CN110008565A - A kind of industrial process unusual service condition prediction technique based on operating parameter association analysis - Google Patents

A kind of industrial process unusual service condition prediction technique based on operating parameter association analysis Download PDF

Info

Publication number
CN110008565A
CN110008565A CN201910244872.5A CN201910244872A CN110008565A CN 110008565 A CN110008565 A CN 110008565A CN 201910244872 A CN201910244872 A CN 201910244872A CN 110008565 A CN110008565 A CN 110008565A
Authority
CN
China
Prior art keywords
parameter
relevance
data
prediction
value
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
Application number
CN201910244872.5A
Other languages
Chinese (zh)
Other versions
CN110008565B (en
Inventor
徐正国
王豆
陈积明
程鹏
孙优贤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201910244872.5A priority Critical patent/CN110008565B/en
Publication of CN110008565A publication Critical patent/CN110008565A/en
Application granted granted Critical
Publication of CN110008565B publication Critical patent/CN110008565B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The industrial process unusual service condition prediction technique based on operating parameter association analysis that the invention discloses a kind of, can be applied to the prognostic and health management of industrial process.The present invention starts with from the relevance between industrial process operating parameter, and the association analysis based on operating parameter carries out unusual service condition prediction.In one-parameter forecast period, the present invention predicts each operating parameter by exponential smoothing method according to existing sensing data.In the association analysis stage, the present invention calculates operating parameter relevance by known each operational parameter value and parameter prediction value, and wherein parameter association is indicated by a series of similitude of indexs of representation parameter curve.In the relevance trend prediction stage, the present invention constructs multivariate autoregressive model and predicts parameter association.Method proposed by the present invention accounts for operating parameter relevance, and the prediction result that can be obtained more complete device exception information and more shift to an earlier date is of practical significance for the failure predication of industrial equipment.

Description

A kind of industrial process unusual service condition prediction technique based on operating parameter association analysis
Technical field
The invention belongs to reliability engineering technique fields, are related to a kind of industrial process based on operating parameter association analysis 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.
For multiple sensors in equipment running process, the operating parameter represented is not self-existent, equipment Operational process in the variation of each operating parameter be essentially all reaction to equipment current operating conditions.Industrial equipment just Under normal operating status, operating parameter is usually relatively more steady, maintains metastable level, thus the relevance of operating parameter It is more stable.But when unusual service condition occurs, each operating parameter is different abnormal response, so as to cause operating parameter Relevance change, relevance and relevance variation tendency necessarily imply the information of unit exception or even failure.
Summary of the invention
Situation in view of the prior art, the purpose of the present invention is for equipment, each operating parameter has pass in the process of running The problem of connection property, the state of equipment running process is judged by the relevance of operating parameter, by being closed to operating parameter The prediction of connection property variation tendency is predicted to carry out the unusual service condition of equipment.
Now design of the invention is described below:
The present invention proposes a kind of industrial process unusual service condition prediction technique based on operating parameter relevance variation tendency, from Relevance between industrial process operating parameter is started with, and it is pre- to carry out unusual service condition by the trend analysis to operating parameter relevance It surveys.In one-parameter forecast period, the present invention according to existing sensing data by exponential smoothing method to each operating parameter into Row prediction.In the association analysis stage, the present invention calculates operation ginseng by known each operational parameter value and parameter prediction value Number relevance, wherein parameter association is indicated by a series of similitude of indexs of representation parameter curve.It is pre- in relevance trend Survey stage, the present invention construct multivariate autoregressive model and predict parameter association.Method proposed by the present invention joins operation Number relevance accounts for, the prediction result that can be obtained more complete device exception information and more shift to an earlier date.
According to the above inventive concept, the present invention proposes a kind of industrial process exception work based on operating parameter association analysis Condition prediction technique, includes the following steps:
Step 1: using Holt exponential smoothing model, each sensor of industrial process measurement value sequence collected is determined Step-ahead prediction；
Step 2: in a manner of sliding window, the triple characterized by industrial process operating parameter variation tendency is constructed, Measurement value sequence in data window is indicated, and is calculated in the window and is appointed by the relevance index based on Euclidean distance The relevance for two operating parameters of anticipating；
Step 3: according to the relevance data configuration relevance prediction model for having calculated acquisition, the relevance predicts mould Type is multivariate autoregressive model, and estimates model parameter by partial least squares algorithm；
Step 4: for newly obtaining data, the prediction of unit exception operating condition is carried out according to relevance prediction model, and pre- It measures before unusual service condition occurs and constantly updates relevance prediction model and model parameter.
Based on above scheme, each step can specifically use following implementation:
Preferably, step 1 is specific as follows:
Step 1.1: for the industrial equipment with multiple sensors, note number of sensors is N, is run when equipment is in Cheng Zhong constantly collects the operational parameter value i.e. sensing data of characterization equipment running status, and each sensor measurement sequence is remembered ForWherein K indicates sequence length,Indicate that sensor i is adopted at k-th The measured value of sample moment point；
Step 1.2: for sensor i, value sequence being measured to it using Holt exponential smoothing model and is predicted, given and survey MagnitudeIts smooth value can calculate according to the following formula:
Wherein,It indicatesSmooth value；It is the linear increase factor, represents smoothed out trend；α and β is Number, value range is (0,1)；The primary condition of Holt model is as follows:
Step 1.3: after the smooth value and the linear increase factor for obtaining measurement data, predicted using the result, Predicted valueAre as follows:
Wherein l represents prediction step, and τ is the sampling interval of device sensor signal.
Preferably, step 2 includes following sub-step:
Step 2.1: operating parameter association analysis is divided into time series segmentation fitting, triple indicates and relevance calculates Three phases；In the time series segmentation fitting stage, the data window for being L for regular length is denoted as window Wj, operating parameter Xi, i=1,2 ..., N, the L data in the window areWherein j is the window start, corresponding Sampling instant is tj；If assuming the m data in the window in one of them periodIt can be by a line Section is fitted, for measured value thereinIts corresponding value in matching line segment isThe then error of fitting of the line segment ERR calculation formula are as follows:
When for carrying out piece-wise linearization as the window data of starting point using j, fromStart to carry out line segment fitting to it；? The process of data sectional fitting carries out according to the following steps, wherein remembering the error of fitting threshold value set as ωE:
Step (1): setting is fitted starting point and isBeing fitted terminal isH=2；Wherein, for using j as the window of starting point Mouth Wj, initial fitting starting point is
Step (2): for dataLine segment fitting is carried out by the way of linear regression, is thus obtained Obtain corresponding segment dataIts error of fitting is calculated according to the error of fitting ERR calculation formula ERR；
Step (3): if ERR≤ωE, then h=h+1 is enabled, and repeat step (2)；If ERR > ωE, then current fitting is saved eventually Point (i.e. data cut-point) resets h=2, and returns to step 1, is that new fitting starting point carries out next portion to be currently fitted terminal The fitting of divided data；
Above step (1)~(3) are repeated until window WjInterior all data line segment, that is, obtain piecewise linearity Data after change
Step 2.2: showing the stage in line segment triple table, a line segment s is described in the form of following triplej,
Wherein, kjIndicate line segment slope,Indicate the length of the line segment on a timeline, rjIndicate the increasing of the line segment numerical value Long rate, i.e., for segment dataFor window WjNew sequence after data line segmentObtaining its triad sequence representation is { s1,s2,…,sn, wherein n indicates number in the window According to the line segment quantity after segmentation；
Step 2.3: in relevance calculation stages, for two operating parameter V in equipment running processAAnd operating parameter VB, need to be split the data after line segment first: in window WjIt is interior, remember parameter VAWaypoint beParameter VBWaypoint beWherein nAAnd nBRespectively indicate operating parameter A and B Measurement data be segmented in the window after line segment number；To parameter VAAnd VBWaypoint merge, removal duplicate keys go forward side by side Row arranges from small to large, and the segmentation point sequence for obtaining two parameters isThen, according to the cut-point Sequence and triple representation obtain parameter VAAnd VBNew triad sequence isWith
Obtain window WjInterior two parameter VAAnd VBTriad sequence after, pass through the relevance index based on Euclidean distance dABTo calculate two parameter V in the windowAAnd VBRelevance:
In formula:For parameter VAThe slope of i-th line section,For parameter VBThe slope of i-th line section,For parameter VAThe The nominal growth rate of i line segment,For parameter VBThe nominal growth rate of i-th line section.
Preferably, step 3 includes following sub-step:
Step 3.1: construction relevance prediction model: setting prediction step is f, determines that model is set according to known data length Meter parameter U and M make U+f+M-1=window WjTotal number, and construct following matrix:
Wherein, { d1,d2,…,dU+f+M-1Expression parameter VAAnd VBRelevance sequence, djIndicate window WjInterior two parameters VAAnd VBRelevance index dAB；
Step 3.2: for each relevance sequence, the relevance value after f step-length is predicted using wherein U relevance value, Construct relevance prediction model:
Fp=Dpθ
Wherein, parameter θ=[θ1,θ2,…,θU]T, can be obtained by partial least squares algorithm.
Preferably, step 4 includes following sub-step:
In prediction, the new matrix of the data configuration newly obtained using sensor:
Then, using constructed relevance prediction model to relevance value carry out f step prediction, i.e.,
WhenIt is abnormal to judge that equipment occurs, wherein dnormalIt is in normal for equipment initial operating stage Parameter V when operating statusAAnd VBRelevance value, ωpIt is relevance value relative to initial normal drift value threshold value；IfThe new data reconstruction model then obtained using sensor is to update model parameter θ；
As the relevance value of certain prediction step is constantly predicted in the update of data, to predict that unit exception operating condition occurs Time.
Industrial process unusual service condition prediction technique proposed by the present invention based on operating parameter association analysis, can be used for having The Complex Industrial Systems of standby multiple sensors.The present invention starts with from the relevance between industrial process operating parameter, based on operation The association analysis of parameter carries out unusual service condition prediction.In one-parameter forecast period, the present invention is logical according to existing sensing data Exponential smoothing method is crossed to predict each operating parameter.In the association analysis stage, the present invention passes through known each operation Parameter value and parameter prediction value calculate operating parameter relevance, wherein parameter association by representation parameter curve a series of fingers Target similitude indicates.In the relevance trend prediction stage, the present invention constructs multivariate autoregressive model and carries out to parameter association Prediction.Method proposed by the present invention accounts for operating parameter relevance, can obtain more complete device exception information with And prediction result more in advance.This will provide strong data supporting to subsequent equipment health control, for high reliability Equipment maintenance and management it is especially valuable, in terms of practical engineering application have bright prospects.
Detailed description of the invention
Fig. 1 steam turbine operation measured value of parameters and prediction result；
Fig. 2 steam turbine vacuum A and other parameters relevance trend prediction result and true value compare；
Fig. 3 steam turbine exception time of origin prediction result.
Specific embodiment
A specific embodiment of the invention is further described now in conjunction with attached drawing, Some principles are chatted in detail in front It states, details are not described herein.This example illustrates tool based on the realistic case of turbine low vacuum protection chaser data with one below Body operating procedure and the validity for verifying proposed method.
The milling machine data record using milling cutter cutting metal material operation degenerative process.The steam turbine operation it is initial Operating condition is load 250MW, condenser vacuum 93kPa, and using vacuum value of condenser as instruction parameter, vacuum A is sampled from the 762nd Point starts instruction exception, when vacuum values drop to 81kPa, the steam turbine chaser.Industrial process unusual service condition prediction technique packet Include following steps:
Step 1: using Holt exponential smoothing model, each sensor of industrial process measurement value sequence collected is determined Step-ahead prediction.This step specifically includes following sub-step:
Step 1.1: for the industrial equipment with multiple sensors, note number of sensors is N, is run when equipment is in Cheng Zhong constantly collects the operational parameter value i.e. sensing data of characterization equipment running status, and each sensor measurement sequence is remembered ForWherein K indicates sequence length,Indicate that sensor i is adopted at k-th The measured value of sample moment point；
Step 1.2: for sensor i, value sequence being measured to it using Holt exponential smoothing model and is predicted, given and survey MagnitudeIts smooth value can calculate according to the following formula:
Wherein,It indicatesSmooth value；It is the linear increase factor, represents smoothed out trend；α and β is Number, value range is (0,1)；The primary condition of Holt model is as follows:
Step 1.3: after the smooth value and the linear increase factor for obtaining measurement data, predicted using the result, Predicted valueAre as follows:
Wherein l represents prediction step, and τ is the sampling interval of device sensor signal.In this example, preset prediction step For l=15.
According to step 1, sequence is measured for each operating parameter, fixed step size prediction is carried out, is as a result provided in Fig. 1, with this Actual status monitoring measurement sequence is given simultaneously.
Step 2: in a manner of sliding window, the triple characterized by industrial process operating parameter variation tendency is constructed, Measurement value sequence in data window is indicated, and is calculated in the window and is appointed by the relevance index based on Euclidean distance The relevance for two operating parameters of anticipating.This step specifically includes following sub-step:
Step 2.1: operating parameter association analysis is divided into time series segmentation fitting, triple indicates and relevance calculates Three phases；In the time series segmentation fitting stage, the data window for being L for regular length is denoted as window Wj, in this example In, sliding window length is 100.Operating parameter Xi, i=1,2 ..., N, the L data in the window areWherein j is the window start, and corresponding sampling instant is tj；If assuming one of them in the window M data in periodJust it can be fitted by a line segment, for measured value therein Its corresponding value in matching line segment isThe then error of fitting ERR calculation formula of the line segment are as follows:
When for carrying out piece-wise linearization as the window data of starting point using j, fromStart to carry out line segment fitting to it；? The process of data sectional fitting carries out (wherein remembering the error of fitting threshold value set as ω according to the following stepsE, ω in this exampleE= 0.025):
Step (1): setting is fitted starting point and isBeing fitted terminal isH=2；Wherein, for using j as the window of starting point Mouth Wj, initial fitting starting point is
Step (2): for dataLine segment fitting is carried out by the way of linear regression, is thus obtained Obtain corresponding segment dataIts error of fitting is calculated according to the error of fitting ERR calculation formula ERR；
Step (3): if ERR≤ωE, then h=h+1 is enabled, and repeat step (2)；If ERR > ωE, then current fitting is saved eventually Point (i.e. data cut-point) resets h=2, and returns to step 1, is that new fitting starting point carries out next portion to be currently fitted terminal The fitting of divided data；
Above step (1)~(3) are repeated until window WjInterior all data line segment, that is, obtain piecewise linearity Data after change
Step 2.2: showing the stage in line segment triple table, a line segment s is described in the form of following triplej,
Wherein, kjIndicate line segment slope,Indicate the length of the line segment on a timeline, rjIndicate the increasing of the line segment numerical value Long rate, i.e., for segment dataFor window WjNew sequence after data line segmentObtaining its triad sequence representation is { s1,s2,…,sn, wherein n indicates number in the window According to the line segment quantity after segmentation；
Step 2.3: in relevance calculation stages, for two operating parameter V in equipment running processAAnd operating parameter VB, need to be split the data after line segment first: with operating parameter VAWith operating parameter VBFor, in window WjIt is interior, note ginseng Number VAWaypoint beParameter VBWaypoint beWherein nAAnd nBTable respectively Show the line segment number after the measurement data of operating parameter A and B are segmented in the window；To parameter VAAnd VBWaypoint merge, Removal duplicate keys are simultaneously arranged from small to large, and the segmentation point sequence for obtaining two parameters isWith Afterwards, parameter V is obtained according to the segmentation point sequence and triple representationAAnd VBNew triad sequence isWith
Obtain window WjInterior two parameter VAAnd VBTriad sequence after, pass through the relevance index based on Euclidean distance dABTo calculate two parameter V in the windowAAnd VBRelevance:
In formula:For parameter VAThe slope of i-th line section,For parameter VBThe slope of i-th line section,For parameter VAThe The nominal growth rate of i line segment,For parameter VBThe nominal growth rate of i-th line section.
Step 3: according to the relevance data configuration relevance prediction model for having calculated acquisition, the relevance predicts mould Type is multivariate autoregressive model, and estimates model parameter by partial least squares algorithm.This step specifically includes following sub-step:
Step 3.1: construction relevance prediction model: setting prediction step is f, determines that model is set according to known data length Meter parameter U and M make U+f+M-1=window WjTotal number, and construct following matrix:
Wherein, { d1,d2,…,dU+f+M-1Expression parameter VAAnd VBRelevance sequence, djIndicate window WjInterior two parameters VAAnd VBRelevance index dAB；
Step 3.2: for each relevance sequence, the relevance value after f step-length is predicted using wherein U relevance value, Construct relevance prediction model:
Fp=Dpθ
Wherein, parameter θ=[θ1,θ2,…,θU]T, can be obtained by partial least squares algorithm.
Step 4: for newly obtaining data, the prediction of unit exception operating condition is carried out according to relevance prediction model, and pre- It measures before unusual service condition occurs and constantly updates relevance prediction model and model parameter.This step specifically includes following sub-step It is rapid:
In prediction, the new matrix of the data configuration newly obtained using sensor:
Then, using constructed relevance prediction model to relevance value carry out f step prediction, i.e.,
WhenIt is abnormal to judge that equipment occurs, wherein dnormalIt is in just for equipment initial operating stage Parameter V when normal operating statusAAnd VBRelevance value, ωpIt is relevance value relative to initial normal drift value threshold value；IfThe new data reconstruction model then obtained using sensor is to update model parameter θ；
As the relevance value of certain prediction step is constantly predicted in the update of data, to predict that unit exception operating condition occurs Time.In this example, prediction step f=10, threshold value ωp=0.1.
According to step 2- step 4, being associated property calculates and relevance trend, as a result as shown in Figure 2.With Afterwards, according to the failure threshold ω of settingp, unusual service condition prediction is carried out, as a result as shown in figure 3, giving use at the same time The out-of-service time that the relevance sequence that real data obtains is obtained according to the threshold test of setting.
Fig. 1 gives steam turbine operation measured value of parameters and prediction result.It can be seen that utilization index smoothing prediction is just It can be to the preferable prediction effect of single gain of parameter.Fig. 2 gives steam turbine vacuum A and other parameters relevance trend prediction knot Fruit and true value compare.As seen from the figure, good prediction effect has been obtained.Fig. 3 gives steam turbine exception time of origin prediction knot Fruit.The 1-6 of Fig. 3 abscissa respectively represents the relevance of vacuum A Yu remaining 6 parameter, and ordinate indicates the unusual service condition of prediction Time of origin.From the figure 3, it may be seen that the prediction using several groups of parameter associations all obtains more accurately prediction result, it can Generation of certain step-ahead prediction to unusual service condition in advance.In addition, passing through the analysis to initial data, it is known that vacuum A is from the 762nd Sampled point starts instruction exception, accelerates decline, and by the analysis to relevance variation tendency, from figure 3, it can be seen that pre- earliest The time for measuring unusual service condition generation is the 590th sampled point (ID6), i.e., captures abnormal generation earlier.It is prior It is, by relevance variation tendency, to find and be abnormal corresponding parameter earliest, facilitate the position for judging to occur extremely, For excluding abnormal play an important role.

Claims (5)

1. a kind of industrial process unusual service condition prediction technique based on operating parameter association analysis, which is characterized in that including with Lower step:
Step 1: using Holt exponential smoothing model, fixed step size is carried out to each sensor of industrial process measurement value sequence collected Prediction；
Step 2: in a manner of sliding window, constructing the triple characterized by industrial process operating parameter variation tendency, logarithm It is indicated according to the measurement value sequence in window, and is calculated any two in the window by the relevance index based on Euclidean distance The relevance of a operating parameter；
Step 3: according to the relevance data configuration relevance prediction model for having calculated acquisition, the relevance prediction model is Multivariate autoregressive model, and model parameter is estimated by partial least squares algorithm；
Step 4: for newly obtaining data, the prediction of unit exception operating condition is carried out according to relevance prediction model, and is being predicted Unusual service condition constantly updates relevance prediction model and model parameter before occurring.
2. a kind of industrial process unusual service condition prediction side based on operating parameter association analysis according to claim 1 Method, it is characterised in that: step 1 includes following sub-step:
Step 1.1: for the industrial equipment with multiple sensors, note number of sensors is N, when equipment is in operational process In, the operational parameter value i.e. sensing data of characterization equipment running status is constantly collected, each sensor measurement sequence is denoted asWherein K indicates sequence length,Indicate that sensor i is sampled at k-th The measured value of moment point；
Step 1.2: for sensor i, value sequence being measured to it using Holt exponential smoothing model and is predicted, give measured valueIts smooth value can calculate according to the following formula:
Wherein,It indicatesSmooth value；It is the linear increase factor, represents smoothed out trend；α and β is smoothing factor, is taken Value range is (0,1)；The primary condition of Holt model is as follows:
Step 1.3: after the smooth value and the linear increase factor for obtaining measurement data, being predicted, predicted using the result ValueAre as follows:
Wherein l represents prediction step, and τ is the sampling interval of device sensor signal.
3. a kind of industrial process unusual service condition prediction side based on operating parameter association analysis according to claim 2 Method, it is characterised in that: step 2 includes following sub-step:
Step 2.1: operating parameter association analysis is divided into time series segmentation fitting, triple indicates and relevance calculates three Stage；In the time series segmentation fitting stage, the data window for being L for regular length is denoted as window Wj, operating parameter Xi,i =1,2 ..., N, the L data in the window areWherein j is the window start, corresponding sampling Moment is tj；If assuming the m data in the window in one of them periodCan by a line segment into Row fitting, for measured value thereinIts corresponding value in matching line segment isThe then error of fitting ERR of the line segment Calculation formula are as follows:
When for carrying out piece-wise linearization as the window data of starting point using j, fromStart to carry out line segment fitting to it；In data The process of piecewise fitting carries out according to the following steps, wherein remembering the error of fitting threshold value set as ωE:
Step (1): setting is fitted starting point and isBeing fitted terminal isWherein, for using j as the window of starting point Wj, initial fitting starting point is
Step (2): for dataLine segment fitting is carried out, by the way of linear regression thus to obtain phase The segment data answeredIts error of fitting ERR is calculated according to the error of fitting ERR calculation formula；
Step (3): if ERR≤ωE, then h=h+1 is enabled, and repeat step (2)；If ERR > ωE, then current fitting terminal is saved, H=2 is reset, and returns to step 1, is the fitting that new fitting starting point carries out next part data to be currently fitted terminal；
Above step (1)~(3) are repeated until window WjInterior all data line segment, that is, after obtaining piece-wise linearization Data
Step 2.2: showing the stage in line segment triple table, a line segment s is described in the form of following triplej,
Wherein, kjIndicate line segment slope,Indicate the length of the line segment on a timeline, rjIndicate the growth rate of the line segment numerical value, I.e. for segment dataFor window WjNew sequence after data line segmentObtaining its triad sequence representation is { s1,s2,…,sn, wherein n indicates data in the window Line segment quantity after segmentation；
Step 2.3: in relevance calculation stages, for two operating parameter V in equipment running processAWith operating parameter VB, first It first needs to be split the data after line segment: in window WjIt is interior, remember parameter VAWaypoint beParameter VBWaypoint beWherein nAAnd nBThe measurement data of operating parameter A and B are respectively indicated in the window Line segment number after segmentation；To parameter VAAnd VBWaypoint merge, removal duplicate keys simultaneously arranged from small to large, obtain The segmentation point sequence of two parameters isThen, it is indicated according to the segmentation point sequence and triple Form obtains parameter VAAnd VBNew triad sequence isWith
Obtain window WjInterior two parameter VAAnd VBTriad sequence after, pass through the relevance index d based on Euclidean distanceABCome Calculate two parameter V in the windowAAnd VBRelevance:
In formula:For parameter VAThe slope of i-th line section,For parameter VBThe slope of i-th line section,For parameter VAI-th The nominal growth rate of line segment,For parameter VBThe nominal growth rate of i-th line section.
4. a kind of industrial process unusual service condition prediction side based on operating parameter association analysis according to claim 3 Method, it is characterised in that: step 3 includes following sub-step:
Step 3.1: construction relevance prediction model: setting prediction step is f, determines that modelling is joined according to known data length Number U and M makes U+f+M-1=window WjTotal number, and construct following matrix:
Wherein, { d1,d2,…,dU+f+M-1Expression parameter VAAnd VBRelevance sequence, djIndicate window WjInterior two parameter VAWith VBRelevance index dAB；
Step 3.2: for each relevance sequence, using the relevance value after wherein U relevance value prediction f step-length, construction Relevance prediction model:
Fp=Dpθ
Wherein, parameter θ=[θ1,θ2,…,θU]T, can be obtained by partial least squares algorithm.
5. a kind of industrial process unusual service condition prediction side based on operating parameter association analysis according to claim 4 Method, it is characterised in that: step 4 includes following sub-step:
In prediction, the new matrix of the data configuration newly obtained using sensor:
Then, using constructed relevance prediction model to relevance value carry out f step prediction, i.e.,
WhenIt is abnormal to judge that equipment occurs, wherein dnormalIt is in and operates normally for equipment initial operating stage Parameter V when stateAAnd VBRelevance value, ωpIt is relevance value relative to initial normal drift value threshold value；IfThe new data reconstruction model then obtained using sensor is to update model parameter θ；
As the relevance value of certain prediction step is constantly predicted in the update of data, thus when predicting that unit exception operating condition occurs Between.
CN201910244872.5A 2019-03-28 2019-03-28 Industrial process abnormal working condition prediction method based on operation parameter correlation analysis Active CN110008565B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910244872.5A CN110008565B (en) 2019-03-28 2019-03-28 Industrial process abnormal working condition prediction method based on operation parameter correlation analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910244872.5A CN110008565B (en) 2019-03-28 2019-03-28 Industrial process abnormal working condition prediction method based on operation parameter correlation analysis

Publications (2)

Publication Number Publication Date
CN110008565A true CN110008565A (en) 2019-07-12
CN110008565B CN110008565B (en) 2020-12-11

Family

ID=67168700

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910244872.5A Active CN110008565B (en) 2019-03-28 2019-03-28 Industrial process abnormal working condition prediction method based on operation parameter correlation analysis

Country Status (1)

Country Link
CN (1) CN110008565B (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111553606A (en) * 2020-05-06 2020-08-18 上海核工程研究设计院有限公司 Evaluation method for nuclear power equipment component design and equipment index relevance
CN111913444A (en) * 2019-09-07 2020-11-10 宁波大学 Chemical process monitoring method based on time sequence multi-block modeling strategy
CN111983478A (en) * 2020-07-07 2020-11-24 江苏大学 Electrochemical energy storage power station SOC anomaly detection method based on Holt linear trend model
CN112464146A (en) * 2020-12-03 2021-03-09 北京航空航天大学 Key subsystem based on historical telemetering data and single-machine correlation health baseline construction method
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
CN113624025A (en) * 2021-08-18 2021-11-09 浙江大学 Condenser vacuum low-jump machine symptom capturing method based on correlation of operation parameters
CN113792988A (en) * 2021-08-24 2021-12-14 河北先河环保科技股份有限公司 Online monitoring data anomaly identification method for enterprise
CN113867262A (en) * 2020-06-30 2021-12-31 华晨宝马汽车有限公司 Apparatus, method and medium for monitoring operation state of spindle device
CN115034094A (en) * 2022-08-10 2022-09-09 南通恒强轧辊有限公司 Prediction method and system for operation state of metal processing machine tool
CN115509187A (en) * 2022-09-20 2022-12-23 北京中佳瑞通科技有限公司 Industrial big data processing method and system
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
CN115829422A (en) * 2023-02-21 2023-03-21 创银科技(南通)有限公司 Industrial equipment operation abnormal state identification method based on big data
WO2023094916A1 (en) * 2021-11-23 2023-06-01 International Business Machines Corporation Identifying persistent anomalies for failure prediction

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090265155A1 (en) * 2008-04-21 2009-10-22 Nec Electronics Corporation Method of predicting reliability of semiconductor device, reliability prediction system using the same and storage medium storing program causing computer to execute the same
CN108037387A (en) * 2017-12-05 2018-05-15 北京能源集团有限责任公司 The equipment fault analysis method and device collected based on cluster
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
CN109086804A (en) * 2018-07-12 2018-12-25 中石化石油机械股份有限公司 A kind of hydraulic device fault forecast method merged based on multi source status monitoring information and reliability characteristic
CN109110608A (en) * 2018-10-25 2019-01-01 歌拉瑞电梯股份有限公司 A kind of elevator faults prediction technique based on big data study

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090265155A1 (en) * 2008-04-21 2009-10-22 Nec Electronics Corporation Method of predicting reliability of semiconductor device, reliability prediction system using the same and storage medium storing program causing computer to execute the same
CN108037387A (en) * 2017-12-05 2018-05-15 北京能源集团有限责任公司 The equipment fault analysis method and device collected based on cluster
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
CN109086804A (en) * 2018-07-12 2018-12-25 中石化石油机械股份有限公司 A kind of hydraulic device fault forecast method merged based on multi source status monitoring information and reliability characteristic
CN109110608A (en) * 2018-10-25 2019-01-01 歌拉瑞电梯股份有限公司 A kind of elevator faults prediction technique based on big data study

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHUNHUI ZHAO 等: "Critical-to-Fault-Degradation Variable Analysis and Direction Extraction for Online Fault Prognostic", 《IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY》 *
卢会国 等: "基于关联规则的故障诊断方法研究", 《成都信息工程学院学报》 *
方名菊 等: "基于形式概念的汽轮机相似故障案例检索", 《山东科技大学学报(自然科学版)》 *

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111913444A (en) * 2019-09-07 2020-11-10 宁波大学 Chemical process monitoring method based on time sequence multi-block modeling strategy
CN111913444B (en) * 2019-09-07 2022-03-18 宁波大学 Chemical process monitoring method based on time sequence multi-block modeling strategy
CN111553606A (en) * 2020-05-06 2020-08-18 上海核工程研究设计院有限公司 Evaluation method for nuclear power equipment component design and equipment index relevance
CN113867262A (en) * 2020-06-30 2021-12-31 华晨宝马汽车有限公司 Apparatus, method and medium for monitoring operation state of spindle device
CN111983478A (en) * 2020-07-07 2020-11-24 江苏大学 Electrochemical energy storage power station SOC anomaly detection method based on Holt linear trend model
CN112464146A (en) * 2020-12-03 2021-03-09 北京航空航天大学 Key subsystem based on historical telemetering data and single-machine correlation health baseline construction method
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
CN113624025A (en) * 2021-08-18 2021-11-09 浙江大学 Condenser vacuum low-jump machine symptom capturing method based on correlation of operation parameters
CN113792988A (en) * 2021-08-24 2021-12-14 河北先河环保科技股份有限公司 Online monitoring data anomaly identification method for enterprise
WO2023094916A1 (en) * 2021-11-23 2023-06-01 International Business Machines Corporation Identifying persistent anomalies for failure prediction
CN115034094A (en) * 2022-08-10 2022-09-09 南通恒强轧辊有限公司 Prediction method and system for operation state of metal processing machine tool
CN115509187A (en) * 2022-09-20 2022-12-23 北京中佳瑞通科技有限公司 Industrial big data processing method and system
CN115509187B (en) * 2022-09-20 2023-04-18 北京中佳瑞通科技有限公司 Industrial big data processing method and system
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
CN115689071B (en) * 2023-01-03 2023-05-02 南京工大金泓能源科技有限公司 Equipment fault fusion prediction method and system based on associated parameter mining
CN115829422A (en) * 2023-02-21 2023-03-21 创银科技(南通)有限公司 Industrial equipment operation abnormal state identification method based on big data
CN115829422B (en) * 2023-02-21 2024-01-02 北京瀚海蓝山智能科技有限公司 Industrial equipment operation abnormal state identification method based on big data

Also Published As

Publication number Publication date
CN110008565B (en) 2020-12-11

Similar Documents

Publication Publication Date Title
CN110008565A (en) A kind of industrial process unusual service condition prediction technique based on operating parameter association analysis
CN104390657B (en) A kind of Generator Unit Operating Parameters measurement sensor fault diagnosis method and system
CN110008253B (en) Industrial data association rule mining and abnormal working condition prediction method
CN110414154B (en) Fan component temperature abnormity detection and alarm method with double measuring points
CN104483575A (en) Self-adaptive load event detection method for noninvasive power monitoring
EP3367198A1 (en) Method for monitoring and diagnosing rotating machines based on signature analysis of electrical signals
CN102789676B (en) Method for designing industrial alarm on basis of alarm evidence fusion
CN112799898B (en) Interconnection system fault node positioning method and system based on distributed fault detection
CN108038044A (en) A kind of method for detecting abnormality towards continuous monitored target
CN113776794B (en) Fault diagnosis method, device and system for embedded edge calculation
CN111860839A (en) Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
CN111665066A (en) Equipment fault self-adaptive upper and lower early warning boundary generation method based on convolutional neural network
CN115455358A (en) Electrical parameter trend early warning and fault diagnosis method based on nonlinear regression model
CN111984705A (en) Precision evaluation management method and device for power quality monitoring device
CN117556366B (en) Data abnormality detection system and method based on data screening
WO2016136391A1 (en) Fault point locating device and method, electric power system monitoring system, and facility planning support system
CN109933031B (en) System and method for automatically correcting soft measuring instrument according to assay data
US11339763B2 (en) Method for windmill farm monitoring
CN109556861A (en) A kind of bearing real-time fault diagnosis system of case-based reasioning
KR102351538B1 (en) System and method for predicting failure of automated facilities using and algorithm that learns and diagnoses patterns
CN108537174B (en) Online monitoring method and system for running state of rotating machinery under time-varying working condition
CN111986469A (en) Intelligent diagnosis method for field terminal fault
KR20010036423A (en) Method and apparatus for detecting fault using General Regression Neural Network
WO2022071189A1 (en) State-monitoring device and state-monitoring method
CN109583124A (en) A kind of HMM failure prediction system based on ADRC

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