JPH01161536A - System for executing inference of artificial intelligence system - Google Patents

Abstract

PURPOSE:To reduce the processing time of a rule conflict dissolving process by selecting one rule out of the separately managing rules in a specific module by comparing relatively added rule priority. CONSTITUTION:A pattern matching processing means 21 fetches the state change of an inference object 11 and finds a rule which meets the current state change in a rule base storing means 13. A conflict setting means 22 collects the rules found by means of the means 21. When plural rules exist, a rule module priority discriminating means 23 decides which module has the highest priority and selects the highest-priority module. When plural rules further exist to another one module, a rule priority discriminating means 24 selects the rule having the highest priority. Thereafter, a time tag discriminating means 25 selects the rule having the new phenomenon and executes the rule when plural rules still exist at the means 24. Therefore, an executed result is obtained by two times of comparison of the module priority comparison and selected module priority comparison.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an inference execution method for an artificial intelligence system that improves the rule conflict resolution processing means during inference execution in the artificial intelligence system.

(Prior Art) In recent years, there has been a growing trend to use electronic computers to control inference objects by providing optimal information to them while making appropriate judgments according to the constantly changing operating conditions of the inference objects.
Along with this, the development of artificial intelligence (hereinafter referred to as AI) systems is proceeding at a rapid pace. In this AI system, a knowledge expert assumes a change in the state of an object to be inferred in advance, creates processing rules to deal with the change in state, and then makes inferences based on the knowledge expert's rules according to the change in the state of the object. This is a system that controls the operation of an inference object while executing. Here, the inference execution of the AI system means comparing the rules that are the knowledge of experts with the current state of the object to be inferred, and when there are multiple rules that match the current state, select one of the most likely rules. It refers to the flow of a series of processes that are executed for one rule after performing so-called conflict resolution processing of rules, and specifically, as shown in Figure 4, rule-based intelligence A pattern matching processing means 1 for determining whether or not there is a rule matching the current state among the rules; a conflict set processing means 2 for collecting rules selected by the pattern matching processing means 1; and a conflict set processing means 2 for collecting rules selected by the pattern matching processing means 1. A rule conflict resolution processing means 3 for selecting one rule from among the rules selected by the means 2 according to the priority, and a rule execution means 4 for executing the one rule selected by the rule conflict resolution processing means 3, etc. It consists of

Conventionally, the rule conflict resolution processing means 3 uses an inference execution method in which all rules are prioritized and the rule with the highest priority is determined according to the priority.

This means that, for example, when five rules are collected in the conflict set processing means 2, it is necessary to compare the priorities of all five rules at the stage of the rule conflict resolution processing means 3.

(Problem to be Solved by the Invention) Therefore, when using the above-described rule conflict resolution processing means, as the objects to be inferred become more and more complex and sophisticated, the conflict set processing means 3 The number of rules to be applied increases, and comparing the priorities of all these rules requires a large amount of time.
Furthermore, there is a problem in that it is not possible to select one rule and quickly reflect the result of the inference execution on the object to be inferred.

The present invention has been made to solve the above problems,
It is an object of the present invention to provide an inference execution method for an AI system that shortens the processing time of rule conflict resolution processing, quickly reflects the inference execution results on the inference target, and ensures stable control.

[Structure of the Invention] (Means and Effects for Solving Problems) In the inference execution method of the AI system according to the present invention, a plurality of rules are collected by a conflict set processing means according to the current operating state of the inference target. Then, a specific module is selected according to the priority of the plurality of modules to be shared and managed for the plurality of rules, and the above-mentioned shared management in this specific module is performed. 1 while comparing the relative priority of each rule.
By selecting one rule, it is possible to reliably select one rule that corresponds to the current operating state of the object to be inferred with a small number of comparisons.

(Example) Hereinafter, an example of the apparatus of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the AI system. In the figure, reference numeral 11 indicates objects to be inferred such as various plants or various types of equipment that undergo state changes. Reference numeral 12 denotes a rule base input means for manually inputting prioritized rules for modularizing the knowledge of what actions to take for each state change of the inference target object 11 and managing this module in a shared manner. Rules input from the base input means 12 are stored in the rule base storage means 13. Therefore, the rule base storage means 13 stores prioritized rules that are managed by the prioritized modules.

Reference numeral 20 denotes an inference execution means that selects one appropriate rule from the rule base storage means 13 in response to a change in the state of the inference object 11, executes inference, and reflects the inference result on the inference object 11. The inference execution means 20 includes a pattern matching processing means 21 for determining whether or not there is a rule that matches the current state change of the inference object 11 from among the rules stored in the rule base storage means 13; The pattern matching processing means 21 includes a conflict set processing means 22 which collects the rules when there are a plurality of rules that match the current state change, and further includes a rule module priority judgment function which corresponds to a rule conflict resolution processing means. The system includes means 23, rule priority determining means 24, time tag determining means 25, etc., and is configured such that a rule executing means 26 executes one rule selected by the rule conflict resolution processing means.

The module priority determining means 23 has the rules collected by the conflict set processing means 22 made into modules with priorities according to purposes, former prisoners, etc.
It has a function that determines that the module with the highest priority among those modules is valid. The rule priority determining means 24 has a function of comparing the rule priorities added relative to the rules in the module that are determined to be valid by the rule module priority determining means 23. The time tag determining means 25 selects one rule from the time tags added to state events.

Next, we will discuss the operation of the device configured as described above in the second section.
This will be explained with reference to the figures. First, the pattern matching processing means 21 takes in the state change of the inference target object 11, and then
A rule matching the current state change is found from among the rules stored in the rule base storage means 13 (step 81).

Thereafter, in the conflict set processing means 22, as shown in step S2, the pattern matching processing means 21 collects any rules that match the current state change. Therefore, the purpose of this conflict set processing means 22 is to collect a large number of rules.
Furthermore, these rules are divided into two modules A and B, for example, as shown in the following table.

Table: Examples of rules that have been grouped. Therefore, the rule module priority determining means 23 uses the rule modules A that have been grouped as shown in step S3 (
It is determined which of the modules B (priority 1) and module B (priority 2) has a higher priority. In this case, module B having a higher priority is determined to be effective.

Here, regarding the module B determined to be valid, the rule priority determining means 24 determines the rule 4 as shown in step S4. Rule 5 Compare priorities between each other. At this time,
In rule 4, the priority is "1", and in rule 5, the priority is "1".
2'', therefore, as a result of comparing these priorities, rule 5 is selected. Thereafter, as shown in step S5, the time tag determining means 25 selects a rule having a new event when the rule priority determining means 24 has a plurality of rules. And if one rule is selected,
The rule is executed (step S6), and after the execution, it is determined whether or not the inference is completed (step S7). If the inference is not completed, the process returns to step S1 and the same process is performed.

Therefore, according to the configuration of the embodiment as described above, the rule module priority determining means 23 selects a module with a high priority among each module with respect to a set of rules collected by pattern matching and conflict set processing. Since the rule priority determination means 24 compares the priorities of each rule in a module with a high priority and selects one rule, the number of rule comparisons can be significantly reduced compared to the conventional method. , the inference execution time can be reduced accordingly. For example, if five rules are collected by the conflict set processing means 22 as shown in the table above, in order to apply this to a conventional device and select one rule, it is necessary to perform four comparisons.
With this device, a total of two comparisons are required: one module priority comparison and one rule priority comparison of the selected module, and the inference execution result can be obtained with just five and a half comparisons compared to the conventional method.

In the above embodiment, the priority of the rule module is determined after the conflict set processing, but for example, as shown in FIG. 3, a certainty processing means is newly provided after the conflict set processing means 22, and the Even if the configuration is such that a rule whose confidence level is higher than the reference value is selected from among the plurality of rules selected by means 22, and selected from among them based on the current module priority, rule priority, and time tag. good. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As described in detail above, according to the present invention, each rule after the Cunflicht processing is selected according to the priority of the rule module, so that the processing of rule conflict resolution processing is improved compared to the conventional device. It is possible to provide an inference execution method for an AI system that can significantly shorten the time, quickly reflect the inference execution results on the object to be inferred, and ensure stable operation control.

4. Detailed Description of the Invention Figures 1 and 2 are shown to explain an embodiment of the inference execution method of the AI system according to the present invention.
Fig. 1 is a block diagram showing the configuration of the AI system, Fig. 2 is an operation flowchart of the inference execution method, Fig. 3 is a block diagram showing the configuration of another example of the device of the present invention, and Fig. 4 is the conventional inference execution method. FIG. 2 is a block diagram illustrating the method.

DESCRIPTION OF SYMBOLS 11... Inference object, 13... Rule base storage means, 20... Inference execution means, 21... Pattern matching processing means, 22... Conflict set processing means, 23... Rule module Priority judgment means, 24
. . . Rule priority determination means, '25 . . . Time tag determination means, 26 . . . Rule execution means.

Applicant's agent: Patent attorney Takehiko Suzue 20th 30th 4th B

Claims (1)

[Claims] An artificial intelligence that, when a plurality of rules matching the current state of an object to be inferred are selected from among predetermined rules, selects one rule using a rule conflict resolution processing means and executes inference. In the inference execution method of the intelligent system, the rule conflict resolution processing means selects a specific module according to module priority among the plurality of modules that respectively manage the plurality of rules, and selects a specific module from among the specific modules. An inference execution method for an artificial intelligence system, characterized in that one rule is selected while comparing rule priorities added to each of the rules to be managed in a shared manner.