JPS61173313A - System of presuming cause of trouble of plant - Google Patents

Abstract

PURPOSE:To presume the cause of troubles of machines simply by comparing an abnormality pattern of a sensor indicated actually with a presumed abnormality pattern by using the causal relation between time lapsed from the first detection of abnormality and the troubles. CONSTITUTION:An abnormality detecting device 107 compares a reference signal stored to judge the normalcy and abnormality of a signal with a detection signal, and outputs a signal of normalcy or abnormality of the detection signal 105 to causes of trouble presuming device 109. The device 109 checks the signal 108 at a fixed time interval, and when the signal changed from normalcy to abnormality, stores the time. Further, the device 109 finds out the time required from the first detection of abnormality in other sensors for each machine based on the signal 113, and stores in the inside. When an operation command signal 111 is inputted from an operator's console 110, the device 109 presumes the causes of troubles based on time of operation command responding to the operation command signal 111 and the time required from the signal 108 to the detection of abnormality and displays 115 a signal 114.

Description

Detailed Description of the Invention [Field of Application of the Invention] The present invention provides a method for detecting a failure when a single failure (one cause of failure) occurs in a system consisting of a plurality of devices such as an atomic couplant or a chemical plant. The device causing the problem,
This paper relates to a method for estimating plant failure causes based on information from a limited number of sensors.

[Background of the invention]

Conventional failure cause estimation methods include a ``list processing method'' in which the observed abnormality pattern of a sensor is determined by which failure cause it is based on a decision table created in advance that associates abnormality patterns with failure causes. There was a method based on causality that traces the propagation path of anomalies (for example, Eiji Oshima: Simulation and Anomaly Diagnosis, Journal of the Institute of Electrical Engineers of Japan, Vol. 99, No. 3 (1973), PP, 19
5 to 199), the list processing method is
Although high-speed processing is possible because the calculations are simple, ``even if the cause of an abnormality is the same, completely different patterns will be observed depending on the time it takes to discover it, so a very large number of patterns must be prepared ( (quoted from the above literature), which limited the scope of application.On the other hand, methods based on causal relationships do not require all patterns to be prepared in advance, but the calculations become complicated. There was a problem that the processing time became long.

[Purpose of the invention]

An object of the present invention is to solve the above-mentioned problems of the prior art, in which calculations are simpler than methods based on causal relationships, and there is no need to prepare all abnormal patterns in advance as in methods based on list processing. Therefore, it is an object of the present invention to provide a plant failure cause estimation method that can be applied to a larger target than conventional methods.

[Summary of the invention]

In order to achieve the above object, the present invention, when an abnormality is detected, uses the elapsed time since the first abnormality detection and the causal relationship of the failure, and currently indicates the cause of the failure for each device. Find the predicted abnormality pattern of the sensor Next, compare the abnormality pattern of the sensor currently displayed with the predicted abnormality pattern, and if they match, the device is assumed to be the cause of the failure. There is.

In this method, after an abnormality is detected, an abnormal pattern corresponding to the elapsed time is generated, so there is no need to prepare all abnormal patterns in advance, unlike the conventional list processing method.

Further, the calculation process of comparing the generated abnormal pattern with the currently displayed abnormal pattern is simpler than the calculation process of conventional methods based on causal relationships.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described in detail with reference to FIGS. 1 to 11.

FIG. 1 shows the configuration of an embodiment of a plant system that implements the plant failure cause estimation method according to the present invention.

In FIG. 1, a plant 101 includes a plurality of component devices 102 and sensors 103 for detecting the status of some of the components. These sensors 103 monitor the operating status of each component 102, such as flow rate and temperature 2.
A signal 104 such as a frequency is detected, and a detection signal 105 is output to an abnormality detection device 107 comprising an abnormality detector 106 corresponding to each sensor 103. The abnormality detection device 107 includes
A reference signal for determining whether each signal is normal or abnormal is stored in advance, the reference signal is compared with the detection signal 105, and a signal 108 indicating whether each detection signal 105 is normal or abnormal is output to the failure cause estimation device 109. do. Failure cause estimation device 10
9, the signal 108 is checked at regular intervals, and if the signal changes from a normal signal to an abnormal signal, the time is stored. In addition, from the initial data input device 112 to the 2
Failure propagation relationship between devices (spread direction, maximum propagation time).

The signal 113 corresponding to the minimum propagation time) is input into the failure cause estimation device 109, and the failure cause estimation bag [109,
Based on the signal 113, the time required for each device from the first abnormality detection in the case of the true cause of failure to the abnormality detection by other sensors is determined and stored internally. When the calculation instruction signal 111 is input from the operator console 110, the failure cause estimation device 109 inputs the calculation instruction signal 11.
1, the time when the signal 108 and the abnormality were first detected, and the time required to detect the abnormality stored internally in advance, the cause of the failure is estimated and the signal 114 is displayed on the display device. The output is displayed on 115.

Note that once the initial data is entered, there is no need to re-enter it unless the data is changed.

FIG. 2 is a flowchart showing an example of the processing procedure in the failure cause estimation device [1109 of FIG. 1.

An embodiment of the present invention will be described below using an example of application to a plant in which failure propagation relationships among plant component equipment are represented by a network in FIG.

In the network shown in Figure 3, the nodes represent plant component equipment, and the direction of the arrow represents the direction in which the effects of a failure will directly spread, and the nodes surrounded by squares can be used to determine whether or not the equipment is abnormal. Represents equipment.

When the influence of a failure directly spreads between two devices in such a plurality of devices (19 in Fig. 3), the direction of the spread, the maximum propagation time, and the minimum propagation time are given, and this relationship '
is expressed in a matrix as shown in FIGS. 4 and 5. For example, 100 in the first row and second column of matrix A indicates that the influence of a failure directly spreads from device 1 to device 2, and a maximum of 100 seconds elapses at that time. 90 in the first row and second column of matrix B indicates that at least 90 seconds have elapsed. Therefore, the direct propagation time of the failure effect from device 1 to device 2 is from 90 seconds to 10 seconds.
It means that it is for 0 seconds. A blank space means that there is no direct influence, and is treated as ψ in calculations.

The maximum failure propagation time input from the initial data input device 112 in FIG. 1 to the failure cause estimation device M109 is the matrix A in FIG.
The minimum failure propagation time is input in the form of matrix B in FIG.

First, in step 116 of FIG. 2, the maximum fault propagation time matrix is calculated. Using the maximum adjacent fault propagation time matrix A in FIG. 4, find the shortest fault propagation time from all devices to the sensor. This is a solution method for finding the known shortest distance and shortest path (Masao Iri, Takashi Furuta: Network Theory 2nd Science and Technology Union Publishing (1976), pp, 34-3
7). The obtained propagation time is expressed in a matrix as shown in Fig. 6. For example, 100 in the first row and first column of matrix C means that the time required for the influence of a failure to spread from device 1 to device 2 is at most 10.
Indicates that it is 0 seconds. A blank part means that there is a possibility that there will be no ripple effect, and it is treated as ω in calculations.

Next, a minimum failure propagation time matrix is calculated in step 117 of FIG. Minimum adjacent fault propagation time matrix B in Figure 5
Similarly to step 116, find the shortest failure propagation time from all devices to the sensor using . The obtained propagation time is expressed in a matrix as shown in Fig. 7. For example, 90 in the first row and first column of the matrix indicates that it takes at least 90 seconds for the influence of a failure to spread from device 1 to device 2. The blank part means that there is no ripple effect, and in calculation, ω
treated as

Next, in step 118 of FIG. 2, a maximum abnormality detection time matrix is calculated. Find the minimum value for each row element of the matrix in Figure 7.6 For example, the minimum value for the first row is 90, which is the minimum value until the first abnormality is detected when an abnormality occurs in device 1. 0 meaning delay time
Next, the minimum value of the uncorresponding rows of the matrix is subtracted from the elements of each row of matrix C in FIG. The obtained results are expressed in a matrix as shown in Figure 8.4. For example, for eta in the 1st row and 1st column, -e12==C,, 1lin (diJ) = 1
00-90=10, which means that it takes at most 10 seconds from when an abnormality occurs in the device 1 and is first detected until the sensor 2 detects the abnormality. The empty sum part means that even if an abnormality occurs, there is a possibility that the abnormality cannot be detected, and is treated as ω in calculations.

Next, in step 119 of FIG. 2, a minimum abnormality detection time matrix is calculated. The minimum value is found for the elements in each row of matrix C in FIG. For example, for the first row, the minimum value is 100, which means the maximum delay time until the first abnormality is detected when an abnormality occurs in device 1. Subtract the minimum value of the corresponding row of matrix C from the elements of . The obtained results are expressed in a matrix as shown in FIG. 9. For example, f if in the 1st row and 2nd column
For f, fxa=dti 1lin (C1J
)=180-100=80, which means that the time required from when an abnormality occurs in the device 1 and the abnormality is detected for the first time to when the abnormality is detected by the sensor 3 is at least 80 seconds. A negative value means that there is a possibility that an abnormality will be detected first. Furthermore, a blank section means that even if an abnormality occurs, it cannot be detected.

As described above, steps 116 to 119 in FIG. 2 are executed when matrix A in FIG. 4 and matrix B in FIG. 5 are input from the initial data input device 112 in FIG. 1.

The failure cause estimation device 1 uses matrix E in FIG. 8 and matrix F in FIG.
109.

In other words, in the network shown in FIG. 3, the devices that are emitting abnormal signals are 2, 3, 9, 18, and 19, and the devices that are emitting normal signals are 7, 10, 11, and 15°17. Further, it is assumed that the elapsed time from the first detection of an abnormality until the instruction signal for starting calculation is input from the operator console 110 to the failure cause estimating device 109 is Tm2O3 seconds.

First, in step 120 of FIG. 2, the minimum abnormality pattern is determined. Elements of the matrix E in FIG. 8 whose elapsed time is less than or equal to T are set to 1, and those whose elapsed time is greater than T are set to O. As a result, matrix G shown in FIG. 10 is obtained. For example, the abnormality pattern in the first row indicates that if equipment 1 is the cause of the failure, sensors 2, 3, 9, and
The column corresponding to 10°11.17 is 1. It is unknown whether the sensor corresponding to the column O indicates abnormality or normality.

Next, in step 121 of FIG. 2, an open fire abnormality pattern is determined. Elements of the matrix F in FIG. 9 that are less than or equal to the elapsed time T are set as 1, and those that are larger than T are set as O. As a result, matrix H shown in FIG. 11 is obtained. For example, in the abnormality pattern in the second row, if equipment 2 is the cause of the failure, the column corresponding to sensor 10°11.15.17 that never shows an abnormality after 300 seconds has passed since the first abnormality detection is O. It has become.

Next, in step 122 of FIG. 2, the abnormality pattern is compared. If the currently displayed abnormality pattern is between the minimum abnormality pattern and the maximum abnormality pattern obtained in steps 120 and 121 of FIG. Find all of them and consider them as the cause of the failure. That is, if sensor j currently indicates an abnormality, z = l, and if it is normal, xJ = o, then equipment i for which the following equation holds true for all sensors j is considered to be the cause of the failure.

g, J≦xJ≦hlJ In this example, (X,, !,, X,, xst x
,. .

x1□,xl,,Xl,,x1@,xl,)=(1,
1,0°1, O, O, O, 0, 1, 1), it can be seen from matrix G in FIG. 10 and matrix H in FIG. 11 that equipment 2 is the cause of the failure.

According to the above-mentioned embodiment, (1) The cause of failure can be estimated even if the number of sensors that can be installed is limited for Plan 1-, which consists of multiple devices. (2) The cause of failure can be estimated even if the number of sensors that can be installed is limited. To use the maximum and minimum values,
Accurate failure cause estimation is possible. (3) After an abnormality is detected, an abnormality pattern is generated according to the elapsed time, so less data is required to be stored in advance. Since only abnormal patterns are compared, it has the effect of being able to be executed at high speed.

As a modification of this embodiment, the scope of application of the present invention can be extended to a system that can be modeled as shown in FIG. 3, for example, a computer network system.

〔Effect of the invention〕

As explained above, according to the present invention, (a) After an abnormality is detected, an abnormality pattern is generated according to the elapsed time, so less data is required to be stored in advance.

(b) Since the calculations after detecting an abnormality consist only of generating an abnormal pattern and comparing the abnormal patterns, it can be executed at high speed.

There is an effect.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of a plant system using the failure cause estimation method of the present invention, FIG. 2 is a flowchart showing the processing procedure of the failure cause estimation device of FIG. 1, and FIG. A diagram showing the relationship and the position of installed sensors, Figure 4 is a diagram showing the maximum adjacent fault propagation time matrix, Figure 5 is a diagram showing the minimum adjacent fault propagation time matrix, and Figure 6 is a diagram showing the maximum fault propagation time matrix. , Fig. 7 shows the minimum fault propagation time matrix, Fig. 8 shows the maximum anomaly detection time matrix, Fig. 9 shows the minimum anomaly detection time matrix, and Fig. 10 shows the minimum anomaly pattern matrix. 11 are diagrams showing the maximum abnormal pattern matrix.・ 1(,j

Claims (1)

[Claims] In a plant failure cause estimation method that estimates the equipment that causes the failure based on information from sensors installed on limited equipment in a system consisting of multiple equipment, the initial input is the failure propagation direction and propagation time between two pieces of equipment. Based on the above, the sensor abnormality is generated when each device is the cause of the failure, according to the elapsed time from the time when the first abnormality was detected by the sensor, and the generated abnormality pattern and the abnormality and normal signals input from the sensor are generated. A plant failure cause estimation method characterized by estimating a failure cause based on the degree of matching with a current abnormality pattern determined from the above.