CN107817784A - A kind of procedure failure testing method based on concurrent offset minimum binary - Google Patents
A kind of procedure failure testing method based on concurrent offset minimum binary Download PDFInfo
- Publication number
- CN107817784A CN107817784A CN201711013418.6A CN201711013418A CN107817784A CN 107817784 A CN107817784 A CN 107817784A CN 201711013418 A CN201711013418 A CN 201711013418A CN 107817784 A CN107817784 A CN 107817784A
- Authority
- CN
- China
- Prior art keywords
- space
- output
- uncorrelated
- data set
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
-
- 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
- G05B2219/24065—Real time diagnostics
Abstract
The present invention provides a kind of procedure failure testing method based on concurrent offset minimum binary, and input variable data and output variable data are handled using the penicillin fermentation process Fault Model based on concurrent offset minimum binary;Calculate Hotelling statistic of the input variable data in output correlation space completely, the Hotelling statistic in the principal component space in the uncorrelated space of output and the SPE statistics in the Hotelling statistic and residual error space of SPE statistics, output variable data in the principal component space for inputting uncorrelated space in residual error space;Calculate combination statistic;Carry out procedure fault judgement.The inventive method extracts the output relevant information of input variable data using orthogonal offset minimum binary, reduce computational load number, output relevant information and output irrelevant information are thoroughly separated using complete orthogonal decomposition, failure is detected exactly, clearly identify whether failure is related to output, so that more preferable operation equipment reduces cost, raising yield and quality.
Description
Technical field
The invention belongs to fault detection technique field, and in particular to a kind of procedure fault inspection based on concurrent offset minimum binary
Survey method.
Background technology
The fermentation process of penicillin is as shown in Figure 1.With the development of modern technologies, penicillin fermentation process increasingly tends to
Maximization, serialization and automation, the structure and composition of equipment are sufficiently complex, and production scale is very huge, each production cycle it
Between contact it is also especially close.On the one hand, this raising for contributing to the decline of cost, penicillin yield and quality;On the other hand,
Device fails and the loss caused by stopping work significantly increases.Therefore, in penicillin fermentation process, as far as possible early
Detect process exception change and reduce wrong report be very necessary.
For penicillin fermentation process, Patent No. ZL201410337732.X patent of invention proposes a kind of penicillin
Fermentation process method for diagnosing faults, although this patent can be diagnosed to be the fault type of penicillin fermentation process.But it is improved
Offset minimum binary when dividing directly related with output principal component space, there is no substantial skill with traditional offset minimum binary
Art difference and progress, it is no completely separated which results in related information and the incoherent information of output is exported, improving
Offset minimum binary detection in:(1) relevant information and output irrelevant information are exported without completely separated;(2) need compared with
High computational load number;(3) the concurrent offset minimum binary detection model obtained, which is relatively difficult to resolve, to be released.So export dependent failure and
Exporting the verification and measurement ratio of uncorrelated failure can reduce or even occur whether failure can not be identified related to output.Patent No.
In ZL201410337732.X patent of invention, failure 1 and failure 2 are the related failure of output, can from accompanying drawing 2 and accompanying drawing 3
To find out:The Hotelling statistic of performance variable principal component space in penicillin fermentation process and process variable residual error space
Hotelling statisticAnd SPE statistics detect failure between 200 to 300 sampled points, this says
The related information of some outputs in uncorrelated space be present in bright export.
The content of the invention
The problem of existing for prior art, the present invention provide a kind of procedure fault detection based on concurrent offset minimum binary
Method.
The technical proposal of the invention is realized in this way:
A kind of procedure failure testing method based on concurrent offset minimum binary, including
The online input variable data and output variable data for obtaining penicillin fermentation process;
Using based on concurrent offset minimum binary penicillin fermentation process Fault Model processing input variable data and
Output variable data:Input variable data are expressed as exporting correlation space data set completely, export the pivot in uncorrelated space
Space data sets and the residual error space data sets sum for exporting uncorrelated space, output variable data are expressed as to input related sky
Between data set, the principal component space data set in the uncorrelated space of input and the uncorrelated space of input residual error space data sets sum;
Calculate Hotelling statistic of the input variable data in output correlation space completely, the master in the uncorrelated space of output
Hotelling statistic in first space, the SPE statistics in the residual error space in the uncorrelated space of output, calculate output variable data
Hotelling statistic in the principal component space for inputting uncorrelated space and the SPE systems in the residual error space in the uncorrelated space of input
Metering;
Calculate Hotelling statistic and input of the output variable data in the principal component space for inputting uncorrelated space not phase
Close the combination statistic of the SPE statistics in the residual error space in space;
Current penicillin fermentation process breakdown judge:
Whether Hotelling statistic of the input variable data for judging to be calculated in output correlation space completely is higher than
Its corresponding control limit:It is that then output dependent failure occurs in current penicillin fermentation process;Otherwise:
Judge Hotelling statistic and input variable of the input variable data in the principal component space for exporting uncorrelated space
Whether data have at least one statistic corresponding higher than its in the SPE statistics in the residual error space for exporting uncorrelated space
Control limit:It is that then current penicillin fermentation process occurs exporting failure uncorrelated but that input is related, otherwise:
Judge Hotelling statistic and output variable of the output variable data in the principal component space for inputting uncorrelated space
Whether the combination statistic of SPE statistic of the data in the residual error space for inputting uncorrelated space is higher than its corresponding control
Limit:It is that then uncertain output dependent failure occurs in current penicillin fermentation process, otherwise continues online acquisition subsequent time
The input variable data and output variable data of penicillin fermentation process.
The input variable data be the concentration of substrate in penicillin fermentation process, the concentration of carbon dioxide, ventilation rate,
Agitator power, pH value and substrate feeding temperature, the output variable data are the dense of the penicillin in penicillin fermentation process
The temperature of degree and fermentation reactor.
Described input variable is handled using the penicillin fermentation process Fault Model based on concurrent offset minimum binary
Before data and output variable data, specification and standard are carried out to the input variable data and output variable data obtained online
Change.
The specification and standardization, specifically make the average of input variable data and output variable data respectively for 0 and side
Difference is 1.
The method for building up of the penicillin fermentation process Fault Model based on concurrent offset minimum binary is as follows:
Gather the offline history normal data set of input variable and the offline history normal data set of output variable;
The offline history normal data set of input variable is divided into output correlation space data using orthogonal offset minimum binary
Collection and output orthogonal intersection space data set;
Correlation space data set and the offline history normal data set difference of output variable will be exported using complete orthogonal decomposition
Carry out complete orthogonal decomposition, obtain the offline history normal data set of input variable complete output correlation space data set, completely
Export the input correlation space data set of the uncorrelated offline history normal data set of space data sets and output variable and input not
Correlation space data set;
Export the output orthogonal intersection space data set of the offline history normal data set of input variable and completely uncorrelated space
Data set merges into the uncorrelated space data sets of output;
Uncorrelated space data sets will be exported and be divided into the principal component space data set for exporting uncorrelated space and output not phase
Close the residual error space data sets in space;
The uncorrelated space data sets of input of the offline history normal data set of output variable are divided into the uncorrelated sky of input
Between principal component space data set and input the residual error space data sets in uncorrelated space;
Input variable data are expressed as exporting correlation space data set completely, export the principal component space number in uncorrelated space
According to the residual error space data sets sum for collecting and exporting uncorrelated space, output variable data are expressed as to input correlation space data
Collect, input the principal component space data set in uncorrelated space and the residual error space data sets sum in the uncorrelated space of input, that is, obtain
Penicillin fermentation process Fault Model based on concurrent offset minimum binary.
The offline history normal data set of input variable is divided into the related sky of output using orthogonal offset minimum binary described
Between data set and output orthogonal intersection space data set before, the offline history normal data set of input variable and output variable are gone through offline
History normal data set carries out specification and standardization.
It is described to export correlation space data set and the offline history normal data set of output variable using complete orthogonal decomposition
Complete orthogonal decomposition is carried out respectively, is specifically:
The offline history normal data set of output variable is mapped to output correlation space and obtains regression coefficient matrix;
Output correlation space data set is mapped into regression coefficient matrix to obtain inputting correlation space data set, output is become
Measure offline history normal data set subtract input correlation space data set obtain inputting uncorrelated space data sets;
The transposed matrix of regression coefficient Matrix Multiplication regression coefficient matrix is obtained into regression matrix, carrying out singular value decomposition to it obtains
The mapping matrix mutually orthogonal to two;
Output correlation space data set is mapped to the mutually orthogonal mapping matrix of two above and obtains output correlation completely
Space data sets and uncorrelated space data sets are exported completely.
Beneficial effect:
The inventive method extracts the output relevant information of input variable data first with orthogonal offset minimum binary, so as to drop
Low computational load number, improves solution to model and releases performance, and it is related that output is then thoroughly separated using complete orthogonal decomposition
Information and output irrelevant information, it can more accurately detect the failure of penicillin fermentation process, more clearly from identification event
Whether barrier is related to output, so as to preferably operate the equipment of penicillin fermentation process to reduce the production of cost, raising penicillin
Amount and quality.Compared with conventional method, the present invention improves the stability and accuracy of fault detect, passes through the simulation experiment result
Illustrate effectiveness of the invention and feasibility.
Brief description of the drawings
Fig. 1 is penicillin fermentation schematic flow sheet;
The fault detect statistics spirogram of the data of failure 1 of the embodiment of Fig. 2 present invention;
(a) it is the Hotelling statistics of the online complete output correlation space for obtaining penicillin fermentation process input variable data
Spirogram;
(b) for the online principal component space for exporting uncorrelated space for obtaining penicillin fermentation process input variable data suddenly
Te Lin counts spirogram;
(c) it is the online residual error space for exporting uncorrelated space for obtaining penicillin fermentation process input variable data
SPE counts spirogram;
(d) for the online principal component space for inputting uncorrelated space for obtaining penicillin fermentation process output variable data suddenly
The SPE in the residual error space in special woods statistic and the uncorrelated space of penicillin fermentation process output variable data input obtained online
The combination statistics spirogram of statistic;
The fault detect statistics spirogram of the data of failure 2 of the embodiment of Fig. 3 present invention;
(a) it is the Hotelling statistics of the online complete output correlation space for obtaining penicillin fermentation process input variable data
Spirogram;
(b) for the online principal component space for exporting uncorrelated space for obtaining penicillin fermentation process input variable data suddenly
Te Lin counts spirogram;
(c) it is the online residual error space for exporting uncorrelated space for obtaining penicillin fermentation process input variable data
SPE counts spirogram;
(d) for the online principal component space for inputting uncorrelated space for obtaining penicillin fermentation process output variable data suddenly
The SPE in the residual error space in special woods statistic and the uncorrelated space of penicillin fermentation process output variable data input obtained online
The combination statistics spirogram of statistic.
Embodiment
The embodiment of the present invention is elaborated below in conjunction with the accompanying drawings.
Because the failure in penicillin fermentation process is the different faults input pair of Protean, different input variable
The influence of output variable is different, but finally can all influence the temperature of fermentation reactor and the yield of penicillin and quality.
In present embodiment there are two kinds of failures in penicillin fermentation process:
Failure 1 is the failure that ventilation rate occurs, and is the slope failure that a kind of slope is 0.5, in 101 sampled points to 200
Occur between individual sampled point.Failure 2 is the failure that agitator power occurs, and is the both phase step fault that a kind of amplitude is+3%, 101
Individual sampled point occurs between 200 sampled points.Using the procedure failure testing method based on concurrent offset minimum binary for logical
The fault type 1 of wind rate and the penicillin fermentation process of 2 two kinds of different faults types of fault type of agitator are detected.
The procedure failure testing method based on concurrent offset minimum binary of present embodiment, including:
Step 1:The online input variable data x for obtaining penicillin fermentation processnewWith output variable data ynew。
In penicillin fermentation process, temperature controller and PH are arranged to closed-loop control, are gradually added into process of production
Glucose, addition be by slope be 0.01 in a manner of straight line increase.Three above condition setting, which can be controlled preferably, fermented
The linear relationship of each variable of journey.Require that each variable meets Gaussian Profile in view of concurrent offset minimum binary, the concentration of penicillin and
The temperature of fermentation reactor is more convenient for detecting, and selects following variable as input and output:
Input variable data are preferably the concentration of the substrate in penicillin fermentation process, carbon dioxide in present embodiment
Concentration, ventilation rate, agitator power, pH value and substrate feeding temperature;
Output variable data are preferably the concentration of the penicillin in penicillin fermentation process and fermented anti-in present embodiment
Answer the temperature of device.
Before step 2 is performed, it is necessary to input variable data to obtaining online and output variable data carry out specification and
Standardization, specifically makes that the average of input variable data and output variable data is 0 and variance is 1 respectively.
Step 2:Input variable is handled using the penicillin fermentation process Fault Model based on concurrent offset minimum binary
Data and output variable data:Input variable data are expressed as to export correlation space data set, the uncorrelated space of output completely
Principal component space data set and export the residual error space data sets sum in uncorrelated space, output variable data are expressed as inputting
The residual error space data sets of correlation space data set, the principal component space data set in the uncorrelated space of input and the uncorrelated space of input
Sum.
The method for building up of penicillin fermentation process Fault Model based on concurrent offset minimum binary is as follows:
Step 2.1:Gather the offline history normal data set of input variableWith the offline history normal number of output variable
According to collectionWherein, subscript N is the number of data sampling, and M, J are variable number.
The each offline history normal data set of input variable and the offline history normal data set of output variable respectively include 200
Sample.
It is normal to the offline history normal data set of input variable and the offline history of output variable before step 2.2 is performed
Data set carries out specification and standardization.
Step 2.2:The offline history normal data set of input variable is divided into output correlation using orthogonal offset minimum binary
Space data sets XcWith output orthogonal intersection space data set Xo;
Incoherent information is exported using orthogonal offset minimum binary as a kind of preprocess method, removal in present embodiment,
The quantity of computational load can be reduced, to reduce the computation complexity of step 2.3.
Step 2.3:Correlation space data set X will be exported using complete orthogonal decompositioncIt is normal with the offline history of output variable
Data set Y carries out complete orthogonal decomposition respectively, obtains the complete output correlation space of the offline history normal data set of input variable
Data setUncorrelated space data sets are exported completelyIt is related to the input of the offline history normal data set of output variable empty
Between data set Yc, the uncorrelated space data sets Y of inputr。
What step 2.3 included comprises the concrete steps that:
Step 2.3.1:The offline history normal data set of output variable is mapped to output correlation space and obtains regression coefficient square
Battle array;
Wherein, Ξ is regression coefficient matrix, and subscript T represents transposition, subscriptRepresent inverse matrix.
After traditional deflected secondary air decomposes, because the principal component space of input variable and residual error space are inclined in the presence of one
Oblique angle, result in output relevant information and output irrelevant information is not completely segregated.Present embodiment is in order to reach complete
Orthogonal Decomposition, regression coefficient matrix is solved using above-mentioned formula.Wherein, uncorrelated space data sets Y is inputtedrIt is with input data
It is completely orthogonal, therefore in above-mentioned formulaShould be 0.
Step 2.3.2:Will output correlation space data set XcRegression coefficient matrix is mapped to obtain inputting correlation space number
According to collection, the offline history normal data set of output variable is subtracted into input correlation space data set and obtains inputting uncorrelated spatial data
Collection.
Step 2.3.3:The transposed matrix of regression coefficient Matrix Multiplication regression coefficient matrix is obtained into regression matrix, it carried out very
Different value decomposes to obtain two mutually orthogonal mapping matrixes.
Wherein, PΞ, cIt is the load using regression coefficient matrix as the space of base vector, PΞ, rIt is using regression coefficient matrix as base
The load of the orthogonal intersection space in the space of vector, Λ is by Ξ ΞTEigenvalue cluster into diagonal matrix,It is wherein one
Individual mapping matrix,It is another mapping matrix.
According to the characteristic of singular value decomposition, it is known thatThat is PΞ, cAnd PΞ, rIt is orthogonal.Therefore with the two to
The space measured as base vector is mutually orthogonal space.
Step 2.3.4:Output correlation space data set is mapped into the mutually orthogonal mapping matrix of two above to have obtained
It is complete to export correlation space data set and export uncorrelated space data sets completely.
Wherein,It is complete output correlation space data set,It is to export uncorrelated space data sets, T completelyΞ, c=
XcPΞ, cIt is the score matrix of complete output correlation space data set, TΞ, r=XcPΞ, rIt is to export uncorrelated space data sets completely
Score matrix.
According to the characteristic of singular value decomposition, it is known thatTherefore
In addition,ThereforeWithIt is orthogonal.It can thus be seen that relative to biography
The offset minimum binary of system, by input variable complete orthogonal decomposition it is output relevant information and output irrelevant information in this method.
Step 2.4:By the output orthogonal intersection space data set of the offline history normal data set of input variable and complete output not
Correlation space data set merges into the uncorrelated space data sets of output.
ConsiderAnd XoIt is in the incoherent information of output, it is not necessary that to its independent analysis, therefore they are closed
And.
Wherein, XrIt is the uncorrelated space data sets of output.
Step 2.5:Uncorrelated space data sets will be exported using pivot analysis be divided into export the pivot in uncorrelated space
Space data sets and the residual error space data sets for exporting uncorrelated space.
Wherein,It is the principal component space data set for exporting uncorrelated space,It is the residual error space for exporting uncorrelated space
Data set, TX, rIt is the score matrix for the principal component space data set for exporting uncorrelated space, PX, rIt is the master for exporting uncorrelated space
The load matrix of first space data sets.
Step 2.6:Using pivot analysis by the uncorrelated spatial data of input of the offline history normal data set of output variable
Collection is divided into the principal component space data set for inputting uncorrelated space and the residual error space data sets for inputting uncorrelated space.
Wherein,It is the principal component space data set for inputting uncorrelated space,It is the residual error space for inputting uncorrelated space
Data set, TY, rIt is the score matrix for the principal component space data set for inputting uncorrelated space, PY, rIt is the master for inputting uncorrelated space
The load matrix of first space data sets.
Step 2.7:Input variable data are expressed as exporting correlation space data set completely, export the master in uncorrelated space
First space data sets and the residual error space data sets sum for exporting uncorrelated space, it is related that output variable data are expressed as input
The residual error space data sets of space data sets, the principal component space data set in the uncorrelated space of input and the uncorrelated space of input it
With obtain the penicillin fermentation process Fault Model based on concurrent offset minimum binary.
Step 3:Calculate Hotelling statistic of the input variable data in output correlation space completely, the uncorrelated sky of output
Between principal component space in Hotelling statistic, the uncorrelated space of output residual error space in SPE statistics, calculate output and become
In the residual error space for measuring Hotelling statistic and the uncorrelated space of input of the data in the principal component space for inputting uncorrelated space
SPE statistics.
Wherein,It is Hotelling statistic of the input variable data in output correlation space completely,It is to export not phase
Close the Hotelling statistic in the principal component space in space, QxIt is the SPE statistics in the residual error space for export uncorrelated space,
The Hotelling statistic for being output variable data in the principal component space for inputting uncorrelated space, QyIt is output variable data defeated
Enter the SPE statistics in the residual error space in uncorrelated space, xNew, cIt is the output correlation space data set of input variable data,
xNew, oIt is the output orthogonal intersection space data set of input variable data,It is the uncorrelated space of output completely of input variable data
Data set, yNew, rIt is the uncorrelated space data sets of input of output variable data,It is input variable data
The uncorrelated space data sets of output,It is obtaining for the complete output correlation space data set of input variable data
Divide vector,It is the online uncorrelated space data sets of output for obtaining penicillin fermentation process input variable data
Score vector,It is the score vector of the uncorrelated space data sets of input of output variable data,It is the residual error space data sets in the uncorrelated space of output of input variable data,It is the residual error space data sets in the uncorrelated space of input of output variable data.
Step 4:Calculate Hotelling statistic of the output variable data in the principal component space for inputting uncorrelated space and defeated
Enter the combination statistic of the SPE statistics in the residual error space in uncorrelated space.
Wherein, ΦyThe Hotelling statistic for being output variable data in the principal component space for inputting uncorrelated space and defeated
Go out the combination statistic of SPE statistic of the variable data in the residual error space for inputting uncorrelated space,It is output variable
Limit is controlled corresponding to Hotelling statistic of the data in the principal component space for inputting uncorrelated space,It is output variable
Limit is controlled corresponding to SPE statistics of the data in the residual error space for inputting uncorrelated space.
Step 5:Current penicillin fermentation process breakdown judge:
Hotelling statistic of the input variable data for judging to be calculated in output correlation space completelyIt is whether high
In its corresponding control limit:It is that then output dependent failure occurs in current penicillin fermentation process;Otherwise step 6 is performed.
Step 6:Judge Hotelling statistic of the input variable data in the principal component space for exporting uncorrelated spaceWith
SPE statistic Q of the input variable data in the residual error space for exporting uncorrelated spacexIn whether have at least one statistic high
In its corresponding control limit:It is that then current penicillin fermentation process occurs exporting failure uncorrelated but that input is related, otherwise
Perform step 7.
Step 7:Judge Hotelling statistic of the output variable data in the principal component space for inputting uncorrelated space and defeated
Go out the combination statistic Φ of SPE statistic of the variable data in the residual error space for inputting uncorrelated spaceyWhether its phase is higher than
The control limit answered:It is that then uncertain output dependent failure occurs in current penicillin fermentation process, otherwise continues online obtain
The input variable data and output variable data of subsequent time penicillin fermentation process, return to step 3.
In present embodiment, 200 sampled points of the offline history normal data of penicillin fermentation process, profit are chosen first
With the penicillin fermentation process Fault Model based on concurrent offset minimum binary, on-line checking fault type 1 and fault type
2, the data of 1~table of table 3 are partial data:
Four groups of data that table 1 is established in the penicillin fermentation process Fault Model based on concurrent offset minimum binary
Table 2 detects two groups of data in the fault data 1 of penicillin fermentation process
Table 3 detects two groups of data in the fault data 2 of penicillin fermentation process
Although Hotelling statistic of the input variable data in output correlation space completely as seen from Figure 2
Exist between 101 to 108 sampled points and fail to report, limited in 100 sampled points not less than control, in 109-200 sampled point
It detected slope failure.And Hotelling statistic of the input variable data in the principal component space for exporting uncorrelated space
SPE statistic Q of the input variable data in the residual error space for exporting uncorrelated spacexNot phase is being inputted with output variable data
Hotelling statistic and output variable data in the principal component space in pass space is in the residual error space for inputting uncorrelated space
The combination statistic Φ of SPE statisticsyLimited in 200 sampled points not less than control.This illustrates that this failure is an output
Related slope failure, is true to life.
Although occurring the point slightly beyond control limit as seen from Figure 3, input variable data are related empty in complete output
Between in Hotelling statisticHotelling statistic of the input variable data in the principal component space for exporting uncorrelated space
Inputted with Hotelling statistic of the output variable data in the principal component space for inputting uncorrelated space and output variable data
The combination statistic Φ of SPE statistics in the residual error space in uncorrelated spaceyNot less than control in 200 sampled points
Limit.And SPE statistic Q of the input variable data in the residual error space for exporting uncorrelated spacexIn failure introduce region (101-
200) control limit has been exceeded.This illustrates that this failure is an incoherent both phase step fault of output, is true to life.
It can illustrate that this method can be good at detecting green grass or young crops by Fig. 2 (a)~(d) He Fig. 3 (a)~(d) testing result
Failure in mould category fermentation process.The testing result of the fault type 2 of fault type 1 and different amplitudes for Different Slope is shown in
4~table of table 5.
The verification and measurement ratio and test point of the fault data 1 of the penicillin fermentation process of table 4
Verification and measurement ratio, rate of false alarm and the test point of the fault data 2 of the penicillin fermentation process of table 5
Wherein, the test point in 4~table of table 5 is the sampled point that continuous five or more samples exceed control limit, is examined
The computational methods of survey rate and rate of false alarm are as follows:
As can be seen from Table 4, for different failure slopes, the input variable data in penicillin fermentation process are complete
Export the Hotelling statistic in correlation spaceVerification and measurement ratio more than 90%.And the input in penicillin fermentation process becomes
Measure Hotelling statistic of the data in the principal component space for exporting uncorrelated spaceInput variable in penicillin fermentation process
SPE statistic Q of the data in the residual error space for exporting uncorrelated spacexWith the output variable data in penicillin fermentation process
The output variable data in Hotelling statistic and penicillin fermentation process in the principal component space for inputting uncorrelated space exist
The combination statistic Φ of the SPE statistics inputted in the residual error space in uncorrelated spaceyVerification and measurement ratio within 10%.Therefore
And, it can be seen that this three statistics do not have detection to be out of order.Further, it can be seen that the input variable in penicillin fermentation process
Hotelling statistic of the data in output correlation space completelyTest point within 110.It can thus be seen that for
The failure of Different Slope, this method can be detected efficiently and accurately.
As can be seen from Table 5, for the fault type 2 of different amplitudes, the input variable data in penicillin fermentation process
SPE statistics Q in the residual error space for exporting uncorrelated spacexVerification and measurement ratio more than 98%, test point 103 with
It is interior, Hotelling statistic of the input variable data in output correlation space completely in penicillin fermentation processRate of false alarm
Within 1.5%.Therefore, this method can efficiently and accurately detect fault type 2.
It can be obtained by the above results, by the inventive method, different type, different faults size can be effectively detected
Output dependent failure and export uncorrelated failure.
Although the foregoing describing the embodiment of the present invention, those skilled in the art in the art should manage
Solution, these are merely illustrative of, and can make various changes or modifications to these embodiments, without departing from the principle of the present invention
And essence.The scope of the present invention is only limited by the claims that follow.
Claims (7)
- A kind of 1. procedure failure testing method based on concurrent offset minimum binary, it is characterised in that includingThe online input variable data and output variable data for obtaining penicillin fermentation process;Utilize the penicillin fermentation process Fault Model processing input variable data based on concurrent offset minimum binary and output Variable data:Input variable data are expressed as exporting correlation space data set completely, export the principal component space in uncorrelated space Data set and the residual error space data sets sum for exporting uncorrelated space, output variable data are expressed as to input correlation space number According to the principal component space data set and the residual error space data sets sum in the uncorrelated space of input for collecting, inputting uncorrelated space;It is empty to calculate Hotelling statistic, the pivot in output uncorrelated space of the input variable data in output correlation space completely Between in Hotelling statistic, the uncorrelated space of output residual error space in SPE statistics, calculate output variable data defeated Enter the SPE statistics in the Hotelling statistic in the principal component space in uncorrelated space and the residual error space in the uncorrelated space of input Amount;Calculate Hotelling statistic and input uncorrelated sky of the output variable data in the principal component space for inputting uncorrelated space Between residual error space in SPE statistics combination statistic;Current penicillin fermentation process breakdown judge:Whether Hotelling statistic of the input variable data for judging to be calculated in output correlation space completely is higher than its phase The control limit answered:It is that then output dependent failure occurs in current penicillin fermentation process;Otherwise:Judge Hotelling statistic and input variable data of the input variable data in the principal component space for exporting uncorrelated space Whether at least one statistic is had in SPE statistics in the residual error space for exporting uncorrelated space higher than its corresponding control Limit:It is that then current penicillin fermentation process occurs exporting failure uncorrelated but that input is related, otherwise:Judge Hotelling statistic and output of the output variable data in the principal component space for inputting uncorrelated space in input not Whether the combination statistic of the SPE statistics in the residual error space of correlation space is higher than its corresponding control limit:Be, then it is current blue or green There is uncertain output dependent failure in mycin fermentation process, otherwise continues online acquisition subsequent time penicillin fermentation process Input variable data and output variable data.
- 2. according to the method for claim 1, it is characterised in that the input variable data are in penicillin fermentation process The concentration of substrate, the concentration of carbon dioxide, ventilation rate, agitator power, pH value and substrate feeding temperature, the output variable number According to the concentration and the temperature of fermentation reactor for the penicillin in penicillin fermentation process.
- 3. according to the method for claim 1, it is characterised in that utilize the penicillin based on concurrent offset minimum binary described Before fermentation process Fault Model handles input variable data and output variable data, to the input variable number obtained online Specification and standardization are carried out according to output variable data.
- 4. according to the method for claim 3, it is characterised in that the specification and standardization, specifically become input respectively The average of amount data and output variable data is 0 and variance is 1.
- 5. according to the method for claim 1, it is characterised in that the penicillin fermentation mistake based on concurrent offset minimum binary The method for building up of journey Fault Model is as follows:Gather the offline history normal data set of input variable and the offline history normal data set of output variable;Using orthogonal offset minimum binary by the offline history normal data set of input variable be divided into output correlation space data set and Export orthogonal intersection space data set;Correlation space data set will be exported using complete orthogonal decomposition and the offline history normal data set of output variable is carried out respectively Complete orthogonal decomposition, obtain the complete output correlation space data set of the offline history normal data set of input variable, output completely The input correlation space data set of the uncorrelated offline history normal data set of space data sets and output variable and input are uncorrelated Space data sets;Export the output orthogonal intersection space data set of the offline history normal data set of input variable and completely uncorrelated spatial data Collection merges into the uncorrelated space data sets of output;Uncorrelated space data sets will be exported and be divided into the principal component space data set for exporting uncorrelated space and the uncorrelated sky of output Between residual error space data sets;The uncorrelated space data sets of input of the offline history normal data set of output variable are divided into the uncorrelated space of input Principal component space data set and the residual error space data sets for inputting uncorrelated space;Input variable data are expressed as exporting correlation space data set completely, export the principal component space data set in uncorrelated space With the residual error space data sets sum for exporting uncorrelated space, by output variable data be expressed as inputting correlation space data set, Input the principal component space data set in uncorrelated space and input the residual error space data sets sum in uncorrelated space, that is, be based on The penicillin fermentation process Fault Model of concurrent offset minimum binary.
- 6. according to the method for claim 5, it is characterised in that it is described using orthogonal offset minimum binary by input variable from Line history normal data set be divided into output correlation space data set and output orthogonal intersection space data set before, to input variable from Line history normal data set and the offline history normal data set of output variable carry out specification and standardization.
- 7. according to the method for claim 5, it is characterised in that described to export correlation space number using complete orthogonal decomposition Complete orthogonal decomposition is carried out respectively according to collection and the offline history normal data set of output variable, is specifically:The offline history normal data set of output variable is mapped to output correlation space and obtains regression coefficient matrix;Will output correlation space data set be mapped to regression coefficient matrix obtain input correlation space data set, by output variable from Line history normal data set subtracts input correlation space data set and obtains inputting uncorrelated space data sets;The transposed matrix of regression coefficient Matrix Multiplication regression coefficient matrix is obtained into regression matrix, carrying out singular value decomposition to it obtains two Individual mutually orthogonal mapping matrix;Output correlation space data set is mapped into the mutually orthogonal mapping matrix of two above to obtain exporting correlation space completely Data set and uncorrelated space data sets are exported completely.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711013418.6A CN107817784B (en) | 2017-10-26 | 2017-10-26 | A kind of procedure failure testing method based on concurrent offset minimum binary |
PCT/CN2018/087693 WO2019080489A1 (en) | 2017-10-26 | 2018-05-21 | Process fault detection method based on concurrent partial least squares |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711013418.6A CN107817784B (en) | 2017-10-26 | 2017-10-26 | A kind of procedure failure testing method based on concurrent offset minimum binary |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107817784A true CN107817784A (en) | 2018-03-20 |
CN107817784B CN107817784B (en) | 2019-07-23 |
Family
ID=61604198
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711013418.6A Active CN107817784B (en) | 2017-10-26 | 2017-10-26 | A kind of procedure failure testing method based on concurrent offset minimum binary |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107817784B (en) |
WO (1) | WO2019080489A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019080489A1 (en) * | 2017-10-26 | 2019-05-02 | 东北大学 | Process fault detection method based on concurrent partial least squares |
CN110928263A (en) * | 2019-12-17 | 2020-03-27 | 中国人民解放军火箭军工程大学 | Fault detection method and system for complex process considering dynamic relationship in advance |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2000077497A1 (en) * | 1999-06-11 | 2000-12-21 | Metso Paper Automation Oy | Method and apparatus for measuring properties of paper web |
CN1602830A (en) * | 2004-11-09 | 2005-04-06 | 清华大学 | Real-time monitoring method for traditional Chinese medicine process |
CN101964021A (en) * | 2010-09-29 | 2011-02-02 | 东北大学 | Penicillin fermentation process failure monitoring method based on recursive kernel principal component analysis |
CN104133991A (en) * | 2014-07-15 | 2014-11-05 | 东北大学 | Penicillin fermentation process fault diagnosis method based on kernel partial least squares reconstitution |
CN104133990A (en) * | 2014-07-15 | 2014-11-05 | 东北大学 | Penicillin fermentation process fault isolation method based on kernel least square regression |
CN104914847A (en) * | 2015-04-09 | 2015-09-16 | 东北大学 | Industrial process fault diagnosis method based on direction kernel partial least square |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AR091412A1 (en) * | 2013-06-11 | 2015-02-04 | Ypf Sa | DEVICE AND METHOD FOR DIAGNOSING ANOMAL SITUATIONS IN PROCESSES |
CN107122611A (en) * | 2017-04-28 | 2017-09-01 | 中国石油大学(华东) | Penicillin fermentation process quality dependent failure detection method |
CN107817784B (en) * | 2017-10-26 | 2019-07-23 | 东北大学 | A kind of procedure failure testing method based on concurrent offset minimum binary |
-
2017
- 2017-10-26 CN CN201711013418.6A patent/CN107817784B/en active Active
-
2018
- 2018-05-21 WO PCT/CN2018/087693 patent/WO2019080489A1/en active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2000077497A1 (en) * | 1999-06-11 | 2000-12-21 | Metso Paper Automation Oy | Method and apparatus for measuring properties of paper web |
CN1602830A (en) * | 2004-11-09 | 2005-04-06 | 清华大学 | Real-time monitoring method for traditional Chinese medicine process |
CN101964021A (en) * | 2010-09-29 | 2011-02-02 | 东北大学 | Penicillin fermentation process failure monitoring method based on recursive kernel principal component analysis |
CN104133991A (en) * | 2014-07-15 | 2014-11-05 | 东北大学 | Penicillin fermentation process fault diagnosis method based on kernel partial least squares reconstitution |
CN104133990A (en) * | 2014-07-15 | 2014-11-05 | 东北大学 | Penicillin fermentation process fault isolation method based on kernel least square regression |
CN104914847A (en) * | 2015-04-09 | 2015-09-16 | 东北大学 | Industrial process fault diagnosis method based on direction kernel partial least square |
Non-Patent Citations (3)
Title |
---|
WENYOU DU, ETC.: "Fault diagnosis of non-Gaussian process based on FKICA", 《JOURNAL OF THE FRANKLIN INSTITUTE》 * |
刘世成 等: "青霉素生产过程的在线统计监测与产品质量控制", 《计算机与应用化学》 * |
张颖伟 等: "基于DKPLS的非线性过程故障检测", 《华中科技大学学报(自然科学版)》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019080489A1 (en) * | 2017-10-26 | 2019-05-02 | 东北大学 | Process fault detection method based on concurrent partial least squares |
CN110928263A (en) * | 2019-12-17 | 2020-03-27 | 中国人民解放军火箭军工程大学 | Fault detection method and system for complex process considering dynamic relationship in advance |
CN110928263B (en) * | 2019-12-17 | 2022-10-28 | 中国人民解放军火箭军工程大学 | Fault detection method and system for complex process considering dynamic relationship in advance |
Also Published As
Publication number | Publication date |
---|---|
WO2019080489A1 (en) | 2019-05-02 |
CN107817784B (en) | 2019-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107357275B (en) | Non-gaussian industrial process fault detection method and system | |
CN105700518B (en) | A kind of industrial process method for diagnosing faults | |
CN106404441B (en) | A kind of failure modes diagnostic method based on non-linear similarity index | |
CN107153409B (en) | A kind of nongausian process monitoring method based on missing variable modeling thinking | |
CN107544477A (en) | Nonlinear industrial processes fault detection method based on core pivot element analysis | |
CN109085805B (en) | Industrial process fault detection method based on multi-sampling-rate factor analysis model | |
CN111580506A (en) | Industrial process fault diagnosis method based on information fusion | |
CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
CN110083860A (en) | A kind of industrial method for diagnosing faults based on correlated variables selection | |
CN111126759A (en) | Electric energy meter state evaluation method based on abnormal event fault correlation degree | |
CN109409425A (en) | A kind of fault type recognition method based on neighbour's constituent analysis | |
CN107817784B (en) | A kind of procedure failure testing method based on concurrent offset minimum binary | |
CN105045220B (en) | A kind of method of quality control based on multivariable Z score quality control chart for being used for laboratory diagnosis field or field of industrial production | |
CN107122611A (en) | Penicillin fermentation process quality dependent failure detection method | |
CN111538655A (en) | Software failure detection method, system, storage medium, computer program, and terminal | |
CN111122811A (en) | Sewage treatment process fault monitoring method of OICA and RNN fusion model | |
CN110244692A (en) | Chemical process small fault detection method | |
CN110209145B (en) | Carbon dioxide absorption tower fault diagnosis method based on nuclear matrix approximation | |
CN107918381A (en) | A kind of class average core pivot method for diagnosing faults based on compound kernel function | |
CN110059010A (en) | The buffer overflow detection method with fuzz testing is executed based on dynamic symbol | |
CN106250937B (en) | A kind of failure modes diagnostic method based on non-index of similarity | |
CN106959397A (en) | A kind of method for designing of small fault diagnostic system for high ferro inverter | |
CN110414086A (en) | A kind of combined stress accelerated factor calculation method based on sensitivity | |
CN111188761A (en) | Monitoring method for pump equipment based on Fourier-CVA model | |
CN114417704A (en) | Wind turbine generator health assessment method based on improved stack type self-coding |
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 |