JP2007048097A - Monitoring system and monitoring method in plant or the like - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a monitoring system capable of detecting occurrence of an abnormal state of a device or the like at an early stage from a change in sound in the site in a plant or the like. <P>SOLUTION: The monitoring system includes: a sound detecting part 11 for detecting sound in the field; a sound processing part 12 for obtaining sound data of N<SB>1</SB>times from input sound; a chaos information processing calculation processing part 13 for preparing an attractor on the basis of the sound data, calculating orbit parallel measurement about k data vectors among data vectors constituting the attractor and calculating the distribution of the orbit parallel measurement; a chaos information standard processing part 14 for calculating the average value and kurtosis of the orbit parallel measurement from the distribution of the orbit parallel measurement and calculating a chaos information standard value from the ratio of the both; a criterion value preparation processing part 16 for preparing a criterion value to be reference for detecting leakage sound on the basis of the obtained chaos information standard value; and a leakage sound detection determining part 17 for determining presence/no presence of leakage sound from the criterion value and the chaos information standard value. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a monitoring system and a monitoring method for determining the presence / absence of abnormality of various devices from a change in sound in a site at a site such as a factory.

In equipment and equipment (hereinafter referred to as equipment) in refineries, chemical factories, power plants, garbage incinerators and other facilities (hereinafter referred to as factories), abnormalities in the equipment etc. are discovered early and dealt with promptly. It is important to.
For this reason, various systems and methods for early detection of abnormalities in these devices have been proposed (see, for example, Patent Documents 1 and 2).
However, in the system and method described in the above document, it is necessary to provide detection devices according to the number and type of devices in a site where a variety of devices are complicatedly mixed. However, the current situation is that the detection device may not be completely detected because it may not be detected when a failure or unexpected abnormality occurs.

JP 2003-51894 A JP 2003-134261 A

By the way, a skilled operator can detect an abnormality of a device in the field where he / she is in charge from a subtle change in operation sound of the device. Therefore, it is possible to detect an abnormality at an early stage based on the intuition and experience of a skilled operator.
However, such skilled operators tend to decrease with automation accompanying rationalization, and development of a system that makes use of the intuition and experience of skilled operators is desired.

  The present invention has been made to meet such a demand, and an object of the present invention is to provide a monitoring system and a monitoring method capable of detecting the occurrence of an abnormality in an apparatus or the like at an early stage from a change in sound in the field.

  As a result of intensive research by the inventors of the present invention, an attractor is obtained from a time-series graph of acoustic data obtained in the field, and the concept of “orbit parallel measure” known in Japanese Patent Laid-Open No. 10-134034 is known for this attractor. Furthermore, the present inventors have conceived that the distinction between background noise and leaked sound can be clarified by expressing the distribution state of the orbital parallel measure by the “chaos information normative value” created by the inventor of the present invention.

Specifically, as described in claim 1, the monitoring system of the present invention is a monitoring system that determines whether or not an abnormality has occurred from the detected sound at a site such as a factory. A sound detection unit to detect, a sound processing unit that processes the sound input from the sound detection unit at regular time intervals, and obtains time series sound data for N ₁ (N ₁ is a natural number greater than ₁ ) times; An attractor is created based on the time-series acoustic data, and a trajectory parallel measure is obtained for each of k data vectors (k is a natural number greater than 1) extracted from the data vectors constituting the attractor. The average value and kurtosis of the trajectory parallel measure are obtained from the chaos information processing calculation processing unit for obtaining the distribution of the measure and the distribution of the trajectory parallel measure obtained by the chaos information processing calculation processing unit. A chaos information norm processing unit for obtaining a chaos information norm value from a ratio of the average value and the kurtosis, and a reference for detecting leaked sound based on the chaos information norm value obtained by the chaos information norm processing unit; From the judgment reference value creation processing unit for creating the judgment reference value, the judgment reference value created by the judgment reference value creation processing unit, and the chaos information norm value obtained by the chaos information norm processing unit, a leaked sound A leakage sound detection determination unit that determines the presence or absence of noise and an alarm output unit that outputs an alarm when the leakage sound detection determination unit determines that there is an abnormality.

According to this configuration, determine the attractor from the time series sound data of N ₁ times, we obtain a trajectory parallel measure for k data vectors extracted from the data vectors constituting the attractor. And the average value and kurtosis are calculated | required from distribution of the obtained orbit parallel measure, and chaos information normative value is calculated | required from the ratio.
Here, the average value of the orbital parallel measure means that there is regularity as the value is small, and the kurtosis means that the structure of the local space is similar as the value is large.
Therefore, by determining the ratio between the two, it becomes easy to analyze the structure of the time-series acoustic data, and to easily distinguish between background noise and leaked sound.

In this case, as described in claim 2, the determination reference value creation processing unit repeats the process by the chaos information norm processing unit N ₂ (N ₂ is a natural number greater than 1) times and obtains the N ₂ pieces The maximum value, the minimum value, and the standard deviation may be obtained from the chaos information standard value, and the upper limit and the lower limit of the determination reference value may be obtained based on the maximum value, the minimum value, and the standard deviation.
According to a third aspect of the present invention, the leakage sound detection determination unit repeats the processing by the chaos information norm processing unit N ₃ (N ₃ is a natural number greater than 1) times, and obtains N ₃ chaos information norms. An average value may be obtained from the values, and the average value may be compared with the determination reference value to determine the presence or absence of leakage sound.
In order to suppress false alarms, as described in claim 4, the alarm output unit issues an alarm when the abnormality determination by the leakage sound detection determination unit is repeated continuously for a preset number of times. It is good to output.

The monitoring method of the present invention is a monitoring method for judging whether or not an abnormality has occurred from the detected sound at a site such as a factory, as described in claim 5, wherein the sound detected in the site is fixed. Processing for each time interval, obtaining N ₁ (N ₁ is a natural number greater than ₁ ) times of time series acoustic data, creating an attractor based on this time series acoustic data, and data constituting this attractor K data vectors are extracted from the vectors, the trajectory parallel measure and its distribution are obtained for these k data vectors, and the average value and kurtosis of the trajectory parallel measure are calculated from the distribution of the trajectory parallel measures. Determining a chaos information normative value that is a ratio of the average value and the kurtosis; and determining a reference for detecting leaked sound based on the chaos information normative value. A method comprising: creating a reference value; determining whether there is a leaked sound from the determination reference value and the chaos information normative value; and outputting an alarm when it is determined that the leaked sound exists It is.

In the step of creating the determination reference value, the determination reference value is obtained by repeating the process of obtaining a chaos information standard value N ₂ (N ₂ is a natural number greater than 1) times. The maximum value, the minimum value, and the standard deviation may be obtained from the N ₂ chaos information standard values obtained, and the upper limit and the lower limit of the determination reference value may be obtained based on the maximum value, the minimum value, and the standard deviation.
Further, as described in claim 7, in the step of determining the presence or absence of the leakage sound, the process of obtaining chaotic information normative processed value N _{3 (N 3} is greater than 1 natural number) repeated times, obtained by this An average value may be obtained from N _three chaos information normative values, and the average value may be compared with the determination reference value to determine the presence or absence of leakage sound.

  According to the present invention, the chaos information normative value is characterized by the ability to detect leaked sound regardless of the magnitude of the sound. It becomes possible to detect the occurrence of the problem at an early stage.

Preferred embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram illustrating a configuration according to an embodiment of the monitoring system of the present invention.
The monitoring system 1 according to the present invention includes a sound collecting microphone 11 that collects sound in the field, a sound processing unit 12 that processes sound collected by the sound collecting microphone 11 at regular time intervals, and a sound processing unit 12. The obtained time-series acoustic data (time-series acoustic data) is processed by the chaotic attractor theory, and the chaotic information calculation processing section 13 for creating an attractor for the time-series acoustic data, and the chaotic attractor information calculation processing section 13 There was a CIC processing unit 14 for obtaining a trajectory parallel measure of the obtained attractor, obtaining a Chaos Information Criteria (hereinafter abbreviated as CIC) value from the obtained orbit parallel velocity distribution, and a reference value creation instruction. In some cases, the mode switching unit 15 that switches the leaked sound detection mode to the determination reference value creation mode, and the CIC The determination reference value creation processing unit 16 that creates a determination reference value based on the CIC value obtained by the processing unit 14, and the determination reference value created by the determination reference value creation processing unit 16 and the CIC value are compared. A leak sound detection determination unit 17 that determines the presence or absence of a leak sound, and an alarm output unit 18 that outputs an alarm notification when the leak sound detection determination unit 17 determines that a leak sound has occurred. ing.

FIG. 2 is a flowchart for explaining the operation of the monitoring system configured as described above.
Sound is collected from the acoustic microphone (step S1) is treated repeatedly N ₁ times at predetermined time intervals in the acoustic processing unit 12, the time series sound data of N ₁ times is created (step S2, S3). At this time, a filtering process may be performed as necessary.
An example of acoustic data collected at regular time intervals is shown in FIGS. 3 and 4, (a), (b), and (c) are the N _n-1 th, N _n th, N _{n + 1} th times (n is a natural number greater than 1) collected repeatedly at regular intervals. ), The vertical axis represents gain, and the horizontal axis represents time. 3 shows background acoustic data, and FIG. 4 shows leaked sound data.

There is no regularity between the sound data of the background noise and the sound data of the leaked sound shown in FIGS. 3 and 4, and it is very difficult to distinguish between the background noise and the leaked sound as it is.
Although there is a method using a power spectrum that is well known as a method for analyzing random signal-like data, the power spectra of the background noise and the leaked sound are similar in shape, and are still difficult to distinguish.
Therefore, in the present invention, time-series acoustic data obtained by N _one- time acoustic processing is processed by the chaos information calculation processing unit 13 according to the chaos attractor theory, and k pieces (k is larger than 1) from the obtained attractors. A natural number) orbit parallel measure is obtained (step S4), and the k orbit parallel measures are processed by the CIC processing unit 14 using the concept of chaos information norm (CIC) (step S5). .
The processing procedure will be described below.

[Building attractors]
The chaos information processing calculation processing unit 13 is described in documents such as Japanese Patent Application Laid-Open No. 2002-73587 (“chaos analysis device”) and Japanese Patent Application Laid-Open No. 2000-292392 (“gas measurement device and gas measurement method”). The attractor is constructed by a well-known technique.
In constructing the attractor, a set of signal values of M time points different from the time series signal x (t) (t is time) by time delay τ {x (t), x (t + τ), x (t + 2τ), ..., x (t + (M-1) τ)} is created, and this is plotted in a two-dimensional space (phase space) to construct an attractor.

The attractor thus constructed is a time-series trajectory (trajectory) of the target signal in the two-dimensional phase space. Each point of the attractor is specified by time t, and the position coordinates represented by M components (that is, x (t + kτ) (k = 0, 1, 2,..., M−1)) are represented in the two-dimensional space. Occupy.
FIG. 5 and FIG. 6 show an example of the background noise and leakage sound attractor constructed in the above procedure. 5 and 6, (a), (b), and (c) show three attractors in different time zones.

[Determine orbit parallel measure]
Next, k data vectors constituting the attractor are selected by random numbers, and a trajectory parallel measure of the k data vectors is obtained.
Here, the trajectory parallel measure is an index indicating the degree of parallelism of adjacent trajectories in the partial space on the attractor.
The procedure for obtaining the trajectory parallel measure for one of the selected k data vectors will be described with reference to FIG.

First, select an arbitrary vector Xi on the orbit that constitutes the attractor, select m neighborhood vectors Xj (j = 1,2,3, ..., m) close to Xi in the Euclidean distance, Respective tangent unit vectors Ti and Tj with respect to the trajectories of Xi and the neighborhood vector Xj are derived. Unit tangent vector Ti, derivation of Tj is assumed ultra circle S _i passing through three points of a point selected Xi and points before and after, to approximately derived by the following procedure.
(1) Obtain normal vectors at Xi and Xi-1, and normal vectors at Xi and Xi + 1.
(2) Find the intersection Oi of these two normal vectors (see FIG. 7).
(3) A correlation vector Li between Xi and Oi is obtained (see FIG. 7).
(4) A unit tangent vector Ti orthogonal to the correlation vector Li is obtained (see FIG. 7).
(5) Similarly, an intersection Oj is obtained from Xj-1, Xj, and Xj + 1, and a correlation vector Lj and a unit tangent vector Tj orthogonal thereto are obtained from Xj and Oj (see FIG. 7). Note that || Ti || = || Tj || = 1.
When the unit tangent vector Ti thus obtained is used as a reference, the variation in the direction of the unit tangent vector of the neighborhood vector is obtained by the following equation.

here,
γ _i : parallelism in local space m: number of neighboring vectors Ti: unit tangent vector of extracted data vector Xi Tj: unit tangent vector of neighboring vector Xj This processing is performed on a local space randomly sampled from the entire attractor, Find the average parallelism of proximity vectors. The trajectory parallel measure (TPM) is 0 if the proximity vectors are completely parallel, and 0.5 if orthogonal.
If the orbital parallel measure (TPM) is represented by Γ, the number of samples is k, and between Γ and k

The relationship is established.
Next, the average value (TPM _svg ), median value (TPM _med ), and kurtosis (TPM _ku ) are obtained for the trajectory parallel measure Γ thus obtained. This is repeated N ₂ times.
TPM _svg and TPM _med indicate regularity as the value is small, and TPM _ku indicates that the value is concentrated as the value is large. That is, the larger the value of TPM _ku, the more similar the structure of the local space.
Therefore, the chaos information norm (CIC) expressed by the following equation is used for leaking sound detection.

この表からわかるように、暗騒音のＣＩＣ値は漏洩音のＣＩＣ値よりも大きく、規則的かつ類似的であるのがわかる。
そこで、この実施形態では、暗騒音と漏洩音のＣＩＣ値の違いに着目し、この値の違いから漏洩音の有無を判断するようにしている。すなわち、現場内の時系列音響データから得られたＣＩＣ値を基に判定基準値を作成し、この判定基準値と実測から得られたＣＩＣ値とを比較することで、漏洩音の有無を判断するようにしているわけである。 CIC indicates that the larger the value, the higher the regularity and the stronger the fractal property (self-similarity).
An example of the average value (TPM _svg ), kurtosis (TPM _kur ) and chaos information norm (CIC) value when the number of samples k = 1000 and N ₂ = 4 times is shown in the following table.

As can be seen from this table, the CIC value of background noise is larger than the CIC value of leaked sound, and is found to be regular and similar.
Therefore, in this embodiment, attention is paid to the difference in the CIC value between the background noise and the leaked sound, and the presence or absence of the leaked sound is determined from the difference in this value. That is, a judgment reference value is created based on the CIC value obtained from the time-series acoustic data in the field, and the presence or absence of leaking sound is judged by comparing this judgment reference value with the CIC value obtained from actual measurement. That is why.

[Create judgment criteria]
The determination reference value is created regularly or appropriately according to an instruction from the operator (step S6). When the determination reference value is created, the mode switching unit 14 is switched to the “reference value creation mode”. Creation of the judgment reference value (step S7) in the judgment reference value creation processing unit 16 is performed in the following procedure.
In obtaining the determination reference value, the maximum value (CIC _max ), the minimum value (CIC _min ), and the standard deviation (CIC _sd ) of the CIC value of the background noise created by the chaos information norm calculation processing unit 13 are obtained. Then, the maximum value (CICref _max ) and the minimum value (CICref _min ) of the determination reference value are obtained using the following equations.
CICref _max = CIC _max + α · CIC _sd
CICref _min = CIC _min + β · CIC _sd
Here, α is a determination criterion maximum value sensitivity coefficient, β is a determination criterion minimum value sensitivity coefficient, and an optimal one is selected from experiments and experience.

Using the above formula, CIC _max , CIC _min in the case of Table 1 above,
The calculation results of CIC _sd , CICref _max and CICref _min are shown in the following table.
Note that α = 2.0 and β = 2.0.

[Detection of leakage sound]
After the determination reference value is created by the above procedure, the mode switching unit 14 is returned to the normal “monitoring mode” and monitoring in the field such as a factory is performed.
In this monitoring mode, the CIC value is obtained by the same procedure as described above, and the average value CIC _avg of the CIC value when N ₃ times is repeated is calculated.
The leakage sound detection determination unit 17 compares CIC _avg with the maximum value (CICref _max ) and minimum value (CICref _min ) of the determination reference value (step S8), and CIC _avg is within the range between the maximum value and the minimum value. It is judged whether it is included in (step S9). If it is out of the range, it is determined that an abnormality has occurred, and an alarm is notified from the alarm output unit 18 (step S10).
In this case, in order to prevent false alarms, the alarm output unit 18 determines whether or not the determination of out of range has been continuously repeated a preset number of times (for example, three times), and outputs an alarm when this number is exceeded. It is good to do.

Although preferred embodiments of the present invention have been described, the present invention is not limited to the above-described embodiments.
For example, in the above description, CIC = TPM _ku / TPM _svg is defined, but the reciprocal, that is, TPM _svg / TPM _ku can be defined as CIC.

  The present invention can be applied to any kind of plant such as an atmospheric distillation tower such as a refinery, a precision distillation tower, a fluidized catalytic fractionator, an ethylene unit cracking furnace of a petrochemical factory, and a facility such as a power plant or a garbage incinerator. Applicable for on-site monitoring.

It is a block diagram explaining the structure concerning one Embodiment of the monitoring system of this invention. It is a flowchart explaining the effect | action of the monitoring system of FIG. It is a graph which shows an example of the acoustic data of background noise. It is a graph which shows an example of the acoustic data of leakage sound. It is a graph which shows an example of the attractor of the background noise in the field. It is a graph which shows an example of the attractor of the leaking sound in the field. It is a figure explaining the procedure which calculates | requires an orbit parallel measure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Monitoring system 11 Sound collecting microphone 12 Acoustic processing part 13 Chaos information processing part 14 CIC processing part 15 Mode switching part 16 Judgment reference value creation processing part 17 Leakage sound detection judgment part 18 Alarm output part

Claims (7)

It is a monitoring system that judges the presence or absence of an abnormality from the detected sound at a site such as a factory,
An acoustic detection unit for detecting acoustics in the field;
An acoustic processing unit that processes the sound input from the acoustic detection unit at regular time intervals to obtain time-series acoustic data for N ₁ (N ₁ is a natural number greater than ₁ ) times;
An attractor is created based on the time-series acoustic data, and a trajectory parallel measure is obtained for each of k data vectors (k is a natural number greater than 1) extracted from the data vectors constituting the attractor. A chaos information processing processor that calculates the distribution of measures;
From the distribution of the orbital parallel measure obtained by the chaos information processing calculation processing unit, an average value and a kurtosis of the orbital parallel measure are obtained, and a chaos information normative value is obtained from a ratio between the average value and the kurtosis. An information norm processing department;
Based on the chaos information norm value obtained by this chaos information norm processing unit, a determination criterion value creation processing unit that creates a criterion value that serves as a reference for leakage sound detection;
From the determination reference value created by this determination reference value creation processing unit and the chaos information norm value obtained by the chaos information norm processing unit, a leaked sound detection determination unit that determines the presence or absence of leaked sound,
An alarm output unit that outputs an alarm when the leakage sound detection determination unit determines that there is an abnormality, and
A monitoring system in a factory or the like, characterized by comprising: Wherein in the determination reference value creation unit, the maximum value processing by chaos information normative processor N _{2 (N 2} is greater than 1 natural number) from times repeated N ₂ pieces of the chaotic information normative value obtained, the minimum value and the standard The monitoring system in a factory or the like according to claim 1, wherein a deviation is obtained, and an upper limit and a lower limit of a determination reference value are obtained based on the maximum value, the minimum value, and the standard deviation. In the leaked sound detection determination unit, an average value is obtained from N ₃ chaos information norm values obtained by repeating the processing by the chaos information norm processing unit N ₃ (N ₃ is a natural number greater than 1) times, and this average value and The monitoring system in a factory or the like according to claim 1 or 2, wherein the presence or absence of leakage sound is determined by comparing with the determination reference value.   4. The alarm output unit according to claim 1, wherein the alarm output unit outputs an alarm when abnormality determination by the leakage sound detection determination unit is repeated continuously for a preset number of times. 5. Monitoring system in factories. It is a monitoring method for judging whether or not an abnormality has occurred from the detected sound at a site such as a factory,
Processing sound detected in the field at regular time intervals to obtain time series sound data for N ₁ (N ₁ is a natural number greater than ₁ ) times;
Creating an attractor based on the time-series acoustic data;
Extracting k data vectors from the data vectors constituting the attractor, obtaining an orbital parallel measure and a distribution thereof for the k data vectors;
From the distribution of the trajectory parallel measure, obtaining an average value and kurtosis of the trajectory parallel measure, obtaining a chaos information norm value that is a ratio of the average value and the kurtosis;
Based on the chaos information norm value, creating a criterion value that serves as a reference for leaked sound detection;
Determining the presence or absence of leaked sound from the criterion value and the chaos information norm value;
Outputting an alarm when it is determined that the leakage sound is present;
A monitoring method in a factory or the like characterized by comprising: In the step of creating the determination reference value, the process of obtaining the chaos information norm value is repeated N ₂ (N ₂ is a natural number greater than 1) times, and the maximum value is obtained from the N ₂ chaos information norm values obtained thereby. 6. A monitoring method in a factory or the like according to claim 5, wherein a minimum value and a standard deviation are obtained, and an upper limit and a lower limit of a determination reference value are obtained based on the maximum value, the minimum value, and the standard deviation. In the step of determining the presence or absence of the leaking sound, the process of obtaining the chaos information norm processing value is repeated N ₃ (N ₃ is a natural number greater than 1) times, and an average is obtained from the N ₃ chaotic information norm values obtained thereby. 7. A monitoring method in a factory or the like according to claim 5 or 6, wherein a value is obtained and the average value is compared with the determination reference value to determine the presence or absence of leaking sound.

Images