CN110018670A - A kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules - Google Patents
A kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules Download PDFInfo
- Publication number
- CN110018670A CN110018670A CN201910244877.8A CN201910244877A CN110018670A CN 110018670 A CN110018670 A CN 110018670A CN 201910244877 A CN201910244877 A CN 201910244877A CN 110018670 A CN110018670 A CN 110018670A
- Authority
- CN
- China
- Prior art keywords
- data
- window
- service condition
- rule
- association rules
- 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
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
- G05B19/406—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form characterised by monitoring or safety
- G05B19/4065—Monitoring tool breakage, life or condition
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/37—Measurements
- G05B2219/37616—Use same monitoring tools to monitor tool and workpiece
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Manufacturing & Machinery (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of industrial process unusual service condition prediction techniques excavated based on dynamic association rules, can be applied to the prognostic and health management of industrial process.The present invention excavates industrial equipment operating parameter dynamic association rules in the way of sliding window, and is introduced into the prediction of industrial process unusual service condition.The time response of present invention consideration correlation rule, data length is limited using sliding window, it is proposed a kind of association rules mining algorithm excavated for operating parameter dynamic association rules two-by-two, then association rule mining result is introduced into wavelet neural network prediction, network is constantly updated with dynamic association rules, to obtain more accurate prediction result.For in engineering failure predication and health control have major application value.
Description
Technical field
The invention belongs to reliability maintenance field of engineering technology, are related to a kind of industrial mistake excavated based on dynamic association rules
Journey unusual service condition prediction technique.
Background technique
As the emergence of complication system and the demand of industrial process real-time monitoring are continuously increased, modern industrial equipment
Multiple sensors are often equipped in the process of running to be monitored its operating status.Meanwhile it may in equipment running process
There is various faults mode, a certain failure may correspond to several signs, and in the case, single-sensor information can not complete body
Existing equipment running status, the failure predication based on multi-sensor information are come into being.Failure predication based on multi-sensor information
It is intended to the operating status using comprehensive sensor information analytical equipment, to carry out more reliable device diagnostic and prediction.With
The sustainable development of sensing technology, using multiple sensors carry out equipment status monitoring, fault diagnosis and prediction have become
Development trend.
There are certain relevances between its operating parameter in equipment running process still rarely has at present in failure predication field
The work that association rule mining is combined with failure predication.Traditional association rules mining algorithm is thought, for certain an object
It is also effective when analyzing other data sets of the system that the correlation rule that system is found, which is constant,.So excavate
Correlation rule do not consider time factor, thus obtained is a kind of correlation rule of static state.And in fact, with number
According to update, may be along with the variation of system running state, consequent is the variation of data characteristic, thus correlation rule
It can also happen that variation.In industrial equipment operational process, operating status is not fixed and invariable.With data update or
When equipment running status changes, the characteristic of operating parameter can also change therewith, so that the relevance of operating parameter can also
It can be changed.In the case, if after using always from the correlation rule excavated in initial data set and instructing
Continuous decision is clearly unreasonable.
Summary of the invention
For the status of the prior art, present invention aim to address rarely have consideration in the Predicting Technique of available data driving
The case where sensing data dynamically associates proposes a kind of unit exception operating condition prediction side based on operating parameter dynamic association rules
Method updates wavelet neural network using dynamic association rules and carries out unusual service condition prediction (failure predication).
Now design of the invention is described below:
The present invention may be with equipment running status for the relevance between operating parameter in industrial equipment operational process
The problem of changing and changing introduces dynamic association rules excavation.The present invention is in the way of sliding window, to each
Data in window are associated rule digging, so that the sliding with window obtains the correlation rule of dynamic change.To obtain
The complete situation of change of parameter association, there is no settings in the association rule mining of parameter level for the mentioned method of the present invention most
Small support and confidence threshold value, but the correlation rule of all two two parameters is saved, and it is described with support and confidence level
Intensity.Then, dynamic association rules Result is introduced into the training and update of wavelet neural network, by by dynamically associating
The relevant parameter that rule digging obtains carries out unusual service condition prediction.
According to the above inventive concept, the invention proposes a kind of industrial process exception works excavated based on dynamic association rules
Condition prediction technique, the specific steps are as follows:
Step 1: it is pre- that data being carried out to the measured value time series collected of sensor in industrial process based on sliding window
Processing carries out piece-wise linearization expression, cluster, symbolism to time series, generates the transaction set for being suitable for association rule mining;
Step 2: strategy generating frequent item set being generated using two stage frequent item set, and excavates the association rule of two two parameters
Then;
Step 3: being associated rule digging using initial data set, be based on initial association rule digging as a result, training is first
Beginning wavelet-neural network model;
Step 4: based on data update carry out data window sliding, for new data window be associated Policy Updates and
Wavelet-neural network model updates;
Step 5: the Wavelet-network model based on update carries out unusual service condition prediction, and before predicting unusual service condition and occurring
Constantly update correlation rule and wavelet-neural network model.
Based on above scheme, each step can specifically use following implementation:
Preferably, the step 1 is realized especially by following sub-step:
Step 1.1: note sensor measurement sequence isN is sensor
Quantity, K are sequence length, k=1,2 ..., K;Note sliding window length is L, and each sliding distance is set as l, sliding window is remembered
For Wk, the data for including in window are
Step 1.2: time series segmentation linearisation expression, cluster, symbolism processing, tool are carried out for each data window
Body process is as follows:
To each data window WkExecute 1.2.1~1.2.5 process:
1.2.1. the initial fitting starting point of setting isTerminal isH=2;Error of fitting threshold value
For ωE;Initialize waypoint count value count=1;
1.2.2. successively to each fitting starting pointExecute step 1.1)-step 1.4):
1.1) end=start+h is calculated first;
1.2) for dataIt is fitted using least square method,
Digital simulation error E RR;
1.3) if error of fitting ERR is not more than error of fitting threshold value ωE, then 1) h=h+1, gos to step again;
1.4) if error of fitting ERR is greater than error of fitting threshold value ωE, then waypoint is saved To Pi;It protects
Deposit matching line segmentThe matching line segment is indicated by the way of triple
And it saves extremelykiIndicate the slope of the line segment,Indicate the span of the line segment on a timeline, riIndicate the segment data
Growth rate;Resetting fitting starting point start=start+h, resets h=2;Update count=count+1;
1.2.3. circulation executes 1.2.2 and terminates greater than k+L-1 until end, the line segment time series after being fittedAnd waypointThe segmentation point sequence P of compositioni;
1.2.4. with Wk-1The cluster centre of window is initial cluster center, using K-means clustering algorithm to triple sequence
ColumnIt is clustered, and distributes distinct symbols for inhomogeneous line segment and obtain symbolism sequenceWherein used in cluster process
Index based on Euclidean distance describes two lines section siAnd sjBetween similarity:
Wherein, dijIndicate line segment siAnd sjSimilarity, dijIt is smaller, then it represents that two lines section has more like variation shape
State, ωkAnd ωrIndicate weight;
1.2.5. for two operating parameter ViAnd Vj, sequence after respectively obtaining symbolismWithMerge two fortune
The segmentation point sequence P of row parameteriAnd Pj, and by the waypoint after merging to its symbolism sequenceWithReconstruct is split,
Symbolism sequence after being reconstructedWithTransaction set is constituted by it,
nij- 1 is PiAnd PjWaypoint number after merging.
Preferably, the step 2 is realized especially by following sub-step:
Step 2.1: each affairs of the transaction set obtained by step 1 are denoted asSeparately willWithIncluded in line segment class code be denoted as respectivelyWith miAnd mjRespectively
ForWithIncluded in line segment classification number;If the minimum support threshold value that frequent item set generates process is minsup1;
Step 2.2: for primary data window W1It is associated rule digging, detailed process is as follows:
2.2.1) the sensor measurement sequence data collection of all the sensors is scanned to calculate each support,
Remember that σ () indicates a certain or item collection support counting, is initially 0;Assuming thatClass code be tk, e expression i or j, t table
Show c or d;
To eachCalculate σ (tk)=σ (tk)+1;
To each tk, judgementIt is whether true, t is thought if setting upkFor frequent 1- item collection, retain
tkAnd record corresponding support counting σ (tk);
2.2.2) using obtained frequent 1- item collection tk2- item collection is constituted, and finds frequent 2- item collection, remembers c respectivelypAnd dq
Respectively pass through step 2.2.1) from former line segment class codeWithThe item of middle reservation;
For each { cp,dqExecute operation: each is present inIn { cp,dq, calculate σ ({ cp,
dq)=σ ({ cp,dq})+1;IfNot less than minsup1, then it is assumed that { cp,dqIt is frequent 2- item collection, retain { cp,
dqAnd record corresponding support counting;
2.2.3) using the frequent 2- item collection { c obtained in step 2.2.2p,dqCalculate every two operating parameter ViAnd Vj
Support in entire data set, calculating process are as follows: to every two operating parameter ViAnd VjItem collection { the V of compositioni,Vj, meter
Calculate σ ({ Vi,Vj)=sum (σ ({ cp,dq)), the correlation rule support of two parametersAnd
Calculate σ (Vi)=sum (σ (cp));σ(Vj)=sum (σ (dq));
2.2.4) for every group of { Vi,Vj, generate following correlation rule: Vj→ViAnd Vi→Vj;For each association rule
Then, calculating its confidence level is
Step 2.3: for each new data window, frequent item set generation being carried out using following procedure and generates association
Rule:
To the data window W of each k > 1kExecute step 2.3.1)~2.3.3):
2.3.1) to each affairs for skidding off windowIt is denoted as { ca1,da2, judge { ca1,da2It whether is frequency
Numerous item collection, if then updating σ ({ ca1,da2)=σ ({ ca1,da2})-1;
2.3.2) to each affairs for sliding into windowIt is denoted as { cb1,db2, judge { cb1,db2It whether is frequency
Numerous item collection, if so, updating σ ({ cb1,db2)=σ ({ cb1,db2)+1, if it is not, then in data window WkInterior calculating should { cb1,
db2Support counting be σ ({ cc1,dc2});
2.3.3) to each { ca1,da2, ifThen update { ca1,da2Corresponding item collection branch
Degree of holding counts, and corresponding item collection is rejected if being unsatisfactory for;To each { cb1,db2, ifThen update
{cb1,db2Corresponding item collection support counting, corresponding item collection is rejected if being unsatisfactory for;To each { cc1,dc2, ifThen update { cc1,dc2Corresponding item collection support counting, corresponding item collection is rejected if being unsatisfactory for;
2.3.4 the support and confidence level of each correlation rule) are updated using step 2.2.3) and 2.2.4).
Preferably, the step 3 is realized especially by following sub-step:
Remember that preset prediction step is lp, the group association parameter extracted in the correlation rule that is excavated by training dataset
For { V1,V2,…,Vu, u is the group association parameter number extracted;For each parameter Vi, measuring value sequence isConstruct following matrix ItrainIt is inputted for initial neural metwork training:
Wherein, ItrainIn each be classified as a trained input sample;Construction training output OtrainFor,
The training sample constructed using above-mentioned formula, is trained wavelet neural network;In netinit, utilize
Correlation rule Vi→VuCorresponding initial confidence level ωiInitial weight between network input layer and hidden layer is set, i=1,
2,…u-1。
Preferably, the step 4 is realized especially by following sub-step:
Step 4.1: association rules updating way are as follows: note sensor in equipment running process newly counts for collected c
According to forc>L;Utilize L data nearest in new collected dataCorrelation rule is updated by step 2.3, to update correlation rule Vi→VuIt is corresponding
Confidence level be denoted as
Step 4.2: model modification way are as follows: combined training data and the new model training of new collected data configuration
Sample, i.e. data set areStructural matrix ItestIt is updated for neural network defeated
Enter, constructs OtestTo export,
New training sample is constructed using above-mentioned formula, and wavelet neural network is trained;In netinit, benefit
With new confidence levelInitial weight between network input layer and hidden layer is set.
Preferably, the step 5 is realized especially by following sub-step:
Remembering that threshold value occurs for preset unusual service condition (failure) is ωp, for newest collected data, using in step 4
The model of update carries out lpStep prediction, if obtained target component predicted value is more than set threshold relative to initial normal drift value
Value, then it is assumed that unusual service condition occurs.
Preferably, before prediction process terminates, every data for updating predetermined quantity return to step 4 and are associated rule
Then with the update of model.
The present invention excavates industrial equipment operating parameter dynamic association rules in the way of sliding window, and is introduced into work
In the prediction of industry process exception operating condition.The present invention considers the time response of correlation rule, limits data length using sliding window, mentions
A kind of association rules mining algorithm excavated for operating parameter dynamic association rules two-by-two out, then by association rule mining knot
Fruit is introduced into wavelet neural network prediction, network is constantly updated with dynamic association rules, to obtain more accurate prediction result.
For in engineering failure predication and health control have major application value.
Detailed description of the invention
Fig. 1 is IDV (13) correlation rule (v13, v16, v36 → v7) confidence level change curve in embodiment;
Fig. 2 is IDV (13) correlation rule (v35, v36 → v11) confidence level change curve in embodiment;
Fig. 3 is 7 prediction result of IDV (13) variable in embodiment;
Fig. 4 is that IDV (13) variable 7 predicts error in embodiment;
Fig. 5 is 11 prediction result of IDV (13) variable in embodiment;
Fig. 6 is that IDV (13) variable 11 predicts error in embodiment.
Specific embodiment
A specific embodiment of the invention is further described now in conjunction with attached drawing.
Below the present embodiment be specifically described by Tennessee-Yi Siman (TE) process simulation data concrete operation step with
And the effect of verification method.
The sampling interval of the data set is 3 minutes, each collected change of sensor under the data set record sampling interval
Measurement.Under each service condition (the failure operation state under normal operating condition and 21 kinds of preset failures), imitate
The measurement data of true process will all generate two class data sets, i.e. training set and test set.Wherein, for the acquisition of training set
Journey is the measured value of all 52 variables obtained in the case of simulation process runs 25 small, wherein except normal operation
Outside the training set that state acquisition arrives, the acquisition of remaining 21 training set data is as a child to introduce failure in simulation process operation 1,
And only record the measurement data of subsequent 24 hours.In other words, the training set of normal operating condition has 500 observation samples,
The training set acquired under remaining 21 malfunction is 480 observation samples.In addition, for 22 test sets, data are
Simulation process runs 48 collected all variable measurements of hour institute, that is to say, that includes 960 in each test set
Sample data.It should be noted that corresponding failure is at simulation run 8 hours when emulating to 21 kinds of procedure faults
It introduces afterwards.Therefore, for the test set under 21 failure operation states, preceding 160 observation samples are normal data, after
800 observation samples are fault data.In TE process simulation model, only IDV (13) is a soft fault, therefore,
In this example, we are tested using the related data of IDV (13).
The specific implementation procedure of industrial process unusual service condition prediction technique is as follows:
Step 1: it is pre- that data being carried out to the measured value time series collected of sensor in industrial process based on sliding window
Processing carries out piece-wise linearization expression, cluster, symbolism to time series, generates the transaction set for being suitable for association rule mining.
This step is realized especially by following sub-step:
Step 1.1: note sensor measurement sequence isN is sensor
Quantity, K are sequence length, k=1,2 ..., K;Note sliding window length is L, and each sliding distance is set as l, sliding window is remembered
For Wk, the data for including in window areIt should be noted that in the present invention, i, j are as subscript
It is the number for indicating sensor, is only to indicate ordinal number as subscript, it is unrelated with sensor number.
Step 1.2: time series segmentation linearisation expression, cluster, symbolism processing, tool are carried out for each data window
Body process is as follows:
To each data window WkExecute 1.2.1~1.2.5 process:
1.2.1. the initial fitting starting point of setting isTerminal isH=2;Error of fitting threshold value
For ωE;Initialize waypoint count value count=1;
1.2.2. successively to each fitting starting pointExecute step 1.1)-step 1.4):
1.1) end=start+h is calculated first;
1.2) for dataIt is fitted using least square method,
Digital simulation error E RR;
1.3) if error of fitting ERR is not more than error of fitting threshold value ωE, then 1) h=h+1, gos to step again;
1.4) if error of fitting ERR is greater than error of fitting threshold value ωE, then waypoint is saved To Pi;It protects
Deposit matching line segmentThe matching line segment is indicated by the way of triple
And it saves extremelykiIndicate the slope of the line segment,Indicate the span of the line segment on a timeline, riIndicate the segment data
Growth rate;Resetting fitting starting point start=start+h, resetting h=2 (resetting fitting starting point and h);Update count=count
+1;
1.2.3. circulation executes 1.2.2 and terminates greater than k+L-1 until end, the line segment after obtaining least square method fitting
Time series(wherein there is a plurality of matching line segment) and waypointThe segmentation point sequence of composition
Pi;
1.2.4. with Wk-1The cluster centre of window is initial cluster center, using K-means clustering algorithm to triple sequence
ColumnIt is clustered, and distributes distinct symbols for inhomogeneous line segment and obtain symbolism sequenceWherein used in cluster process
Index based on Euclidean distance describes two lines section siAnd sjBetween similarity:
Wherein, dijIndicate line segment siAnd sjSimilarity, dijIt is smaller, then it represents that two lines section has more like variation shape
State, ωkAnd ωrIndicate weight;
1.2.5. for two operating parameter ViAnd Vj, sequence after respectively obtaining symbolismWithMerge two fortune
The segmentation point sequence P of row parameteriAnd Pj, and by the waypoint after merging to its symbolism sequenceWithReconstruct is split,
Symbolism sequence after being reconstructedWithTransaction set is constituted by it,
nij- 1 is PiAnd PjWaypoint number after merging.
Step 2: strategy generating frequent item set being generated using two stage frequent item set, and excavates the association rule of two two parameters
Then.This step is realized especially by following sub-step:
Step 2.1: each affairs of the transaction set obtained by step 1 are denoted asSeparately willWithIncluded in line segment class code be denoted as respectivelyWith miAnd mjRespectively
ForWithIncluded in line segment classification number;If the minimum support threshold value that frequent item set generates process is minsup1;?
In this example, setting minimum support threshold value is 0.2;
Step 2.2: for primary data window W1It is associated rule digging, detailed process is as follows:
2.2.1) the sensor measurement sequence data collection of all the sensors is scanned to calculate each support,
Remember that σ () indicates a certain or item collection support counting, is initially 0;Assuming thatClass code be tk, e expression i or j, t table
Show c or d (i.e. ckOr dk);
To eachIt updates and calculates σ (tk)=σ (tk)+1;
To each tk, judgementIt is whether true, t is thought if setting upkFor frequent 1- item collection, retain
tkAnd record corresponding support counting σ (tk);
2.2.2) using obtained frequent 1- item collection tk2- item collection is constituted, and finds frequent 2- item collection, remembers c respectivelypAnd dq
Respectively pass through step 2.2.1) from former line segment class codeWithThe item of middle reservation;
For each { cp,dqExecute operation: each is present inIn { cp,dq, calculate σ ({ cp,
dq)=σ ({ cp,dq})+1;IfNot less than minsup1, then it is assumed that { cp,dqIt is frequent 2- item collection, retain { cp,
dqAnd record corresponding support counting;
2.2.3) using the frequent 2- item collection { c obtained in step 2.2.2p,dqCalculate every two operating parameter ViAnd Vj
Support in entire data set, calculating process are as follows: to every two operating parameter ViAnd VjItem collection { the V of compositioni,Vj, meter
Calculate σ ({ Vi,Vj)=sum (σ ({ cp,dq)), the correlation rule support of two parametersAnd
Calculate σ (Vi)=sum (σ (cp));σ(Vj)=sum (σ (dq));
2.2.4) for every group of { Vi,Vj, generate following correlation rule: Vj→ViAnd Vi→Vj;For each association rule
Then, calculating its confidence level is
In this example, for IDV (13), initial correlation rule is excavated by training set.Support threshold, which is arranged, is
0.2, confidence threshold value 0.7 obtains several groups of different correlation rules, and result can be shown in Table 1, wherein include two group association parameters,
That is (variable 13, variable 16,36 → variable of variable 7) and (variable 35,36 → variable of variable 11).
Step 2.3: for each new data window, frequent item set generation being carried out using following procedure and generates association
Rule.In this example, the sliding window length that setting dynamic association rules excavate is 100, and 1 data point of sliding every time.
To the data window W of each k > 1kExecute step 2.3.1)~2.3.3):
2.3.1) to each affairs for skidding off windowIt is denoted as { ca1,da2, judge { ca1,da2It whether is frequency
Numerous item collection, if then updating σ ({ ca1,da2)=σ ({ ca1,da2})-1;
2.3.2) to each affairs for sliding into windowIt is denoted as { cb1,db2, judge { cb1,db2It whether is frequent
Item collection, if so, updating σ ({ cb1,db2)=σ ({ cb1,db2)+1, if it is not, then in data window WkInterior calculating should { cb1,db2}
Support counting be σ ({ cc1,dc2});
2.3.3) to each { ca1,da2, ifThen update { ca1,da2Corresponding item collection branch
Degree of holding counts, and corresponding item collection is rejected if being unsatisfactory for;To each { cb1,db2, ifThen update
{cb1,db2Corresponding item collection support counting, corresponding item collection is rejected if being unsatisfactory for;To each { cc1,dc2, ifThen update { cc1,dc2Corresponding item collection support counting, corresponding item collection is rejected if being unsatisfactory for;
2.3.4 the support and confidence level of each correlation rule) are updated using step 2.2.3) and 2.2.4).
Using the test set of (13) IDV, since the 100th point, 1 sampled data of every update, this example is just with including this
Preceding 100 points of a point carry out Mining Association Rules, and obtained confidence level situation of change is as depicted in figs. 1 and 2.As seen from the figure, it closes
Join regular (confidence level) as the update of data is constantly changing, but for this two groups of parameters of IDV (13), numerical value is protected
It holds in a higher level.
Step 3: being associated rule digging using initial data set, be based on initial association rule digging as a result, training is first
Beginning wavelet-neural network model.This step is realized especially by following sub-step:
Remember that preset prediction step is lp, in this case, it is 10.It is extracted in the correlation rule excavated by training dataset
A group association parameter be { V1,V2,…,Vu, u is the group association parameter number extracted;For each parameter Vi, survey
Magnitude sequence isConstruct following matrix ItrainFor initial neural metwork training
Input:
Wherein, ItrainIn each be classified as a trained input sample;Construction training output OtrainFor,
The training sample constructed using above-mentioned formula, is trained wavelet neural network;In netinit, utilize
Correlation rule Vi→VuCorresponding initial confidence level ωiInitial weight between network input layer and hidden layer is set, i=1,
2,…u-1。
Step 4: based on data update carry out data window sliding, for new data window be associated Policy Updates and
Wavelet-neural network model updates.This step is realized especially by following sub-step:
Step 4.1: association rules updating way are as follows: note sensor in equipment running process newly counts for collected c
According to forc>L;Utilize L data nearest in new collected dataCorrelation rule is updated by step 2.3, to update correlation rule Vi→VuIt is corresponding
Confidence level be denoted asIt should be noted that wherein c need to be greater than preset sliding window length L.
That is, just starting the update for being associated rule only when equipment brings into operation and collected data length is greater than L.
Step 4.2: model modification way are as follows: combined training data and the new model training of new collected data configuration
Sample, i.e. data set areStructural matrix ItestIt is updated for neural network defeated
Enter, constructs OtestTo export,
New training sample is constructed using above-mentioned formula, and wavelet neural network is trained;In netinit, benefit
With new confidence levelInitial weight between network input layer and hidden layer is set.In this example, before using test set
300 data, 10 data of every update, are just updated correlation rule and model.
Step 5: the Wavelet-network model based on update carries out unusual service condition prediction, and before predicting unusual service condition and occurring
Constantly update correlation rule and wavelet-neural network model.This step is realized especially by following sub-step:
Remembering that threshold value occurs for preset unusual service condition (failure) is ωp, the threshold value that unusual service condition (failure) occurs in this example is
ωp=0.015.For newest collected data, l is carried out using the model updated in step 4pStep prediction, if obtained target
Parameter prediction value is more than set threshold value relative to initial normal drift value, then it is assumed that unusual service condition occurs.In prediction process knot
Shu Qian, U data of every update then return to the update that step 4 is associated rule and model, and the numerical value of U is adjustable, U in this example
It is 10.
1 correlation rule of table
Regular preceding paragraph | Rule is consequent | Confidence level |
Variable 13 | Variable 7 | 0.7527 |
Variable 16 | Variable 7 | 0.7446 |
Variable 36 | Variable 7 | 0.7017 |
Variable 35 | Variable 11 | 0.7513 |
Variable 36 | Variable 11 | 0.7390 |
Table 2 always predicts error rate
Introduce dynamic association rules | Introduce correlation rule | It is not introduced into correlation rule | |
Variable 7 | 0.7602 | 1.0247 | 1.7423 |
Variable 11 | 0.8075 | 0.8309 | 1.1884 |
This example is to verify the validity of this chapter method therefor, by the prediction result for introducing dynamic association rules and is not introduced into pass
The case where connection rule and introducing static association rule, compares, and Fig. 3-Fig. 6 is shown in the comparison of prediction result and prediction error.?
In Fig. 3 and Fig. 5, erecting solid line is the considered repealed moment under preset failure threshold, and perpendicular chain-dotted line is to introduce to dynamically associate rule
Failure moment predicted value then, perpendicular dotted line are the failure moment predicted value for introducing static association rule, and perpendicular dotted line is to be not introduced into pass
Join the failure moment predicted value of rule.In addition, in order to which preferably quantized prediction error, this example calculates two parameter predictions in difference
Under the conditions of total error rate, as shown in table 2.
From the point of view of prediction result (Fig. 3 and Fig. 5), compared with other two kinds of situations, introducing dynamic association rules can be obtained
More accurate unusual service condition prediction result can from prediction curve especially compared with the case where being not introduced into correlation rule
Dynamic association rules are introduced out to have a clear superiority.It should be noted that introducing static association rule in Fig. 5 and not drawing
Enter predicted value and true value in the case where correlation rule and biggish deviation occur, this is since the failure of the parameter is real in fact
Actual value is exactly in a certain extreme point, if therefore there is subtle prediction error here will cause unusual service condition prediction result to go out
Existing relatively large deviation.Parameter is monitored indeed, it is possible to which multiple and different out-of-limit threshold values is arranged, to avoid such situation band
The risk come.In addition, compared with the case where introducing static association rule, drawing from the point of view of the angle (Fig. 4 and Fig. 6) of prediction error
Prediction error can be reduced to a certain extent by entering dynamic association rules, from table 2 it can be seen that variable 7 is introduced dynamic and closed
The promotion effect for joining rule is more obvious, and for variable 11, due to correlation rule variation and little, so its effect is not shown
It writes;Compared with the case where being not introduced into correlation rule, it can be seen that introducing dynamic association rules can obviously drop from Fig. 4 and Fig. 6
Low prediction error, by table 2 can also be more intuitive find out introduce dynamic association rules compared to be not introduced into correlation rule have it is bright
Aobvious advantage.
Claims (7)
1. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules, which is characterized in that specific steps
It is as follows:
Step 1: data prediction is carried out to the measured value time series collected of sensor in industrial process based on sliding window,
Piece-wise linearization expression, cluster, symbolism are carried out to time series, generate the transaction set for being suitable for association rule mining;
Step 2: strategy generating frequent item set being generated using two stage frequent item set, and excavates the correlation rule of two two parameters;
Step 3: being associated rule digging using initial data set, be based on initial association rule digging as a result, training is initial small
Wave neural network model;
Step 4: updating the sliding for carrying out data window based on data, Policy Updates and small echo are associated for new data window
Neural network model updates;
Step 5: the Wavelet-network model based on update carries out unusual service condition prediction, and before predicting unusual service condition and occurring constantly
Update correlation rule and wavelet-neural network model.
2. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 1,
It is characterized by: the step 1 is realized especially by following sub-step:
Step 1.1: note sensor measurement sequence isN is sensor number
Amount, K is sequence length, k=1,2 ..., K;Note sliding window length is L, and each sliding distance is set as l, sliding window is denoted as
Wk, the data for including in window are
Step 1.2: time series segmentation linearisation expression, cluster, symbolism processing, specific mistake are carried out for each data window
Journey is as follows:
To each data window WkExecute 1.2.1~1.2.5 process:
1.2.1. the initial fitting starting point of setting isTerminal isH=2;Error of fitting threshold value is
ωE;Initialize waypoint count value count=1;
1.2.2. successively to each fitting starting pointExecute step 1.1)-step 1.4):
1.1) end=start+h is calculated first;
1.2) for dataIt is fitted, is calculated using least square method
Error of fitting ERR;
1.3) if error of fitting ERR is not more than error of fitting threshold value ωE, then 1) h=h+1, gos to step again;
1.4) if error of fitting ERR is greater than error of fitting threshold value ωE, then waypoint is saved To Pi;It saves
Matching line segmentThe matching line segment is indicated by the way of tripleAnd
It saves extremelykiIndicate the slope of the line segment,Indicate the span of the line segment on a timeline, riIndicate the increasing of the segment data
Long rate;Resetting fitting starting point start=start+h, resets h=2;Update count=count+1;
1.2.3. circulation executes 1.2.2 and terminates greater than k+L-1 until end, the line segment time series after being fittedAnd waypointThe segmentation point sequence P of compositioni;
1.2.4. with Wk-1The cluster centre of window is initial cluster center, using K-means clustering algorithm to triad sequence
It is clustered, and distributes distinct symbols for inhomogeneous line segment and obtain symbolism sequenceIt wherein uses and is based in cluster process
The index of Euclidean distance describes two lines section siAnd sjBetween similarity:
Wherein, dijIndicate line segment siAnd sjSimilarity, dijIt is smaller, then it represents that two lines section has more like change shape,
ωkAnd ωrIndicate weight;
1.2.5. for two operating parameter ViAnd Vj, sequence after respectively obtaining symbolismWithMerge two operation ginsengs
Several segmentation point sequence PiAnd Pj, and by the waypoint after merging to its symbolism sequenceWithIt is split reconstruct, is obtained
Symbolism sequence after reconstructWithTransaction set, n are made of itij-1
For PiAnd PjWaypoint number after merging.
3. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 2,
It is characterized by: the step 2 is realized especially by following sub-step:
Step 2.1: each affairs of the transaction set obtained by step 1 are denoted asSeparately willWithIncluded in line segment class code be denoted as respectivelyWithmiAnd mjRespectivelyWithIncluded in line segment classification number;If the minimum support threshold value that frequent item set generates process is minsup1;
Step 2.2: for primary data window W1It is associated rule digging, detailed process is as follows:
2.2.1) the sensor measurement sequence data collection of all the sensors is scanned to calculate each support, remembers σ
() indicates a certain or item collection support counting, is initially 0;Assuming thatClass code be tk, e expression i or j, t indicate c
Or d;
To eachCalculate σ (tk)=σ (tk)+1;
To each tk, judgementIt is whether true, t is thought if setting upkFor frequent 1- item collection, retain tkAnd remember
Record corresponding support counting σ (tk);
2.2.2) using obtained frequent 1- item collection tk2- item collection is constituted, and finds frequent 2- item collection, remembers c respectivelypAnd dqRespectively
To pass through step 2.2.1) from former line segment class codeWithThe item of middle reservation;
For each { cp,dqExecute operation: each is present inIn { cp,dq, calculate σ ({ cp,dq)=σ
({cp,dq})+1;IfNot less than minsup1, then it is assumed that { cp,dqIt is frequent 2- item collection, retain { cp,dqAnd remember
Record corresponding support counting;
2.2.3) using the frequent 2- item collection { c obtained in step 2.2.2p,dqCalculate every two operating parameter ViAnd VjEntire
Support in data set, calculating process are as follows: to every two operating parameter ViAnd VjItem collection { the V of compositioni,Vj, calculate σ
({Vi,Vj)=sum (σ ({ cp,dq)), the correlation rule support of two parametersAnd it counts
Calculate σ (Vi)=sum (σ (cp));σ(Vj)=sum (σ (dq));
2.2.4) for every group of { Vi,Vj, generate following correlation rule: Vj→ViAnd Vi→Vj;For each correlation rule, meter
Calculating its confidence level is
Step 2.3: for each new data window, frequent item set generation being carried out using following procedure and generates association rule
Then:
To the data window W of each k > 1kExecute step 2.3.1)~2.3.3):
2.3.1) to each affairs for skidding off windowIt is denoted as { ca1,da2, judge { ca1,da2It whether is frequent episode
Collection, if then updating σ ({ ca1,da2)=σ ({ ca1,da2})-1;
2.3.2) to each affairs for sliding into windowIt is denoted as { cb1,db2, judge { cb1,db2It whether is frequent episode
Collection, if so, updating σ ({ cb1,db2)=σ ({ cb1,db2)+1, if it is not, then in data window WkInterior calculating should { cb1,db2?
Support counting is σ ({ cc1,dc2});
2.3.3) to each { ca1,da2, ifThen update { ca1,da2Corresponding item collection support meter
Number, rejects corresponding item collection if being unsatisfactory for;To each { cb1,db2, ifThen update { cb1,db2Phase
The support counting for answering item collection rejects corresponding item collection if being unsatisfactory for;To each { cc1,dc2, if
Then update { cc1,dc2Corresponding item collection support counting, corresponding item collection is rejected if being unsatisfactory for;
2.3.4 the support and confidence level of each correlation rule) are updated using step 2.2.3) and 2.2.4).
4. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 3,
It is characterized by: the step 3 is realized especially by following sub-step:
Remember that preset prediction step is lp, the group association parameter extracted in the correlation rule that is excavated by training dataset is
{V1,V2,…,Vu, u is the group association parameter number extracted;For each parameter Vi, measuring value sequence isConstruct following matrix ItrainIt is inputted for initial neural metwork training:
Wherein, ItrainIn each be classified as a trained input sample;Construction training output OtrainFor,
The training sample constructed using above-mentioned formula, is trained wavelet neural network;In netinit, association is utilized
Regular Vi→VuCorresponding initial confidence level ωiInitial weight between network input layer and hidden layer, i=1,2 ... u- are set
1。
5. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 4,
It is characterized by: the step 4 is realized especially by following sub-step:
Step 4.1: association rules updating way are as follows: collected c new data is sensor note in equipment running processc>L;Utilize L data nearest in new collected dataCorrelation rule is updated by step 2.3, to update correlation rule Vi→VuIt is corresponding
Confidence level be denoted as
Step 4.2: model modification way are as follows: combined training data and the new model training sample of new collected data configuration
This, i.e., data set isStructural matrix ItestFor neural network update input,
Construct OtestTo export,
New training sample is constructed using above-mentioned formula, and wavelet neural network is trained;In netinit, using new
Confidence levelInitial weight between network input layer and hidden layer is set.
6. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 5,
It is characterized by: the step 5 is realized especially by following sub-step:
Remembering that threshold value occurs for preset unusual service condition is ωp, for newest collected data, using the model updated in step 4 into
Row lpStep prediction, if obtained target component predicted value is more than set threshold value relative to initial normal drift value, then it is assumed that different
Normal operating condition occurs.
7. a kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules according to claim 6,
It is characterized by: before prediction process terminates, every data for updating predetermined quantity return to step 4 and are associated rule and mould
The update of type.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910244877.8A CN110018670B (en) | 2019-03-28 | 2019-03-28 | Industrial process abnormal working condition prediction method based on dynamic association rule mining |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910244877.8A CN110018670B (en) | 2019-03-28 | 2019-03-28 | Industrial process abnormal working condition prediction method based on dynamic association rule mining |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110018670A true CN110018670A (en) | 2019-07-16 |
CN110018670B CN110018670B (en) | 2020-07-10 |
Family
ID=67190170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910244877.8A Active CN110018670B (en) | 2019-03-28 | 2019-03-28 | Industrial process abnormal working condition prediction method based on dynamic association rule mining |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110018670B (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111060652A (en) * | 2019-11-21 | 2020-04-24 | 西南交通大学 | Method for predicting concentration of dissolved gas in transformer oil based on long-term and short-term memory network |
CN111858712A (en) * | 2020-07-20 | 2020-10-30 | 上海仪电(集团)有限公司中央研究院 | In-situ water quality inspection data time-space analysis and anomaly detection method and system |
CN112016097A (en) * | 2020-08-28 | 2020-12-01 | 重庆文理学院 | Method for predicting time of network security vulnerability being utilized |
CN112329828A (en) * | 2020-10-26 | 2021-02-05 | 北京旋极信息技术股份有限公司 | Fault correlation analysis method and device |
CN112380274A (en) * | 2020-11-16 | 2021-02-19 | 北京航空航天大学 | Control process-oriented anomaly detection system |
CN112632151A (en) * | 2020-12-25 | 2021-04-09 | 南京邮电大学 | Mobile object infection behavior mining method based on binary filtering |
CN112749370A (en) * | 2021-04-06 | 2021-05-04 | 广东际洲科技股份有限公司 | Fault tracking and positioning method and system based on Internet of things |
CN112801426A (en) * | 2021-04-06 | 2021-05-14 | 浙江浙能技术研究院有限公司 | Industrial process fault fusion prediction method based on correlation parameter mining |
CN112800686A (en) * | 2021-03-29 | 2021-05-14 | 国网江西省电力有限公司电力科学研究院 | Transformer DGA online monitoring data abnormal mode judgment method |
CN112862347A (en) * | 2021-03-02 | 2021-05-28 | 同济大学 | Equipment abnormity monitoring method and system based on federal learning, storage medium and terminal |
CN113297272A (en) * | 2021-05-30 | 2021-08-24 | 福建中锐网络股份有限公司 | Bridge monitoring data association rule mining and health early warning method and system |
CN113515554A (en) * | 2020-04-09 | 2021-10-19 | 华晨宝马汽车有限公司 | Anomaly detection method and system for irregularly sampled time series |
CN113591994A (en) * | 2021-08-03 | 2021-11-02 | 北京邮电大学 | Terminal behavior prediction method based on automatic labeling |
CN115689071A (en) * | 2023-01-03 | 2023-02-03 | 南京工大金泓能源科技有限公司 | Equipment fault fusion prediction method and system based on correlation parameter mining |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050177482A1 (en) * | 2004-02-10 | 2005-08-11 | V.S. Subrahmanian | Method and system for optimal data diagnosis |
CN101408754A (en) * | 2008-10-30 | 2009-04-15 | 中山大学 | Intelligent house optimizing system based on data excavation |
CN101667197A (en) * | 2009-09-18 | 2010-03-10 | 浙江大学 | Mining method of data stream association rules based on sliding window |
CN201898519U (en) * | 2010-09-01 | 2011-07-13 | 燕山大学 | Equipment maintenance early-warning device with risk control |
CN102818642A (en) * | 2012-07-18 | 2012-12-12 | 辽宁省海洋水产科学研究院 | Disease pre-warning system for stichopus japonicus |
CN103440248A (en) * | 2013-07-22 | 2013-12-11 | 西南交通大学 | Network video event mining framework based on dynamic association rules |
CN103676645A (en) * | 2013-12-11 | 2014-03-26 | 广东电网公司电力科学研究院 | Mining method for association rules in time series data flows |
CN104636479A (en) * | 2015-02-15 | 2015-05-20 | 西安电子科技大学 | Industrial big data driven total completion time prediction method |
CN108873859A (en) * | 2018-05-31 | 2018-11-23 | 浙江工业大学 | Based on the bridge-type grab ship unloader fault prediction model method for improving correlation rule |
-
2019
- 2019-03-28 CN CN201910244877.8A patent/CN110018670B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050177482A1 (en) * | 2004-02-10 | 2005-08-11 | V.S. Subrahmanian | Method and system for optimal data diagnosis |
CN101408754A (en) * | 2008-10-30 | 2009-04-15 | 中山大学 | Intelligent house optimizing system based on data excavation |
CN101667197A (en) * | 2009-09-18 | 2010-03-10 | 浙江大学 | Mining method of data stream association rules based on sliding window |
CN201898519U (en) * | 2010-09-01 | 2011-07-13 | 燕山大学 | Equipment maintenance early-warning device with risk control |
CN102818642A (en) * | 2012-07-18 | 2012-12-12 | 辽宁省海洋水产科学研究院 | Disease pre-warning system for stichopus japonicus |
CN103440248A (en) * | 2013-07-22 | 2013-12-11 | 西南交通大学 | Network video event mining framework based on dynamic association rules |
CN103676645A (en) * | 2013-12-11 | 2014-03-26 | 广东电网公司电力科学研究院 | Mining method for association rules in time series data flows |
CN104636479A (en) * | 2015-02-15 | 2015-05-20 | 西安电子科技大学 | Industrial big data driven total completion time prediction method |
CN108873859A (en) * | 2018-05-31 | 2018-11-23 | 浙江工业大学 | Based on the bridge-type grab ship unloader fault prediction model method for improving correlation rule |
Non-Patent Citations (5)
Title |
---|
刘伯红等: "一种面向时空数据的关联规则更新算法", 《计算机与数字工程》 * |
张建明等: "基于关联规则的故障诊断方法及研究", 《化工自动化及仪表》 * |
汪成亮等: "基于神经网络的铝电解混合控制模型研究", 《计算机应用研究》 * |
熊腾飞: "基于滑动窗口的多元时间序列数据动态关联规则挖掘", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
鲍文: "基于关联规则的火电厂传感器故障检测", 《中国电机工程学报》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111060652A (en) * | 2019-11-21 | 2020-04-24 | 西南交通大学 | Method for predicting concentration of dissolved gas in transformer oil based on long-term and short-term memory network |
CN113515554A (en) * | 2020-04-09 | 2021-10-19 | 华晨宝马汽车有限公司 | Anomaly detection method and system for irregularly sampled time series |
CN111858712A (en) * | 2020-07-20 | 2020-10-30 | 上海仪电(集团)有限公司中央研究院 | In-situ water quality inspection data time-space analysis and anomaly detection method and system |
CN112016097A (en) * | 2020-08-28 | 2020-12-01 | 重庆文理学院 | Method for predicting time of network security vulnerability being utilized |
CN112016097B (en) * | 2020-08-28 | 2024-02-27 | 深圳泓越信息科技有限公司 | Method for predicting network security vulnerability time to be utilized |
CN112329828A (en) * | 2020-10-26 | 2021-02-05 | 北京旋极信息技术股份有限公司 | Fault correlation analysis method and device |
CN112380274A (en) * | 2020-11-16 | 2021-02-19 | 北京航空航天大学 | Control process-oriented anomaly detection system |
CN112380274B (en) * | 2020-11-16 | 2023-08-22 | 北京航空航天大学 | Abnormality detection method for control process |
CN112632151A (en) * | 2020-12-25 | 2021-04-09 | 南京邮电大学 | Mobile object infection behavior mining method based on binary filtering |
CN112632151B (en) * | 2020-12-25 | 2023-02-10 | 南京邮电大学 | Mobile object infection behavior mining method based on binary filtering |
CN112862347A (en) * | 2021-03-02 | 2021-05-28 | 同济大学 | Equipment abnormity monitoring method and system based on federal learning, storage medium and terminal |
CN112800686A (en) * | 2021-03-29 | 2021-05-14 | 国网江西省电力有限公司电力科学研究院 | Transformer DGA online monitoring data abnormal mode judgment method |
CN112749370B (en) * | 2021-04-06 | 2021-07-02 | 广东际洲科技股份有限公司 | Fault tracking and positioning method and system based on Internet of things |
CN112801426A (en) * | 2021-04-06 | 2021-05-14 | 浙江浙能技术研究院有限公司 | Industrial process fault fusion prediction method based on correlation parameter mining |
CN112749370A (en) * | 2021-04-06 | 2021-05-04 | 广东际洲科技股份有限公司 | Fault tracking and positioning method and system based on Internet of things |
CN113297272B (en) * | 2021-05-30 | 2023-01-06 | 福建中锐网络股份有限公司 | Bridge monitoring data association rule mining and health early warning method and system |
CN113297272A (en) * | 2021-05-30 | 2021-08-24 | 福建中锐网络股份有限公司 | Bridge monitoring data association rule mining and health early warning method and system |
CN113591994A (en) * | 2021-08-03 | 2021-11-02 | 北京邮电大学 | Terminal behavior prediction method based on automatic labeling |
CN113591994B (en) * | 2021-08-03 | 2023-06-06 | 北京邮电大学 | Terminal behavior prediction method based on automatic labeling |
CN115689071A (en) * | 2023-01-03 | 2023-02-03 | 南京工大金泓能源科技有限公司 | Equipment fault fusion prediction method and system based on correlation parameter mining |
Also Published As
Publication number | Publication date |
---|---|
CN110018670B (en) | 2020-07-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110018670A (en) | A kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules | |
CN110008253B (en) | Industrial data association rule mining and abnormal working condition prediction method | |
CN102789545B (en) | Based on the Forecasting Methodology of the turbine engine residual life of degradation model coupling | |
JP6661839B1 (en) | Time series data diagnosis device, additional learning method, and program | |
CN102042848B (en) | Prediction method of multi-functional parameter accelerated degradation testing product life based on multivariate hybrid time sequence analysis | |
CN108985380B (en) | Point switch fault identification method based on cluster integration | |
CN105205288A (en) | Mode evolution-based forecasting method for satellite long-term on-orbit running state | |
CN105607631B (en) | The weak fault model control limit method for building up of batch process and weak fault monitoring method | |
CN114167838B (en) | Multi-scale health assessment and fault prediction method for servo system | |
CN104915434A (en) | Multi-dimensional time sequence classification method based on mahalanobis distance DTW | |
CN113344134A (en) | Data acquisition abnormity detection method and system for low-voltage power distribution monitoring terminal | |
CN116448419A (en) | Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization | |
CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
CN111680407B (en) | Satellite health assessment method based on Gaussian mixture model | |
CN114492642A (en) | Mechanical fault online diagnosis method for multi-scale element depth residual shrinkage network | |
CN115186762A (en) | Engine abnormity detection method and system based on DTW-KNN algorithm | |
CN111693726B (en) | Ventilation system fault diagnosis wind speed sensor arrangement method based on neighborhood rough set | |
CN110134676B (en) | Method for monitoring data quality of sensor | |
CN110308713A (en) | A kind of industrial process failure identification variables method based on k neighbour reconstruct | |
CN113092083A (en) | Machine pump fault diagnosis method and device based on fractal dimension and neural network | |
CN113687411B (en) | Earth stress azimuth prediction method based on microseism | |
CN117235489A (en) | Graph variation self-coding feature extraction method for multi-source monitoring data of transmission system | |
Yen et al. | Winner take all experts network for sensor validation | |
CN111639621A (en) | Method for diagnosing fault by sensor signal | |
CN115636103B (en) | Working condition separation method, device and equipment of PIU subsystem and storage medium |
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 |