system is obtained by using a ridge regression method, lambda is a constraint parameter, I is a unit matrix, A n =[Z n |H j ]Y is the label output obtained by training;
5) According to the steps, training the width learning model by utilizing different fault type data to obtain over-complete width model subsystems of C different fault types;
B. on-line monitoring:
6) Collecting current sewage treatment process data X test As a test dataset.
7) Average and standard deviation vs. X of normal data obtained off-line test Standardized test data are obtained by standardized processing
8) Calculation ofIndependent component matrix S of (2) test :/>Wherein D is a unmixed matrix obtained in an off-line stage;
9) Matrix S of independent components test Respectively input into the C overcomplete width model subsystems. In the C overcomplete width model subsystems, test data S is obtained by utilizing network parameters obtained through offline training test Mapping to characteristic nodes and enhancement nodes, and then utilizing a network weight matrix W obtained by offline training n For online data S test Reconstructing to obtain a reconstructed data tag
10 Reconstructed data tag obtained by calculating C overcomplete width model subsystemsIf->If the value of (1) is 1, the fault is considered to occur, and an alarm is given; otherwise, it is normal.
The steps are specific application of the method in the field of fault monitoring of the BSM1 simulation platform. To verify the effectiveness of the method, each training was performed using one batch of normal data and 3 batches of fault data, and experiments at the online monitoring stage were performed using 4 batches of test data, respectively. The experimental results obtained are shown in fig. 1 to 4. If the value of the ordinate in the graph is 1, the sewage treatment process is considered to be faulty at the moment; if the ordinate value is 0, the sewage treatment process is normally operated. From the monitoring graph, it can be seen that the overcomplete width learning model responds more rapidly to the monitoring of four faults. Meanwhile, as can be seen from table 3, fault monitoring of the overcomplete width learning model is free from false alarm and less in missing alarm.
TABLE 3 monitoring Effect
Claims (3)
1. A sewage treatment process fault monitoring method based on an overcomplete width learning model is characterized by comprising two stages of off-line training and on-line monitoring, and comprises the following specific steps:
A. offline training stage:
1) Collecting normal data and fault data in sewage treatment process to form training data set X train Wherein, the normal data is marked with a '0' label, the fault data is marked with a '1' label, and the label matrix is marked with Y train ;X train Comprises N groups of data, each group of data comprises K sampling moments and M variables, and each sampling momentx i,j The value of the jth variable representing the ith sample time, the training data is marked as follows:
normal and normal-label represent normal data and labels thereof, and fault-label represent fault data and labels thereof, respectively;
2) The training data are standardized:
3) Calculating standardized training data by using OICA algorithmThe unmixed matrix D of normal data, and the independent component matrix S of training data is obtained by using the D train Will S train Data tag Y corresponding to the data tag Y train Labeling:
wherein ,snomal And normal-label represent the independent component matrix of normal data and its label, s, respectively fault And fault-label respectively represent an independent component matrix of fault data and a label thereof;
4) Independent component matrix S obtained for OICA algorithm by using width learning model train And corresponding data tag Y train Training to obtain offline weight W m :
Input matrix A of width learning model n Is formed by mapping node Z n And reinforcing node H j Two parts. Is provided with n mapping nodes and j enhancement nodes, which are formed by independent component matrix S train Is mapped to obtain:Z i =φ(S train W ei +β ei ) Phi is a linear activation function, W ei Is a random weight, beta ei Is a random bias; mapping to obtain a mapping node Z n =(Z 1 ,Z 2 ,...,Z n ). Similarly, enhance node H j =(H 1 ,H 2 ,...,H j ) Is formed by mapping node Z n =(Z 1 ,Z 2 ,...,Z n ) The mapping is carried out to obtain: h j =ζ(Z n W hj +β hj ) ζ is an activation function tanh, W hj Is a random weight, beta hj Is a random bias; a is that n =[Z n |H j ]Is an extended input matrix formed by combining all mapping nodes and enhancement nodes; 100 mapping nodes and 100 enhancement nodes are determined as the number of network nodes during offline training; thus, the width learning model can be expressed as:
Y=[φ(S train W e1 +β e1 ),...,φ(S train W en +β en )|ζ(Z n W h1 +β h1 ),...,ζ(Z n W hj +β hj )]W n
=[Z 1 ,Z 2 ,...,Z n |H 1 ,...,H j ]W n
=[Z n |H j ]W n
wherein ,W n the connection weight of the width system is obtained by using a ridge regression method, lambda is a constraint parameter, I is a unit matrix, A n =[Z n |H j ]Y is the label output obtained by training;
5) According to the steps, training the width learning model by utilizing different fault type data to obtain over-complete width model subsystems of C different fault types;
B. on-line monitoring:
6) Collecting current sewage treatmentProcess data X test As a test dataset;
7) Average and standard deviation vs. X of normal data obtained off-line test Standardized test data are obtained by standardized processing
8) Calculation ofIndependent component matrix S of (2) test :/>Wherein D is a unmixed matrix obtained in an off-line stage;
9) Matrix S of independent components test Respectively inputting the model into C overcomplete width model subsystems; in the C overcomplete width model subsystems, test data S is obtained by utilizing network parameters obtained through offline training test Mapping to characteristic nodes and enhancement nodes, and then utilizing a network weight matrix W obtained by offline training n For online data S test Reconstructing to obtain a reconstructed data tag
2. The method for monitoring the fault of the sewage treatment process based on the overcomplete width learning model as claimed in claim 1, wherein the method comprises the following steps:
the standardized processing concretely comprises the following steps:
3. The method for monitoring the fault of the sewage treatment process based on the overcomplete width learning model as claimed in claim 1, wherein the method comprises the following steps:
the number of network nodes in offline training is preferably 100 mapped nodes and 100 enhanced nodes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911402093.XA CN111079857B (en) | 2019-12-30 | 2019-12-30 | Sewage treatment process fault monitoring method based on overcomplete width learning model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911402093.XA CN111079857B (en) | 2019-12-30 | 2019-12-30 | Sewage treatment process fault monitoring method based on overcomplete width learning model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111079857A CN111079857A (en) | 2020-04-28 |
CN111079857B true CN111079857B (en) | 2023-06-02 |
Family
ID=70320060
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911402093.XA Active CN111079857B (en) | 2019-12-30 | 2019-12-30 | Sewage treatment process fault monitoring method based on overcomplete width learning model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111079857B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112215281A (en) * | 2020-10-12 | 2021-01-12 | 浙江大学 | Fan blade icing fault detection method |
CN113779773B (en) * | 2021-08-20 | 2024-05-28 | 北京工业大学 | Fermentation process monitoring method of overcomplete width learning system based on slow feature analysis |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
CN106907927A (en) * | 2017-04-05 | 2017-06-30 | 东北大学 | The flexible manifold insertion electric melting magnesium furnace fault monitoring method of one seed nucleus |
CN108960422A (en) * | 2018-06-19 | 2018-12-07 | 河南工业大学 | A kind of width learning method based on principal component analysis |
CN109240274A (en) * | 2018-11-05 | 2019-01-18 | 浙江大学 | A kind of industrial process method for diagnosing faults based on higher order dependencies |
CN110243590A (en) * | 2019-06-25 | 2019-09-17 | 中国民航大学 | A kind of Fault Diagnosis Approach For Rotor Systems learnt based on principal component analysis and width |
CN110297475A (en) * | 2019-07-23 | 2019-10-01 | 北京工业大学 | A kind of batch process fault monitoring method based on Fourth-order moment singular value decomposition |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107025338B (en) * | 2017-03-27 | 2020-04-03 | 北京工业大学 | Recursive RBF neural network-based sludge bulking fault identification method |
-
2019
- 2019-12-30 CN CN201911402093.XA patent/CN111079857B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
CN106907927A (en) * | 2017-04-05 | 2017-06-30 | 东北大学 | The flexible manifold insertion electric melting magnesium furnace fault monitoring method of one seed nucleus |
CN108960422A (en) * | 2018-06-19 | 2018-12-07 | 河南工业大学 | A kind of width learning method based on principal component analysis |
CN109240274A (en) * | 2018-11-05 | 2019-01-18 | 浙江大学 | A kind of industrial process method for diagnosing faults based on higher order dependencies |
CN110243590A (en) * | 2019-06-25 | 2019-09-17 | 中国民航大学 | A kind of Fault Diagnosis Approach For Rotor Systems learnt based on principal component analysis and width |
CN110297475A (en) * | 2019-07-23 | 2019-10-01 | 北京工业大学 | A kind of batch process fault monitoring method based on Fourth-order moment singular value decomposition |
Non-Patent Citations (2)
Title |
---|
《独立成分相关分析的自适应故障监测方法》;王培良等;《控制理论与应用》;第1331-1338页 * |
Process Monitoring Based on Independent Component Contribution;lÜ Xiaotiao等;《Journal of Donghua University》;第349-354页 * |
Also Published As
Publication number | Publication date |
---|---|
CN111079857A (en) | 2020-04-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107025338B (en) | Recursive RBF neural network-based sludge bulking fault identification method | |
CN104914723B (en) | Industrial process soft-measuring modeling method based on coorinated training partial least square model | |
CN105740619B (en) | Weighting extreme learning machine sewage disposal on-line fault diagnosis method based on kernel function | |
CN109495296B (en) | Intelligent substation communication network state evaluation method based on clustering and neural network | |
CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
CN110928187B (en) | Sewage treatment process fault monitoring method based on fuzzy width self-adaptive learning model | |
CN103914064B (en) | Based on the commercial run method for diagnosing faults that multi-categorizer and D-S evidence merge | |
CN104537415B (en) | A kind of non-linear process industry failure predication and recognition methodss based on compressed sensing and DROS ELM | |
CN111079857B (en) | Sewage treatment process fault monitoring method based on overcomplete width learning model | |
CN102854296A (en) | Sewage-disposal soft measurement method on basis of integrated neural network | |
CN110647911A (en) | Bearing fault diagnosis method based on principal component analysis and deep belief network | |
CN106950945A (en) | A kind of fault detection method based on dimension changeable type independent component analysis model | |
CN111122811A (en) | Sewage treatment process fault monitoring method of OICA and RNN fusion model | |
Metenidis et al. | A novel genetic programming approach to nonlinear system modelling: application to the DAMADICS benchmark problem | |
CN110032799A (en) | A kind of the angle similarity divided stages and monitoring method of microbiological pharmacy process | |
Deng et al. | Sparse stacked autoencoder network for complex system monitoring with industrial applications | |
CN112214006A (en) | Intermittent process fault detection method and system considering two-dimensional dynamic characteristics | |
CN108198268A (en) | A kind of production equipment data scaling method | |
AU2021100614A4 (en) | A novel regression prediction method for electronic nose based on broad learning system | |
CN110929809B (en) | Sewage key water quality index soft measurement method of characteristic self-enhanced cyclic neural network | |
CN106769748B (en) | Intelligent detection system for water permeability of membrane bioreactor-MBR (Membrane bioreactor) | |
CN111204867B (en) | Membrane bioreactor-MBR membrane pollution intelligent decision-making method | |
Lu et al. | Industrial process data visualization based on a deep enhanced t-distributed stochastic neighbor embedding neural network | |
CN116702580A (en) | Fermentation process fault monitoring method based on attention convolution self-encoder | |
CN109034366B (en) | Application of ELM integrated model based on multiple activation functions in chemical engineering modeling |
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 |