CN113237332B - Method for identifying working condition of electro-fused magnesia furnace - Google Patents
Method for identifying working condition of electro-fused magnesia furnace Download PDFInfo
- Publication number
- CN113237332B CN113237332B CN202110401590.9A CN202110401590A CN113237332B CN 113237332 B CN113237332 B CN 113237332B CN 202110401590 A CN202110401590 A CN 202110401590A CN 113237332 B CN113237332 B CN 113237332B
- Authority
- CN
- China
- Prior art keywords
- data
- working condition
- points
- new class
- point
- 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.)
- Active
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27B—FURNACES, KILNS, OVENS, OR RETORTS IN GENERAL; OPEN SINTERING OR LIKE APPARATUS
- F27B14/00—Crucible or pot furnaces
- F27B14/08—Details peculiar to crucible or pot furnaces
- F27B14/20—Arrangement of controlling, monitoring, alarm or like devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Mechanical Engineering (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method for identifying the working condition of an electric smelting magnesium furnace, which comprises the following steps: s1, acquiring online data in a preset period in the working condition of the electro-fused magnesia furnace; s2, clustering all continuous attribute data in the online data by using a CURD clustering algorithm; s3, screening data points outside the data extraction circle in each initial cluster according to a preset data extraction circle, forming a new class by the data points in the data extraction circle in each initial cluster, and classifying the screened data points into the new class closest to the selected data points to obtain a discretization result of continuous attribute data; and S4, matching the discretization result of the continuous attribute data and the original discretization attribute data in the on-line data with a working condition decision table, and taking the matching result as the working condition identification result of the current electro-fused magnesia furnace. The working condition identification result of the electro-fused magnesia furnace can be more accurate, and the identification method is faster and safer.
Description
Technical Field
The invention relates to the technical field of electric smelting magnesium furnace fault diagnosis, in particular to a working condition identification method for an electric smelting magnesium furnace.
Background
The smelting process of the electro-fused magnesia furnace is complex and changeable, the switching among a plurality of different working conditions is involved, the change of the smelting state can be caused due to the change of the impurity components and the particle length of the furnace burden, and if the position of the electrode is not adjusted timely, the abnormal working conditions can be caused, and the normal production of enterprises is influenced. Therefore, the method has important significance for safe and efficient production of enterprises by timely identifying the working condition of the electro-fused magnesia furnace.
The working condition identification of the electro-fused magnesia furnace in the actual smelting process can cause the following problems: firstly, in the smelting process of the electric smelting magnesium furnace, the field environment is severe, the current fluctuation is frequent, whether abnormal working conditions occur or not is judged by simply adopting manual experience, the long smelting process has great test for workers, and once a fault condition occurs, field operators often cannot accurately distinguish the fault condition in time; secondly, the electric smelting magnesium furnace is complex and changeable in operation, a large amount of information can be generated in the production process, the severe smelting environment enables the data acquired on site to be accompanied by much interference and noise, and when a fault occurs, the components of the system may not have problems, but because the particle size and impurity components of the raw materials are changed in the complex production process, the accurate and timely operation of the electric smelting magnesium furnace is not caused; the current working condition identification method of the electric smelting magnesium furnace only utilizes the current value generated by the electric smelting magnesium furnace to establish an expert knowledge base according to the change trend of the current, but the complicated and changeable field environment often causes the problem of low accuracy rate of working condition identification or untimely identification, thereby reducing the product yield.
Therefore, in order to solve the problem that the identification of the abnormal working conditions in the smelting process of the electric smelting magnesium furnace is not intelligent and accurate, a method for identifying the working conditions of the electric smelting magnesium furnace is urgently needed.
Disclosure of Invention
Technical problem to be solved
In view of the problems in the art described above, the present invention is at least partially addressed. Therefore, the invention provides the method for identifying the working condition of the electro-fused magnesia furnace, which can ensure that the identification result of the working condition of the electro-fused magnesia furnace is more accurate and the identification method is quicker and safer.
(II) technical scheme
In order to achieve the aim, the invention provides a method for identifying the working condition of an electro-fused magnesia furnace, which comprises the following steps:
and S1, acquiring online data in a preset period in the working condition of the electro-fused magnesia furnace.
And S2, clustering all continuous attribute data in the online data by adopting a CURD clustering algorithm to obtain initial clusters.
S3, screening data points outside the data extraction circle in each initial cluster according to a preset data extraction circle, wherein the data points inside the data extraction circle in each initial cluster form a new class; and classifying the screened data points into a new class closest to the data points to obtain a discretization result of the continuous attribute data.
S4, matching the discretization result of the continuous attribute data and the original discretization attribute data in the on-line data with a working condition decision table, and taking the matching result as the working condition identification result of the current electro-fused magnesia furnace; the working condition decision table comprises discrete attribute values and corresponding information of various working conditions, which are established in advance according to historical data of the working conditions of the electric smelting magnesium furnace.
Optionally, after S1 and before S2, further comprising: and S20, normalizing the data of each continuous attribute in the online data by using a linear conversion function.
Wherein the linear transfer function comprises:
in the formula, s' is the value after sample data normalization, s is sample data, and s isminIs the minimum value, s, in the data of the continuous attribute to be processedmaxIs the maximum value in the data of the continuous attribute to be processed.
Optionally, the data extraction circle includes a circle with a cluster center of the initial cluster as a center and a preset value as a radius.
Optionally, the preset value is α · Ri max(i ═ 1,2,. n); in the formula, alpha is more than or equal to 0<1，Ri maxAnd (i ═ 1,2,. n) is the farthest distance between a data point and the cluster center in the initial cluster, and n is the number of clusters.
Optionally, in S2, the classifying the screened data points into the new class closest thereto includes: and selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, and classifying the screened data points into the new class to which the representative point closest to the screened data points belongs.
Optionally, selecting a representative point of each new class from the data points located on the boundary of each new class includes: and sequentially selecting a preset number of data points from the data points on the boundary of each new class as representative points of each new class according to the distance from the data points to the clustering center of the new class and the descending order.
Optionally, S2 includes:
and S21, obtaining a reference point set according to the continuous attribute data, the distance threshold and the density threshold.
And S22, calculating the distance between each data point and each reference point, and mapping the data point and the reference point closest to the data point.
S23, determining adjacent reference points, describing a reference point set by using an undirected graph, enabling an edge to be arranged between the adjacent reference points, searching the reference points positioned on the same connected subgraph by adopting a breadth-first search algorithm of the graph, classifying the reference points positioned on the same connected subgraph into the same class, and obtaining a classification result of the reference points.
And S24, obtaining initial clustering of the continuous attribute data according to the mapping relation between the data points and the reference points and the classification result of the reference points.
Optionally, the working condition decision table includes corresponding information between discrete attribute values and various working conditions, which are established in advance according to historical data of the working conditions of the electric smelting magnesium furnace and based on the rough probability set.
Optionally, before S1, the method further includes:
and S01, acquiring historical data in a preset period in the working condition of the electro-fused magnesia furnace.
S02, processing all continuous attribute data in the history data in S2 and S3, and obtaining the discretization result of the continuous attribute data in the history data.
S03, establishing a discrete data table according to the discretization result of the continuous attribute data, original discrete attribute data in the historical data and corresponding decision attributes, and extracting each group of data and the working condition of each group of data according to the discrete data table to obtain an initial decision table; and modifying the target objects with the same condition attributes and different decision attributes in the initial decision table based on the probability rough set to obtain a condition decision table.
Optionally, modifying the target objects with the same condition attribute and different decision attributes in the initial decision table based on the rough probability set includes: and calculating the probability of the target object belonging to each working condition, and selecting the working condition with the maximum probability as the working condition identification result of the target object.
(III) advantageous effects
The invention has the beneficial effects that:
1. by adopting the CURD clustering algorithm to cluster all continuous attribute data in the electric smelting magnesium furnace working condition on-line data instead of clustering each continuous attribute data in the electric smelting magnesium furnace working condition on-line data, the information loss and errors are reduced, the discretization result is improved, and the effect of attribute reduction is achieved at the same time. And screening out data points outside the data extraction circle in each initial cluster through the data extraction circle, and reclassifying the screened data points, so that the problem of classification errors caused by a distance threshold in the CURD clustering algorithm is solved. Therefore, the method for identifying the working condition of the electro-fused magnesia furnace provided by the embodiment of the invention can identify the data cluster with any shape generated in the working condition of the electro-fused magnesia furnace, and the accuracy of data clustering is greatly improved. The method has important significance for reducing the energy consumption of a single ton of products, improving the smelting safety and improving the enterprise benefits.
2. In the method, the distance between each data point and each reference point is calculated, and the data point and the reference point closest to the data point are mapped, so that the problem of excessive invalid points in the CURD clustering algorithm is solved, the influence of isolated point data on a clustering result can be effectively eliminated, and the clustering result is more accurate.
3. The method of the invention aims at the defect that the rough set can not identify the target objects with the same condition attribute and different decision attributes, the probability of identifying the target objects as a certain abnormal working condition is obtained by utilizing the probability rough set, and the working condition with the maximum probability is taken as an identification result to identify the working condition. The working condition identification is more accurate.
Drawings
The invention is described with the aid of the following figures:
fig. 1 is a schematic flow chart of a method for identifying the working condition of an electric smelting magnesium furnace according to one embodiment of the present invention;
FIG. 2 is a schematic diagram of a first simulation result of a simulation test according to the present invention;
FIG. 3 is a schematic diagram of a second simulation result of a simulation test according to the present invention;
FIG. 4 is a diagram illustrating a third simulation result of a simulation test according to the present invention.
Detailed Description
For the purpose of better explaining the present invention and to facilitate understanding, the present invention will be described in detail by way of specific embodiments with reference to the accompanying drawings.
Clustering is a branch of data mining that partitions physical objects or abstract objects based on similarity to form similar and valuable groupings. The essence of the clustering algorithm is to divide the data set into clusters according to the density of the data set on the features, so that the similarity between the clusters is as small as possible and the similarity in the clusters is as large as possible. The K-means algorithm (K-means) is a classical partitional clustering algorithm that uses distance metric similarity to perform clustering according to the principle of objective function minimization. But the k-means algorithm is susceptible to the shape of the data cluster to be processed and is sensitive to isolated point data. Aiming at the problems faced by the k-means algorithm, a quick clustering algorithm based on reference points and density, namely a CURD (clustering using References and density) algorithm, uses a certain number of reference points with unfixed number to represent a clustering area and the shape thereof, finds a candidate reference point set, and screens out candidate reference points which do not meet the preset density threshold condition.
The method comprises the steps of firstly clustering all continuous attribute data in on-line data of the working condition of the electric smelting magnesium furnace by adopting a CURD clustering algorithm to obtain initial clusters, then screening out data points outside a data extraction circle in each initial cluster according to a preset data extraction circle, forming a new class by the data points in the data extraction circle in each initial cluster, classifying the screened data points into the new class closest to the data extraction circle to obtain a discretization result of the continuous attribute data, finally matching the discretization result of the continuous attribute data and original discretization attribute data in the on-line data with a working condition decision table, and taking the matching result as the working condition recognition result of the current electric smelting magnesium furnace.
By adopting the CURD clustering algorithm to cluster all continuous attribute data in the electric smelting magnesium furnace working condition on-line data instead of clustering each continuous attribute data in the electric smelting magnesium furnace working condition on-line data, the information loss and errors are reduced, the discretization result is improved, and the effect of attribute reduction is achieved at the same time. And screening out data points outside the data extraction circle in each initial cluster through the data extraction circle, and reclassifying the screened data points, so that the problem of classification errors caused by a distance threshold in the CURD clustering algorithm is solved. Therefore, the method for identifying the working condition of the electro-fused magnesia furnace provided by the embodiment of the invention can identify the data cluster with any shape generated in the working condition of the electro-fused magnesia furnace, and the accuracy of data clustering is greatly improved. The method has important significance for reducing the energy consumption of a single ton of products, improving the smelting safety and improving the enterprise benefits.
In order to better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The method for identifying the working condition of the electric smelting magnesium furnace provided by the embodiment of the invention is described below with reference to the attached drawings.
Fig. 1 is a schematic flow chart of a method for identifying the working condition of an electric magnesium melting furnace according to an embodiment of the present invention.
As shown in fig. 1, the method for identifying the working condition of the electro-fused magnesia furnace comprises the following steps:
and step S1, acquiring online data in a preset period in the working condition of the electro-fused magnesia furnace.
Step S2, the data of each continuous attribute in the online data is normalized by the linear conversion function.
Wherein the linear transfer function comprises:
in the formula, s' is the value after sample data normalization, s is sample data, and s isminIs the most in the data of the continuous attribute to be processedSmall value of smaxIs the maximum value in the data of the continuous attribute to be processed. The mapping of all variables to the value 0,1 is realized]And the subsequent data processing is convenient.
And step S3, clustering all continuous attribute data in the online data by using a CURD clustering algorithm to obtain initial clusters.
Because the production data of the electro-fused magnesia furnace has 9 continuous attributes, if the data of each continuous attribute are clustered respectively and then discretized, the process needs to be performed for 9 times for the production data of the electro-fused magnesia furnace, so that the whole discretization process is too complex, the calculation amount is huge, the loss of data information can be caused, and some unnecessary errors are caused. Therefore, in order to reduce information loss, reduce errors and improve discretization results, the method clusters all continuous attribute data together, realizes attribute discretization through one-time clustering, and achieves the effect of attribute reduction.
To illustrate the CURD algorithm, some definitions of the CURD algorithm are first given:
definition 1.1 (density of dots): let R be the distance threshold, and for any point a in space, the number of points in the radial region around point a, R being referred to as the density (density) of point a based on the distance threshold R, is denoted as Dens (a, R).
Definition 1.2 (reference point): and setting T as a density threshold, and if the condition Dens (a, R) is more than or equal to T for any point a in the space, calling the point a as a reference point. The reference point is not a point in the actual input data but a virtual point or a so-called phantom point.
Definition 1.3 (representative region): each reference point represents a circular area with the center of the point and the radius as the distance threshold R, and we refer to this area as a representative area of the reference point.
Definition 1.4 (adjacent reference point): for any point a, b in space, given a distance threshold R and a density threshold T, a reference point a, b is said to be an adjacent reference point if the reference point a, b satisfies the condition that the distance between a and b is less than or equal to 2 times the distance threshold.
Definition 1.5 (breadth first search): the idea of the method is that the method starts from a node on a graph, accesses the child nodes directly connected with the node, if the child nodes do not conform to the child nodes, asks the child nodes of the child nodes to access the child nodes in sequence according to the level order until the target node is accessed.
As an example, a CURD clustering algorithm is adopted to cluster all continuous attribute data in online data, and the method comprises the following steps:
and S31, obtaining a reference point set according to the continuous attribute data, the distance threshold and the density threshold.
Specifically, S31 includes:
s311, setting the data set M to be clustered as M1,...,mnThe first set of data m in1As a candidate set of reference points.
S312, calculating the next group of data m2If the distances between the data points and all the candidate reference points are greater than the distance threshold value R, adding the data points into the candidate reference point set D ═ D1,...,dxAnd if the distance between the data point and the target candidate reference point is smaller than a distance threshold value R, adding the data point into a representative area of the target candidate reference point, and according to the distance threshold value R, adding the data point into the representative area of the target candidate reference point
And averaging to update the positions of the target candidate reference points, and iteratively performing S312 until all data in the set M are compared, thereby obtaining a final candidate reference point set.
Where x is the number of uncertain candidate reference points,
for the updated coordinates of the reference point, diIs the coordinate of the original reference point.
S313, comparing the density of each candidate reference point in the final candidate reference point set with a density threshold T, and adding the candidate reference points meeting the condition that Dens (a, R) is more than or equal to T into the reference point set.
And S32, calculating the distance between each data point and each reference point, and mapping the data point and the reference point closest to the data point.
Compared with the conventional CURD algorithm, the method has the advantages that the distance between each data point and each reference point is calculated and compared with the distance threshold, if the distance is smaller than the distance threshold, mapping is established between the corresponding data point and the reference point, and if the distance is larger than the distance threshold, the data point is marked as an invalid point. In the method, the distance between each data point and each reference point is calculated, and the data point and the reference point closest to the data point are mapped, so that the problem of excessive invalid points in the CURD clustering algorithm is solved, the influence of isolated point data on a clustering result can be effectively eliminated, and the clustering result is more accurate.
S33, determining adjacent reference points, describing a reference point set by using an undirected graph, enabling an edge to be arranged between the adjacent reference points, searching the reference points positioned on the same connected subgraph by adopting a breadth-first search algorithm of the graph, classifying the reference points positioned on the same connected subgraph into the same class, and obtaining a classification result of the reference points.
Specifically, determining the adjacent reference points includes: if any two reference points d in the reference point set1And d2Is less than or equal to 2R, then d1And d2Are adjacent reference points.
And S34, obtaining initial clustering of the continuous attribute data according to the mapping relation between the data points and the reference points and the classification result of the reference points.
S4, screening out data points outside the data extraction circle in each initial cluster according to a preset data extraction circle, wherein the data points inside the data extraction circle in each initial cluster form a new class; and classifying the screened data points into a new class closest to the data points to obtain a discretization result of the continuous attribute data.
Preferably, the data extraction circle comprises a circle which takes the cluster center of the initial cluster as the center of a circle and takes the preset value as the radius. Further, the preset value is α · Ri max(i ═ 1,2,. n); in the formula, alpha is more than or equal to 0< 1，Ri maxN is the data point in the initial cluster and the cluster center thereofThe farthest distance, n is the number of clusters.
Preferably, the selected data points are classified into the new class closest thereto, including: and selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, and classifying the screened data points into the new class to which the representative point closest to the screened data points belongs. Further, selecting a representative point of each new class from the data points located on the boundary of each new class includes: and sequentially selecting a preset number of data points from the data points on the boundary of each new class as representative points of each new class according to the distance from the data points to the clustering center of the new class and the descending order.
Step S5, matching the discretization result of the continuous attribute data and the original discretization attribute data in the on-line data with a working condition decision table, and taking the matching result as the working condition identification result of the current electro-fused magnesia furnace; the working condition decision table comprises discrete attribute values and corresponding information of various working conditions, which are established in advance according to historical data of the working conditions of the electric smelting magnesium furnace.
Preferably, the working condition decision table comprises corresponding information of discrete attribute values and various working conditions established in advance according to historical data of the working conditions of the electric smelting magnesium furnace and based on the rough probability set.
The rough set is a qualitative model, and the definition of the upper approximation set and the lower approximation set is limited to the case that the conditional probability takes 0 and 1, and no distinction is given to the case that the conditional probability is between 0 and 1, so that the classical rough set model lacks the fault tolerance capability in practical application. The probabilistic rough set model makes up for the classical rough set model's inadequacies in solving the knowledge uncertainty decision problem, in other words, the probabilistic rough set enhances the processing power of objects that are in the boundary threshold. Therefore, the method of the invention aims at the defect that the rough set can not identify the target objects with the same condition attribute and different decision attributes, the probability of identifying the target objects as a certain abnormal working condition is obtained by utilizing the rough probability set, and the working condition with the maximum probability is taken as an identification result to identify the working condition. The working condition identification is more accurate.
Therefore, the method further includes the following steps before step S1:
and S01, acquiring historical data in a preset period in the working condition of the electro-fused magnesia furnace.
S02, processing all the continuous attribute data in the history data in the steps S2 and S3, and obtaining the discretization result of the continuous attribute data in the history data.
S03, establishing a discrete data table according to the discretization result of the continuous attribute data, original discrete attribute data in the historical data and corresponding decision attributes, and extracting each group of data and the working condition of each group of data according to the discrete data table to obtain an initial decision table; and modifying the target objects with the same condition attributes and different decision attributes in the initial decision table based on the probability rough set to obtain a condition decision table.
Specifically, modifying the target objects with the same condition attributes and different decision attributes in the initial decision table based on the rough probability set includes: and calculating the probability of the target object belonging to each working condition, and selecting the working condition with the maximum probability as the working condition identification result of the target object.
Simulation test
The method for identifying the working condition of the electro-fused magnesia furnace is applied to an electro-fused magnesia furnace working condition identification simulation platform, and the platform comprises three parts: the system comprises a database, a working condition identification interface and a working condition identification algorithm. The database adopts a version SQL services 2017 and is responsible for accessing the real-time information generated in the smelting process of the electro-fused magnesia furnace; making a working condition identification interface by using Visual Studio2017, wherein the working condition identification interface is used for displaying real-time information and an identification result; the algorithm is compiled and debugged by using MATLAB2017, the real-time information of the smelting process is identified by the working condition identification algorithm, and the result is returned to the interface.
2500 groups of samples are collected from the working condition of the electro-fused magnesia furnace to be used as a total sample, 1500 groups of samples are selected from the total sample to be used as a training set to establish a working condition decision table, and the remaining 1000 groups of samples are used as a test set. The training set comprises 5 working conditions shown in table 1, each working condition comprises 300 groups and comprises 3 discrete attributes and 9 continuous attributes, and the discrete attributes and the continuous attributes are respectively shown in tables 2 and 3 below. The test set also included 5 conditions as shown in table 1, with 200 sets for each condition.
TABLE 1 working conditions and numbering of the fused magnesia furnace production process
TABLE 2 discrete attributes and discrete values
TABLE 3 continuous Properties in the production of electro-fused magnesia furnaces
Clustering all continuous attribute data in the training set by using a CURD clustering algorithm to obtain initial clusters; calculating the cluster center of each initial cluster, screening out data points in each initial cluster, wherein the distance between each initial cluster and the cluster center exceeds a first preset threshold, and the rest data points in each initial cluster form a new class; selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, classifying the screened data points into the new class to which the representative point closest to the screened data points belongs, and obtaining a discretization result of the continuous attribute data, wherein the obtained result can be represented as follows:
in the formula, D*Representing discretization results of continuous attribute data in training set, m1、m2、 ...、mxRespectively represent the number of samples of the discretization result, and m1+m2+...+mx1500. The discretization result of the specifically obtained continuous attribute data is shown in table 4.
TABLE 4 discretization results of continuous Attribute data in training sets
The discretized 9 continuous attributes can be used as a discrete attribute for a decision table, which is equivalent to the completion of attribute reduction, so that the step of attribute reduction can be omitted.
Establishing a discrete data table according to a discretization result of the continuous attribute data, original discrete attribute data in the training set and corresponding decision attributes, and extracting each group of data and the working condition of each group of data according to the discrete data table to obtain an initial decision table; and modifying the target objects with the same condition attributes and different decision attributes in the initial decision table based on the probability rough set to obtain a first condition decision table, as shown in table 5.
TABLE 5 decision table for first working condition of electro-fused magnesia furnace
The discretization result of all continuous attribute data in the training set obtained by the K-means clustering algorithm is combined with the original discrete attribute data, the corresponding decision attribute and expert knowledge in the training set, and the obtained second working condition decision table is shown in table 6.
TABLE 6 decision table for second working condition of electro-fused magnesia furnace
1. Discretizing the continuous attribute data in the test set by a K-means clustering method to obtain a discretization result of the continuous attribute data, matching the discretization result of the continuous attribute data and the original discretization attribute data in the test set with a second working condition decision table, and taking the matching result as a working condition identification result of the current electro-fused magnesia furnace, as shown in table 7, and as shown in a simulation chart of fig. 2.
Table 7 operating mode identification result 1 of electric magnesium melting furnace test data
2. Clustering all continuous attribute data in the test set through a CURD clustering algorithm to obtain initial clusters; calculating the cluster center of each initial cluster, screening out data points in each initial cluster, wherein the distance between each initial cluster and the cluster center exceeds a first preset threshold, and the rest data points in each initial cluster form a new class; and selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, classifying the screened data points into the new class to which the representative point closest to the screened data points belongs, and obtaining the discretization result of the continuous attribute data. And matching the discretization result of the continuous attribute data and the original discretization attribute data in the test set with a first working condition decision table, taking the matching result as the working condition identification result of the current electro-fused magnesia furnace, as shown in table 8, and as shown in a simulation graph of fig. 3.
Table 8 operating mode identification result 2 of electric magnesium melting furnace test data
3. The method comprises the steps of obtaining an online data set in the working condition of the electro-fused magnesia furnace, wherein the online data set comprises 250 groups of samples, namely 150 groups of normal smelting data, 50 groups of exhaust abnormal data and 50 groups of charging abnormal data. Clustering all continuous attribute data in the online data set through a CURD clustering algorithm to obtain initial clusters; calculating the cluster center of each initial cluster, screening out data points in each initial cluster, wherein the distance between each initial cluster and the cluster center exceeds a first preset threshold, and the rest data points in each initial cluster form a new class; and selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, classifying the screened data points into the new class to which the representative point closest to the screened data points belongs, and obtaining the discretization result of the continuous attribute data. And matching the discretization result of the continuous attribute data and the original discretization attribute data in the test set with a first working condition decision table, taking the matching result as the working condition identification result of the current electro-fused magnesia furnace, as shown in table 9, and as shown in a simulation graph of fig. 4.
TABLE 9 Condition recognition results of on-line data of fused magnesia furnaces
From the results in table 7 and table 8, it can be seen that the test data condition identification rate obtained by the attribute discretization method based on the improved cut plus inner circle representative point algorithm is higher than the test data condition identification rate obtained by the attribute discretization method based on the K-means cluster, and the online condition identification rate obtained through table 9 indicates that the attribute discretization method based on the improved cut plus inner circle representative point algorithm is effective.
The attribute discretization based on the improved CURD and inner circle representative point algorithm reduces the calculated amount, only one discretization is needed for all continuous attributes, the generated final decision table has fewer influence variables, the complexity of identifying the working condition in the production process is effectively reduced, the time required by the working condition identifying process is reduced, the reaction is quicker when the method is applied to actual production, abnormal working condition signals can be reflected to a control system more quickly, and the abnormality is eliminated in time. Meanwhile, the working condition identification method based on the rough probability set has high accuracy in online identification, is a feasible working condition identification method, and has certain practical application value.
It should be understood that the above description of specific embodiments of the present invention is only for the purpose of illustrating the technical lines and features of the present invention, and is intended to enable those skilled in the art to understand the contents of the present invention and to implement the present invention, but the present invention is not limited to the above specific embodiments. It is intended that all such changes and modifications as fall within the scope of the appended claims be embraced therein.
Claims (9)
1. A working condition identification method for an electro-fused magnesia furnace is characterized by comprising the following steps:
s1, acquiring online data in a preset period in the working condition of the electro-fused magnesia furnace;
s2, clustering all continuous attribute data in the online data by using a CURD clustering algorithm to obtain initial clusters;
s3, screening data points outside the data extraction circle in each initial cluster according to a preset data extraction circle, wherein the data points inside the data extraction circle in each initial cluster form a new class;
classifying the screened data points into a new class closest to the data points to obtain a discretization result of the continuous attribute data;
s4, matching the discretization result of the continuous attribute data and the original discretization attribute data in the on-line data with a working condition decision table, and taking the matching result as the working condition identification result of the current electro-fused magnesia furnace; the working condition decision table comprises discrete attribute values and corresponding information of various working conditions, which are established in advance according to historical data of the working conditions of the electro-fused magnesia furnace;
wherein S2 includes:
s21, obtaining a reference point set according to the continuous attribute data, the distance threshold and the density threshold;
s22, calculating the distance between each data point and each reference point, and establishing mapping between the data point and the reference point closest to the data point;
s23, determining adjacent reference points, describing a reference point set by using an undirected graph, enabling an edge to be arranged between the adjacent reference points, searching the reference points positioned on the same connected subgraph by adopting a breadth-first search algorithm of the graph, classifying the reference points positioned on the same connected subgraph into the same class, and obtaining a classification result of the reference points;
and S24, obtaining initial clustering of the continuous attribute data according to the mapping relation between the data points and the reference points and the classification result of the reference points.
2. The method of claim 1, further comprising, after S1 and before S2:
s20, normalizing the data of each continuous attribute in the online data by using a linear conversion function;
wherein the linear transfer function comprises:
in the formula, s' is the value after sample data normalization, s is sample data, and s isminIs the minimum value, s, in the data of the continuous attribute to be processedmaxIs the maximum value in the data of the continuous attribute to be processed.
3. The method of claim 1, wherein the data extraction circle comprises a circle with a cluster center of the initial cluster as a center and a preset value as a radius.
4. Method according to claim 3, characterized in that the preset value is α -Ri max(i ═ 1,2,. n); in the formula, alpha is more than or equal to 0<1，Ri maxAnd (i ═ 1,2,. n) is the farthest distance between a data point and the cluster center in the initial cluster, and n is the number of clusters.
5. The method of claim 1, wherein the step of classifying the screened data points into the nearest new class at S2 comprises:
and selecting a representative point of each new class from the data points on the boundary of each new class, calculating the distance from each screened data point to each new class representative point, and classifying the screened data points into the new class to which the representative point closest to the screened data points belongs.
6. The method of claim 5, wherein selecting a representative point for each new class from the data points that lie on the boundary of each new class comprises:
and sequentially selecting a preset number of data points from the data points on the boundary of each new class as representative points of each new class according to the distance from the data points to the clustering center of the new class and the descending order.
7. The method according to claim 1, wherein the working condition decision table comprises corresponding information of discrete attribute values and various working conditions which are established in advance according to historical data of the working conditions of the electric smelting magnesium furnace and based on a probability rough set.
8. The method of claim 1, further comprising, prior to S1:
s01, acquiring historical data in a preset period in the working condition of the electro-fused magnesia furnace;
s02, processing all continuous attribute data in the historical data in S2 and S3 to obtain discretization results of the continuous attribute data in the historical data;
s03, establishing a discrete data table according to the discretization result of the continuous attribute data, original discrete attribute data in the historical data and corresponding decision attributes, and extracting each group of data and the working condition of each group of data according to the discrete data table to obtain an initial decision table;
and modifying the target objects with the same condition attributes and different decision attributes in the initial decision table based on the probability rough set to obtain a condition decision table.
9. The method of claim 8, wherein modifying the target objects with the same condition attribute and different decision attributes in the initial decision table based on the rough set of probabilities comprises:
and calculating the probability of the target object belonging to each working condition, and selecting the working condition with the maximum probability as the working condition identification result of the target object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110401590.9A CN113237332B (en) | 2021-04-14 | 2021-04-14 | Method for identifying working condition of electro-fused magnesia furnace |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110401590.9A CN113237332B (en) | 2021-04-14 | 2021-04-14 | Method for identifying working condition of electro-fused magnesia furnace |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113237332A CN113237332A (en) | 2021-08-10 |
| CN113237332B true CN113237332B (en) | 2022-02-25 |
Family
ID=77128287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110401590.9A Active CN113237332B (en) | 2021-04-14 | 2021-04-14 | Method for identifying working condition of electro-fused magnesia furnace |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113237332B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116110516B (en) * | 2023-04-14 | 2023-07-21 | 青岛山青华通环境科技有限公司 | Method and device for identifying abnormal working conditions in sewage treatment process |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599450B (en) * | 2016-12-12 | 2020-04-07 | 东北大学 | Priori knowledge-based fault monitoring method for flexible manifold embedded electric smelting magnesium furnace |
| CN107092923B (en) * | 2017-03-22 | 2019-10-11 | 东北大学 | The electric melting magnesium furnace process monitoring method of method is locally linear embedding into based on improvement supervision core |
| CN107065828B (en) * | 2017-03-24 | 2019-05-21 | 东北大学 | The semi-supervised kernel discriminant analysis method for diagnosing faults of knowledge based data |
| CN106959662B (en) * | 2017-05-10 | 2019-03-08 | 东北大学 | A kind of identification of electric melting magnesium furnace unusual service condition and control method |
-
2021
- 2021-04-14 CN CN202110401590.9A patent/CN113237332B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113237332A (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101927910B1 (en) | System and method for predicting disease inforamtion using deep neural network | |
| CN109273096B (en) | Medicine risk grading evaluation method based on machine learning | |
| CN111028224A (en) | Data labeling method, model training device, image processing method, image processing device and storage medium | |
| CN112288706A (en) | Automatic chromosome karyotype analysis and abnormality detection method | |
| Arbin et al. | Comparative analysis between k-means and k-medoids for statistical clustering | |
| CN112987666B (en) | Power plant unit operation optimization regulation and control method and system | |
| CN111638707B (en) | Intermittent process fault monitoring method based on SOM clustering and MPCA | |
| Wang et al. | Imbalance data processing strategy for protein interaction sites prediction | |
| CN113237332B (en) | Method for identifying working condition of electro-fused magnesia furnace | |
| CN108416373A (en) | A kind of unbalanced data categorizing system based on regularization Fisher threshold value selection strategies | |
| CN112837754B (en) | Method and device for automatically classifying single cells based on characteristic genes | |
| Chakradeo et al. | Breast cancer recurrence prediction using machine learning | |
| CN111128300B (en) | Protein interaction influence judgment method based on mutation information | |
| CN116738551B (en) | Intelligent processing method for acquired data of BIM model | |
| CN111414930A (en) | Deep learning model training method and device, electronic equipment and storage medium | |
| CN116365519B (en) | Power load prediction method, system, storage medium and equipment | |
| CN115378000A (en) | Power distribution network operation state evaluation method based on interval two-type fuzzy clustering analysis | |
| CN112148605B (en) | Software defect prediction method based on spectral clustering and semi-supervised learning | |
| CN111125082B (en) | Cement process parameter data analysis method based on association rule mining | |
| CN113505818A (en) | Aluminum melting furnace energy consumption abnormity diagnosis method, system and equipment with improved decision tree algorithm | |
| CN110533080B (en) | Fuzzy rule set-based breast cancer cell image classification method | |
| CN114597886A (en) | Power distribution network operation state evaluation method based on interval type two fuzzy clustering analysis | |
| Ressom et al. | Applications of fuzzy logic in genomics | |
| CN106778848A (en) | A kind of Zn finery method for diagnosing faults based on multiclass Probit models | |
| To et al. | Analysis of cardiac imaging data using decision tree based parallel genetic programming |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |