CN110554667A - convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis - Google Patents
convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis Download PDFInfo
- Publication number
- CN110554667A CN110554667A CN201910747241.5A CN201910747241A CN110554667A CN 110554667 A CN110554667 A CN 110554667A CN 201910747241 A CN201910747241 A CN 201910747241A CN 110554667 A CN110554667 A CN 110554667A
- Authority
- CN
- China
- Prior art keywords
- data
- layer
- industrial process
- fault diagnosis
- convolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 26
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 25
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 19
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims description 16
- 238000011176 pooling Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010923 batch production Methods 0.000 claims description 2
- 238000009776 industrial production Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 abstract description 7
- 238000013135 deep learning Methods 0.000 abstract 1
- 229930182555 Penicillin Natural products 0.000 description 6
- JGSARLDLIJGVTE-MBNYWOFBSA-N Penicillin G Chemical compound N([C@H]1[C@H]2SC([C@@H](N2C1=O)C(O)=O)(C)C)C(=O)CC1=CC=CC=C1 JGSARLDLIJGVTE-MBNYWOFBSA-N 0.000 description 6
- 238000000855 fermentation Methods 0.000 description 6
- 230000004151 fermentation Effects 0.000 description 6
- 229940049954 penicillin Drugs 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 238000004088 simulation Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 150000001875 compounds Chemical class 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012567 pattern recognition method Methods 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000010364 biochemical engineering Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41885—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by modeling, simulation of the manufacturing system
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32339—Object oriented modeling, design, analysis, implementation, simulation language
-
- 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)
- Manufacturing & Machinery (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an intermittent industrial process fault diagnosis method based on a Convolutional Neural Network (CNN), which applies a deep learning method to industrial field fault diagnosis. The method specifically comprises the following steps: 1) collecting field data of an intermittent industrial process; 2) carrying out gridding processing on the acquired data according to a time axis, and standardizing the data into a gray level picture between 0 and 255; 3) training a built CNN network model by using a large amount of off-line data under known working conditions to obtain various parameters of the network; 4) inputting online data and carrying out fault diagnosis.
Description
Technical Field
The invention relates to a fault diagnosis method based on a convolutional neural network for an intermittent industrial process, and mainly relates to a strategy for converting an original signal of the intermittent process into a two-dimensional gray picture.
background
the intermittent process is an extremely important production mode in the modern industrial process, and is widely applied to the production of various high-value-added products such as medicines, foods, biochemical engineering, semiconductors and the like. However, in the actual process, a series of problems such as equipment aging and sudden change of the external environment cause a fault. Therefore, fault diagnosis of an intermittent process becomes critical to ensure the safety of the production process and to improve the quality of the product. For the fault diagnosis of the intermittent process, the conventional methods include a contribution diagram and a pattern recognition method, some scholars use a multivariate statistical method to monitor the intermittent process on line and trace fault variables by using the contribution diagram method, but the method adopts normal data to carry out the fault diagnosis, cannot truly reflect fault information, ignores the correlation among the variables and can only diagnose the fault diagnosis of the single-variable fault intermittent process, and the pattern recognition method determines the fault category to which a new data sample belongs on the basis of a training set of known fault types. The support vector machine and Fisher discriminant analysis are widely applied to fault classification in the intermittent process as a linear classification technology. However, they can only obtain good classification effect under a small sample, and meanwhile, the robustness of the model is weak, and the diagnosis accuracy is relatively low.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an intermittent process fault diagnosis method based on a Convolutional Neural Network (CNN), which can automatically extract deep features among variables, fully consider the continuity of the variables in time and provide reference for production decision.
The invention adopts the following technical scheme and implementation steps.
A. Data acquisition and processing
1) Collecting batch data under normal working conditions and different fault working conditions in intermittent industrial production processwhereinmRepresents the number of the process variables,nrepresenting the number of samples in a batch process;
2) within each batch, stacking the data to form two-dimensional grid data. Due to the difference of the statistical units of the original variables, normalization is needed, namely, normalization is carried out throughThe grid data signals are converted into a two-dimensional grayscale picture with pixel values between 0 and 255. WhereinAndRespectively representiLine ofjValues before and after the grid data of the columns are normalized;Represents the firstiand (6) row data.
B. Building a Convolutional Neural Network (CNN) model
1) Setting two layers of convolution-pooling networks, randomly initializing weight matrixWAnd biasBand setting a network regularization parameter kenel-regularization =0.00001, to sparsify the matrix. Setting a parameter dropout =0.1 behind each convolution layer and each pooling layer to prevent overfitting;
2) Inputting the preprocessed two-dimensional gray level picture, and extracting picture characteristics by the convolution layer according to the following formula:
，
In the formula (I), the compound is shown in the specification,IRepresenting the input mesh data;Krepresents a convolution kernel, also a grid of data;SRepresenting the data subjected to feature mapping, namely the extracted feature grid data;
3) The sigmoid activation function used in the convolutional layer is:
；
4) Deep layer characteristic process obtained by convolutionThe maximum pooling of the training data is reduced by half;
5) Repeating the steps 2) to 4) to extract a final characteristic diagram;
6) Converting the obtained characteristic diagram into the characteristic diagram with the length ofIs output vector of。
C. Use ofSoftmaxThe classifier performs fault classification
to this output vectorSoftmaxRegression processing, the processing procedure uses the following formula:
，
A probability distribution is obtained. The network outputs the fault category corresponding to the maximum probability value.
D. fault diagnosis
1) Preprocessing actual data to be used as test data, and inputting the test data into a trained network;
2) And comparing with the label data, and outputting the diagnosis result of the model for each type of fault.
method advantages
compared with other methods, the method has the following advantages: (1) the method completely starts from process data, does not need prior knowledge and mechanism model of the process, and has strong applicability; (2) the off-line modeling learning training speed is high, the on-line calculation amount is small, and the real-time performance is strong; (3) and the correlation between variables and the correlation of the variables in time are fully considered, and the characteristics related to the fault are better extracted.
Drawings
FIG. 1 is a block diagram of the process of the present invention.
Fig. 2 shows the conversion of original data into a grayscale image under normal operating conditions.
Fig. 3 shows that the original data is converted into a gray picture under the fault 1 condition.
Fig. 4 shows the conversion of the original data into a grayscale picture under the fault 2 condition.
Fig. 5 shows that the original data is converted into a grayscale picture under the fault 3 condition.
FIG. 6 is a graph of actual failure information of the penicillin fermentation process over time.
FIG. 7 is a diagnostic trouble picture of the penicillin fermentation process.
Detailed Description
The penicillin fermentation process is a typical intermittent industrial process, and the Pensim software is a software for simulating penicillin fermentation, which is a simulation software developed by a research group of the U.S. Illinois academy of technology and can better simulate the penicillin fermentation process.
The initial values of various parameters are input according to given values, the sampling time is 0.01 hour, and 200 sampling points are used as one batch. Three fault variables were introduced, air flow rate, make-up matrix flow rate, and agitator power. The types of faults are classified into steps and ramps. After one batch of simulation is finished, the whole-course sampling data points of all variables of the batch are used as the output quantity of the matrix, the results of the program under the normal working condition and the three fault working conditions are respectively used for 1000 times, and the results of 4000 times are used as the data set for CNN network training, and then the results are randomly operated for 400 times to be used as the data set for CNN testing.
The method applied to the penicillin fermentation process simulation object comprises three steps of data preprocessing, model training and fault diagnosis, and is specifically set out as follows.
A. Data preprocessing stage
Using the formula of 16 x 200 groups of data obtained from each batchThe row-wise normalization is to a gray picture with pixel values between 0 and 255, i.e., the mesh data input to the convolutional neural network.
B. Building a CNN network model
1) Setting two layers of convolution-pooling networks, randomly initializing weight matrixWand biasBAnd setting a network sparsity parameter kenel-regularization =0.00001 to sparsify the matrix. A dropout =0.1 parameter is set behind each convolution layer and each pooling layer to prevent overfitting;
2) inputting the preprocessed two-dimensional gray picture, and passing the convolution layerAnd extracting the depth features of the picture. In the formula (I), the compound is shown in the specification,IRepresenting the input mesh data;KRepresents a convolution kernel, also a grid of data;SRepresenting the data subjected to feature mapping, namely the extracted feature grid data;
3) sigmoid activation function used in convolutional layer is formula ；
4) the deep features obtained after convolution are subjected to 2-by-2 maximum pooling, and half of parameters needing training are reduced;
5) repeating the steps 2) to 4) to extract a final characteristic diagram;
6) Converting the obtained characteristic diagram into a one-dimensional characteristic output vector through a full connection layer；
7) To this feature vectorSoftmaxRegression processinga probability distribution is obtained. The network outputs the fault category corresponding to the maximum probability value;
C. And (5) fault diagnosis.
1) and preprocessing actual data to be used as test data, and inputting the test data into the trained network.
and comparing with the label data, and outputting the diagnosis result of the model for each type of fault. The results of the network diagnosis on the simulation data are shown in table 1, and the data results are the average of 10 random trials.
TABLE 1 diagnosis of simulation data by the network
| Sample classes | number of samples | Correct number of diagnoses | accuracy of diagnosis |
| Is normal | 100 | 97 | 0.97 |
| Failure 1 | 100 | 95 | 0.95 |
| failure 2 | 100 | 95 | 0.95 |
| Failure 3 | 100 | 96 | 0.96 |
Claims (4)
1. An intermittent industrial process fault diagnosis based on a Convolutional Neural Network (CNN), wherein the intermittent industrial process refers to an industrial process of which the operation steps are performed at the same position and at different times, the operation state is unstable, and the parameters are changed along with the time; the process data is characterized by huge data volume, uncertainty, dynamics and time-varying property; the convolutional neural network mainly comprises an input layer, a convolutional layer, a pooling layer, a full-connection layer and a pooling layer; the input layer and the convolution layer are connected in front and back; the convolution layer and the pooling layer are connected in front and back; the full connecting layer and the pooling layer are connected in front and back; the full connection layer is used as the previous layer of the output layer.
2. The Convolutional Neural Network (CNN) -based intermittent industrial process fault diagnosis according to claim 1, wherein raw industrial data collected in an intermittent industrial process is composed into two-dimensional grid data (picture) on time axis; the fault diagnosis of the original industrial process is realized by the convolution operation of the grid data.
3. the Convolutional Neural Network (CNN) -based intermittent industrial process fault diagnosis according to claim 2, characterized by comprising the steps of:
1) collecting batch data under normal working conditions and different fault working conditions in intermittent industrial production process whereinmrepresents the number of the process variables,nRepresenting the number of samples in a batch process;
2) within each batch, stacking the data to form two-dimensional grid data Normalization is required due to the difference in the statistical units of the original variables, i.e. by Converting the grid data signal into a two-dimensional grayscale picture with pixels between 0 and 255, wherein And respectively representiLine ofjValues before and after the grid data of the columns are normalized; represents the firstiRow data;
3) Setting two layers of convolution-pooling networks, randomly initializing weight matrixWAnd biasBAnd setting kenel-regularization =0.00001 to sparsify the matrix, and setting dropout =0.1 parameters behind each convolutional and pooling layer to prevent overfitting;
4) inputting the preprocessed two-dimensional gray picture, and passing the convolution layerthe depth features of the picture are extracted, in the formula,IRepresenting the input mesh data;KRepresents a convolution kernel, also a grid of data;SRepresenting the data subjected to feature mapping, namely the extracted feature grid data;
5) Performing convolution-pooling operation on the grid data obtained in the step 4) for multiple times to obtain a final characteristic diagram;
6) Carrying out full connection operation on the characteristic diagram obtained in the step 5) to obtain a final output vector;
7) performing softmax regression processing on the output vector obtained in the step 6) to obtain final probability distribution;
8) and after preprocessing, inputting the actual data into the trained network, comparing the actual data with the label data, and outputting the diagnosis result of the network model on each type of fault.
4. The Convolutional Neural Network (CNN) -based intermittent industrial process fault diagnosis of claim 3, wherein:
1) the activation function used for the convolutional layer is ；
2) The regression function used by softmax is ；
3) The network contains multiple convolution-pooling layers, each layer of which parameters can be set separately.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910747241.5A CN110554667A (en) | 2019-08-14 | 2019-08-14 | convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910747241.5A CN110554667A (en) | 2019-08-14 | 2019-08-14 | convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110554667A true CN110554667A (en) | 2019-12-10 |
Family
ID=68737516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910747241.5A Pending CN110554667A (en) | 2019-08-14 | 2019-08-14 | convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110554667A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860775A (en) * | 2020-07-03 | 2020-10-30 | 南京航空航天大学 | Ship fault real-time diagnosis method based on CNN and RNN fusion |
| CN113283443A (en) * | 2020-02-20 | 2021-08-20 | 中国石油天然气股份有限公司 | Working condition identification method and device, computer equipment and storage medium |
| CN114184861A (en) * | 2021-11-28 | 2022-03-15 | 辽宁石油化工大学 | Fault diagnosis method for oil-immersed transformer |
-
2019
- 2019-08-14 CN CN201910747241.5A patent/CN110554667A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283443A (en) * | 2020-02-20 | 2021-08-20 | 中国石油天然气股份有限公司 | Working condition identification method and device, computer equipment and storage medium |
| CN111860775A (en) * | 2020-07-03 | 2020-10-30 | 南京航空航天大学 | Ship fault real-time diagnosis method based on CNN and RNN fusion |
| CN111860775B (en) * | 2020-07-03 | 2024-05-03 | 南京航空航天大学 | Ship fault real-time diagnosis method based on CNN and RNN fusion |
| CN114184861A (en) * | 2021-11-28 | 2022-03-15 | 辽宁石油化工大学 | Fault diagnosis method for oil-immersed transformer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109765053B (en) | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index | |
| CN110361176B (en) | Intelligent fault diagnosis method based on multitask feature sharing neural network | |
| CN111709292B (en) | Compressor vibration fault detection method based on recursion diagram and deep convolution network | |
| CN102707256B (en) | Fault diagnosis method based on BP-Ada Boost nerve network for electric energy meter | |
| CN113203566B (en) | Motor bearing fault diagnosis method based on one-dimensional data enhancement and CNN | |
| CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
| CN110554667A (en) | convolutional Neural Network (CNN) based intermittent industrial process fault diagnosis | |
| CN110929765A (en) | Convolution self-coding fault monitoring method based on batch imaging | |
| CN106017876A (en) | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN110297479B (en) | Hydroelectric generating set fault diagnosis method based on convolutional neural network information fusion | |
| CN109740687B (en) | Fermentation process fault monitoring method based on DLAE | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN117290800B (en) | Timing sequence anomaly detection method and system based on hypergraph attention network | |
| CN105607631B (en) | The weak fault model control limit method for building up of batch process and weak fault monitoring method | |
| CN111275108A (en) | Method for performing sample expansion on partial discharge data based on generation countermeasure network | |
| CN111324110A (en) | Fermentation process fault monitoring method based on multiple shrinkage automatic encoders | |
| CN116593157A (en) | Complex working condition gear fault diagnosis method based on matching element learning under small sample | |
| CN111122811A (en) | Sewage treatment process fault monitoring method of OICA and RNN fusion model | |
| CN115358259A (en) | Self-learning-based unsupervised cross-working-condition bearing fault diagnosis method | |
| Xue et al. | A novel framework for motor bearing fault diagnosis based on multi-transformation domain and multi-source data | |
| CN114548295A (en) | Bearing fault classification system and method based on multi-scale domain adaptive network | |
| CN118500981A (en) | Intelligent quality inspection monitoring system and method for paint production process | |
| CN112163474B (en) | Intelligent gearbox diagnosis method based on model fusion | |
| CN114295967A (en) | Analog circuit fault diagnosis method based on migration neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191210 |