JPS58222311A - Abnormality diagnostic device - Google Patents

Abstract

PURPOSE:To attain abnormality diagnosis with high accuracy and reliability, by studying the past result of abnormality diagnosis and setting a threshold value where rate of mis-information and rate of missing information are equal as an optimizing value. CONSTITUTION:When an objective plant 1 is normal, a detecting value (y) of a detector 2 is not so much different from a reference value (x), an output ¦x-y¦ of an absolute value generator 5 is smaller than a threshold value epsilon, and a comparator 6 outputs an off-signal representing the normality. If the objective plant 1 is failed, the detected value (y) differs largely from the reference value (x), becomes larger than the threshold epsilon. and the comparator 6 outputs an on-signal representing the abnormality for displaying the abnormality.

Description

[Detailed Description of the Invention] The present invention is carried out in a thermal power plant or a nuclear power plant by comparing standard values in positive and dense states of pressure and temperature variables with detected values from an actual one runt. The present invention relates to a type of abnormality diagnosing device for diagnosing whether a vehicle has entered an abnormal state or is heading toward an abnormal state.

Assuming the judgment index QJ for abnormality diagnosis, various values can be considered for J. For example, the reference value obtained by a model representing normal state 1 is expressed as In other words, the absolute value of the difference between the calculated value and the detected value at the time is used as the determination index.

The judgment logic is such that if this J becomes larger than a predetermined threshold, it is judged as "abnormal".
, Now, in this type of abnormality diagnosis device, no matter what the judgment index is, it must be given to the threshold value 1' for determining "abnormality", but even in normal conditions, j) the reference value due to dynamic lag. (i) deviations due to variations in detected values due to noise in a); and deviations due to unavoidable model errors when creating normal reference values using a model. (Fl) The difference between the reference value and the detected value due to an abnormality occurring in the prevention of fire alarms is It is necessary to set a threshold value g=1 so that it will not be overlooked.

Conventionally, it was necessary to conduct many tests to see the error between the reference value and the detected value under normal conditions, and it was not possible to set all appropriate thresholds until after many tests, or it was not possible to perform a large number of tests. Therefore, it was not possible to determine an appropriate threshold value ti.

The present invention was proposed in view of the above circumstances, and has the objective of providing a highly reliable abnormality diagnosis device.
A method for diagnosing abnormalities in p1 runt by comparing the difference between a total of 7L electric current based on a model trap representing the behavior of plant variables in a normal state and measurement E measured from an actual plant with a set threshold value, and based on the magnitude relationship. Then, from the false alarm rate distribution and the fire alarm rate distribution corresponding to various past threshold values stored cumulatively, a threshold value at which the false alarm rate and the fire alarm rate are equal to y is determined, and this is output as the all set threshold value. A threshold setting device, and an updating circuit for updating the past false alarm rate distribution and/or fire alarm rate distribution cumulatively stored in the threshold setting device, which includes all three manual leads for false alarm, fire alarm, and linked axis. The next step will be tq # position.

One embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing its circuit configuration, and FIG. 2 is a diagram showing the operation of the threshold setting device of FIG. 1.

In the above diagram, the detected value ty is the value detected by the plant variable sound detector 2 of the plant 1 to be diagnosed for abnormality), while the reference value xf of the one runt variable in the normal state is generated by the reference signal generator 3, and the subtracter 4 By these 2
The difference in value, that is,
The output signal ε of the threshold setting device 7 is inputted to the input terminal of
The output of the comparator 6 is input to the display device 8.

The threshold setting device 7 receives all the outputs of the monitoring switch 9, and the monitoring switch 9 has three input capacitors corresponding to false alarm, fire alarm, and linked shaft, respectively. False alarm caused by being pressed 11. The fire alarm 12 and link shaft 13 are output, and when neither button is pressed, O1 is output. In other words, the monitoring switch 9 is an input terminal for learning whether the present abnormality diagnosis device has operated properly or has generated a false alarm or fire alarm.

In such a device, when an abnormality occurs in the target 1 runt 1, the detection value y detected by the detector 2 deviates by a large amount from the reference value The output of the comparator 5 is p larger than the output 6 of the abnormality judgment threshold setting device 7, and the output of the comparator 6 is an ONN signal ratio output representing 1 abnormality 1, and the display device 8 indicates that the output is 1.
Displays that there is an abnormality 1.

On the other hand, if target 1 runt 1 is in a normal state, detector 2
The detected value y is not much different from the output of the signal generator 3 that represents the normal state, that is, the reference @X, and the output 1 of the absolute value generator 5 is
x-yl is smaller than threshold 6, and the output of comparator 6 is 1
An OFF signal indicating "normal" is output, and the display device 8 displays "1 normal".

Next, to explain the calculation details of the threshold setter 7, the output ε of the threshold setter 7 is not a continuous value, but a discrete value ε1.8! ...Choose t from n of εn. Now, ε = t l (1 = xl @ 2..., n) t and the number of vA notifications when using this abnormality diagnosis 6 equipment ft - t
−k j , the number of fire alarms 2/l, the number of presses of the monitoring button kmle, the false alarm rate tαi, and the skipping rate tβsu.

At this time α1-kl/m1 β1-no1/mi (1
) There is a relationship.

Now, the performance of Threshold E setting device 7! The following are 1) A calculation that learns all the abnormality diagnosis results every time an input is received from the monitoring switch 9. 2) A calculation that calculates the optimal threshold %r from the learned result.The contents are 1) Learning calculation monitoring switch If the man power from 9 is 0, do nothing.

If the input from the monitoring switch 9 is v4 report 11, ki,
αi and β1 are calculated using equation (1) with mi@increased by 1 and All left unchanged.

If the input from the monitoring switch 9 is open phase 12,
m i is increased by 11-1, kl is set to 11, and par and β1 are calculated using equation (1).

Monitoring switch 9 r) r6 Human power is suitable @13! Raba, ml
Increase by 11-1 and k 1.11 that! As a hawk, αi and βit are calculated using equation (1).

2) Optimal! l calculation operation This calculation content is Iα amount - β for i-1, 2...n
Find the I that minimizes 11 (2) For example, if you omit it in Fortran style, IMIN-1 Dφ1i=s2*n IP(4α i −β 11.GT, l α quantity −1
-β i-11) The calculation result of GφTφ I IMIN-Flea CφNTINLIEf (DIM・, IN is l that satisfies (2)′.

1 To describe the operation of the above operations, the premise is that the threshold value C
is large, that is, unless the deviation between the detected value y and the reference value X becomes significantly large, it is not determined that there is an abnormality, the false alarm rate is small, but the fire alarm rate is high.

On the other hand, if the threshold (ml) is small, even a slight deviation between The relationship of Kisen 6i is as shown in Figure 2.

From the perspective of setting the threshold value, to reduce the false alarm rate, the threshold value should be set thicker, and to reduce the fire alarm rate by 1, the threshold value should be made smaller. This is what happens when a false alarm occurs, αM increases and βI decreases. Therefore, since the α1 curve in FIG. 2 is upward and the β1 curve is downward, the intersection point of these moves to the right and Mk increases.

When a fire alarm occurs, α5 decreases and β1 increases. Therefore, since the cci song #μ in the same figure has a bottom of 9% and a 11 curve rises to the top, the defect moves to the left and ε becomes smaller.

In the case of a continuous phase, both αl and βl decrease, and the αi and β1 curves in the same figure both fall downward, and tl changes to e. 8 also becomes difficult to move when an alarm or fire alarm occurs.In other words, the 8i learns from the past experience of being in a continuous phase.

A device like Jin learns all past abnormality diagnosis results and solves the problem of setting thresholds for abnormality judgment, which has traditionally been difficult to set appropriate values and required many preparation tests, to calculate false alarm rates and fire alarms. Is there a function to set the optimal value as a compromise between the two rates? By adding this, it is possible to automatically set the optimal threshold value and eliminate the drawbacks of conventional methods, such as being able to deal with changes such as changes over time in the diagnostic target.

It produces an improving effect.

In the above embodiment, 1x-
I took ylk, but there is no particular meaning to it.% IX
-F+'', the temporal deviation between the true temporal change curve and the y change curve, or the distance between double-sided lines, etc. Any bone may be used as long as it represents the deviation in x and y.

In short, according to the present invention, the value calculated by the model representing the behavior of the plant xi in the normal state and the measured value measured from the actual 1 runt are compared with the difference sound setting threshold, and the magnitude relationship (depending on whether the plant When diagnosing a problem, find the threshold value at which the false alarm rate and the fire alarm rate are equal to y from the false alarm rate distribution and the fire alarm rate distribution corresponding to various accumulated past threshold values. It has a threshold setting device that outputs as a standard, and a three-manpower Qw for false alarms, fire alarms, and weekly reports, and updates the past false alarm rate distribution and/or fire alarm rate distribution cumulatively stored in the threshold setting device. The present invention is extremely useful in industry because it provides a highly accurate and highly reliable abnormality diagnosis device by providing an update circuit.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention, and FIG. 2 is a diagram showing the operation of the threshold setting device of FIG. 1. DESCRIPTION OF SYMBOLS 1...Target plant, 2...Detector, 3...Reference value signal generator, 4...Subtractor, 5...Absolute value generator, 6...Comparator, 7...・Threshold E setting device, 8...
・Display device, 9... Monitoring switch, 11... False alarm,
12...fire alarm, 13...coupled axis. Applicant Sub-Agent Takehiko Hito Suzue 1 (:1゜Figure 1 Figure 2

Claims (1)

[Claims] Calculated values from a model representing the behavior of one runt summer number under normal conditions and an actual plant! Dtl-i11 The difference between the measured value and the set threshold value is compared with the set threshold value to diagnose abnormalities in the plant based on the magnitude relationship, and the false alarm rate distribution and It has a threshold setting device that calculates the threshold value at which the false alarm rate and the fire alarm rate are equal to each other from the fire alarm rate distribution and outputs this as the meeting-set threshold, and three input buttons for false axis, fire alarm, and weekly report. and an update circuit for updating the past false alarm rate distribution and/or fire alarm rate distribution cumulatively stored in the threshold setting device.
``Abnormality diagnostic device that detects % retention.