CN113688895A - Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA - Google Patents

Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA Download PDF

Info

Publication number
CN113688895A
CN113688895A CN202110957181.7A CN202110957181A CN113688895A CN 113688895 A CN113688895 A CN 113688895A CN 202110957181 A CN202110957181 A CN 202110957181A CN 113688895 A CN113688895 A CN 113688895A
Authority
CN
China
Prior art keywords
data set
keca
simplified
euclidean distance
roller kiln
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110957181.7A
Other languages
Chinese (zh)
Inventor
许潇
杨海东
徐康康
雷绍俊
程明阳
印四华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN202110957181.7A priority Critical patent/CN113688895A/en
Publication of CN113688895A publication Critical patent/CN113688895A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Evolutionary Biology (AREA)
  • Mathematical Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Optimization (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computing Systems (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention discloses a method and a system for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA, wherein the method comprises the following steps: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set; constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit; and carrying out anomaly detection on the test data set by utilizing a simplified KECA model. According to the invention, the simplified KECA model is constructed by reconstructing the kernel matrix of the data set, so that the complexity and the calculation cost of the model are reduced, and the real-time performance and the accuracy of detection are improved while the detection performance of the model is improved.

Description

Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA
Technical Field
The invention relates to the technical field of ceramic roller kiln burning zone abnormity detection, in particular to a simplified KECA-based ceramic roller kiln burning zone abnormity detection method and system.
Background
High energy consumption and low energy efficiency in the ceramic production process are major factors that restrict the development of the ceramic industry. In the ceramic production process, the firing process of the ceramic product in the roller kiln accounts for 70-80% of the total production energy consumption. As the roller kiln serving as core equipment in the ceramic firing process has complex firing zone working conditions and is in a full-load operation state for a long time, the possibility of abnormity is high, and the abnormity is difficult to diagnose; the problems of low production efficiency, poor product quality and the like can be caused in the production process, and the energy utilization rate of the ceramic industry is seriously influenced. Researches find that two methods for improving the production efficiency and the product quality on the basis of safe operation are mainly used, namely, the existing control technology is improved and an advanced prediction maintenance method is adopted. After the existing control technology is improved, the improvement of the production efficiency is lower than 5%, the production efficiency can be improved by 20-40% by the advanced predictive maintenance method, and the anomaly detection is a supporting technology of the predictive maintenance method. Therefore, the adoption of the anomaly detection method for timely and effectively detecting and diagnosing the anomaly condition in the firing process is the key for ensuring the normal and efficient operation of the roller kiln and improving the energy efficiency of the roller kiln.
At present, the detection method and means of the ceramic industry aiming at the abnormal working condition of the roller kiln are very limited. The method is used for investigating and analyzing the operation site of the roller kiln of a certain ceramic enterprise, and finds that although a digital acquisition device and a central control system are introduced into the enterprise, the enterprise still depends on manual work when setting the operation of kiln process parameters, roller bed maintenance, fan adjustment and the like, and particularly, the abnormal maintenance of gas, roller bed operation state, heat insulation state and the like in the sintering process is completely dependent on manual inspection, so that the method has slow response to sudden abnormal events and lower efficiency of processing the abnormal events. The development of the artificial intelligence technology provides a new idea for the abnormity detection of the ceramic roller kiln. By constructing the roller kiln sintering zone abnormity detection model, the monitoring of the roller kiln sintering zone abnormity state and the positioning of the abnormity position are realized, the energy loss in the abnormity state is reduced, and the method has important significance for reducing the production cost of enterprises and promoting the sustainable development of the ceramic industry.
In the prior art, the publication numbers are: the CN110457550A chinese invention patent discloses a method for correcting abnormal operation data in the sintering process. Firstly, carrying out anomaly detection on operation data in a sintering process by adopting a box chart method to obtain a plurality of anomaly data; clustering and dividing the normally running historical data by using a kernel fuzzy C-means clustering algorithm to obtain a plurality of categories; then, Euclidean distances between the abnormal operation data and different clustering centers are calculated, and a plurality of categories of the abnormal operation data are obtained according to the category corresponding to the minimum Euclidean distance; and finally, obtaining the historical data of normal operation which is most similar to the abnormal operation data in the category of the abnormal operation data by using a nearest neighbor algorithm based on the Mahalanobis distance to correct the abnormal operation data, and storing the corrected abnormal operation data into a historical database of the sintering process. The method is used for detecting and correcting the abnormity based on the fuzzy mean value clustering idea, and the energy consumption abnormity cannot be accurately detected in real time.
Disclosure of Invention
The invention provides a method and a system for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA, aiming at overcoming the defects that in the roller kiln abnormal detection method in the prior art, the detection model is high in complexity, the detection cost is high, and the abnormality cannot be detected accurately in real time.
The primary objective of the present invention is to solve the above technical problems, and the technical solution of the present invention is as follows:
the invention provides a method for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA, which comprises the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
s2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
acquiring a test data set and preprocessing the test data set;
s3: and carrying out anomaly detection on the test data set by utilizing a simplified KECA model.
Further, the step S1 specifically includes:
collecting process variable data of a roller kiln sintering zone system in a normal state as a training data set, and recording the process variable data as a first data set;
and carrying out normalization processing on the first data set, calculating the Euclidean distance of the first data set, and simplifying the training data set according to a preset Euclidean distance threshold value to obtain a second data set.
Further, the normalization processing is performed on the first data set, the euclidean distance of the first data set is calculated, the training data set is simplified according to a preset euclidean distance threshold, and a second data set is obtained, and the specific process is as follows:
constructing the first data set into a sample matrix, and carrying out normalization processing on the sample matrix;
calculating Euclidean distances between two different rows of the normalized sample matrix;
comparing the Euclidean distances between two different rows with a preset Euclidean distance threshold value one by one;
if the Euclidean distance between two different rows is smaller than a preset Euclidean distance threshold value, deleting or combining two rows of samples corresponding to the Euclidean distance; if the Euclidean distance between two different rows is larger than a preset Euclidean distance threshold value, two rows of sample data corresponding to the Euclidean distance are reserved, and a new data set consisting of all the reserved sample data is recorded as a second data set.
Further, the euclidean distance threshold is calculated by the following formula:
Figure BDA0003220781630000031
wherein d islRepresenting the Euclidean distance threshold, N representing the number of rows of the sample matrix, di,jRepresenting the euclidean distance between the ith and jth lines.
Further, in step S2, the step of constructing the KECA model by using the second data set, that is, the step of reducing the dimension of the second data set in the principal component direction by using the KECA method, includes the specific steps of:
normalizing the second data set;
calculating the Renyi entropy of the samples in the second data set after normalization processing;
calculating the contribution rate of each entropy by using each Renyi entropy, determining the contribution rate threshold value of the entropy,
and obtaining the number k of the principal components according to the contribution rate threshold of the entropy, and mapping the data to the directions of the k kernel principal components to obtain a second data set after dimension reduction.
Further, the specific process for determining the number of the principal components is as follows:
x is set in the data set D1,…,xNIs generated by the probability density function p (x), the second order Renyi entropy of the samples in the dataset is defined as:
H(p)＝-log∫p2(x)dx (2)
from equation (2), let:
V(p)＝∫p2(x)dx (3)
the logarithmic function in the formula (2) is an increasing type function, then the estimated value of H (p) can be obtained through the estimated value of V (p), and V (p) is obtained through the estimation by calling a Parzen window estimator
Figure BDA0003220781630000032
Probability density function when invoking Parzen window estimator
Figure BDA0003220781630000033
As shown in equation 4:
Figure BDA0003220781630000034
wherein k isσ(x,xt) A kernel function for the feature space, usually a radial basis function, can be expressed as:
Figure BDA0003220781630000035
wherein σ is a parameter of the kernel function, and the desired operator is estimated by the sample mean, so that:
Figure BDA0003220781630000041
Figure BDA0003220781630000042
represents an estimated value of V (p); convert equation (6) to:
Figure BDA0003220781630000043
where K is a sample kernel matrix of N × N, 1 is a vector of N × 1, and the Renyi entropy is represented by eigenvalues of the kernel matrix and corresponding eigenvectors, where the kernel matrix may be characterized as:
K＝EDET (8)
wherein D is a characteristic value lambda1,…,λNE is a feature vector E1,…,eNIs a matrix of columns, ETIs the transpose of E, so the estimate of V (p) can be expressed as:
Figure BDA0003220781630000044
mapping of sample points to a new data set phi formed in the direction of k kernel principal componentsecaComprises the following steps:
Figure BDA0003220781630000045
calculating the contribution rate eta of Renyi entropy:
Figure BDA0003220781630000046
wherein m is a mean vector of the kernel feature space data set, and the expression of m is as follows:
Figure BDA0003220781630000047
mecais indicative of phiecaMean vector of transformed data, mecaThe expression is as follows:
Figure BDA0003220781630000048
calculating the contribution rate of Renyi entropy of each sample, sequencing the entropy contribution rates from large to small, and performing addition calculation until the accumulated entropy contribution rate value is greater than a preset threshold value, wherein the number of the feature vectors is the number of the selected principal components.
Further, the detection index is T in step S22Statistics and SPE statistics; wherein, T2The statistical quantity is calculated by the formula:
T2＝[tk,1,tk,2,…,tk,p]Λ-1[tk,1,tk,2,…,tk,p]T (14)
wherein, tk＝[tk,1,tk,2,…,tk,p]TIs a score vector, Λ-1A diagonal matrix which is the reciprocal of the eigenvalue;
T2the control limit calculation formula of the statistic is as follows:
Figure BDA0003220781630000051
the SPE statistic calculation formula is as follows:
Figure BDA0003220781630000052
wherein n is the number of non-zero eigenvalues, and p is the number of principal elements;
the control limit calculation formula of the SPE statistic is as follows:
Figure BDA0003220781630000053
wherein g is theta/2 omega, h is 2 omega2And/theta, where omega is the k-sample SPE mean and theta is the k-sample SPE variance.
Further, the step S3 of using the simplified KECA model to perform anomaly detection on the test data set specifically includes the steps of:
collecting process variable data of a roller kiln sintering zone system in a normal state as a test data set;
carrying out normalization processing on the test data set;
constructing a test data set core matrix of the RKECA method;
and calculating the value of the detection index based on the test set, and comparing the calculated value with a preset detection index threshold value to judge whether the abnormity occurs.
Further, the value of the detection index is calculated based on the test set, and the calculated value is compared with a preset statistic control limit to judge whether the abnormality occurs, wherein the specific process comprises the following steps:
when the value of the detection statistic is larger than the control limit of the statistic and 3 or more data exceeding the control limit exist, judging that the abnormality exists at the moment;
when the SPE statistic is larger than the control limit, judging that the abnormality exists at the moment;
when SPE statistic is less than control limit, T2When the statistic is larger than the control limit, judging according to the actual situation;
if SPE statistic and T2And if the statistics are not larger than the control limit, judging that no abnormity exists at the moment.
The invention provides a system for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA, which comprises: the detection method comprises a memory and a processor, wherein the memory comprises a simplified KECA-based ceramic roller kiln burning zone abnormity detection method program, and the processor executes the simplified KECA-based ceramic roller kiln burning zone abnormity detection method program to realize the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
s2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
acquiring a test data set and preprocessing the test data set;
s3: and carrying out anomaly detection on the test data set by utilizing a simplified KECA model.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
according to the invention, the simplified KECA model is constructed by reconstructing the kernel matrix of the data set, so that the complexity and the calculation cost of the model are reduced, and the real-time performance and the accuracy of detection are improved while the detection performance of the model is improved.
Drawings
FIG. 1 is a flow chart of the method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA.
FIG. 2 is a comparison graph of SPE statistic changes in an embodiment of the present invention.
FIG. 3 shows an embodiment of the present invention T2The statistical change is compared to a graph.
FIG. 4 is a graph comparing the variation contribution ratios according to the embodiment of the present invention.
FIG. 5 is a block diagram of a system for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
Example 1
As shown in FIG. 1, the invention provides a method for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA, which comprises the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
the specific process of step S1 is:
collecting process variable data of a roller kiln sintering zone system in a normal state as a training data set, and recording the process variable data as a first data set;
and carrying out normalization processing on the first data set, calculating the Euclidean distance of the first data set, and simplifying the training data set according to a preset Euclidean distance threshold value to obtain a second data set.
The normalization processing is carried out on the first data set, the Euclidean distance of the first data set is calculated, the training data set is simplified according to a preset Euclidean distance threshold value, and a second data set is obtained, and the specific process is as follows:
constructing the first data set into a sample matrix, and carrying out normalization processing on the sample matrix;
calculating Euclidean distances between two different rows of the normalized sample matrix;
comparing the Euclidean distances between two different rows with a preset Euclidean distance threshold value one by one;
wherein the Euclidean distance threshold is calculated by the following formula:
Figure BDA0003220781630000071
wherein d islRepresenting the Euclidean distance threshold, N representing the number of rows of the sample matrix, di,jRepresenting the euclidean distance between the ith and jth lines.
If the Euclidean distance between two different rows is smaller than a preset Euclidean distance threshold value, deleting or combining two rows of samples corresponding to the Euclidean distance; if the Euclidean distance between two different rows is larger than a preset Euclidean distance threshold value, two rows of sample data corresponding to the Euclidean distance are reserved, and a new data set consisting of all the reserved sample data is recorded as a second data set.
It should be noted that, for an N × M normalized sample matrix X, the euclidean distance between different rows in the matrix can be calculated by the following formula:
di,j＝||xi-xj||2，i,j＝1，…,N；i≠j
the Euclidean distance between two opposite sample matrixes X is obtained from the Euclidean distance formula
Figure BDA0003220781630000072
And (4) respectively. If d is obtainedi,jThe larger the value is, the farther the distance between the two rows i and j of data in the sample matrix X is, namely the difference is larger; if d is obtainedi,jThe smaller the value, the closer the distance between the two rows i and j in the sample matrix X, i.e. the smaller the difference. Thus, the euclidean distance provides information on the very similar pairs of data in the sample matrix X. For pairs of data similar to each other in the sample data set, the calculation amount and the memory requirement of the established KECA model are greatly increased. Therefore, it is necessary to simplify the processing of the initial sample data and to retain some representative non-redundant sample data. It is a key objective of the present invention to reduce the sample matrix X with dimension NxM to a matrix X with dimension nxM (N < N), and to build a simplified KECA model.
In an ideal situation, when the Euclidean distance di,jWhen the value is 0, it means that the data in the two rows i and j in the sample matrix X are completely similar, and one row of data can be directly deleted in the simplified processing of the sample matrix. While in practice the Euclidean distance di,jThe case of 0 does not exist basically, so a threshold value is set, and the Euclidean distance d is seti,jDeleting or merging the data of the rows smaller than the set threshold value to obtain the Euclidean distance di,jSample data larger than the threshold value is reserved, and all the reserved sample data form a new data set, so that the data matrix x is simplified. Since the initial sample data set is a normalized data set, the euclidean distance value of the sample matrix can be set according to an empirical threshold valueIs measured.
S2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
acquiring a test data set and preprocessing the test data set;
it should be noted that, in step S2, constructing the simplified KECA model by using the second data set, that is, reducing the dimension of the second data set in the principal component direction by using the KECA method, includes the specific steps of:
normalizing the second data set;
calculating the Renyi entropy of the samples in the second data set after normalization processing;
calculating the contribution rate of each entropy by using each Renyi entropy, determining the contribution rate threshold value of the entropy,
and obtaining the number k of the principal components according to the contribution rate threshold of the entropy, and mapping the data to the directions of the k kernel principal components to obtain a second data set after dimension reduction.
The specific process for determining the number of the main components comprises the following steps:
x is set in the data set D1,…,xNIs generated by the probability density function p (x), the second order Renyi entropy of the samples in the dataset is defined as:
H(p)＝-log∫p2(x)dx (2)
from equation (2), let:
V(p)＝∫p2(x)dx (3)
the logarithmic function in the formula (2) is an increasing type function, and then the estimated value of H (p) is obtained through the estimated value of V (p), and V (p) is obtained through calling a Parzen window estimator for estimation
Figure BDA0003220781630000081
Probability density function when invoking Parzen window estimator
Figure BDA0003220781630000091
As shown in equation 4:
Figure BDA0003220781630000092
wherein k isσ(x,xt) A kernel function for the feature space, usually a radial basis function, can be expressed as:
Figure BDA0003220781630000093
wherein σ is a parameter of the kernel function, and the desired operator is estimated by the sample mean, so that:
Figure BDA0003220781630000094
Figure BDA0003220781630000095
represents an estimated value of V (p); convert equation (6) to:
Figure BDA0003220781630000096
where K is a sample kernel matrix of N × N, 1 is a vector of N × 1, and the Renyi entropy is represented by eigenvalues of the kernel matrix and corresponding eigenvectors, where the kernel matrix may be characterized as:
K＝EDET (8)
wherein D is a characteristic value lambda1,…,λNE is a feature vector E1,…,eNIs a matrix of columns, ETIs the transpose of E, so the estimate of V (p) can be expressed as:
Figure BDA0003220781630000097
mapping of sample points to a new data set phi formed in the direction of k kernel principal componentsecaComprises the following steps:
Figure BDA0003220781630000098
calculating the contribution rate eta of Renyi entropy:
Figure BDA0003220781630000099
wherein m is a mean vector of the kernel feature space data set, and the expression of m is as follows:
Figure BDA00032207816300000910
mecais indicative of phiecaMean vector of transformed data, mecaThe expression is as follows:
Figure BDA0003220781630000101
calculating the contribution rate of Renyi entropy of each sample, sequencing the entropy contribution rates from large to small, and performing addition calculation until the accumulated entropy contribution rate value is greater than a preset threshold value, wherein the number of the feature vectors is the number of the selected principal components.
It should be noted that the invention is based on the constructed simplified KECA model, and adopts T2Anomaly detection is performed for statistics and SPE statistics, where T2The statistical quantity is calculated by the formula:
T2＝[tk,1,tk,2,…,tk,p]Λ-1[tk,1,tk,2,…,tk,p]T (14)
wherein, tk＝[tk,1,tk,2,…,tk,p]TIs a score vector, Λ-1A diagonal matrix which is the reciprocal of the eigenvalue;
T2the control limit calculation formula of the statistic is as follows:
Figure BDA0003220781630000102
the SPE statistic calculation formula is as follows:
Figure BDA0003220781630000103
wherein n is the number of non-zero eigenvalues, and p is the number of principal elements;
the control limit calculation formula of the SPE statistic is as follows:
Figure BDA0003220781630000104
wherein g is theta/2 omega, h is 2 omega2And/theta, where omega is the k-sample SPE mean and theta is the k-sample SPE variance.
S3: and carrying out anomaly detection on the test data set by utilizing a KECA model.
It should be noted that the specific steps of using the simplified KECA model to perform anomaly detection on the test data set are as follows:
collecting process variable data of a roller kiln sintering zone system in a normal state as a test data set;
carrying out normalization processing on the test data set;
constructing a test data set core matrix of the RKECA method;
and calculating the value of the detection index based on the test set, and comparing the calculated value with a preset detection index threshold value to judge whether the abnormity occurs.
More specifically, the value of the detection index is calculated based on the test set, and the calculated value is compared with a preset statistic control limit to judge whether an abnormality occurs, and the specific process is as follows:
when the value of the detection statistic is larger than the control limit of the statistic and 3 or more data exceeding the control limit exist, judging that the abnormality exists at the moment;
when the SPE statistic is larger than the control limit, judging that the abnormality exists at the moment;
when SPE statistic is less than control limit, T2Big statisticWhen the control time is limited, judging according to the actual condition;
if SPE statistic and T2And if the statistics are not larger than the control limit, judging that no abnormity exists at the moment.
As shown in fig. 5, a second aspect of the present invention provides a simplified KECA-based ceramic roller kiln burn-in anomaly detection system, comprising: the detection method comprises a memory and a processor, wherein the memory comprises a simplified KECA-based ceramic roller kiln burning zone abnormity detection method program, and the processor executes the simplified KECA-based ceramic roller kiln burning zone abnormity detection method program to realize the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
s2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
acquiring a test data set and preprocessing the test data set;
s3: and carrying out anomaly detection on the test data set by utilizing a KECA simplified model.
Verification and analysis
The present example will be specifically described with reference to the detection of an abnormality in the firing system of the ceramic roller kiln.
Step 1: summary of abnormal conditions of ceramic roller kiln firing system
The working process of the roller kiln burning zone is very complex, and the thermal process relates to the scientific fields of pneumatics, fuel combustion science, heat transfer science and the like. In the ceramic firing process, each system device completes the whole firing process under the combined action of a formulated temperature system, an atmosphere system and a pressure system. The quality and the yield of the ceramic product can be directly influenced when any link is abnormal, the performance of the whole roller kiln burning zone system is reduced, a large amount of energy loss and unnecessary resource waste can be caused, and the considerable economic loss is brought to enterprises. In order to ensure that the roller kiln is normally, safely and efficiently operated in a sintering zone, various influence factors and possible abnormal conditions in the operation process of the roller kiln need to be analyzed.
The common abnormal types of the roller kiln sintering zone in the production process can be roughly divided into two types of process abnormality and equipment abnormality. The process abnormality is mainly an abnormal working condition caused by improper control of a temperature system, an atmosphere system and a pressure system, and the equipment abnormality is mainly an abnormal phenomenon caused by aging of hardware equipment such as a roller way, a burner, a motor and an actuating mechanism or instrument faults and is shown in table 1.
TABLE 1 abnormal phenomena and causes
Figure BDA0003220781630000121
In the operation process of the roller kiln burning zone system, measurable variables such as temperature, flow and the like exist, the measurable variables are data generated by equipment which has the most important influence on the roller kiln burning zone, are key variables influencing the operation condition of the roller kiln burning zone in the production process, and can reflect the operation condition of the roller kiln burning zone system. The details of the measurable variables are given in table 2:
TABLE 2 descriptio of measurable variables
Figure BDA0003220781630000122
Figure BDA0003220781630000131
Common abnormal phenomena and measurable variable analysis of the roller kiln burning zone are combined with the actual production condition of an enterprise, and the common abnormal phenomena and measurable variable analysis of the roller kiln burning zone are summarized. The types of anomalies considered by the present invention mainly include three major categories:
(1) the material flow is abnormal, mainly the flow of each gas in the burning zone and the ceramic body is abnormal;
(2) the temperature in the kiln is abnormal, mainly the temperature of the flue gas at the entrance and the exit of the burning zone is abnormal;
(3) abnormal atmosphere in kiln, mainly natural gas leakage
The invention considers only a single anomaly, the anomalies considered being as in table 3:
TABLE 3 Exception description Table
Exception numbering Description of anomalies
1 The mass flow of the natural gas becomes smaller
2 The mass flow of the natural gas becomes larger
3 The mass flow of the combustion-supporting gas becomes smaller
4 The mass flow of the combustion-supporting gas is increased
5 The mass flow of the flue gas becomes smaller
6 The mass flow of the flue gas is increased
7 The temperature of the inlet flue gas becomes lower
8 Inlet flue gas temperature becomes high
9 The mass flow of the green brick becomes small
10 The mass flow of the green brick is increased
11 Leakage of natural gas
Step 2: data acquisition
Data are collected from the operation process of 2019 and 10 months of firing zone equipment of a roller kiln of a certain ceramic production enterprise, 20 state variables are totally obtained, and as all monitoring values of some control parameters and state parameters are not changed and some parameters irrelevant to abnormity exist, the variables are selected, 11 variables are selected from the variables, as shown in table 4, wherein t is1～t6Respectively the temperature of the ceramic tile green body, combustion-supporting gas, input flue gas, brick firing, output flue gas and the surface of the outer wall of the kiln body; m is1～m5Respectively the flow rates of the ceramic tile green body, the natural gas, the combustion-supporting gas, the input flue gas and the output flue gas. State variable of roller kiln
TABLE 4 State variable table
Variable of state Meaning of variables
t1–t6 Temperature measurement
m1–m5 Flow rate measurement
By collecting the obtained sample data, a simulation model is established, summarized abnormal conditions are introduced, wherein 11 abnormal types are included, each type of abnormal data comprises 800 groups, and finally, 12 data sets are obtained, 11 data sets in 12 data sets are abnormal data types, and 1 data set is a normal data type. The method comprises the steps that a training data set is normal working condition data and comprises 1000 groups of data, the Euclidean distance of the training data set is calculated by the method, the training data set is simplified according to a preset Euclidean distance threshold value, and a simplified KECA model is constructed; the test data set consists of 11 independent data matrixes, each data matrix comprises 200 groups of normal working condition data and 800 groups of abnormal data, and the test data set is used for testing and verifying the constructed simplified KECA model.
And step 3: model parameter setting
Since the roller kiln sintering belt system contains 11 anomalies, 11 RKECA models will be built, one for each anomaly. In the embodiment, two methods of RKECA and KECA are respectively used for the abnormity detection of the sintering belt equipment of the roller kiln, and the performances of the equipment are compared and verified. Before reconstruction processing is carried out on training set data, a KECA model (kernel entropy component analysis method) and a KPCA model (kernel principal component analysis method) are respectively established by utilizing original training set data; a comparative study was carried out in the same context as the RKECA process proposed by the present invention (simplified KECA process, i.e. the process of the present invention).
In the abnormal simulation of the RKECA method (namely, the simplified KECA), the KECA method and the KPCA method, the selection of the kernel function types is the same and all the kernel function types are radial basis functions; selecting the number of principal elements by adopting an accumulated contribution rate of 85 percent; t is selected as the monitoring statistic2The control limit of the statistic and SPE statistic and monitoring statistic is set to 95%, and if 3 or more sample data exceed the limit, the abnormal condition can be determined. All the required parameters in the simulation are set according to experiments, and when the parameters take the values, the anomaly monitoring shows better performance and high reliability.
And 4, step 4: analysis of results
The invention analyzes the detection result aiming at the abnormity 1, and the obtained result is shown as a figure 2-3, wherein (a) in the figure 2 shows the SPE statistic change of the invention method (RKECA method), and (b) in the figure 2 shows the SPE statistic change by adopting a KECA method; wherein (a) in FIG. 3 represents the process of the invention (RKECA process) T2Variation of statistic amount, and (b) in FIG. 3 shows that KECA method T is used2A change in the statistics; in which (a) in fig. 4 represents the variation contribution rate of the method of the present invention (the RKECA method), and (b) in fig. 4 represents the variation contribution rate by the KECA method.
In order to verify the effectiveness of the RKECA method, the invention adopts the Computation Time (CT), the detection rate (FDR) and the False Alarm Rate (FAR) as evaluation indexes, and respectively contrasts and analyzes the RKECA, the KECA and the KPCA based on 11 abnormal data sets of the roller kiln sintering belt equipment, wherein the detection rate (FDR) and the False Alarm Rate (FAR) results are derived from SPE monitoring statistics. The results of abnormal detection of the SPE monitoring statistics are shown in table 5, and the comparison tables of the calculation times of the three methods are shown in table 6.
From table 5, it can be seen that the detection rate of the RKECA method is 86.85%, and the false detection rate is 1.05%; the detection rate of the KECA method is 52.83%, the false detection rate is 0.86%, the detection rate of the KPCA method is 46.11%, and the false detection rate is 4.09%.
Table 511 type anomaly detection rate and false alarm rate table
Figure BDA0003220781630000151
Figure BDA0003220781630000161
TABLE 6 comparison of calculated times for three methods
RKECA KECA KPCA
CT(s) 2.89 169.83 175.64
In addition, the data of the training set reconstructed by the Euclidean distance similarity in the RKECA method is reduced from the original 1000 groups of data to 156 groups, and the reduction rate is 84.4%. As can be seen from Table 6, the calculation time of RKECA method is greatly shortened by 98.3% compared with KECA method and 98.35% compared with KPCA method. Significantly reducing computation time and memory space costs.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (10)

1. A method for detecting abnormal firing zone of a ceramic roller kiln based on simplified KECA is characterized by comprising the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
s2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
s3: and carrying out anomaly detection on the test data set by utilizing a simplified KECA model.
2. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 1, wherein the step S1 comprises the following steps:
collecting process variable data of a roller kiln sintering zone system in a normal state as a training data set, and recording the process variable data as a first data set;
and carrying out normalization processing on the first data set, calculating the Euclidean distance of the first data set, and simplifying the training data set according to a preset Euclidean distance threshold value to obtain a second data set.
3. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 2, wherein the normalization process is performed on the first data set, the Euclidean distance of the first data set is calculated, and the training data set is simplified according to the preset Euclidean distance threshold to obtain the second data set, and the specific process is as follows:
constructing the first data set into a sample matrix, and carrying out normalization processing on the sample matrix;
calculating Euclidean distances between two different rows of the normalized sample matrix;
comparing the Euclidean distances between two different rows with a preset Euclidean distance threshold value one by one;
if the Euclidean distance between two different rows is smaller than a preset Euclidean distance threshold value, deleting or combining two rows of samples corresponding to the Euclidean distance; if the Euclidean distance between two different rows is larger than a preset Euclidean distance threshold value, two rows of sample data corresponding to the Euclidean distance are reserved, and a new data set consisting of all the reserved sample data is recorded as a second data set.
4. The method of claim 3, wherein the Euclidean distance threshold is calculated by the following formula:
Figure FDA0003220781620000011
wherein d islRepresenting the Euclidean distance threshold, N representing the number of rows of the sample matrix, di,jRepresenting the euclidean distance between the ith and jth lines.
5. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 1, wherein the step S2 of constructing the KECA model by using the second data set, i.e. reducing the dimension of the second data set in the principal component direction by using the KECA method, comprises the following steps:
normalizing the second data set;
calculating the Renyi entropy of the samples in the second data set after normalization processing;
calculating the contribution rate of each entropy by using each Renyi entropy, determining the contribution rate threshold value of the entropy,
and obtaining the number k of the principal components according to the contribution rate threshold of the entropy, and mapping the data to the directions of the k kernel principal components to obtain a second data set after dimension reduction.
6. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 5, wherein the specific process for determining the number of main components is as follows:
x is set in the data set D1,…,xNIs generated by the probability density function p (x), the second order Renyi entropy of the samples in the dataset is defined as:
H(p)＝-log∫p2(x)dx (2)
from equation (2), let:
V(p)＝∫p2(x)dx (3)
the logarithmic function in the formula (2) is an increasing type function, then the estimated value of H (p) can be obtained through the estimated value of V (p), and V (p) is obtained through the estimation by calling a Parzen window estimator
Figure FDA0003220781620000021
Probability density function when invoking Parzen window estimator
Figure FDA0003220781620000022
As shown in equation 4:
Figure FDA0003220781620000023
wherein k isσ(x,xt) A kernel function for the feature space, usually a radial basis function, can be expressed as:
Figure FDA0003220781620000024
wherein σ is a parameter of the kernel function, and the desired operator is estimated by the sample mean, so that:
Figure FDA0003220781620000025
Figure FDA0003220781620000031
represents an estimated value of V (p); convert equation (6) to:
Figure FDA0003220781620000032
where K is a sample kernel matrix of N × N, 1 is a vector of N × 1, and the Renyi entropy is represented by eigenvalues of the kernel matrix and corresponding eigenvectors, where the kernel matrix may be characterized as:
K＝EDET (8)
wherein D is a characteristic value lambda1,…,λNE is a feature vector E1,…,eNIs a matrix of columns, ETIs the transpose of E, so the estimate of V (p) can be expressed as:
Figure FDA0003220781620000033
mapping of sample points to a new data set phi formed in the direction of k kernel principal componentsecaComprises the following steps:
Figure FDA0003220781620000034
calculating the contribution rate eta of Renyi entropy:
Figure FDA0003220781620000035
wherein m is a mean vector of the kernel feature space data set, and the expression of m is as follows:
Figure FDA0003220781620000036
mecais indicative of phiecaMean vector of transformed data, mecaThe expression is as follows:
Figure FDA0003220781620000037
calculating the contribution rate of Renyi entropy of each sample, sequencing the entropy contribution rates from large to small, and performing addition calculation until the accumulated entropy contribution rate value is greater than a preset threshold value, wherein the number of the feature vectors is the number of the selected principal components.
7. The method as claimed in claim 1, wherein the index detected in step S2 is T22Statistics and SPE statistics; wherein, T2The statistical quantity is calculated by the formula:
T2＝[tk,1,tk,2,…,tk,p]Λ-1[tk,1,tk,2,…,tk,p]T (14)
wherein, tk＝[tk,1,tk,2,…,tk,p]TIs a score vector, Λ-1A diagonal matrix which is the reciprocal of the eigenvalue;
T2the control limit calculation formula of the statistic is as follows:
Figure FDA0003220781620000041
the SPE statistic calculation formula is as follows:
Figure FDA0003220781620000042
wherein n is the number of non-zero eigenvalues, and p is the number of principal elements;
the control limit calculation formula of the SPE statistic is as follows:
Figure FDA0003220781620000043
wherein g is theta/2 omega, h is 2 omega2And/theta, where omega is the k-sample SPE mean and theta is the k-sample SPE variance.
8. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 1, wherein the step S3 of using the simplified KECA model to detect the abnormality of the test data set comprises the following steps:
collecting process variable data of a roller kiln sintering zone system in a normal state as a test data set;
carrying out normalization processing on the test data set;
constructing a test data set core matrix of the RKECA method;
and calculating the value of the detection index based on the test set, and comparing the calculated value with a preset detection index threshold value to judge whether the abnormity occurs.
9. The method for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA as claimed in claim 8, wherein the method comprises calculating the value of the detection index based on the test set, and comparing the calculated value with the preset statistic control limit to determine whether an abnormality occurs, and comprises the following steps:
when the value of the detection statistic is larger than the control limit of the statistic and 3 or more data exceeding the control limit exist, judging that the abnormality exists at the moment;
when the SPE statistic is larger than the control limit, judging that the abnormality exists at the moment;
when SPE statistic is less than control limit, T2When the statistic is larger than the control limit, judging according to the actual situation;
if SPE statistic and T2And if the statistics are not larger than the control limit, judging that no abnormity exists at the moment.
10. A simplified KECA-based ceramic roller kiln firing zone anomaly detection system, comprising: the detection method comprises a memory and a processor, wherein the memory comprises a simplified KECA-based ceramic roller kiln burning zone abnormity detection method program, and the processor executes the simplified KECA-based ceramic roller kiln burning zone abnormity detection method program to realize the following steps:
s1: acquiring training data to construct a training data set, and preprocessing the training data set to obtain a second data set;
s2: constructing a simplified KECA model by using the second data set, and determining a detection index and a detection index control limit;
acquiring a test data set and preprocessing the test data set;
s3: and carrying out anomaly detection on the test data set by utilizing a simplified KECA model.
CN202110957181.7A 2021-08-19 2021-08-19 Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA Pending CN113688895A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110957181.7A CN113688895A (en) 2021-08-19 2021-08-19 Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110957181.7A CN113688895A (en) 2021-08-19 2021-08-19 Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA

Publications (1)

Publication Number Publication Date
CN113688895A true CN113688895A (en) 2021-11-23

Family

ID=78580794

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110957181.7A Pending CN113688895A (en) 2021-08-19 2021-08-19 Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA

Country Status (1)

Country Link
CN (1) CN113688895A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358297A (en) * 2022-07-27 2022-11-18 广东工业大学 Injection molding machine abnormity detection method and system based on improved MKECA method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106094749A (en) * 2016-06-17 2016-11-09 浙江大学 Based on the nonlinear fault detection method and the application that improve nuclear entropy component analysis
CN110687895A (en) * 2019-10-24 2020-01-14 上海工程技术大学 Chemical process fault detection method based on self-adaptive kernel principal component analysis
JP2020027342A (en) * 2018-08-09 2020-02-20 富士電機株式会社 Information processing apparatus, monitoring device and information processing method
CN111914889A (en) * 2020-06-13 2020-11-10 宁波大学 Rectifying tower abnormal state identification method based on brief kernel principal component analysis
CN112066724A (en) * 2020-08-18 2020-12-11 广东工业大学 Roller kiln energy consumption abnormity detection method based on self-adaptive principal component analysis

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106094749A (en) * 2016-06-17 2016-11-09 浙江大学 Based on the nonlinear fault detection method and the application that improve nuclear entropy component analysis
JP2020027342A (en) * 2018-08-09 2020-02-20 富士電機株式会社 Information processing apparatus, monitoring device and information processing method
CN110687895A (en) * 2019-10-24 2020-01-14 上海工程技术大学 Chemical process fault detection method based on self-adaptive kernel principal component analysis
CN111914889A (en) * 2020-06-13 2020-11-10 宁波大学 Rectifying tower abnormal state identification method based on brief kernel principal component analysis
CN112066724A (en) * 2020-08-18 2020-12-11 广东工业大学 Roller kiln energy consumption abnormity detection method based on self-adaptive principal component analysis

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
F. BENCHEIKH等: "New reduced kernel PCA for fault detection and diagnosis in cement rotary kiln", 《CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS》 *
张文涛等: "基于小波去噪与KECA的工业过程故障检测研究", 《化工自动化及仪表》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358297A (en) * 2022-07-27 2022-11-18 广东工业大学 Injection molding machine abnormity detection method and system based on improved MKECA method

Similar Documents

Publication Publication Date Title
CN110441065B (en) Gas turbine on-line detection method and device based on LSTM
CN112070109B (en) Water chestnut kiln energy consumption abnormality detection method based on improved density peak value clustering
CN106709662B (en) Power equipment operation condition division method
US7251582B2 (en) Fault diagnosis
CN107562696B (en) On-line detection and control method for tire product quality
CN110738274A (en) nuclear power device fault diagnosis method based on data driving
CN110532674B (en) Method for measuring temperature of boiler furnace of coal-fired power plant
JP2010504501A (en) Kernel-based method for detecting boiler tube leaks
CN116559598B (en) Smart distribution network fault positioning method and system
CN111160457B (en) Scroll engine fault detection method based on soft-class extreme learning machine
CN110779745A (en) Heat exchanger early fault diagnosis method based on BP neural network
CN108897354B (en) Aluminum smelting process hearth temperature prediction method based on deep belief network
CN114297918A (en) Aero-engine residual life prediction method based on full-attention depth network and dynamic ensemble learning
Chen et al. A whole building fault detection using weather based pattern matching and feature based PCA method
CN116825243B (en) Multi-source data-based thermal barrier coating service life prediction method and system
CN114757269A (en) Complex process refined fault detection method based on local subspace-neighborhood preserving embedding
CN110942258B (en) Performance-driven industrial process anomaly monitoring method
CN115375026A (en) Method for predicting service life of aircraft engine in multiple fault modes
CN110427019B (en) Industrial process fault classification method and control device based on multivariate discriminant analysis
CN113688895A (en) Method and system for detecting abnormal firing zone of ceramic roller kiln based on simplified KECA
CN111507374A (en) Power grid mass data anomaly detection method based on random matrix theory
CN113090420B (en) EGR flow diagnosis method based on multi-classification logistic regression algorithm
Lyu et al. Image-based process monitoring using deep belief networks
CN115017978A (en) Fault classification method based on weighted probability neural network
CN114548701A (en) Process early warning method and system for analyzing and estimating coupling structure of full-scale measuring point

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20211123