JPH01100608A - Method for assuming first cause of abnormal condition in plant operation - Google Patents

Abstract

PURPOSE:To promptly and easily decide the first cause of an abnormal condition by tabling relating alarms at every alarm item into an electronic computer, etc., filing them, automatically retrieving them when plural abnormal alarms simultaneously occur, and tracing causal relationship. CONSTITUTION:When the three alarms of alarm windows A7, B4 and C1 peal, the electronic computer automatically retrieves the three tables of A7, B4 and C1, and decides as follows: (a) A7 occurs because of B4, and affects C1; (b) B4 affects A7, while, C1 is not the cause of B4; (c) C1 occurs because of A7, while, it does not affect B4; (d) as this result, the affecting path of the occurred alarms is B4 A7 C1; (e) consequently, the first cause of the occurred alarm is B4. When the alarms, whose causal relationship cannot be judged, occur, they are decided to simultaneously occur by chance.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to alarm technology for notifying operators of the operating status of the plant or abnormal status of equipment in various plants such as thermal power plants, and The present invention relates to an abnormality cause estimation method suitable for estimating the first cause of an alarm item when two alarms occur simultaneously.

[Conventional technology]

The conventional plant condition monitoring/alarm device was disclosed in Japanese Patent Application Laid-open No. 58-1.
No. 17008, JP-A-58-119007, JP-A-5
As described in No. 8-119008, the operating state of the plant is detected and displayed on a tree diagram when an abnormal state occurs to determine the cause of the abnormality, or the state quantity of the plant is detected and calculated using a high-speed simulator model. The cause of the abnormality is judged to be the one that has a good degree of agreement with the actual measured data, but in addition to requiring complex logic, if multiple abnormality alarms are activated at the same time,3) It was difficult to estimate which of the abnormalities was the primary cause.

If the first cause can be immediately determined, appropriate measures can be taken quickly to respond to the abnormality alarm. However, the longer it takes to determine the first cause, the more the response will be delayed, increasing the risk of secondary and tertiary damage.

The time wasted in this case cannot be compared with the value of time in a general production process, and it is unclear how a one-second delay could lead to millions of yen in damages.

The present invention has been made in view of the above-mentioned circumstances, and it is an object of the present invention to provide a method that can quickly and easily estimate the first cause when multiple abnormality alarms occur simultaneously (including almost simultaneously, the same applies hereinafter). With the goal.

[Means for solving problems]

First, the basic principle of the method of the present invention created to achieve the above object is summarized as follows.

The above purpose is to create a table in an internal file of a computer, etc. for each alarm item, ``alarms that lead to the occurrence of the alarm'' and ``alarms that are caused by the adverse effects of the alarm''. This is achieved by automatically searching and automatically tracking the table using the generated alarm as a key when an alarm occurs to determine the first cause.

Based on the above-mentioned principle, the estimation method of the present invention has a specific configuration for applying the same to a practical aspect, and for each of a large number of information items representing the operating status of a plant,
The operating status of the plant is monitored using a large number of indicators that activate an alarm when the numerical value representing the operating status corresponding to the information item becomes an abnormal value, and when an alarm for multiple items is activated, Applicable to the case of estimating the first cause of the abnormality of the plurality of items, (a) When an abnormality of the item occurs for each of the plurality of information items, the abnormality item that is induced as a result. (b) For each of the large number of information items, assume in advance an abnormal item that may be the cause of the abnormality in the item, (c) For each abnormal item, determine the cause of the abnormality. (d) When an alarm indicating multiple abnormalities is issued during plant status monitoring, , for each of the plurality of abnormality items, search and collate the cause abnormality item and the resultant abnormality item, and (e) search for cause/effect, cause/response, etc. in all permutations of the plurality of abnormality items. This method selects one set in which the relationship between effect and cause/effect holds true without contradiction, and infers the first cause by tracing the causal relationship of the selected set in reverse. However, in the present invention, a large number refers to an integer of 3 or more.

[Effect]

The reason why the first cause can be estimated by the method according to the above configuration will be explained using the simplest case.

Generally, when an abnormality occurs in plants, the first
There are extremely many cases in which secondary abnormality B is induced as a result of cause A, and tertiary abnormality C is further induced as a result of the above-mentioned secondary abnormality.

Moreover, the actual situation is even more complicated, with multiple secondary abnormalities occurring from one primary cause. B2. ...Bn, and each of these secondary abnormalities Bl to Bn induces a plurality of tertiary abnormalities, resulting in a large number of tertiary abnormalities C1-1゜,
Cx-xt C2, -z, C2-2, C2-8~C
There are also examples where nn is expanded in a pyramidal manner.

In this "action" section, as a simple example, A→B
Let us consider a case where the occurrence of abnormality progresses in a chain like -+C.

However, when implementing the present invention, it is possible to estimate the first cause using a computer, and when using a computer in this way, the arrows (→) of A-)B-)C above are Since it is just a calculation factor, it is possible to replace it with any mathematical symbol.

In the example described below, A-)B is expressed as "A is (IN) for rB when viewed with B as a reference."

Similarly, when looking at B-+C with B as a reference, it is expressed as "C is (OUT) for B." Applying this to Japanese, (IN) is cause and (OUT) is effect.

Now, it is assumed that the above-mentioned table, which is filed on a computer, etc., is created for each alarm item, and now, "A
”, “B”, and “C” are activated simultaneously, the method of the present invention described above is implemented as follows.

(i) Using each of rAJ, rBJ, and rCJ as search keys, check whether there are any other generated alarms in each file (that is, "B" in the file rAJ).

Isn't there a "C"? , "A" in the file "B".

Is there no rCJ? , rc” file is “A”.

Don't you have rBJ? ). In this case, an alarm J (IN) which is the cause of the occurrence of the alarm and an alarm j (OUT) which is the cause of the alarm and which spreads are searched separately.

(ii) IN searched by section (i) above.

Connect OUT. In other words, rBJ to IN of rAJ
, if there is rCJ at OUT, rAJ at rBJ○UT, and rAJ at IN of "C", the propagation path is as follows.
The causal relationship among the three becomes clear as B-4A-40.

(iii) Through the processing in item (...) above, it can be determined that rBJ is the first cause of the alarm, and this is displayed on a display or the like.

〔Example〕

Hereinafter, examples of the present invention will be explained using a thermal power plant as an example.
This will be explained using drawings.

Figure 1 shows an example of the arrangement of alarm windows in a thermal power plant, and related alarms selected using each window (alarm item) Nα (window number) as a key, and compiled into a table on a computer file. It is.

Assuming that the alarm windows are arranged in M rows and N columns, each window is numbered A1 to AN, Bl to BN, . . . M1 to MN. In this case, for example, the contents of the above-mentioned file in the alarm window A4 are as shown in FIG. In this example,
“The alarm items (windows) that cause alarm A4 are C8, A8.
M5. C4, C3°A4. J (I
N) and that "the alarm items (windows) that spread and occur due to alarm A4 are C1, M4.A2.B3.M2."

Next, using the table described in FIG. 1, a method for tracing the alarm propagation route when a plurality of alarms occur and determining the first cause will be explained with reference to FIG.

In the alarm window of FIG. 1, A7. B4. When three alarms C1 occur at the same time, the electronic computer etc. will display rA7'' and rBl as shown in Figure 2.

The three tables of "C1" are automatically searched and the following results are obtained.

(b) rA7J occurs due to "B4" and also spreads to "C1".

(0) rB4J spreads to rA7JL, but "cl" is not the cause.

(c) rcIJ is caused by rA7J, but
It will not affect "B4".

(d) From the above (a), (b), and (c), "B4" →
“A7” → “C1” is the propagation route of the alarm that occurred this time.

(e) Therefore, the first cause of multiple alarms this time is “B
4".

The electronic computer caused this result (rB4J is the first cause)
By displaying this on a display (not shown), the operator is not confused by multiple alarms and only needs to take action against "B4", which is the first cause.

In this embodiment, the case where three alarms occur at the same time has been described as an example, but the same method can be applied to the case where four or more alarms occur at the same time.

In addition, if there is an alarm for which a causal relationship cannot be established even after performing the processing shown in Figure 2 (an isolated alarm that is not connected to other alarms), the alarm is considered to be an alarm that "happened to occur at the same time."
It can be determined that

According to this embodiment, when multiple alarms occur simultaneously, the cause-and-effect relationship can be clarified and the first cause can be determined, so operators can take effective measures immediately without being confused by multiple alarms. can.

FIG. 3 shows a different embodiment from the above regarding the tabulation method.

What is explained in Figure 1 is a list of related alarms arranged at random for each IN and OUT, and if related alarms are added later, they will be added to the blank area at the rear. Become.

The table of this embodiment shown in FIG. 3 is obtained by automatically sorting (data conversion) the table of FIG. 1, and rearranging each of the OUT and IN with the alarm window arrangement in mind. As a result, when adding data (related alarms &), the operator who inputs the data does not have to be aware of the arrangement of the alarm windows, and the computer automatically rearranges them, making automatic search even easier.

As shown in FIG. 4(a), the alarm in the embodiment described in FIG. The search process shown in FIG. 2 is started only after a plurality of such "alarm occurrences" occur. However, a causal relationship is not necessarily established until multiple alarms occur, and in the actual process, although it is abnormal in relation to the above limit value (the ripple factor to other alarms is However, there are cases where the limit value is not exceeded and an "alarm is not issued". For this reason, we can predict abnormalities before they become "alarms", trace the cause-and-effect relationship, and use the 1st Ji! The fourth one was created based on the idea of estimating the cause.
This is the abnormal area determination function shown in FIG.

In FIG. 4(b), the vertical axis shows Δ (deviation between the measured process amount and the normal process value), and the horizontal axis shows the temporal change rate of this Δ. Normally, an alarm occurs when Δ>P or Δ<Q is below, but in the example shown in this figure, the time rate of change of Δ is taken into account, and even if Δ>P and Δ<Q are not present, the value suddenly changes. If there is a change, it is considered to be in an abnormal region. As a result, even if the limit value is not exceeded at that time, C
13) It is possible to detect abnormal conditions that would cause an "alarm to occur." Therefore, if the search process in Figure 2 is started when there are multiple objects that have "entered the warning area," causality tracing can be carried out quickly, and the operator will know the first cause before the alarm occurs. It is possible to quickly respond to warnings.

In addition, Fig. 5(a), (b) and Fig. 6 are similar to Fig. 4(b).
), and it is preferable to define the warning area by taking into account the characteristics of the process corresponding to each warning item.

〔Effect of the invention〕

According to the present invention, alarms related to each alarm item are stored in a table on a computer, etc., and when multiple abnormalities occur at the same time, they are automatically searched and the causal relationship is traced to find the first cause. can be determined. This allows operators to quickly and easily take appropriate countermeasures without being confused by multiple abnormalities, which greatly contributes to improved plant operability and maintainability.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing the arrangement of alarm windows and the relationship between related alarm tables in an embodiment in which the present invention is applied to a thermal power plant, and FIG.
A schematic diagram to explain the method for processing the first cause determination, Figure 3 is an explanatory diagram showing automatic sorting of the contents of the table shown in Figure 1, and Figure 4 is a warning before "alarm occurrence". FIGS. 5 and 6 are charts showing functions for determining regions, and charts showing other examples of the functions.

Claims (1)

[Claims] 1. For each of a large number of information items representing the operating status of the plant, a large number of indicators are used that activate an alarm when a numerical value representing the operating status corresponding to the information item becomes an abnormal value. In a method for monitoring the operating state of a plant using a computer, and estimating the first cause of an abnormality in a plurality of items when a plurality of alarms are activated, (a) for each of the plurality of information items; (b) For each of the large number of information items, predict in advance the abnormal items that may be the cause of the abnormality in the item when an abnormality occurs in the item. (c) For each abnormal item, record the abnormal item that is the cause of the abnormality and the abnormal item that is the result induced by the abnormality by a means that can be immediately retrieved. (d) When an alarm indicating multiple abnormalities is issued during plant status monitoring, search and collate the cause abnormality item and resultant abnormality item for each of the multiple abnormality items, (e) Among all the permutations of the plurality of abnormal items, at least one set is selected in which the relationship between cause and effect, cause and result to cause and result is established without contradiction, and the causal relationship of the selected set is reversed. 1. A method for estimating a first cause of an abnormality in a plant operating state, the method comprising tracing the first cause and estimating the first cause. 2. Abnormal conditions that cause the above-mentioned alarm to be activated are determined for the relevant information item: (f) when the numerical value representing the operating state reaches a preset dangerous value; and (g) the numerical value representing the operating state. The plant operation according to claim 1, characterized in that when the alarm value reaches a preset alarm value at a rate higher than a predetermined rate of change, this is determined to be an abnormal state. A method for estimating the first cause of abnormal conditions.