CN106778848A - A kind of Zn finery method for diagnosing faults based on multiclass Probit models - Google Patents
A kind of Zn finery method for diagnosing faults based on multiclass Probit models Download PDFInfo
- Publication number
- CN106778848A CN106778848A CN201611097388.7A CN201611097388A CN106778848A CN 106778848 A CN106778848 A CN 106778848A CN 201611097388 A CN201611097388 A CN 201611097388A CN 106778848 A CN106778848 A CN 106778848A
- Authority
- CN
- China
- Prior art keywords
- fault
- finery
- multiclass
- data
- probit
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/245—Classification techniques relating to the decision surface
- G06F18/2453—Classification techniques relating to the decision surface non-linear, e.g. polynomial classifier
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Nonlinear Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Zn finery method for diagnosing faults based on multiclass Probit models, its step is:Fault alarm data to gathering typing, carry out preliminary data pretreatment;With PCA, i.e. principal component analysis, data are standardized, and extract the principal character information of fault data;With reference to Monte Carlo method, by Probit models, discriminant classification is carried out to fault data;Mishap Database and fault knowledge storehouse are utilized simultaneously, and renewal adjustment is iterated to Probit models and fault category.The present invention shortens the fault information analysis time by above procedure, improves fault message classification accuracy, reduce cost, so as to improve the degree of accuracy of failure modes, the reliability and stability of Zn finery are improve, also achieve carries out failure predication with model.
Description
Technical field
The present invention relates to method for diagnosing faults field, specifically, it is related specifically to a kind of based on multiclass Probit models
Zn finery method for diagnosing faults.
Background technology
With the fast development of national economy, zinc powder as raw material be increasingly being applied to metallurgy, chemical industry, building,
The fields such as traffic, medicine, electronics and food.Zinc powder production generally carries out rectifying condensation using rectifying column, and the method is using different
The boiling point of metal is different, such as the low boiling of zinc, vapor pressure characteristic high, is made by the rapid condensation of continuous rectification and zinc powder condenser
With foreign metal being separated, so as to obtain impurity content low, even particle size distribution, the good smart zinc of chemism.With zinc powder
Market demand increasingly increases, the normal work of Zn finery in Making Zinc powder industry in occupation of critical role, thus to rectifying
During occur failure carry out effectively identification be particularly important.
As the increasing number of zinc refining equipment, control system scale are increasing, structure becomes increasingly complex.Any equipment
The shadow of uncertain factor and running environment factor in quality or design defect, and large complicated smelting process control system
Ring etc., all often causing the reliability of operational outfit and control system reduces, and rate of breakdown is high, and phenomenon of the failure, event
The characteristics of barrier reason and fault type show variation and ambiguity.At present, the fault diagnosis technology of most of rectifying column is all
It is expert system based on Heuristics or to the exigent ray scanning technology of appointed condition, but these diagnostic methods
Difficult point is the foundation renewal and the required precision of equipment of expert knowledge library, and the data for being gathered are difficult to judge complex fault point
Class.
The content of the invention
It is an object of the invention to be directed to deficiency of the prior art, there is provided a kind of simple and easy to apply, low cost is fireballing
Zn finery method for diagnosing faults, i.e. the Zn finery method for diagnosing faults based on multiclass Probit models, so as to improve event
Hinder the degree of accuracy of diagnosis, improve the reliability and stability of Zn finery, it is achieved thereby that the classification of fault data, and use mould
Type carries out failure predication.
Technical problem solved by the invention can be realized using following technical scheme:
2. a kind of Zn finery method for diagnosing faults based on multiclass Probit models, it is characterised in that including following step
Suddenly:
1) fault data is gathered by rectifying column data acquisition device, and preliminary pre- place is carried out to the fault data
Reason;
2) PCA PCA is used, the principal character information of the fault data of Zn finery is extracted;
3) with the fault grader based on multiclass Probit models, the principal character of the fault data of Zn finery is believed
Breath carries out classification analysis as training sample;
4) with the fault grader based on multiclass Probit models, fault data type is differentiated;Meanwhile, by life
Into grouped data sample, be input to Mishap Database and be updated, and as historical experience to multiclass Probit models therefore
Barrier grader is trained adjustment, and the result for differentiating fault type iterates renewals, and realization is at utmost using number of faults
According to collection;
5) it is that the result of fault type module is compared with fault knowledge storehouse, comparison result is included man-machine in real time
In interactive interface.
2. the Zn finery method for diagnosing faults based on multiclass Probit models according to claim 1, its feature
Be, the step 2) flow it is as follows:Fault data is standardized, correlation matrix R and R is then asked
Characteristic value and characteristic vector, are designated as λi, (i=1,2 ..., p), its character pair vector is ei(i=1,2 ..., p), connect
The main composition contribution rate of calculating, contribution rate of accumulative total and determine principal component number.
3. the Zn finery method for diagnosing faults based on multiclass Probit models according to claim 1, its feature
Be, the step 3) flow it is as follows:
3.1) the principal character information of the fault data, it is assumed that Normal Distribution, its distribution density function is:
If (η 1, η 2) is equally distributed stochastic variable on [0,1], for standardized normal distribution N (0,1), using binary
Functional transformation can obtain sample of random variable value:
Ε=(- 2In η1)1/2cos(2πη2);
Then normal distribution N (μ, σ2) sample of random variable value be:
Y=ε σ+u;
The principal character information of the Zn finery fault data generated with reference to PCA, carries out next step calculating;
3.2) according to Monte Carlo method is utilized, one group of random function of match state variable distribution is produced, i.e., with reference to zinc essence
The principal character information of tower fault data is evaporated, multiclass Probit Model Parameters are determined;
Detailed process is as follows:
3.21) construct or describe probabilistic process, i.e., first construct artificial probabilistic process, parameter is with reference to Zn finery event
Hinder the parameter in the multiclass Probit models of data;
3.22) realize being distributed from known probability sampling, realized by means of random sequence;
3.23) various estimators are set up, so as to therefrom obtain the solution of multiclass Probit Model Parameters.
Compared with prior art, beneficial effects of the present invention are as follows:
1) the principal character information in fault message is extracted by PCA, the Zn finery fault data to gathering is carried out
Standardization pretreatment, reduces fault characteristic information difference numerically, reduces the complexity of data, realizes data not phase
Close and model simplification.
2) Probit models are a nonlinear regression model (NLRM)s, and its normal distribution for representing can almost represent all of number
According to so Probit is one of model for being most widely used in modern statistics.It both can carry out linear unbias to parameter
Estimate, nonlinear response variable can also be simulated.
3) parameter of Probit models is solved using Monte Carlo Method, advantage has that calculating is simple, is easily achieved;Not factor data
Dimension and influence its convergence rate;Error easily determines;Limited small etc. by geometrical condition.
4) data sample after training is input into data fault database to be updated, as historical experience to multiclass
The fault grader of Probit models is trained adjustment, and the result for differentiating fault type iterates renewal, so as to realize
Fault data collection, and being continuously increased with all kinds fault data are at utmost utilized, Mishap Database system is constantly entered
Row updates, and fault grader is also more perfect.
5) present invention incorporates fault knowledge storehouse system, by the various of fault alarm sample data and fault knowledge storehouse system
Fault type data sample carries out matching combination, it is achieved thereby that quick diagnosis fault type, improves the accurate of fault diagnosis
Degree, improves the reliability and stability of Zn finery, so as to reduce the loss that failure is caused, with value higher.
Brief description of the drawings
Fig. 1 is the main flow chart of the Zn finery method for diagnosing faults based on multiclass Probit models of the invention.
Fig. 2 is the specific content resolution schematic diagram in fault knowledge storehouse of the invention.
Specific embodiment
For technological means, creation characteristic, reached purpose and effect for making present invention realization are easy to understand, with reference to
Specific embodiment, is expanded on further the present invention.
Referring to Fig. 1, the Zn finery method for diagnosing faults based on multiclass Probit models of the present invention, its step is such as
Under:
Step one is by rectifying column data acquisition device, to gather fault alarm data, and data is carried out preliminary pre-
Treatment.The reason for rectifying column causes failure in the process of running is many, wherein main the reason for main initiation failure
Equipment and control are concentrated on, is gone into operation or is stopped work, operating process, reboiler and condenser install error, and column plate is designed with downspout
With decompressor etc..Accordingly, the fault data of collection is tentatively pre-processed.
Step 2 is to use PCA, i.e. principal component analysis, extracts the principal character information of Zn finery fault data.Its
Flow data normalization first, then seeks the characteristic value and characteristic vector of correlation matrix R and R, is designated as λi, (i=1,
2 ..., p), its character pair vector is ei(i=1,2 ..., p).Then main composition contribution rate, contribution rate of accumulative total are calculated
And determine principal component number.
Step 3 is with the fault grader based on multiclass Probit models, by the main spy of Zn finery fault data
Reference breath carries out classification analysis as training sample, comprises the following steps that:
1) the principal character information of the fault data, it is assumed that Normal Distribution, its distribution density function is:
If (η 1, η 2) is equally distributed stochastic variable on [0,1], for standardized normal distribution N (0,1), using binary
Functional transformation can obtain sample of random variable value:
Ε=(- 2In η1)1/2cos(2πη2);
Then normal distribution N (μ, σ2) sample of random variable value be:
Y=ε σ+u;
The principal character information of the Zn finery fault data generated with reference to PCA, carries out next step calculating.
2) according to Monte Carlo method is utilized, one group of random function of match state variable distribution is produced, i.e., with reference to zinc fractionating
The principal character information of tower fault data, determines multiclass Probit Model Parameters.Detailed process is as follows:
(1) construct or describe probabilistic process, i.e., first construct artificial probabilistic process, parameter is with reference to Zn finery failure
Parameter in the multiclass Probit models of data.
(2) realize being distributed from known probability sampling, realized by means of random sequence.
(3) various estimators are set up, so as to therefrom obtain the solution of multiclass Probit Model Parameters.
Step 4 differentiates with the fault grader of multiclass Probit models to fault data type.Meanwhile, by life
Into grouped data sample, be input to Mishap Database and be updated, and as historical experience to multiclass Probit models therefore
Barrier grader is trained adjustment, and the result for differentiating fault type iterates renewals, so as to realize at utmost utilizing former
Barrier data set, and being continuously increased with all kinds fault data, Mishap Database system are constantly updated, failure modes
Device is also more perfect.
Step 5 is that the result of fault type module is compared with fault knowledge storehouse, so as to realize the essence of fault type
Really differentiate.Specific fault knowledge storehouse content is as shown in Figure 2.
Comparison result is included in human-computer interaction interface, can timely receive fault type differentiation and divide by step 6 in real time
Class result, and output result is controlled and adjustment in time.
Explanation two parts of renewal of the fault knowledge storehouse comprising knowledge and knowledge in Fig. 2.Wherein, renewal of knowledge part needs
Expert is wanted to provide experience and complete jointly to the summary of historical experience;And knowledge interpretation need to formulate a series of correlations
Compare with the data newly trained with rule and search mechanisms, in the hope of being capable of more accurately failure judgement type.
General principle of the invention and principal character and advantages of the present invention has been shown and described above.The technology of the industry
Personnel it should be appreciated that the present invention is not limited to the above embodiments, simply explanation described in above-described embodiment and specification this
The principle of invention, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appending claims and its
Equivalent thereof.
Claims (3)
1. a kind of Zn finery method for diagnosing faults based on multiclass Probit models, it is characterised in that comprise the following steps:
1) fault data is gathered by rectifying column data acquisition device, and preliminary pretreatment is carried out to the fault data;
2) PCA PCA is used, the principal character information of the fault data of Zn finery is extracted;
3) with the fault grader based on multiclass Probit models, the principal character information of the fault data of Zn finery is made
For training sample carries out classification analysis;
4) with the fault grader based on multiclass Probit models, fault data type is differentiated;Meanwhile, by what is generated
Grouped data sample, is input to Mishap Database and is updated, and the failure of multiclass Probit models is divided as historical experience
Class device is trained adjustment, and the result for differentiating fault type iterates renewals, realizes at utmost using fault data collection;
5) it is that the result of fault type module is compared with fault knowledge storehouse, comparison result is included in man-machine interaction in real time
In interface.
2. the Zn finery method for diagnosing faults based on multiclass Probit models according to claim 1, it is characterised in that
The step 2) flow it is as follows:Fault data is standardized, the feature of correlation matrix R and R is then sought
Value and characteristic vector, are designated as λi, (i=1,2 ..., p), its character pair vector is ei(i=1,2 ..., p), then count
Calculate main composition contribution rate, contribution rate of accumulative total and determine principal component number.
3. the Zn finery method for diagnosing faults based on multiclass Probit models according to claim 1, it is characterised in that
The step 3) flow it is as follows:
3.1) the principal character information of the fault data, it is assumed that Normal Distribution, its distribution density function is:
If (η 1, η 2) is equally distributed stochastic variable on [0,1], for standardized normal distribution N (0,1), using binary function
Conversion can obtain sample of random variable value:
Ε=(- 2In η1)1/2cos(2πη2);
Then normal distribution N (μ, σ2) sample of random variable value be:
Y=ε σ+u;
The principal character information of the Zn finery fault data generated with reference to PCA, carries out next step calculating;
3.2) according to Monte Carlo method is utilized, one group of random function of match state variable distribution is produced, i.e., with reference to Zn finery
The principal character information of fault data, determines multiclass Probit Model Parameters;
Detailed process is as follows:
3.21) construct or describe probabilistic process, i.e., first construct artificial probabilistic process, parameter is with reference to Zn finery number of faults
According to multiclass Probit models in parameter;
3.22) realize being distributed from known probability sampling, realized by means of random sequence;
3.23) various estimators are set up, so as to therefrom obtain the solution of multiclass Probit Model Parameters.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611097388.7A CN106778848A (en) | 2016-12-02 | 2016-12-02 | A kind of Zn finery method for diagnosing faults based on multiclass Probit models |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611097388.7A CN106778848A (en) | 2016-12-02 | 2016-12-02 | A kind of Zn finery method for diagnosing faults based on multiclass Probit models |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106778848A true CN106778848A (en) | 2017-05-31 |
Family
ID=58883055
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611097388.7A Pending CN106778848A (en) | 2016-12-02 | 2016-12-02 | A kind of Zn finery method for diagnosing faults based on multiclass Probit models |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106778848A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107238508A (en) * | 2017-07-21 | 2017-10-10 | 浙江中控技术股份有限公司 | A kind of equipment state diagnostic method and device |
CN111912638A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectifying tower fault diagnosis method for online fault source identification |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1987697A (en) * | 2006-12-22 | 2007-06-27 | 浙江大学 | Fault diagnostic system and method for under industrial producing process small sample condition |
CN102254193A (en) * | 2011-07-16 | 2011-11-23 | 西安电子科技大学 | Relevance vector machine-based multi-class data classifying method |
-
2016
- 2016-12-02 CN CN201611097388.7A patent/CN106778848A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1987697A (en) * | 2006-12-22 | 2007-06-27 | 浙江大学 | Fault diagnostic system and method for under industrial producing process small sample condition |
CN102254193A (en) * | 2011-07-16 | 2011-11-23 | 西安电子科技大学 | Relevance vector machine-based multi-class data classifying method |
Non-Patent Citations (5)
Title |
---|
丁云飞 等: "锌精馏塔的混合故障诊断方法", 《计算机工程》 * |
吴养会 等: "《农村金融计量研究方法及应用》", 30 November 2015, 中国金融出版社 * |
易富君: "《一种基于多元Probit模型的随机网络配流加载算法》", 《交通运输工程与信息学报》 * |
赵旭: "基于统计学方法的过程监控与质量控制研究", 《中国博士学位论文全文数据库信息科技辑》 * |
邓良基: "《统筹城乡发展 建设美好家园 高层论坛论文集》", 30 September 2009, 经济日报出版社 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107238508A (en) * | 2017-07-21 | 2017-10-10 | 浙江中控技术股份有限公司 | A kind of equipment state diagnostic method and device |
CN111912638A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectifying tower fault diagnosis method for online fault source identification |
CN111912638B (en) * | 2020-06-13 | 2021-12-21 | 宁波大学 | Rectifying tower fault diagnosis method for online fault source identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN100470417C (en) | Fault diagnostic system and method for under industrial producing process small sample condition | |
CN104931960B (en) | Trend message and radar target situation information full flight path segment data correlating method | |
CN104390657B (en) | A kind of Generator Unit Operating Parameters measurement sensor fault diagnosis method and system | |
CN109766583A (en) | Based on no label, unbalanced, initial value uncertain data aero-engine service life prediction technique | |
CN108985380B (en) | Point switch fault identification method based on cluster integration | |
CN107065834B (en) | The method for diagnosing faults of concentrator in hydrometallurgy process | |
Liu et al. | Unevenly sampled dynamic data modeling and monitoring with an industrial application | |
CN102282552A (en) | System, method and computer program for pattern based intelligent control, monitoring and automation | |
WO2011034805A1 (en) | Supervised fault learning using rule-generated samples for machine condition monitoring | |
CN111340110B (en) | Fault early warning method based on industrial process running state trend analysis | |
CN108875771A (en) | A kind of failure modes model and method being limited Boltzmann machine and Recognition with Recurrent Neural Network based on sparse Gauss Bernoulli Jacob | |
CN107463993A (en) | Medium-and Long-Term Runoff Forecasting method based on mutual information core principle component analysis Elman networks | |
Wang et al. | Fault detection and identification using a Kullback-Leibler divergence based multi-block principal component analysis and Bayesian inference | |
CN108875118B (en) | Method and device for evaluating accuracy of prediction model of silicon content of blast furnace molten iron | |
CN110297207A (en) | Method for diagnosing faults, system and the electronic device of intelligent electric meter | |
CN109308063A (en) | Fermentation process stage division method based on score matrix | |
CN109507972A (en) | Industrial processes fault monitoring method based on layer-stepping non-gaussian monitoring algorithm | |
CN109558993A (en) | Prediction technique, device, storage medium and the server of theory of crime prison term | |
CN111160393B (en) | Modularized modeling method of carrier rocket health evaluation model based on data driving | |
He et al. | Intelligent Fault Analysis With AIOps Technology | |
CN106778848A (en) | A kind of Zn finery method for diagnosing faults based on multiclass Probit models | |
CN107943966A (en) | Abnormal individual character decision method and device based on microblogging text | |
CN115186935B (en) | Electromechanical device nonlinear fault prediction method and system | |
CN110320802A (en) | Complication system signal sequence recognition methods based on data visualization | |
CN103268329A (en) | Plasma display screen manufacturing process data mining system |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |
|
RJ01 | Rejection of invention patent application after publication |