JPS60107111A - Estimating system for fault influence range of plant - Google Patents

Abstract

PURPOSE:To decide easily the order of countermeasure at the time of fault by calculating the probability of the temporary fault influence simultaneously in a step of revising the temporary fault affecting time to the assembly of non affected elements. CONSTITUTION:A plant 101 consists of plural component devices 102 and sensors 103, and the detection signal 105 of the action state (flow rate, etc.) of the plant is supplied to a fault detectors 107 to be compared with a reference signal for decision of fault which is previously stored. This comparison signal 108 is supplied to a fault influence range estimating device 109 and checked at every fixed period of time. Then the changing time point into a fault signal is stored. When an estimated time point designation signal 111 is delivered from an operator console 110, the device 109 obtains the devices which are included within a fault influence range at a designated and estimated time point in accordance with the signal 111 and an influence probability signal 113 supplied from an initial data input device 112 as well as a stored time point. Then the device 109 delivers the numbers of the obtained devices to a display device 115.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention provides a method for determining whether, when a failure occurs in some of the component equipment in a plant, the influence of the failure spreads to any range of equipment in the plant after a certain period of time. This invention relates to a system for predicting the extent of failure of a plant, which predicts whether a failure will occur and displays the state of the predicted influence.

[Background of the invention]

The closest known example to the present invention is "
"Plant failure influence range prediction/indication device"
-160474) and “Plant Failure Impact Range Prediction Method” by Satoshi Miyazaki and 4 others (Patent Application No. 11546/1982)
No.). However, the former method required (1) sensors to be installed in all components of the plant;

(2) It is assumed that there is only one cause of failure, but it cannot be applied when there are multiple causes of failure.

(3) Compound failures caused by a combination of two or more failure causes cannot be handled.

There was a problem.

Furthermore, the latter method could not calculate the failure propagation probability, which could be calculated using the former method. As a result, it is difficult to distinguish between a fault spread range with a high spread probability and a fault spread range with a low spread probability.
There was a possibility that the order of measures would be incorrect.

[Purpose of the invention]

An object of the present invention is to solve the above-mentioned problems of the prior art by using information from a limited number of installed sensors to determine the range in which the influence of the failure will spread from the abnormality detection point after a certain period of time, even in the event of a complex failure. An object of the present invention is to provide a plant failure influence range prediction method that can predict equipment and calculate its influence probability.

[Summary of the invention]

In order to achieve the above object, the present invention is characterized in that it is able to handle complex failures but is unable to calculate the failure propagation probability in the conventional method (% Application No. 58-11546).

Conventional method that can calculate failure propagation probability
No. 0474) targeted only single failures (one cause of failure). However, since the present invention deals with multiple failures (multiple failure causes) and complex failures, the failure propagation prediction path from the abnormality detection point to the target device (in the case of a single failure device, one shortest failure propagation time path) By finding the union of the above branches and calculating the product of all the failure propagation probabilities corresponding to the branches, we can calculate the probability of failure propagation to the predicted device. For example, the failure propagation relationship between plant components is the first
The network shown in the figure (arrows indicate failure propagation directions, ○ indicates single failure equipment, ◎ indicates complex failure equipment), and the abnormality detection point is equipment 2.16. , if the predicted failure propagation path is shown by the bold line, the failure propagation probability from device 1 to j is P
+j, a single faulty device (behind the adjacent device,
When the effect of a failure spreads from at least one device, the probability of the effect spreading to the next device) is P2
13 XPs+e, complex faulty device (device in which the effect of a failure from that device spreads to other devices only when the effect of a failure spreads from all the devices that directly affect the target device) 19. is P2+3 ×P3+
8×PM+19×PIL+9 The present invention is characterized by the method of calculating the failure propagation probability described above.

[Embodiments of the invention]

Embodiments of the present invention will be described in detail below with reference to FIGS. 1 to 13.

FIG. 2 shows the configuration of an embodiment of a plant system that implements the plant failure influence range prediction method according to the present invention.

In FIG. 2, 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 monitor the operating status of each component 102, such as flow rate, temperature,
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, and the reference signal and the detection signal 105 are compared, and the signal 108 indicating whether each detection signal 105 is normal or abnormal is determined by the failure influence range prediction device. Output to 109. The failure influence range prediction device 109 checks the signal 108 at regular intervals, and if the signal changes from a normal signal to an abnormal signal, the time is stored. When the predicted time designation signal 111 is output from the operator console 110, the failure influence range prediction device 109 calculates the signal 108, the time at which the normal signal changes to the abnormal signal, and the initial data input. Based on the failure propagation direction, propagation time, propagation probability, and signal 113 corresponding to the distinction between complex failure devices and single failure devices for each plant component device inputted from the device 112, the failure propagation range is determined at the specified predicted time. A signal 114 indicating the equipment number and failure propagation probability is output to the display device 115.

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

FIG. 3 is a flowchart showing an example of the flow of processing in the failure influence range prediction device 109 of FIG. 2.

DESCRIPTION OF THE PREFERRED EMBODIMENTS 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 the network shown in FIG.

In the network shown in FIG. 1, the nodes represent plant components, and the direction of the arrow represents the direction in which the influence of a failure directly spreads. Also, ○ indicates a device with a single failure, and ◎ indicates a device with multiple failures.

When the influence of a failure directly spreads between two devices in such a plurality of devices (19 in Figure 1), the direction and time of the influence are given, and this relationship is expressed as shown in Figure 4. Express a matrix. Let this matrix be A=[atj] l i, J
:=1.2. -+19, for example, 100 in the first row and second column of the matrix N indicates that the influence of the failure directly spreads from device 1 to device 2, and that the time is 100 seconds.

A blank space means that there is no direct influence, and is treated as ■ in calculations.

Furthermore, the propagation direction and propagation probability are given when a failure directly propagates between two devices, and this relationship is expressed in a matrix as shown in FIG. This matrix is B=[bB].

'+3==l, 2. ..., 19, for example, 1 in the first row and second column of matrix B indicates that the probability that the influence of a failure will directly spread from device 1 to device 2 is 1. A blank space means that there is no direct influence, and is treated as 0 in calculations.

As data from the initial data input device 112 in FIG.
The failure propagation direction, propagation time, propagation probability between plant component devices, and the distinction between complex failure devices and single failure devices for each component device are input, and these data are stored in the storage device in the failure spread prediction device 1109. I'll keep it. By checking the sensor signal at regular intervals, the sensor records the distinction between normality and abnormality of the device S whose abnormality can be detected, and the time when it changes from normal to abnormal. Even when returning from abnormality to normal, the value of 1.1 does not change.

When a predicted time designation signal is input from the operator console 110 in FIG. 2, the input predicted time 1o is stored and the processes from step 116 in FIG. 3 are executed.

Assume now that abnormal signals are being output from devices 2 and 16 in FIG. It is assumed that the elapsed time from a certain reference time to the abnormality detection time is t2-0 seconds and t+g=150 seconds, respectively, and the elapsed time from the reference time to the predicted time specified by the operator console 110 is 1o''=300 seconds.

FIG. 6 shows the relationship between these times.

If, in step 116 of FIG. 3, the input branch of the abnormality detection device is removed. That is, in the network of FIG. 1, for the abnormality detection device 2, branches 1-2 are removed, and for the abnormality detection device 16, branches 11-16 are removed.

Next, in step 117 in FIG. 3, the reference device is set. Here, the device with the earliest abnormality detection time is defined as the reference device. In this example, device 2 is the reference device.

Next, in step 118 in FIG. 3, a dummy branch is added from the reference device i to another abnormality detection device j, and the device i
Let the direct failure propagation time from to j be alj=tj tls, and the direct failure propagation probability be b+j==1. In this example, a branch is added between devices 2 and 16, and a211g=T111
Let fg=150 and bz++s=1. As a result, the network shown in FIG. 1 becomes as shown in FIG. 7.

In step 119 of FIG. 3, the initial values of the predicted failure propagation time and propagation probability are set, and the
The provisional failure propagation prediction time tjj is pj = flash, and the provisional failure propagation prediction probability Qj is Qj = 0 (in this example, j = x, a
,4. ..., 19). Furthermore, for the reference device i, pl=t1°Q 1=1. In this example, pz
=fzmi01Q2=1.

At step 120 in FIG. 3, a failure-unaffected equipment set is set, and all devices other than the reference equipment are set as a failure-unaffected equipment set. In this example, the set of failure-unaffected devices is L = (1, 3
, 4, ..., 19).

At step 121 in FIG. 3, it is determined whether the failure-unaffected device set is an empty set, and if it is an empty set, the calculation is terminated. In this case, the influence of the failure spreads to all devices by the specified predicted time. If it is not an empty set, step 1
Proceed to step 22.

At step 122 in FIG. 3, the following processing is performed. From the latest device i (in this example, we will consider device 2) removed from the failure-unreached device set, the next device j that belongs to the failure-unaffected device set among all the devices to which the influence of the failure will directly spread is removed. The provisional failure propagation prediction time and probability are updated. For example, if device j is a single failure device, if the provisional failure propagation prediction time pj is pj>p 1+a, then p 1 = p ++ aB, q j = q + The affected device i is r
3-r. If pj=pt+aB, then qj
If <q+Xb+j, then Q''=qIX b+j.

Let r 1 = i, and leave it as is if qs≧q+XbB. I) If j < p l +aIj, leave as is. When device j is a device with multiple failures, if all devices k whose failures directly affect device j do not belong to the failure-unaffected device set, the failure propagation time from device k to j is within pk0ak1. Let the latest time in , be the provisional failure propagation prediction time pj of device j, find the sum of the branches on the failure propagation path from the reference device to device j, and let Qj be the product of all the direct failure propagation probabilities corresponding to the branches. . Otherwise,
Leave it as is. In this example, from device 2, devices 3 and 1
6. Since devices 3 and 16 are single failure devices, p3=flash>pi+F12. s = 100, therefore ps = 100, q3 = 02 ×b2+3
=l, "3=2, py5-oO>p2+az+1g=1
50, therefore f)xs=150.916=Q2
X bz+1g=1, rI6-2. As a result, the provisional failure propagation prediction time and probability are as shown in FIG.

At step 123 in FIG. 3, the devices with the earliest provisional failure propagation prediction time are narrowed down to one among the set of failure-unaffected devices. In this example, the set of failure-unaffected devices is L=, (1, 3, 4
, . . . , 19), so from FIG. 81, it is device 3.

In step 124 of FIG. 3, the tentative failure propagation prediction time pl of device i determined in step 123 is changed to the input specified failure propagation prediction time 1. y) Determine whether it is large or not, and py)t
If o, the calculation ends.

In this case, for devices that do not belong to the failure-unaffected device set,
It is predicted that the influence of the failure will spread by time io. Further, if pI≦to, the process proceeds to step 125. In this example, since ts=150 (t.=30G), the process proceeds to step 125.

In step 125 of FIG. 3, the device i selected in step 123 is removed from the set of failure-unaffected devices, and step 12
Return to J4. In this example, device 3 is removed. Therefore, the set of failure-unaffected devices is L-(1, 4, 5,..., 1
It becomes 91.

Below, the procedure until the end of calculation for the plant represented by the network in FIG. 1 will be described.

Since the determination result in step 121 is NO, the process advances to step 122. In step 122, the failure directly spreads from device 3 to devices 5, 8.9, and ps −■>
ps+”3+li=200 * ps=Sen>p3+
aass =200 + l)9 ==oo:)p3
Since 133.9=200, ps=200. q.

=qs Xbs, 5 ==0.2. rs −3+ p
m = 200°QB ””Q3 ×1)3+8” 1
+ '8 "'3 + p9 ""200+q9 =
q3 Xb3+e =11 '9 =3, as shown in Figure 9. In step 123, device 16 is selected and 1r
s=150 (to=300, so step 124
The result of the determination is NO, and the process proceeds to step 125. In step 125, the failure-unaffected device set is Lmi(
1,4,5,...,15°17.18.19), and the process returns to step 121. Since the determination result in step 121 is No, the process advances to step 122. Step 1
In 22, the failure directly spreads from device 16 to device 19, but device 19 is a compound failure device, and device 8, whose failure affects device 19, belongs to the set of devices that do not have failure, so l) + s is It remains as it is. Therefore, there is no change in the provisional failure propagation prediction time. In step 123, device 5 is selected, and since ts=2'00 (to=3oo), the determination result in step 124 is Nohtυ, and the process proceeds to step 125.

At step 125, L=(it 4.6, 7. . . ).

15.17, 18.191, and returns to step 121. Step 121 is No, so step 122
Proceed to. In step 122, the failure occurs from device 5 to device 6.
Directly spreads to p6 = (capital) > ps + as, s =
Since it is 300, p = = aoo.

Qa = Qs X bs + a = 0.2 + ra =
5, it will look like Figure 10. In step 123, device 8 is selected, and since ts = 200 (to = 300), the result is No in step 124, and step 125
Proceed to. At step 125, L=(1,4,6°7,9
．． 10. ..., 15, 17, 18, 19), and the process returns to step 121. Step 121 is No,
Proceed to step 122. In step 122, the failure propagates directly from device 8 to device 19. Device 19 is a complex failure device, and devices 8 and 16 whose failure directly affects device 19 do not belong to the failure-unaffected device set, so device 1
9. Calculate the provisional failure propagation prediction time and probability. l)s
10agBg=300 + p+g+au++xg =2
50, p+e=300. Reference device 2 to device 1
The failure propagation route at 94 is from device 19 to device 2 with the arrow reversed.
You can get it by going back to . Since 16=2, one of the ripple paths is 2→16→19. rIt”3+r3=2, the other ripple path stops at 2→3→8→19. Therefore, qn=p2.+gXp16 ++9 ×p2+3
×p3+8 ×pH+I9 ”” 1st next to O-1
It will look like Figure 1. At step 123, device 8 is selected, and since t9=20 o<to=300, the answer to step 124 is No, υ, and the process proceeds to step 125. In step 125, L=il.

4.6.7.10.11. ..., 15, 17, 18° 19), and the process returns to step 121. Step 121
Then, the answer is No, and the process proceeds to step 122.

In step 122, the failure directly spreads from device 9 to device 18, but device 18 is a compound failure device, and device 18
Since the device 17 to which the influence of the failure spreads belongs to the failure-unreached device set, pI8 remains unchanged. therefore,
There is no change in the provisional failure propagation prediction time. In step 123, device 6 is selected and t6-1o = 300, so the determination result in step 124 is No, and the process proceeds to step 125. In step 125, L=(1,4°7
．． 10,11. ..., 15, 17, 18, 19), and the process returns to step 121. In step 121, the result is No, and the process proceeds to step 122. In step 122, the failure directly spreads from device 6 to device 7, p7-■>I)6
+ as, 1 =: 310, so Tart =
3 1 01 Q7 = Qa X p a Mockery 720・
2.

At r7-6, it becomes as shown in Fig. 12. In step 123, the device 19 is selected, and since t19=1°-380, the determination result in step 1.24 is N.

Then, the process proceeds to step 125. In step 125, L
=i1, 4, 7, 10, 11. ..., 15°17.1
8), and the process returns to step 121. If the answer to step 121 is No, the process proceeds to step 122. Step 122
In this case, since there is no device to which the influence of the failure spreads from the device 19, there is no change in the tentative failure spread prediction time. Step 12
3, device 7 is selected and t, ==310)to=3
00, the determination result at step 124 is Ye.
S, and the calculation ends. Devices 2, 3, 5, 6, 8, 9, and 16゜19 that do not belong to the failure-unaffected device group are
This is a device that will be affected by the failure by the specified predicted time to = 300 seconds, that is, a device that is within the failure influence range. FIG. 13 shows the relationship between predicted failure propagation times for these devices. Note that the predicted probability of failure spread for device i is 91 in FIG.

Finally, the result of this calculation is displayed on the display device 115 shown in FIG.

According to the above-mentioned embodiment, it can be predicted that the number of devices in the failure influence range after 300 seconds is 2, 3, 5, 6, 8, 9 degrees 16.19 degrees. Furthermore, among the devices within the failure spread range, the devices with a high possibility of spillover are devices 2, 3, 8, 9.16, and the devices with a low possibility of spillover are devices 5, 6, 7. 19
It turns out that it is. Therefore, there is an effect that the operator can easily determine the order of troubleshooting.

〔Effect of the invention〕

As explained above, according to the present invention, (1) Targeting a plant consisting of multiple devices, it is possible to predict the range of failure influence from a point where an abnormality is detected after a specified time based on information from a limited number of sensors. , (2) It is possible to predict the fault propagation range even when there is a complex fault in the fault propagation relationship between plant components, and (3) it is possible to predict the abnormality even when there is a complex fault in the fault propagation relationship between plant components. The effect is that it is possible to calculate the predicted probability of failure propagation from the detection point to each device.

[Brief explanation of the drawings] Fig. 1 is a network diagram showing the fault propagation relationship of plant component equipment, Fig. 2 is a configuration diagram of an example of a plant system that realizes the fault spread range prediction method of the present invention, and Fig. 3 is a flowchart showing an example of processing in the fault spread range prediction device shown in Fig. 2, Fig. 4 is a matrix representation of the direct fault spread time between devices, and Fig. 5 is a matrix representation of the direct fault spread probability between devices. Figure 6 is a relational diagram showing the relationship between reference time, abnormality detection time, and specified predicted time; Figure 7 is a network diagram showing the failure propagation relationship of plant component equipment after preprocessing of calculations; and Figure 8 - Figure 12 is a diagram showing provisional failure propagation prediction times, propagation prediction probabilities, and propagation source equipment numbers for each plant component device at the beginning and during the performance, and Figure 13 is a related diagram %r diagram showing the relationship between failure propagation prediction times. 'fez Diagram No. 3 Diagram Z4 Diagram ■ 5 Diagram ■ 〆 口f Z t (6t. ■1 Sugi 11) Fig. 7 B Fig. 9 Fig. 10 Fig. 11 Kasumi No. 1? Diagram line / 3 Fig. F9 PI3 Continuation of page 1 ■Inventor Shigeo Hashimoto Hitachi University Mika Kaiten-chome 2-1 Inside Hitachi, Ltd. Omika Factory

Claims (1)

[Claims] In the plant failure spread prediction method, which predicts the range of devices to which the effects of a failure will spread based on sensor information corresponding to a limited number of devices in a system consisting of multiple devices, two devices are the initial input. The direction of failure propagation between
The predicted time is determined based on the propagation time, the propagation probability, whether the device is a complex failure device or a single failure device, the abnormality signal input from the sensor, the abnormality detection time, and the predicted time input from the operator. A step of removing the input branch of the anomaly detection device, a step of setting a reference device, and a step of setting a dummy branch from the reference device to other anomaly detection devices. a step of adding a failure propagation prediction time and an initial value of a failure propagation prediction probability; a step of setting a failure propagation range prediction; a step of determining whether the set is an empty set; If not, updating the provisional failure propagation prediction time and provisional failure propagation prediction probability in the failure propagation direction, determining the equipment to which the failure will spread earliest in the above-mentioned sharing, and the failure propagation prediction time of the requested equipment. By performing the step of comparing whether or not is greater than the predicted time input from the operator, and the step of removing the attached device from the set if the dog does not listen and returning to the step of determining the above. A plant failure influence range prediction method characterized by prediction.