JP3131659B2 - Equipment abnormality monitoring device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device abnormality monitoring device for determining whether an operating state of a device used in a plant is normal or abnormal.
The present invention relates to a device abnormality monitoring device that uses a learning function to determine abnormality.

[0002]

2. Description of the Related Art As a conventional monitoring device having a learning function, for example, as described in Japanese Patent Application Laid-Open No. 58-189514, an algorithm for determining an abnormality is determined in advance in accordance with the mechanism of the device. To use the learning function. Also, Japanese Patent Application Laid-Open No. 59-114
As described in Japanese Patent Application Laid-Open No. 07-2007, a learning function of a method of acquiring an optimal control parameter value during device operation while performing learning control and rewriting a diagnostic reference value based on the information of the learning control has been used.

[0003]

In the above-mentioned prior art, the former is based on the fact that an algorithm for determining an abnormality is created in accordance with the mechanism of operation of a device, and therefore cannot detect an abnormality that is not considered by the determination algorithm. There was a problem with the point. In the latter case, by monitoring a plurality of monitoring parameters simultaneously and grasping the operation state of the device with high accuracy, the point of detecting an unexpected abnormality and the storage capacity for storing the values of the monitoring parameters are reduced. Points are not considered. In view of the above circumstances, the object of the present invention is to detect a device operating state with high accuracy,
It is another object of the present invention to provide a device abnormality monitoring device suitable for judging an unexpected abnormality by reducing the storage capacity and improving the device abnormality monitoring performance.

[0004]

In order to achieve the above object, a plurality of values of a main monitoring parameter and a main monitoring parameter
Multiple dependent monitoring parameters for each value of
Storage means for attaching and storing the
Slave monitoring parameters and slave monitoring stored in the storage means
And the representative value of the dependent monitoring parameter based on the
And the dependent monitoring parameters stored in the storage means.
Together with an update means for updating the representative values, monitoring
Update corresponding to the value of the main monitoring parameter detected from the target
Multiple dependent monitoring parameters updated by new means
Extraction means for extracting values from storage means and detection from monitored objects
The value of the plurality of subordinate monitoring parameters issued and the
Of the dependent monitoring parameters extracted
Abnormality detection means for detecting an abnormality of a monitoring target by comparing
You. Alternatively, the dependent monitoring parameter determined by the update
The representative value of the meter is stored periodically, and multiple stored
The monitoring target changes according to the change of the monitoring parameter value over time.
It has abnormality detecting means for detecting the normal state. Where the monitored
Of the monitoring parameter in the operating state
And the dependent monitoring parameters stored in the storage means.
Set the distance difference from the representative value pattern
When the value is exceeded, it is judged as abnormal . Also, the main monitoring parameters
Data and multiple values for the main monitoring parameters.
To store the number of dependent monitoring parameters in association with each other
Read from storage means in storage mode and learning mode
Input the subordinate monitoring parameters for normal
Learn to be the quantization value of the main monitoring parameter determined in advance.
Learning multi-layer neural network and monitoring mode
And the value of the dependent monitoring parameter detected from the monitoring target
And the output when input to the multilayer neural network
Comparing the quantization values of predetermined monitoring parameters
Abnormality is detected based on the comparison means and the magnitude of the comparison result
It has abnormality detection means . Also, when the monitoring target is started or
Indicates the multiple values of the main monitoring parameters
Multiple dependent monitoring parameters for each value of viewing parameters
First and second storing means for each association only stores the bets
During startup or steady state detected from the monitoring target.
Genus monitoring parameters and stored in the first and second storage means.
Depending on the dependent monitoring parameters
A representative value of a monitoring parameter is determined, and first and second storages are performed.
Start-up or steady-state subordinate monitoring stored in the means
First and second parameters to be updated to the respective representative values.
And a second updating means, and a startup time detected from the monitoring target.
The first is the first one corresponding to the value of the main monitoring parameter in the steady state.
And a plurality of dependent monitors updated by the second updating means
Retrieve parameter values from first and second storage means
Extraction method and the time of startup or
The values of a plurality of dependent monitoring parameters at regular times and the first and second
A plurality of subordinate monitoring parameters extracted by the second extracting means.
An abnormality that detects an abnormality of the monitoring target by comparing with the meter value
Detecting means .

[0005]

In order to grasp the normal state of the device, it is necessary to know the normal range of the monitoring parameter such as the absolute value of the measured quantity or the amplitude of a specific frequency component. However, the normal range of each monitoring parameter may fluctuate greatly depending on the operating conditions, and if each monitoring parameter is used independently, the accuracy of grasping the normal state is poor. On the other hand, if the normal state is grasped by a combination of the values of many monitoring parameters, that is, the monitoring parameter pattern, the accuracy of grasping the normal state can be improved. For example, in a rotating machine, it is assumed that the number of rotations during normal operation is 0 to 100% and the normal range of its vibration is 0 to 100 μm. However, if the vibration amplitude of 100 μm occurs only at the critical speed of 50% rotation (for example, at the time of resonance phenomenon),
If a vibration of 100 μm is observed at a rotation speed other than the critical speed, it is abnormal. For this reason, if the normal amplitude is grasped for each rotation speed, the accuracy of grasping the normal operation state is further improved.
As a result, the detection sensitivity in a state other than the normal operation state can be improved, and an unexpected change in the critical speed due to an unexpected change in the rigidity of the rotating machine base can be detected. In other words, the reason for simultaneously monitoring a plurality of monitoring parameters in the present invention is that the range of the normal state is limited to that of the conventional method by more accurately grasping the operating state of the device, and consequently the detection range of the abnormal state is reduced. This is because an unexpected abnormality can be detected by widening. Since the normal state of the device usually changes during a periodic inspection or the like, it is necessary to grasp the normal state each time the periodic inspection is performed.
For this reason, in a nuclear power plant or the like having many devices, there is a possibility that a lot of time may be spent on the work of grasping the normal state.
Therefore, the problem of working time can be avoided by automatically grasping the normal state. The simplest method for automatically grasping the normal state is to store a combination of all of the plurality of monitored parameter values in the normal operation state once, and to combine the combination of all of the stored plurality of monitored parameter values, What is necessary is just to compare the monitoring parameter values in the operating state. For example, a vibration amplitude of 10 μ at a rotation speed of 10%
m and 20% are stored as 11 μm, and if the monitored parameter value during normal operation is 10% of rotation speed and the vibration amplitude is 10 μm, it is determined to be normal, and if it is 11 μm, it is determined to be abnormal. However, this method increases the number of monitoring parameters,
It is impossible to memorize all the normal states because of the huge amount of memory. Even if a neural network used in so-called learning theory is used, it is not possible to expect effective learning for monitoring. In a supervised learning mechanism, the concept of how to classify many normal states is unknown, and in an unsupervised learning mechanism, an evaluation function for learning related to monitoring is unknown. Therefore,
In the present invention, in order to avoid the above-mentioned problem of storage capacity, a monitoring parameter that governs the state of a device to be monitored is selected from among a plurality of monitoring parameters, and this is set as a main monitoring parameter, and other monitoring parameters are selected. Is stored as a subordinate monitoring parameter, and a representative value of the subordinate monitoring parameter is stored according to the value of the main monitoring parameter. The representative value here is the average or median value of the monitoring parameters,
It is a standard deviation or a maximum / minimum value indicating the magnitude of the monitoring parameter fluctuation. By providing the main monitoring parameters, the classification of parameter patterns in learning is performed in advance,
By storing the representative values of the dependent monitoring parameters while updating them, the storage capacity can be realized as an apparatus. In the monitoring mode of the device abnormality monitoring device, the representative value of the subordinate monitoring parameter is read according to the value of the predetermined main monitoring parameter, and the value of the subordinate monitoring parameter is compared with the value of the subordinate monitoring parameter of the device in the operating state. Otherwise, it is determined to be abnormal. Thus, substantially the same abnormality detection can be performed without storing all combinations of the monitoring parameters in the normal state. From a learning perspective,
The normal state differs for each main monitoring parameter, which is equivalent to classifying the subordinate monitoring parameter pattern according to the value of the main monitoring parameter. From this viewpoint, the following device abnormality monitoring device can be constructed. That is, the main monitoring parameter value is used as a teacher signal, and the subordinate monitoring parameter is used as an input signal of the neural network. In the learning, the coefficient of the middle layer of the neural network is tuned, and when the operation state is normal, when the dependent monitoring parameter is input, the output substantially matches the value of the main monitoring parameter. The tuning of the coefficient of the intermediate layer and the calculation of the representative value of the monitoring parameter in the above-described method have the same meaning. In the abnormality determination, when the subordinate monitoring parameter during normal operation of the device is input to the learned neural network, it is determined to be normal when the output substantially matches the main monitoring parameter, and to be abnormal when the output does not match.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to FIG. FIG. 1 shows an example in which the present invention is used for monitoring an abnormality of a recirculation pump loop of a nuclear power plant. In FIG. 1, heat generated in a core 102 inside a reactor 101 is transmitted to cooling water flowing inside the core 102, and the cooling water boils and passes through a main steam pipe 103 to rotate a turbine / generator to generate electric power. Get. The steam that has turned the turbine / generator turns into water again, passes through the water supply pipe 104, and
It is led to 1. The calorific value of the core 102 depends on the recirculation pump 1
05, it changes depending on the flow rate of the recirculation pump loop composed of the recirculation pipe 106 and the jet pump 107. In order to maintain the purity of the cooling water passing through the reactor core 102 and pipes, the cooling water is branched from the recirculation pipe 106, purified by the reactor water purification device 108, and returned to the water supply pipe 104. The device abnormality monitoring device of the present embodiment is installed for monitoring a recirculation pump loop for controlling the calorific value of the core 102. For monitoring, it is equipped with various measuring devices for knowing the operation state of the recirculating pump loop, such as a rotation speed measuring device 121, a vibration measuring device 122, a flow measuring device and a bearing temperature measuring device of the recirculation pump 105. . In order to monitor also the water quality having information on the operation state of the recirculation pump loop, a sensor is disposed in the reactor water purification device 108 to measure the conductivity.
Measured at 3. A signal measured by each measuring device is input to a monitoring parameter extractor 200 having a function of converting an AC signal into a DC level, filtering an appropriate time constant, and converting an analog signal into a digital signal. The monitoring mode setting unit 250 instructs switching between the two modes of the device abnormality monitoring device of the present embodiment, and can set a learning mode and a monitoring mode. The input of the monitoring parameter pattern selector 220 is a rotation speed which is a main monitoring parameter, and the output thereof is a value obtained by quantizing the monitoring parameter value. The quantization referred to here is, for example, the number of rotations 117
This is an operation of converting 5 rpm into 1170, and is an operation of converting the value of the monitoring parameter into a discrete value at a preset interval instead of a continuous value. When the output of the monitoring mode setting unit 250 is in the learning mode, the monitoring parameter learning unit 230 learns the normal values of the dependent monitoring parameters, which are the other monitoring parameters, according to the quantization value of the input main monitoring parameter. . When the output of the monitoring mode setting unit 250 is in the monitoring mode, a representative value of the learned subordinate monitoring parameter is output according to the quantization value of the main monitoring parameter. The abnormality detector 260 operates when the output of the monitoring mode setting unit 250 is in the monitoring mode, compares the representative value of the learned normal monitoring parameter with the value of the dependent monitoring parameter at the current time, and determines the monitoring parameter at the current time. If the value is out of the normal range, it is determined to be abnormal, and the determination result is output to the alarm / display 280. The alarm / display 280 performs a predetermined alarm / display according to the determination result.

The configuration of the apparatus abnormality monitoring apparatus for the recirculation pump loop according to the embodiment of the present invention and the general functions of each section have been described above. Hereinafter, the structure and operation of the main part will be described in detail. Monitoring parameter learning unit 230
2 is shown in FIG. A storage 231 for storing the learning data, a data switch 232 for changing the output destination of the learning data according to the monitoring mode, and an accumulator 233 for updating the representative value of the dependent monitoring parameter which is the learning data. Become. Here, learning refers to an operation for determining a range that can be taken when the dependent monitoring parameter is normal from the value of the dependent monitoring parameter when the recirculation pump loop is normal. FIG. 3 shows a data flow in the monitoring parameter learning unit 230 in the monitoring mode. In FIG. 3, the storage unit 231 stores, in accordance with the quantization value i of the main monitoring parameter, the total number of learned data that is the representative value of the dependent monitoring parameter, the simple sum of the monitoring parameter values, the square sum,
The average value, standard deviation, maximum value, and minimum value are stored.
The storage area of the representative value of the dependent parameter in the storage unit 231 is selected in accordance with the quantization value i of the main monitoring parameter every time a predetermined time elapses, and the storage content of the selected storage area is sent to the accumulator 233. The accumulator 233 calculates a new representative value from the representative value sent from the storage unit 231 and the value X _jG of the dependent monitoring parameter at the current time, and calculates the contents of the storage area corresponding to the quantized value i of the main monitoring parameter. To update. By repeating such an operation in the learning mode, the range of the subordinate monitoring parameter during normal operation of the device is stored in the storage unit 231 for each value of the main monitoring parameter. FIG. 4 is a flowchart showing a flow of data processing of the device abnormality monitoring device according to the present embodiment. The rotational speed, which is a main monitoring parameter, is A / D-converted (converts an analog signal to a digital signal) at predetermined time intervals and quantized at predetermined steps. Read the contents of the storage area corresponding to the quantized value i,
The value of the dependent monitoring parameter at the current time is A / D converted. If the currently set mode is the learning mode, the content of the storage area is updated after the update operation of the representative value. After updating the content,
Wait for the input of the main monitoring parameter again. On the other hand, learning
At the time of monitoring mode determination, if the mode is the monitoring mode, the following processing is performed. First, an allowable maximum value and an allowable minimum value for each dependent monitoring parameter are calculated from the learning data by using a predetermined allowable width coefficient a. The calculation formula is shown on the flowchart. Using the allowable range obtained in this calculation, compare whether the value of the dependent parameter at the current time is within the allowable range, and if at least one of the dependent monitoring parameters is outside the allowable range, perform the predetermined procedure. Performs alarm and display.
If the value is within the allowable range, a normal display is performed, and input of a main monitoring parameter is waited. As described above, the operation of the device abnormality detection device of the present embodiment is such that the normal range of the monitoring parameter is learned in the learning mode during the operation check test after the periodic inspection, and the monitoring device is operated in the monitoring mode during normal operation. Among the monitoring parameters, for example, the conductivity etc. may be different from the values during normal operation immediately after the periodic inspection, etc.Therefore, it is not possible to have a function to partially correct the learning result with human intervention. It is possible. Further, it is possible to perform learning newly or additionally during driving with the configuration of FIG.

FIG. 5 shows an example of abnormality detection in this embodiment. For explanation, the normal range set for each conventional single monitoring parameter is also shown in the figure. In the recirculation pump loop, it is the number of revolutions of the recirculation pump 105 that may dominate the fluctuation of various measured quantities,
The rotation speed is selected as the main monitoring parameter. The dependent monitoring parameters include the vibration, flow rate, bearing temperature conductivity, etc. of the recirculation pump 105. The monitoring parameter value is expressed as a percentage based on the maximum value of each monitoring parameter previously considered. “━” is the normal range of the monitoring parameter defined by the conventional monitoring method, and “← →” is the normal range obtained by learning of the device abnormality monitoring device of the present embodiment. In the past, the normal range was widened because the value of the vibration that could be taken in the entire range of the rotation speed was a normal value, but the normal range was narrowed by adopting the learning mechanism of the present invention, and as a result, abnormality detection was performed. Sensitivity has been improved. “○” is the monitoring parameter value at the current time, and only the bearing temperature is out of the normal range obtained by learning. By investigation after abnormality detection,
It was discovered that the bearing lubrication system was defective, and measures were taken before the vibration increased. The conventional abnormality monitoring device cannot detect a slight abnormality as shown in FIG. 5, and it can be seen that the abnormality detection sensitivity is improved in this regard. Also,
According to the present embodiment, when the values of the monitoring parameters are out of the normal range, all are determined to be abnormal, so that it can be seen that unexpected abnormalities can also be detected.

FIG. 6 shows an alarm / display 2 according to the present embodiment.
80 shows an example of the display screen 80. The time change of the observed pattern at 50% rotation speed, which is the main monitoring parameter, is defined as the minimum value of the difference between the learned normal pattern and the observed monitoring parameter pattern distance (considering the standard deviation of the normal pattern and the allowable variation coefficient). , Shows the temporal change of the pattern distance. In addition, the current value of each monitoring parameter and the learning were obtained so that it was possible to know which monitoring parameter in the observation parameter pattern was closer to the alarm level (considering the average value and standard deviation and the allowable variation width coefficient). The relation between the alarm value and the average value is also displayed. Thereby, the trend of the margin from the alarm value of the observation parameter pattern can be grasped, and what the monitoring parameter which causes the margin is clear at a glance. In addition, these displays have meaning as separate screens. In the data processing of the device abnormality monitoring device described above, the average value is used as one of the representative values of the dependent monitoring parameters, but this is changed to a median value, or when the abnormality is determined, the monitoring parameter is learned. It is possible to use maximum and minimum values. In the abnormality detection, it is also possible to determine that an abnormality has occurred when the difference between the pattern distance between the observation monitoring parameter pattern and the learning pattern exceeds a set range. The hardware used in the above device configuration can be all realized by a combination of conventional techniques.

Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 7 is an example in which the present invention is applied to abnormality monitoring of a turbine / generator system of a nuclear power plant. In FIG. 7, high-temperature / high-pressure steam is guided to the turbine 110 through the main steam pipe 103, and the turbine 110 is rotated.
Electric power is generated by a generator 111 directly connected to the turbine 110. Steam from the turbine 110 passes through a condenser 112 to become water and is sent to a water supply system. The measured quantities for abnormality monitoring include generator output, vibration, rotation speed (not shown),
Lubrication system instrumentation such as bearing temperature (not shown), degree of vacuum of condenser 112, main steam flow rate, and the like. These measured quantities are guided to the monitoring parameter extractor 201 through various measuring devices 125 to 128. The learning control of the two monitoring parameter pattern learning units 230 and 240 is performed in accordance with the setting mode of the monitoring mode setting unit 250. The monitoring parameter pattern selector 220 has the same function as in the first embodiment. In the monitoring mode, the abnormality detector 265 controls the two monitoring parameter pattern learning units 230 and 24.
The output of 0 is compared with the monitoring parameter pattern at the current time to detect the presence or absence of an abnormality. In addition, the long-term fluctuation detector 27
0 is added to detect an abnormality from a long-term fluctuation of the monitoring parameter pattern, and the details will be described later. The alarm / display 285 is for displaying an abnormality determination result of the abnormality detector 250 or the long-term fluctuation detector 270.
Characteristic portions in the embodiment of FIG. 7 are that there are a monitoring parameter pattern learning device 230 for activation and a monitoring parameter pattern learning device 240 as learning devices, and that a long-term fluctuation detector 270 is added. . The start-up monitoring parameter learning unit 230 is the same as that described in the first embodiment, and learns the normal range of the monitoring parameter at the time of starting-up after inspection of equipment such as regular inspection. The monitoring parameter pattern learning unit 240 increases the learning weight to the current time, assuming that the normal range of the monitoring parameter pattern changes due to aging of the device. The long-term fluctuation detector 270 periodically records the learning data of the monitoring parameter pattern learning unit 240,
It has a function to detect abnormalities from the change tendency. This makes it possible to realize a device abnormality detection device that can cope with aging of plant devices.

The operation of the apparatus abnormality monitoring apparatus according to the present embodiment will be described with reference to the flow of data processing shown in FIGS. The turbine speed is controlled to a constant value, and the generator output is used as the main monitoring parameter in this embodiment. The dependent monitoring parameters are other than the generator output described above. Main monitoring parameters are fetched at regular intervals, quantized, and representative values are read. Next, the subordinate monitoring parameters are read to determine whether the mode is the monitoring mode or the learning mode. In the learning mode, the learning data is updated as in the first embodiment. Since the two learning devices 230 and 240 perform the same learning, the learning results are the same if the initial values are the same. on the other hand,
In the case of the monitoring mode, an allowable value is calculated. Since there are two learning data, to calculate the allowable value corresponding to both,
It is necessary to set two allowable width coefficients in advance. In the abnormality determination, if the value of the monitoring parameter at the current time is out of the allowable range of one of the learning data, it is determined that the abnormality is abnormal.
In the abnormal alarm / display process, related information including which learning data exceeds the allowable range is displayed. Wait for learning / monitoring time after abnormal alarm / display processing. When the value of the monitoring parameter at the current time is within the allowable range of the two learning data, the monitoring parameter pattern learning unit 2 shown in FIG.
40 learning data are updated. The learning with the weight increased at the current time is realized by the time weight l. After updating the learning data, it is determined whether or not filing of the learning data is performed in the long-term fluctuation detector 270. If unnecessary, the normal display processing shown in FIG. 8 is performed. At the time of filing the learning data (in the present embodiment, it is performed every 10 days), the learning data is additionally recorded in the long-term fluctuation detection file. The fluctuation tendency is extracted from the recorded data for each monitoring parameter, and if the fluctuation tendency is determined to be significant, an abnormal alarm / display process is performed. In the extraction of the fluctuation tendency, the average value and the standard deviation among the representative values of the monitoring parameters are used.
If there is no significant fluctuation tendency, the normal display processing shown in FIG. 8 is performed, and the learning / monitoring time is waited.

FIG. 10 shows an example of detecting an abnormality of the long-term fluctuation detector 270. The generator output which is the main monitoring parameter is 10
At 0%, there is a gradual increase in vibration level. As a result of investigating the cause, the sinking of the turbine / generator base slowly progressed,
As a result, it was found that the vibration increased slowly. Such a gradual change tendency cannot be detected in the first embodiment. In the device abnormality monitoring apparatus of the present embodiment described above, if there is a difference between the learning data of the activation parameter pattern learning unit 230 and the value of the monitoring parameter due to aging or the like, it is always determined that an abnormality has occurred. As in the first embodiment, this point can be avoided by the monitoring personnel rewriting the learning data or masking the learning data so as not to make a comparison at the time of abnormality detection. Further, in order to simplify the device configuration, if two learning mechanisms having no time weight are prepared and one of them is learned at the time of monitoring, an erroneously operated device abnormality monitoring device can be realized by normal aging.

Next, a third embodiment of the present invention will be described with reference to FIG. FIG. 11 is an example in which the present description is used for monitoring abnormal sound of a steam generator of a fast breeder reactor. A typical anomalous noise is the sound that leaks into the sodium and is associated with the resulting sodium water reaction. Hot sodium passes through the sodium inlet pipe 701, passes around the heat transfer pipe 707, and returns from the sodium outlet pipe 702 to the pump side. A heat transfer tube 707 through which the water supplied from the water supply inlet pipe 703 passes is disposed in the container copper 705, and the heat transfer tube 707 is supported by a heat transfer tube support structure 706. A plurality of acoustic sensors 711 to 716 are arranged on the outer wall of the container copper 705, and the output is input to the monitoring parameter extractor 202 through the amplifiers 720 to 725. Ma
The output of the sodium flow meter 726 is also monitored.
It is input to the parameter extractor 202 . When the setting of the monitoring mode setting unit 250 is the learning mode, the monitoring parameter pattern selector 221 and the neural network 235 learn the normal monitoring parameter pattern, and in the monitoring mode, compare the output of the neural network 235 with the main monitoring parameter quantization value. Is performed by the abnormality detector 266, and when the difference is larger than the set value, it is determined that the abnormality is abnormal. The judgment result and the related information are displayed on the alarm / display 286. The neural network 235 uses a multilayer neural network having a supervised learning mechanism.

The operation of the device abnormality monitoring apparatus of the present embodiment will be described with reference to the flowchart of FIG. In this embodiment, the sodium flow rate is used as the main monitoring parameter, and the sound at each position on the outer wall of the container copper 705 is used as the subordinate monitoring parameter. The sound generated by the steam generator is mainly a flow sound of steam or sodium, and the generated sound level varies depending on the flow rate of sodium. Wait for the learning / monitoring time and fetch the main monitoring parameters. After quantizing the main monitoring parameters and reading the dependent monitoring parameters, it is determined whether the mode is the learning mode or the monitoring mode. In the case of the learning mode, the input of the neural network is set as the dependent monitoring parameter, and at that time, the back propagation calculation is performed so that the output becomes the quantization value i of the main monitoring parameter.
Tune the coefficients of the hidden layer of the neural network. Then, it waits for the learning / monitoring time. In the monitoring mode, the input of the neural network is used as the dependent monitoring parameter, and the difference between the output of the neural network at that time and the quantization value i of the main monitoring parameter is determined. When the difference is equal to or larger than a preset value, an abnormality occurs. Alarm and display processing. If the status is normal, a normal display process including the determination result and the display of the related information is performed. After the display processing, it waits for the learning / monitoring time. By the above abnormality monitoring operation, the body container 70 for each sodium flow rate is automatically set.
(5) The acoustic level of the outer wall can be learned, and abnormal sound detection with higher sensitivity than before can be performed. The sound generated inside the steam generator also depends on the surface condition of the inner wall of the tube with which the fluid contacts. Since the surface state may change to some extent over time, to deal with this, use two neural networks,
If one trained neural network is used for monitoring and the other initialized neural network is trained and operated alternately at appropriate time intervals, it is possible to cope with aging. It is needless to say that the third embodiment can be realized by the learning mechanism shown in the first and second embodiments even if the learning machine is not the neural network of this embodiment.

[0015]

As described above, according to the present invention, the following specific effects can be obtained. (1) The range of normal monitoring parameters is automatically clarified by learning, and anything outside the normal range is determined to be abnormal.
Detection of unexpected abnormalities becomes possible. (2) Since a learning method that employs a constant value such as an average value that can be updated every moment as a representative value of the monitoring parameter is realized, the storage capacity can be reduced, and the hardware scale of the device abnormality detection device can be reduced. This has the effect of improving economic efficiency. (3) Since the learning result is a representative value of the monitoring parameter, that is, the average value, the standard deviation, or the like, the evaluation of the learning result is easy, and the correction or the like can be easily determined. Therefore, the operation of the device abnormality monitoring device is facilitated. The usability is significantly improved. (4) Since the normal range of the monitoring parameters during operation is automatically learned, the work required for the examination and test for setting the normal range can be reduced, and the manpower related to operation can be reduced. effective. (5) Since the normal range including the mutual relation of the monitoring parameters is grasped, the abnormality detection sensitivity is improved, and the performance as the device abnormality detection device is improved. (6) In the abnormality detection, the allowable range for determining the normal range is set to the allowable range proportional to the standard deviation of the monitoring parameter, so that the allowable range of the monitoring range is automatically determined depending on the magnitude of the normal fluctuation. Therefore, the abnormality detection sensitivity of the apparatus can be adjusted only by changing the allowable width coefficient (see FIGS. 4 and 9), and the usability is significantly improved. (7) By displaying the abnormality determination result as the time change of the pattern distance and the current value of each monitoring parameter, it is possible to intuitively grasp the time variation of the monitoring parameter pattern and the margin to the alarm level of each monitoring parameter. As a result, the usability is significantly improved. (8) By providing the monitoring parameter pattern learning device, it is possible to avoid erroneous alarms for aging of normal devices and the like. (9) By providing the long-term fluctuation detector, abnormality detection that progresses slowly becomes possible, and the accuracy of device abnormality monitoring is improved. (10) Since a multi-layer neural network with a learning function is used as a learning machine, which is relatively general-purpose, it is possible to reduce the price associated with the production of the device abnormality monitoring device, and to improve the economical efficiency of the device abnormality monitoring device. effective. (11) By alternately using the two parameter pattern learning devices for learning and monitoring, it is possible to avoid false alarms due to aging of the parameter pattern in a normal state.

[Brief description of the drawings]

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

FIG. 2 is a detailed configuration of a monitoring parameter learning device.

FIG. 3 is a flow of data around a monitoring parameter learning device.

FIG. 4 is a flowchart showing a flow of data processing of the device abnormality monitoring device.

FIG. 5 is an example of abnormality detection in the first embodiment.

FIG. 6 is a display example of an alarm / display unit.

FIG. 7 shows the configuration of a device abnormality monitoring device according to a second embodiment of the present invention.

FIG. 8 is a flowchart illustrating a flow of data processing of the device abnormality monitoring device according to the second embodiment;

FIG. 9 is a flowchart illustrating a flow of data processing of the device abnormality monitoring device according to the second embodiment;

FIG. 10 is an example of abnormality detection by the long-term fluctuation detector.

FIG. 11 shows a configuration of a device abnormality monitoring device according to a third embodiment of the present invention.

FIG. 12 is a configuration diagram of a device abnormality detection device according to a third embodiment.

[Explanation of symbols]

 Reference Signs List 200 monitoring parameter extractor 220 monitoring parameter pattern selector 230 monitoring parameter learning device 250 monitoring mode setting device 260 abnormality detector 280 alarm / display 231 storage device 233 accumulator 240 monitoring parameter pattern learning device (for monitoring) 265 abnormality detector 270 Long-term fluctuation detector 285 Alarm display 221 Monitoring parameter pattern selector 235 Neural network 266 Abnormality detector 286 Alarm display

──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Shunsuke Uchida 1168 Moriyama-cho, Hitachi City, Ibaraki Prefecture Inside Energy Laboratory, Hitachi, Ltd. Hitachi, Ltd. Hitachi factory (72) Inventor Kouji Obebe 3-1-1, Sakaimachi, Hitachi, Ibaraki Pref. Hitachi, Ltd. Hitachi factory (56) References JP-A-5-12587 (JP, A) JP Hei 4-93610 (JP, A) JP-A-58-129495 (JP, A) JP-A-58-100876 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G21C 17 / 00 G01D 21/00 G05B 23/02

Claims (5)

(57) [Claims] A plurality of values of a main monitoring parameter;
Multiple dependent monitoring parameters for each value of the monitoring parameter
Storage means for associating data with the
And the dependent monitoring parameters detected from the storage means are stored in the storage means.
Dependent monitoring parameters based on the dependent monitoring parameters
A representative value of the parameter is obtained and stored in the storage means.
Update means for updating a dependent monitoring parameter to the representative value
And a main monitoring parameter detected from the monitoring target.
A plurality of subordinates updated by the updating means corresponding to the values.
Taking out the value of the generic monitoring parameter from the storage means
Means, and a plurality of subordinate monitoring parameters detected from the monitoring target.
Parameter values and a plurality of values extracted by the extraction means
Of the monitoring target by comparing the value of the dependent monitoring parameter of
An apparatus abnormality monitoring device comprising: abnormality detection means for detecting abnormality . 2. The method according to claim 1 , further comprising:
Multiple dependent monitoring parameters for each value of the monitoring parameter
Storage means for associating data with the
And the dependent monitoring parameters detected from the storage means are stored in the storage means.
Dependent monitoring parameters based on the dependent monitoring parameters
A representative value of the parameter is obtained and stored in the storage means.
Update means for updating a dependent monitoring parameter to the representative value
And the dependent monitoring parameters determined by the updating means.
Data is periodically stored, and a plurality of stored subordinate supervisors are stored.
Of the monitoring target according to the temporal change of the value of the visual parameter
An apparatus abnormality monitoring device comprising: abnormality detection means for detecting abnormality . 3. The method according to claim 1, wherein
The abnormality of the monitoring target is determined by the subordinate monitoring parameter in the operating state.
Data and the dependent supervisor stored in the storage means.
The distance difference from the pattern of the representative value of visual parameters
A device abnormality monitoring device that determines that an abnormality has occurred when a predetermined value is exceeded . 4. The method according to claim 1 , wherein a plurality of values of the main monitoring parameter are provided.
Multiple dependent monitoring parameters for each value of the monitoring parameter
And a learning mode for storing the data in association with each other
In the above, the normal condition read from the storage means is
Enter the parameter monitoring parameters and output
Multi-layer training that learns to be the quantization value of the monitoring parameter
Monitor in the neural network and monitor mode
The values of the dependent monitoring parameters detected from the elephant
Output and input when input to neural network
Comparing means for comparing the quantized values of the determined monitoring parameters
And anomaly detection that detects anomalies based on the magnitude of the comparison result
Equipment malfunction monitoring apparatus characterized by having means. 5. The method according to claim 1, wherein the monitoring object is used at the time of starting or at a steady state.
Multiple values of the main monitoring parameter and the main monitoring parameter
Multiple dependent monitoring parameters for each value of
First and second storage means for linking and storing;
Start-up or steady-state subordinate monitoring detected from the visual target
Parameters and stored in the first and second storage means.
Dependent monitoring parameters based on the dependent monitoring parameters
A representative value of a visual parameter is obtained, and the first and second descriptions are obtained.
Start-up or steady-state subordinate supervision stored in memory
The first step is to update the visual parameters to the respective representative values.
And a second updating unit, and an activation detected from the monitoring target.
Corresponds to the value of the main monitoring parameter during operation or steady state
The plurality updated by the first and second updating means
Of the dependent monitoring parameter of the first and second storages
Means for removing from the means, and
Multiple start-up or steady-state dependent monitoring parameters
And the value extracted by the first and second extracting means.
The value of the plurality of dependent monitoring parameters
An apparatus abnormality monitoring device comprising: an abnormality detection unit configured to detect abnormality of a visual target .