JPH07107719B2 - Alarm processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to an alarm processing method when an abnormality occurs in a large-scale plant system.

(Prior art) In a large-scale system such as a nuclear power plant or a thermal power plant, when an abnormality occurs, a large number of alarms (ANN) that occur are reasonably processed, and these alarms are mainly displayed on the CRT display screen. It is desirable that information is integrated and alarm processing is performed to reduce the burden on the operator for judging the situation in the event of an abnormality.

Conventionally, human beings have assumed various causes of abnormality from a large number of alarms that are generated based on knowledge about the function and structure of the plant, estimated the spread route, and dealt with the cause of abnormality. In this way, it is difficult for the human method to list all possible causes of abnormalities and abnormal transmission paths when the system becomes large-scale, so there is no particular impact on the operation of the plant among the numerous alarms that occur, Alarms that have relatively allowance for operator's response are deleted (filtered), various abnormal causes are assumed from the remaining alarms, the propagation route is estimated, and the action method at the time of abnormalities is set for each event or alarm. As a result, alarm processing has been performed to reduce and support the operator's situational judgment in the event of an abnormality.

(Problems to be Solved by the Invention) In the above-described conventional alarm processing method, even if the number of generated alarms can be reduced by deleting alarms that have a relatively large margin without particularly affecting plant operation, It is difficult to enumerate all possible causes of abnormalities and abnormal transmission paths when the system becomes large-scale by the method described in (1), and when these abnormal transmission paths form a closed loop or due to multiple faults. When it occurs, it is difficult to summarize for each initiating event or each alarm, and it is difficult to determine the cause of abnormality and the abnormal spread route.
There was a problem that it was difficult to take measures during abnormal times.

The present invention has been made to solve the above-mentioned problems, and even when the abnormal spread path constitutes a closed loop or when the abnormal spread path due to multiple failures occurs, the corresponding measures are automatically taken. An object is to provide an alarm processing method that can be selected.

[Structure of Invention]

(Means for solving problems) Knowledge about attributes including the importance of alarms in large-scale plants such as nuclear power plants, knowledge about causes of alarms and actions to be taken in order to provide more natural knowledge to humans. It is stored using a frame for expressing, and based on the stored knowledge, the alarm propagation path generated from the plant data at the time of an abnormality is connected in a tree shape, and the connected tree-shaped alarm propagation path is connected. If there are multiple abnormal events depending on the abnormal event, the ripple route is traced upstream and downstream according to the abnormal event, and if there is no closed loop in the uppermost stream of the ripple path, the uppermost alarm is used as the cause alarm of the abnormality. Estimate and determine the action that corresponds to the cause of this abnormality.If there is a closed loop in the uppermost stream of the transmission route, the alarm that forms that closed loop has the earliest occurrence time. It is presumed as a cause alarm and the action corresponding to the cause of this abnormality is determined.If the generated alarm has an alarm of a predetermined importance level or higher, the action corresponding to that alarm from the most important alarm is treated as the cause of the error. It prioritizes decisions and assists the operator in judging the situation.

(Operation) The state of a large-scale plant is detected by a large amount of analog signals and digital signals such as alarms. In the present invention, for analog signals, an abnormality detection method (for example, comparison with an AR model) suitable for the characteristics of each signal based on safety analysis and operational experience during plant design, and a threshold value for determining normality / abnormality The output of a monitoring system that indicates normality / abnormality as ON / OFF (hereinafter referred to as ON / OFF) by performing threshold value determination and the operation of alarms and equipment installed in the plant control room / Using a signal indicating a pause, grasp the state of the plant as a combination with a signal represented by a binary value. (Hereinafter, all of these signals are called alarms in a broad sense.) Knowledge of plant interlock (for example, automatic start logic of auxiliary motor at cooling system pump trip in nuclear power plant) and physical causal relationship of abnormal phenomenon (Discharge pressure drop due to pump trip, etc.)
Can be obtained from the plant structure and design data. Therefore, for each alarm, the attribute (such as importance) indicated by the alarm, the cause of the alarm and the corresponding action are stored in a frame format as shown in FIG. 5, for example. Here, the frame is proposed in the frame theory proposed by Minsky, and refers to a framework for expressing more natural knowledge for humans.

In this case, one frame corresponds to one alarm, and the alarm name slot indicating the attribute of the alarm, the importance slot, the cause alarm slot indicating the causal relationship of the alarm, and the pointer pointing to the upstream alarm frame It has a slot, a pointer slot that points to a downstream alarm frame, a treatment slot that indicates a treatment, and the like. By using these frames to form a network, the hardware alarm logic can be realized as it is.

Since these are described by dividing them into alarm units, there is an advantage that data can be easily created, modified, changed, and added, and a hierarchical structure such as an abnormal route of alarm is easily expressed ( For a detailed definition, refer to "Introduction to Knowledge Engineering" issued by Ohmsha Co., Ltd.).

As shown in Fig. 5, for example,
Im, the generation condition is Am and (&) An, or Am '&An',
When the cause alarm Am & An occurs, the action is Dm, similarly Am ′
The treatment for & An 'is Dm', and these data are entered and saved by humans for each alarm.

When an abnormality occurs in the plant, create a table that describes the plant status with a collection of ON / OFF binary signals by collecting the alarms provided in the plant control room and the signals indicating the operation / stop of the equipment. , The abnormal condition is read from the created table, and the spread route of the alarm generated based on the stored knowledge is connected like a tree using a frame, and the spread route from any of the generated alarms to the upstream side. By tracing the alarm toward the downstream side, the state of the abnormality is grasped, and the alarm at the most upstream side of the transmission route is determined as the alarm causing the abnormality. In addition, some alarms including alarms of abnormalities form a closed loop, and even if the uppermost alarm of the ripple route is not found, the alarms forming the closed loop are extracted from the generated alarms and the maximum of the ripple route is extracted. It judges whether it is a closed loop in the upstream, and if it is the most upstream closed loop, it judges the alarm causing the abnormality by comparing the time of the alarms forming the extracted closed loop. Is.

The abnormal plant status includes some steps that can restore the plant to normal status by taking measures, and some steps that cannot be recovered from the abnormality and must be stopped safely. However, it is important to take measures corresponding to each stage. Therefore, the alarms are ranked as importance levels indicating abnormal stages, and, for example, in a nuclear power plant, the importance levels shown in FIG. Among them, the most important and most upstream alarm is determined as the alarm that should be the most noticeable when the operator takes action, and the corresponding action is displayed as a processing guide given to the operator based on the stored knowledge.

By connecting the spillover route of the generated alarm like a tree,
Since the processing can be performed for each abnormal event, even if an abnormal transmission route due to multiple failures occurs, it is possible to determine the cause alarm of the abnormality for each abnormal event and display the action corresponding to the abnormal stage.

(Embodiment) FIG. 1 shows the configuration of an apparatus for carrying out the present invention.
The monitoring data obtained from the plant 1 is input to the alarm processing device 10 and another monitoring device 20. This alarm processing device 10
Is an alarm table creation unit 11, a knowledge base 12, an inference unit
13 and a display unit 14.

Among these, in the alarm table creation unit, an alarm signal in the original meaning provided on the control panel, a signal indicating the operation / stop of the device, and another monitoring device 20 for monitoring the analog amount to detect an abnormality. Create a table that describes the state of the plant 1 with the collection of binary signals that are ON / OFF together with the output signals.

The knowledge base 12 stores knowledge about attributes such as the name and importance of each alarm, knowledge about the cause of the alarm and the corresponding action for each alarm in a frame format. This frame is made up of a collection of knowledge about each alarm and the conditions under which it is connected to other alarms, so it is easy to create, and corrections, changes and additions after completion are made. Can be done easily.

The inference unit 13 reads the state of the plant from the alarm table creation unit 11, searches for an alarm that has occurred using the knowledge base 12, connects the propagation route of the alarm that has occurred in a tree-like manner, and processes each abnormal event. By doing so, the cause alarm of the abnormality is measured for each abnormal event, and the action corresponding to the abnormal stage is selected.

The display unit 14 displays an alarm for the cause of the abnormality and a treatment corresponding to the stage of the abnormality to assist the operator 2 in determining the situation.

FIG. 2 shows a concrete processing procedure of the alarm processing apparatus 10. In step 101, the plant state is periodically taken in, and in step 102, the presence or absence of a newly generated alarm is determined. And if there is a new alarm, step
In 103, search for an alarm that has occurred using knowledge about attributes such as the name and importance of the alarm, knowledge about the cause of the alarm and the corresponding action, and in step 104, frame the spread route of the alarm. Connect in a tree-like manner using and determine the uppermost alarm of the transmission route as the cause alarm of the abnormality. If it is determined in step 105 that there are closed loop alarms by performing processing for each abnormal event using the tree-like connected alarm propagation path, they are released, and in step 106 the propagation path. It is determined whether the closed loop is in the uppermost stream, and if it is in the uppermost stream, the times of occurrence of the extracted alarms are compared and the cause alarm in the closed loop is determined by the earliest alarm.

Furthermore, by using the importance of the alarm for each abnormal event, the most important and most upstream alarm among the alarms forming the abnormal spread route is measured, and the corresponding action is selected in step 108,
Display the action.

The knowledge base, tree-like alarm spread path and alarm measurement will now be described in more detail.

Consider a case where knowledge bases (frames) 121 to 132 as shown in FIG. 3 are prepared for many alarms A1 to A12. The importance levels shown in these knowledge bases are classified into I1 to I5. Among them, I1 has the highest level of importance, and I2, ..., I5 have lower levels of importance.

Now, assuming that the alarms A1 to A12 are generated at one time, based on the knowledge bases 121 to 132, a tree-like alarm propagation route as shown in FIGS. 4 (a) and 4 (b) is created. Any alarm
Check the upstream warning from A2. Alarm A2 finds the cause alarm A1 which is the upstream alarm by searching the frame for alarm A2, and checks alarm A1.
Even if the frame corresponding to A1 is searched, there is no cause alarm on the upstream side that causes alarm A1, so it is judged to be the cause alarm for this abnormal event, the uppermost flag is set, and the alarm A2 that was first investigated is returned. . Next, in the same manner, the alarm on the downstream side from the alarm A2 is checked, and the alarm A3 which is the downstream side alarm is found.
Although the alarm A3 is checked, even if the frame corresponding to the alarm A3 is searched, there is no alarm having A3 on the upstream side, and therefore, the lowest downstream flag is set, and the process returns to A2 which is the first to be checked. Among the alarms A1 to A3 examined in the series of processes, the alarm A2 having the highest importance is estimated to be the most important alarm, and the corresponding action is displayed.

Next, since there are some alarms that have not been investigated yet, the alarms upstream from the arbitrary alarm A5 are investigated in the same manner as described above. Alarm A5 to alarm A4, alarm A7, and alarm A6 are searched.Alarm A5, which is an alarm on the upstream side of alarm A6, has already been checked, and an alarm that forms a loop is input from an alarm other than an alarm that forms a loop. Therefore, these alarms are registered as the alarms forming the loop in the uppermost stream. Next, the alarms on the downstream side from the alarm A5 are also examined in the same manner. For the alarms A8, A9, A10 and A11, the alarms A5 other than the alarms forming the loop are
Since it has been entered, the last remaining alarm A1 was not registered and was not registered.
By investigating 2 we return to A5, which was the first to investigate with the most downstream flag. Furthermore, the time of occurrence of alarm A4, alarm A5, alarm A6, and alarm A7 forming the loop in the uppermost stream is examined, and the earliest one is regarded as the starting point. For example,
If the time of occurrence of alarm A5 is the earliest, alarm A5 is presumed to be the cause alarm. Furthermore, similarly to the above, among the alarms A4 to A12 examined in a series of processes, the alarm A12 having the highest importance is determined to be the most important alarm, and the corresponding action is displayed. Since all the generated alarms have been checked, the process ends.

[Effects of the Invention] As described above, according to the present invention, the transmission route is traced toward the upstream side and the downstream side for each abnormal event, the alarm for the cause of the abnormality is estimated, and the corresponding action is determined. Therefore, it is possible to deal with multiple failures, and if there is no closed loop in the uppermost stream when tracing the ripple path, the uppermost stream alarm is presumed to be the cause of the abnormality, and if there is a closed loop in the uppermost stream, the closed loop is configured. Since the alarm with the earliest alarm occurrence time is estimated as the cause alarm of the abnormality, it is possible to take corrective action according to the identification of the cause alarm. In this case, since the alarm corresponding to the most important alarm is prioritized over the action for the cause of the abnormality, it is possible to allow the operator to make a quick situation judgment, and Compared to the method In this case, there is an effect that the operator can be appropriately treated and the safety of the plant can be remarkably improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an apparatus for carrying out the present invention,
FIG. 2 is a flow chart for explaining the operation of the device, FIG. 3 is a frame diagram showing the detailed configuration of the main elements of the device, and FIGS. 4 (a) and 4 (b) are for explaining the operation of the device. 5 and 6 are framework diagrams for explaining the principle of the present invention. 10 …… Alarm device, 11 …… Alarm table, 12 …… Knowledge base, 13 …… Inference section, 14 …… Display section.

Claims (1)

[Claims] 1. A frame for expressing more natural knowledge for human beings on knowledge about attributes including importance of alarms in a large-scale plant such as a nuclear power plant, and knowledge about causes of alarms and actions corresponding to the alarms. Based on the stored knowledge, the alarm propagation path generated from the abnormal plant data is connected in a tree shape, and the connected tree-like alarm propagation path becomes an abnormal event. Therefore, if there are more than one, the spillover route is traced upstream and downstream for each abnormal event, and if there is no closed loop in the uppermost stream of the spillover route, the uppermost stream alarm is presumed to be the cause alarm of the abnormality. Determine the action that corresponds to the cause of the abnormality, and if there is a closed loop in the uppermost stream of the transmission route, the alarm with the earliest occurrence time of the alarm forming the closed loop is recommended as the alarm causing the abnormality. And determine the action corresponding to the cause of the abnormality, and if the occurrence alarm has an alarm of a predetermined importance level or higher, the action corresponding to the alarm from the alarm with the highest importance is taken as the action for the error cause. An alarm processing method characterized by prioritizing decisions and assisting the operator in judging the situation.