AU2019333948B2 - Interference alarm identification method based on alarm duration feature - Google Patents
Interference alarm identification method based on alarm duration feature Download PDFInfo
- Publication number
- AU2019333948B2 AU2019333948B2 AU2019333948A AU2019333948A AU2019333948B2 AU 2019333948 B2 AU2019333948 B2 AU 2019333948B2 AU 2019333948 A AU2019333948 A AU 2019333948A AU 2019333948 A AU2019333948 A AU 2019333948A AU 2019333948 B2 AU2019333948 B2 AU 2019333948B2
- Authority
- AU
- Australia
- Prior art keywords
- alarm
- nuisance
- duration
- alarm duration
- distribution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012544 monitoring process Methods 0.000 claims abstract description 19
- 238000009826 distribution Methods 0.000 claims description 71
- 238000005070 sampling Methods 0.000 claims description 11
- 230000000295 complement effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims 2
- 238000004519 manufacturing process Methods 0.000 description 15
- 230000005856 abnormality Effects 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 101100379081 Emericella variicolor andC gene Proteins 0.000 description 1
- 241000845077 Iare Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B29/00—Checking or monitoring of signalling or alarm systems; Prevention or correction of operating errors, e.g. preventing unauthorised operation
- G08B29/12—Checking intermittently signalling or alarm systems
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Alarm Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
An interference alarm identification method based on an alarm duration feature, comprising: step S1, comparing an obtained monitoring process signal with an alarm threshold thereof to generate an alarm signal; step S2, defining an alarm duration set; step S3, if C
Description
Technical Field
The present invention belongs to the field of process industry production monitoring, and in particular, to a nuisance alarm identification method based on the feature of alarm duration distribution.
Related Art
In modern industrial production processes, alarm systems are widely used in production process monitoring to ensure the production process safety and economic running. The alarm system monitors the production process by comparing current values of monitoring process variables to their alarm thresholds, respectively. Alarm thresholds are boundaries of process variables' variation ranges in normal situation, which are usually determined according to requirements of the production process. When the value of a process variable exceeds its alarm threshold, an alarm message is generated by the alarm system and sent to notify operators about the abnormality or failure, which is indicated in the alarm message in the production process. If an alarm message has been received, the operators determine the cause of the failure or abnormality by analyzing alarm signals, as well as the related process variables, and then take effective actions to troubleshoot or address abnormalities, so that the production process returns to its normal state.
However, there is a large number of nuisance alarms (or false alarms) in current industrial alarm systems, that is, there is no fault or abnormality in the production process but the alarm system generates alarms. Numerous nuisance alarms will cause true alarms to be submerged in a large amount of nuisance alarm information, which is very harmful to the effectiveness of the alarm system, and also make the operator lose their trust in alarm systems. This situation makes the safety and economy of the production process at great risk. Therefore, it is very necessary to eliminate nuisance alarms in alarm systems, in order to enhance the performance of the alarm systems and make the true alarms found easily, thereby monitoring the production processes effectively.
Despite the serious harm of nuisance alarms, the existing methods about nuisance alarm identification are mainly about qualitative descriptions. For example, 9 or more alarms within 5 minutes are used as a rule to determine the nuisance alarm. Lacking of a theoretical nuisance alarm identification method that has prevented the nuisance alarm elimination, alarm system performance optimization, production process safety, and economic monitoring of the production process, resulting in production safety risks.
By considering the shortcomings of the prior art, the present invention provides a nuisance alarm identification method based on alarm duration distribution, which can effectively identify nuisance alarms. The method is effective to eliminate nuisance alarms and optimize alarm system performance.
The technical solutions used in the present invention to resolve the technical problems are as follows.
The nuisance alarm identification method based on an alarm duration distribution feature includes the following steps:
step Si: comparing a historical data sequence x(n), which is associated with a process
signal, to its alarm threshold x, to generate an alarm data sequence x,(n), n=1,2,- ;
step S2: defining an alarm duration set T k =IT(1), T(2), ---, T(Cj), k = 0,1,2,- -,and
C being the number of alarm durations in the alarm duration set T;
when k = 0, the alarm duration set is initialized as T° ={7T(1), T (2),- -,7(C°) }being
all the alarm durations in the alarm data sequence x,(n), n =1,2,-- -, where C° indicates
the number of elements in T°, that is, C° alarms are occurred in the alarm duration set
T°; step S3: defining RC as a required number of elements in the alarm duration set during performing algorithm of the method; if C" does not satisfy an inequality Ck>RCA, concluding that there is no nuisance alarm induced by the monitoring process signal x(n), n=1,2,---, and then ending the entire algorithm; if C" satisfies the inequality C) >RCA , obtaining sample probability distribution of
Tk ={T (1), T(2),., T (Ck) and geometric distribution estimated with it, k = 0,1,2, -- ;and
step S4: calculating a fitting degree R2 between the sample probability distribution and the estimated geometric distribution, obtained with the alarm duration set
Tk*={T(1),Tif(2),.-,T(C )j;
if the fitting degree R2 satisfies an inequality R2 >0.95, concluding that alarms
corresponding to the elements in alarm duration set Tk are all nuisance alarms; and if the
removed alarm durations max(T-)=m, concluding that m is an upper limit of the
nuisance alarm duration, and then ending the entire algorithm;
if the fitting degree R2 satisfies the inequality R2 < 0.95, removing the largest alarm
duration elements from Tk*, and performing assignment of k=k+1 to form a subset of
the alarm duration set Tk - : -= {T(0T(i)< max(T-)}, and performing step S3.
Further, step Si is specifically implemented as follows:
when the alarm threshold x, denotes a high alarm threshold, the monitoring process
signal x(n), n =1,2,-- , is compared to x, to obtain an alarm data sequence as
(n .1, x(n) > x,, 0O, x(n) < x,,
Without loss of generality, the high alarm threshold is discussed herein, and the alarm data sequence x,(n) is composed of binary data '0' and '1', n=1,2,---,.
Further, in step S2, an alarm duration T is defined as the time duration from the time instant of an alarm occurring to the time instant of the alarm ending; for a given alarm, assuming that it occurs at a sampling instant n, and ends at a sampling instant n2
, n 2 >n, then the corresponding alarm data sequence is x,(nj)=1 and x,(n 2 +1)=0
respectively. Therefore, the alarm duration T is calculated as 7= n2-n 1+1. All the
alarm durations can formulate the alarm duration set T,={j;(1),j(2),---,j(Ck)},
k = 0,1,2,- --. It is the initialized set when k = 0 in the algorithm.
Further, step S3 is specifically implemented as follows:
nuisance alarms are almost all caused by noise, and the length of alarm duration caused by noise is a random variable, therefore, the nuisance alarm duration has a corresponding probability distribution feature. In the present invention, identification of nuisance alarms is achieved by identifying the nuisance alarm caused by noise, based on their alarm duration probability distribution features.
Due to the randomness of alarm durations corresponding to nuisance alarms, by taking the central limit theorem into consideration, it can be concluded that the alarm duration of nuisance alarm obeys geometric distribution on condition that noise obeys independent and identically normal distribution, that is: T (n)= q"-1 (1- q) ,where n indicates that the alarm duration is n sampling period, and q, indicates a false alarm probability caused by noise;
the false alarm probability q, being estimated according to the alarm duration set c4 Tk ={;(1),7;(2),- -,(C )}as I =1-TC T(n); and n=1
the estimated geometric distribution associated with the alarm duration set Tk ={;(1),7;(2),---,;(C ) being obtained with 41 as ZA(n)= 4n"-1(1- 41); and
the sample probability distribution being obtained from the alarm duration set
Tk ={T(1), T(2),---, T(C ),whichbeing: p(n)=c,/C), where c, indicates numbers of
T(l)=n l=1,2,---,Ck in the alarm duration set Tn =etT (1),T(2)T(2, ---, T (C ), that is, the
number of alarms with an alarm duration TI in the alarm duration set
Tk*=IT(1), T(2),---,T(C ) is n.
Further, step S4 is specifically implemented as follows:
if all the alarm durations in the alarm duration set T, ={T(1), T(2),---,T(C) Iare
induced by independent and identically distributed normal distribution noise, the estimated geometric distribution and sample probability distribution should tend to be equal when the
value of Ck tends to infinity, that is, the sample probability distribution is equivalent to
the estimated geometric distribution. Therefore, if the values of the sample probability distribution and estimated geometric distribution are approximately equal, all the alarms
related to the alarm duration set T"=T(1),T(2),---,T(C ) are considered as nuisance
alarms.
To measure the goodness of fit between the sample probability distribution and the
estimated geometric distribution, the fitting degree R2 is adopted,
R 2 =1-(p(n)-TAD(n)Y/ rTD(n)-TAD(n) , where TAD(n) indicates the mean of
TAD(n), T m (n)= I , AD(n)M, M = ma{Tk
It should be noted herein that a real alarm is a reflection of abnormal conditions in the production process, which can be eliminated only after an operator (operators) takes effective actions. Therefore, a relatively long time duration is generally related to a true alarm. However, the nuisance alarms are almost caused by noise and can be eliminated without operators' actions, and thus they are associated with short alarm durations. In most practice cases, the true alarms with longer alarm durations and the nuisance alarms with shorter alarm durations are mixed together. Therefore, in the nuisance alarm identification process, the elements associated with larger alarm durations in the alarm duration set should be excluded so that all the rest elements in the alarm duration set are noise caused nuisance alarms. Therefore, when an alarm sequence is obtained, the alarm duration set T° is obtained firstly, the sample probability distribution and estimated geometric distribution of elements in the alarm duration set T° are calculated, as well as their fitting degree R2
. When k =0 and the fitting degree R 2 0.95, it is concluded that the sample probability of the alarm duration set T° obeys the estimated geometric distribution, and it can be concluded that the alarms corresponding to the elements in the alarm duration set T°={T(1),T(2),---,JT(C°) are all nuisance alarms.
Further, when the fitting degree R 2 does not satisfy the inequality R 2 >0.95, the elements in the alarm duration set T are removed in descending order to form an alarm
duration subsetk Tk = T(i)|)<aX( -1)}, k is positive integers and k 1.
For each specific value of k, the sample probability distribution and estimated geometric distribution of the elements in the alarm duration set Tk are calculated, and the fitting
degree R2 between the sample probability distribution and estimated geometric distribution is calculated correspondingly. If the fitting degree R 2 dose not satisfy the inequality R 2 >0.95 yet, let k=k+1, and repeat the steps of removing the maximum alarm durations and calculating the fitting degree R 2 . Until the fitting degree R2 satisfies R 2 >0.95, the repeating process is stopped. The end value of k indicates the repeating times.
Further, if the fitting degree R 2 does not satisfy R 2 >0.95 after the most elements of T are removed, that is, if the elements number C 1 of the alarm duration set T,
satisfies C 1 >RCA and the elements number Cjkof the alarm duration set T satisfies
Ck < RCA of elements in the alarm duration set T hold, it can be concluded that there is
no noise caused nuisance alarm in the process signal x(n).
Further, if the inequality R 2 >0.95 is still not satisfied when the data length is no less than 30 days. It is considered that there is no noise caused nuisance alarm in the process signal x(n), n=1,2,---.
Further, the alarm interval To is defined as time span between two adjacent alarms.
Suppose that there are the two adjacent alarms, the previous alarm ends at the time instant
n, that is, x,(n+1) =0; for the next alarm, it is triggered at a sampling instant n2
(n 2 n,), x,(n 2 )=1, then the alarm interval To is represented as: To = n 2 -n 1 +1. In the
present invention, the alarm duration TI serves as the basis for nuisance alarm
identification. Due to the fact that the alarm interval To and the alarm duration TI are
complementary, the method is also suitable for nuisance alarm identification with alarm intervals.
Further, the method provided in the present invention is suitable for not only analog monitoring variables but also digital monitoring variables. This merit is based on the fact that the nuisance alarms are identified with their alarm durations, which are extracted from the alarm signal.
The present invention has the following beneficial effects. (1) In the present invention, the collected data of monitored process signal is compared to its alarm threshold to obtain the corresponding alarm signal, the alarm duration set is formulated based on the alarm signal. The nuisance alarms are identified by calculating the fitting degree of the sample probability distribution and the estimated geometric distribution of the alarm duration set or its subsets. (2) The present invention provides basic guidance for the nuisance alarm identification in the alarm system through the alarm duration distribution features. (3) The present invention provides a basis for performance evaluation of alarm systems.
FIG. 1 is the flowchart of nuisance alarm identification according to an embodiment of the present invention.
The following further describes the present invention with reference to accompanying drawings.
Embodiment 1:
As shown in FIG. 1, a nuisance alarm identification method based on the feature of alarm duration distribution in the present invention, includes the following steps:
step SI: collecting one month historical data of a monitoring process signal x(n) with
sampling period 1 second, n=1,2,---, and its alarm threshold x,,; generating the
corresponding alarm data sequence x,(n), n =1,2,-- -, by comparing the historical data to its alarm threshold;
step S2: defining an alarm duration set Tk"={T(1),T(2),---,Tj(Ck)}, k=1,2,-,
where C" is a number of alarms corresponding to the alarm duration set T
when k = 0, the alarm duration set T,={(1), T(2), ---, (C°)} being the set of all
the alarm durations in the alarm data sequence x,(n), n=1,2,---, C° indicating the
number of elements in T°, that is, indicating a total of C° alarms occurring in the alarm
duration set T°;
step S3: defining RCA as a required number of elements in the alarm duration set T
and R CA= 500 taking in the present invention as the default;
if C" does not satisfy the inequality C) >RCA , concluding that there is no nuisance alarm in the monitoring process signal data sequence x(n), n =1,2,---, and ending the entire algorithm;
if Cj satisfies the inequality C) >RCA , obtaining the sample probability distribution and estimated geometric distribution of the alarm duration set with Tk,={T(1),T(2),---,T(Ck)}, k=0,1,2,---;and
step S4: calculating a fitting degree R2 between the sample probability distribution and the estimated geometric distribution of the alarm duration set Tk ={T(1),T(2),---,T(C)) if the fitting degree R2 >0.95, concluding that alarms corresponding to elements in the alarm duration set Tk are all nuisance alarms; if the removed alarm duration max(T-)= m, selecting m as the upper limit of the nuisance alarm duration, and ending the entire algorithm; if the fitting degree R2 < 0.95, removing the largest elements from the alarm duration set Tk and performing assignment of k=k+1 to form a subset of the alarm duration set
Tk: = (i) (i) < max( -') , and performing step S3.
In step Si of this embodiment, the alarm threshold x, is taken as a high alarm
threshold, the historical data of the monitoring process signal x(n) , n=1,2,---, is
compared to the alarm threshold x, to obtain an alarm data sequence as
rx,(n)= (n .1, X x(n) >x,, "" (1) 1 0O, x(n) < x,,
Hence, the high alarm threshold is discussed and the alarm data sequence x,(n),
n =1,2,-- -, only consists of '0' and '1'.
in step S2, for the alarm data sequence x,(n), n =1,2,-- -, the time span from the time
instant of an alarm occurring to the time instant of the alarm ending is referred to as an alarm duration, and denoted by TI. For a given alarm process, the alarm occurs at a
sampling instant nj, that is, x,(n) =1, and it ends at a sampling instant n2 (n 2 >n),
x,(n 2 +1)= 0, and thus the alarm duration can be calculated as:
T = n2 -n +1 (2)
All alarm durations obtained from an alarm data sequence formulate the alarm duration
set T =IT(1), T(2),---, T(C°) .
In step S3, based on the nuisance alarm caused by noise, the length of the alarm duration is a random variable. Hence, the nuisance alarm duration has its corresponding probability distribution feature. In the present invention, identification of nuisance alarms is achieved by identifying the nuisance alarm caused by noise, based on their alarm duration probability distribution features.
A random duration nuisance alarm refers to the nuisance alarm with random alarm duration. The random duration nuisance alarm is almost all caused by noise. According to the central limit theorem, if noise obeys independent identical distribution, then its caused nuisance alarm durations obeys geometric distribution, that is:
T,(n) = q "-1(1 qi) (3)
Where n indicates the length of the alarm duration, and q, indicates the false alarm probability caused by noise. The false alarm probability q, can be estimated according to
the alarm duration set Tk=IT(1), T(2),- -, T(C ), as
C* 4 c1 (4) T((n)
The estimated geometric distribution of the alarm duration set
T," =IT (1), T (2), - -, T(Ck)j can be obtained from the false alarm probability 41 as
TA, (n) = 4n"-I(1 - qj) (5)
T1k={T(1),if(2),--,T(C ), that is, c,~ is the number of elements in the alarm duration
set Tk"={T(1),JT(2),.-.-.,J(C )}whose alarm duration is n.
If all the elements in the alarm duration set Tk=(T(1),T(2),---,T(C ) are induced
by independent and identically distributed noise, the probability calculated in formula (5)
and formula (6) should tend to be equal when C" tends to infinity. Hence, if the
probabilities obtained from formula (5) and formula (6) are approximately equal, it is
concluded that all alarms related to the the alarm duration set Tk=T (1), T(2),---, T(C
) are nuisance alarms.
In order to measure the goodness offit between the sample probability distribution and the estimated geometric distribution obtained with formula (5) and formula (6), in the
present invention, step S4 is performed, that is, the fitting degree R2 is used as a
approximation measurement. R2 is calculated as:
R2 n- i,, (n) R_=1-(p(n)-tn (7) J!(n) - !A,(n)y
Here, TAD (n) is the mean of T (n) and T,(n)= f, T_ (n)/M , M = max{T,.
It should be noted here that a real alarm is a reflection of abnormal conditions in the production process, which can be eliminated only after operators take effective actions. Therefore, a relatively long time duration is generally required for the true alarm. However, nuisance alarms are almost caused by noise and can be eliminated without operators' actions, and thus the alarm duration of nuisance alarm is short. In most practice cases, alarm durations associated with true alarms and nuisance alarms are mixed together. Therefore, during identification of nuisance alarms, the larger elements in the alarm
duration set T" should be excluded. Because the upper duration time limit of the nuisance
alarm in T, is unknown, a step-by-step elimination and identification method is adopted
in the present invention. Therefore, when an alarm sequence is obtained, the alarm duration
set T" isfirst obtained; next, the sample probability distribution and estimated geometric
distribution of elements in the alarm duration set T, are calculated, as well as the fitting
degree R2
When the fitting degree R 2 0.95, it can be said that the sample probability of the
alarm duration set Tk obeys the estimated geometric distribution, and it is concluded that
the alarms corresponding to the alarm durations in set 7 = (1), T(2),--.,7(C) }are all
nuisance alarms.
When R 2 -0.95 is not satisfied, the elements in the alarm duration set T shouldbe
excluded in descending order, and the elements corresponding to the maximum value in the
alarm duration set T, is excluded every time. After the maximum value element is
excluded, it is specified that k=k+1, so that a subset of the alarm duration set T"kis
formulated and denoted as: 7 k 7=,(i |I(i)< Ma . Based on the updated T, the sample probability distribution and estimated geometric distribution are calculated
again, as well as the fitting degree R2. This process is repeated until the fitting degree
R2 0.95 , the end value of k indicates the repeating times. Then the alarms
corresponding to the remaining alarm durations in the alarm duration set
Tk = |(i)(I)<mI (7-1)} are all nuisance alarms, and suppose the last excluded
alarm duration value is max(Tl-)=m, and thus m is regarded as the upper limit of
nuisance alarm durations.
If the inequality R 2 0.95 is still not satisfied after most elements in the alarm
duration set T are removed, that is, if C->RC and CkcCRC, hold, it can be
concluded that there is no noise caused nuisance alarm at the measuring point. The required numbers of the alarms RCA = 500 is taken as a default value in the present invention.
In addition, an alarm interval is defined as time span between two adjacent alarms,
which is represented as To . Suppose that there are the two adjacent alarms, the previous
alarm ends at the time instant n, that is, x,(n,+1)= 0; for the next alarm, it is triggered at
a sampling instant n2 (n2 n,), x,(n 2 )=1, then the alarm interval To is calculated as:
To =n2 -n,+1. In the present invention, the alarm duration TI serves as the basis for nuisance alarm identification. Due to the fact that the alarm interval To and the alarm duration T are complementary, the method is also suitable for nuisance alarm identification with alarm intervals.
The method provided in the present invention is suitable for not only analog monitoring variables but also digital monitoring variables. This merit is based on the fact that the nuisance alarms are identified with their alarm durations, which are extracted from the alarm signal.
The following is the method of the present invention process in a specific scenario.
For a given process variable x(n) which is monitored by the alarm system, n =1,2,- -, the nuisance alarm identification method in the present invention is applied. The first step is to obtain 1 month historical data samples of x(n) and its (upper) alarm
threshold x, to generate alarm data x,(n) according to formula (1), n =1,2,---. The
second step is to obtain alarm durations in the alarm data sequence x,(n) to form the
alarm duration set {T° (1), T(2),- -, (C )1, C' is the number of alarms in x,(n). In
the third step, if the inequality C > RCA is not satisfied, it is considered that there is no alarm caused by noise at the measuring point, and the algorithm is ended. If the inequality C >RCA is satisfied, nuisance alarm is identified on the alarm duration set
T* =T (1), T(2),.- --, T (C) according to formulas (4) to (7), k = 0,1,2- --. If the fitting degree R2 of the sample probability distribution and the estimated geometric distribution associated with the alarm duration set satisfies R 2 >0.95, it is concluded that the
alarms corresponding to the alarm durations in the alarm duration set T, are all nuisance
alarms. Otherwise, the largest elements in the alarm duration set T" are removed to
formulate a subset of T -= |(i)o(i)< ax(r'-1)} and update k as k=k+1.
The fitting degree R 2 of the sample probability distribution and the estimated geometric distribution associated with the alarm duration set T k =1,2,-- -, is calculated according to formulas (4) to (7) until R2 >0.95, and then the alarms corresponding to the alarm durations in the alarm duration set Tk are regarded as nuisance alarms.
If the fitting degree R2 >0.95 is still not satisfied after most elements in the alarm
duration set T are removed, that is, if C' > RCA andC<RC hold,itisconcluded
that there is no nuisance alarm in x,(n).
The foregoing descriptions are embodiments of the present invention, and the protection scope of the present invention is not limited thereto. All equivalent structure or process changes made according to the content of this specification and accompanying drawings in the present invention or by directly or indirectly applying the present invention in other related technical fields shall fall within the protection scope of the present invention.
Claims (9)
1. A nuisance alarm identification method based on an alarm duration distribution feature, comprising the following steps:
step Si: comparing a historical data sequence x(n), which is associated with a process signal, to its alarm threshold x, to generate an alarm data sequence
x,(n), n=1,2,---.
step S2: formulating an alarm duration set T={7;(1),T(2),---,7;(Ck))
k 0,1,2,-- -, and Ck being a number of alarms in the alarm duration set j;
when k = 0, the alarm duration set is initialized as T° = T (1),7 (2),- T (C°)
being all the alarm durations in the alarm data sequence x,(n), n =1,2,--, wherein C°
indicates the number of elements in T°;
step S3: defining RCA as a required number of elements in the alarm duration set during performing algorithm of the method;
if Ck does not satisfy Ck>RCA, concluding that there is no nuisance alarm induced by the monitoring process signal x(n) n=1,2,---, and then ending the entire
algorithm; and
if Ck satisfies the inequality C) >RCA , obtaining sample probability distribution of
Tk ={7(1), T(2),. ,7(Cjk) and geometric distribution estimated with it, k = 0,1,2, -- ;and
step S4: calculating a fitting degree R2 between the sample probability distribution and the estimated geometric distribution, obtained with the alarm duration set
T,k= IT(1),7T.(2),.-, T(C k)j.
if the fitting degree R 2 satisfies an inequality R 2 >0.95, concluding that alarms corresponding to the elements in alarm duration set T, are all nuisance alarms; and if the removed alarm durations max(Tl-)=m, concluding that m is an upper limit of the nuisance alarm duration, and then ending the entire algorithm; if the fitting degree R2 satisfies the inequality R2 < 0.95, removing the largest alarm duration elements from T, and performing assignment of k=k+1 to form a subset of the alarm duration set Tk*-l: = f (i(i)< max(-)}, and performing step S3 again to continue calculation through the algorithm.
2. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 1, wherein step S is specifically implemented as follows:
when the alarm threshold x, denotes a high alarm threshold, the historical data
sequence x(n), n =1,2,... of the monitoring process signal is compared to the alarm
threshold x, to obtain an alarm data sequence as x, (n)= r .x(n)x o, x(n)< xP
3. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 1, wherein in step S2, an alarm duration TI is defined as the time duration from the time instant of an alarm occurring to the time instant of the alarm ending; and
for a given alarm, assuming that it occurs at a sampling time instant n, and ends at a sampling time instant n 2 , n 2 n , then the corresponding alarm data sequence is
x,(ni)=1 and x,(n 2 +1)= respectively; therefore, the alarm duration TI iscalculated
as T =n 2 -n+1.
4. The nuisance alarm identification method based on the feature of alarm duration distribution according to claim 1, wherein step S3 is specifically implemented as follows:
nuisance alarms are almost all caused by noise, the length of the alarm duration caused by noise is a random variable, and by taking the central limit theorem into consideration, it can be concluded that the alarm duration of nuisance alarm obeys geometric distribution on condition that noise obeys independent and identically distributed normal distribution, that is: TA(n)=q"-'(1-q,), wherein n indicates that the alarm duration is n sampling period, and q, indicates a false alarm probability caused by noise; the false alarm probability rate q, being estimated according to the alarm duration c set T,=7(1), T(2),-.,7(Ck) as 4.=1-C JT(n); n=1 the estimated geometric distribution associated with the alarm duration set T ={;(1),7;(2),. ,;(C ) being obtained from 41, which being: Z,(n)="(1-ii); and the sample probability distribution being obtained with the alarm duration set
T,={7 (1),T;(2),.,7(C ) as p(n)= , wherein c, indicates numbers of T(l)=n CA
and l=1,2,---,Ck in the alarm duration set Tk={;(1),7;(2),---,;(C ) , that is, the
number of alarms with an alarm duration T in the alarm duration set
Tk={;(1),7;(2),. -., ,(C ) is n .
5. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 4, wherein step S4 is specifically implemented as follows:
a specific calculation formula of the fitting degree R2 is:
R 2 =_ _ rt , D(n) , wherein T(n) indicates an mean of
T(n), T (n)= T,(n)/M, and M=maxT,.
6. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 5, wherein if the fitting degree does not satisfy the inequality R 2 >0.95 after most elements in the alarm duration set T, are removed, that is, if a number of elements included in the alarm duration set T*-1 and the number of elements
CA in the alarm duration set T' satisfied C <RCA , it can be concluded that there is no noise caused nuisance alarm in the process variable x(n).
7. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 6, wherein if the fitting degree does not satisfy the inequality R 2 >0.95 when the historical data length is no less than 30 days, it can be concluded that there is no noise caused nuisance alarm in the process signal x(n), n =1,2,- -.
8. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 1, wherein an alarm interval To is defined as time span between
two adjacent alarms, and suppose that the alarm interval To and the alarm duration T are complementary, the method is also suitable for nuisance alarm identification with alarm intervals.
9. The nuisance alarm identification method based on an alarm duration distribution feature according to claim 1, wherein the method is suitable for not only analog monitoring variable but also digital monitoring variables.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811019040.5 | 2018-09-03 | ||
| CN201811019040.5A CN109087490B (en) | 2018-09-03 | 2018-09-03 | A kind of interference alarm recognition methods based on alarm duration features |
| PCT/CN2019/078713 WO2020048114A1 (en) | 2018-09-03 | 2019-03-19 | Interference alarm identification method based on alarm duration feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| AU2019333948A1 AU2019333948A1 (en) | 2020-10-22 |
| AU2019333948B2 true AU2019333948B2 (en) | 2021-02-25 |
Family
ID=64840532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2019333948A Active AU2019333948B2 (en) | 2018-09-03 | 2019-03-19 | Interference alarm identification method based on alarm duration feature |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN109087490B (en) |
| AU (1) | AU2019333948B2 (en) |
| WO (1) | WO2020048114A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109087490B (en) * | 2018-09-03 | 2019-09-03 | 山东科技大学 | A kind of interference alarm recognition methods based on alarm duration features |
| CN111915858B (en) * | 2020-08-04 | 2022-03-29 | 山东科技大学 | Alarm method and system fusing analog quantity and digital quantity related information |
| CN114005256B (en) * | 2021-10-25 | 2023-06-20 | 山东科技大学 | Alarm delay design method and system based on alarm duration probability function |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040133402A1 (en) * | 2002-11-29 | 2004-07-08 | Nissin Ion Equipment Co., Ltd. | Alarm management method and apparatus therefor |
| CN105931438A (en) * | 2016-06-27 | 2016-09-07 | 北京协同创新研究院 | Method and system for determining alarm duration statistical distribution stability |
| CN106971521A (en) * | 2017-05-03 | 2017-07-21 | 山东科技大学 | A kind of online elimination method and system for shaking alarm and repetition of alarms |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8792949B2 (en) * | 2008-03-31 | 2014-07-29 | Covidien Lp | Reducing nuisance alarms |
| EP2711907B1 (en) * | 2012-09-24 | 2015-01-21 | Hekatron Vertriebs GmbH | Detector, test device and method for testing a detector |
| CN104574729B (en) * | 2015-01-28 | 2020-07-24 | 无锡感知金服物联网科技有限公司 | Alarm method, device and system |
| CN107092654B (en) * | 2017-03-20 | 2019-02-15 | 山东科技大学 | Based on Change in Mean detection alarm normally with abnormal deviation data examination method and device |
| CN109087490B (en) * | 2018-09-03 | 2019-09-03 | 山东科技大学 | A kind of interference alarm recognition methods based on alarm duration features |
-
2018
- 2018-09-03 CN CN201811019040.5A patent/CN109087490B/en active Active
-
2019
- 2019-03-19 AU AU2019333948A patent/AU2019333948B2/en active Active
- 2019-03-19 WO PCT/CN2019/078713 patent/WO2020048114A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040133402A1 (en) * | 2002-11-29 | 2004-07-08 | Nissin Ion Equipment Co., Ltd. | Alarm management method and apparatus therefor |
| CN105931438A (en) * | 2016-06-27 | 2016-09-07 | 北京协同创新研究院 | Method and system for determining alarm duration statistical distribution stability |
| CN106971521A (en) * | 2017-05-03 | 2017-07-21 | 山东科技大学 | A kind of online elimination method and system for shaking alarm and repetition of alarms |
Also Published As
| Publication number | Publication date |
|---|---|
| AU2019333948A1 (en) | 2020-10-22 |
| CN109087490A (en) | 2018-12-25 |
| CN109087490B (en) | 2019-09-03 |
| WO2020048114A1 (en) | 2020-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2019333948B2 (en) | Interference alarm identification method based on alarm duration feature | |
| CN111143102B (en) | Abnormal data detection method and device, storage medium and electronic equipment | |
| US20190140778A1 (en) | Information processing method, information processing system, and recording medium | |
| CN110362612B (en) | Abnormal data detection method and device executed by electronic equipment and electronic equipment | |
| CN108123849B (en) | Method, device, equipment and storage medium for determining threshold value for detecting network flow | |
| CN107202604B (en) | A kind of alert processing method and system | |
| CN107888441B (en) | Network traffic baseline self-learning self-adaption method | |
| CN112348237B (en) | Abnormal trend detection method for dynamic drilling data | |
| US9524223B2 (en) | Performance metrics of a computer system | |
| JP2007310665A (en) | Process monitoring apparatus | |
| US20190164067A1 (en) | Method and device for monitoring a process of generating metric data for predicting anomalies | |
| US8386849B2 (en) | Noisy monitor detection and intermittent fault isolation | |
| CN116308285A (en) | Intelligent terminal safety supervision system based on data processing | |
| CN110971488A (en) | Data processing method, device, server and storage medium | |
| JP2023520066A (en) | Data processing for industrial machine learning | |
| CN118051863B (en) | Health data acquisition system and method based on digital metering technology | |
| CN117614978A (en) | Information security communication management system for digital workshop | |
| CN117235460B (en) | Data transmission processing method and system based on power time sequence data | |
| US10895872B2 (en) | Detection of temperature sensor failure in turbine systems | |
| Andrade et al. | Analysis of software aging impacts on plant anomaly detection with edge computing | |
| CN113052272B (en) | Abnormity detection method and device, electronic equipment and storage medium | |
| CN110245844B (en) | Abnormal index detection method and device | |
| CN114900359A (en) | Network security event backtracking method and system | |
| CN109714213B (en) | Website alarm method and system | |
| CN116027771B (en) | Abnormality detection method for industrial personal computer control system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGA | Letters patent sealed or granted (standard patent) |