JP3161844B2 - Plant abnormality detection method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting an abnormality in a plant, and more particularly to an abnormality detection method and an apparatus suitable for accurately detecting an abnormality in a plant.

[0002]

2. Description of the Related Art In various plants such as a nuclear power plant, various sensors such as a TV camera, an infrared camera, and a microphone are arranged in various parts of the plant or mounted on a mobile monitoring robot to detect a sign of a plant abnormality in advance. There is a demand for the development of a system that does this.

For example, JP-A-57-208495,
JP-A-58-184585, JP-A-60-105
In the prior art described in Japanese Patent Application Publication No. 994, the state of the sound of the plant is detected, and this sound is compared with pre-registered sound data when the plant is normal to detect an abnormality of the plant. However, the normal sound data registered in advance is detection data in one operation state of the plant or reference data having a predetermined bandwidth including the operation state of several plants. For this reason, there has been a problem that every time the operating state of the plant changes, it is detected as an abnormality, and because the bandwidth in the normal range is wide, the accuracy of the abnormality detection is not improved.

Therefore, in the prior art described in Japanese Patent Application Laid-Open No. 1-270623, the data of the normal sound to be compared is updated according to the operating state of the plant, and the sound captured by the sensor is used for the plant operation at that time. Compared with the normal sound corresponding to the state, the detection accuracy is improved.

In the prior art described in Japanese Patent Application Laid-Open No. 58-124995, if there is no reference value to be compared with the sound captured by the sensor, a new reference value is created by learning and used for plant diagnosis. I have to.

[0006]

The prior art described above is
Each of them has a configuration as shown in FIG. That is, the sensor 10 of the abnormality detection system is attached to the plant 1 of the nuclear power plant.
In this configuration, many microphones are set in a containment vessel in a nuclear power plant as 0, and a detection signal of the microphone 100 is taken into an abnormality detection device 200 installed outside the containment vessel to perform abnormality detection. Naturally, the data of the acoustic spectrum during normal operation of the plant is stored in the abnormality detection device 200, and the data is input to the abnormality detection device 200 in advance offline.

[0007] Normal sound data input off-line cannot be said to be normal sound data accurately corresponding to the plant operating state during the actual operation of the plant. Is anxious. For example, in the prior art described in Japanese Patent Application Laid-Open No. 1-270623, a sensor for abnormality detection and a sensor for plant operation state detection are used in common. Higher precision in production cannot be expected.

An object of the present invention is to provide a plant abnormality detection method and apparatus capable of always detecting a plant abnormality with high accuracy regardless of the operation state of the plant.

[0009]

The object of the present invention is to provide a plant type
State is detected by a sensor attached to the mobile robot, and the
Data that shows the plant status when the plant is normal
The detection signal of the sensor is stored in advance in the
When judging plant abnormalities by comparing
Turn on the plant operation state separately from the detection signal of the sensor
Plant operation taken online and taken online
From the sensor in the running state and the plant operating state
The position and shape of the mobile robot
To the storage device as normal data along with robot information such as
The sensor signal is compared with the normal data.
The same operating condition and the same robot
Extracting normal data of information from the storage device
This is achieved by comparing with a detection signal. Also, the device configuration
As a plant operation separately from the detection signal of the sensor
A means to capture the status online,
The plant operating state and the plant operating state
The sensor detection signal captured from the sensor is
Normal data along with robot information such as bot position and posture
Means for storing in a storage device as
Same operation when judging abnormality by comparing with normal data
Status, normal data of the same robot information stored in the storage device
And a means for taking out the sensor detection signal and comparing with the sensor detection signal.
This is achieved by tightening .

[0010] The above object is also aimed at the sound, temperature and vibration of the plant.
When the plant is normal
The sensor signal is stored as reference data
Abnormalities in the plant by comparing the signal with the reference data
To determine plant operation status online.
And the detection signal from the sensor
Characteristic parameters showing the aging of the plant in the virtual model
Meters are supplemented as needed based on the operation data when the plant is normal.
The virtual model of the plant and the temporary
The plant operating state and sensor output from the virtual model,
Import the result obtained from the virtual model and the online
Actual plant operating conditions or the actual
Detecting plant abnormalities by comparing detection signals
Is achieved in. As the equipment configuration, the plant operation status
Capture online and capture the detection signal of the sensor.
Of the plant and the virtual model of the plant
Operation during normal plant operation with characteristic parameters indicating aging
Means to make corrections as needed based on data and virtual plant
Plant operation status from the model and the virtual model of the sensor
Means for obtaining the state and the sensor output, and obtaining from the virtual model.
Results and actual plant captured online
Compare the operating state or the actual detection signal of the sensor
Means for detecting plant abnormalities
It is a good thing.

[0011]

[Function] The abnormality detection device is a mobile robot by taking the operation status of the plant into the abnormality detection device online.
It is possible to recognize in real time what kind of operating state of the plant the sound, temperature, vibration and the like of the plant currently detected by the sensor attached to the plant are. For this reason, when state data such as sound, temperature, and vibration when the plant is normal is acquired from the sensor and stored in the storage device, the state data can be stored corresponding to the operating state of the plant at that time. Furthermore, when a plant abnormality is detected from state data such as the currently detected plant sound, temperature, and vibration, the plant information stored when the operating state is the same as the current operating state of the plant that enters online Since the data can be compared with the data in the normal state, the abnormality of the plant can always be detected with high accuracy regardless of the operation state of the plant.

In addition, when the operation state of the plant and the detection signal of the sensor are taken on-line into the abnormality detection device and the output of the virtual model is compared, even if the operation of the plant changes transiently, Irrespective of the operating state of the plant, abnormalities in the plant can always be detected with high accuracy.
This is because the virtual model can also determine the transient phenomenon of the plant. In this case, it provided in the abnormality detecting apparatus al
The, based on the characteristic parameters of the virtual model on the operating data of the normal plant, any time, depending on the correction to take functions it more accurate, depending on the accuracy of the operating conditions of the plant obtained from the simulation of the virtual model the error of the characteristic parameters, by gradually reducing with plant operation record, to detect the abnormality of the plant more precisely, regardless of the operating conditions of the plant, will be detectable by the very high accuracy of the abnormality of the plant .

[0013]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a plant abnormality detection device according to one embodiment of the present invention. In the plant abnormality detection device of the present embodiment, compared to the configuration of FIG.
In addition to the sensor 100 that detects the state of the plant 1, a path is provided for taking the operating state of the plant 1 online.

If the sensor 100 is a microphone installed in the containment vessel, the abnormality detection device 200
First, data of the acoustic spectrum when the plant is normal is stored. At this time, the abnormality detection device 200 stores acoustic spectrum data when the plant is normal, corresponding to the operating state of the plant input online. This can, of course, be performed automatically and without any need for human intervention. In addition, when storing the normal data in the storage device, it is preferable to first determine whether or not the data is normal based on the judgment of the plant operator.

When a large amount of data is accumulated in the abnormality detection device 200, the abnormality detection device 200 analyzes the spectrum of the newly detected acoustic data, and further, in the past, based on the current operating state of the plant that has entered simultaneously. The stored data is retrieved from the stored acoustic spectrum data when the driving state is the same or close to the current one, and the two are compared and evaluated.
Here, if the current operating state of the plant is between the operating states of the two reference data stored in the past, if the data to be referenced is obtained by interpolating from the two data, the number of references can be reduced. Data can be used to more accurately detect plant abnormalities. If the abnormality detection device does not detect “abnormal”, the reference data may be automatically stored as normal data in association with the operation state of the plant at that time, and “abnormal” If detected, an alarm is issued and notified to the plant operator, and the abnormality detection device is determined to be `` abnormal '', but the plant operator determines whether or not it is really abnormal. The sound data determined to be “abnormal” is also stored as reference data in a normal state.

The above steps, ie, the step of accumulating normal data, need to be performed frequently at the stage of starting plant operation, but the frequency decreases as the reference data is accumulated. . More this method and adopted child, small abnormality is eliminated miss.
Although an acoustic sensor using a microphone is described here as an example, examples of the sensor include a vibrometer, a thermometer, a hygrometer, a radiation dosimeter, a TV camera, an infrared camera, a pressure gauge, and an odor sensor. The same method can be applied when using.

FIG. 3 is a configuration diagram of an abnormality detection device when the present invention is applied to a mobile inspection robot in a containment vessel of a nuclear power plant. In the containment vessel 1d of the nuclear power plant, there are a nuclear reactor 1a, a cooler 1c in a drywell, a recirculation pump motor 1b, and the like as main equipment. The nuclear power plant uses the steam generated from the reactor 1a to
The power is generated by the generator 1f by turning e, and the operation of the power plant equipment is controlled by the control / operation panel 1g in the central control room. In addition, during the power plant operation man is in a storage container 1d is in the environment process is not performed to enter. Therefore, a mobile inspection robot system was adopted,
The robot main body 102 is moved along the rail 101 installed in the storage container d, and the main equipment in the storage container 1d is automatically inspected and patrolled.

The control and operation of the robot are performed by an operation panel 200 installed outside the storage container 1d. The robot main body 102 is equipped with a TV camera, a microphone, and an infrared camera, and can perform visual inspection, abnormal noise inspection, and temperature inspection on behalf of humans. The determination as to whether or not there is an abnormality is made by storing detection data in a normal state in the operation panel 200 in advance.

The robot body 102 is provided with the sensor 10 shown in FIG.
0, and the operation panel 200 corresponds to the abnormality detection device 20 of FIG.
Corresponds to 0. Here, the operation state of the plant is input to the operation panel 200 online from the control / operation panel 1g of the plant. The TV image data, acoustic data, and temperature data (including temperature distribution) detected by the robot body 102. If these data are data that depends on the position, posture, and the like of the robot, the position, posture, and the like are also related data. The operating conditions of the plant that mainly affect the plant output (the power generation amount of the generator 1f), the operating status of the cooling system cooler 1c in the drywell, the rotation speed of the recirculation pump motor 1b, and the like are mainly included. Can be considered. Therefore, the plant output (generator 1
f), the operation status of the cooling system cooler 1c in the drywell, and the rotation speed of the recirculation pump motor 1b are input to the operation panel 200 online.

When the system is configured in this manner, the normal detection data in the operation panel 200 includes the plant output (the amount of power generated by the generator 1f), the operation status of the cooling system cooler 1c in the drywell, and the recirculation pump. The number of revolutions of the motor 1b is stored in association with the plant operation state, and the comparison with the detected data is performed under the same operation conditions of those plants. Therefore, as described above, it is possible to always detect the abnormality of the plant with high accuracy by the mobile inspection robot regardless of the operation state of the plant. If the data to be compared is data that depends on the position, posture and the like of the robot, it goes without saying that the comparison is performed using the data at the same position and posture. Here, as the operation state of the plant, the ambient temperature in the containment vessel 1d, the temperature of the coolant entering the recirculation pump motor 1b, and the like may be taken in so that the abnormality can be detected more accurately.

FIG. 4 is a block diagram of an abnormality detection device according to a second embodiment of the present invention, and is a basic configuration diagram of an abnormality detection device provided with a virtual model. In the abnormality detection device 200,
Model 1m of plant simulating plant 1, sensor 1
There is a virtual model 201 having a sensor unit model 100m that simulates the virtual model 201, and the sensor 100 of the virtual model 201
m is compared with the actual detection data of the sensor 100 by the comparator 202, and the plant 1 of the virtual model 201 is compared.
The comparison between the plant state of m and the plant state sent from the actual plant 1 is performed by the comparator 203, and the presence or absence of inconsistency between the virtual model 201 and the actual data is detected. A signal is output to the alarm 204 to issue an alarm.

Here, the virtual model 201 of the plant is
It may be a large-scale model of the entire nuclear power plant, a model of one device in the power plant, or a model of a part of the device. The virtual model may be a simple mathematical expression described by a computer program or data, a transfer function described on the basis of a block diagram, a program itself for executing a difference process, an input and an output. May be stored. Here, in order to explain the concept of the virtual model, only one embodiment of the virtual model will be introduced using a reactor recirculation pump as an example.

FIG. 6 shows a control system configuration of the reactor recirculation pump. The pump 20 is driven by the electric motor 10.
Since the output of the reactor is performed by changing the flow rate Q of the pump, the flow rate control system 31 controls the rotation speed of the electric motor 10. Since the motor 10 is an induction motor, its rotation speed is as shown in the following equation 1.

[0024]

(Equation 1)

Since the frequency is proportional to the frequency, the output of the variable frequency generator 32 rotated by the drive motor 34 via the fluid coupling 33 is used as the power source of the motor 10. The flow control system 31 outputs an operation signal to the fluid coupling 33 using the rotation speed of the variable frequency generator 32 as a feedback signal. Here, when this reactor recirculation pump is viewed as one plant, what can be easily output as the operating state of the plant is the rotation speed of the variable frequency generator 32 detected as a feedback signal. This can be taken online as an operation state of the plant by the abnormality detection device. In addition, the more detailed the state of the plant to be taken online, the more accurately the virtual model can be simulated, and the more accurate the abnormality detection can be. Therefore, if the operation state of the plant is to be taken in more detail, the oil of the electric motor 10 is required. The level of the surface gauges 11 and 12 is output as an electric signal, or the small generator 24 is directly connected to the motor 10 to directly detect the rotation speed of the motor 10 and output it as the operating state of the plant. The flow rate Q may be converted into an electric signal by the meter 23 and output, or the suction pressure and the discharge pressure of the pump 20 may be converted into electric signals by the pressure gauges 21 and 22 and output.

FIG. 7 is a graph showing an example of a model of a reactor recirculation pump represented by a characteristic curve. As the motor characteristic 10m in FIG. 7A, the relationship between the primary current I, the torque T, and the mechanical output P with respect to the slip s is given.
As the pump characteristic 20m of (b), the rotation speed N and the flow rate Q
Is given by the total head H and the resistance curve R. This may be stored in a computer as a characteristic expression having a coefficient, or may be stored as characteristic data itself. Here, the rotation speed of the variable frequency generator 32 and the small generator 24
The slip s is obtained from the relationship of the above equation 1 from the actual rotational speed detected by the above, and the torque T, the primary current I, and the mechanical output P at that time are determined from the characteristic 10 m of the motor. Further, the current total head is obtained from the suction pressure and the discharge pressure detected by the pressure gauges 21 and 22 of the pump 20, and the flow rate Q is determined from the current head and the current rotational speed N. Assuming that the actual primary current is detected as the state of the plant and is taken online in the abnormality detection device, the primary current determined from the detected primary current and the characteristic 10 m of the motor (this is the output of the virtual model 201). ) Is significantly different from "abnormal"
If the flow rate Q detected by the flow meter 23 is significantly different from the flow rate Q (this is the output of the virtual model 201) obtained from the pump characteristic 20m, the "abnormality" May be determined. Describing a virtual model with a characteristic curve in this way can be used as an effective means when modeling the characteristics of a device that is difficult to formulate in a nonlinear equation since it is possible to actually take data and use it as initial characteristics. is there.

Further, a temperature sensor 111 may be attached to the electric motor 10 and its electric signal may be taken into the abnormality detecting device, or a vibrometer 112 may be attached to the pump 20 so as to take the electric signal into the abnormality detecting device. Although the vibration model may be faithfully described based on the structure of the pump, the vibration data of the vibrometer 112 is analyzed into spectrum data by the abnormality detecting device, and the actual rotation speed N of the pump is detected. , The flow rate Q is stored as a parameter that regulates the operating state of the plant in association with the operating state of the plant, and is stored in the virtual model 1 of the sensor unit.
It is advisable to set it to 00m.

Next, an example of a model of the temperature sensor unit to be compared with the temperature sensor 111 will be described. In this case, it is preferable that the ambient temperature around the reactor recirculation pump is also detected by the thermometer 110 and taken into the abnormality detection device. FIG.
Fig. 7 shows a model of the motor surface temperature. The average temperature of the internal space of the rotor 10R is T0, the average weight is G0, the average specific heat is C0, the average weight inside the thermal center of the coil portion 10co of the stator is Gc1, the average specific heat is Cc1, the thickness is dc1,
The average thermal conductivity is λc1, the outer average weight is Gc2, the average specific heat is Cc2, the thickness is dc2, the average thermal conductivity is λc2, the temperature of the inner surface of the coil 10co is T1, the surface area of the inner surface is A1, and the heat of the inner surface is A1. The transmission rate is α0, the temperature of the thermal center of the coil section 10co is Tc, and the heat passage cross-sectional area is Ac.
T2 is the temperature of the inner surface of the casing 10ca outside, and A2 is the heat passage cross-sectional area of that portion. T3 is the surface temperature of the casing 10ca, A3 is the surface area, G2 is the average weight, C2 is the average specific heat, d2 is the thickness. The average thermal conductivity is λ2, the heat transfer coefficient of the surface is αa, the ambient temperature is Ta, the mechanical output of the rotor 10R is P, the input of electric energy to the coil 10co is qIN, and the heat radiation from the surface of the casing 10ca To q
If OUT, the heat transfer model of this motor is represented by the following equation (2).

[0029]

(Equation 2)

The ambient temperature Ta can be detected by a thermometer 110. Electric energy input qIN can also be detected by measuring voltage and current. The mechanical output P of the rotor 10R can also be known from the characteristic 10m of the electric motor. △ T
i is the change of the temperature Ti during the short time Δt. Here, the weight G0, Gc1, Gc2, Gc2 of each part, each cross-sectional area A1,
Ac, A2, A3 and the thicknesses dc1, dc2, d2 are constants that can be accurately known in advance. Also, specific heats C0, Cc1,
Cc2, C2, thermal conductivity λc1, λc2, λ2, heat conductivity α0,
α1 can also be obtained by calculation or a prior characteristic test. Therefore, it is possible to sequentially calculate the temperature change ΔTi of each part and the temperature Ti of each part by using the difference method by the equation of Expression 2.

On the other hand, the actual surface temperature T3 of the casing 10ca can be detected by the temperature sensor 111. Alternatively, the surface temperature T3 can be detected by a non-contact method using an infrared camera mounted on the mobile robot 102.
Therefore, the surface temperature T3 obtained from Equation 2 (virtual model 1)
00m output) and the actual surface temperature T detected by the sensor.
By comparing with 3, it is possible to easily detect the abnormality of the electric motor. In this way, the current temperature is not determined only by the current operation state of the plant, such as when the plant includes a delay element or the surface temperature of the equipment, etc.
In the case where it also affects the history of the past operation states, providing the virtual model 201 of the plant in the abnormality detection device 200 is an effective method for more correctly detecting the abnormality of the plant. .

FIG. 5 is a block diagram of an abnormality detecting device for correcting characteristic parameters of a virtual model. Abnormality detection device 20
In 0, comparators 202 and 203 for detecting an abnormality
In addition, comparison correctors 205 and 206 are provided for comparing the output of the virtual model with the actual detection data in order to obtain the correction amount of the characteristic parameter. Here, the characteristic parameter is the characteristic curve itself of the electric motor characteristic 10 m and the pump characteristic 20 m in FIG. 7, or the heat transfer coefficient α0, the heat conductivity λc1, λc2, etc. in Equation 2. Comparing and detecting the contradiction between the output of the model 201 and the actual detection data means detecting an abnormality of the plant by detecting that a characteristic parameter of the real object changes due to a failure or deterioration. When a large number of operation results of the plant are accumulated, it becomes possible to describe the secular change of the characteristic parameter of the plant as a model using the accumulated operation time of the plant as a parameter. In this case, a change in the characteristic parameter exceeding the normal aging is detected, and an abnormality in the plant is detected. Therefore, there is no problem in correcting the characteristic parameter of the model 201 to a more accurate one when no "abnormal" is detected. However, when "abnormal" is detected, the switches 207 and 208 are opened and the model 201 is opened. It is necessary not to unconditionally correct these characteristic parameters. In this case, the abnormality detection device 200 issues an alarm and leaves the judgment to the plant operator. Only when the operator determines that it is not “abnormal”, the characteristic parameters on the plant side are corrected, and the model 201 is rewritten to a more appropriate one. According to this method, the abnormality of the plant can be detected more accurately, and the abnormality of the plant can always be detected with high accuracy regardless of the operation state of the plant.

[0033]

According to the present invention, abnormalities in a plant can always be detected with high accuracy regardless of the operating state of the plant.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an abnormality detection device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of a conventional abnormality detection device.

FIG. 3 is a configuration diagram of an abnormality detection device of the mobile inspection robot.

FIG. 4 is a configuration diagram of an abnormality detection device according to a second embodiment of the present invention.

FIG. 5 is a configuration diagram of an abnormality detection device having a model correction function.

FIG. 6 is a control system configuration diagram of a reactor recirculation pump.

FIG. 7 is an explanatory diagram of a model of a reactor recirculation pump.

FIG. 8 is an explanatory diagram of a model of a motor surface temperature.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Plant, 100 ... Sensor, 200 ... Abnormality detection device, 201 ... Virtual model, 1m ... Plant virtual model, 100m ... Virtual model of sensor part, 202, 203
... Comparator, 204 ... Alarm.

Continuation of the front page (72) Inventor Tomiji Yoshida 3-1-1 Kochi-cho, Hitachi-shi, Ibaraki Hitachi, Ltd. Hitachi Plant (56) References JP-A-3-220498 (JP, A) JP-A-3 JP-A-248015 (JP, A) JP-A-3-100425 (JP, A) JP-A-3-92718 (JP, A) JP-A-63-302397 (JP, A) JP-A-58-100786 (JP, A) JP-A-57-198894 (JP, A) JP-A-57-110991 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G21C 17/00

Claims (4)

(57) [Claims] 1. A plant state is detected by a sensor attached to a mobile robot, data indicating a plant state when the plant is normal is stored in a storage device in advance, and a detection signal of the sensor is compared with the data. In the plant abnormality detection method of judging the abnormality of the plant in, the plant operation state is captured online separately from the detection signal of the sensor, the plant operation state captured online and the sensor detection captured from the sensor under the plant operation state The signals are stored in a storage device as normal data together with the robot information such as the position and posture of the mobile robot, and when the detection signal of the sensor is compared with the normal data to determine the abnormality, the same operation state and the same operation are performed. The normal data of the robot information is retrieved from the storage device and compared with the sensor detection signal. And a method for detecting abnormality of a plant. 2. The method according to claim 1, wherein a plant state is detected by a sensor attached to the mobile robot, data indicating the plant state when the plant is normal is previously stored in a storage device, and a detection signal of the sensor is compared with the data. In a plant abnormality detection device that determines a plant abnormality with
Means for capturing the plant operating state online separately from the detection signal of the sensor, and the position and posture of the mobile robot based on the plant operating state captured online and the sensor detection signal captured from the sensor under the plant operating state. Means for storing in the storage device as normal data together with the robot information, etc., and when the abnormality is determined by comparing the detection signal of the sensor with the normal data, the normal data of the same operating state and the same robot information is taken out from the storage device. And a means for comparing with a sensor detection signal. 3. A plant state such as a sound, a temperature, and a vibration of the plant is detected by a sensor, and a sensor signal when the plant is in a normal state is stored as reference data, and the sensor signal is compared with the reference data. in the abnormality detection method of the plant to determine the abnormal, takes in the detection signal of the sensor fetches the plant operating conditions online, in a virtual model of the plant, the aging of the plant
The characteristic parameters shown in the
While being corrected from time to time, the plant operating state and the sensor output are obtained from the virtual model of the plant and the virtual model of the sensor, and the result obtained from the virtual model and the actual plant operating state taken online or the actual state of the sensor are obtained. A plant abnormality detection method, wherein a plant abnormality is detected by comparing the detection signal with a detection signal of the plant. 4. A plant state such as a sound, a temperature, and a vibration of the plant is detected by a sensor, and a sensor signal in a normal state of the plant is stored as reference data, and the sensor signal is compared with the reference data to obtain a plant signal. In a plant abnormality detection device for determining abnormality, means for capturing a plant operation state online and capturing a detection signal of the sensor, and a plant aging model using a virtual model of the plant.
The characteristic parameters that indicate the change
Means for correcting at any time based on data, means for determining a plant operating state and a sensor output from a virtual model of a plant and the virtual model of the sensor, and a result obtained from the virtual model and the actual plant operating state captured online. Means for comparing a detection signal of the sensor with an actual detection signal to detect a plant abnormality.