CN113885371B - Mixed variable process monitoring method based on health state data - Google Patents
Mixed variable process monitoring method based on health state data Download PDFInfo
- Publication number
- CN113885371B CN113885371B CN202111097150.5A CN202111097150A CN113885371B CN 113885371 B CN113885371 B CN 113885371B CN 202111097150 A CN202111097150 A CN 202111097150A CN 113885371 B CN113885371 B CN 113885371B
- Authority
- CN
- China
- Prior art keywords
- variable
- binary
- calculating
- health state
- steps
- 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 49
- 230000036541 health Effects 0.000 title claims abstract description 33
- 238000012544 monitoring process Methods 0.000 title claims abstract description 17
- 230000008569 process Effects 0.000 title claims abstract description 12
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 230000003862 health status Effects 0.000 claims description 13
- 238000005316 response function Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 description 12
- 230000000052 comparative effect Effects 0.000 description 9
- 238000000513 principal component analysis Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012360 testing method 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
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
- G05B19/042—Programme control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
- G05B19/0428—Safety, monitoring
-
- 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
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24024—Safety, surveillance
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a mixed variable process monitoring method based on health state data, which belongs to the field of fault detection, and under the condition that the health state data can be obtained, continuous variable and binary variable information are utilized to detect faults at the same time; the offline modeling stage acquires health state data and trains to obtain a fault detection model, wherein the fault detection model comprises calculation of variable weights, calculation of statistics and control limits and the like; the online detection stage can collect a new sample online in real time, then calculate the statistic of the new sample by using the fault detection model, and compare the statistic with a control limit so as to judge whether a fault occurs currently. Compared with the traditional continuous variable-based monitoring method, the continuous variable-based monitoring method can simultaneously utilize continuous variable information and binary variable information, and has stronger fault detection capability.
Description
Technical Field
The invention belongs to the field of fault diagnosis, and particularly relates to a mixed variable process monitoring method based on health state data.
Background
Process monitoring is a precondition and guarantee for safe and stable operation of industrial systems. With the continuous upsizing, integration and complexity of modern industrial processes, the monitored variables have binary variables in addition to continuous variables. In the case where only health status data is available, the conventional monitoring method is performed based on only continuous variables, and the fault detection capability is low.
Disclosure of Invention
Aiming at the situation that a mixed variable system simultaneously comprises a continuous variable and a binary variable and only health state data is available, the invention provides a mixed variable process monitoring method based on the health state data. The method utilizes continuous variable and binary variable information simultaneously, and has stronger fault detection capability.
The technical scheme of the invention is as follows:
a method of hybrid variable process monitoring based on health status data, comprising the steps of:
step 1: the offline modeling specifically comprises the following steps:
step 1.1: collecting healthState training data setWherein n samples are included, ">For the ith sample and containing d samples, i represents the sampling time; x is x i Co-comprise d c Each continuous feature d b A binary feature; record x j Is the j-th variable;
step 1.2: when x is j When the variable is continuous, the variable is set to follow Gaussian distribution under the health state
Wherein the method comprises the steps ofμ j Sum sigma j The mean value and standard deviation of the jth variable in the health state are respectively shown;
step 1.3: when x is j In the case of binary variables, it is set to follow Bernoulli distribution in health state
Wherein eta j A response function of the j variable under the health state;
step 1.4: determining the prior normal probability P (N), taking P (N) as the confidence levelI.e. < ->Wherein δ is a significant level;
step 1.5: for continuous variableConstructing a binary variable in a healthy state according to
Then all the calculation of the weight values are carried out by replacing the original continuous variable with the constructed binary variable;
step 1.6: calculating probability in health status
Wherein the method comprises the steps ofAs an indication function, when x is calculated j Probability of =1->Otherwise->
Step 1.7: calculating joint probabilities in health status
Wherein the method comprises the steps of
Step 1.8: calculating condition mutual information under health state
Step 1.9: if x j And x j′ When the variable is a continuous variable, the correction is performed according to the following formula
Wherein the method comprises the steps ofx′ j And x' j′ Respectively x j And x j′ A binary variable constructed according to step 1.5;
step 1.10: calculating the weight of a variable
Step 1.11: when x is j When the binary variable is used, the response function of the binary variable in the health state is calculated
Step 1.12: when x is j As continuous variable, calculate its mean value under health condition
Step 1.13: when x is j As continuous variable, calculate standard deviation under health condition
Step 1.14: calculating sample x i Conditional probability of (2)
P(N|x i )=P(N)P(x i |N) (12)
Wherein the method comprises the steps ofj c J of continuous variable set c Individual variables j b J of the binary variable set b A number of variables;
step 1.15: calculating sample x i Is a function of f (x i )
f(x i )=lnP(N|x i ) (13)
Step 1.16: calculating sample x i Statistics s of i
s i =f 2 (x i ) (14)
Step 1.17: calculating s= [ s ] by nuclear density estimation 1 ,…,s i ,...s n ]Control limit s of (2) lim ;
Step 2: the on-line detection method specifically comprises the following steps:
step 2.1: with new samples x a Upon arrival, calculate its conditional probability
P(N|x a )=P(N)P(x a |N) (15)
Step 2.2: calculating sample x a Is a function of f (x a )
f(x a )=lnP(N|x a ) (16)
Step 2.3: calculating sample x a Statistics s of a
s a =f 2 (x a ) (17)
Step 2.4: will statistic s a And control limit s lim And comparing to judge the occurrence of the fault.
Preferably, the specific judgment criteria of the fault occurrence condition are as follows: if statistics s of the newly sampled data a Not exceeding the control limit s lim And if not, the system is considered to be normal, otherwise, the system is considered to be faulty.
The invention has the beneficial technical effects that:
for a mixed variable system comprising continuous variables and binary variables, process monitoring is performed by utilizing continuous variable and binary variable information, and compared with a traditional continuous variable-based method, the fault detection capability is stronger only when health state data is available.
Drawings
FIG. 1 is a flow chart of a hybrid variable process monitoring method based on health status data of the present invention;
FIG. 2 shows the T of PCA in the comparative experiment of the present invention 2 A statistical chart;
FIG. 3 is a graph of Q statistics of PCA in a comparative experiment of the present invention;
FIG. 4 shows the T of DPCA in a comparative experiment of the present invention 2 A statistical chart;
FIG. 5 is a graph of the Q statistics of DPCA in a comparative experiment of the present invention;
FIG. 6 is I of ICA in comparative experiments of the present invention 2 A statistical chart;
FIG. 7 is a diagram of ICA in a comparative experiment of the present inventionA statistical chart;
FIG. 8 is a graph of Q statistics for ICA in a comparative experiment of the present invention;
FIG. 9 is a statistical plot of MD in a comparative experiment of the present invention;
fig. 10 is a statistical plot of HVM in a comparative experiment of the invention.
Detailed Description
The invention is described in further detail below with reference to the attached drawings and detailed description:
as shown in fig. 1, a method for monitoring a process of a mixed variable based on health status data includes the following steps:
step 1: the offline modeling specifically comprises the following steps:
step 1.1: collecting health training data setsWherein n samples are included, ">For the ith sample and containing d samples, i represents the sampling time; x is x i Co-comprise d c Each continuous feature d b A binary feature; record x j Is the j-th variable;
step 1.2: when x is j As a continuous variable, it is assumed that it follows a gaussian distribution under healthy conditions
Wherein the method comprises the steps ofμ j Sum sigma j The mean value and standard deviation of the jth variable in the health state are respectively shown;
step 1.3: when x is j As a binary variable, it is assumed that it obeys Bernoulli's distribution in a healthy state
Wherein eta j A response function of the j variable under the health state;
step 1.4: determining a priori normal probability P (N), typically taking P (N) as the confidence levelI.e. < ->Wherein δ is a significant level;
step 1.5: for continuous variableConstructing a binary variable in a healthy state according to
All calculations on weights are then performed by replacing the original continuous variable with the constructed binary variable.
Step 1.6: calculating probability in health status
Wherein the method comprises the steps ofAs a function of the readiness (when calculating x j Probability of =1->Otherwise->);
Step 1.7: calculating joint probabilities in health status
Wherein the method comprises the steps of
Step 1.8: calculating condition mutual information under health state
Step 1.9: if x j And x j′ When the variable is a continuous variable, the correction is performed according to the following formula
Wherein the method comprises the steps ofx′ j And x' j′ Respectively x j And x j′ A binary variable constructed according to step 1.5;
step 1.10: calculating the weight of a variable
Step 1.11: when x is j When the binary variable is used, the response function of the binary variable in the health state is calculated
Step 1.12: when x is j As continuous variable, calculate its mean value under health condition
Step 1.13: when x is j As continuous variable, calculate standard deviation under health condition
Step 1.14: calculating sample x i Conditional probability of (2)
P(N|x i )=P(N)P(x i |N) (12)
Wherein the method comprises the steps ofj c J of continuous variable set c Individual variables j b J of the binary variable set b A number of variables;
step 1.15: calculating sample x i Is a function of f (x i )
f(x i )=lnP(N|x i ) (13)
Step 1.16: calculating sample x i Statistics s of i
s i =f 2 (x i ) (14)
Step 1.17: calculating s= [ s ] by nuclear density estimation 1 ,...,s i ,...s n ]Control limit s of (2) lim 。
Step 2: the on-line detection method specifically comprises the following steps:
step 2.1: with new samples x a Upon arrival, calculate its conditional probability
P(N|x a )=P(N)P(x a |N) (15)
Step 2.2: calculating sample x a Is a function of f (x a )
f(x a )=lnP(N|x a ) (16)
Step 2.3: calculating sample x a Statistics s of a
s a =f 2 (x a ) (17)
Step 2.4: will statistic s a And control limit s lim Comparing and judging the occurrence condition of the fault; if statistics s of the newly sampled data a Not exceeding the control limit s lim And if not, the system is considered to be normal, otherwise, the system is considered to be faulty.
In order to fully demonstrate the feasibility and superiority of the invention, simulation studies were performed. The simulation example contains 10 variables, 5 of which are continuous variable x 1 ,x 2 ,x 3 ,x 4 ,x 5 5 binary variables x 6 ,x 7 ,x 8 ,x 9 ,x 10 . The continuous variable obeys a gaussian distribution under normal and fault conditions with parameters as set forth in table 1Shown. The numerical values and adjustment ratios of the binary variables under normal and fault conditions are shown in table 2, for example. Under normal conditions 4000 data of health status data are generated for training the model and 4000 data are then generated for testing, wherein the first 2000 are normal data and the second 2000 are fault data.
TABLE 1 continuous variable distribution
Variable(s) | Normal state | Failure of |
x 1 | N(1.50,0.76 2 ) | N(0.55,0.55 2 ) |
x 2 | N(3.00,0.68 2 ) | N(2.55,1.01 2 ) |
x 3 | N(1.70,0.85 2 ) | N(2.20,1.00 2 ) |
x 4 | N(0.80,1.01 2 ) | N(1.45,0.91 2 ) |
x 5 | N(0.89,0.64 2 ) | N(1.30,0.55 2 ) |
Table 2 binary variable parameters
The invention selects four methods of Principal Component Analysis (PCA), dynamic Principal Component Analysis (DPCA), independent principal component analysis (ICA) and Markov Distance (MD) to carry out comparison experiments with the mixed variable monitoring (HVM) method provided by the invention, and the statistic results of the methods are shown in figures 2-10. Wherein FIG. 2 is the T of PCA 2 Statistical plot, FIG. 3 is a Q statistical plot of PCA, FIG. 4 is T of DPCA 2 Statistical graph, FIG. 5 is a Q statistical graph of DPCA, FIG. 6 is I of ICA 2 Statistical diagram, FIG. 7 is ICA I e 2 Statistical graphs, fig. 8 is a Q statistical graph of ICA, fig. 9 is a statistical graph of MD, and fig. 10 is a statistical graph of HVM of the method of the invention. The cumulative contribution of PCA and DPCA was 0.8, the time superposition of DPCA was 2, and the independent principal component of ICA was 3.
100 experiments were independently repeated for each of the five methods, and the failure False Alarm Rate (FAR) and Failure Detection Rate (FDR) means for each method are shown in table 3. It can be seen that the mixed variable monitoring (HVM) method according to the invention maintains the false alarm rate on a lower horizontal line on the premise of ensuring the highest fault detection rate.
Table 3 Experimental methods and comparison of detection results
It should be understood that the above description is not intended to limit the invention to the particular embodiments disclosed, but to limit the invention to the particular embodiments disclosed, and that the invention is not limited to the particular embodiments disclosed, but is intended to cover modifications, adaptations, additions and alternatives falling within the spirit and scope of the invention.
Claims (2)
1. A method for monitoring a process of a hybrid variable based on health status data, comprising the steps of:
step 1: the offline modeling specifically comprises the following steps:
step 1.1: collecting health training data setsWherein n samples are included, ">For the ith sample and containing d samples, i represents the sampling time; x is x i Co-comprise d c Each continuous feature d b A binary feature; record x j Is the j-th variable;
step 1.2: when x is j When the variable is continuous, the variable is set to follow Gaussian distribution under the health state
Wherein the method comprises the steps ofμ j Sum sigma j The mean value and standard deviation of the jth variable in the health state are respectively shown;
step 1.3: when x is j In the case of binary variables, it is set to follow Bernoulli distribution in health state
Wherein eta j A response function of the j variable under the health state;
step 1.4: determining the prior normal probability P (N), taking P (N) as the confidence levelI.e. < ->Wherein δ is a significant level;
step 1.5: for continuous variableConstructing a binary variable in a healthy state according to
Then all the calculation of the weight values are carried out by replacing the original continuous variable with the constructed binary variable;
step 1.6: calculating probability in health status
Wherein the method comprises the steps ofAs an indication function, when x is calculated j Probability of =1->Otherwise->
Step 1.7: calculating joint probabilities in health status
Wherein the method comprises the steps of
Step 1.8: calculating condition mutual information under health state
Step 1.9: if x j And x j′ When the variable is a continuous variable, the correction is performed according to the following formula
Wherein the method comprises the steps ofx′ j And x' j′ Respectively x j And x j′ A binary variable constructed according to step 1.5;
step 1.10: calculating the weight of a variable
Step 1.11: when x is j When the binary variable is used, the response function of the binary variable in the health state is calculated
Step 1.12: when x is j As continuous variable, calculate its mean value under health condition
Step 1.13: when x is j As continuous variable, calculate standard deviation under health condition
Step 1.14: calculating sample x i Conditional probability of (2)
P(N|x i )=P(N)P(x i |N) (12)
Wherein the method comprises the steps ofj c J of continuous variable set c Individual variables j b J of the binary variable set b A number of variables;
step 1.15: calculating sample x i Is a function of f (x i )
f(x i )=ln P(N|x i ) (13)
Step 1.16: calculating sample x i Statistics s of i
s i =f 2 (x i ) (14)
Step 1.17: calculating s= [ s ] by nuclear density estimation 1 ,...,s i ,...s n ]Control limit s of (2) lim ;
Step 2: the on-line detection method specifically comprises the following steps:
step 2.1: with new samples x a Upon arrival, calculate its conditional probability
P(N|x a )=P(N)P(x a |N) (15)
Step 2.2: calculating sample x a Is a function of f (x a )
f(x a )=ln P(N|x a ) (16)
Step 2.3: calculating sample x a Statistics s of a
s a =f 2 (x a ) (17)
Step 2.4: will statistic s a And control limit s lim And comparing to judge the occurrence of the fault.
2. The method for monitoring a process of a hybrid variable based on health status data according to claim 1, wherein the specific judgment criteria of the occurrence of the fault are: if statistics s of the newly sampled data a Not exceeding the control limit s lim And if not, the system is considered to be normal, otherwise, the system is considered to be faulty.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111097150.5A CN113885371B (en) | 2021-09-18 | 2021-09-18 | Mixed variable process monitoring method based on health state data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111097150.5A CN113885371B (en) | 2021-09-18 | 2021-09-18 | Mixed variable process monitoring method based on health state data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113885371A CN113885371A (en) | 2022-01-04 |
CN113885371B true CN113885371B (en) | 2024-01-19 |
Family
ID=79009882
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111097150.5A Active CN113885371B (en) | 2021-09-18 | 2021-09-18 | Mixed variable process monitoring method based on health state data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113885371B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111414943A (en) * | 2020-03-11 | 2020-07-14 | 山东科技大学 | Anomaly detection method based on mixed hidden naive Bayes model |
CN112131516A (en) * | 2020-09-01 | 2020-12-25 | 山东科技大学 | Anomaly detection method based on feature weight mixed naive Bayes model |
CN112651444A (en) * | 2020-12-29 | 2021-04-13 | 山东科技大学 | Self-learning-based non-stationary process anomaly detection method |
-
2021
- 2021-09-18 CN CN202111097150.5A patent/CN113885371B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111414943A (en) * | 2020-03-11 | 2020-07-14 | 山东科技大学 | Anomaly detection method based on mixed hidden naive Bayes model |
CN112131516A (en) * | 2020-09-01 | 2020-12-25 | 山东科技大学 | Anomaly detection method based on feature weight mixed naive Bayes model |
CN112651444A (en) * | 2020-12-29 | 2021-04-13 | 山东科技大学 | Self-learning-based non-stationary process anomaly detection method |
Also Published As
Publication number | Publication date |
---|---|
CN113885371A (en) | 2022-01-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111507376B (en) | Single-index anomaly detection method based on fusion of multiple non-supervision methods | |
CN109165504B (en) | Power system false data attack identification method based on anti-generation network | |
CN103914064B (en) | Based on the commercial run method for diagnosing faults that multi-categorizer and D-S evidence merge | |
WO2019080367A1 (en) | Method for evaluating health status of mechanical device | |
CN111780800B (en) | Method and system for monitoring, isolating and reconstructing sensor fault | |
CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
CN103901880A (en) | Industrial process fault detection method based on multiple classifiers and D-S evidence fusion | |
CN116739829B (en) | Big data-based power data analysis method, system and medium | |
CN112149750A (en) | Water supply network pipe burst identification data driving method | |
WO2021114320A1 (en) | Wastewater treatment process fault monitoring method using oica-rnn fusion model | |
CN108257365A (en) | A kind of industrial alarm designs method based on global nonspecific evidence dynamic fusion | |
CN108763048B (en) | Hard disk fault early warning and reliability evaluation method based on particle filtering | |
CN116205265A (en) | Power grid fault diagnosis method and device based on deep neural network | |
CN110119579B (en) | OICA-based complex industrial process fault monitoring method | |
CN115526258A (en) | Power system transient stability evaluation method based on Spearman correlation coefficient feature extraction | |
CN115496108A (en) | Fault monitoring method and system based on manifold learning and big data analysis | |
CN111080484A (en) | Method and device for monitoring abnormal data of power distribution network | |
CN109211564B (en) | Self-adaptive threshold detection method for health assessment of ball screw pair | |
CN113885371B (en) | Mixed variable process monitoring method based on health state data | |
CN112651444B (en) | Self-learning-based non-stationary process anomaly detection method | |
CN111459697A (en) | Excitation system fault monitoring method based on deep learning network | |
CN116561692A (en) | Dynamic update real-time measurement data detection method | |
CN116151799A (en) | BP neural network-based distribution line multi-working-condition fault rate rapid assessment method | |
CN115184734A (en) | Power grid line fault detection method and system | |
CN116048036A (en) | Fault detection method based on dynamic non-stationary projection structure |
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 |