CN101799369A - Grey modeling method in failure monitoring and predicting of power equipment - Google Patents
Grey modeling method in failure monitoring and predicting of power equipment Download PDFInfo
- Publication number
- CN101799369A CN101799369A CN201010101342A CN201010101342A CN101799369A CN 101799369 A CN101799369 A CN 101799369A CN 201010101342 A CN201010101342 A CN 201010101342A CN 201010101342 A CN201010101342 A CN 201010101342A CN 101799369 A CN101799369 A CN 101799369A
- Authority
- CN
- China
- Prior art keywords
- modeling
- grey
- data
- condition
- level
- 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.)
- Pending
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a grey modeling method in failure monitoring and predicting of power equipment, comprising the following steps: (1) judging whether non-equidistant sampled data meet the requirements based on the class ratio condition and the modeling condition; (2) when meeting the modeling condition, carrying out grey modeling according to the grey modeling method by a difference quotient method, if not, carrying out equidistant treatment; (3) judging whether the sampled data rules are smooth; and (4) carrying out fitting precision inspection after grey modeling by various methods. The initial data sequence is smoothened, and when determining the grey moment, the difference method is not adopted but the grey data sequence is determined according to the actual variation of the sampled data, thereby improving the modeling and predicating precision. The invention can be widely applied to various reciprocating machines.
Description
Technical field
The present invention relates to a kind of mechanical fault forecast monitoring modeling method, particularly about the grey modeling method in a kind of failure monitoring and predicting of power equipment that is used for the reciprocating machine field.
Background technology
The reciprocating mechanical failure forecasting procedure is confined to the dummy run phase in the laboratory more, and practicality also reaches field requirement far away.Main difficulty shows: non-stationary, the plyability of signal of signal, the diversity of diagnostic model, suitable complexity of reciprocating machine structure and motion state and model are a lot, are difficult to summarize general character.At present, GM model (gray model) successfully has been applied to many fields such as engineering prediction, control, society, economy and agricultural as the core of gray system theory.The application of gray theory in failure prediction comprises gray system modeling, correlation analysis, gray model prediction etc.The principle that gray theory is used for failure prediction is that predicted system is regarded as a gray system, utilizes the Given information that exists to go to know by inference characteristic, state and the development trend that contains fixed mode unknowable information, and development in future is made prediction and made a strategic decision.Gray theory is applied to reciprocating mechanical failure diagnosis to be based on sample data amount finite sum sample value and to change that to have randomicity characteristics selected.The GM model of making prediction usefulness is generally GM (n, 1) model (n represents the order of a differential equation number, and 1 represents the number of variable), and the wherein of paramount importance while also is that using maximum in practice is GM (1,1) model.Because GM (1, the 1) model of establishment is on the basis of Ullage when being based upon etc., and the sample of being got in the actual tests process much all right and wrong are equally spaced, so can only adopt the unequal interval modeling method.Can reduce two classes for the unequal interval grey modeling method: a class is difference coefficient type grey modeling, promptly the unequal interval sequence carried out the differential equation being approximately the difference equation modeling finding the solution after one-accumulate generates; Another kind of is sequential match type, promptly earlier goes out uniformly-spaced sequence according to the unequal interval sequence structure, utilizes uniformly-spaced sequence modeling to find the solution again.Difference coefficient type grey modeling requires the sample interval of modeling data to differ can not be too big, otherwise model accuracy is not high; Sequential match type grey modeling requirement data wave momentum can not be too big, otherwise the error that predicts the outcome is bigger, when above-mentioned these two kinds of methods are respectively applied in the reciprocating mechanical failure prediction, because reciprocating machine oil analysis data often do not satisfy sample interval and the less requirement of data wave momentum, therefore range of application has certain limitation, and diagnostic accuracy is not high.
Summary of the invention
At the problems referred to above, the purpose of this invention is to provide a kind of be suitable for comparatively extensively and failure monitoring and predicting of power equipment that diagnostic accuracy is higher in grey modeling method.
For achieving the above object, the present invention takes following technical scheme: the grey modeling method in a kind of failure monitoring and predicting of power equipment, its step is as follows: (1) is for the sampled data of non-equidistance, whether satisfy level than requiring according to level than condition judgment sampled data, if do not satisfy, then sampled data is carried out judging again after logarithmetics is handled; If satisfy, then continue to judge whether to satisfy the modeling condition; (2) when satisfying the modeling condition, then carry out grey modeling according to the method for difference coefficient method grey modeling; If do not satisfy the modeling condition, then processing is uniformly-spaced changed in sampling constantly; (3) after compartmentation is handled, judge whether the sampled data rule is level and smooth,, then can carry out modeling according to the grey modeling method of sequential fitting process if data rule is level and smooth; If data rule has undulatory property, then utilize improved sequential fitting process to carry out grey modeling based on the modeling data Changing Pattern; (4) behind described difference coefficient method, sequential fitting process and the sequential fitting process grey modeling, all to carry out the fitting precision check,, then can carry out gray prediction if be up to the standards based on the modeling data Changing Pattern; If disqualified upon inspection then changes the modeling dimension and carries out modeling again, until satisfying the fitting precision requirement.
Described level than condition is:
Wherein, δ be the level than function, n is the quantity of sample data.
Described modeling condition is: max (Δ t
k)/min (Δ t
k)＜1.5, wherein Δ t
kBe sampling interval time.
The present invention is owing to take above technical scheme, and it has the following advantages: 1, the present invention is owing to pass through initial data sequence x
(0)(k) carry out smooth treatment, and at definite grey moment point t
i' (i=1,2 ..., in the time of k), do not adopt differential technique, but determine the grey data sequence, thereby improved the modeling and forecasting precision according to the actual change rule of sample data.2, the present invention is because the modeling information needed is less, and for unequal interval grey GM (1, the 1) modeling that improves sequential match type, when dimension was 4, the model prediction accuracy reached the highest.3, the present invention is not owing to need know the priori features that raw data distributes, to the discrete original series of arbitrary smooth random or that obey any distribution, handle by making logarithmetics, make it to change into regular sequence, thereby provide intermediate information and the randomness of the former random series that weakens for modeling.4, the present invention since each predicted data all be based on the several sampled datas in adjacent front according to etc. the method for dimension modeling carry out obtaining after the prediction and calculation, and only predict a next step after each modeling, adopt then and wait the dimension modeling method, new sampled result is added, eliminate the oldest information simultaneously, therefore again next value of modeling and forecasting has again kept the feature of original system, can reflect the actual conditions of system preferably.Therefore the present invention can be widely used in the reciprocating machines such as various large combustion machines, compressor and reciprocating pump.
Description of drawings
Fig. 1 is a grey modeling process flow diagram of the present invention.
Embodiment
Below in conjunction with drawings and Examples the present invention is described in detail.
As shown in Figure 1, because the failure prediction in the reciprocating machine experimental system method for diagnosing faults of the present invention is based on grey GM (1, the 1) modeling method of modeling data Changing Pattern, for the sampling interval time Δ t of the oil analysis data of reciprocating machine
kBe unequal, grey GM (1,1) modeling method may further comprise the steps:
(1) for the sampled data of non-equidistance, judge than condition A (level is than the ratio that is conjoint data) whether the sampled data of fluid data and each sensor acquisition satisfies level than requiring according to level, if do not satisfy, then concentration of element in the fluid and various data are carried out judging again after logarithmetics is handled; If satisfy level than condition A, then continue to judge whether to satisfy modeling condition max (Δ t
k)/min (Δ t
k)＜1.5; Δ t wherein
kBe sampling interval time;
Level than condition A is:
Wherein, δ compares function for level; N is the quantity of sample data;
(2) when satisfying modeling condition max (Δ t
k)/min (Δ t
kGrey modeling is then carried out according to the method for difference coefficient method grey modeling in)＜1.5 o'clock; If do not satisfy the modeling condition, then processing is uniformly-spaced changed in sampling constantly;
(3) after compartmentation is handled, judge whether the sampled data rule is level and smooth,, then can carry out modeling according to the grey modeling method of sequential fitting process if data rule is level and smooth; If data rule has undulatory property, then utilize modified sequential fitting process to carry out grey modeling based on the modeling data Changing Pattern;
(4) behind the difference coefficient method in the above steps, sequential fitting process and the sequential fitting process grey modeling, all to carry out the fitting precision check,, then can carry out gray prediction if be up to the standards based on the modeling data Changing Pattern; If disqualified upon inspection then changes the modeling dimension and carries out modeling again, until satisfying the fitting precision requirement.
Below by specific embodiment, the present invention is done further introduction.
Embodiment 1: with certain reciprocating engine road data is example, carries out Grey Prediction Modeling.Modeling mainly comprises following process:
(1) the raw data level is than judging that Fe concentration of element analysis result and data sequence level are than as shown in table 1 in the actual road test process.
Table 1
In the above table, multinomial satisfied level is arranged than requiring in the video data, therefore, must carry out pre-service to this data sequence, adopt the log-transformation method at this, the data sequence after the conversion is as shown in table 2.
Table 2
(2) sample interval is judged constantly, and through judging, the data sequence mileage does not satisfy max (Δ t at interval
k)/mix (Δ t
kThe condition of)＜1.5 is so can only carry out the match prediction according to sequential match type grey modeling method; The conversion mileage intervening sequence of determining according to least square method and as shown in table 3 based on the grey data sequence of data variation rule.
Table 3
(3) model fitting precision judges that two kinds of sequential match type grey modeling method accuracy test values are as shown in table 4.
Table 4
By above table as can be known, traditional sequential match type is the definite grey sequence of values x of method (least square method) in proportion
(0)(t
k'), so that its modeling fitting precision is wanted is high, but we to set up the final purpose of gray model be to predict, require just passable as long as the model of foundation satisfies fitting precision.
Because the fitting precision of two kinds of methods all reaches extra fine grade in the table 4, satisfies the fitting precision requirement, therefore can predict.
(4) gray prediction according to above-mentioned two kinds of modeling methods, predicts that to the follow-up test wear extent it is as shown in table 5 to predict the outcome.
Table 5
The demonstration that predicts the outcome, when the prediction mileage was 623km, traditional sequential match type gray model predicated error was 3.427%, and follow-on gray model predicated error is 1.715%; When the prediction mileage is 1076km, traditional sequential match type gray prediction error is 10.293%, and follow-on predicated error is 7.977%, and the result shows, and is little based on the improvement sequential match type grey forecasting model predicated error of data variation rule, accuracy is higher.
Embodiment 2:, another reciprocating machine oil analysis data of the same type are predicted that original experiment data and prediction verification msg are as shown in table 6 for further verification model prediction accuracy.
Table 6
(the Δ t of max during modeling
k)/mix (Δ t
kThe condition of)＜1.5, and level is than also not satisfying
Condition, therefore can only carry out the log-transformation pre-service to the data sequence respectively, the mileage sequence is equidistantly changed processing.
Raw data is carried out the result of three kinds of gray prediction models, and then show the influence relation of data pre-service and gray prediction model accuracy, as shown in table 7.The result who raw data is carried out three kinds of gray prediction models after logarithmetics is handled is as shown in table 8.
Table 7
Table 8
Find that through the contrast back directly raw data is carried out modeling, the error that predicts the outcome is bigger; After raw data carried out pre-service, carry out modeling and forecasting under than situation about requiring satisfying level, precision of prediction improves a lot.The demonstration that predicts the outcome of three kinds of methods, the improvement sequential match type grey forecasting model precision of prediction that the present invention proposes is the highest.For this test data sequence, because increase, so cause precision of prediction to descend to some extent at the late time data rate of change, therefore, for grey modeling, not necessarily to carry out modeling based on whole given datas, should select to concern that with the later stage wear trend closer pass tuple is according to modeling.
For the selection of grey modeling grey data, should be according to principle nearby, promptly the nearer several data of selected distance predicted value are carried out modeling, but what are selected actually, need determine according to the data sample Changing Pattern.Can predict that it predicts the outcome as shown in table 9 to the follow-up test data according to the modeling curve.
Table 9
In the above table, each predicted data all be based on the several sampled datas in adjacent front according to etc. the method for dimension modeling carry out obtaining after the prediction and calculation, and only predict a next step after each modeling, adopt then and wait the dimension modeling method, new sampled result is added, eliminate the oldest information simultaneously, again next value of modeling and forecasting again.The demonstration that predicts the outcome, for unequal interval grey GM (1, the 1) modeling that improves sequential match type, when dimension was 4, the model prediction accuracy was the highest.
The various embodiments described above only are preferred implementations of the present invention, and are every based on the changes and improvements on the technical solution of the present invention in the present technique field, should not get rid of outside protection scope of the present invention.
Claims (3)
1. the grey modeling method in the failure monitoring and predicting of power equipment, its step is as follows:
(1) for the sampled data of non-equidistance, whether satisfies level than requiring than condition judgment sampled data,, then sampled data is carried out judging again after logarithmetics is handled if do not satisfy according to level; If satisfy, then continue to judge whether to satisfy the modeling condition;
(2) when satisfying the modeling condition, then carry out grey modeling according to the method for difference coefficient method grey modeling; If do not satisfy the modeling condition, then processing is uniformly-spaced changed in sampling constantly;
(3) after compartmentation is handled, judge whether the sampled data rule is level and smooth,, then can carry out modeling according to the grey modeling method of sequential fitting process if data rule is level and smooth; If data rule has undulatory property, then utilize improved sequential fitting process to carry out grey modeling based on the modeling data Changing Pattern;
(4) behind described difference coefficient method, sequential fitting process and the sequential fitting process grey modeling, all to carry out the fitting precision check,, then can carry out gray prediction if be up to the standards based on the modeling data Changing Pattern; If disqualified upon inspection then changes the modeling dimension and carries out modeling again, until satisfying the fitting precision requirement.
2. the grey modeling method in a kind of failure monitoring and predicting of power equipment as claimed in claim 1 is characterized in that: described level than condition is:
Wherein, δ be the level than function, n is the quantity of sample data.
3. the grey modeling method in a kind of failure monitoring and predicting of power equipment as claimed in claim 1 is characterized in that: described modeling condition is: max (Δ t
k)/min (Δ t
k)＜1.5, wherein Δ t
kBe sampling interval time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010101342A CN101799369A (en) | 2010-01-27 | 2010-01-27 | Grey modeling method in failure monitoring and predicting of power equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010101342A CN101799369A (en) | 2010-01-27 | 2010-01-27 | Grey modeling method in failure monitoring and predicting of power equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101799369A true CN101799369A (en) | 2010-08-11 |
Family
ID=42595139
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010101342A Pending CN101799369A (en) | 2010-01-27 | 2010-01-27 | Grey modeling method in failure monitoring and predicting of power equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101799369A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824137A (en) * | 2014-03-20 | 2014-05-28 | 北京信息科技大学 | Multi-condition fault prediction method for complex mechanical equipment |
| CN105938575A (en) * | 2016-04-13 | 2016-09-14 | 山东毅康科技股份有限公司 | Multivariable-grey-neural-network-based prediction system for residual life of industrial equipment |
| CN106339588A (en) * | 2016-08-25 | 2017-01-18 | 华南理工大学 | Discrete modeling method of accelerated degradation data based on grey system theory |
| CN109034225A (en) * | 2018-07-16 | 2018-12-18 | 福州大学 | A kind of combination stochastic variable ash and the modified uncertain parameters estimation method of Bayesian model |
| CN109657398A (en) * | 2018-12-29 | 2019-04-19 | 中国人民解放军92942部队 | A kind of non-equidistant Ship Structure residual thickness prediction technique based on gray theory |
| CN113849540A (en) * | 2021-09-22 | 2021-12-28 | 广东电网有限责任公司 | Fault prediction model training and prediction method, device, electronic equipment and medium |
-
2010
- 2010-01-27 CN CN201010101342A patent/CN101799369A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824137A (en) * | 2014-03-20 | 2014-05-28 | 北京信息科技大学 | Multi-condition fault prediction method for complex mechanical equipment |
| CN103824137B (en) * | 2014-03-20 | 2016-08-17 | 北京信息科技大学 | A kind of complex mechanical equipment multi-state failure prediction method |
| CN105938575A (en) * | 2016-04-13 | 2016-09-14 | 山东毅康科技股份有限公司 | Multivariable-grey-neural-network-based prediction system for residual life of industrial equipment |
| CN106339588A (en) * | 2016-08-25 | 2017-01-18 | 华南理工大学 | Discrete modeling method of accelerated degradation data based on grey system theory |
| CN109034225A (en) * | 2018-07-16 | 2018-12-18 | 福州大学 | A kind of combination stochastic variable ash and the modified uncertain parameters estimation method of Bayesian model |
| CN109034225B (en) * | 2018-07-16 | 2021-01-29 | 福州大学 | Uncertainty parameter estimation method combining random variable grayness and Bayesian model correction |
| CN109657398A (en) * | 2018-12-29 | 2019-04-19 | 中国人民解放军92942部队 | A kind of non-equidistant Ship Structure residual thickness prediction technique based on gray theory |
| CN109657398B (en) * | 2018-12-29 | 2023-02-21 | 中国人民解放军92942部队 | Grey theory-based method for predicting residual thickness of unequally-spaced ship structure |
| CN113849540A (en) * | 2021-09-22 | 2021-12-28 | 广东电网有限责任公司 | Fault prediction model training and prediction method, device, electronic equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110779746B (en) | Diagnosis method for improving composite fault of deep sparse self-encoder network rotating machinery | |
| CN101799369A (en) | Grey modeling method in failure monitoring and predicting of power equipment | |
| CN104598736B (en) | A kind of multi-core adaptive combines the rolling bearing life forecast model of Method Using Relevance Vector Machine | |
| CN102509023B (en) | Modeling method for combined stress accelerated life test damage accumulation model of space driving assembly | |
| CN104899608B (en) | The Weighted Fusion Method Using Relevance Vector Machine model of rolling bearing predicting residual useful life | |
| Altunkaynak | A predictive model for well loss using fuzzy logic approach | |
| CN102033994A (en) | Steering engine reliability simulation sampling method based on Markova chain Monte Carlo | |
| CN104020401A (en) | Cloud-model-theory-based method for evaluating insulation thermal ageing states of transformer | |
| CN100349135C (en) | Test vector generation method | |
| CN103455658A (en) | Weighted grey target theory based fault-tolerant motor health status assessment method | |
| Huang et al. | Application of optimized GM (1, 1) model based on EMD in landslide deformation prediction | |
| Tian et al. | An outliers detection method of time series data for soft sensor modeling | |
| Wu et al. | Time-dependent reliability model of components with strength degradation based-on gamma process | |
| CN103603794A (en) | Method and device for adaptive fault diagnosis of gas storage injection-production compressor unit | |
| Wong et al. | Modelling and prediction of diesel engine performance using relevance vector machine | |
| CN102749584A (en) | Prediction method for residual service life of turbine generator based on ESN (echo state network) of Kalman filtering | |
| CN105426991A (en) | Transformer defect prediction method and transformer defect prediction system | |
| CN102789445A (en) | Establishment method for wavelet analysis and rank set pair analysis of medium and long-term hydrological forecast model | |
| Guo et al. | The variable sampling interval S 2 chart with known or unknown in-control variance | |
| CN107256333A (en) | A kind of Argo profile anomaly detection methods based on prediction and dynamic threshold | |
| CN116341705A (en) | Long-period memory network water quality parameter prediction method based on sparse label | |
| Niu et al. | A hybrid bearing prognostic method with fault diagnosis and model fusion | |
| CN102033989A (en) | Back propagation (BP) neural network-based chloridion solidified amount prediction method | |
| CN102156802A (en) | Method for forecasting evenly distributed live data | |
| Cui et al. | Multipoint vibration response prediction under uncorrelated multiple sources load based on elastic-net regularization in frequency domain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| DD01 | Delivery of document by public notice |
Addressee: Xu Xiaoli Document name: Notification of Passing Preliminary Examination of the Application for Invention Addressee: Xu Xiaoli Document name: Notification of Passing Examination on Formalities |
|
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20100811 |