JPH03287017A - Plant abnormality diagnostic system - Google Patents

Abstract

PURPOSE:To find the cause of abnormality of facilities and a countermeasure for it speedily by estimating the cause for each of plural abnormal events if the abnormal events occur and their causes contradict that of a single event. CONSTITUTION:Information which is inputted to an input part 1 is sent to a reasoning part 2 and checked by a logical contradiction check part 5. Then the input information is matched with logical contradiction data and when even part of it is coincident, it is judged that there is a contradiction, thereby actuating an individual cause reasoning part 7. When no coincidence is found, a cause reasoning part 6 is actuated. The cause reasoning part 6 calculates likely causes from past generation frequencies of respective extracted cause items and displays them in the likelihood decreasing order. Further, the individual cause reasoning part 7 does not search for the common cause of the input events and reasons the causes of individual input abnormal events one by one. In such a case, estimated cause and countermeasure plans are displayed one by one, similarly to the case of the reasoning part 6.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a plant abnormality diagnosis system that investigates the cause of an abnormality and presents countermeasures when an abnormal event occurs in a plant.

[Conventional technology]

An alarm is issued when an abnormality occurs in the plant.

Dealing with this problem appropriately is important both for safety and for efficient plant operation. Due to the recent increase in the scale of plants, it has become difficult to quickly identify the cause and take appropriate measures when an abnormality occurs. To this end, various systems have been developed that do not rely on human labor but utilize knowledge engineering technology. ing. Conventional devices of this type include, for example,
No. 196947, if there is a conditional statement stored in the knowledge base that matches the event that occurred in that conditional ring, the description of the conclusion term of that conditional statement is changed to the next event or result. The cause of the abnormality is estimated by repeating this process. In this case, a confidence threshold is given to the conditional ring, and a confidence level indicating the probability of the −-side event is also determined, so that the confidence level exceeds the confidence level of the conditional ring. When the threshold is exceeded, the result is adopted with the confidence given along with the result.

[Problem to be solved by the invention]

Depending on the plant, multiple abnormalities may occur from a single cause, and in this case, even if the confidence level is below the threshold, there is a good chance that the conclusion of that condition is true. is left in. However, in the conventional technology, if the confidence level is lower than the threshold value, the conditional statement is not applied, and there is a problem that appropriate inference cannot be made. Furthermore, in the conventional technology, only one common cause is inferred for multiple abnormal events, so if multiple abnormal events have different causes and occur simultaneously, those events will be inferred within the inference mechanism. The problem was that it was rejected as a contradiction and no further inferences were made.

An object of the present invention is to provide a plant abnormality diagnosis system that can reliably estimate the cause of an abnormality even when multiple abnormalities occur due to a single cause or when multiple abnormalities occur simultaneously due to multiple causes. The system is to provide.

[Means to solve the problem]

In order to achieve the above object, in the present invention, 11
[Investigate whether it is inconsistent to assume that a single cause exists when several abnormal events occur.]

If there is a discrepancy, try to estimate the cause for each individual abnormal event, and when estimating the cause, consider what could be the cause of each abnormal event or multiple abnormal events, and consider the degree of the possibility that it could be the cause. Displayed in order for diagnosis. Furthermore, the entire plant is divided into several subsystems according to function, abnormal events that may occur in each subsystem are displayed, and when the occurrence of an abnormal event is notified by an alarm, the display for each subsystem The event can now be entered.

[For production]

Even if multiple abnormal events are contradictory when attributed to a single cause,
Individual causes can be estimated through individual diagnosis.

It will no longer be ignored like before. In addition, the possible causes of an abnormal event are displayed on the screen in descending order of the estimated cause of the abnormality, so if you proceed with inference of secondary causes based on this, you will not miss the true cause of the abnormality. The cause can be quickly estimated. In addition, when an abnormal event occurs, it is now possible to input it for each subsystem, so when an abnormal event occurs only in one subsystem, input operations can be performed from one input screen without scrolling. Operation can be made faster.

〔Example〕

The present invention will be explained below using examples. FIG. 1 is a block diagram showing an embodiment of the present invention, in which an input section 1.1 is used by an operator to input an abnormal event when it occurs. An inference unit 2 that infers the cause of an abnormality based on input; a knowledge base 3 that stores knowledge used in this inference; It consists of an output section 4 that displays inference results and countermeasure plans. The inference unit 2 includes a logical contradiction check unit 5 that determines whether there is an inference contradiction between input events, and a cause inference unit 6 that executes inference if there is no logical contradiction, and a cause inference unit 6 that performs inference when a logical contradiction has occurred. and an individual cause inference unit 7 that individually executes cause inference for each input event.

The input unit 1 divides the plant into subsystems.

Abnormal events can be detected using two methods: inputting abnormalities that occur in each subsystem while looking at one screen, and inputting alarm items for the entire plant using multiple screens when an abnormality occurs throughout the plant. It is possible to input the following information.

FIG. 4 shows a gaseous waste treatment facility for an atomic couplant, and this will be explained below as an example. When this equipment is viewed as one plant, it is divided into subsystems as surrounded by dotted lines in the same figure. Then, the abnormality alarms related to each subsystem are classified in advance as the first system to the third system in FIG. 5.

FIG. 5 shows a part of the matrix used for estimating the cause of an abnormal event for the system shown in FIG. 4, and is stored in the knowledge base 3. The input screen of the first system displayed by the input unit 1 is shown in FIG. 6. The initial state of this screen is NO for each item, and the operator inputs YES to the corresponding item in response to the generated alarm. In the system shown in Figure 4, there are 36 abnormal events, that is, alarms, and if these were always displayed and input from the beginning, it would take time to input them, but as shown in Figure 6, If we could make it possible to display and input information for each system at once,
Especially when the abnormality is localized to one subsystem and it is known which one.

Input operations can be performed extremely efficiently.

The input information is then sent to the reasoning section 2 in FIG. 1 and checked by the logical contradiction checking section 5.

Figure 7 is a schematic flowchart of this check.

Among abnormal events that can occur (and therefore can be input) other than instrument abnormalities, combinations of contradictory events, such as one occurring from one cause but not the other, are set in advance as logical contradiction data. I'll keep it. Then, check the input information against logically contradictory data.

If even part of the input information and logically inconsistent data match,
It is determined that there is a contradiction and the individual cause inference unit 7 is activated. Further, if no match is found in the comparison between the input information and the logical contradiction data, the cause inference section 6 is activated.

FIG. 8 shows a schematic processing flowchart of the cause inference unit 6 that is activated when an abnormal event with no logical contradiction is input. Here, first, by referring to a matrix stored in the knowledge base 3 whose elements are the frequency of occurrence of each cause of abnormal events, the cause item corresponding to the input abnormal event is extracted.

If there are multiple pieces of input information, the common cause items are extracted. Next, the probability of the cause is calculated from the past frequency of occurrence (the value of the matrix element) for each of the extracted cause items, and the causes are displayed in descending order of probability. In the case of the system shown in FIG. 4, it is assumed that, for example, a high temperature at the outlet of the air extractor and a high pressure at the outlet of the air extractor occur simultaneously.

The matrix in Figure 5 shows "hydrogen explosion", "main steam pipe residual heat", "hydrogen explosion", "main steam pipe residual heat",
"Measurement control system malfunction" is set, and "hydrogen explosion J, rSJAE back pressure increase" and "measurement control system malfunction" are set for the air extractor outlet pressure high, so the air extractor For the simultaneous occurrence of high outlet temperature and high air extractor outlet pressure, ``hydrogen explosion'' and ``malfunction of the measurement control system'' are extracted as the primary cause inference result. Furthermore, based on numerical calculations using the frequency of occurrence of the cause as a parameter, the first candidate for the cause is "hydrogen explosion" and the second candidate is "malfunction of the measurement control system", and these are displayed as the first probable cause. . For example, as shown in Figure 9, Knowledge Base 3 includes three causes (this is the same as the matrix in Figure 5) for the high pressure at the air extractor outlet (exhaust gas preheater inlet) and the secondary causes that occur for each of them. Tree-shaped data such as causes, causes, and countermeasures for these causes are also stored, and when one cause is selected on the display screen of the above-mentioned first-order estimated factors, lower-order causes are displayed according to the tree shown in Figure 9. By repeating this, the final inference result of the cause of the abnormality and countermeasures are displayed on the output unit 4. In this way, even if two causes have a low frequency of occurrence, they are always displayed and can be selected, so even when multiple events occur from a single cause, the cause can be reliably estimated.

If the logical contradiction checking section 5 determines that there is a logical contradiction in the input information, the individual cause inference section 7 is activated and the process of the flowchart schematically shown in FIG. 10 is executed. In this process, a search for a common cause for the input events is not performed, and the inference of the cause is made for each abnormal input event one by one using the matrix shown in Figure 5 and the matrix shown in Figure 9.
Estimation is made based on tree-shaped knowledge as shown in the figure. The specific method of estimation is the same as that of the cause inference section 6.

Estimated causes and countermeasures are determined and displayed for each abnormal input event. For example, when the air extractor outlet temperature and the same outlet pressure become high at the same time, the primary probable causes of the high air extractor outlet pressure are rSJA, E exhaust pressure rise, hydrogen explosion, and measurement. The primary probable cause of the high air extractor outlet temperature is "main steam pipe residual heat,""hydrogenexplosion," and "measurement control system malfunction" are displayed on the output section 4 in this order. . In the following, we will select from these and investigate lower-level causes. In this way, even when a plurality of contradictory abnormal events occur simultaneously, the cause can be estimated.

FIG. 2 shows a second embodiment of the invention.

This is a modification of the inference section 2 in FIG. Logical contradiction check section 5. The processing contents of the cause inference section 6 and the individual cause inference section 7 are the same as in FIG.

Their connection relationships are different. In other words, the logical contradiction check unit 5
Even when a logical contradiction is once discovered, the causal inference section 6 executes the inference, and when the result is displayed on the output section 4, the occurrence of the logical contradiction is also displayed on the screen. At that time, we next performed individual cause inference. In this embodiment, when performing an abnormality diagnosis, there is almost no logical contradiction, so
The main focus is on activating a system that estimates a single cause as much as possible.

FIG. 3 shows a third embodiment of the present invention, in which the range of abnormality cause estimation is further narrowed based on the inference result performed by the cause inference section 6, and the true abnormality cause can be identified. , which has a re-input section 8 added to the embodiment shown in FIG.

This re-input unit 8 is prepared with input items for specifying the cause of the abnormality based on the estimation results performed by the cause inference unit 6. Using the loop that is formed, it is possible to identify a single faulty device.

〔Effect of the invention〕

According to the present invention, when an abnormal event occurs, the cause of equipment abnormality and its countermeasures can be quickly determined even when a skilled operator is not available, and the cause can be reliably estimated even when multiple abnormalities occur. There is an effect that it can be done.

[Brief explanation of drawings]

Figures 1 to 3 are functional block diagrams showing embodiments of the present invention, Figure 4 is a diagram showing an example of division of gaseous waste treatment equipment into subsystems, and Figure 5 shows abnormal events and their causes. The figure which shows the example of the matrix shown in FIG. Fig. 6 is a diagram showing an example of an abnormal event input screen, Fig. 7 is a flowchart showing the processing of the logical contradiction check section, Fig. 8 is a flowchart showing the processing of the cause inference section, and Fig. 9 is a diagram showing the cause of the abnormality in a hierarchical manner. FIG. 10 is a flowchart showing the processing of the individual cause inference section. DESCRIPTION OF SYMBOLS 1... Input part, 2... Inference part, 3... Knowledge base, 4... Output part, 5... Logical contradiction check part, 6...
...Causal inference section, 7. Individual cause inference section, 8. Re-input section.

Claims (1)

[Claims] 1. Input means for inputting abnormal events in the plant;
In a plant abnormality diagnosis system comprising an inference means for inferring the cause and determining a countermeasure from the input abnormal event, and a display means for displaying the cause inferred by the means and the countermeasure determined, the above inference is provided. The means includes a list of combinations of a plurality of abnormal events that cannot occur simultaneously from one cause, and a check means for checking whether or not among the input combinations of the plurality of abnormal events, there is one that matches a combination in the list. In addition, if a match is not detected by the checking means, the cause is estimated as if multiple input abnormal events were caused by one cause, and if a match is detected by the checking means, the input abnormality is A plant abnormality diagnosis system is characterized in that each event is assumed to have occurred due to a different cause and its individual causes are estimated. 2. The abnormal event of the plant is divided into subsystems of the plant that cause the event, and the input means inputs the abnormal event by displaying a list of abnormal events for each subsystem and selecting one. The plant abnormality diagnosis system according to claim 1, wherein the plant abnormality diagnosis system is capable of diagnosing a plant abnormality. 3. The inference means infers the cause of each abnormal event using a matrix that records the cause of the event and the frequency with which the event occurred in the past based on the cause, and also determines whether each of the estimated causes 2. The plant abnormality diagnosis system according to claim 1, wherein the probability of being the true cause of the occurrence of the abnormal event is calculated from the frequency of the matrix and displayed on the display means in order of the probability.