KR20200125384A - Method and System for Real-time Leakage State Recognition of Tubes - Google Patents

Abstract

According to the present invention, disclosed are a real-time tube leak state recognition method and a system thereof. According to the present invention, the method includes the following steps of: extracting features from a signal measured by a sensor; forming clusters from the features; calculating an evaluation index about the clusters; determining the number of clusters based on a result of the calculation; and recognizing a leak state of a tube based on the number of clusters. Moreover, the present invention has an effect of enabling an inspector to accurately determine whether a corresponding tube has a leak state without checking whether there is a leak on the site in person.

Description

Method and System for Real-time Leakage State Recognition of Tubes}

The present invention relates to a real-time tube leakage state recognition method and system, and more particularly, by measuring a signal with a sensor, etc., and through a mathematical analysis of the signal characteristic, it is possible to quickly and accurately detect whether the tube leaks in real time. It relates to a method and system for real-time tube leakage status recognition.

Globally, energy consumption and demand are increasing exponentially, and electricity demand is also steadily increasing in Korea. In addition, due to the recent nuclear accident in Japan, anxiety about nuclear power plants has increased, and interest in thermal power plants that use coal, oil, natural gas, etc. as fuels has increased, and thermal power plants are being built a lot.

In general, a thermal power plant uses fossil fuels such as coal and natural gas as fuel, and generates steam by burning coal, which is a main raw material, in a boiler, and generates electric power by rotating a turbine using the generated steam.

When analyzing the failure rate of each facility in such a thermal power plant, the importance of early detection of tube leakage is increasing day by day as boiler facilities account for 32.3%, and boiler tube leakage accounts for 61% of them.

A boiler is a facility that generates superheated steam by heating water flowing in a plurality of tubes disposed therein, and generates electricity by operating a turbine with superheated steam in a thermal power plant, or is used for heating in general homes.

In such a boiler, if the tube disposed inside it cannot withstand the high-pressure vapor pressure and ruptures, it causes a large-scale accident. In particular, if such an accident occurs in a thermal power plant, the power supply is not smooth, and the assumption that electricity is used. It will be a big blow to the factory or factory.

In general, boiler tube leakage occurs when pin holes or cracks occur due to defects in tube manufacturing, and high temperature and high pressure steam is ejected from micro holes and cracks due to defects during boiler operation. Over time, the leaked hot and high pressure steam can lead to damage to adjacent tubes, eventually causing the generator to trip, resulting in enormous economic losses.

Therefore, before leakage occurs in the tube, the weak part of the tube vibrates for a considerable period of time to produce an abnormal sound, thereby showing abnormal symptoms.

When energy is transferred by the pressure difference and high-pressure steam is ejected, the steam passing through the micropores or cracks rubs against the boiler tube material, generating a high-noise leakage sound. The generated leak noise is detected by the acoustic sensor installed in the boiler, sends a signal to the boiler tube leak detection system, and sends an alarm to the operator. That is, the boiler tube leak detection system is a facility that predicts and diagnoses a failure of the boiler tube by attaching a plurality of acoustic sensors to the side of the boiler tube and detecting a leak sound.

However, the conventional Patent Publication No. 10-2013-0035102 (Boiler Tube Leak Detection System) simply detects when the signal strength exceeds a certain level, and when an alarm is suspected due to leakage, the inspector directly uses the probe at the site. Therefore, there was a problem that it was difficult to listen, check for leaks, and compare with the constant driving noise.

Unexamined Patent Publication No. 10-2013-0035102 (2013.04.08.)

The present invention was devised to solve the above problems, and when a new cluster other than the existing cluster is formed, the center point of the cluster is detected to detect whether a signal is a new type, and through this, the leak state of the tube can be recognized. An object of the present invention is to provide a method and system for real-time tube leakage status recognition that can determine whether or not there is.

In addition, the present invention was conceived to solve the above problems, and the real-time tube leakage state recognition that allows an inspector to accurately determine whether a leakage state exists in the corresponding tube without directly checking whether there is a leakage at the site. It is an object to provide a method and system.

In order to solve the above problems, the present invention includes the steps of extracting features from a signal; A clustering step of forming a cluster from the features; Calculating an evaluation index for the cluster; Determining the number of clusters based on the evaluation index; And recognizing a leak state of the tube based on the number of clusters.

The features include Root-mean-square, Shape factor, Kurtosis value, Square-mean-root, Peak-to-peak value. value), a skewness value, an impulse factor, and a crest factor.

In the step of extracting the feature, the feature may be extracted from an existing signal and an acquired signal.

The clustering step may include: a first clustering step of performing clustering by setting the number of clusters to K; And a second clustering step of performing clustering in which the number of clusters is set to K+1.

The first clustering step may be to obtain an existing center point by averaging the features extracted from the existing signal.

In this case, the first clustering step may be performing clustering by mixing the features extracted from the acquired signal with the features extracted from the existing signal.

In the second clustering step, the existing center point is fixed as it is, and only a new center point for a feature extracted from the acquired signal is updated.

The calculating of the evaluation index may include setting the number of clusters to K and then calculating a first evaluation index using clustering and an average probability; And calculating a second evaluation index using clustering and an average probability after setting the number of clusters to K+1.

In this case, the calculating of the first evaluation index may include: selecting one target cluster from among the clusters and calculating a first average probability; Calculating a second average probability of two clusters in which the selected target cluster and another cluster are mixed; And calculating the first evaluation index based on the first average amplification and the second average probability.

Further, the calculating of the second evaluation index may include: selecting one target cluster from among the clusters and calculating a first average probability; Calculating a second average probability of two clusters in which the selected target cluster and another cluster are mixed; And calculating the second evaluation index based on the first average amplification and the second average probability.

In this case, the evaluation index includes an average probability ratio, and the average probability ratio may be a value obtained by dividing the second average probability with respect to the first average probability.

The determining of the number of clusters may include selecting the number of clusters having a larger evaluation index among the first evaluation index and the second evaluation index.

In this case, the step of recognizing may recognize that the tube is in a leak state when the number of clusters increases, and recognize that the tube is in a normal state when the number of clusters is maintained.

In addition, the present invention in order to solve the above-described problem, the tube flowing material flows therein; A sensor installed around the tube; A determination unit that extracts features from the sensor, forms a cluster from the features, calculates an evaluation index for the cluster, determines the number of clusters based on the evaluation index, and recognizes the leak state of the tube in real time Includes;

The tube may be used in a circulating fluidized bed boiler.

The real-time tube leakage state recognition method and system of the present invention as described above has an effect of allowing an inspector to accurately determine whether a leakage state exists in a corresponding tube without directly checking whether there is a leakage in the field.

1 is a schematic diagram of a system for recognizing a real-time tube leakage state according to a first embodiment of the present invention.
2 is a diagram schematically illustrating a real-time tube leakage state recognition method according to a second embodiment of the present invention.
3 is a diagram showing in more detail steps from S100 to S400 according to the second embodiment of the present invention.
4 is a diagram showing a first clustering step according to a second embodiment of the present invention.
5 is a diagram showing in detail a second clustering step according to a second embodiment of the present invention.
6 is a diagram showing in detail a method of calculating an evaluation index value using clustering and average probability according to a second embodiment of the present invention.
7 is a diagram showing a method of calculating a probability density function according to a second embodiment of the present invention.
8 is a diagram showing a method of calculating an average probability according to a second embodiment of the present invention.
9 is a diagram showing an example of analyzing a value obtained by calculating a probability density function according to a second embodiment of the present invention.
10 is a diagram showing an example of analyzing a value obtained by calculating a probability density function according to a second embodiment of the present invention.

In the present invention, various transformations may be applied and various embodiments may be provided, and specific embodiments will be illustrated in the drawings and described in detail. However, this is not intended to be limited to a specific embodiment of the present invention, it is to be understood to include all conversions, equivalents, and substitutes included in the spirit and scope of the present invention. In describing the present invention, when it is determined that a detailed description of a related known technology may obscure the subject matter of the present invention, a detailed description thereof will be omitted.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. These terms are used only for the purpose of distinguishing one component from another component.

The terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present application, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof, does not preclude in advance.

Therefore, the spirit of the present invention is limited to the described embodiments and should not be defined, and all things that are equivalent or equivalent to the claims as well as the claims to be described later fall within the scope of the spirit of the present invention. .

Hereinafter, a system for recognizing a real-time tube leakage state according to a first exemplary embodiment of the present invention will be described in detail.

1 is a schematic diagram of a system for recognizing a real-time tube leakage state according to a first embodiment of the present invention.

Referring to FIG. 1, a system for recognizing a real-time tube leakage state according to the present invention includes a tube 10, a sensor 20, and a determination unit 30.

A flowing material flows inside the tube 10. In this case, the flowing material may generally mean water used in a boiler. In addition, the tube 10 may be used for a circulating fluidized bed boiler. At this time, the circulating fluidized bed boiler refers to an eco-friendly power generation facility that completely burns coal by continuously circulating heat, unlike conventional boilers that simply burn coal to run a generator.

The circulating fluidized bed boiler has the effect of greatly reducing the emission of pollutants such as nitrogen oxides and sulfur oxides by circulating combustion by simultaneously injecting air and lime. Given this effect, demand is expected to continue to increase in the future.

The sensor 20 may collect a noise signal generated inside the boiler, in particular, the tube 10. To this end, the sensor 20 may include a pre-amplifier, a filter, a main-amplifier, and an analog digital converter.

The sensor 20 converts the noise signal into an electric signal, and the sensor 20 may be formed in plural on the outer wall of the tube 10 or around the tube 10, an acoustic emission sensor, an acoustic sensor, a microphone, and It may include at least one of the acceleration sensors.

In addition, the sensor 20 may not collect a noise signal, and may collect a vibration signal generated inside the boiler, in particular, the tube 10.

The determination unit 30 extracts features from the sensor 20, forms a cluster from the extracted features, calculates an evaluation index for the formed cluster, determines the number of clusters based on the calculated evaluation index, and determines the tube ( 10) leakage status is recognized in real time.

), 형상 계수(Shape factor,

), 첨도 값(Kurtosis value,

), 제곱 평균 루트(Square-mean-root,

), 피크 투 피크 값(Peak-to-peak value,

), 왜도 값(Skewness value,

), 임펄스 인자(Impulse factor,

) 및 파고율(Crest factor,

)로 이루어진 군으로부터 선택되는 적어도 하나인 것일 수 있다.The feature extracted from the sensor 20 is the root-mean-square,

), Shape factor,

), Kurtosis value,

), Square-mean-root,

), Peak-to-peak value,

), Skewness value,

), Impulse factor,

) And Crest factor,

) It may be at least one selected from the group consisting of.

The above features can be expressed as an equation as follows.

In mathematics, a cluster refers to a unit in which several elements of a population subject to statistical investigation are collected, and clustering refers to forming a cluster as described above.

The determination unit 30 may perform the step of clustering based on each cluster, and may calculate a probability density function (PDF) for features in the formed cluster. The determination unit 30 may calculate the probability density function, and use the calculation result to calculate an evaluation index. The determination unit 30 may determine in real time whether there is a leak in the tube 10 (whether it is a leak state) using the evaluation index.

At this time, a method for the determination unit 30 to perform clustering, a method for calculating a probability density function, and a method for determining whether or not a leak state by using an evaluation index will be described in detail through the second embodiment of the present invention. , Omitted here.

Hereinafter, a method of recognizing a real-time tube leakage state according to a second exemplary embodiment of the present invention will be described in detail.

For reference, in the method for detecting a real-time tube leakage state according to the second embodiment of the present invention, descriptions of components having the same or similar characteristics as in the first embodiment of the present invention will be omitted and only differences will be described.

However, the subject performing the real-time tube leakage state recognition method according to the second embodiment of the present invention may be a real-time tube leakage state recognition system according to the first embodiment of the present invention.

FIG. 2 is a diagram schematically showing a method for real-time tube leakage status recognition according to a second embodiment of the present invention, and FIG. 3 is a view showing in more detail steps from S100 to S400 according to the second embodiment of the present invention. .

According to FIG. 2, the method for real-time tube leakage state recognition according to the present invention includes a step of extracting features from a signal measured by a sensor 20 (S100), a clustering step of forming a cluster from the extracted features (S200), Calculating the evaluation index for the formed cluster (S300), determining the number of clusters based on the calculated evaluation index (S400), and recognizing the leakage state of the tube 10 based on the determined number of clusters It includes a step (S500).

In the step of extracting features from the signal measured by the sensor 20 (S100), when a noise signal, a vibration signal, etc. are measured by the sensor 20 installed in the boiler or around the tube 10, the A feature (or feature vector) is extracted using equations.

In this case, the step of extracting the feature (S100) may include extracting the feature from the existing signal (S110) and the step of extracting the feature from the acquired signal (S120). That is, the step of extracting features (S100) refers to a step of extracting features from an existing signal and an acquired signal.

Existing signals are those whose properties are all analyzed and known. If only the existing signal is measured, the boiler or tube can be considered to be in a normal state.

In addition, the acquired signal refers to new data measured by the sensor 20 as the property of the signal has not been analyzed. That is, the acquisition signal may mean all new data newly measured in addition to the existing signal.

), 형상 계수(Shape factor,

), 첨도 값(Kurtosis value,

), 제곱 평균 루트(Square-mean-root,

), 피크 투 피크 값(Peak-to-peak value,

), 왜도 값(Skewness value,

), 임펄스 인자(Impulse factor,

) 및 파고율(Crest factor,

)로 이루어진 군으로부터 선택되는 적어도 하나인 것일 수 있다. 이에 대한 설명은 본 발명의 제1 실시예에 기재된 수학식으로 대체한다. The feature extracted from the sensor 20 is the root-mean-square,

), Shape factor,

), Kurtosis value,

), Square-mean-root,

), Peak-to-peak value,

), Skewness value,

), Impulse factor,

) And Crest factor,

) It may be at least one selected from the group consisting of. Description of this is replaced by the equation described in the first embodiment of the present invention.

The clustering step (S200) of forming a cluster from the extracted features refers to a step of using a clustering technique. The clustering technique refers to a method of reconfiguring all nodes into small virtual groups according to local adjacency. In the present invention, the clustering step (S200) is a clustering technique that averages features to obtain a center value, a K-means clustering technique, and K-Medoids. Various clustering techniques such as (PAM) clustering techniques can be used.

Among clustering techniques, K-means clustering refers to clustering a given data into K clusters. In this K-means clustering, after randomly selecting one sample value from a given data sample, the distance from the sample value to the other sample data is measured. At this time, the closest center of each data sample is selected, and the distance to another data sample value is calculated around the corresponding sample again. By repeating this process, clustering is achieved. That is, a plurality of clusters are created. This clustering technique can be mainly implemented in Python.

According to FIG. 3, the clustering step (S200) includes a first clustering step (S210) of performing clustering in which the number of clusters is set to K, and a second clustering step of performing clustering in which the number of clusters is set to K+1. It may include (S220). However, the first clustering step S210 and the second clustering step S220 may be simultaneously performed in parallel.

That is, according to FIG. 3, the first clustering step (S210) may refer to a step of calculating a center value (hereinafter, an existing center value) of existing data, and the second clustering step (S220) is a center of acquired data. It may refer to the step of calculating the value (hereinafter, the new central value).

According to FIG. 3, in the step of calculating the evaluation index for the formed cluster (S300), the first evaluation index value is calculated using clustering and average probability after setting the number of clusters to the existing cluster number K (S310). , And setting the number of clusters to the number of existing clusters K+1, and calculating a second evaluation index value using clustering and an average probability (S320). That is, the first evaluation index refers to the evaluation index when the number of clusters is K, and the second evaluation index refers to the evaluation index when the number of clusters is K+1.

According to FIG. 3, determining the number of clusters based on the calculated evaluation index (S400) comprises selecting the number of clusters having a larger evaluation index among the first evaluation index and the second evaluation index (S410). It may include.

In addition, the step of recognizing the leak state of the tube 10 based on the determined number of clusters (S500), if the number of clusters increases, the tube 10 is recognized as a leak state, and if the number of clusters is maintained, The tube 10 is recognized as being in a normal state.

As such, in the real-time tube leakage state recognition method according to the present invention, the leakage state of the tube 10 may be recognized through the steps of FIGS. 2 and 3.

4 is a diagram showing a first clustering step according to a second embodiment of the present invention.

According to FIG. 4, the first clustering step (S210) is a step of performing a clustering technique with the total number of clusters as K even when new data is added in addition to the existing clusters (number K) based on existing data. Say.

According to FIG. 4, in the first clustering step (S210 ), a center value may be calculated by obtaining an average of features extracted from existing signals, and the center value at this time is referred to as an existing center value. In other words, the first clustering step S210 may refer to a step of calculating an existing center value from features of the existing signals.

According to FIG. 4, when new data is input, the total number of clusters is left as K, and clustering can be newly performed by mixing new data with existing data (ie, existing signals), that is, existing clusters. Therefore, according to FIG. 4, there is no change in the center point obtained previously, and new data belong to the nearest cluster.

At this time, the new data refers to an acquisition signal that is data different from the existing data, and the present invention is an invention for mathematically determining whether such new data can be interpreted as meaning leakage of the tube 10.

5 is a diagram showing in detail the second clustering step according to the second embodiment of the present invention, and FIG. 6 is a detailed diagram of a method of calculating an evaluation index value using clustering and average probability according to the second embodiment of the present invention. It is a diagram shown below.

Referring to FIG. 5, in the second clustering step (S220 ), when the number of clusters is K+1, an initial center point of a new cluster may be arbitrarily selected from within new data. At this time, the existing cluster center point is fixed and only the center point of the new cluster is updated. In addition, the average value of the clusters is updated by calculating the distance between the center point of each cluster and the features (or feature vectors), and updating may be repeated until the center point of new data is not changed.

According to FIG. 6, the second clustering step (S220) is a step of calculating a distance between each data and an existing center value and a new center value (S221), and classifying the data into clusters of closer center values (S222). , Recalculating and updating the central value for the cluster of the new central value (S223), and determining whether the updated new central value has changed (S224). In this case, the second clustering step S220 may use a K-means clustering technique.

If it is determined that the updated new center value has changed, it may return to step S221 again. Conversely, if it is determined that the updated new center value has not changed, the process may proceed to the next step S310.

In addition, according to FIG. 6, the step of calculating the first evaluation index value (S310) includes selecting one target cluster from among the formed clusters and calculating the first average probability (S311), one different from the selected target cluster. It may include calculating a second average probability of the two clusters in which the clusters of are mixed (S312), and calculating a first evaluation index based on the first average probability and the second average probability (S313). .

In addition, the step of calculating the second evaluation index value (S320) includes selecting one target cluster from among the formed clusters and calculating the first average probability (S321), by mixing the selected target cluster with the other cluster. It may include calculating a second average probability of the two clusters (S322), and calculating a second evaluation index based on the first average probability and the second average probability (S323).

According to FIG. 6, in the calculating of the first average probability (S311, S321), calculating a probability density function (PDF) of the target cluster by selecting one cluster not selected as the target cluster as the target cluster (S3111 , S3211), and calculating the average probability of data of each cluster (S3112 and S3212).

According to FIG. 6, the steps of calculating the second average probability (S312, S322) include calculating a probability density function (PDF) for two clusters in which the target cluster and the other cluster are mixed (S3121, S3221). , And calculating the maximum value of the average probability of the target cluster for data of two clusters of all combinations (S3122 and S3222). At this time, the maximum value of the average probability is referred to as the second average probability.

According to FIG. 6, calculating the first evaluation index or the second evaluation index based on the first average probability and the second average probability (S313, S323) includes a second average probability compared to the first average probability for all clusters. Calculating the average probability ratio by calculating the ratio of (S3131, S3231), determining whether all clusters have been selected as target clusters (S3132, S3232), and evaluating the minimum value of all the calculated average probability ratios for the cluster It may include the steps (S3133, S3233) designating the index.

In steps S3132 and S3232, if it is determined that all clusters have not been selected as target clusters, calculating a probability density function (PDF) of the target cluster by selecting one cluster not selected as the target cluster as the target cluster (S3111, S3211) can be returned. In addition, in steps S3132 and S3232, if it is determined that all clusters have been selected as target clusters, the process may proceed to the next steps S3133 and S3233.

7 is a diagram showing a method of calculating a probability density function according to a second embodiment of the present invention.

Referring to FIG. 7, one cluster may be selected as a target cluster, and probability density functions of features within the target cluster may be calculated. In addition, according to FIG. 7, a probability density function can be calculated for all two clusters in which a target cluster and one other cluster are mixed. In this case, the method of calculating the probability density function used may be used in the step of calculating the probability density function according to FIG. 6.

8 is a diagram showing a method of calculating an average probability according to a second embodiment of the present invention.

Referring to FIG. 8, the average probability may be calculated by substituting feature vectors in the target cluster into the probability density function of the target cluster.

In addition, according to FIG. 8, the average probability may be calculated by substituting feature vectors in the target cluster into the probability density function of two clusters.

At this time, the average probability ratio refers to calculating the ratio between the calculated average probability of the target cluster and the average probability of the target cluster in two clusters, as shown in [Equation 9] below.

The ratio of the average probability as described above may be used in the step of calculating the ratio of the average probability according to FIG. 6.

9 and 10 are diagrams showing an example of analyzing a value obtained by calculating a probability density function according to a second embodiment of the present invention.

According to FIG. 9, when the values of the average probability obtained by substituting the probability density function of two clusters are relatively small, it may mean that the two clusters are relatively far apart.

According to FIG. 10, when the values of the average probability obtained by substituting the probability density function of two clusters are relatively large, it may mean that the two clusters are relatively even. In this case, since the two clusters are close together, they can be recognized as one cluster rather than being separated into each cluster.

In this way, it is possible to determine whether the number of clusters is increased or maintained as it is by comparing the value of the average probability obtained by substituting the probability density function.

So far, the present invention has been looked at around its preferred embodiments. Those of ordinary skill in the art to which the present invention pertains will be able to understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative point of view rather than a limiting point of view. The scope of the present invention is shown in the claims rather than the above description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

10: tube
20: sensor
30: judgment

Claims (15)

Extracting features from the signal;
A clustering step of forming a cluster from the features;
Calculating an evaluation index for the cluster;
Determining the number of clusters based on the evaluation index; And
Recognizing a leak state of the tube based on the number of clusters; including, real-time tube leak state recognition method.
The method of claim 1,
The above features are:
Root-mean-square, Shape factor, Kurtosis value, Square-mean-root, Peak-to-peak value, Why It is at least one selected from the group consisting of a Skewness value, an impulse factor, and a Crest factor.
The method of claim 1,
The step of extracting the features,
A method for real-time tube leakage status recognition by extracting features from an existing signal and an acquired signal.
The method of claim 3,
The clustering step,
A first clustering step of performing clustering by setting the number of clusters to K; And
And a second clustering step of performing clustering in which the number of clusters is set to K+1.
The method of claim 4,
The first clustering step,
The method for real-time tube leakage status recognition by averaging the features extracted from the existing signal to obtain an existing center point.
The method of claim 4,
The first clustering step,
And performing clustering by mixing the features extracted from the acquired signal with the features extracted from the existing signal.
The method of claim 5,
The second clustering step,
The existing center point is fixed as it is, and only a new center point for a feature extracted from the acquired signal is updated.
The method of claim 1,
The step of calculating the evaluation index,
Setting the number of clusters to K and calculating a first evaluation index using clustering and average probability; And
Comprising a second evaluation index using clustering and an average probability after setting the number of clusters to K+1; to include, real-time tube leakage state recognition method.
The method of claim 8,
The step of calculating the first evaluation index,
Calculating a first average probability by selecting one target cluster from among the clusters;
Calculating a second average probability of two clusters in which the selected target cluster and another cluster are mixed; And
Comprising the first evaluation index based on the first average amplification and the second average probability; containing, real-time tube leakage state recognition method.
The method of claim 8,
The step of calculating the second evaluation index,
Calculating a first average probability by selecting one target cluster from among the clusters;
Calculating a second average probability of two clusters in which the selected target cluster and another cluster are mixed; And
Comprising the second evaluation index based on the first average amplification and the second average probability; containing, real-time tube leakage state recognition method.
The method of claim 9 or 10,
The evaluation index includes an average probability ratio,
The average probability ratio is,
It is a value obtained by dividing the second average probability with respect to the first average probability.
The method of claim 8,
The step of determining the number of clusters,
To select the number of clusters having a large evaluation index among the first evaluation index and the second evaluation index, real-time tube leakage state recognition method.
The method of claim 12,
The step of recognizing,
When the determined number of clusters increases, it is recognized that the tube is in a leak state,
When the determined number of clusters is maintained, the tube is recognized as being in a normal state.
A tube through which the flowing material flows;
A sensor installed around the tube;
A determination unit that extracts features from the sensor, forms a cluster from the features, calculates an evaluation index for the cluster, determines the number of clusters based on the evaluation index, and recognizes the leak state of the tube in real time Including; real-time tube leakage state recognition system.
The method of claim 14,
The tube,
A system for real-time tube leakage status recognition that is used in a circulating fluidized bed boiler

Images