KR102303406B1 - Method for something wrong diagnosis of industrial equipment and the device - Google Patents

Abstract

The present invention relates to an apparatus and method for diagnosing the presence or absence of abnormalities such as abnormal driving conditions or defects of industrial equipment. The present invention provides a first processing unit that receives a normal signal offline from a sensor installed in an industrial facility, a second processing unit that receives a sensor signal in real time from a sensor installed in the industrial facility, and the first processing unit and the second processing unit. and a similarity calculating unit for calculating the similarity of signals, and a determining unit for determining whether there is an abnormality in the industrial equipment according to the calculated similarity result. And the first processing unit and the second processing unit include an entropy calculation unit for calculating the entropy of the normal signal and the sensor signal, and a probability distribution calculation unit for calculating the probability distribution of the calculated entropy. The similarity between the probability distribution of entropy and the probability distribution of entropy of the sensor signal is calculated.

Description

An apparatus and method for diagnosing abnormalities in industrial facilities

The present invention relates to an apparatus and method for detecting an abnormal driving state of an industrial facility, and more particularly, to an apparatus and method capable of diagnosing an abnormality or defect of an industrial facility through a signal of a sensor installed in the industrial facility during operation of the industrial facility will be.

Management of various industrial facilities has long been recognized as a human domain. Workers with a lot of experience analyze the data acquired from various industrial facilities to find out the abnormal state of the facility, find the cause of the defect, and take action before a serious failure occurs or the facility stops suddenly.

However, there are several problems that it is difficult to accurately collect data generated from complex and diverse facilities, and that it takes a considerable amount of time to evaluate and analyze it. In addition, there is a problem in that a misdiagnosis occurs when a situation that the operator has not experienced occurs or the data is misinterpreted.

Therefore, companies are introducing new methods to more efficiently and safely manage various industrial facilities as a way to improve productivity and secure corporate competitiveness. has been

For example, Korea Patent No. 10-1302519 (Prior Document 1) is an invention for diagnosing the presence or absence of an abnormality in equipment, and a configuration for predicting the presence or absence of abnormality and lifespan of an equipment by analyzing the vibration of an industrial equipment is disclosed. However, Prior Document 1 is a technology for diagnosing abnormalities based on only mechanical vibration information that is essentially generated in the process of actually driving industrial facilities, and has a very limited application range. .

As another example, Korean Patent No. 10-1271993 discloses an abnormal signal display method for monitoring industrial plant equipment, and analyzes the signal of a detection sensor installed in the equipment system to find a malfunctioning device.

However, in Prior Document 2, it is a method of checking whether an abnormal signal occurs in the operating state information of facilities, and when such an abnormal signal occurs, the abnormal signal is traced in the reverse and forward directions of the system flow at the corresponding location to find the malfunctioning device. It can be cumbersome to track the abnormal signal, and when an abnormal signal occurs in two or more facilities in a given system, it is difficult to find the faulty device.

In addition, in the prior literature, when instantaneous noise unrelated to normal operation occurs, it is mistakenly judged as an abnormal signal, and equipment operation is stopped. In this case, the operation of the equipment is stopped automatically or manually, and there was a cumbersome point that it had to be operated again after a series of inspections were completed. I had a problem I couldn't.

Accordingly, it is an object of the present invention to solve the above problems, and to provide an apparatus and method for diagnosing abnormalities in industrial equipment, which can accurately diagnose the presence or absence of abnormalities in industrial equipment.

Another object of the present invention is to provide an apparatus and method for diagnosing abnormalities in industrial equipment so that some generated noise is determined as noise rather than a defect so that continuous and stable operation of industrial equipment can be performed.

The present invention for achieving the above object, a first processing unit receiving a normal signal off-line from a sensor installed in an industrial facility; a second processing unit receiving a sensor signal in real time from a sensor installed in an industrial facility; a similarity calculator configured to calculate a degree of similarity between the signals processed by the first processor and the second processor; and a determination unit for determining whether there is an abnormality in the industrial equipment according to the calculated similarity result.

The first and second processing units include an entropy calculation unit for calculating entropy of a normal signal and a sensor signal, and a probability distribution calculation unit for calculating a probability distribution of the calculated entropy, wherein the similarity calculation unit includes: Calculate the similarity between the probability distribution and the probability distribution of the entropy of the sensor signal.

에 의해 엔트로피를 계산한다. 여기서 p1은 m 길이의 거리 행렬에서, 행렬의 각 수가 r보다 큰 개수/전체개수, p2는 m+1 길이의 거리 행렬에서 행렬의 각 수가 r보다 큰 개수/전체 개수를 의미하고, 상기 r은 표준편차 ×0.2이다.The entropy calculation unit,

Calculate the entropy by Here, p1 is the number/total number of each matrix greater than r in a distance matrix of length m, p2 is the number/total number of each matrix greater than r in a distance matrix of length m+1, wherein r is It is standard deviation × 0.2.

에 의해 유사도를 계산한다. 여기서 상기 pdf1, pdf2는 각각 정상신호 및 실시간신호로부터 소정 개수의 엔트로피 계산 값을 기반으로 한 정규분포 계산값을 의미한다.The similarity calculator,

Calculate the similarity by Here , pdf 1 and pdf 2 refer to normal distribution calculated values based on a predetermined number of entropy calculation values from a normal signal and a real-time signal, respectively.

The determination unit determines whether there is an abnormality by counting the number of times the current similarity value is greater than the 'average + standard deviation × 3' of the recent number of similarity values.

According to another feature of the present invention, a signal measuring step of measuring a normal signal and a real-time signal through a sensor installed in an industrial facility; an entropy calculation step of calculating entropy for the normal signal and the real-time signal; a probability distribution calculation step of calculating a probability distribution of the calculated entropy, respectively; a probability distribution similarity calculation step of calculating a similarity of probability distributions of the normal signal and the real-time signal; and a determination step of determining whether there is an abnormality in the industrial equipment through the calculated similarity.

에 의해 계산하며, 여기서 p1은 m 길이의 거리 행렬에서, 행렬의 각 수가 r보다 큰 개수/전체개수, p2는 m+1 길이의 거리 행렬에서 행렬의 각 수가 r보다 큰 개수/전체 개수를 의미하고, 상기 r은 표준편차 ×0.2이다.The entropy calculation step is,

where p1 is the number/total number of each matrix greater than r in a distance matrix of length m, and p2 is the number/total number of each matrix greater than r in a distance matrix of length m+1. and r is the standard deviation × 0.2.

에 의해 유사도를 계산하면, 여기서 상기 pdf1, pdf2는 각각 정상신호 및 실시간 신호로부터 소정 개수의 엔트로피 계산 값을 기반으로 한 정규분포 계산값을 의미한다.In the step of calculating the probability distribution similarity,

When the similarity is calculated by , where pdf 1 and pdf 2 refer to a normal distribution calculated value based on a predetermined number of entropy calculation values from a normal signal and a real-time signal, respectively.

In the judging step, if the current similarity value is more than a predetermined number of times than the 'average + standard deviation ×3' of the recent predetermined number of similarity values, it is determined as an abnormal occurrence of industrial equipment, and if less than the number, it is determined as a temporary outlier do.

According to the apparatus and method for diagnosing industrial equipment abnormality of the present invention as described above, when an abnormality occurs during operation of industrial equipment, there is an effect of diagnosing the abnormality.

In addition, the present invention operates the diagnostic device in a pattern in which, when a predetermined condition is satisfied, even if noise occurs, it is judged as general noise rather than an abnormality or defect of industrial equipment and is ignored, so that noise not related to normal operation of industrial equipment, etc. It is possible to prevent the problem that the operation of industrial equipment is stopped due to this.

1 is a configuration diagram illustrating an apparatus for detecting abnormalities in industrial facilities according to a preferred embodiment of the present invention;
2 is a flowchart illustrating a method for detecting abnormalities in industrial facilities according to a preferred embodiment of the present invention;
3 to 5 are exemplary views for explaining the entropy calculation process of FIG. 2

Objects and effects of the present invention, and technical configurations for achieving them will become clear with reference to the embodiments described below in detail in conjunction with the accompanying drawings. In describing the present invention, if it is determined that a detailed description of a well-known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, the terms described below are terms defined in consideration of functions in the present invention, which may vary according to intentions or customs of users and operators.

However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. Only the present embodiments are provided so that the disclosure of the present invention is complete, and to fully inform those of ordinary skill in the art to which the present invention belongs, the scope of the invention, the present invention is defined by the scope of the claims will only be Therefore, the definition should be made based on the content throughout this specification.

The present invention is to determine the presence or absence of an abnormality occurring during operation of an industrial facility and the degree thereof, and the present invention will be described in more detail below based on the embodiments shown in the drawings.

As shown in FIG. 1 , the industrial facility abnormality detection apparatus includes a first processing unit 120 and a second processing unit 130 . The first processing unit 120 is a part that processes the normal signal acquired by the sensor 101 installed in the industrial facility in an offline state, and the second processing unit 130 is the sensor 101 acquired in real time according to the driving of the industrial facility. It refers to the part that processes real-time signals.

The first processing unit 120 and the second processing unit 130 include entropy calculation units 122 and 132 and probability distribution calculation units 124 and 134, respectively.

The entropy calculation unit 122 and the probability distribution calculation unit 124 of the first processing unit 120 each have entropy and probability for normal signals (limited to 20 in the embodiment) acquired by the sensors 101 in an offline state. The distribution has been calculated in advance. An entropy value and a probability distribution value by the first processing unit 120 may be a reference signal.

The entropy calculation unit 132 and the probability distribution calculation unit 134 of the second processing unit 130 obtain a sensor signal online in real time and calculate an entropy value and a probability distribution value.

The entropy calculators 122 and 132 calculate an entropy value according to Equation 1 below.

Here, p1 denotes the number/total number of each matrix greater than r in a distance matrix of length m, and p2 denotes the number/total number of each matrix greater than r in a distance matrix of length m+1. And r is the standard deviation × 0.2.

A similarity calculation unit 140 is provided for calculating a similarity between the probability distribution value of the first processing unit 120 and the probability distribution value of the second processing unit 130 . The similarity calculator 140 is obtained by Equation 2 below.

where pdf 1 above, pdf 2 refers to a normal distribution calculated value based on a predetermined number of entropy calculation values from a normal signal and a real-time signal, respectively. In addition, the pdf 1 and pdf 2 can be obtained using the mean (μ) and the standard deviation (σ), and the mean and the standard deviation can be found from the 20 entropy calculation values described above.

A determination unit 150 is provided for judging the presence or absence of abnormalities or defects in industrial equipment based on the degree of similarity calculated by the similarity calculation unit 140 . Here, the determination unit 150 does not simply determine the abnormality or defect of the industrial equipment, but determines that the signal that is suddenly generated is noise unrelated to the normal operation of the industrial equipment, so that it is not determined that the abnormality of the industrial equipment has occurred. That is, if all of the instantaneous and sudden noises are judged as abnormal or defective signals, it may be difficult to operate industrial equipment normally. Of course, such noise should be a signal that is not substantially related to the normal operation of industrial equipment.

The determination unit 150 according to an embodiment of the present invention determines that a signal having the calculated similarity value greater than 'average + standard deviation × 3' of a predetermined number of recent similarity values is less than a predetermined number of times as noise and ignores it. . That is, it is regarded as having no defects or abnormalities. In addition, it is determined that there is an abnormality or a defect only when the calculated similarity value is greater than the 'mean + standard deviation x 3' of all similarity values more than a predetermined number of times. In the embodiment, the number of times is based on 5 times, but it is natural that it may be applied differently depending on the characteristics of industrial facilities.

Next, a method for diagnosing an abnormal state or a defective state of equipment using the industrial equipment abnormality diagnosis apparatus configured as described above will be described.

The industrial facility abnormality diagnosis method of the present invention is largely a signal measurement step (s100), an entropy calculation step (s110), a probability distribution calculation step (s120), a similarity calculation step for a probability distribution (s130), and the presence or absence of abnormalities in industrial facilities It is performed including the determination step (s140) of determining, and each of these steps will be described in detail.

In the non-driving state (eg, offline state) of the industrial equipment, the first processing unit 120 acquires a predetermined number, for example, 20 normal signals from the sensors 101 installed in the industrial equipment (s101), and based on this Thus, the entropy calculation unit 122 and the probability distribution calculation unit 124 calculate an entropy value and a probability distribution calculation value, respectively (s111) (s121). The calculated values are stored in a series of memories (not shown) or the like.

As described above, when the industrial equipment is driven in a state in which the entropy value and the probability distribution value for the normal signal are stored offline, the sensors 101 installed in the industrial equipment will sense various signals of the industrial equipment in real time (s102). And based on the sensed real-time signal, the entropy calculation unit 132 of the second processing unit 130 calculates entropy using Equation 1 (s112).

Entropy comes out high when the real-time sensor signal is applied randomly instead of according to a certain pattern, whereas when it has a certain period, it comes out low in the case of a repetitive pattern. Therefore, it can be seen that the more normal the signal, the higher the entropy, and the lower the entropy when there is an abnormality or defect. That is, the normal signal is randomly applied from the sensor, and when there is an abnormality or defect, a corresponding signal is periodically generated.

The reason for obtaining such an entropy value is to extract the characteristics of the normal signal and the defect signal.

In the entropy calculation process, as shown in FIG. 3 , in the state where there are N sample signals, 1/5 is set as the m signal. Then, as shown in FIG. 4, the difference between the two signals is calculated while continuously increasing the value by 1 for the i signal for the entire signal length and the j signal which is 1/5 of the total signal length. If this is expressed as a matrix, it is shown in FIG. 5 . 5 shows a distance matrix of length m.

And looking at the calculation process using the matrix of FIG. 5, the diagonal components (A) are the same signals [ie, (1,1), (2,2)... ] does not need to be calculated as This is unnecessary calculation. Also, the signals B and C on the upper/lower side with respect to the diagonal are identical to each other. Therefore, it is not necessary to calculate all of these signals. That is, it is sufficient to calculate only a part of the entire matrix, such as part B or part C. As such, it can be seen that the present invention can reduce the calculation time by performing unnecessary calculations and calculations of overlapping parts only once.

Accordingly, the entropy calculator 132 may calculate the entropy of the real-time signal using Equation (1).

Then, when the entropy of the real-time signal is calculated, the probability distribution calculator 134 calculates the probability distribution of the calculated entropy (s122). That is, the probability distribution value of the calculated entropy for the last 20 signals is calculated. According to this process, both the entropy probability distribution value of the normal signal and the entropy probability distribution value of the real-time signal are obtained.

The similarity calculator 140 calculates a similarity between the obtained entropy probability distribution value of the normal signal and the entropy probability distribution value of the real-time signal (s130). The degree of similarity may be obtained using Equation 2 above.

Then, the calculated similarity value is transmitted to the determination unit 150 . Accordingly, the determination unit 150 determines whether there is an abnormality or a defect in the industrial equipment according to a predetermined determination condition (s140). That is, the determination unit 150 determines that there is an abnormality or a defect in the industrial equipment when the current similarity value is a predetermined number of times, for example, 5 times or more, than the 'average + standard deviation × 3' of the recent 20 similarity values. do. The judgment result will be provided so that the operator or the like can know.

However, if the number is less than a predetermined number, it is determined as a temporary noise rather than an abnormality or defect and is ignored. That is, if the number of times that the current similarity value is greater than the 'mean + standard deviation ×3' of the 20 recent similarity values is less than 5 times, the industrial equipment is not considered to have an abnormality or defect, but is determined as a temporary outlier. Therefore, under these conditions, the operation of the industrial equipment can be continuously performed without stopping.

Although described with reference to the illustrated embodiments of the present invention as described above, these are merely exemplary, and those of ordinary skill in the art to which the present invention pertains can use various functions without departing from the spirit and scope of the present invention. It will be apparent that modifications, variations, and other equivalent embodiments are possible. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

101: sensor
120, 130: first and second processing units
122, 132: entropy calculator
124, 134: probability distribution calculator
140: similarity calculator
150: judgment unit

Claims (9)

a first processing unit receiving a normal signal off-line from a sensor installed in an industrial facility;
a second processing unit receiving a sensor signal in real time from a sensor installed in an industrial facility;
a similarity calculator configured to calculate a degree of similarity between the signals processed by the first processor and the second processor; and
A determination unit for determining whether there is an abnormality in the industrial equipment according to the calculated similarity result;
The first processing unit and the second processing unit,
an entropy calculation unit for calculating the entropy of the normal signal and the sensor signal; and
It is configured to include a probability distribution calculator for calculating the probability distribution of the calculated entropy,
The similarity calculator calculates a similarity between a probability distribution of entropy of a normal signal and a probability distribution of entropy of a sensor signal,
The judging unit,
If the current similarity value is more than a predetermined number of times than the 'average + standard deviation × 3' of the recent predetermined number of similarity values, it is determined that an abnormality occurs in the industrial equipment, and if it is less than the number, it is determined as a temporary outlier. An abnormality diagnosis device for industrial equipment. delete The method of claim 1,
The entropy calculation unit,

An abnormality diagnosis device for industrial equipment that calculates entropy by
Here, p1 is the number/total number of each matrix greater than r in a distance matrix of length m, p2 is the number/total number of each matrix greater than r in a distance matrix of length m+1, wherein r is Standard deviation × 0.2. The method of claim 1,
The similarity calculator,

An abnormality diagnosis device for industrial equipment that calculates the degree of similarity by
Here , pdf 1 and pdf 2 mean normal distribution calculated values based on a predetermined number of entropy calculation values from normal signals and real-time signals, respectively. delete A signal measurement step of measuring a normal signal and a real-time signal through a sensor installed in an industrial facility;
an entropy calculation step of calculating entropy for the normal signal and the real-time signal;
a probability distribution calculation step of calculating a probability distribution of the calculated entropy, respectively;
a probability distribution similarity calculation step of calculating the similarity of probability distributions of the normal signal and the real-time signal; and
Comprising a determination step of determining whether there is an abnormality in the industrial equipment through the calculated similarity,
The judging step is
If the current similarity value is more than a predetermined number of times than the 'average + standard deviation × 3' of the recent predetermined number of similarity values, it is determined that an abnormality occurs in the industrial equipment, and if it is less than the number, it is determined as a temporary outlier. Method of diagnosing the presence of abnormalities in industrial facilities. 7. The method of claim 6,
The entropy calculation step is,
ceremony

where p1 is the number/total number of each matrix greater than r in a distance matrix of length m, and p2 is the number/total number of each matrix greater than r in a distance matrix of length m+1. and, wherein r is a standard deviation × 0.2, an abnormality diagnosis method of industrial equipment. 7. The method of claim 6,
The probability distribution similarity calculation step includes:

, where pdf 1 and pdf 2 mean normal distribution calculated values based on a predetermined number of entropy calculated values from normal signals and real-time signals, respectively. delete

Images