JP2003344122A - Status monitoring method, status monitoring system, status monitoring device, computer program, and recording medium - Google Patents

Abstract

(57) [Summary] [Problem] A state monitoring method, a state monitoring system, a state monitoring device, and a computer capable of quickly estimating a parameter of a model that is theorized about a relationship between time-series data indicating a state of a monitoring target A program and a recording medium are provided. A state monitoring device stores a first data string composed of N time-series data and a first estimated value of a parameter. When a second data string composed of m pieces of time-series data is received (S4), a third data string composed by adding the second data string to the first data string is created (S5). A second estimated value of the parameter related to the third data string is calculated from the first estimated value and the third data string using an equation corresponding to an equation obtained by repeating the calculation m times (S7).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a state monitoring method, a state monitoring system, a state monitoring device, and a computer for monitoring the state of a monitored object such as equipment in a factory and detecting the abnormality in real time when the abnormality occurs. The present invention relates to a computer program for realizing the above as a status monitoring device, and a recording medium.

[0002]

2. Description of the Related Art A method for regularly inspecting and maintaining various facilities such as production facilities used in factories or public facilities used in society in order to prevent occurrence of failure ( TBM: Time Based preventive Maint
enance) has been used in the past. In addition to TBM,
A method of constantly monitoring the condition of the equipment and performing maintenance when the condition of the equipment deteriorates to a predetermined standard (CBM: Conditio
Based Preventive Maintenance) is known. C
The BM has advantages over the TBM in that it is possible to avoid an early failure or easily reduce the maintenance cost. In recent years, the CBM has begun to prevail instead of the TBM.

In order to realize a CBM, data such as the number of operations, temperature, or vibration indicating the state of equipment is acquired in real time, the acquired data is accumulated, and the status of the equipment is estimated using the accumulated data. However, a facility monitoring system is required to detect the abnormality if it occurs. As a method of estimating the state of the equipment, there is a method of processing the acquired data, calculating parameters relating to the state of the equipment, and determining that the equipment is abnormal when the change in the calculated parameters becomes large. As an example of this method, there is a method of using an RMS (Root Mean Square) value that is a square root of a value obtained by squaring the change amount of the acquired data. For example, when acquiring the vibration of a machine as data, the change in the position of the machine due to the vibration is accumulated and the average position of the vibration is calculated.
When the RMS value, which is the square root of the squared value of the deviation of the position from the average position, is obtained and the RMS value becomes large,
It is determined that the equipment is in an abnormal state because the vibration width becomes large.

Another method for estimating the state of equipment is as follows:
Create a model that shows the relationship between the acquired data,
There is a method of applying the acquired data to a model to estimate a parameter indicating a relationship between the data and determining that the equipment is abnormal when the estimated parameter largely changes.
The Kalman filter is known as this method,
In the Kalman filter, parameters are estimated for up to N pieces of data based on the model, and when the N + 1th data is acquired, the parameters are estimated using the N + 1 pieces of data.

As a simple model showing the relationship between data
AR (Autoregres) which is one of the linear models
sive) model is often used. The status of the monitoring target
Data string {x
_n : N = 1, 2, ...} is a stationary ergodic normal process
When the sample time series is used, in the p-order AR model,
{X _n } The relationship between elements is

[0006]

[Equation 4]

It is represented by Here, {e _n : n = 1,
2, ...} are normal noises with a mean value of 0 and variance σ _e ² , and {a _n : n = 1, 2, ..., P} are parameters of the AR model. Data string {x _n } and parameter {a
_{When n} } is set as Zn = [− x _n−1 , −x _n−2 , ..., −x _np ] ^T θ = [a ₁ , a ₂ ..., a _p ] ^T , the equation (1) is it can be expressed as ^{_{x n = θ T Z n +}} e n. When a data string {x _n : n = 1, 2, ..., N} composed of N time-series data is obtained, in order to obtain the parameter θ from the data string by the least square method,

[0008]

[Equation 5]

It is only necessary to find θ that minimizes the right side of. Therefore, by partially differentiating the right side of the equation (2) by θ and making it equal to 0, the estimated value θ ^{^} _N estimated from the data string of the parameter θ is

[0010]

[Equation 6]

[0011] Next, in addition to the data string (N
Consider the case where the +1) th data x _{N + 1} is obtained. x
A data string to which _{N + 1} is added {x _n : n = 1, 2, ..., N,
Estimated value θ of parameter θ that can be estimated from N + 1}
^{^} _{N + 1} is calculated from the equation (3),

[0012]

[Equation 7]

[0013] Where the p × p matrix A _N is

[0014]

[Equation 8]

Let us say that. n × n matrix Q, n × 1 matrix R
And the matrix R ′ of 1 × n, the following formula (Q + RR ′) ⁻¹ = Q ⁻¹ −Q ⁻¹ RR′Q ⁻¹ / (1 + R ′)
Q ⁻¹ R), the matrix A _{N + 1} is: A _{N + 1} ⁻¹ = A _N ⁻¹ −A _N ⁻¹ Z _{N + 1} Z _{N + 1} ^T A _N ⁻¹ / (1 + Z
_{N + 1} ^T A _N ^-1 Z _{N + 1} ). By applying this equation to the equation (4), I
Is an identity matrix, the estimated value θ ^{^} _{N + 1} is θ ^{^} _{N + 1} = {I- _AN ^-1 Z _{N + 1} Z _{N + 1} ^T / (1 + Z _{N + 1} ^T _AN ^-1 Z _{N + 1} )} θ ^{^} _N + A _N ^-1 Z _{N + 1} x _{N + 1} / (1 + Z _{N + 1} ^T A _N ^-1 Z _{N + 1} ) ... (5) Further, if K _{N + 1} = A _N ⁻¹ Z _{N + 1} / (1 + Z _{N + 1} ^T A _N ⁻¹ Z _{N + 1} ), then equation (5) yields θ ^{^} _{N + 1} = (I− K _{N + 1} Z _{N + 1} ^T ) θ ^{^} _N + K _{N + 1} x _{N + 1} (6) By substituting the value of the data string into the equation (6), the model parameter can be estimated. The above is the method of estimating the parameter using the Kalman filter, and the state of the monitoring target can be determined based on the estimated parameter.

[0016]

When performing CBM in a factory, each facility is equipped with a sensor, and the data acquired by the sensor is used as an arithmetic unit such as a personal computer (PC), a microcomputer, a minicomputer or a process computer. Accumulates online and processes the accumulated data to monitor equipment. In the parameter estimation using the Kalman filter described above, each time a new piece of data is obtained, the accumulated data is processed to calculate an estimated value of the parameter. The behavior of the equipment in the factory is fast, and when using a computing device with low computing capacity, the speed at which data is accumulated is faster than the processing speed performed by the computing device for each new data. , Failure to detect equipment abnormalities, or detection of abnormalities may be delayed. In addition, there is a problem that it is difficult to determine the change in the state of the equipment because the parameter hardly changes even if a new piece of data is added to many data.

The present invention has been made in view of such circumstances, and its purpose is to estimate a parameter from a large number of data when a monitoring target is normal and use a Kalman filter. Using the formula corresponding to the estimated value obtained by repeating the parameter estimation multiple times,
When a plurality of new data are obtained, the obtained data are collectively processed to estimate the change in the parameter, so that the parameter is sufficiently fast compared with the speed at which the plurality of data are accumulated. For realizing a state monitoring method, a state monitoring system, a state monitoring device, and a computer as the state monitoring device. It is to provide a computer program and a recording medium.

[0018]

A state monitoring method according to a first aspect of the present invention uses a computer provided with a receiving unit for receiving information from the outside, a storage unit, and a computing unit, and uses a plurality of time series indicating a state of a monitoring target. In the method of monitoring the state of a monitoring target according to a model that theoreticalizes the relationship between data, a first data string composed of a plurality of time-series data indicating the state of the monitoring target is stored in a storage unit, and A plurality of other time series indicating the state of the monitoring target are stored in the storage unit, which are first estimated values of parameters indicating the relationship between the plurality of time series data included in the first data sequence according to the model. The second data string composed of data is externally received by the reception unit, and each time the plurality of time-series data included in the second data string is sequentially added to the first data string, the parameter related to the data string is formed. A third data string formed by adding the second data string to the first data string by using an equation corresponding to an equation in which the estimation of the data is repeated using the Kalman filter for the number of time series data included in the second data row. From the data string and the first estimated value, the second estimated value, which is the estimated value of the parameter indicating the relationship between the plurality of time series data included in the third data string according to the model, is calculated by the calculation unit and calculated. The arithmetic unit determines the state of the monitoring target based on the second estimated value.

A state monitoring system according to a second aspect of the present invention theoretically describes a relationship between a data acquisition device that acquires a plurality of time-series data indicating a state of a monitoring target and a plurality of time-series data acquired by the data acquisition device. In a status monitoring system including a status monitoring device that monitors the status of the monitoring target according to a model that has been modeled, the status monitoring device generates a first data string composed of a plurality of time-series data indicating the status of the monitoring target. The data acquisition device comprises: a storage unit; and a storage unit that stores a first estimated value that is an estimated value of a parameter that indicates a relationship between a plurality of time series data included in the first data string according to the model. The state monitoring apparatus further comprises means for inputting a second data string composed of a plurality of acquired time series data to the state monitoring apparatus, and the state monitoring apparatus further includes the input second data. Second using a Kalman filter to estimate the parameter a plurality of time-series data sequentially according to the data sequence constructed each time applied to the first data string contained in
The model is calculated from a third data string and a first estimated value, which are formed by adding the second data string to the first data string, using an expression corresponding to an expression repeated by the number of time series data included in the data string. Which is the estimated value of the parameter indicating the relationship between the plurality of time series data included in the third data string according to
It is characterized by comprising means for calculating an estimated value and means for determining the state of the monitoring target based on the calculated second estimated value.

A condition monitoring system according to a third aspect of the present invention theoretically describes a relationship between a data acquisition device that acquires a plurality of time series data indicating a condition of a monitoring target and a plurality of time series data acquired by the data acquisition device. In a status monitoring system including a status monitoring device that monitors the status of the monitoring target according to a model that has been modeled, the status monitoring device includes N time-series data {x _n : n = 1, 1 indicating the status of the monitoring target. Two
, N} for storing a first data string, and p coefficients indicating the relationship between a plurality of time-series data included in the first data string according to the p-order AR model. And a means for storing a first estimated value θ ^{^} _N which represents the estimated value of the parameter consisting of a matrix of p rows and 1 column, and the data acquisition device acquires m time-series data {x
_n : n = N + 1, N + 2, ..., N + m} is provided with a means for inputting a second data string to the status monitoring device, and the status monitoring device is further configured to input the input second data string as a first data string. An estimated value of a parameter indicating a relationship between a plurality of time series data included in the third data string according to the model, using the third data string configured in addition to the data string and the first estimated value θ ^{^} _N. The second estimated value θ ^{^} _{N + m,} which is a p-by-1 matrix,

[0021]

[Equation 9]

Means for calculating and the calculated second estimated value θ
Means for determining the state of the monitoring target based on ^{^} _{N + m} .

According to a fourth aspect of the present invention, there is provided a state monitoring device for monitoring the state of a monitoring target according to a model that theoreticalizes the relationship between a plurality of time series data indicating the state of the monitoring target. Means for storing a first data string composed of a plurality of time-series data indicating a plurality of time-series data and a first parameter that is an estimated value of a parameter indicating a relationship between a plurality of time-series data included in the first data string according to the model. A means for storing the estimated value; a means for externally receiving a second data string composed of a plurality of other time series data indicating the state of the monitoring target; and a plurality of time series data included in the second data string. The estimation of the parameters related to the data string that is formed each time it is sequentially added to the first data string is repeated by the Kalman filter by the number of timeseries data included in the second data string. The relationship between the plurality of time-series data included in the third data sequence according to the model from the third data sequence configured by adding the second data sequence to the first data sequence and the first estimated value using the formula And means for calculating a second estimated value that is an estimated value of the parameter, and means for determining the state of the monitoring target based on the calculated second estimated value.

According to a fifth aspect of the present invention, there is provided a state monitoring device for monitoring the state of a monitoring target according to a model that theoreticalizes the relationship between a plurality of time series data indicating the state of the monitoring target. Means for storing a first data string composed of N time-series data {x _n : n = 1, 2, ..., N}, and a first data string according to the model that is a p-order AR model. Means for storing a first estimated value θ ^{^} _N , which represents an estimated value of a parameter consisting of p coefficients showing a relationship between a plurality of time-series data included in, in a matrix of p rows and 1 column, and the monitoring target. And means for externally receiving the second data string composed of the other m pieces of time series data {x _n : n = N + 1, N + 2, ..., N + m}.
A relationship between a plurality of time-series data included in the third data string according to the model using the third data string configured by adding the second data string to the first data string and the first estimated value θ ^{^} _N. The second estimated value θ ^{^} _{N + m in} which the estimated value of the parameter indicating

[0025]

[Equation 10]

Means for calculating and the calculated second estimated value θ
Means for determining the state of the monitoring target based on ^{^} _{N + m} .

A computer program according to a sixth aspect of the present invention comprises a plurality of time-series data indicating a state of a monitoring target according to a model that theoreticalizes a relationship between a plurality of time-series data indicating a state of a monitoring target. A first data string and a first parameter that is an estimated value of a parameter indicating a relationship between a plurality of time series data included in the first data string according to the model.
A computer program for causing a computer storing an estimated value to monitor the state of the monitoring target, wherein the computer sequentially adds a plurality of time-series data included in a second data sequence received from the outside to the first data sequence. Using the equation corresponding to the equation in which the estimation of the parameter related to the data string that is constructed each time is repeated by the Kalman filter for the number of time series data included in the second data string,
An estimated value of a parameter indicating a relationship between a plurality of time series data included in the third data string according to the model, from the third data string and the first estimated value configured by adding the second data string to the data string. A procedure for calculating a certain second estimated value,
A step of causing a computer to determine the state of the monitoring target based on the calculated second estimated value.

The computer program according to the seventh aspect of the invention is a set of N time-series data {x _n : {x _n : p = 1, which indicates the relationship between the first data string composed of n = 1, 2, ..., N} and a plurality of time-series data included in the first data string according to the model that is a p-order AR model. A computer program for storing a first estimated value θ ^{^} _N , which represents an estimated value of a parameter composed of a coefficient of a matrix of p rows and 1 column, for causing the computer to monitor the state of the monitoring target, M time-series data {x _n : n received from
= N + 1, N + 2, ..., N + m}, a third data string configured by adding a second data string composed of the first data string and a first estimated value θ ^{^} _N to a third data string according to the model. The second estimated value θ ^{^} _{N + m} , which represents the estimated value of the parameter indicating the relationship between the plurality of time-series data included in the data string, in a p-row by 1-column matrix,

[0029]

[Equation 11]

And a step of causing a computer to determine the state of the monitoring target based on the calculated second estimated value θ ^{^} _{N + m} .

A recording medium according to an eighth aspect of the present invention comprises a plurality of time-series data indicating a state of a monitoring target according to a model that theoreticalizes a relationship between a plurality of time-series data indicating a state of a monitoring target. The state of the monitoring target is monitored in a computer that stores one data string and a first estimated value that is an estimated value of a parameter that indicates a relationship between a plurality of time-series data included in the first data string according to the model. A computer-readable recording medium that stores a computer program for causing the computer to read
The time series included in the second data sequence using the Kalman filter for estimating the parameter related to the data sequence configured each time a plurality of time series data included in the second data sequence received from the outside is sequentially added to the first data sequence. Included in the third data string according to the model from the third data string and the first estimated value configured by adding the second data string to the first data string, using an expression corresponding to the expression repeated for the number of data And a step of causing a computer to determine a state of the monitoring target based on the calculated second estimated value, and a step of calculating a second estimated value which is an estimated value of a parameter indicating a relationship between a plurality of time series data. A computer program including is recorded.

The present inventor does not calculate the estimated values of the parameters sequentially as in the prior art, but estimates the parameters by using a Kalman filter when m pieces of time series data are added to N pieces of time series data. The formula for calculating the value is derived.

Data string {x _n : n = 1, 2, ..., N, N
When (N + 2) th data x _{N + 2} is obtained in addition to +1}, a data string {x _n : n = 1, x _{N + 2} is added.
The estimated value θ ^{^} _{N + 2} of the parameter θ that can be estimated from 2, ..., N + 2} is θ ^{^} _{N + 2} = (I-K _{N + 2} Z _{N + 2} ^T ) (I-K _{N + 1ZN +} ^1T )? ^{^} _N + (I-KN _{+ 2ZN +} ^2T ) _{KN + 1xN + 1} + _{KN + 2xN + 2} . Similarly, the estimated value θ ^{^} _{N + 3} of the parameter θ that can be estimated from the data string obtained by adding the (N + 3) th data x _{N + 3} to the data string is θ ^{^} _{N + 3} = (I-K _{N + 3} Z _{N + 3} ^T ) θ ^{^} _{N + 2} + K _{N + 3} x _{N + 3} = (I-K _{N + 3} Z _{N + 3} ^T ) (I-K _{N + 2} Z _{N + 2} ^T ) (I- K _{N + 1} Z _{N + 1} ^T ) θ ^{^} _N + (I-K _{N + 3} Z _{N + 3} ^T ) (I-K _{N + 2} Z _{N + 2} ^T ) K _{N + 1} x _{N + 1} + ( I-K _{N + 3} Z _{N + 3} ^T ) K _{N + 2} x _{N + 2} + K _{N + 3} x _{N + 3} . Generalizing the above equation, the first data string {x _n : n =
1, 2, ..., N}, and a second data string {x _n : n = N + 1, N + 2, which is composed of m pieces of time-series data.
, N + m} is obtained, a third data string {x _n : n = 1, 2, ..., In which the second data string is added to the first data string
The estimated value θ ^{^} _{N + m} of the parameter θ that can be estimated from N, ..., N + m} is

[0034]

[Equation 12]

It can be expressed as Expression (7) corresponds to an expression in which the Kalman filter calculation is repeated m times. Now consider the nature of K _{N + 1} .

[0036]

[Equation 13]

And R _i , A _N is

[0038]

[Equation 14]

It can be expressed as Therefore, when N → ∞, 1 + Z _{N + 1} ^T _AN ⁻¹ Z _{N + 1} = 1 + 0 (1 / N) → 1. Here, 0 (1 / N) indicates that the expression is expressed in the order of (1 / N). Therefore, N
→ ∞, K _{N + 1} is K _{N + 1} = A _N ^-1 Z _{N + 1} . Therefore, when N is sufficiently large, the equation (6) is expressed by θ ^{^} _{N + 1} = (I- _AN ^-1 Z _{N + 1} Z _{N + 1} ^T ) θ ^{^} _N + _AN ^-1 Z _{N + 1} x
It becomes _{N + 1} . Similarly, K _{N + i} , K included in equation (7)
_{N + m + 1-i} , K _{N + j} and K _{N + m} are K _{N + i} = A _{N-1 + i} ^-1 Z _{N + i} K _{N + m + 1-i} = A _{N + mi} ^-1 Z _{N + m + 1-i} K _{N + j} = A _{N-1 + j} ^-1 Z _{N + j} K _{N + m} = A _{N-1 + m} ^-1 Z _{N + m} , respectively. Therefore, when N is sufficiently large, equation (7) becomes

[0040]

[Equation 15]

It becomes The equation (8) has a smaller amount of calculation than the equation (7), and when N is sufficiently large, the parameter θ is estimated when m pieces of time series data are added to N pieces of time series data. This is a calculation formula for calculating the value θ ^{^} _{N + m} .

In the first, second, fourth, sixth and eighth inventions, the estimation of the parameter indicating the relationship between the plurality of time series data included in the first data string indicating the state of the monitored object according to the model. It corresponds to an equation in which the first estimated value, which is a value, is stored, and when a second data string is newly obtained, parameter estimation by the Kalman filter is repeated for the number of time-series data included in the second data string. Using the formula
The second estimated value, which is the estimated value of the parameter related to the third data string, is calculated from the third data string configured by adding the second data string to the first data string and the first estimated value. The state of the monitoring target is determined based on. By collectively processing a plurality of newly obtained time-series data, it becomes possible to perform parameter estimation processing sufficiently faster than the speed at which the plurality of time-series data are accumulated.

In the third, fifth and seventh inventions, the third data obtained by adding the second data string composed of m time series data to the first data string composed of N time series data. The second estimated value, which is the estimated value of the parameter indicating the relationship between the plurality of time-series data included in the sequence according to the AR model, is the first estimated value and the third data, which are the estimated values of the parameters related to the first data sequence. From the column, calculation is performed using the above-described approximate expression that holds when N is sufficiently large, and the state of the monitoring target is determined based on the second estimated value. The parameter estimation process can be performed at high speed, and the abnormality of the monitoring target can be reliably determined.

[0044]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be specifically described below with reference to the drawings showing the embodiments thereof. FIG. 1 is a block diagram showing the configuration of the condition monitoring system of the present invention. The condition monitoring system of the present invention monitors the condition of equipment in a factory,
When an abnormality occurs in equipment, it is operated to issue an alarm. The facility in the factory is provided with a sensor 31 for measuring the condition of the facility such as vibration or temperature. Sensor 31
Is connected to the data acquisition device 32 according to the present invention, and inputs measurement data to the data acquisition device 32. The data acquisition device 32 has a function of sampling the measurement data input from the sensor 31 at a predetermined cycle and storing time-series data obtained as a result of the sampling. The data acquisition device 32 is connected to a communication network N provided in the factory, the communication network N is connected to the state monitoring device 1 of the present invention, and the data acquisition device 32 is
The stored time series data is transmitted to the state monitoring device 1 via the communication network N.

The state monitoring device 1 is constructed by using a computer, and has a CPU 11 for performing calculations, a RAM 12 for storing temporary information generated by the calculations, and a CD-RO.
An external storage device 13 such as an M drive and an internal storage device 14 such as a hard disk are provided. The external storage device 13 reads the computer program 20 of the present invention from the recording medium 2 of the present invention such as a CD-ROM. The read computer program 20 is stored in the internal storage device 14, the computer program 20 is loaded into the RAM 12, and the CPU 11 executes the necessary processing for the state monitoring device 1 based on the computer program 20. The state monitoring device 1 also includes an input unit 15 connected to the communication network N in the factory, and the input unit 15 receives time-series data from the data acquisition device 32 via the communication network N. Further, the state monitoring device 1 includes an output unit 16 that outputs information to the outside, the output unit 16 is connected to the alarm device 4, and the state monitoring device 1 outputs the information indicating the abnormality of the equipment. To the alarm device 4. The alarm device 4 includes a buzzer, a lamp, a display unit that displays the content of the alarm, and the like, and notifies an abnormality of the equipment according to the information received from the state monitoring device 1.

The state monitoring device 1 may be configured so that the computer program 20 according to the present invention is downloaded from an external server device (not shown) connected to the communication network N and the CPU 11 executes the process. .

The internal storage device 14 uses the first data string composed of N time-series data acquired by the data acquisition device 32 when the equipment to be monitored is normal, and the first data string using the equation (3). The first estimated value θ ^{^} _N calculated from one data string is stored.

FIG. 2 is a flow chart showing the operation performed by the condition monitoring system of the present invention. Data acquisition device 32
Measures the measurement data input from the sensor 31 at a predetermined cycle (S1), and determines whether m pieces of time-series data acquired as a result of sampling are accumulated (S1).
2). If m pieces of time series data are not accumulated (S
2: NO), the data acquisition device 32 returns the process to step S1 and continues sampling, and when m pieces of time series data are accumulated (S2: YES), the data acquisition device 32 is composed of m pieces of time series data. The second data string is transmitted to the status monitoring device 1 via the communication network N (S3).

The state monitoring device 1 receives the second data string from the data acquisition device 32 (S4), reads the first data string stored in the internal storage device 14, and reads the second data string in the first data string. To create a third data string (S5). The state monitoring device 1 then determines the first estimated value θ ^{^} _N.
From the internal storage device 14 (S6), using the equation (8), from the third data string and the first estimated value θ ^{^} _N to the second
The estimated value θ ^{^} _{N + m} is calculated (S7).

The status monitoring apparatus 1, then a vector of a difference between the first estimated value theta ^{^} _N and the second estimated value θ ^{^} _{N + m} is a vector of p dimensions together (θ ^{^} _{N + m} -θ ^{^} _N ) is calculated (S8), and it is determined whether or not the absolute value | θ ^{^} _{N + m-} θ ^{^} _N | of the calculated difference vector is larger than a predetermined value (S).
9). When the absolute value | θ ^{^} _{N + m} −θ ^{^} _N | is larger than the predetermined value (S9: YES), the state monitoring device 1 monitors the vibration, temperature, etc. It is judged that the equipment of is abnormal, and the abnormality information indicating the abnormality of the equipment,
The output unit 16 transmits to the alarm device 4 (S10), and the process ends. According to the abnormality information received from the state monitoring device 1, the alarm device 4 notifies the abnormality of the equipment such as sounding a buzzer, turning on a lamp, or displaying the contents of the abnormality information on the display unit, and determines that there is an abnormality. Measures are taken such as manually stopping the installed equipment.

In step S9, the absolute value | θ ^{^} _{N + m-} θ ^{^} _N |
Is less than the predetermined value (S9: NO), the state monitoring device 1 determines that the equipment to be monitored is normal,
From the third data string composed of (N + m) time-series data, the latest N time-series data including the m time-series data of the second data string is selected, and a new first data string is selected. Is stored in the internal storage device 14 (S11). The state monitoring device 1 then uses the stored new first data string to
A new first estimated value θ ^{^} _N is calculated using the equation (3) (S
12), the calculated first estimated value θ ^{^} _N is stored in the internal storage device 14 (S13), and the process ends.

In the above process, the vector (θ ^{^} _{N + m-} θ ^{^} _N ) of the difference between the first estimated value θ ^{^} _N and the second estimated value θ ^{^} _{N + m} is calculated in step S8. , Absolute value in step S9 |
Although the process of judging the equipment abnormality by comparing θ ^{^} _{N + m} − θ ^{^} _N | with a predetermined value is used, the present invention is not limited to this, and the Euclidean distance (θ ^{^} _{N + m} − θ ^{^} _N ) ² may be used, and a process of determining an abnormality of equipment by comparing the calculated Euclidean distance with a predetermined value may be used, and the second estimated value θ ^{^} _{N + m}
You may use the process which determines the abnormality of an equipment by comparing the absolute value of with a predetermined value. If the equipment is normal, the same first data string is repeatedly used without updating the first data string stored in the internal storage device 14, and the processing in steps S11 to S13 described above is performed. You may use the process which omitted.

As described in detail above, in the present invention, the first estimated value θ ^{^} _N of the parameter of the model is calculated from the first data string including N time-series data indicating the state of the equipment, and m When a second data string including a plurality of time-series data is obtained, the change in the parameter due to the addition of the second data string to the first data string is expressed by an equation in which the parameter estimation using the Kalman filter is repeated m times. Using the corresponding equation, m time-series data is accumulated by calculating from the third data string configured by adding the second data string to the first data string and the first estimated value θ ^{^} _N. It is possible to calculate the change in the parameter at a sufficiently high speed as compared with the speed at which the parameter changes.
Further, by using the approximate expression derived under the condition that N is sufficiently large, the change in the parameter can be calculated in a shorter time. In this case, N must be at least 1000 or more in order to maintain the accuracy of parameter estimation, and when N is 10,000 or more, the parameter can be estimated with sufficiently high accuracy. Since the parameter change is calculated sufficiently faster than the speed at which m time-series data is accumulated, even if the state monitoring device 1 has a low computing capacity, it fails to detect the abnormality of the equipment or the abnormality. It is possible to stably monitor the state of equipment without delaying the detection of.

Further, since m time-series data are collectively processed to calculate the change in the parameter, the change in the parameter at the time of abnormality becomes large as compared with the conventional method of sequentially estimating the parameter, and the facility can be reliably operated. Changes in state can be determined.

In the present embodiment, one sensor 31
Shows the form of monitoring the state of the equipment based on the measured data, but is not limited to this, one equipment,
Alternatively, each of the plurality of facilities is provided with a plurality of sensors 31, each of the plurality of sensors 31 is connected to the state monitoring device 1 via the data acquisition device 32 and the communication network N, and each data of each sensor 31 is measured. The state of one or a plurality of facilities may be monitored based on the above. In addition, in the present embodiment, the data acquisition device 32 stores the time-series data, but the data acquisition device 3
2 is the status monitoring device 1 without accumulating the acquired time series data
Alternatively, the time series data received by the status monitoring device 1 may be accumulated, and the abnormality determination may be performed when m pieces of time series data are accumulated. Further, in the present embodiment, the form in which the alarm device 4 is connected to the output unit 16 of the condition monitoring device 1 is shown. However, when the condition monitoring device 1 determines that the equipment is abnormal, the equipment is abnormal. It is also possible to connect a control device that controls the equipment such as stopping the power supply.

Further, in the present embodiment, the mode in which the equipment in the factory is monitored is shown, but the mode is not limited to this, and the mode in which the state of other general apparatus is monitored may be used. Alternatively, it may be in a form of monitoring a phenomenon such as fluctuation of sales data or stock price.

[0057]

According to the first, second, fourth, sixth and eighth inventions, a plurality of newly obtained time series data are collectively processed to accumulate a plurality of the time series data. It is possible to perform parameter estimation processing at a speed sufficiently higher than the speed.

In the third, fifth and seventh inventions, when m time-series data is newly added to N time-series data, N is sufficient as the estimated value of the model parameter. By using an approximate expression that holds when it is large, the parameter estimation process can be performed at high speed, and stable detection is possible without failing to detect facility abnormalities or delaying detection of abnormalities. The condition of the equipment can be monitored. Further, the present invention has an excellent effect that the change of the parameter at the time of abnormality becomes large as compared with the conventional method of sequentially estimating the parameter, and the change of the state of the equipment can be reliably judged.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a condition monitoring system of the present invention.

FIG. 2 is a flowchart showing an operation performed by the condition monitoring system of the present invention.

[Explanation of symbols]

1 state monitoring device 2 recording medium 20 computer program 31 sensor 32 data acquisition device 4 alarm device θ ^{^} _N first estimated value θ ^{^} _{N + m} second estimated value

Claims (8)

[Claims] 1. The monitoring using a computer having a receiving unit for receiving information from the outside, a storage unit, and a computing unit according to a model that theoreticalizes the relationship between a plurality of time-series data indicating the state of a monitoring target. In a method of monitoring a state of a target, a first data string composed of a plurality of time series data indicating a state of the monitoring target is stored in a storage unit, and a plurality of time series data included in the first data string according to the model. A first estimated value, which is an estimated value of a parameter indicating the relationship between the two, is stored in the storage unit, and a second data string including a plurality of other time-series data indicating the state of the monitoring target is externally received from the reception unit. , The plurality of time-series data included in the second data string are sequentially received in the first
Using a Kalman filter to estimate the parameters associated with the data string that is configured each time the data string is added to the first data string, the parameter corresponding to the number of timeseries data included in the second data string is repeated. A second estimated value of a parameter indicating a relationship between a plurality of time series data included in the third data string according to the model, from a third data string configured by adding the second data string and the first estimated value. A state monitoring method, wherein an estimated value is calculated by a calculation unit, and the state of the monitoring target is determined by the calculation unit based on the calculated second estimated value. 2. A data acquisition device that acquires a plurality of time-series data indicating a state of a monitoring target, and a model of the monitoring target according to a model that theoreticalizes the relationship between the plurality of time-series data acquired by the data acquisition device. In a state monitoring system including a state monitoring device for monitoring a state, the state monitoring device stores a first data string composed of a plurality of time series data indicating the state of the monitoring target, and according to the model. Means for storing a first estimated value, which is an estimated value of a parameter indicating a relationship between a plurality of time series data included in the first data string, wherein the data acquisition device A means for inputting a configured second data string to the status monitoring device, wherein the status monitoring device is further provided with a plurality of time series data included in the input second data string. Is sequentially added to the first data sequence, the parameter estimation for the data sequence configured is repeated by using a Kalman filter for the number of time-series data included in the second data sequence. An estimated value of a parameter indicating a relationship between a plurality of time-series data included in the third data string according to the model, from a third data string configured by adding the second data string to one data string and the first estimated value. And a means for determining the state of the monitored object based on the calculated second estimated value. 3. A data acquisition device that acquires a plurality of time-series data indicating the status of a monitoring target, and a model of the monitoring target according to a model that theoreticalizes the relationship between the plurality of time-series data acquired by the data acquisition device. In a status monitoring system including a status monitoring device that monitors a status, the status monitoring device is configured to provide N time-series data {x _n :
means for storing a first data string composed of n = 1, 2, ..., N}, and a relationship between a plurality of time series data included in the first data string according to the model that is a p-order AR model. Means for storing a first estimated value θ ^{^} _N , which represents an estimated value of a parameter made up of p coefficients shown in a matrix of p rows and 1 column, and the data acquisition device acquires m time series. Means for inputting a second data string composed of data {x _n : n = N + 1, N + 2, ..., N + m} to the status monitoring device, wherein the status monitoring device is further provided with the input second data string. Using a third data string configured by adding the first data string and the first estimated value θ ^{^} _N , a parameter indicating a relationship between a plurality of time series data included in the third data string according to the model. Second estimated value θ ^{^} _N that represents the estimated value of the _{+ m} is given by And a means for determining the state of the monitoring target based on the calculated second estimated value θ ^{^} _{N + m} . 4. An apparatus for monitoring the status of a monitoring target according to a model that theoreticalizes the relationship between a plurality of time series data indicating the status of a monitoring target, from a plurality of time series data indicating the status of the monitoring target. A unit configured to store a first data string configured, a unit configured to store a first estimated value that is an estimated value of a parameter indicating a relationship between a plurality of time series data included in the first data sequence according to the model, A means for externally receiving a second data sequence composed of a plurality of other time-series data indicating the state of the monitoring target, and a plurality of time-series data included in the second data sequence in order of the first
Using a Kalman filter to estimate the parameters associated with the data string that is configured each time the data string is added to the first data string, the parameter corresponding to the number of timeseries data included in the second data string is repeated. A second estimated value of a parameter indicating a relationship between a plurality of time series data included in the third data string according to the model, from a third data string configured by adding the second data string and the first estimated value. A state monitoring device comprising: a means for calculating an estimated value; and a means for determining the state of the monitoring target based on the calculated second estimated value. 5. A plurality of time-series data indicating the status of the monitoring target
According to a model that theorizes the relationship between
In the device that monitors the condition of the elephant, N time-series data indicating the status of the monitoring target {x_n : N =
Store a first data string composed of 1, 2, ..., N}
Means, First data according to the model that is a p-order AR model
P indicating the relationship between multiple time series data included in a column
The estimated value of the parameter consisting of the coefficient is a p-by-1 matrix
First estimated value θ^{^} _NMeans for storing The other m time-series data {x
_n : N = N + 1, N + 2, ..., N + m}
Means for externally receiving the second data string, A third data structure formed by adding the second data string to the first data string.
Data string and first estimated value θ^{^} _NAccording to the model
Is the relationship between the multiple time series data included in the third data string.
The estimated parameter values are expressed in a p-by-1 matrix.
Second estimated value θ ^{^} _{N + m}To [Equation 2] And means to calculate Calculated second estimated value θ^{^} _{N + m}Based on the condition of the monitored object
State monitoring, characterized by comprising means for determining the state
apparatus. 6. A first data string composed of a plurality of time-series data indicating a state of a monitoring target, and the model according to a model that theoreticalizes a relationship between a plurality of time-series data indicating a state of a monitoring target. A computer program storing a first estimated value, which is an estimated value of a parameter indicating a relationship between a plurality of time series data included in the first data string, according to A second data string is included in the second data string by using a Kalman filter for estimation of the parameter related to the data string that is configured each time the computer sequentially adds a plurality of time series data included in the second data string to the first data string. Using an equation equivalent to the number of repeated time series data,
Estimation of a parameter indicating a relationship between a plurality of time-series data included in the third data string according to the model, from the third data string configured by adding the second data string to the first data string and the first estimated value. A computer program comprising: a procedure for calculating a second estimated value that is a value; and a procedure for causing a computer to determine the state of the monitoring target based on the calculated second estimated value. 7. N time-series data {x _n : n = 1, 2, 1 indicating the state of the monitoring target, according to a model that theoreticalizes the relationship between a plurality of time-series data indicating the state of the monitoring target.
, N} and a parameter consisting of p coefficients showing the relationship between a plurality of time-series data included in the first data string according to the model that is a p-order AR model. A computer program for storing a first estimated value θ ^{^} _N , which represents an estimated value in a matrix of p rows and 1 column, for monitoring the state of the monitoring target. A third data string configured by adding a second data string composed of time-series data {x _n : n = N + 1, N + 2, ..., N + m} to the first data string, and a first estimated value θ ^{^} _N Using the above, the second estimated value θ ^{^} _{N + m} , which represents the estimated value of the parameter indicating the relationship between the plurality of time-series data included in the third data string according to the model in the matrix of p rows and 1 column, Number 3] And a step of causing a computer to determine the state of the monitoring target based on the calculated second estimated value θ ^{^} _{N + m} . 8. A first data string composed of a plurality of time-series data indicating a state of a monitoring target, and the model according to a model that theoreticalizes a relationship between a plurality of time-series data indicating a state of a monitoring target. According to the above, a computer program storing a computer program for monitoring the state of the monitoring target is stored in a computer that stores a first estimated value that is an estimated value of a parameter that indicates a relationship between a plurality of time series data included in the first data string. A recording medium readable by a certain computer, wherein the computer is provided with a plurality of time-series data included in a second data sequence received from the outside, the data sequence being configured to be added to the first data sequence in order. A parameter corresponding to the number of time-series data included in the second data string is used for parameter estimation using the Kalman filter, and the first data is calculated. Data from the third data sequence formed by adding the second data sequence to the data sequence and the first estimated value, the estimated value of the parameter indicating the relationship between the plurality of time series data included in the third data sequence according to the model. A computer program is recorded, which includes a procedure for calculating a certain second estimated value and a procedure for causing the computer to determine the state of the monitoring target based on the calculated second estimated value. A readable recording medium.