KR102492514B1 - Alarm method using sequential probability ratio - Google Patents

Alarm method using sequential probability ratio Download PDF

Info

Publication number
KR102492514B1
KR102492514B1 KR1020220136639A KR20220136639A KR102492514B1 KR 102492514 B1 KR102492514 B1 KR 102492514B1 KR 1020220136639 A KR1020220136639 A KR 1020220136639A KR 20220136639 A KR20220136639 A KR 20220136639A KR 102492514 B1 KR102492514 B1 KR 102492514B1
Authority
KR
South Korea
Prior art keywords
residuals
alarm
state
distribution
probability ratio
Prior art date
Application number
KR1020220136639A
Other languages
Korean (ko)
Inventor
조만영
Original Assignee
가온플랫폼 주식회사
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 가온플랫폼 주식회사 filed Critical 가온플랫폼 주식회사
Priority to KR1020220136639A priority Critical patent/KR102492514B1/en
Application granted granted Critical
Publication of KR102492514B1 publication Critical patent/KR102492514B1/en

Links

Images

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18Status alarms
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/18Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B31/00Predictive alarm systems characterised by extrapolation or other computation using updated historic data

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Mathematical Optimization (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Emergency Management (AREA)
  • Mathematical Analysis (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Operations Research (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Algebra (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The present invention discloses an alert method using a sequential probability ratio, which includes the steps of: calculating the residual between a measured value and an expected value using operation data of an operation target in an alarm device; calculating a sequential probability ratio using the residual; determining the state of an operation target by applying the sequential probability ratio to an alarm condition according to the probability distribution of residuals; and generating an alarm if the condition is abnormal.

Description

순차확률비를 이용한 경보 방법{ALARM METHOD USING SEQUENTIAL PROBABILITY RATIO}Alarm method using sequential probability ratio {ALARM METHOD USING SEQUENTIAL PROBABILITY RATIO}

본 발명은 순차확률비를 이용한 경보 방법에 관한 것으로서, 더욱 상세하게는 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하고, 비정상 상태이면 경보를 발생하는 기술에 관한 것이다.The present invention relates to an alarm method using a sequential probability ratio, and more particularly, to a technique for determining the state of an operating target by applying the sequential probability ratio to an alarm condition according to a probability distribution of residuals, and generating an alarm if the condition is abnormal. it's about

종래에는 플랜트 등 산업설비에 다양한 설비가 존재하고, 설비 구동에 따른 운전데이터를 생성하며, 운전데이터를 분석하여 산업설비에 이상이 있는지 판별하였고, 이상이 있으면 조기 경보를 발생하였다.Conventionally, various facilities exist in industrial facilities such as plants, operation data is generated according to the operation of the facilities, and operation data is analyzed to determine whether there is an abnormality in the industrial facilities, and an early warning is generated if there is an abnormality.

예를 들어 특허문헌 1 또는 특허문헌 2와 같이 운전데이터의 실측값과 기대값(또는 예측값)을 이용하여 산업설비에 이상이 있는지 경보를 발생할 수 있다.For example, as in Patent Document 1 or Patent Document 2, it is possible to generate an alarm whether there is an abnormality in an industrial facility by using the actual measured value and expected value (or predicted value) of operation data.

특허문헌 1은 실측값과 예측값을 이용하여 태그인덱스를 결정하고, 태그인덱스를 지수화하여 경보를 지정하는 방식이다. 특허문헌 2는 실측값이나 잔차와 설정값(또는 임계값)을 비교하고, 설정값의 범위에 벗어나는 횟수에 대응하여 알람의 규칙을 설정하는 방식이다.Patent Literature 1 is a method of determining a tag index using an actual measured value and a predicted value, and specifying an alarm by indexing the tag index. Patent Literature 2 is a method of comparing an actual value or residual with a set value (or threshold) and setting an alarm rule corresponding to the number of times out of the range of the set value.

그러나 종래에는 특허문헌 1 또는 특헌문헌 2의 방식을 활용하면 유효하지 않은 조기경보가 너무 발생하는 문제점이 있다.However, conventionally, when the method of Patent Document 1 or Patent Document 2 is used, there is a problem in that invalid early warning occurs too much.

한국공개특허 제10-2013-0042523호Korean Patent Publication No. 10-2013-0042523 한국공개특허 제10-2017-0125237호Korean Patent Publication No. 10-2017-0125237

상기 문제점을 해결하기 위하여 본 발명은 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하고, 비정상 상태이면 경보를 발생하는 순차확률비를 이용한 경보 방법을 제공한다.In order to solve the above problem, the present invention provides an alarm method using a sequential probability ratio that applies the sequential probability ratio to an alarm condition according to a probability distribution of residuals to determine the state of a driving target and generates an alarm if it is in an abnormal state.

상기의 해결하고자 하는 과제를 위한 본 발명의 실시예에 따른 순차확률비를 이용한 경보 방법은, 경보 장치에서 운전대상의 운전데이터를 이용하여 실측값과 기대값 간의 잔차를 연산하는 단계; 상기 잔차를 이용하여 순차확률비를 연산하는 단계; 상기 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하는 단계 및 상기 상태가 비정상이면 경보를 발생하는 단계를 포함하고, 상기 운전대상의 비정상 상태에 대한 잔차의 확률분포는 잔차의 평균이 증가된 상태의 분포, 잔차의 평균이 감소된 상태의 분포 및 잔차의 편차가 증가된 상태의 분포인 것을 특징으로 한다.An alarm method using a sequential probability ratio according to an embodiment of the present invention for the above problem to be solved includes the steps of calculating a residual between an actual value and an expected value using driving data of a driving target in an alarm device; calculating a sequential probability ratio using the residual; Determining the state of the operating target by applying the sequential probability ratio to an alarm condition according to the probability distribution of the residual, and generating an alarm if the state is abnormal, and the residual probability for the abnormal state of the operating target The distribution is characterized by being a distribution in which the mean of the residuals is increased, a distribution in which the mean of the residuals is decreased, and a distribution in which the variance of the residuals is increased.

상기 경보조건은 비정상, 보류 및 정상 상태로 분류되고, [수식 1] 및 [수식 2]에 의해 결정되며, 상기 순차확률비는 [수식 3], [수식 4] 및 [수식 5]에 의해 결정되는 것을 특징으로 할 수 있다.The alarm conditions are classified into abnormal, pending, and normal states, and are determined by [Equation 1] and [Equation 2], and the sequential probability ratio is determined by [Equation 3], [Equation 4], and [Equation 5]. can be characterized as being

[수식 1][Formula 1]

Figure 112022111425798-pat00001
Figure 112022111425798-pat00001

[수식 2][Formula 2]

Figure 112022111425798-pat00002
Figure 112022111425798-pat00002

[수식 3][Formula 3]

Figure 112022111425798-pat00003
Figure 112022111425798-pat00003

[수식 4][Formula 4]

Figure 112022111425798-pat00004
Figure 112022111425798-pat00004

[수식 5][Formula 5]

Figure 112022111425798-pat00005
Figure 112022111425798-pat00005

상기 경보 장치는

Figure 112022111425798-pat00006
,
Figure 112022111425798-pat00007
,
Figure 112022111425798-pat00008
Figure 112022111425798-pat00009
보다 작으면 정상 상태로 판별하고,
Figure 112022111425798-pat00010
,
Figure 112022111425798-pat00011
Figure 112022111425798-pat00012
보다 크면 잔차의 평균이 증가된 비정상 상태로 판별하며,
Figure 112022111425798-pat00013
,
Figure 112022111425798-pat00014
Figure 112022111425798-pat00015
보다 크면 잔차의 평균이 감소된 비정상 상태로 판별하고,
Figure 112022111425798-pat00016
,
Figure 112022111425798-pat00017
Figure 112022111425798-pat00018
보다 크면 잔차의 편차가 증가된 비정상 상태로 판별하는 것을 특징으로 할 수 있다.the warning device
Figure 112022111425798-pat00006
,
Figure 112022111425798-pat00007
,
Figure 112022111425798-pat00008
go
Figure 112022111425798-pat00009
If it is less than, it is determined as a normal state,
Figure 112022111425798-pat00010
,
Figure 112022111425798-pat00011
go
Figure 112022111425798-pat00012
If it is greater than this, it is determined as an abnormal state with an increased average of the residuals,
Figure 112022111425798-pat00013
,
Figure 112022111425798-pat00014
go
Figure 112022111425798-pat00015
If it is greater than this, it is determined as an abnormal state with a reduced average of the residuals,
Figure 112022111425798-pat00016
,
Figure 112022111425798-pat00017
go
Figure 112022111425798-pat00018
If it is greater than , it can be characterized in that it is determined as an abnormal state in which the deviation of the residual is increased.

본 발명은 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하고, 비정상 상태이면 경보를 발생함으로써, 유효하지 않은 조기경보의 횟수를 감소시킬 수 있다.The present invention determines the state of the driving target by applying the sequential probability ratio to the alarm condition according to the probability distribution of the residuals, and generates an alarm if it is in an abnormal state, thereby reducing the number of invalid early warnings.

도 1은 본 발명의 실시예에 따른 순차확률비를 이용한 경보 장치를 도시한 블록도이다.
도 2는 조기 경보의 판별 방법을 도시한 예이다.
도 3은 잔차에 따른 확률분포를 도시한 예이다.
도 4는 정상 상태의 잔차 데이터를 그래프로 도시한 예이다.
도 5는 정상 상태의 순차확률비 가능도를 그래프로 도시한 예이다.
도 6은 잔차의 평균이 증가된 데이터를 그래프로 도시한 예이다.
도 7은 잔차의 평균이 증가된 상태의 순차확률비 가능도를 그래프로 도시한 예이다.
도 8은 잔차의 평균이 감소된 데이터를 그래프로 도시한 예이다.
도 9는 잔차의 평균이 감소된 상태의 순차확률비 가능도를 그래프로 도시한 예이다.
도 10은 잔차의 편차가 증가된 데이터를 그래프로 도시한 예이다.
도 11은 잔차의 편차가 증가된 상태의 순차확률비 가능도를 그래프로 도시한 예이다.
도 12는 임계값과 순착확률비를 비교한 조기 경보를 그래프로 도시한 예이다.
도 13은 본 발명의 실시예에 따른 순차확률비를 이용한 경보 방법을 도시한 흐름도이다.
1 is a block diagram illustrating an alarming device using sequential probability ratios according to an embodiment of the present invention.
2 is an example illustrating a method for determining an early warning.
3 is an example of a probability distribution according to residuals.
4 is an example of a graph of residual data in a steady state.
5 is an example of a graph showing the probability of a sequential probability ratio in a steady state.
6 is an example of a graph showing data in which the average of the residuals is increased.
7 is an example of a graph illustrating the probability of sequential probability ratios in a state in which the mean of residuals is increased.
8 is an example of a graph showing data in which the average of the residuals is reduced.
9 is an example of a graph showing the probability of a sequential probability ratio in a state in which the average of the residuals is reduced.
10 is an example of a graph showing data with increased residual deviation.
11 is an example of a graph showing the probability of a sequential probability ratio in a state in which the deviation of the residual is increased.
12 is an example of a graph showing an early warning in which a threshold value and a sequential arrival probability ratio are compared.
13 is a flowchart illustrating an alerting method using sequential probability ratios according to an embodiment of the present invention.

이하 첨부 도면들 및 첨부 도면들에 기재된 내용들을 참조하여 본 발명의 실시예를 상세하게 설명하지만, 본 발명이 실시예에 의해 제한되거나 한정되는 것은 아니다.Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and the contents described in the accompanying drawings, but the present invention is not limited or limited by the embodiments.

운전데이터에 대한 실시간 기대값 계산을 기반으로 이상 상태 탐지 시 주로 사용되는 임계값(threshold) 기반의 경보 발생 기능은 각 신호(Tag)별로 설정한 임계값에 따라 경보 발생 여부가 결정된다. 즉, 경보를 발생시키는 기준인 임계값을 적정수준 이하로 설정하면 오경보가 발생하거나, 적정수준 이상으로 설정하면 이상상태를 탐지하지 못하는 경우가 발생할 수 있다.Based on the real-time expected value calculation for operation data, the threshold-based alarm function, which is mainly used when detecting anomalies, determines whether an alarm is generated according to the threshold set for each tag. That is, if the threshold value, which is a criterion for generating an alarm, is set below an appropriate level, a false alarm may occur, or if it is set above an appropriate level, an abnormal state may not be detected.

따라서 종래에는 적정 수준의 임계값을 설정하는 것이 매우 중요하나 임계값의 설정 후 다른 신호 또는 외부 환경 등에 따라 계산되는 기대값이 다르기 때문에, 영구적으로 사용할 수 있는 임계값을 찾기가 어렵다. 본 발명의 순차 확률비(SPR: Sequential Probability Ratio)는 정상 상태일 때의 데이터의 분포와 비정상 상태일 때의 데이터의 분포의 비율을 기반으로 현재 상태를 판단할 수 있는 방법으로, 별도의 임계값에 대한 지정이 없이 경보를 발생시킬 수 있는 방법이다.Therefore, in the prior art, it is very important to set an appropriate threshold value, but it is difficult to find a threshold value that can be used permanently because the expected value calculated according to other signals or external environments after setting the threshold value is different. The Sequential Probability Ratio (SPR) of the present invention is a method for determining the current state based on the ratio of the distribution of data in a normal state and the distribution of data in an abnormal state. This is a method that can generate an alarm without specifying it.

도 1은 본 발명의 실시예에 따른 순차확률비를 이용한 경보 장치를 도시한 블록도로서, 경보 장치(10)는 순차확률비를 이용하여 경보를 발생할 수 있다. 경보 장치(10)는 수집부(110), 판별부(120) 및 경보부(130)로 구성된다.1 is a block diagram showing an alarm device using a sequential probability ratio according to an embodiment of the present invention, and the alarm device 10 may generate an alarm using a sequential probability ratio. The alarm device 10 is composed of a collection unit 110, a determination unit 120 and an alarm unit 130.

수집부(110)는 운전대상의 운전데이터를 수집한다. 운전대상은 플랜트 또는 산업설비에 구비된 다양한 설비일 수 있고, 설비를 감시하는 센서일 수 있다. 운전데이터는 설비 또는 센서의 구동과 관련된 데이터일 수 있고, 설비 또는 센서의 구동에 따른 결과와 관련된 데이터일 수 있다.The collection unit 110 collects driving data of a driving subject. An operation target may be various facilities provided in a plant or industrial facility, and may be a sensor that monitors the facility. The operation data may be data related to driving of facilities or sensors, or data related to results of driving of facilities or sensors.

판별부(120)는 운전데이터를 분석하여 운전대상이 정상 또는 비정상 상태인지 판별하고, 경보부(130)는 비정상 상태이면 조기 경보를 발생한다. 경보를 출력하는 방식은 소리, 빛 또는 데이터 전송 등 종래의 다양한 방식을 채용할 수 있다.The determining unit 120 analyzes the driving data to determine whether the driving target is in a normal or abnormal state, and the warning unit 130 generates an early warning if the driving target is in an abnormal state. As a method of outputting an alarm, various conventional methods such as sound, light, or data transmission may be employed.

도 2는 조기 경보의 판별 방법을 도시한 예로서, 판별부(120)는 운전대상의 운전데이터를 이용하여 실측값과 기대값 간의 잔차를 연산하고, 잔차를 이용하여 순차확률비를 연산하며, 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별한다. 판별부(120)는 특허문헌 1 또는 특허문헌 2와 같은 방식으로 잔차를 연산할 수 있다.2 is an example of an early warning determination method, in which the determination unit 120 calculates a residual between an actual value and an expected value using driving data of a driving target, and calculates a sequential probability ratio using the residual, The sequential probability ratio is applied to the alarm condition according to the probability distribution of the residuals to determine the state of the operating target. The determination unit 120 may calculate the residual in the same manner as in Patent Document 1 or Patent Document 2.

Figure 112022111425798-pat00019
은 순차확률비를 의미하고,
Figure 112022111425798-pat00020
는 허위 경보 확률을 의미하며,
Figure 112022111425798-pat00021
는 누락 경보 확률을 의미하고,
Figure 112022111425798-pat00022
는 허위 경보율을 의미하며,
Figure 112022111425798-pat00023
는 누락 경보율을 의미한다.
Figure 112022111425798-pat00019
is the sequential probability ratio,
Figure 112022111425798-pat00020
is the false alarm probability,
Figure 112022111425798-pat00021
denotes the missed alert probability,
Figure 112022111425798-pat00022
is the false alarm rate,
Figure 112022111425798-pat00023
is the missing alarm rate.

도 3은 잔차에 따른 확률분포를 도시한 예로서, 순차확률비를 잔차의 확률분포에 적용하기 위해서는, 정상과 비정상 상태의 확률분포를 정의해야 한다.3 is an example of a probability distribution according to residuals. In order to apply a sequential probability ratio to the probability distribution of residuals, it is necessary to define probability distributions in normal and abnormal states.

정상 상태의 잔차 분포는

Figure 112022111425798-pat00024
로 정의되고, 잔차의 평균이 증가된 상태의 분포는
Figure 112022111425798-pat00025
으로 정의되며, 잔차의 평균이 감소된 상태의 분포는
Figure 112022111425798-pat00026
로 정의되고, 잔차의 편차가 증가된 상태의 분포는
Figure 112022111425798-pat00027
으로 정의된다. 즉 설비의 특성상 비정상 상태의 분포는 상황에 따라 3가지 경우가 존재한다.The steady-state residual distribution is
Figure 112022111425798-pat00024
, and the distribution of states with increased mean of residuals is
Figure 112022111425798-pat00025
, and the distribution of the state in which the average of the residuals is reduced is
Figure 112022111425798-pat00026
, and the distribution of states with increased deviation of the residuals is
Figure 112022111425798-pat00027
is defined as In other words, there are three cases of abnormal distribution according to the characteristics of the facility.

순차확률비에 따른 경보조건은 비정상, 보류 및 정상 상태로 분류되고, [수식 1] 및 [수식 2]에 의해 결정된다. 즉 본 발명은 순차확률비를 계산하기 위하여 정상 상태를 가정하는

Figure 112022111425798-pat00028
의 분포
Figure 112022111425798-pat00029
와 비정상 상태를 가정하는
Figure 112022111425798-pat00030
의 분포
Figure 112022111425798-pat00031
의 가능도의 비율(Likelihood ratio)을 계산한다.The alarm condition according to the sequential probability ratio is classified into abnormal, pending, and normal status, and is determined by [Equation 1] and [Equation 2]. That is, the present invention assumes a steady state to calculate the sequential probability ratio
Figure 112022111425798-pat00028
distribution of
Figure 112022111425798-pat00029
assuming an abnormal state with
Figure 112022111425798-pat00030
distribution of
Figure 112022111425798-pat00031
Calculate the likelihood ratio of .

[수식 1][Formula 1]

Figure 112022111425798-pat00032
Figure 112022111425798-pat00032

[수식 2][Formula 2]

Figure 112022111425798-pat00033
Figure 112022111425798-pat00033

여기서

Figure 112022111425798-pat00034
은 가능도의 비율을 의미하고,
Figure 112022111425798-pat00035
은 정상 상태에 대한 잔차의 편차를 의미하며,
Figure 112022111425798-pat00036
은 비정상 상태에 대한 잔차의 편차를 의미하고,
Figure 112022111425798-pat00037
은 측정된 잔차를 의미하며,
Figure 112022111425798-pat00038
은 정상 상태에 대한 잔차의 평균을 의미하고,
Figure 112022111425798-pat00039
은 비정상 상태에 대한 잔차의 평균(평균이 증가된 상태)을 의미한다.here
Figure 112022111425798-pat00034
is the probability ratio,
Figure 112022111425798-pat00035
is the deviation of the residuals from the steady state,
Figure 112022111425798-pat00036
is the deviation of the residuals for the non-steady state,
Figure 112022111425798-pat00037
is the measured residual,
Figure 112022111425798-pat00038
is the mean of the residuals for the steady state,
Figure 112022111425798-pat00039
means the average of the residuals for the non-steady state (the state in which the mean is increased).

순차확률비는 [수식 3], [수식 4] 및 [수식 5]에 의해 결정된다. 즉 경보 발생 여부를 결정하기 위해서는, 순차확률비

Figure 112022111425798-pat00040
를 계산해야 하며, 세 가지 경우에 대해 각각 산출
Figure 112022111425798-pat00041
해야 한다.The sequential probability ratio is determined by [Equation 3], [Equation 4] and [Equation 5]. That is, in order to determine whether an alarm is generated, the sequential probability ratio
Figure 112022111425798-pat00040
must be calculated, and for each of the three cases
Figure 112022111425798-pat00041
Should be.

[수식 3][Formula 3]

Figure 112022111425798-pat00042
Figure 112022111425798-pat00042

[수식 4][Formula 4]

Figure 112022111425798-pat00043
Figure 112022111425798-pat00043

[수식 5][Formula 5]

Figure 112022111425798-pat00044
Figure 112022111425798-pat00044

여기서

Figure 112022111425798-pat00045
은 비정상 상태에 대한 잔차의 평균(평균이 감소된 상태)을 의미하고,
Figure 112022111425798-pat00046
은 비정상 상태에 대한 잔차의 편차(편차가 증가된 상태)를 의미한다.here
Figure 112022111425798-pat00045
Means the average of the residuals for the non-steady state (the state with the mean reduced),
Figure 112022111425798-pat00046
denotes the variance of the residual for the non-steady state (state with increased variance).

도 4는 정상 상태의 잔차 데이터를 그래프로 도시한 예이고, 도 5는 정상 상태의 순차확률비 가능도를 그래프로 도시한 예로서, 판별부(120)는 데이터의 잔차 분포가 정상 잔차 분포

Figure 112022111425798-pat00047
에 가까운 경우 등
Figure 112022111425798-pat00048
가 모두
Figure 112022111425798-pat00049
보다 작기 때문에 정상으로 판별한다.4 is an example of a graph of residual data in a steady state, and FIG. 5 is an example of a graph of the probability of sequential probability ratios of a steady state.
Figure 112022111425798-pat00047
close to
Figure 112022111425798-pat00048
is all
Figure 112022111425798-pat00049
It is judged normal because it is smaller than

도 6은 잔차의 평균이 증가된 데이터를 그래프로 도시한 예이고, 도 7은 잔차의 평균이 증가된 상태의 순차확률비 가능도를 그래프로 도시한 예로서, 판별부(120)는 데이터의 잔차 분포가 증가하여

Figure 112022111425798-pat00050
에 가까운 경우 등
Figure 112022111425798-pat00051
Figure 112022111425798-pat00052
보다 크기 때문에 비정상으로 판별한다.6 is an example of a graph showing data with an increased average of residuals, and FIG. 7 is an example of a graph showing the probability of sequential probability ratios in a state in which the average of residuals is increased. As the distribution of residuals increases,
Figure 112022111425798-pat00050
close to
Figure 112022111425798-pat00051
go
Figure 112022111425798-pat00052
It is determined as abnormal because it is larger than

도 8은 잔차의 평균이 감소된 데이터를 그래프로 도시한 예이고, 도 9는 잔차의 평균이 감소된 상태의 순차확률비 가능도를 그래프로 도시한 예로서, 판별부(120)는 데이터의 잔차 분포가 감소하여

Figure 112022111425798-pat00053
에 가까운 경우 등
Figure 112022111425798-pat00054
Figure 112022111425798-pat00055
보다 크기 때문에 비정상으로 판별한다.8 is an example of a graph showing data with a reduced average of residuals, and FIG. 9 is an example of a graph showing the probability of a sequential probability ratio in a state in which the average of residuals is reduced. The distribution of the residuals decreases,
Figure 112022111425798-pat00053
close to
Figure 112022111425798-pat00054
go
Figure 112022111425798-pat00055
It is determined as abnormal because it is larger than

도 10은 잔차의 편차가 증가된 데이터를 그래프로 도시한 예이고, 도 11은 잔차의 편차가 증가된 상태의 순차확률비 가능도를 그래프로 도시한 예로서, 판별부(120)는 데이터의 잔차 분포의 편차가 증가하여

Figure 112022111425798-pat00056
에 가까운 경우 등
Figure 112022111425798-pat00057
가 크게 증가하며,
Figure 112022111425798-pat00058
또한 증가하여
Figure 112022111425798-pat00059
보다 크기 때문에 비정상으로 판별한다.FIG. 10 is an example of graphing data with increased deviation of residuals, and FIG. 11 is an example of graphing the likelihood of sequential probability ratios in a state in which deviation of residuals is increased. As the variance of the residual distribution increases,
Figure 112022111425798-pat00056
close to
Figure 112022111425798-pat00057
increases significantly,
Figure 112022111425798-pat00058
also increased
Figure 112022111425798-pat00059
It is determined as abnormal because it is larger than

도 12는 임계값과 순착확률비를 비교한 조기 경보를 그래프로 도시한 예로서, 임계값을 기반으로 경보를 발생하는 경우 임계값(=1)보다 큰 구간에서 모두 경보가 발생하지만, 순차확률비의 경우 이러한 불필요한 알람은 발생하지 않고, 잔차가 증가하는 1500 step이후 구간에 적절한 경보가 발생함을 볼 수 있다.12 is an example of a graph showing an early warning by comparing a threshold and a sequential probability ratio. When an alarm is generated based on the threshold, all alarms occur in a section greater than the threshold (= 1), but the sequential probability In the case of rain, such an unnecessary alarm does not occur, and it can be seen that an appropriate alarm occurs in the interval after 1500 steps when the residual increases.

특허문헌 1과 특허문헌 2의 경우 기대값(또는 예측값) 계산 모델 초기 설정 후 태그 인덱스를 기반으로 경보 발생 기준이 정해져, 경보 발생 기준이 정해진 뒤 특정 잔차를 초과하면 경보가 발생된다.In the case of Patent Literature 1 and Patent Literature 2, after the initial setting of the expected value (or predicted value) calculation model, an alarm generation standard is determined based on the tag index, and an alarm is generated when a specific residual is exceeded after the alarm generation standard is set.

그러나 본 발명은 일정구간 동안 발생한 잔차의 분포가 정상잔차의 확률과 비정상잔차 확률의 비율(확률비)을 기반으로 계산하기 때문에, 해당 기간동안 잔차가 높은 현상이 발생하여도 정상 잔차를 유지하는 비율이 많아 정상잔차의 확률로 계산되면 경보를 발생시키지 않아 불필요한 경보를 억제할 수 있다.However, since the present invention calculates the distribution of residuals generated during a certain period based on the ratio (probability ratio) of the probability of stationary residuals and the probability of abnormal residuals, the ratio of maintaining stationary residuals even when a phenomenon with high residuals occurs during the period If it is calculated as the probability of a stationary residual, an alarm is not generated and unnecessary alarms can be suppressed.

도 13은 본 발명의 실시예에 따른 순차확률비를 이용한 경보 방법을 도시한 흐름도로서, 경보 장치(10)는 운전대상의 운전데이터를 이용하여 실측값과 기대값 간의 잔차를 연산하는 단계, 상기 잔차를 이용하여 순차확률비를 연산하는 단계, 상기 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하는 단계 및 상기 상태가 비정상이면 경보를 발생하는 단계로 구동된다.13 is a flowchart illustrating an alarming method using a sequential probability ratio according to an embodiment of the present invention, in which the alarming device 10 calculates a residual between an actual value and an expected value using driving data of a driving target; The step of calculating the sequential probability ratio using the residual, the step of determining the state of the operating target by applying the sequential probability ratio to the alarm condition according to the probability distribution of the residual, and the step of generating an alarm if the state is abnormal .

10: 경보 장치 110: 수집부
120: 판별부 130: 경보부
10: alarm device 110: collection unit
120: determination unit 130: warning unit

Claims (3)

경보 장치에서 운전대상의 운전데이터를 이용하여 실측값과 기대값 간의 잔차를 연산하는 단계;
상기 잔차를 이용하여 순차확률비를 연산하는 단계;
상기 순차확률비를 잔차의 확률분포에 따른 경보조건에 적용하여 운전대상의 상태를 판별하는 단계 및
상기 상태가 비정상이면 경보를 발생하는 단계를 포함하고,
상기 운전대상의 비정상 상태에 대한 잔차의 확률분포는 잔차의 평균이 증가된 상태의 분포, 잔차의 평균이 감소된 상태의 분포 및 잔차의 편차가 증가된 상태의 분포이며,
상기 경보조건은 비정상, 보류 및 정상 상태로 분류되고, [수식 1] 및 [수식 2]에 의해 결정되며,
상기 순차확률비는 [수식 3], [수식 4] 및 [수식 5]에 의해 결정되는 것을 특징으로 하는 순차확률비를 이용한 경보 방법.
[수식 1]
Figure 112022135717529-pat00060

[수식 2]
Figure 112022135717529-pat00061

[수식 3]
Figure 112022135717529-pat00062

[수식 4]
Figure 112022135717529-pat00063

[수식 5]
Figure 112022135717529-pat00064
calculating a residual between an actual value and an expected value using driving data of a driving target in an alarm device;
calculating a sequential probability ratio using the residual;
Determining the state of an operating target by applying the sequential probability ratio to an alarm condition according to a probability distribution of residuals; and
generating an alarm if the condition is abnormal;
The probability distribution of the residuals for the abnormal state of the driving target is a distribution in which the mean of the residuals is increased, a distribution in which the mean of the residuals is decreased, and a distribution in which the deviation of the residuals is increased,
The alarm conditions are classified into abnormal, pending, and normal states, and are determined by [Equation 1] and [Equation 2],
The sequential probability ratio is determined by [Equation 3], [Equation 4] and [Equation 5].
[Formula 1]
Figure 112022135717529-pat00060

[Equation 2]
Figure 112022135717529-pat00061

[Formula 3]
Figure 112022135717529-pat00062

[Formula 4]
Figure 112022135717529-pat00063

[Formula 5]
Figure 112022135717529-pat00064
삭제delete 제1항에 있어서,
상기 경보 장치는
Figure 112022135717529-pat00065
,
Figure 112022135717529-pat00066
,
Figure 112022135717529-pat00067
Figure 112022135717529-pat00068
보다 작으면 정상 상태로 판별하고,
Figure 112022135717529-pat00069
,
Figure 112022135717529-pat00070
Figure 112022135717529-pat00071
보다 크면 잔차의 평균이 증가된 비정상 상태로 판별하며,
Figure 112022135717529-pat00072
,
Figure 112022135717529-pat00073
Figure 112022135717529-pat00074
보다 크면 잔차의 평균이 감소된 비정상 상태로 판별하고,
Figure 112022135717529-pat00075
,
Figure 112022135717529-pat00076
Figure 112022135717529-pat00077
보다 크면 잔차의 편차가 증가된 비정상 상태로 판별하는 것을 특징으로 하는 순차확률비를 이용한 경보 방법.
According to claim 1,
the warning device
Figure 112022135717529-pat00065
,
Figure 112022135717529-pat00066
,
Figure 112022135717529-pat00067
go
Figure 112022135717529-pat00068
If it is less than, it is determined as a normal state,
Figure 112022135717529-pat00069
,
Figure 112022135717529-pat00070
go
Figure 112022135717529-pat00071
If it is greater than this, it is determined as an abnormal state with an increased average of the residuals,
Figure 112022135717529-pat00072
,
Figure 112022135717529-pat00073
go
Figure 112022135717529-pat00074
If it is greater than this, it is determined as an abnormal state with a reduced average of the residuals,
Figure 112022135717529-pat00075
,
Figure 112022135717529-pat00076
go
Figure 112022135717529-pat00077
An alarm method using a sequential probability ratio, characterized in that it is determined as an abnormal state in which the deviation of the residual is increased when it is greater than .
KR1020220136639A 2022-10-21 2022-10-21 Alarm method using sequential probability ratio KR102492514B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020220136639A KR102492514B1 (en) 2022-10-21 2022-10-21 Alarm method using sequential probability ratio

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020220136639A KR102492514B1 (en) 2022-10-21 2022-10-21 Alarm method using sequential probability ratio

Publications (1)

Publication Number Publication Date
KR102492514B1 true KR102492514B1 (en) 2023-01-30

Family

ID=85106043

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020220136639A KR102492514B1 (en) 2022-10-21 2022-10-21 Alarm method using sequential probability ratio

Country Status (1)

Country Link
KR (1) KR102492514B1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008186472A (en) * 2001-04-10 2008-08-14 Smartsignal Corp Diagnostic systems and methods for predictive condition monitoring
JP2010049359A (en) * 2008-08-19 2010-03-04 Toshiba Corp Plant monitoring device and plant monitoring method
KR20130042523A (en) 2013-03-21 2013-04-26 비앤에프테크놀로지 주식회사 Method measuring of healthy indicia of plant reflected status of sub-component and storage media thereof
KR20170125237A (en) 2016-05-04 2017-11-14 두산중공업 주식회사 Plant system, and fault detecting method thereof
US20200272140A1 (en) * 2019-02-21 2020-08-27 Oracle International Corporation Adaptive sequential probability ratio test to facilitate a robust remaining useful life estimation for critical assets

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008186472A (en) * 2001-04-10 2008-08-14 Smartsignal Corp Diagnostic systems and methods for predictive condition monitoring
JP2010049359A (en) * 2008-08-19 2010-03-04 Toshiba Corp Plant monitoring device and plant monitoring method
KR20130042523A (en) 2013-03-21 2013-04-26 비앤에프테크놀로지 주식회사 Method measuring of healthy indicia of plant reflected status of sub-component and storage media thereof
KR20170125237A (en) 2016-05-04 2017-11-14 두산중공업 주식회사 Plant system, and fault detecting method thereof
US20200272140A1 (en) * 2019-02-21 2020-08-27 Oracle International Corporation Adaptive sequential probability ratio test to facilitate a robust remaining useful life estimation for critical assets

Similar Documents

Publication Publication Date Title
CN109325692B (en) Real-time data analysis method and device for water pipe network
CN112348237B (en) Abnormal trend detection method for dynamic drilling data
BR112021013786A2 (en) METHOD AND SYSTEM TO DETECT ANOMALIES OR PREMATURE INDICATIONS OF EQUIPMENT FAILURE
NO170957B (en) FIRE ALARMSYSTEM
CN110057406B (en) Multi-scale self-adaptive mechanical equipment trend early warning method
KR20200041515A (en) Early warning system and method of power plants using multiple prediction model
JP5824333B2 (en) Stick-slip detection device and detection method
CN115311829A (en) Accurate alarm method and system based on mass data
US5670938A (en) Fire alarm device
KR102492514B1 (en) Alarm method using sequential probability ratio
JP2006163517A (en) Abnormality detector
CN115240353A (en) Fire-fighting dangerous case processing method based on Internet of things sensing
CN106652393B (en) False alarm determination method and device
CN111207306B (en) Heating power pipe network leakage monitoring method based on digital pressure signal processing
KR20180075889A (en) alarm occurring method for using big data of nuclear power plant
KR20180057829A (en) Method for controlling sensor based on statistical process control
JP4402613B2 (en) Plant abnormality monitoring system and plant abnormality monitoring method
KR102547605B1 (en) Method for calculating dynamic threshold value
US11976419B2 (en) Health assessment of a mechanical system
JPH04152220A (en) Method and device for failure sensing
CA3114157C (en) Process management device, process management method, and process management program storage medium
JP4842547B2 (en) Gas detector
KR102450030B1 (en) System and method for sensing failure of power generation facilities
TWI783360B (en) Abnormal detection device and abnormal detection method
JPH0610835B2 (en) Fire detector

Legal Events

Date Code Title Description
E902 Notification of reason for refusal
E701 Decision to grant or registration of patent right
GRNT Written decision to grant