JPH0235526A - Knowledge structuring method for expert system - Google Patents

Abstract

PURPOSE:To easily generate a rule even when the number of events increase and the number of times of the establishment of sequence relation increases by storing respective events in sequence relation separately as prevents and postevents. CONSTITUTION:The except system is provided with a inference mechanism 1, to which a knowledge base 2 and a data base 3 are connected. Further, a user interface 4 is connected to the inference mechanism 1 and input information is sent to the inference mechanism 1. A knowledge acquisition assistant module 5 is provided between the knowledge base 2 and user interface 4 and knowledge which is inputted through the user interface is stored in the knowledge base 2. In the knowledge base 2, an event table which has fields of prevents and postevents is formed.

Description

[Detailed description of the invention]

[Object of the Invention] (Industrial Application Field) The present invention relates to a knowledge construction method for an expert system that has knowledge of each event having a context. (Prior Art) Some expert system knowledge is constructed by linking a plurality of events that have a context. To explain this knowledge, for example, when an input signal is supplied to a certain circuit and its output signal is obtained, there are three events: one input signal circuit and an output signal. In this case, the event after the input signal is the circuit, and the event after the circuit becomes the output signal. Also, the preceding event of the output signal becomes a circuit. Therefore,
If the output signal is normal, it is estimated that the circuit that is the previous event of this output signal is normal, and if the output signal is abnormal, it is estimated that the circuit is abnormal. Furthermore, if the output signal is abnormal and the circuit is normal, the cause of the abnormality is presumed to be the output signal. Here, if we generalize such knowledge and explain with reference to FIG. 6, events A, B, C, and D. There is an event E, and the event after events A and B is C, and this event C
The subsequent event is E, the previous event of H is C, and the previous events of C are A and B. By the way, when building knowledge based on each event that has such a context, the following format rules (1) to (
12) will be created. (1) If D is abnormal and C is normal, D is the cause. (2) If E is abnormal and C is normal, E is the cause. (3) If C is abnormal and A is normal, C is the cause. (4) ) If C is abnormal and B is normal, C is the cause. (5) If A is abnormal, A is the cause. (6) If B is abnormal, B is the cause. (7) If D is normal, C is normal. (8) If E is normal, C is normal (9) If C is normal, A and B are normal (10) If A is abnormal, C is abnormal (11) If B is abnormal, C is abnormal (12) If C is abnormal, E is abnormal, and each of the above rules (1) to (12) is stored in the knowledge base. Here, when the expert system receives an input such as ``Event C is abnormal,'' the inference mechanism infers that the event occurs according to rule (12) and that E is abnormal.Next, the expert system infers that rule (3) In order to determine whether or not this applies, the question is asked, "What about event A?" On the other hand, if you manually say "Event A is normal", the rule (3
), it is inferred that "C is the cause." Also, in order to determine whether rule (4) is applied, the expert system asks the question "What about event B?".In response, if you input "event B is abnormal", rule (4) is applied. 6), it is inferred that "B is the cause." However, when constructing knowledge consisting of events that are related before and after, each rule (1) to (
12) must be created. These rules (1)
~(12) cannot be created easily by anyone, and cannot be created unless one knows how to create rules. Furthermore, as the number of events increases and the number of context relationships increases, the number of rules increases, making it difficult to create rules. (Problems to be Solved by the Invention) As described above, it was difficult to create rules for each event that has a relation between the two. SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a knowledge construction method for an expert system that can easily construct knowledge consisting of events that have a sequential relationship. [Structure of the Invention] (Means and Effects for Solving the Problems) The present invention is characterized in that a plurality of events in a sequential relationship are connected in succession, and one of these events is designated as normal or abnormal. normal for the events before and after this event,
In the knowledge construction method of an expert system that infers anomalies and their causes, the knowledge construction method of an expert system that attempts to achieve the above purpose by storing each event in a context relationship separately into a preceding event and a subsequent event. be. (Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of the expert system. The system is provided with an inference mechanism 1, and this inference mechanism 1
A knowledge base 2 and a database 3 are connected to. Further, a user ink face 4 is connected to the inference mechanism 1 so that input information is sent to the inference mechanism 1. On the other hand, a knowledge acquisition support module 5 is connected between the knowledge base 2 and the user interface 4, so that knowledge input through the user interface 4 is stored in the knowledge base 2. Incidentally, in the knowledge base 2, an event table having columns for preceding events and subsequent events as shown in FIG. 2 is formed. Therefore, when fG exists as a preceding event of event G, the knowledge acquisition support module 5 stores event fG as a preceding event and also stores event G as a subsequent event. Further, the inference mechanism 1 has the following functions. Here, if the preceding event fG and the subsequent event G are 1, if G is normal, then

[G is normal; 2. If G is abnormal, fG is abnormal; 3. If 0 is abnormal and fG is normal, G is the cause 4; if fG is abnormal, G is abnormal. ing. With such a configuration, if the function indicating the preceding event of event G is expressed as fG-pre(G), the inference operation is executed according to the flowchart of the inference process shown in FIG. Post-event G in step sl
is input to the inference mechanism 1 through the user interface 4 as an abnormal event GNG, in step s2 the inference mechanism 1 calculates fG-pre(GNG), and in the next step s3, asks ``What is the event fGN6?
” questions. When the fact that the event fGNG is normal in this state is input to the inference mechanism through the user interface 4, the inference mechanism 1 infers that the subsequent event GNG is the cause and outputs it as a conclusion. Also, the rear 1f elephant fG
When NG is manually determined to be abnormal, the subsequent event GN G - the previous event fGN is determined in step s6. Then, the process returns to step s2. Next, a case where inference is executed based on the event relationship shown in FIG. 6 will be described. As mentioned above, the preceding event of event 1c is event <A, B, and the preceding event of event 1c is event C. Also, event A. The event after B is C, and the event after this event C is C. It is E. Therefore, the knowledge acquisition support module 5 records events A and B in the previous event column of the knowledge base 2. C1C is memorized and these events A and B. Events C, C, and D are in the post-event column corresponding to C1C. E is memorized. When each event is stored in this way, the inference operation is executed according to the flowchart of the inference process shown in FIG. When an event is input to the inference mechanism 1 through the user interface 4 as an abnormal event in step el,
In the next step e2, the inference mechanism 1 creates C-pre(D), and in step e3 asks the question "Is the event C abnormal or normal?" through the user interface 4. When inputting "Event C is normal" in response to this question, the inference mechanism 1 moves from step e4 to step 05 and infers that "the cause of the abnormality is the event." On the contrary, [Is event C abnormal or normal?] When "event C is abnormal" is input in response to the question ", the inference mechanism 1 moves to step 06 and further calculates the functions A, B - previous (C) regarding the previous events A and H. Then, the inference mechanism 1 asks the question, "Are events A and B normal or abnormal?" For example, in response to this question,
When the inference mechanism 1 inputs “Event A is abnormal”, the inference mechanism 1 proceeds from step c7 to step 09 and inputs “!+1 elephant A”.
It is the cause.'' Further, when "Event B is normal" is input to the inference mechanism 1, the inference mechanism 1 proceeds from step e7 to C13 and infers that "Event C is the cause." In this way, in the -I-2 embodiment, each event that has a sequential relationship is stored separately as a preceding event and a subsequent event, so each event that has a sequential relationship also has a preceding event and a subsequent event. Knowledge can be constructed simply by registering each event in each column of the table. Therefore, with such a knowledge base, anyone can easily register. Furthermore, even if the number of events increases, it is only necessary to additionally register only the increased number of events, so it is possible to easily cope with the increase in the number of events. Note that the present invention is not limited to the above-mentioned embodiment, and may be modified without departing from the spirit thereof. [Effects of the Invention] As described in detail above, according to the present invention, it is possible to provide a knowledge construction method for an expert system that can easily construct a process consisting of events having a sequential relationship.

[Brief explanation of the drawing]

1 to 5 are diagrams for explaining the knowledge construction method of the expert system according to the present invention, in which FIG. 1 is a configuration diagram of the expert system, FIG. 2 is a schematic diagram of an event table, and FIG. The figure is a flowchart of inference processing, Figure 4 is a schematic diagram when each event is registered in the event table, Figure 5 is a flowchart of inference processing for the event table in Figure 4, and Figure 6 is a schematic diagram of each event in the event table. It is a schematic diagram showing an event. 1... Reasoning mechanism, 2... Knowledge base, 3... Database, 4... User interface, 5... Knowledge acquisition support module □ Figure 2 Figure 6LA Figure

Claims (1)

[Claims] When a plurality of events that have a sequential relationship are consecutively connected, and one of these events is designated as normal or abnormal, the events before and after this event are designated as normal, abnormal, and the cause of the abnormality. A knowledge construction method for an expert system that makes inferences, characterized in that each of the events in which a context is established is stored separately as a preceding event and a subsequent event.