JP2023105972A - FAILURE SIGNS MONITORING SYSTEM AND FAILURE SIGNS MONITORING METHOD - Google Patents

Abstract

Kind Code: A1 A failure sign monitoring system and a failure sign monitoring method are provided that make it possible to prevent detection omissions and erroneous detections when monitoring signs of plant failure using a monitoring model created by an automatic update function. A monitoring model is used for a plant equipped with A1-Am sensors, . A plant monitoring system to be updated wherein, before updating the monitoring model with each sensor value for the current period, each sensor value for the current period is updated relative to each sensor value for a period prior to the current period. The variability is examined, and if the variability is large relative to a preset range, the monitoring model is modified using sensor values obtained by removing the highly variable sensor values or using sensor values obtained from another period of time. A management server 43 to be updated is provided. [Selection drawing] Fig. 1

Description

The present invention relates to a failure sign monitoring system and a failure sign monitoring method for monitoring signs before a failure occurs, and more particularly to a failure sign monitoring system and a failure sign monitoring method for monitoring various plants.

Various types of equipment such as steam turbines and boilers are used in power plants of power plants, and the power plants generate power using these equipment. In such a power plant, installed devices and the like are monitored for signs of failure. (See Patent Document 1, for example). According to the monitoring system described in this document, System Invariant Analysis Technology (SIAT) is used to monitor signs of failure. The invariant analysis technology utilizes a network fault handling engine. The network failure handling engine monitors the entire system by detecting a break in the relationship between observation points.

These invariant analysis techniques are applied for plant monitoring. In other words, there is a strong correlation between the measured values of the sensors in the plant (hereinafter referred to as "sensor values"). Conventional monitoring systems automatically analyze sensor values collected from observation points, comprehensively find correlations (invariants) that are invariant relationships between sensor values between two points, model them, and monitor devices, etc. Create a correlation model (hereinafter referred to as a “monitoring model”) for After that, the monitoring system examines the correlation between the sensors installed in the power plant, for example, the piping, and the sensors installed in the steam turbine. Then, it is possible to detect abnormalities in the power plant by the monitoring system from the "unusual" behavior, and it is possible to monitor the entire plant in real time.

Japanese Patent Application Laid-Open No. 2017-021702

By the way, the conventional monitoring system described above has the following problems. In other words, this monitoring system has an automatic update function that automatically updates the monitoring model on a regular basis. However, when using the automatic update function, there are concerns about "missed detection" and "false positives". For example, there is a detection omission (example 1) as shown in FIG. It is possible to detect "trouble I" when the monitoring system evaluates the period of "3rd week of month xx, year 20xx" with the monitoring model based on the sensor values of "2nd week of month xx, year 20xx". . In FIG. 6, the sensor value is represented by two upper and lower curves. The lower curve g1 shows the sensor value in a normal state without trouble, and the upper curve g2 shows the sensor value when a trouble occurs. It shows the sensor value when

However, when the monitoring system evaluates the period of "4th week of xx month, 20xx" with the monitoring model based on the sensor values of "20xx year, xx month, 3rd week", "Trouble I" in the 3rd week Since sensor value fluctuations are learned as normal, there is a possibility that "Trouble II" involving similar sensor value fluctuations cannot be detected.

Also, there is an erroneous detection (example 2) as shown in FIG. For example, even if the sensor value from any sensor is constant due to maintenance in "20xx year xx month 3rd week", the monitoring system creates a model and evaluates "20xx month xx month 4th week" If so, it learns that the sensor value from the sensor is a constant value as normal behavior. Therefore, there is a possibility that the monitoring system will detect a state change in the "fourth week of month xx, year 20xx" even if there is nothing.

An object of the present invention is to solve the above-mentioned problems, and to prevent omissions and false detections when monitoring signs of plant failure using a monitoring model created by an automatic update function. An object of the present invention is to provide a system and a failure sign monitoring method.

In order to solve the above-mentioned problems, the invention of claim 1 is used for a plant equipped with various types of equipment and with various types of sensors for measuring the states of these equipments, and is based on the correlation between the sensors. A plant monitoring system for monitoring a plant using a monitoring model created in and periodically updating the monitoring model, wherein before updating the monitoring model with each sensor value for the current period, the Examine the variation of each sensor value in the current period relative to each sensor value in the period before the current period. and processing means for updating the monitoring model using one of each sensor value obtained from the period and each sensor value obtained from another period.

In claim 1, before updating the monitoring model with each sensor value of the current period, the processing means is configured to update each sensor value of the current period with reference to each sensor value of the period prior to the current period. Examine variability. If the variation is larger than the preset range, the processing means uses one of the sensor values obtained by removing the sensor values with large variation and the sensor values obtained from another period. to update the monitoring model.

According to a second aspect of the present invention, in the failure sign monitoring system according to the first aspect, the processing means calculates a variance from each sensor value in the current period, and compares the calculated variance with a preset reference value. It is characterized by judging from the results whether the variation in the sensor values in the current period is large.

The invention of claim 3 is the failure sign monitoring system of claim 1 or 2, wherein the processing means calculates the variance from each sensor value in the current period and calculates the variance from each sensor value in the period prior to the current period. ), calculate the variance ratio, which is the ratio of the two calculated variances, and compare the preset reference value with the calculated variance ratio to determine whether the variation in the sensor values in the current period is large. It is characterized by judging.

The invention of claim 4 is used for a plant comprising various types of equipment and various sensors for measuring the state of these equipments, and using a monitoring model created based on the correlation between the sensors A predictive failure monitoring method for monitoring a plant and periodically updating the monitoring model, wherein before updating the monitoring model with each sensor value for the current period, Using each sensor value as a reference, the processing means examines the variation in each sensor value in the current period, and if the variation is greater than the preset range, each sensor obtained by removing the sensor value with a large variation and updating the monitoring model by the processing means using one of the sensor values obtained from another period.

According to the invention of claim 1, when the monitoring model is periodically updated, if the variation in the sensor values in the current period is large, each sensor value obtained by removing the sensor value with the large variation or another A monitoring model is created from each sensor value obtained from the period. This eliminates the need to use a monitoring model created using sensor values with large variations when monitoring signs of failure of equipment installed in the plant in the next period using the updated monitoring model. Therefore, it is possible to prevent omissions and erroneous detection of failure signs, thereby preventing a decrease in monitoring accuracy.

According to the invention of claim 2, by calculating the variance of the sensor values in the current period, it becomes possible to check whether the variation in the sensor values in the current period is large.

According to the invention of claim 3, by calculating the variance ratio between the variance of the sensor values in the previous period and the variance of the sensor values in the current period, whether the variation in the sensor values in the current period is large, It becomes possible to investigate by a simple calculation method.

According to the invention of claim 4, similarly to claim 1, when the monitoring model is periodically updated, if the variation in sensor values in the current period is large, sensor values with large variation are removed. A monitoring model is created from each sensor value obtained from each time period or from another time period. This eliminates the need to use a monitoring model created using sensor values with large variations when monitoring signs of failure of equipment installed in the plant in the next period using the updated monitoring model. Therefore, it is possible to prevent omissions and erroneous detection of failure signs, thereby preventing a decrease in monitoring accuracy.

1 is a configuration diagram showing a plant monitoring system according to Embodiment 1 of the present invention; FIG. It is a figure which shows an example of sensor data. It is a figure which shows an example of a time-series sensor value. It is a flow chart which shows an example of judgment processing. FIG. 4 is a diagram showing part of calculation for calculating variance from sensor values; It is an explanatory view explaining an example of omission of detection. It is an explanatory view explaining an example of erroneous detection.

Next, each embodiment of the present invention will be described in detail with reference to the drawings.

(Embodiment 1)
A plant monitoring system according to this embodiment is shown in FIG. The plant monitoring system of FIG. 1 is used in a power plant in which a power plant 10 is installed, and mainly includes a sensor monitoring device 20 and a model monitoring device 40 . The sensor monitoring device 20 is connected to the model monitoring device 40 via an in-house communication network 30 such as an intranet so as to enable data communication with the model monitoring device 40 .

The power plant 10 generates power using nuclear power or thermal power, and in this embodiment, a case of generating power using nuclear power is taken as an example. Although not shown, the power plant 10 uses a large number of devices such as a nuclear reactor, a turbine, a generator, pumps, and piping. The power plant 10 generates power using these devices. The power plant 10 also has various sensors (not shown) for checking the status of each device, such as A ₁ sensor to _Am sensor . . .
B ₁ sensor to B _n sensor are installed. Each sensor has sensor values a ₁ to a _m
・・・
_b1 to _bn
to the sensor monitor 20 .

Sensor monitor ₂₀ receives sensor values a _{1 -a m} , . When _the sensor monitoring device 20 receives the sensor values a ₁ to a _m , _. In the sensor data, the installation point where the sensor is installed is recorded in correspondence with the sensor identification information for representing the sensor and identifying the sensor. In addition, the sensor data includes sensor values, measurement date and time, and the like corresponding to the sensor identification information. In FIG. 2, the sensor values a ₁ to a _m are those of the A ₁ sensor to A _m sensors installed in the condenser of the power plant 10, and the sensor values b ₁ to b _n are those of the feed water pump of the power plant 10. A case is taken as an example of the B ₁ sensor to B _n sensor installed in the system. The configuration will be described below based on this example.

The sensor monitoring device 20 transmits the created sensor data to the model monitoring device 40 via the in-house communication network 30 each time the transmission time elapses. The transmission time is set in the sensor monitoring device 20 as required, such as in seconds, minutes, hours, days, or the like.

The model monitoring device 40 receives sensor data from the sensor monitoring device 20 via the in-house communication network 30 . Model monitoring device 40 monitors power plant 10 using the received sensor data. For this purpose, the model monitoring device 40 includes a communication control section 41, a data server 42, a management server 43, and clients 44 ₁ to 44 _k . The communication control unit 41, the data server 42, the management server 43, and the clients 44 ₁ to 44 _k of the model monitoring device 40 are connected by a LAN (Local Area Network) or the like so as to be able to transmit and receive data.

The communication control unit 41 performs communication control for connecting the data server 42, the management server 43, and the clients 44 ₁ to 44 _k to the in-house communication network 30. FIG. For example, when receiving sensor data from the sensor monitoring device 20 via the in-house communication network 30 , the communication control unit 41 sends the sensor data to the data server 42 .

Data server 42 is a storage device that stores data related to power plant 10 . For example, when the data server 42 receives sensor data from the sensor monitoring device 20 via the in-house communication network 30 and the communication control unit 41, the data server 42 stores the sensor data. Also, upon receiving a data transmission request from the management server 43 , the data server 42 extracts the corresponding sensor data and sends it to the management server 43 .

The clients 44 ₁ to 44 _k are computers operated by a person in charge of operating the power plant 10 and are for operating the power plant 10 . Various instructions required for the operation of the power plant 10 are input to the clients 44 ₁ to 44 _k by the person in charge. For example, a model creation instruction is input to the clients 44 ₁ to 44 _k in order to create each monitored model of the power plant 10 as required. When the client 44 ₁ to 44 _k inputs the model creation instruction, it sends the model creation instruction to the management server 43 .

Further, when the clients 44 ₁ to 44 _k receive various data from the management server 43, for example, alarm data to be described later, they output an alarm. Furthermore, when the clients 44 ₁ to 44 _k receive the sensor list data from the management server 43, they display this data.

The management server 43 is a computer that performs various processes for monitoring the power plant 10 and each device. First, the management server 43 creates a monitoring model of the power plant 10 . Next, the management server 43 monitors the power plant 10 and equipment based on the created monitoring model. Below, creation of a monitoring model of the power plant 10 and monitoring of the power plant 10 and the like by the monitoring model will be described in order.

The management server 43 creates a monitoring model during normal operation in order to monitor the power plant 10 . The monitoring model is obtained by deriving the correlation between sensors by invariant analysis technology based on the time-series data of each sensor attached to the power plant 10 . The monitoring model during normal operation is created by the management server 43 based on the sensor values when the power plant 10 was operating normally. That is, since the state is stable during the constant temperature operation of the power plant 10, the distribution of the sensor values is also stable. The management server 43 creates a monitoring model on the premise of such a state.

Furthermore, the management server 43 automatically updates the monitoring model every predetermined period, for example, every week in this embodiment. Note that the predetermined period for automatically updating the monitoring model is arbitrary, such as several days or several weeks. The management server 43 sends a data transmission request to the data server 42 at preset timing. The management server 43 also sends a data transmission request to the data server 42 when it receives a model creation instruction manually input by the person in charge from the clients 44 ₁ to 44 _k .

After that, the management server 43 receives each sensor data from the data server 42 . Each sensor data received by the management server 43 is data during normal operation of the power plant 10, and is, for example, as shown in FIG. Then, the management server 43 creates a time-series sensor value list based on the sensor values obtained from each sensor data. An example of this time series sensor value list is shown in FIG. FIG. 3 shows a list of sensor values separated by one week as each predetermined period, and for the sensor value a ₁ of the A ₁ sensor, the time-series sensor values a ₁ (p ₁ ), . . . _a1 ( _pt )
_a1 ( _q1 ), ..., _a1 ( _qt )
_a1 ( _r1 ), ..., _a1 ( _rt )
a ₁ (s ₁ ), . . . , a ₁ (s _t )
is shown. In other words, the time-series sensor values for the first week of "month 20xx" are
a ₁ (p ₁ ), . . . , a ₁ (p _t )
and the time-series sensor values for the second week are
_a1 ( _q1 ), ..., _a1 ( _qt )
is. In addition, the time-series sensor values for the third week are
_a1 ( _r1 ), ..., _a1 ( _rt )
and the time-series sensor values for the 4th week are
a ₁ (s ₁ ), . . . , a ₁ (s _t )
is. The time-series sensor value a ₁ (s _t ) for the fourth week is the last value for this week.

When the management server 43 completes creating such a time-series list of sensor values, conventionally, the monitoring model to be used in the next week is replaced with the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ). In other words, conventionally, the fourth week, which is the current period, is used as the model learning period in order to create a monitoring model.

On the other hand, in this embodiment, the following is done. in short. When the management server 43 finishes creating the time-series sensor value list, the management server 43 performs determination processing. An example of this determination process is shown in FIG. When the management server 43 receives the last sensor value a ₁ (s _t ) of this week, which is the current model learning period (fourth week), it starts determination processing before updating the monitoring model. _When starting the determination process, _the management server 43 calculates the variance σ _i of the sensor values a ₁ (s ₁ ), . (step S1). That is _, in step S1, the management server ₄₃ determines that the distribution of the sensor values a ₁ (s ₁ ), . find out if there are Therefore, the management server 43 calculates the average value a _1av of the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ) using the following formula.

Next, as shown in FIG. 5, the management server 43 calculates the difference between the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ) and the average value a _1av . Thereafter, the management server 43 calculates the square of the difference between the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ) and the average value a _1av . Next, the management server 43 adds the value of the square of the difference from the average value a _1av . Furthermore, the management server 43 divides the calculated added value by the value t to calculate the variance σ _i for this week (fourth week), which is the model learning period.

The value t here is the data number t of the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ).

After step S1 is finished, the management server 43 similarly calculates the variance σ _i-1 of the third week, which is the previous period of the current model learning period, as the sensor values a ₁ (r ₁ ), . . . Calculate using a ₁ (r _t ) (step S2).

After step S2 ends, the management server 43 performs F-test on the distribution of sensor values (step S3). That is, the management server 43 calculates the variance ratio F _i that is the ratio of the variance σ _i in the model learning period (fourth week) and the variance σ _i−1 in the previous period (third week). The variance ratio F _i is a value for determining whether the model learning period (fourth week) and the previous period (third week) have equal variances. for example,

, the management server 43

Calculate the variance ratio F _i by and conversely,

, the management server 43

Calculate the variance ratio F _i by

The management server 43 holds a predetermined criterion value _Fs for performing the F-test. The judgment reference value F _s is a value that serves as a judgment reference for the F-test, and is calculated based on two periods when the power plant is normal, for example, the variance in the first week and the variance in the second week. Also, it is a value input to the management server 43 by the person in charge operating the client 44 ₁ , that is, a judgment reference value of the dispersion ratio F _i set according to the characteristics of the sensor of the power plant.

The management server 43, regarding the variance ratio F _i calculated as above,

If the relationship holds, it is determined in step S3 that the variance _σi of this week (fourth week), which is the model learning period, and the variance _σi-1 of the previous week (third week), which is the previous period, are the same. do. That is, the management server 43 calculates the sensor values a ₁ (s ₁ ), . . . , a ₁ (s _t ) during the model learning period
and sensor values a ₁ (r ₁ ), . . . , a ₁ (r _t ) in the previous period
is the same as the variation of

After the determination in step S3, the management server ₄₃ creates an optimum monitoring _model based on the sensor values _a1 ( _s1 ), . S4).

In step S4, the management server 43 creates _a monitoring _model using the sensor values a ₁ (s ₁ ), . In other words, the management server 43 examines the strength of the correlation between the sensors and creates a monitoring model using the correlated sensors. At this time, the management server 43 prepares a monitoring model that focuses on the strength of the correlation between sensors according to the state of the plant (during inspection, starting, operating at rated thermal output, during periodic testing, and stopping). create it.

On the other hand, the management server 43, regarding the variance ratio F _i calculated in step S3,

does not hold, for example, the clients 44 ₁ to 44 _k are notified of an abnormality in the sensor value (step S5). In step S5, the management server 43 creates alarm data indicating that the sensor value distribution is not stable as an abnormality, and sends this alarm data to the clients 44 ₁ to 44 _k .

After step S5, the management server 43 modifies or changes the model learning period as necessary by changing the model learning period or excluding from the model learning period sensor values larger than the judgment reference value set by the variance ratio _Fi . and create a monitoring model based on the model learning period after correction/change (step S6). For example, when the sensor value fluctuates during the model learning period due to trouble as shown in FIG. 6 above, or when the sensor value becomes constant during the model learning period due to maintenance as shown in FIG. is detected as a variation in the variance ratio F _i . As a result, the management server 43 corrects/changes the model learning period. Therefore, in step S6, when the monitoring model is updated, the management server 43 updates the optimum monitoring model equivalent to when the sensor value distribution is stable. can be created.

When step S4 or step S6 ends, the management server 43 ends the determination process. Then, the management server 43 uses the next week's optimum monitoring model to perform real-time monitoring or monitoring by the core invariant method.

The above is the configuration of the failure sign monitoring system according to this embodiment. Next, a failure sign monitoring method by this failure sign monitoring system will be described.

Generally, the model monitoring device 40 monitors the state of the power plant 10 using a monitoring model. The management server 43 of the model monitoring device 40 automatically updates the monitoring model every model learning period, which is a predetermined period, for example, every week. At this time, the management server 43 determines whether or not the sensor values for this period, that is, the fourth week, are different from the sensor values for the third week by the determination process shown in FIG. .

If there is no variation in the sensor values for the fourth week compared to the sensor values for the third week, the management server 43 creates a monitoring model based on the sensor values for the fourth week and updates the monitoring model.

If the sensor values of the fourth week are different from the sensor values of the third week, the management server 43 corrects/changes the model learning period, and the sensor values of the model learning period after correction/change to create and update a monitoring model. At the same time, the management server 43 creates alarm data and sends it to the clients 44 ₁ to 44 _k . When the clients 44 ₁ to 44 _k receive the alarm data, they inform the person in charge that the distribution of the sensor values during the model learning period is not stable.

Then, in the next week, the management server 43 of the model monitoring device 40 uses the updated monitoring model to perform real-time monitoring or monitoring of the power plant 10 by the core invariant method.

Thus, according to this embodiment, a function of performing a test (F-test) based on a variance ratio is added to the sensor value distribution. That is, the sensor value distribution is filtered by a test (F-test) based on the variance ratio. As a result, when the distribution of sensor values is not stable, the model learning period is changed or the sensor values larger than the judgment reference value set by the variance ratio _Fi are excluded from the model learning period, so the monitoring model is automatically updated. Sometimes, it makes it possible to prevent "detection omission" and "false positive". As a result, deterioration in monitoring accuracy can be prevented.

(Embodiment 2)
In this embodiment, the variation of each sensor value for the current period is determined as follows. In addition, in this embodiment, the same reference numerals are given to the same or the same constituent elements as those of the first embodiment described above, and the description thereof will be omitted.

In this embodiment, the judgment process (FIG. 4) is simplified. That is, the management server 43 calculates the variance σ _i of this week (fourth week), which is the model learning period, as the sensor value a ₁ (s ₁ ), . . , a ₁ (s _t ).

The management server 43 stores in advance a variance reference value that is calculated from each variance at the time of past automatic update and represents the normal state of sensor values. Then, the management server 43 compares this variance reference value with the variance σ _{i of} this week (fourth week), and finds the sensor values _{a 1 (} s ₁ ), . Determine if the distribution of is stable.

Thereafter, the management server 43 performs step S4 or steps S5 and S6 similar to the judgment processing (FIG. 4).

In this way, according to this embodiment, by adding processing for simplified judgment by dispersion to the sensor value distribution, "detection omission" and "false detection" can be prevented at the time of automatic update of the monitoring model. make it possible to prevent As a result, deterioration in monitoring accuracy can be prevented.

10 power plant 20 sensor monitoring device 40 model monitoring device 41 communication control unit 42 data server 43 management server (processing means)
44 ₁ to 44 _k clients

Claims (4)

A monitoring model that is used for a plant that is equipped with various types of equipment and has various types of sensors that measure the state of these equipments, and monitors the plant using a monitoring model that is created based on the correlation between the sensors. A plant monitoring system for periodically updating a model, comprising:
Before updating the monitoring model with each sensor value for the current period, determine the variation in each sensor value for the current period relative to each sensor value for the period prior to the current period, and determine if this variation is updating the monitoring model using one of each sensor value obtained by removing sensor values that vary widely when compared to a preset range, and each sensor value obtained from another time period; processing means for
A failure sign monitoring system comprising: The processing means calculates the variance from each sensor value in the current period, and determines whether the variation in the sensor value in the current period is large based on the comparison result between the preset reference value and the calculated variance.
The failure sign monitoring system according to claim 1, characterized in that: The processing means calculates the variance from each sensor value in the current period, calculates the variance from each sensor value in the period before the current period, calculates the variance ratio that is the ratio of the two calculated variances, Judging whether the variation of the sensor value in the current period is large from the comparison result between the preset reference value and the calculated variance ratio,
3. The failure sign monitoring system according to claim 1 or 2, characterized in that: A monitoring model that is used for a plant that is equipped with various types of equipment and has various types of sensors that measure the state of these equipments, and monitors the plant using a monitoring model that is created based on the correlation between the sensors. A failure sign monitoring method for periodically updating a model,
before updating the monitoring model with each sensor value for the current period, the processing means examines the variation of each sensor value for the current period relative to each sensor value for the period prior to the current period;
If the variation is larger than the preset range, the monitoring is performed using one of the sensor values obtained by removing the sensor values with large variation and the sensor values obtained from another period. updating the model by the processing means;
A failure sign monitoring method characterized by: