JPH08328865A - Rule acquisition method for expert system - Google Patents

Abstract

PURPOSE: To reduce the time and labor needed for repetition of the same correction in an expert system by fetching the correction given to the inference result into a knowledge base as a new rule as long as the correction is repeatable. CONSTITUTION: An inference part 2 performs the inferences based on the rules stored in a knowledge base 1 and outputs these inference results to an input/ output device 3. When the inference result is changed via the device 3, this change history information is stored in a change history store part 5 via a change history collection part 4. A rule extraction part 6 produces a rule from the change history information that is stored in the part 5 and repeated in several times and then stores the rule in an extracted rule store part 7. A rule addition part 8 stores the produced rule in the base 1 by an instruction of the device 3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a knowledge acquisition method in an expert system, and in particular, obtains a change history of changes made to an inference result and newly adds it to a knowledge base based on this change history. Regarding how to extract rules.

[0002]

2. Description of the Related Art An expert system is a system in which experience knowledge and know-how of experts are accumulated in a knowledge base in the form of rules, and a computer makes inferences based on this knowledge base. Actually, the work of constructing and maintaining a knowledge base requires a lot of man-hours, and some proposals have been made on how to efficiently acquire knowledge according to reality.

For example, Japanese Patent Laid-Open No. 63-311441, "Knowledge Acquisition Method Using Time History", accumulates the result of inference by an expert system as an inference history, and from this inference history the certainty factor of a conclusion description belonging to the same category. A knowledge acquisition method is disclosed in which a calculation procedure is calculated, a processing procedure (rule) is generated from a conclusion description according to a certainty factor, and is registered in a knowledge base.

[0004]

According to the prior art, a rule to be added is created based on the inference result inferred by the expert system and registered in the knowledge base. However, there is a problem that the missing condition or the missing rule that was overlooked when constructing the knowledge base cannot be added to the knowledge base. Especially when the user artificially corrects the inference result output by the expert system, if such correction is reproducible, there is a problem that the same correction must be repeated many times. is there.

An object of the present invention is to reduce the time and effort for making similar corrections many times by incorporating them into the knowledge base as new rules when the corrections made to the inference results are reproducible. It is in.

[0006]

There is a scheduling problem in which a plurality of lots are sequentially allocated to production equipment in a production plan, and an expert system is applied to the scheduling of the lots. The information about each lot is composed of a plurality of data items such as a product number and the number of stocks. The knowledge base also stores rules for setting the priority of allocation to production equipment according to the values of data items related to scheduling. The inference mechanism of the expert system applies this rule to the input information about the lot and infers that a plurality of lots should be sequentially allocated to the production equipment.

According to the present invention, when a lot allocation order is changed with respect to an inference result, information about a lot changed in relation to another lot is accumulated as change history information, and this change history information is assigned. When the change history information that has the same value of the data item related to the priority of is accumulated a predetermined number of times or more, a rule is created to include the data item having the same value that appears in the change history information in the condition part. It is characterized by a rule acquisition method that additionally stores it in the knowledge base.

[0008]

The fact that the same change occurs repeatedly with respect to the inference result under the same condition with respect to scheduling is considered to be a lack of rules stored in the knowledge base. According to the present invention, a rule is created from the change history information such that the condition section includes a data item related to the lacking priority of allocation, and the rule is registered in the knowledge base. This makes it unnecessary for the user to change the inference result thereafter.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a diagram showing the configuration of the expert system of this embodiment. Knowledge base 1 is a database that registers rules used for inference. The inference unit 2
This is a processing unit that draws out rules from the knowledge base 1 and infers them, and displays the results on the input / output device 3. The change history collecting unit 4 monitors whether or not the inference result displayed on the input / output device 3 is changed by the input / output device 3, and when the change is made, stores the change history in the change history storage unit 5. It is a department. The rule extraction unit 6 is a processing unit that refers to the change history storage unit 5 to extract a rule from the change history and stores the rule in the extraction rule storage unit 7. The rule addition unit 8 is a processing unit that adds the rules stored in the extraction rule storage unit 7 to the knowledge base 1 according to an instruction from the input / output device 3. Inference unit 2,
Change history collection unit 4, rule extraction unit 6, and rule addition unit 8
Is realized by executing a program stored in the storage device of the information processing device. The knowledge base 1, the change history storage unit 5, and the extraction rule storage unit 7 are files or databases stored in the storage device of this information processing apparatus. The input / output device 3 is a peripheral device or a terminal device connected to the information processing device.

FIG. 2 is a diagram showing an example of applying this expert system to a production planning problem, and is a diagram showing information on lots to be scheduled. The product number of the product or part produced for each lot, the number of products produced, and the current inventory quantity are shown. As such, lot information is generally composed of a plurality of data items, some of which are related to scheduling priority and others are not.

FIG. 3 is a diagram showing an example of rules registered in the knowledge base 1. The rules handled in this embodiment are generally expressed in the following description format. (Rule Name Processing Procedure Division if Condition Section) FIG. 3 shows an example of three rules. Rule 01, rule 02 and rule 03 are uniquely set rule names. Each rule is composed of a condition section and a processing procedure section. The example of FIG. 3 shows the priorities of lots assigned to the machines, and the priorities of 0.7, 0.6 and 0.5 are set for the product numbers a, b and c, respectively.
In general, the rule puts the value of the data item related to the scheduling priority in the condition part, and accordingly puts the process of setting the priority in the processing procedure part.

The inference unit 2 starts scheduling with three lots, lot A, lot B, and lot C as allocation targets. Read rules from knowledge base 1,
Lot A is "Rule 01", Lot B is "Rule 02"
And lot C satisfies the condition part of "rule 03",
Lot A is given a priority of 0.7, lot B is given a priority of 0.6, and lot C is given a priority of 0.5. The inference unit 2 first allocates lots A to machines, then allocates lots B and C to machines sequentially, and displays the results on the input / output device 3 in order to allocate the machines to the machines in order from the highest priority.

FIG. 4 is a Gantt chart showing an example of the inference result displayed on the input / output device 3, in which the horizontal axis indicates the date scale and the vertical axis indicates the production equipment used for production, and the lots are arranged in chronological order. Indicates the status assigned to production equipment. In this example, lot A, lot B, and lot C to be scheduled are assigned to machine A, which is a production facility.

In FIG. 5, the user changes the allocation order of the priority of the lot C on the Gantt chart displayed on the input / output device 3 by the user with respect to the inference result shown in FIG. It is a figure which shows the appearance.

The change history collection unit 4 detects that the inference result has been changed, acquires lot information through the inference unit 2, and stores the change history in the change history storage unit 5.

FIG. 6 is a diagram showing an example of the data format of the change history storage unit 5. The change history is composed of lot information, scheduling results, and subsequent lot information. The change history information 37 is the change history when the inference result is changed in the past, and the change history information 43 is the change history when the inference result is changed this time, as shown in FIG. The change history information 43 is assigned as a result of the change of the allocation of the lot C, the information 39 including the information of the lot A and the information of the lot C allocated before the lot C, and the information of the lot C and the information after the lot C. It is composed of information 41 including information on lot B.

FIG. 7 is a flow chart showing the flow of processing of the rule extraction unit 6. The rule extraction unit 6 inputs the change history information stored in the change history storage unit 5 (step 21) and determines whether or not there are n or more identical change history information items (step 22). The same change history information is change history information such that the values of data items related to scheduling among a plurality of data items match. Further, n is a predetermined integer of 1 or more. In the case of the present embodiment, the data items of product number and inventory = 0 match, and the product number and inventory = 0 are data items related to the priority of allocation. Further, there are two pieces of the same change history information in the above meaning. If the number of changes is n or more (YES in step 22), a rule is created from the corresponding change history information and displayed on the input / output device 3 (step 2).
3). If the user specifies that the rule created through the input / output device 3 is added to the extraction rule storage unit 7 (YES in step 24), the rule name input from the input / output device 3 is obtained (step 25), The extracted rule is stored in the extraction rule storage unit 7 (step 26). If the same change history information is less than n (step 22 NO) and the user does not select the addition of the rule (step 24)
If NO), the process ends.

FIG. 8 is a diagram showing an example of a message displayed on the input / output device 3 and a message for allowing the user to select whether or not to add the rule. The rule is created in a format in which the same content of the change history information is used as the condition section and the procedure for setting the priority of the target lot is used as the processing procedure section.

FIG. 9 is a diagram showing an example of a screen displayed on the input / output device 3 for displaying the message for prompting the setting of the rule name and the set rule name.

FIG. 10 is a flow chart showing the flow of processing of the rule adding section 8. The rule adding unit 8 reads out the rules stored in the extraction rule storage unit 7, and displays a list of the rule names on the input / output device 3 (step 51). The user inputs the knowledge base 1 via the input / output device 3.
When you select the rule you want to add to (step 52),
Add the selected rules to Knowledge Base 1 (step 5
3), delete the rule in the extraction rule storage unit 7 (step 54). When there are no more rules to add, the process ends. After that, if the inference unit 2 is executed,
The rules added along with the existing rules are executed.

In FIG. 11, the rule adding section 8 is used for the input / output device 3
FIG. 6 is a diagram showing an example of a list of rule names displayed in FIG. When the user requests to display the content of each rule, the content of the specified rule is displayed. User is knowledge base 1
If a rule to be added to is selected, the selected rule is added to the knowledge base 1.

According to the above-described embodiment, after allocating lots of product number = "a" and stock = 0 to the machine, lots of product number = "b" and product number = "c" are to be allocated. In the knowledge base 1, the know-how of allocating the lot with the product number = “c” is reflected in the knowledge base 1. Thereafter, in such a case, it is not necessary to change the inference result.

Another embodiment of the present invention will be described below.

FIG. 12 shows that the color data item is added to the case shown in FIG. A color is a color that colors a product or part.

When the inference unit 2 infers the scheduling target by applying the rule shown in FIG. 3, the inference result shown in FIG. 4 is obtained. When the inference result obtained here is changed as shown in FIG. 5, the change history collection unit 4 stores the change history information in the change history storage unit 5.

FIG. 13 is a diagram showing an example of the data format of the change history storage unit 5 in this embodiment. Color is a data item related to scheduling, and there are two pieces of change history information whose color matches the product number. Rule extraction unit 6
Creates a rule based on these two identical change history information.

FIG. 14 is a diagram showing rules created from this change history information. According to this, product number = ”
After allocating a lot of “a” and color = “yellow” to the machine, if there is a candidate lot of part number = “c” and color = “blue”, it is assigned to this machine with the highest priority.

[0029]

According to the present invention, if the change made to the inference result is reproducible, it is considered that the rules of the knowledge base are insufficient and a new rule is added to the knowledge base. Therefore, the lack of knowledge can be quickly supplemented and a more sophisticated knowledge base can be constructed.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an expert system according to an embodiment.

FIG. 2 is a diagram showing an example of lot information input to an expert system.

FIG. 3 is a diagram showing an example of rules registered in a knowledge base 1.

FIG. 4 is a Gantt chart showing an example of an inference result.

FIG. 5 is a Gantt chart showing an example of an inference result changed by a user.

FIG. 6 is a diagram showing an example of change history information.

FIG. 7 is a flowchart showing a processing flow of a rule extraction unit 6 of the embodiment.

FIG. 8 is a diagram showing a display example of created rules and messages.

FIG. 9 is a diagram showing a display example of setting a rule name.

FIG. 10 is a flowchart showing a processing flow of a rule adding unit 8 of the embodiment.

FIG. 11 is a diagram showing a display example showing a list of rule names and urging the user to select a rule to be registered.

FIG. 12 is a diagram showing an example of lot information according to another embodiment.

FIG. 13 is a diagram showing an example of change history information of another embodiment.

FIG. 14 is a diagram showing an example of a rule created according to another embodiment.

[Explanation of symbols]

1 ... Knowledge base, 2 ... Inference unit, 4 ... Change history collection unit, 5 ... Change history storage unit, 6 ... Rule extraction unit

Claims (1)

[Claims] 1. A plurality of lots are sequentially produced into production equipment based on information about a lot composed of a plurality of data items and a rule stored in a knowledge base and setting an allocation priority according to a value of the data item. In the knowledge acquisition method of the expert system that infers to allocate, when the allocation order of the lots allocated to the production equipment, which is the result of the inference through the input means, is changed, the lots changed in relation to other lots Information is accumulated as change history information, and when change history information having the same data item value related to the priority of allocation for the change history information is accumulated a predetermined number of times or more, it appears in the change history information. Create a rule that includes data items with the same value in the condition part, and store the created rule in the knowledge base additionally. Rule acquisition method in the expert system according to claim.