CN105843212B - A kind of blast furnace fault diagnosis system and method - Google Patents
A kind of blast furnace fault diagnosis system and method Download PDFInfo
- Publication number
- CN105843212B CN105843212B CN201610187770.0A CN201610187770A CN105843212B CN 105843212 B CN105843212 B CN 105843212B CN 201610187770 A CN201610187770 A CN 201610187770A CN 105843212 B CN105843212 B CN 105843212B
- Authority
- CN
- China
- Prior art keywords
- blast furnace
- fault state
- hypersphere
- feature
- attribute data
- 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.)
- Expired - Fee Related
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000012706 support-vector machine Methods 0.000 claims abstract description 60
- 239000011159 matrix material Substances 0.000 claims abstract description 31
- 238000010276 construction Methods 0.000 claims abstract description 6
- 238000004519 manufacturing process Methods 0.000 claims description 60
- 238000012549 training Methods 0.000 claims description 21
- 238000013480 data collection Methods 0.000 claims description 5
- 238000009529 body temperature measurement Methods 0.000 claims description 4
- 239000000463 material Substances 0.000 claims description 4
- 230000035699 permeability Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000002360 preparation method Methods 0.000 claims 1
- 230000008569 process Effects 0.000 abstract description 5
- 230000007257 malfunction Effects 0.000 abstract 2
- 238000005303 weighing Methods 0.000 abstract 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 4
- 230000009977 dual effect Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 229910052742 iron Inorganic materials 0.000 description 2
- 238000003723 Smelting Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
A kind of blast furnace fault diagnosis system of present invention offer and method, the system include：Historical data acquisition module, real data acquisition module, feature weight matrix construction module, model building module, blast furnace fault diagnosis module.This method includes：Acquire blast fumance situation actual attribute data, historical status data and its corresponding operation of blast furnace malfunction type；The feature weight of each attribute, construction feature weight matrix are determined for the significance level of fault diagnosis according to each attribute；Establish the twin Hypersphere Support Vector Machine model of characteristic weighing for blast furnace fault diagnosis；Blast fumance situation actual attribute data are brought into the twin Hypersphere Support Vector Machine model of each characteristic weighing of foundation, the affiliated operation of blast furnace malfunction type of blast fumance situation actual attribute data is obtained, complete blast furnace fault diagnosis.The present invention quantifies each feature importance of blast furnace failure, and each feature importance is incorporated in the building process of learning machine, to improve the precision of fault diagnosis.
Description
Technical Field
The invention belongs to the technical field of blast furnace fault diagnosis, and particularly relates to a blast furnace fault diagnosis system and method.
Background
The diagnosis of the fault of the blast furnace has important significance for reducing accidents and economic losses. The blast furnace smelting is to reduce the iron ore into iron, and is a continuous production process with complex process; in order to prevent abnormal conditions, a large number of parameters in the production process need to be monitored, such as hot air temperature, hot air quantity, hot air pressure, furnace top pressure, total pressure difference, upper pressure, lower pressure, oxygen-rich quantity, permeability index, cross temperature measurement, material speed, physical heat, Si content and the like; therefore, the fault state is characterized by the comprehensive embodiment of the signals with high-dimensional multi-characteristics; therefore, the conventional fault diagnosis method is not ideal for the fault diagnosis effect of the blast furnace. The fault diagnosis method based on machine learning and artificial intelligence is an effective method for diagnosing the faults of the blast furnace because the method does not depend on an accurate mathematical model. At present, an artificial intelligence method for diagnosing the faults of the blast furnace is mainly based on a neural network and a support vector machine. The support vector machine provides a new way for the intelligent development of the blast furnace fault diagnosis technology due to the good performance of the support vector machine, especially the good performance of the small sample identification problem.
The twin support vector machine is different from the traditional support vector machine, a pair of non-parallel planes is constructed by solving two small-scale quadratic programming problems, but not parallel planes in the support vector machine, and experimental results show that the learning efficiency is improved by 4 times on the premise of ensuring the precision. The twin hypersphere support vector machine is the latest improvement of the twin support vector machine, the model constructs two non-concentric hyperspaces in a characteristic space by solving two small-scale quadratic programming problems, each hypersphere contains one type of sample as far as possible and is far away from the other type of sample, and similar to the twin support vector machine, the model has higher learning efficiency because the model solves two small-scale quadratic programming problems but not one large-scale quadratic programming problem; compared with a twin support vector machine, the twin hypersphere support vector machine avoids the operation of inversion of two large-scale matrixes in the twin support vector machine, so that the twin hypersphere support vector machine has higher learning efficiency; because the twin hyper-sphere support vector machine uses a pair of hyper-spheres instead of a pair of non-parallel hyper-planes to describe the characteristics of the two types of samples, the strategy is more reasonable in practical application. Experimental results show that compared with a support vector machine and a twin support vector machine, the twin hypersphere support vector machine has higher learning efficiency and classification precision, and provides a new idea for a blast furnace fault diagnosis technology.
In the blast furnace fault diagnosis, the effect of each dimensional feature of a fault sample on classification is often different and sometimes very different. The traditional classification models comprise a support vector machine, a twin support vector machine and a twin hypersphere support vector machine, and the influence of the difference of the importance of each feature on classification is ignored when the classification model is constructed.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a blast furnace fault diagnosis system and method.
The technical scheme of the invention is as follows:
a blast furnace fault diagnosis system comprising:
a historical data acquisition module: acquiring historical attribute data of the production condition of the blast furnace and the corresponding operating fault state type of the blast furnace;
the actual data acquisition module: acquiring actual attribute data of the production condition of the blast furnace;
a feature weight matrix construction module: determining the weight of each attribute according to the importance degree of each attribute on fault diagnosis, and constructing a characteristic weight matrix;
a model building module: training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data by using a characteristic weight matrix, establishing a characteristic weighting twin supersphere support vector machine model, wherein the input of the model is the blast furnace production condition historical attribute data, the output of the model is one of the two blast furnace operation fault state types, a pair of characteristic weighting superspheres in the model respectively represent the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the characteristic weighting supersphere, the higher the probability of the corresponding blast furnace operation fault state;
the blast furnace fault diagnosis module: and (4) bringing the actual attribute data of the production condition of the blast furnace into the built characteristic weighted twin hypersphere support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and completing the fault diagnosis of the blast furnace.
The blast furnace fault diagnosis system also comprises a normalization processing module which is used for normalizing the blast furnace production condition historical attribute data collected by the historical data collection module, the blast furnace operation fault state type corresponding to the blast furnace production condition historical attribute data and the blast furnace production condition actual attribute data collected by the actual data collection module.
The attributes include: air quantity, air pressure, top pressure, pressure difference, air permeability, top temperature, cross temperature measurement, material speed, Si content and physical heat; the fault status types include: cool, hot, suspend and disintegrate.
The characteristic weight of the attribute is the difference between the information entropy of the fault state type and the information entropy of the fault state of the attribute.
The method for diagnosing the blast furnace faults by using the system comprises the following steps:
acquiring actual attribute data and historical attribute data of the production condition of the blast furnace and the corresponding operating fault state type of the blast furnace;
determining the characteristic weight of each attribute according to the importance degree of each attribute to fault diagnosis, and constructing a characteristic weight matrix;
establishing a feature weighted twin hypersphere support vector machine model for blast furnace fault diagnosis: training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data by using a characteristic weight matrix, establishing a characteristic weighting twin supersphere support vector machine model, wherein the input of the model is the blast furnace production condition historical attribute data, the output of the model is one of the two blast furnace operation fault state types, a pair of characteristic weighting superspheres in the model respectively represent the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the characteristic weighting supersphere, the higher the probability of the corresponding blast furnace operation fault state;
and (3) blast furnace fault diagnosis: and (4) bringing the actual attribute data of the production condition of the blast furnace into the built characteristic weighted twin hypersphere support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and completing the fault diagnosis of the blast furnace.
The types of the operating fault states of the blast furnace to which the obtained actual attribute data of the production condition of the blast furnace belong are as follows: and substituting the actual attribute data of the production condition of the blast furnace into each characteristic weighting twin hypersphere support vector machine model, and counting the running fault state types of the blast furnace output by each characteristic weighting twin hypersphere support vector machine model, wherein the running fault state type with the largest occurrence frequency of the blast furnace is the running fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs.
Has the advantages that:
the invention quantifies the importance of each characteristic of the blast furnace fault and integrates the importance of each characteristic into the construction process of the learning machine so as to improve the precision of fault diagnosis. On the basis of the twin hypersphere support vector machine, the feature importance, namely the feature weighting matrix, is introduced into the twin hypersphere support vector machine, the feature weight is calculated by using an information gain method, on the basis of ensuring the fault diagnosis training time, the fault diagnosis process is prevented from being influenced by some weakly related or unrelated features, and the fault diagnosis performance is more excellent.
Drawings
FIG. 1 is a block diagram of a blast furnace fault diagnosis system according to an embodiment of the present invention;
FIG. 2 is a flow chart of a blast furnace fault diagnosis method according to an embodiment of the present invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings.
In this embodiment, the attribute data includes: air volume (m3/min), air pressure (Pa), top pressure (MPa), pressure difference, air permeability, top temperature (including four-point temperature), cross temperature measurement (including center and edge, unit is ℃), material speed (unit is batch/hour), Si content, and physical heat (unit is ℃). Fault status types, including: cool, hot, suspend and disintegrate.
A blast furnace fault diagnosis system, as shown in fig. 1, comprising:
a historical data acquisition module: and acquiring historical attribute data of the production condition of the blast furnace and the corresponding operating fault state type of the blast furnace.
The actual data acquisition module: and acquiring actual attribute data of the production condition of the blast furnace.
A feature weight matrix construction module: and determining the characteristic weight of each attribute according to the importance degree of each attribute on fault diagnosis, and constructing a characteristic weight matrix.
A model building module: training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data by using a characteristic weight matrix, establishing a characteristic weighting twin supersphere support vector machine model, wherein the input of the model is the blast furnace production condition historical attribute data, the output of the model is one of the two blast furnace operation fault state types, a pair of characteristic weighting superspheres in the model respectively represent the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the characteristic weighting supersphere, the higher the probability of the corresponding blast furnace operation fault state is.
The blast furnace fault diagnosis module: and (4) bringing the actual attribute data of the production condition of the blast furnace into the built characteristic weighted twin hypersphere support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and completing the fault diagnosis of the blast furnace.
A normalization processing module: and normalizing the historical attribute data of the production condition of the blast furnace, which are acquired by the historical data acquisition module, the type of the operation fault state of the blast furnace corresponding to the historical attribute data of the production condition of the blast furnace, which are acquired by the actual data acquisition module.
The method for diagnosing the blast furnace fault by using the system comprises the following steps as shown in FIG. 2:
step 1, acquiring actual attribute data and historical attribute data of the production condition of the blast furnace and corresponding operating fault state types of the blast furnace;
in the embodiment, the acquired data is normalized by adopting a mean-variance standardization method, samples are marked manually, and the acquired historical attribute data of the production condition of the blast furnace and the corresponding type of the running fault state of the blast furnace are used as a training sample set T { (x)1，y1)，(x2，y2)，...，(xn，ym) In which xi∈X＝RdRepresenting the ith historical attribute data, namely the data vector of the ith training sample, wherein the value of n is 14 in the embodiment; y isiE, Y, represents a fault state type corresponding to the ith training sample, m represents a fault type, and the value of m is an integer, and m is 4; rdRepresenting the sample space, d representing the feature dimension of the sample;
the fault state types of the embodiment are four types, namely cold, hot, suspended and collapsed;
the cold fault state types are totally 800 training samples and are distributed by Gaussian N (mu)1，c1) Is generated, whereinCovariance matrix
The total number of 1000 training samples for a thermal fault condition type is 500 with a Gaussian distribution of N (μ)21，c21) Is generated, whereinCovariance matrixThe rest is composed of N (mu)22，c22) Is generated, whereinCovariance matrix
The suspension fault state types are 500 training samples in total and are distributed by Gaussian N (mu)3，c3) Is generated, whereinCovariance matrix
The types of the breakdown failure states total 400 training samples and are distributed by Gaussian N (mu)4，c4) Is generated, whereinA covariance matrix.
Step 2, determining the characteristic weight of each attribute according to the importance degree of each attribute to fault diagnosis, and constructing a characteristic weight matrix;
the larger the feature weight is, the more important the attribute is for fault diagnosis, and a feature weight matrix is constructed according to the feature weight of each attribute;
the characteristic weight of the attribute is the difference between the information entropy of the fault state type and the information entropy of the fault state of the attribute.
Feature weight matrix
Wherein,information entropy of fault state type corresponding to attribute
Of the fault state type, P (C)i) As a fault state type CiThe probability of occurrence;is characterized by ApHas a value ofiThe set of samples of (1).
Step 3, establishing a feature weighted twin hypersphere support vector machine model for blast furnace fault diagnosis: training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data by using a characteristic weight matrix, establishing a characteristic weighting twin supersphere support vector machine model, wherein the input of the model is the blast furnace production condition historical attribute data, the output of the model is one of the two blast furnace operation fault state types, a pair of characteristic weighting superspheres in the model respectively represent the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the characteristic weighting supersphere, the higher the probability of the corresponding blast furnace operation fault state;
a classification surface between the ith class of blast furnace fault state and the jth class of blast furnace fault state is constructed by adopting a feature weighting twin hypersphere support vector machine, the classification surface is determined by a pair of feature weighting hypersphere, and in order to solve the problem that the hypersphere is not distinguishable in the original space, a training sample is transformed by a certain amountMapping into a high-dimensional feature space Z such that the training samples are hyperspectral in the feature space. In a high-dimensional feature space, a feature weight matrix is introduced, and a feature weighted hypersphere of the ith fault state and a feature weighted hypersphere of the jth fault state are defined as follows:
and
wherein,as a feature weight matrix, ci(ij)And Ri(ij)Respectively the centre of sphere and radius of the i-th fault state in the high-dimensional feature space, cj(ij)And Rj(ij)Respectively the sphere center and radius of the jth fault state in the high-dimensional feature space.
The sphere centers and radii of two feature-weighted hypersphere in the high-dimensional feature space are determined by solving the following quadratic programming problem:
wherein ξi(ij)and xij(ij)Are relaxation factors and are used for respectively restraining positive and negative singular points l±Number of positive and negative type samples, c1、c2、v1、v2For the penalty factor, a cross-validation method is used for determination.
I(i)＝{k|yk＝i}，I(j)＝{k|ykJ denotes an index of attribute data corresponding to the i-th failure state and an index of attribute data corresponding to the j-th failure state, respectively.
As can be seen from the formulas (1) and (2), the obtained feature weighted hypersphere surrounds the attribute data corresponding to the ith fault state as much as possible, and is far away from the attribute data corresponding to the jth fault state, wherein the attribute with the larger feature weight is more important for the calculation of the distance, and the attribute with the smaller feature weight is less in contribution to the calculation of the distance; at the same time, the radius of the characteristic weighted hyper-sphere is required to be as small as possible, so that the characteristic weighted hyper-sphere is as tight as possible.
For convenience, the optimal values of equations (1) and (2) are not directly solved, but are obtained by solving the dual problem of equations (1) and (2). The dual problem of equations (1) and (2) is as follows:
wherein,andare Lagrange multipliers, K, respectivelyw(xi，xi) Referred to as the feature-weighted kernel function,
solving the dual problems (3) and (4) can result in
Center of sphere of feature weighted hypersphere for class i fault condition in high dimensional feature space
Radius of a feature weighted hypersphere for a class i fault condition in a high dimensional feature space
Center of sphere of feature weighted hypersphere for class j fault state in high dimensional feature space
Radius of a feature weighted hypersphere for a class j fault condition in a high dimensional feature space
Wherein,andare respectively provided withFor optimal values for the dual problems (3) and (4),
for any historical attribute data x, decision function Dii(x) Namely, the feature weighted twin hypersphere support vector machine model is defined as follows:
the decision function indicates that: the closer the historical attribute data to be classified is to which sphere center, and the larger the radius of the corresponding feature-weighted hypersphere is, the type of fault state to which the historical attribute data belongs.
Weighting the center c of the ball for the characteristics of the actual attribute data x to i-th failure statesi(ij)Characteristic weighted distance of (2):
weighting the center c of the ball for the characteristics of the actual attribute data x to the jth fault statej(ij)Characteristic weighted distance of (2):
by constructing the feature weighted twin hypersphere support vector machine model by using the method in any two fault states, a feature weighted twin hypersphere support vector machine model can be obtainedFeature weighted twin hypersphere support vector machine model Dij(x)。
Step 4, blast furnace fault diagnosis: and (4) bringing the actual attribute data of the production condition of the blast furnace into the built characteristic weighted twin hypersphere support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and completing the fault diagnosis of the blast furnace.
And substituting the actual attribute data of the production condition of the blast furnace into each characteristic weighting twin hypersphere support vector machine model, and counting the running fault state types of the blast furnace output by each characteristic weighting twin hypersphere support vector machine model, wherein the running fault state type with the largest occurrence frequency of the blast furnace is the running fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs.
Step 4-1, initializing a voting array Vote [ m ] ═ 0 };
step 4-2, substituting actual data of the production condition of the blast furnace into each characteristic weighted twin hypersphere support vector machine model Dij(x) In (1).
Step 4-3, if Dij(x)＝i，Vote[i]Self-increasing; otherwise, Vote [ j ]]Self-increasing;
4-4, the fault state type to which the actual attribute data x of the production condition of the blast furnace finally belongs is the fault state type with the highest ticket number, namely the fault state type to which the actual attribute data x belongs
Claims (4)
1. A blast furnace fault diagnosis system comprising:
a historical data acquisition module: acquiring historical attribute data of the production condition of the blast furnace and the corresponding operating fault state type of the blast furnace;
the actual data acquisition module: acquiring actual attribute data of the production condition of the blast furnace;
a model building module: establishing a support vector machine model, wherein the input of the model is the historical attribute data of the production condition of the blast furnace;
the blast furnace fault diagnosis module: the actual attribute data of the production condition of the blast furnace is brought into a support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and the fault diagnosis of the blast furnace is completed;
it is characterized by also comprising:
a feature weight matrix construction module: determining the characteristic weight of each attribute according to the importance degree of each attribute to fault diagnosis, and constructing a characteristic weight matrix; the characteristic weight of the attribute is the difference between the information entropy of the fault state type and the information entropy of the fault state of the attribute;
the model establishing module is a feature weighting twin hypersphere support vector machine model established by training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data by using a feature weight matrix, the output of the model is one of the two blast furnace operation fault state types, a pair of feature weighting hypersphere in the model respectively represents the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the feature weighting hypersphere, the higher the probability of the corresponding blast furnace operation fault state is;
a classification surface between the ith class of blast furnace fault state and the jth class of blast furnace fault state is constructed by adopting a feature weighted twin hypersphere support vector machine, the classification surface is determined by a pair of feature weighted hypersphere, and a training sample is transformed by a certain amountMapping the training samples into a high-dimensional feature space so that hypersphere of the training samples can be separated in the feature space, introducing a feature weight matrix into the high-dimensional feature space, and defining a feature weighted hypersphere of the ith fault state and a feature weighted hypersphere of the jth fault state respectively as followsAnd
where W is a feature weight matrix, ci(ij)And Ri(ij)Are respectively provided withCenter and radius of the i-th fault state in the high-dimensional feature space, cj(ij)And Rj(ij)Respectively the sphere center and the radius of the jth fault state in the high-dimensional characteristic space; determining the sphere centers and the radii of two feature-weighted hypersphere in a high-dimensional feature space, wherein the obtained feature-weighted hypersphere surrounds the attribute data corresponding to the ith fault state as much as possible and is far away from the attribute data corresponding to the jth fault state, wherein the attribute with the larger feature weight is more important for calculating the distance, and the attribute with the smaller feature weight is less in contribution to calculating the distance; meanwhile, the radius of the obtained characteristic weighted hypersphere is required to be as small as possible, so that the characteristic weighted hypersphere is as tight as possible; for any historical attribute data x, decision function Dij(x) Namely, the feature weighted twin hypersphere support vector machine model is defined as follows:the blast furnace fault diagnosis module is used for calculating the type of the blast furnace operation fault state output by each characteristic weighting twin hypersphere support vector machine model by bringing the actual attribute data of the production condition of the blast furnace into each established characteristic weighting twin hypersphere support vector machine model, and obtaining the type of the blast furnace operation fault state with the largest occurrence frequency, namely the type of the blast furnace operation fault state to which the actual attribute data of the production condition of the blast furnace belongs.
2. The blast furnace fault diagnosis system according to claim 1, further comprising a normalization processing module for normalizing the historical attribute data of the production condition of the blast furnace collected by the historical data collection module, the type of the operation fault state of the blast furnace corresponding to the historical attribute data of the production condition of the blast furnace collected by the actual data collection module, and the actual attribute data of the production condition of the blast furnace collected by the actual data collection module.
3. The blast furnace fault diagnostic system of claim 1, wherein the attributes comprise: air quantity, air pressure, top pressure, pressure difference, air permeability, top temperature, cross temperature measurement, material speed, Si content and physical heat; the fault status types include: cool, hot, suspend and disintegrate.
4. A method for blast furnace fault diagnosis using the system of claim 1, comprising:
acquiring actual attribute data and historical attribute data of the production condition of the blast furnace and the corresponding operating fault state type of the blast furnace;
establishing a support vector machine model, wherein the input of the model is the historical attribute data of the production condition of the blast furnace;
and (3) blast furnace fault diagnosis: the actual attribute data of the production condition of the blast furnace is brought into a support vector machine model to obtain the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs, and the fault diagnosis of the blast furnace is completed;
it is characterized in that the preparation method is characterized in that,
determining the characteristic weight of each attribute according to the importance degree of each attribute to fault diagnosis, and constructing a characteristic weight matrix; the characteristic weight of the attribute is the difference between the information entropy of the fault state type and the information entropy of the fault state of the attribute;
the method comprises the following steps that a support vector machine model is built, wherein the support vector machine model is a feature weighting twin hypersphere support vector machine model built by training any two blast furnace operation fault state types and corresponding blast furnace production condition historical attribute data through a feature weight matrix, the output of the model is one of the two blast furnace operation fault state types, a pair of feature weighting hypersphere in the model respectively represents the two blast furnace operation fault state types, and the closer the blast furnace production condition historical attribute data to be classified is to the feature weighting hypersphere, the higher the probability of the corresponding blast furnace operation fault state is;
a classification surface between the ith class of blast furnace fault state and the jth class of blast furnace fault state is constructed by adopting a feature weighted twin hypersphere support vector machine, the classification surface is determined by a pair of feature weighted hypersphere, and a training sample is transformed by a certain amountMapping to high-dimensional feature space to make training sample be hypersphere separable in feature space, introducing feature weight matrix in high-dimensional feature space,defining a characteristic weighted hypersphere for a class i fault condition and a characteristic weighted hypersphere for a class j fault condition as followsAnd
where W is a feature weight matrix, ci(ij)And Ri(ij)Respectively the centre of sphere and radius of the i-th fault state in the high-dimensional feature space, cj(ij)And Rj(ij)Respectively the sphere center and the radius of the jth fault state in the high-dimensional characteristic space; determining the sphere centers and the radii of two feature-weighted hypersphere in a high-dimensional feature space, wherein the obtained feature-weighted hypersphere surrounds the attribute data corresponding to the ith fault state as much as possible and is far away from the attribute data corresponding to the jth fault state, wherein the attribute with the larger feature weight is more important for calculating the distance, and the attribute with the smaller feature weight is less in contribution to calculating the distance; meanwhile, the radius of the obtained characteristic weighted hypersphere is required to be as small as possible, so that the characteristic weighted hypersphere is as tight as possible; for any historical attribute data x, decision function Dij(x) Namely, the feature weighted twin hypersphere support vector machine model is defined as follows:
the blast furnace fault diagnosis method specifically includes the steps of bringing actual attribute data of the production condition of the blast furnace into each established characteristic weighted twin hypersphere support vector machine model, counting the operation fault state types of the blast furnace output by each characteristic weighted twin hypersphere support vector machine model, and obtaining the operation fault state type of the blast furnace with the largest occurrence frequency, namely the operation fault state type of the blast furnace to which the actual attribute data of the production condition of the blast furnace belongs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610187770.0A CN105843212B (en) | 2016-03-29 | 2016-03-29 | A kind of blast furnace fault diagnosis system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610187770.0A CN105843212B (en) | 2016-03-29 | 2016-03-29 | A kind of blast furnace fault diagnosis system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105843212A CN105843212A (en) | 2016-08-10 |
| CN105843212B true CN105843212B (en) | 2018-09-28 |
Family
ID=56584045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610187770.0A Expired - Fee Related CN105843212B (en) | 2016-03-29 | 2016-03-29 | A kind of blast furnace fault diagnosis system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105843212B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109791401B (en) * | 2016-09-16 | 2022-07-22 | 西门子股份公司 | Generating fault models for embedded analytics and diagnostic/prognostic reasoning |
| CN107526784A (en) * | 2017-07-27 | 2017-12-29 | 上海电力学院 | A kind of method for diagnosing faults based on matrix fill-in |
| CN110059359A (en) * | 2019-03-21 | 2019-07-26 | 江苏东方国信工业互联网有限公司 | System and method for controlling furnace body process based on big data analysis |
| CN117030264B (en) * | 2023-08-10 | 2024-04-16 | 盐城工学院 | Rolling bearing fault diagnosis method based on feature weighted twin support vector machine |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9730663B2 (en) * | 2007-08-31 | 2017-08-15 | Koninklijke Philips N.V. | Uncertainty maps for segmentation in the presence of metal artifacts |
| CN101699432B (en) * | 2009-11-13 | 2015-01-21 | 黑龙江工程学院 | Ordering strategy-based information filtering system |
| CN103745234B (en) * | 2014-01-23 | 2017-01-25 | 东北大学 | Band steel surface defect feature extraction and classification method |
| CN104462846B (en) * | 2014-12-22 | 2017-11-10 | 山东鲁能软件技术有限公司 | A kind of equipment fault intelligent diagnosing method based on SVMs |
-
2016
- 2016-03-29 CN CN201610187770.0A patent/CN105843212B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN105843212A (en) | 2016-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021135630A1 (en) | Rolling bearing fault diagnosis method based on grcmse and manifold learning | |
| WO2020244134A1 (en) | Multi-task feature sharing neural network-based intelligent fault diagnosis method | |
| CN105843212B (en) | A kind of blast furnace fault diagnosis system and method | |
| CN109300111B (en) | Chromosome recognition method based on deep learning | |
| WO2021007801A1 (en) | Aluminum oxide comprehensive production index decision-making method based on multi-scale deep convolutional network | |
| CN111810124B (en) | Oil pumping well fault diagnosis method based on characteristic recalibration residual convolutional neural network model | |
| CN113988126B (en) | Rolling bearing fault diagnosis method based on few-label data feature migration | |
| CN105550744A (en) | Nerve network clustering method based on iteration | |
| CN105550426B (en) | A kind of multiple dimensioned binary tree blast furnace method for diagnosing faults based on sample decomposition | |
| CN108445752A (en) | A kind of random weight Artificial neural network ensemble modeling method of adaptively selected depth characteristic | |
| CN103258130B (en) | Sintered ore rotary drum strength Forecasting Methodology | |
| CN106778846A (en) | A kind of method for forecasting based on SVMs | |
| CN108460213A (en) | Based on the T-S models of multi-cluster prototype to the prediction technique and program of gas flowrate in bosh | |
| CN104899425A (en) | Variable selection and forecast method of silicon content in molten iron of blast furnace | |
| CN108181891A (en) | Industrial big data method for diagnosing faults based on intelligent core pivot element analysis | |
| CN114239404A (en) | Intelligent material optimization design method based on multi-objective optimization | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| Hou et al. | Improved multiclass support vector data description for planetary gearbox fault diagnosis | |
| CN110427019B (en) | Industrial process fault classification method and control device based on multivariate discriminant analysis | |
| Xu et al. | A novel hyperspectral image clustering method with context-aware unsupervised discriminative extreme learning machine | |
| CN115204041A (en) | Large aircraft approach and approach stage oil consumption prediction method | |
| CN110298385A (en) | One kind is based on * information and the online incipient fault detection method of increment SVDD | |
| CN111914471A (en) | Rectification tower fault detection method based on rapid nuclear independent component analysis | |
| Hu et al. | Research on the fault identification method of oil pumping unit based on residual network | |
| CN107066984A (en) | Algorithm for gait recognition based on subspace integrated study |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180928 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |