CN107798353A - A kind of batch process monitoring data processing method - Google Patents
A kind of batch process monitoring data processing method Download PDFInfo
- Publication number
- CN107798353A CN107798353A CN201711134478.3A CN201711134478A CN107798353A CN 107798353 A CN107798353 A CN 107798353A CN 201711134478 A CN201711134478 A CN 201711134478A CN 107798353 A CN107798353 A CN 107798353A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msubsup
- period
- msup
- sub
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 55
- 238000010923 batch production Methods 0.000 title claims abstract description 35
- 238000003672 processing method Methods 0.000 title claims abstract description 11
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 22
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 17
- 238000013461 design Methods 0.000 claims abstract description 13
- 230000007704 transition Effects 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000007423 decrease Effects 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 4
- SPSSULHKWOKEEL-UHFFFAOYSA-N 2,4,6-trinitrotoluene Chemical compound CC1=C([N+]([O-])=O)C=C([N+]([O-])=O)C=C1[N+]([O-])=O SPSSULHKWOKEEL-UHFFFAOYSA-N 0.000 claims description 3
- 238000010408 sweeping Methods 0.000 claims 2
- 238000000034 method Methods 0.000 abstract description 30
- 230000008569 process Effects 0.000 abstract description 8
- 230000003247 decreasing effect Effects 0.000 abstract description 7
- 238000004458 analytical method Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000009514 concussion Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Complex Calculations (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The present invention relates to Control Science and Engineering field, especially a kind of batch process monitoring data processing method, comprise the following steps:A. improved affine propagation clustering algorithm is used, b. designs similarity is successively decreased scanning algorithm, and c. determines outlier and misclassified gene and separately designs solution;Improved affine propagation clustering algorithm can more accurately disclose the Mode-switch process between each sub-period of batch process, similarity is successively decreased the trend that scanning algorithm can further reflect that mode changes over time inside each sub-period, determine outlier and misclassified gene and respectively solution method corresponding to proposition, improve the degree of accuracy of statistical modeling, algorithm small scale, computing are simple, are easy to Project Realization.
Description
Technical field
The present invention relates to Control Science and Engineering field, especially a kind of batch process monitoring data processing method.
Background technology
For a long time, as active demand of the market to more categories, small lot and high value added product, Batch Process turn into
The main modes of production of many industrial circles.But the complexity of Batch Process can inevitably lead to the reliable of production
The appearance of property and safety problem.In order to more accurately capture the mechanism of batch process to improve monitoring performance, supervised in time with this
Potential safety problem is controlled, sphere of learning has carried out in-depth study to batch process, and proposes many data and drive
Dynamic multivariate statistical analysis method.For example, typical PCA and least square method and their extension form are all
Through being widely used.However, these traditional methods, which are substantially all, assumes that the variation tendency of whole batch process is
Smoothly, therefore to whole batch process single model is established.Thus since, the correlation of process will be hidden, and be entered
And weaken the performance of on-line monitoring so that the risk that the on-line monitoring of batch process is subject to report by mistake and failed to report.
In view of the particularity of batch process, single model can not truly react the variation tendency of whole process,
Related researcher has also been proposed the thought of divided stages, entirely will be entered according to the difference of procedural correlation batch production process
One step is divided into independent sub-stage, then again to model corresponding to the foundation of each sub-stage.In the research of divided stages thought
Cheng Zhong, Zhao Chunhui et al. propose K mean cluster algorithm, and Qi Yongsheng et al. proposes Fuzzy C-Means Cluster Algorithm.But this
Two methods all limitation of various degrees when carrying out divided stages.K mean cluster is entered to initial cluster center
Randomness is bigger and more sensitive to noise and outlier during row selection, and this will cause cluster result unstable, clusters matter
Amount is therefore limited.And fuzzy C-means clustering is a kind of local search algorithm, it needs the artificial number for determining cluster centre, and
And its iterative search is easily absorbed in locally optimal solution, therefore cluster result is often unsatisfactory.It is in fact, multiple for one
Miscellaneous batch process, it is difficult to the number of cluster centre is accurately estimated in no priori or in the case of having no experience, because
This Accurate Model for being unfavorable for realizing batch process and monitoring.
The content of the invention
The technical problems to be solved by the invention are to overcome the deficiencies in the prior art, there is provided a kind of algorithm scale
Small, computing is simple, is easy to the batch process monitoring data processing method of Project Realization.
The present invention is to be achieved by the following technical programs:
A kind of batch process monitoring data processing method, it is characterised in that comprise the following steps:
A. data are divided into advance using improved affine propagation clustering algorithm;
B. design similarity successively decrease scanning algorithm carry out period of the day from 11 p.m. to 1 a.m segment data carefully divide, judge the property of each sub-period, really
Sub-period stablizes sub-period or transition sub-period, or determines both when existing concurrently with steady component and transition portion
Separation;
C. outlier and misclassified gene are determined, outlier is rejected before off-line modeling, misclassified gene is divided from current
Place sub-period takes out and puts back to sub-period where former sample sequence.
Further, the input data similarity formula in the step a in improved affine propagation clustering algorithm is:
In formula, s (i, k) is the input data similarity of improved affine propagation clustering algorithm, Ci(q, r) and Ck(q, r) table
Show monitoring data any two timeslice matrix XiAnd XkCovariance matrix, J is the dimension of covariance matrix.
Further, similarity successively decreases scanning algorithm when implementing scanning in the step b, scans obtained sample every time
Average similarity formula is:
In formula, m represents period of the day from 11 p.m. to 1 a.m segment number, and n represents scanning times, and k represents sample number,Represent m-th of sub-period
The average similarity of n-th scanning,The quantity of the sample obtained in being scanned from m-th of sub-period n-th is represented,Represent
The similarity of m-th of sub-period, k-th of sample,
To m-th of period of the day from 11 p.m. to 1 a.m, the distribution situation of the average similarity obtained according to scanning, with Current Scan step scan knot
After beam, implement tolerance scanning, i.e., tried again scanning with the step-length in the tolerance less than Current Scan step-length, swept afterwards
The process of retouching terminates, and scans, puts down when m-th of sub-period experienced n' scanning and finally implement tolerance altogether in addition to tolerating and scanning
Equal similarity variation tendency is present
Wherein, n and n' refers both to scanning times and n'> n, now, two kinds of situations as shown in formula (4) be present:
For case 1, the sample in region before steady component is scanned by n-th forms, and transition portion is by positioned at n-th
It is secondary scanning tolerance scan between region sample into;
For case 2, the sample in region before steady component is still scanned by n-th forms, transition portion then by positioned at
N-th scan and (n-1) ' secondary scanning between region sample into,
Wherein,It is the average similarity of tolerance scanning,
When meeting n' βmLm≤Lmax,m, and exist
I.e. when average similarity is continuous equal three times, n-th ' the secondary end of scan no longer carries out tolerance scanning, steady component
By sweep radius (n'-2) βmLmWithin sample composition, the remaining sample point not scanned is considered as outlier,
When meeting n' βmLm≥Lmaxm, and exist
I.e. when the equal situation of any average similarity does not occur in whole sweep phase, then whole sub-period can be with
It is considered as stable sub-period,
In formula, m is period of the day from 11 p.m. to 1 a.m segment number, βmFor step factor, βm∈ [0.5,1], LmIt is the cluster centre of m-th of sub-period
The distance of the sample of cluster centre, L are only second to similaritymax,mBe m-th of sub-period cluster centre it is farthest to the sub-period
The distance of sample, as include the maximum radius of all samples of m-th of sub-period.
Further, in the step c after the end of scan, scanned when having been carried out tolerance, and (n' β be presentmLm)tol<
Lmax,m, then tolerate that the remaining sample beyond sweep radius is outlier, (n' βmLm)tolIt is the tolerance scanning half of m-th of sub-period
Footpath,
When the end of scan, shown in the average similarity such as formula (5) of sample, occur continuous equal three times and n' β be presentmLm< Lmax,m, then the remaining sample not scanned is also considered as outlier,
Outlier constitutes threat to the Accurate Model of nominal situation down-sampled data, is removed before off-line modeling,
In the step c, carefully divided through step b period of the day from 11 p.m. to 1 a.m segment data, the sub-period that some samples occur not is them
Sub-period corresponding to sample data sequence, these samples are referred to as misclassified gene,
The sub-period that sub-period where misclassified gene from current division is taken out and put back to where former sample sequence.
The beneficial effects of the invention are as follows:
(1) improved affine propagation clustering algorithm can more accurately disclose the mould between each sub-period of batch process
State handoff procedure;
(2) similarity scanning algorithm of successively decreasing can further reflect that what mode inside each sub-period changed over time becomes
Gesture;
(3) outlier and misclassified gene and respectively solution method corresponding to proposition are determined, improves the accurate of statistical modeling
Degree;
(4) algorithm small scale, computing are simple, are easy to Project Realization.
Brief description of the drawings
Fig. 1 is batch process monitoring data processing method flow chart provided by the invention.
Fig. 2 is that the original monitoring data of batch process pre-processes schematic diagram.
Fig. 3 is that monitoring data divides implementing procedure figure in advance.
Fig. 4 is that similarity is successively decreased scanning algorithm implementation principle figure.
Fig. 5 is sub-period carefully Different Results figure caused by division.
Fig. 6 is the schematic diagram for determining and handling misclassified gene.
Fig. 7 is off-line modeling and on-line monitoring flow chart.
Fig. 8 (a) is that monitoring data divides design sketch in advance.
Fig. 8 (b) is that sub-period carefully divides design sketch.
Fig. 9 is traditional affine propagation clustering algorithm effect figure.
Figure 10 (a) is the T of method provided by the present invention under normal batch2Monitoring effect figure.
Figure 10 (b) is the SPE monitoring effect figures of method provided by the present invention under normal batch.
Figure 11 (a) is the T of the multidirectional principal component analytical method of tradition under normal batch2Monitoring effect figure.
Figure 11 (b) is the SPE monitoring effect figures of the multidirectional principal component analytical method of tradition under normal batch.
Figure 12 (a) is the T of method provided by the present invention2Failure monitoring design sketch.
Figure 12 (b) is the SPE failure monitoring design sketch of method provided by the present invention.
Figure 13 (a) is the T of the multidirectional principal component analytical method of tradition2Failure monitoring design sketch.
Figure 13 (b) is the SPE failure monitoring design sketch of the multidirectional principal component analytical method of tradition.
Embodiment
In order that those skilled in the art more fully understand technical scheme, below in conjunction with the accompanying drawings and most
The present invention is described in further detail for good embodiment.
As shown in figure 1, batch process monitoring data division provided by the invention and statistical modeling method include entering in order
Capable the following steps:
(1) data are carried out using improved affine propagation clustering algorithm to divide in advance;
Division needs to do original sampling data pretreatment as shown in Figure 2 first described data in advance, i.e., by original three
Dimension monitoring data X (I × J × K) deploys along time shaft, obtains K timeslice matrix Xk(I × J), k=1,2 ..., K, then will
Xk(I × J) standardization, centralization, obtain covariance matrix C corresponding to each timeslice successivelyk(J×J)。
Wherein, I represents batch, and J represents variable, and K represents the sampling time.
Described improved affine propagation clustering algorithm is to set the similarity of the input data of affine propagation clustering algorithm
It is calculated as shown in formula (1):
In formula, s (i, k) is the input data similarity of improved affine propagation clustering algorithm, Ci(q, r) and Ck(q, r) table
Show monitoring data any two timeslice matrix XiAnd XkCovariance matrix, J is the dimension of covariance matrix.
The described process divided in advance using improved affine propagation clustering algorithm progress data is as shown in Figure 3, it is assumed that institute
State algorithm and monitoring data is divided into M sub-period altogether.The similarity of input data is calculated by formula (1).It is noted that formula
(1) i ≠ k in, therefore, the similarity matrix diagonal entry s (k, k) of input is vacancy, then sets Λ (k)=s (k, k), Λ
(k) represent deviation degree and value is identical.Λ (k) initial value is set to Λ0(k)=min (s (i, k)).
Shown in information transmission and rule of iteration such as formula (2)-(6):
rt(i, k)=(1- λ) × rt(i,k)+λ×rt-1(i,k) (5)
at(i, k)=(1- λ) × at(i,k)+λ×at-1(i,k) (6)
Wherein, t is iterations, and λ represents the concussion factor, and r (i, k) represents XkAs XiCluster centre evidence, a (i,
K) X is referred toiSelect XkAppropriate level as its cluster centre.S (i, k'), a (i, k') and r (i', k) are samples in iterative process
The renewal of this generation.Optimum cluster result can be obtained by Adaptive adjusting algorithm parameter, realize pre- stroke to monitoring data
Point, obtain each cluster centre and corresponding sub-period.
(2) design similarity successively decrease scanning algorithm carry out period of the day from 11 p.m. to 1 a.m segment data carefully divide;
Described similarity successively decreases scanning algorithm as shown in figure 4, its principle is:
Pre- division is done to monitoring data using the method described in step (1), the result is that whole batch process data are drawn
It is divided into different sub-periods.In order to further disclose the variation tendency of each sub-period built-in variable correlation, similarity is designed
Scanning algorithm of successively decreasing investigates each sub-period, that is, does sub-period and carefully divide, and then judges the property of each sub-period, that is, determines every
Individual sub-period stablizes sub-period or transition sub-period, or determines both when existing concurrently with steady component and transition portion
Separation.Described similarity successively decreases scanning algorithm when doing sub-period and carefully dividing, by scanning and sub-period property determination two
It is grouped into.
The successively decrease scanning theory of scanning algorithm of described similarity is:
Generally, sampled data is nearer apart from cluster centre, and similarity is higher, and correlation is stronger, and variance is smaller, turns into stable
Partial possibility is bigger, based on this principle, illustrates that similarity is successively decreased scanning algorithm by taking m-th of sub-period as an example.M-th of son of note
The cluster centre of period is Em, with EmFor the center of circle, EmThe distance L of the farthest sample data included to the sub-periodmax,mFor maximum
Bound radius, thus form a circle for including all sample datas of the sub-period.In the sub-period, cluster centre is corresponding
Sample point be self similarity, namely circle EmIt is interior, the sample that a similarity is only second to cluster centre is found out, is denoted as Am, and will
EmAnd AmThe distance between be designated as Lm, then β is definedmLm,βm∈ [0.5,1] is with respect to fixed length scanning step, βmFor step factor.It is right
In circle Em, cluster centre EmIt is set to the starting point of scanning, βmLmIt is set to unit sweep radius, namely the radius of first scan, it is later every
Secondary scanning, sweep radius all increase a unit sweep radius.Each end of scan, record sample corresponding to Current Scan number
The average similarity of data, is denoted as:
In formula, m represents period of the day from 11 p.m. to 1 a.m segment number, and n represents scanning times, and k represents sample number,Represent m-th of sub-period
The average similarity of n-th scanning,The quantity of the sample obtained in being scanned from m-th of sub-period n-th is represented,Represent
The similarity of m-th of sub-period, k-th of sample,
When the average similarity for finding adjacent twice sweep first is equal, it is designated asContinue to scan on, until
Occur the equal situation of the average similarity of adjacent twice sweep for the second time, be designated asTolerance is now done to sweep
Retouch.Tolerance factor α (α≤min { β are setILI,βIILII,…,βmLm,…,βMLM), and be added to Current Scan radius, i.e.,
The sweep radius of tolerance scanning is (n' βmLm)tol=n' βmLm+ α, the tolerance end of scan mean the scanning knot of whole sub-period
Beam.
Illustrate that similarity is successively decreased scanning theory in conjunction with Fig. 4.In Fig. 4, stain and ash point represent different samples, βmTemporarily
When value be 1.When with sweep radius 3LmWhen carrying out the 3rd scanning, any new sample is not scanned, and now existContinue to scan on until occurringThen tolerance scanning is done.Two dashdotted circle is enclosed
It is tolerance region in region, the remainder beyond this region will be scanned no longer.Meanwhile if occur in scanning process continuous
Scanning result is equal three times, i.e.,Scanning also terminates.
The successively decrease sub-period property determination principle of scanning algorithm of described similarity is:
According to scanning situation, each sub-period can further determine that out steady component and transition portion.Occur when first time
When adjacent twice sweep result is equal, currently the variation tendency of scanning area sample is relatively more steady and correlation is of a relatively high,
These samples can be considered as stablizing sample, collectively form a steady component.When adjacent second of the phase of twice sweep result of appearance
Deng when, if it by sweep radius is (n' β to be situations such as this phasemLm)tol=n' βmLm+ α tolerance scanning then will come what is changed
Transition portion is defined as by between sweep radius n βmLm(n' βmLm)tolBetween region sample composition;If such case
Condition with change is that sweep radius meets (n' βmLm)tol′> n' βmLm+ C α, C are arbitrary constants, then transition portion
By sweep radius n βmLm(n'-1) βmLmBetween region include sample composition.Do not swept in Fig. 4, during the 5th end of scan
Retouch any new sample and presenceIt is (5L then to implement sweep radiusm)tol=5Lm+ α tolerance is swept
Retouch, and scan to a sample point, in tolerance region.After tolerating the end of scan, positioned at sweep radius 2LmWithin
The sample composition steady component that stain represents.Sweep radius 2LmWhat the point into the region between tolerance sweep radius n' represented
Sample collectively constitutes transition portion.If tolerance scanning does not obtain any new sample point, transition portion is only by scanning half
Footpath 2LmAnd 4LmBetween sample into.The sample that two Grey Points represent in Fig. 4 is described further below.
Assuming that m-th of sub-period experienced n' scanning and finally implement tolerance scanning altogether in addition to tolerating and scanning, then
If average similarity variation tendency is present
Wherein, n and n' refers both to scanning times and n'> n.Now, two kinds of situations as shown in formula (9) be present.For feelings
Condition 1, the sample in region before steady component is scanned by n-th form, and transition portion positioned at n-th by scanning and tolerating scanning
Between region sample into;For situation 2, the sample in region before steady component is still scanned by n-th forms, mistake
Cross part then by positioned at n-th scanning (n-1) ' secondary scanning between region sample into.
Wherein,It is the average similarity of tolerance scanning.Meanwhile if average similarity variation tendency is expressed
Shown in formula such as formula (10) and (11), i.e., when average similarity is continuous equal three times, n-th ' the secondary end of scan is no longer held
Bear scanning, steady component is by sweep radius (n'-2) βmLmWithin sample composition, the remaining sample point not scanned be considered as from
Group's point;Or when the equal situation of any average similarity does not occur in whole sweep phase, illustrate the similar of m-th sub-period
Degree distribution is steady smooth, then whole sub-period can be considered as stable sub-period.
When meeting n' βmLm≤Lmax,m, and exist
Or work as and meet n' βmLm≥Lmax,m, and exist
Wherein, m is period of the day from 11 p.m. to 1 a.m segment number, βmFor step factor, βm∈ [0.5,1], LmIt is the cluster centre of m-th of sub-period
The distance of the sample of cluster centre, L are only second to similaritymax,mBe m-th of sub-period cluster centre it is farthest to the sub-period
The distance of sample, as include the maximum radius of all samples of m-th of sub-period.
The similarity successively decreases scanning algorithm when carefully being divided to sub-period, steady component and transition part in division result
Five kinds of situations as shown in Figure 5 be present in the distribution divided:
1. transition portion occurs over just the starting position of sub-period.
2. transition portion occurs over just the end position of sub-period.
3. transition portion occurs in the beginning and end position of sub-period simultaneously.
4. whole sub-period is transition portion.After the end of scan, if it find that fall into the sample of stability region
Quantity very little, is not enough to form independent steady component, then these stable sample points are included into current transition portion.
5. whole sub-period is steady component.If the quantity for falling into the sample of transitional region is few enough, structure is not enough to
Into independent transition portion, then these transition samples are included into current steady component, or cluster result is good enough so that sweep
There is not transition portion in the result retouched, the sub-period forms independent stable sub-period.
(3) outlier and misclassified gene are determined and separately designs its corresponding solution method;
The determination of described outlier and solution method are:
After the end of scan, if having been carried out tolerance scanning, and (n' β be presentmLm)tol< Lmax,m, then sweep radius is tolerated
Remaining sample in addition is outlier.Wherein, (n' βmLm)tolIt is the tolerance sweep radius of m-th of sub-period.If the end of scan
When, shown in the average similarity such as formula (10) of sample, occur continuous equal three times and n' β be presentmLm< Lmax,m, then do not scan
To remaining sample be also considered as outlier.With reference to Fig. 4, the sector scanning radius position for the sample appearance that one of Grey Point represents
In tolerance sweep radius (5Lm+ α) and maximum boundary radius Lmax,mBetween, then the sample point is identified as outlier.Another ash
Color dot is apart from cluster centre EmFarthest point, is not also scanned, is considered as outlier.Because outlier is to nominal situation down-sampling
The Accurate Model of data constitutes threat, therefore should be removed before off-line modeling.
Determination and the solution method of described misclassified gene are as shown in fig. 6, its concrete principle is:
Under normal circumstances, cluster result generally meets sample sequence.Because being compared with given sampled point, difference compared with
Small sampled point is typically situated near given sampled point, and sample sequence is sufficiently spaced-apart the similar of two remote sampled points
Degree is very low, and in most cases, they are the same sub-periods for not appearing in cluster, but is also occurred indivedual anti-
Reason condition.Divided, the sub-period that some sample points occur not is sub-period corresponding to their sample sequence, then by these
Sample point is referred to as misclassified gene.With reference to Fig. 6, illustrate by taking the monitoring data with four sub-periods as an example what be misclassified gene and
How misclassified gene is handled.Fig. 6 shows any two misclassified gene, is designated as A respectivelyIII,I,AIII,IV.For sampled point
AIII,I, it belongs to sub-period III according to sample sequence originally, but appears in sub-period I after cluster.Son is appeared in compared to it
Interval I II situation, it is unusual, therefore referred to as misclassified gene.Sampled point AIII,IVAnd same situation.Due to all
Monitoring data be all the batch process picked up under nominal situation, it is therefore necessary to by the son where misclassified gene from current division
The sub-period that period takes out and put back to where former sample sequence.Presence in view of indivedual misclassified genes does not interfere with sub-period
Thin division result, then the division of current sub-period and the atom period where it will not be changed by returning to the misclassified gene of atom period
As a result, a sample under nominal situation is only corresponded to.The Fault Identification for being easy to batch process is so done, and then reduces failure
Rate of false alarm.
(4) off-line modeling and on-line monitoring method are designed;
Described off-line modeling and on-line monitoring process is as shown in fig. 7, described off-line modeling is:
If m-th of subdivision includes kmIndividual timeslice, then these timeslices formed a three-dimensional matriceIt is deployed by variable direction, forms a two-dimensional matrixBy data processing,
It can establish such as the analysis model of formula (12)-(16):
Wherein, TmAnd PmPrincipal component matrix and load matrix, t are represented respectivelymIt is score vector, EmRepresent residual matrix, em
It is residual vector,And SPEmIt is statistic, SmIt is TmDiagonal matrix.
Described on-line monitoring is:
After off-line modeling completion, described on-line monitoring is realized according to the following steps, and then determine real-time sampling number
According to whether unusual:
1. determine which subdivision is present sample data belong to;
2. standardize present sample data;
3. calculate T2With SPE statistics;
4. judge T2Whether transfinited with SPE statistics:If new statistic does not transfinite, production process is normal;Such as
Fruit is transfinited, then triggers alarm and show that production process is abnormal.
(5) simulation result
In order to verify the validity of the batch process monitoring data division and statistical modeling method, the present invention uses steering wheel
Platform carries out confirmatory experiment.The amplitude and frequency of this experiment current loop control are respectively set to 1.0A, 1.00HZ.It is real for protection
Test for the sake of platform, this tests the data of 10 batches acquired altogether under nominal situation, 6 variables and 246 sampled points, shape
Into monitoring data matrix X (10 × 6 × 246)
Fig. 8 is monitoring data division design sketch.In Fig. 8 (a), divided in advance by data, the running status of steering wheel can divide
For 6 sub-periods, Fig. 8 (b) is the result that sub-period carefully divides.Fig. 9 is to use traditional affine propagation clustering algorithm implementing monitoring
The contrast experiment of data division.Can be seen that improved affine propagation clustering algorithm from Fig. 8 (a) and Fig. 9 has good property
Can, the monitoring data variation characteristic more suitable for disclosing complicated batch process.
The result divided according to data implements monitoring strategies, and does contrast in fact with the multidirectional principal component analytical method of tradition
Test, experimental result is as shown in Figure 10 and Figure 11.From experimental result, divided and counted by the batch process monitoring data
Modeling method, whole flight course is divided into different sub-periods, and each sub-period is further identified and divided,
Steady component and transition portion are determined, and then improves the precision of data monitoring.Moreover, pass through contrast, the batch process
Monitoring data divides and statistical modeling method is substantially better than the multidirectional principal component analytical method of tradition.
Further to verify the failure monitoring ability of the batch process monitoring data division and statistical modeling method, by one
The sample that the sample sequence that kind of steering wheel fault type adds to a monitoring data batch at random is 109th-227th, described in implementation
Algorithm and the failure monitoring for contrasting algorithm.It can be seen from Figure 12 (b) the SPE statistical values of the algorithm to added failure very
Sensitivity, just transfinited since first fault sample.By contrast, in Figure 13 (b), divide when with the multidirectional principal component of tradition
When analysis method does failure monitoring, SPE values just transfinite after two fault samples are postponed, and go out among fault sample sequence
Failing to report phenomenon is showed, this explanation multidirectional principal component analytical method of tradition is to failure monitoring and insensitive.Pass through pair of two kinds of algorithms
Understand that performance of the algorithm in batch process fault diagnosis is gratifying than experiment.
The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (4)
1. a kind of batch process monitoring data processing method, it is characterised in that comprise the following steps:
A. data are divided into advance using improved affine propagation clustering algorithm;
B. design similarity successively decrease scanning algorithm carry out period of the day from 11 p.m. to 1 a.m segment data carefully divide, the property of each sub-period is judged, it is determined that sub
Period stablizes sub-period or transition sub-period, or determines both points when existing concurrently with steady component and transition portion
Boundary's point;
C. outlier and misclassified gene are determined, outlier is rejected before off-line modeling, by where misclassified gene from current division
Sub-period takes out and puts back to sub-period where former sample sequence.
2. a kind of batch process monitoring data processing method according to claim 1, it is characterised in that in the step a
Input data similarity formula in improved affine propagation clustering algorithm is:
<mrow>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mo>-</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>q</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>J</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>r</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>J</mi>
</munderover>
<mo>|</mo>
<msub>
<mi>C</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>q</mi>
<mo>,</mo>
<mi>r</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>C</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>q</mi>
<mo>,</mo>
<mi>r</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<msup>
<mi>J</mi>
<mn>2</mn>
</msup>
</mfrac>
<mo>,</mo>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>K</mi>
<mi> </mi>
<mi>a</mi>
<mi>n</mi>
<mi>d</mi>
<mi> </mi>
<mi>i</mi>
<mo>&NotEqual;</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, s (i, k) is the input data similarity of improved affine propagation clustering algorithm, Ci(q, r) and Ck(q, r) represents prison
Control data any two timeslice matrix XiAnd XkCovariance matrix, J is the dimension of covariance matrix.
A kind of 3. batch process monitoring data processing method according to claim 1 or 2, it is characterised in that the step b
Middle similarity successively decrease scanning algorithm implement scanning when, the average similarity formula for scanning obtained sample every time is:
<mrow>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<msubsup>
<mi>N</mi>
<mi>m</mi>
<mi>n</mi>
</msubsup>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msubsup>
<mi>N</mi>
<mi>m</mi>
<mi>n</mi>
</msubsup>
</munderover>
<msubsup>
<mi>S</mi>
<mi>m</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, m represents period of the day from 11 p.m. to 1 a.m segment number, and n represents scanning times, and k represents sample number,Represent m-th of sub-period n-th
The average similarity of scanning,The quantity of the sample obtained in being scanned from m-th of sub-period n-th is represented,Represent m-th
The similarity of k-th of sample of sub-period,
To m-th of period of the day from 11 p.m. to 1 a.m, the distribution situation of the average similarity obtained according to scanning, it is being terminated with Current Scan step scan
Afterwards, implement tolerance scanning, i.e., tried again scanning with the step-length in the tolerance less than Current Scan step-length, it is scanned afterwards
Journey terminates, and is scanned when m-th of sub-period experienced n' scanning and finally implement tolerance altogether in addition to tolerating and scanning, average phase
Exist like degree variation tendency
<mrow>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>1</mn>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mo>></mo>
<mn>...</mn>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mn>...</mn>
<mo>></mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, n and n' refers both to scanning times and n'> n, now, two kinds of situations as shown in formula (4) be present:
For case 1, the sample in region before steady component is scanned by n-th forms, and transition portion positioned at n-th by sweeping
Retouch and tolerate scanning between region sample into;
For case 2, the sample in region before steady component is still scanned by n-th forms, and transition portion is then by positioned at n-th
It is secondary scanning (n-1) ' secondary scanning between region sample into,
<mrow>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>></mo>
<msup>
<mrow>
<mo>(</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>)</mo>
</mrow>
<mrow>
<mi>t</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msup>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>c</mi>
<mi>a</mi>
<mi>s</mi>
<mi>e</mi>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<mrow>
<mo>(</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>)</mo>
</mrow>
<mrow>
<mi>t</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msup>
<mo>.</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>c</mi>
<mi>a</mi>
<mi>s</mi>
<mi>e</mi>
<mn>2</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein,It is the average similarity of tolerance scanning,
When meeting n' βmLm≤Lmax,m, and exist
<mrow>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>1</mn>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mo>></mo>
<mn>...</mn>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>></mo>
<mn>...</mn>
<mo>></mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
I.e. when average similarity is continuous equal three times, n-th ' the secondary end of scan no longer carries out tolerance scanning, and steady component is by sweeping
Retouch radius (n'-2) βmLmWithin sample composition, the remaining sample point not scanned is considered as outlier;
When meeting n' βmLm≥Lmax,m, and exist
<mrow>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>1</mn>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mo>></mo>
<mo>...</mo>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>></mo>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>></mo>
<mo>...</mo>
<mo>></mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>></mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>></mo>
<msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>e</mi>
<mo>,</mo>
<mi>m</mi>
</mrow>
<mi>n</mi>
</msubsup>
<mo>&prime;</mo>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
I.e. when the equal situation of any average similarity does not occur in whole sweep phase, then whole sub-period can be considered as
Stable sub-period;
In formula, m is period of the day from 11 p.m. to 1 a.m segment number, βmFor step factor, βm∈ [0.5,1], LmIt is the cluster centre of m-th of sub-period to phase
The distance of the sample of cluster centre, L are only second to like degreemax,mIt is the cluster centre of m-th of sub-period to the farthest sample of the sub-period
Distance, as include the maximum radius of all samples of m-th of sub-period.
4. a kind of batch process monitoring data processing method according to claim 3, it is characterised in that in the step c
After the end of scan, scanned when having been carried out tolerance, and (n' β be presentmLm)tol< Lmax,m, then the remaining sample beyond sweep radius is tolerated
This is outlier, (n' βmLm)tolIt is the tolerance sweep radius of m-th of sub-period,
When the end of scan, shown in the average similarity such as formula (5) of sample, occur continuous equal three times and n' β be presentmLm<
Lmax,m, then the remaining sample not scanned is also considered as outlier,
Outlier constitutes threat to the Accurate Model of nominal situation down-sampled data, is removed before off-line modeling,
In the step c, carefully divided through step b period of the day from 11 p.m. to 1 a.m segment data, the sub-period that some samples occur not is their sampling
Sub-period corresponding to data sequence, these samples are referred to as misclassified gene,
The sub-period that sub-period where misclassified gene from current division is taken out and put back to where former sample sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711134478.3A CN107798353B (en) | 2017-11-16 | 2017-11-16 | Intermittent process monitoring data processing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711134478.3A CN107798353B (en) | 2017-11-16 | 2017-11-16 | Intermittent process monitoring data processing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107798353A true CN107798353A (en) | 2018-03-13 |
CN107798353B CN107798353B (en) | 2020-10-30 |
Family
ID=61535206
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711134478.3A Active CN107798353B (en) | 2017-11-16 | 2017-11-16 | Intermittent process monitoring data processing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107798353B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108732931A (en) * | 2018-05-17 | 2018-11-02 | 北京化工大学 | A kind of multi-modal batch process modeling method based on JIT-RVM |
CN110531729A (en) * | 2019-10-15 | 2019-12-03 | 南京科思倍信息科技有限公司 | A kind of storage tank operating risk intelligent recognition and operating method based on big data |
CN111811818A (en) * | 2020-06-02 | 2020-10-23 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080131237A1 (en) * | 2003-11-10 | 2008-06-05 | Van Der Meulen Peter | Batch Wafer Alignment |
CN104298213A (en) * | 2014-10-30 | 2015-01-21 | 清华大学 | Index time varying gain type iterative learning control algorithm based on reference batch |
CN105334831A (en) * | 2015-11-24 | 2016-02-17 | 江南大学 | Batch process double-dimensional online optimization method driven by data difference |
CN105955214A (en) * | 2016-05-26 | 2016-09-21 | 华东理工大学 | Batch process fault detection method based on sample timing sequence and neighborhood similarity information |
CN106354889A (en) * | 2016-11-07 | 2017-01-25 | 北京化工大学 | Batch process unequal-length time period synchronization method based on LWPT-DTW (lifting wavelet package transform-dynamic time warping) |
-
2017
- 2017-11-16 CN CN201711134478.3A patent/CN107798353B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080131237A1 (en) * | 2003-11-10 | 2008-06-05 | Van Der Meulen Peter | Batch Wafer Alignment |
CN104298213A (en) * | 2014-10-30 | 2015-01-21 | 清华大学 | Index time varying gain type iterative learning control algorithm based on reference batch |
CN105334831A (en) * | 2015-11-24 | 2016-02-17 | 江南大学 | Batch process double-dimensional online optimization method driven by data difference |
CN105955214A (en) * | 2016-05-26 | 2016-09-21 | 华东理工大学 | Batch process fault detection method based on sample timing sequence and neighborhood similarity information |
CN106354889A (en) * | 2016-11-07 | 2017-01-25 | 北京化工大学 | Batch process unequal-length time period synchronization method based on LWPT-DTW (lifting wavelet package transform-dynamic time warping) |
Non-Patent Citations (2)
Title |
---|
RUNXIA GUO ET AL.: "Phase Partition and Online Monitoring for Batch Process Based on Multiway BEAM", 《IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING》 * |
张子羿 等: "一种基于聚类方法的多阶段间歇过程监控方法", 《化工学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108732931A (en) * | 2018-05-17 | 2018-11-02 | 北京化工大学 | A kind of multi-modal batch process modeling method based on JIT-RVM |
CN110531729A (en) * | 2019-10-15 | 2019-12-03 | 南京科思倍信息科技有限公司 | A kind of storage tank operating risk intelligent recognition and operating method based on big data |
CN110531729B (en) * | 2019-10-15 | 2020-03-20 | 南京科思倍信息科技有限公司 | Storage tank operation risk intelligent identification and operation method based on big data |
CN111811818A (en) * | 2020-06-02 | 2020-10-23 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
CN111811818B (en) * | 2020-06-02 | 2022-02-01 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
Also Published As
Publication number | Publication date |
---|---|
CN107798353B (en) | 2020-10-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ge et al. | Multimode process monitoring based on Bayesian method | |
Du et al. | An adaptive support vector machine-based workpiece surface classification system using high-definition metrology | |
CN107798353A (en) | A kind of batch process monitoring data processing method | |
CN101650838A (en) | Point cloud simplification processing method based on resampling method and affine clustering algorithm | |
CN104346481A (en) | Community detection method based on dynamic synchronous model | |
CN114386466B (en) | Parallel hybrid clustering method for candidate signal mining in pulsar search | |
CN106780552A (en) | Anti-shelter target tracking based on regional area joint tracing detection study | |
CN101908213A (en) | SAR image change detection method based on quantum-inspired immune clone | |
CN105893716A (en) | Structure fracture non-probability reliability analysis method based on fractal theory | |
CN112613542B (en) | Bidirectional LSTM-based load identification method for enterprise decontamination equipment | |
Li et al. | A new approach for manufacturing forecast problems with insufficient data: the case of TFT–LCDs | |
CN105607631A (en) | Batch process weak fault model control limit establishment method and weak fault monitoring method | |
CN110910021A (en) | Method for monitoring online defects based on support vector machine | |
Decke et al. | DADO–low-cost query strategies for deep active design optimization | |
CN106295691A (en) | The feature selection of single classification SVM and parameter synchronization optimization method | |
Jiang et al. | Ensemble Data Augmentation for Imbalanced Fault Diagnosis | |
You et al. | A new robust least squares support vector machine for regression with outliers | |
CN115114985A (en) | Sensor system distributed fusion method based on set theory | |
CN115034005A (en) | Model analysis visualization method for component residual service life prediction | |
Tao et al. | A Detection Approach for Wafer Detect in Industrial Manufacturing based on YOLOv8 | |
Sagar et al. | Error evaluation on k-means and hierarchical clustering with effect of distance functions for iris dataset | |
Zhou et al. | Imbalanced data classification for defective product prediction based on industrial wireless sensor network | |
CN110046708A (en) | A kind of credit-graded approach based on unsupervised deep learning algorithm | |
Shi et al. | Key Process Protection of High Dimensional Process Data in Complex Production. | |
Tokuyama et al. | Etch Profile Prediction Model Using Convolutional Neural Network |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20231126 Address after: Room 306, Building C2, Science and Technology Park, Civil Aviation University of China, Zone C, Guangxuan Road Aviation Business Center, Dongli District, Tianjin, 300300 Patentee after: Tianjin Ruichi Aviation Technology Co.,Ltd. Address before: 300300 Tianjin city Dongli District North Road No. 2898 Patentee before: CIVIL AVIATION University OF CHINA |