JPS5890122A - Inferring system for in-plant trouble cause - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a method for identifying the source of failure when a single failure (one failure source) occurs in a system consisting of a plurality of devices such as an atomic couplant, a chemical plant, and a water and sewage plant. This invention relates to a method for estimating the source of a plant failure from information from a limited number of sensors.

Various methods have been known to date as a fault diagnosis method for investigating failures in plant equipment, but these methods (1) are based on the assumption that the sensor can be installed on the equipment that is the source of the failure; (2) ) A causal relationship between state quantities such as pressure and flow rate and an abnormal phenomenon (failure source). For example, if the pressure is low and the flow rate is low, create the relationship K (crack) in the piping in advance and check the relationship between the sensor and the abnormal phenomenon (failure source). The characteristic is that abnormalities are diagnosed by comparing them with the current values of pressure, flow rate, etc., which are currently available.Therefore, under the assumption (1), the abnormal phenomena that can be diagnosed based on the types of existing sensors are limited to a small number. , (2) has the disadvantage that it is difficult to create causal relationships when dealing with large-scale and complex plants.

The purpose of the present invention is to solve the problems of the above-mentioned prior art by using information from a limited number of installed sensors.
The object of the present invention is to provide a complete fault source estimation method that can accurately estimate the true fault source even in a shoddy plant and efficiently order the possibilities of the fault source.

°Such a purpose’! f- To achieve, the present invention:
In a system consisting of multiple devices, by creating a direct fault propagation relationship between two devices and performing matrix operations,
Direct and indirect ripple relationships between all devices are determined, abnormal information and normal information obtained from sensors are applied to this ripple relationship, and all sources of failure are estimated. In addition, in the case where there are multiple estimated failure sources, the present invention can identify the failure source between the multiple estimated failure sources based on the failure rate of each device and the probability of failure spreading between two devices. This allows questions to be prioritized in order of likelihood.

1''J, Do, This study's fruit Mli N All figures 1 to 7 are used for revision purposes.

FIG. 1 shows the configuration of an embodiment of a plant system using the failure source + ifl determination method according to the present invention.

, 11, a plant 101 includes a plurality of component devices 102 and sensors 103 for detecting the status of some of the devices. These sensors 103
detects the operating status of each +tit component 102, such as a current meter, temperature, frequency, etc. signal 104, and detects the detection signal 1
05 is output to an abnormality detection device 107 consisting of an abnormality detector 106 corresponding to each sensor 103. Abnormality detection device t1
07 stores in advance a reference signal for determining whether each signal is normal or abnormal, and the reference signal and detection signal 10
Signal 1 indicates whether each detection signal 105 is normal or abnormal.
08 is output to the failure source estimation device 109. The failure source estimation device 109 uses the signal 108 and the initial data input device 110.
Based on the failure propagation direction, propagation time, propagation probability, and the signal 111 corresponding to the failure rate of each device inputted from the Output.

Note that once the initial data is input, there is no need to input it again unless the data is changed.

Figure 2 shows the processing in the failure source estimating device f109 in Figure 1.
It is a flowchart which shows an example of (At, re) of ul.

Figure 3 shows IJPG (Squifie) consisting of two systems.
dpetroleum Qas) is a system diagram showing equipment names and system names in an example of pump equipment of a plant.

As shown in Figure 3, the fl of the pump equipment is
PG main pipe 1.18, 19, sluice valve 2.5゜10
, 13, LI'G pump 3, 11, power supply 4゜12,
Flow Transmitter Mitsutaro, 14, flow & indicator controller 7.1
5, Valve with air motor handle 8, 16, Air operated cutoff valve 9
, 17.

In such cases, if a failure directly spreads between two devices, give the direction and time of the spread,
This function f,%' is expressed in a matrix as shown in FIG. This matrix *A=[a++:], ' l J = 11
2+..., 21, for example, 1st row 2 of matrix A
100 in the column indicates that the influence of the failure directly spreads from device 1 to device 2, and the time period is 100 seconds. Blank areas indicate that there is no direct influence, and are treated as O in calculations. - Also, 20 and 21 represent a dummy manual node and a dummy output node, respectively, and are added for convenience of calculation.

When the fault propagation relationship between devices represented by matrix A is rewritten in a network form, it becomes as shown in FIG. In the diagram,
1121..., 21 correspond to each component 1, and the number above the arrow indicates the time during which the influence of the failure spreads.

The above matrix A represents the direct failure propagation relationship (propagation time and propagation direction) between two devices. In the following, as the first step in estimating the fault source, only the propagation direction will be displayed in its entirety, including indirect propagation relationships in which the fault spreads to other elements via all the elements affected by the fault. To do this, first, in order to simplify the following calculations, the matrix Ak is converted into a matrix B whose elements have values of 0 or 1. That is, in matrix A, the value of an element corresponding to an element having a certain value is set to 1, and the other values are set to 0. This operation is performed in block 114 of FIG. 2, and the result is shown in FIG.

Next, add the identity matrix ■ to B, and (Il+I)j = (
B+I)J", and all fault propagation paths are constructed as shown in block 115 of FIG. 2. The result is shown in FIG. 7, and this matrix is called the reachability matrix and B do.

An element of matrix ■3 with a value of 1 is transferred from the faulty element corresponding to the row number to the faulty element corresponding to the column number up to j
-This shows that the influence of a failure will definitely spread through one failure element. On the other hand, those with a value of 0 are row i
This shows that the influence of a failure never spreads from the element corresponding to the column number 'lT to the element corresponding to the column number.

In the following, signal 10, as block 116 in FIG.
A method for estimating the source of failure using abnormality information as shown in Section 8 will be described. However, as shown in the box in Figure 5,
All the devices equipped with sensors are 9.19 and 15.17, and the devices emitting abnormal signals are i9 and 19, and the devices emitting normal signals are 15.17.

To estimate all the failure sources, the abnormality is detected, X, and
. Follow the arrows that lead to , and select all of the common rocky shores and knees that follow each route. As a method for group performance, examine the 9th column corresponding to xo and the 19th column corresponding to II9 in Figure 7, and find the rows that have a value of 1 in common, that is, 1, 2, 3, 4.゜1, 2, 3 corresponding to 20 lines
, 4.20 is presumed to be the source of the failure. Excluding 20, which is the 116 point of the dummy in Hata et al., the estimated failure sources are 1 and 2.
, 3.4.

Next, in block 117 of FIG. 2, it is determined whether or not any of the estimated failure sources should be excluded from the estimated number 1 umbrella sources using the following method based on the equipment 15.17 that is operating normally. Check it out. To check, first check 15.17
Follow all the arrows in the opposite direction to find all the devices that are concentric with the above estimated failure source. Next, the shortest time t for several routes from device 1 to abnormal device 9? ＋－“＝
Ask for 300. Similarly, calculate 1711~=400. Thirdly, “Kat” 7’+”of!: k ratio soft L
, Maximum (D (fll ”°8tTIn = 400 is determined. maze, min, , In the case of failure mk Then, the longest time t , m〕1 = 41 for the route from device 1 to normal device 15
Find 0. Similarly, find 'I'l'7=420 (
If you want to shorten the calculation time, it is possible to find the shortest total time that is easy to calculate, but in this case, 11. Insufficient checking of fixed failure sources.
b Le. )o iiJ'45K,'7'lxlto17"
l"7 (!: 'e ratio 1 bite, minimum) Hani m"t",
Find ``= 410= , I+1m8.ml ,
tllaK is, if X is acid 11? This indicates that if it is an S source, an abnormality will occur in X5w after 410 seconds at the latest. However, whether an abnormality will definitely occur after 410 seconds is determined by the failure propagation probability of 1 from Xl to XII+.
．． Since there is a difference between 0 and 0, next we perform all tests N=J from the propagation probability. That is, as the sixth factor, the probability II J' that the influence of the failure will spread from equipment 1 to equipment 15- is N
-13'J. II P (1, (In the case of l, failure occurrence i. Since there are cases where an abnormality does not occur in X4 even after 410 seconds, equipment 1) is included in the I1mo constant fault 11?F source.

, /7 p = 1. , 0 (7) If m” t7
1+t"",", axt gold ratio i1RL, maw 1.
If ≧n□Ink, then device 15 should have an abnormality, but it shows a normal value, so we consider that device 1 is not the source of the failure, and determine whether device 1 is the source of the 1111 failure (9) and 1dfL. ra*x tT"< ml°t7X i
Whether or not an abnormality occurs will depend on the passage of time, and it cannot be determined at this point that it is not the source of the failure, so it shall be included in the total estimated failure sources of equipment 1.

In this example, if pHP = 1.0, maxtr”
-400 ("tT" = 410 and 7Th, so device 1
are included in the probable failure sources.

Next, in block 118 of FIG. 2, the estimated failure source is calculated using the failure rate PI of each device and the probability of failure spread between devices P+ j
k is used to prioritize the probability of the failure source using the following method.

First, the element X! which is the most downstream among the multiple estimated failure sources! (= Select (equipment 3).Next, calculate the down payment below for the route from the estimated failure source other than X3 to
: It blows up its own money. H-1')'L ILL:I
It is assumed that the thicker the value, the higher the priority (the possibility that the acid is 13? 50?).

In the example, (10) Pl・P3,2・P2,.

P,・P2. .

3 P4 ・P, 1゜ The values of will be compared.

Although the above prioritization method is strictly a 1 calculation, it requires a long calculation time as the network becomes large.

The following two methods can be considered in order to shorten the number of chanting times, although they are not very precise.

(1) Prioritize each of the four sources based on the failure rate. In the example, PI I P21 Pg 1P
The down payment ratio of 4 is soft.

(2) Due to the 11F foot of multiple states (II; Elements X, A, which are the most downstream from the C source.Next,
and X4, t's, s l J'4. Check whether m is greater than a given threshold ID'j.

Those below the threshold are ・chiru1ru)i1, for example l,',
, s = (15(threshold=(), 8, I's, 5-I)
-9>Threshold = 0.8, x2 and
For X4 and X4, the method of prioritizing strictness (11) is applicable.

According to the above-mentioned embodiment, (1) the source of the failure can be estimated even if the number of sensors that can be installed is limited even if the target is a plant consisting of a plurality of 11+'i fittings, and (2) the source of the failure can be estimated. This has the effect of being able to display items in descending order of probability.

As described above, according to the present invention, (1) failures can be estimated even in systems that include equipment in which sensors and abnormality detection devices are not installed, (2) normal signals and failure propagation probability are used. Then, the estimated failure sources obtained in (1) can be narrowed down to the awns. (3) Using the failure IS rate and failure propagation probability of each safety device, it is possible to prioritize the possibility of being the failure source. Possible.

Therefore, it is possible to determine the source of the failure, which was not possible with conventional methods, and it is possible to easily take countermeasures in the event of a failure.

[Brief explanation of the drawing]

(12) Figure 1 is a configuration diagram of an example of a plant system that fully implements the fault σ≦ estimation method of the present invention, and Figure 2 is the fault source J11L in Figure 1.
A flowchart showing an example of the process Jl in a fixed installation jh: Figs. Equipment system diagram, Figure 4 is a diagram showing the failure propagation probability matrix, Figure 5 is a diagram showing the position of the failure propagation network and installed sensors of pump equipment, and Figure 6 is a diagram showing the pool matrix of the failure propagation related matrix. , No. 7181 shows the town arrival matrix of the failure propagation related matrix.

Claims (1)

[Claims] 1. In a plant failure source estimation method that estimates a device that is a failure source based on sensor information corresponding to a limited number of devices that are made up of a plurality of devices, A system for estimating the source of failure in a plant, in which the number of failures and their relationships are determined, and the source of the failure is estimated based on the determined failure propagation probability and the abnormality and normality information obtained from the sensor information. . 2. A plant failure source estimation method, which is characterized in that the estimated failure sources are ranked based on at least one of the failure rate of each device and the failure propagation probability between devices.