JPH10326191A - Expert system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce keyboard operation and to obtain a variety of descriptive expressions by actualize a rule defining operation on a fuzzy rule definition screen in a rule table style and a metarule definition screen by the four rules of arithmetic and rule processing through mouse operation. SOLUTION: The functions of an operation support system are divided into that of rule editing computers 17 to 19 and a rule execution computer 16, and GUI developing libraries and rule editing software is mounted on the OS and GUI systems of the rule editing computers 17 to 19, which are connected to the rule execution computer 16 by a network 20. Those rule editing computers 17 to 19 enable rule definition on the fuzzy rule definition screen in a rule table style and a metarule definition screen by the four rules of arithmetic and rule processing through mouse operation which is all intuitive except numeral input. Consequently, the keyboard operation is reducible and a variety of descriptive expressions are obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an expert system using fuzzy rules and meta rules in plant control.

[0002]

2. Description of the Related Art FIG. 11 shows a fuzzy rule editing screen in a conventional expert system, in which 27 is a fuzzy rule definition screen, 28 is an operation menu showing keys corresponding to operation contents, and 29 is Is a fuzzy rule description area. Reference numerals 2, 3, and 30 denote lists of a condition part, a conclusion part, and a splice necessary for describing in the fuzzy rule description area, respectively, and the user describes a fuzzy rule by selecting a word from these lists.

[0003] F1 shown in the operation menu 28
By inputting the F3 key, a list of corresponding condition parts, conclusion parts, and junctions can be selected. The user sequentially switches the selection of items in each list by inputting the space key, and inputs a RETURN key when the target item is reached. The input of the RETURN key causes the selected item to be written into the fuzzy rule description area 29.

FIG. 12 shows a screen for editing a meta-rule in a conventional expert system. Here, the meta-rule defines the execution procedure of the fuzzy rule defined in FIG. 11, and uses a conditional branch by a rule variable. You can also.

[0005] In the figure, 31 is a meta-rule definition screen,
Reference numeral 32 denotes an operation menu similar to the above-described operation menu 28, in which keys corresponding to operation contents are shown. 33 is a meta-rule description area. Reference numerals 13, 12, 30, and 34 are lists of fuzzy rules, rule variables, splicers, and operators required to describe the meta-rule. Here, the rule variable is an internal variable used at the time of executing the rule. By linking the rule variable with the plant data, the plant data value is reflected in the rule processing.

As in the case of the fuzzy rule definition, the user selects and decides an item in each list by using a combination of the F1 to F4 keys, the space key, and the RETURN key, and describes the selected item in the meta rule description area 33.

[0007] As another example of the conventional rule editing user I / F, a "knowledge base construction tool" disclosed in JP-A-3-269630 and an "expert system" disclosed in JP-A-5-224935 are disclosed. Meta-rule generation method ". In the "knowledge base construction tool", expert rules based on if-then-else expressions are described using menu selection with a mouse. In the "method of generating a meta-rule of an expert system", the flow of starting and ending the expert rule itself is defined by an effective graph expression, and matrix format data for internal processing is automatically generated from the effective graph.

Next, the configuration of a conventional expert system will be described. FIG. 13 is a system configuration diagram of a dedicated computer and a dedicated terminal. In FIG. 13, reference numeral 35 denotes a rule execution computer as a computer dedicated to the expert system, and reference numeral 36 denotes a dedicated terminal provided as a rule editing computer. A user I / F for rule editing is displayed on the dedicated terminal 36 display. 21 is a plant bus.

At the time of rule editing, the user performs rule editing processing using the dedicated terminal 36 as an operation terminal. Further, at the time of rule execution, the rule execution computer 35 performs inference processing while receiving plant data from the plant bus 21.

FIG. 14 shows a conventional expert system composed of one computer or a computer such as EWS.
In FIG. 14, reference numeral 37 denotes a PC or EWS as a general-purpose computer, and a rule editing user I / F is displayed on a display of the general-purpose computer 37 at the time of rule editing. At the time of rule execution, the general-purpose computer 37 performs inference processing based on plant data obtained from the plant bus 21.

In the rule editing screen, the user has to use a lot of complicated keyboard operations when switching between various lists of the elements constituting the rule and selecting rules in each list. Also, in order to learn keyboard operations, it is necessary to remember the correspondence between each operation and a key. Furthermore, when executing the actual rules, there may be a case where, in addition to the data directly obtained from the plant bus, a processing operation such as a totaling and averaging process in which these are subjected to four arithmetic operations is required. To cope with these problems, a method of allocating the machining operation data itself as one data to a plant database commonly used in the entire inference process can be considered. However, in this method, every time processing data is added locally generated at the time of creating an individual rule, a change must be made to a database common to all the rules, which is expected to be an expensive operation. It is desirable that the use of local processing data be absorbed within the rules.

For example, it is assumed that data 1, data 2,..., Data 10 are registered in a plant database. According to a certain rule, the total value of data 1 to data 6,
In another rule, it is assumed that an average value of data 2 to data 9 is required for an operation inside the rule (processed data generated locally). At this time, registering the following data in the plant database means changing the database common to all the above. Data 11 = Data 1 + Data 2... + Data 6 Data 12 = (Data 2+... Data 9) / 8

Although it depends on the implementation of the system, it is generally considered that changing the database common to the entire system often requires stopping and restarting the entire system. Registering data that is used only in individual rules in a database that aims to organize data that is frequently shared,
Deviates from the original operation policy. Here, such an adverse effect on the entire system is referred to as “costly work”.

However, the description of the conventional fuzzy rules and meta-rules shown in the "knowledge base construction tool" in FIG. 11 and the "method for generating meta-rules in the expert system" in FIG. 12 is only conditional branching by if-then-else. It is difficult to define a processing operation using the four arithmetic operations inside a rule.

In general, in defining a fuzzy rule, a plurality of conditions and conclusions are defined in one fuzzy rule. Then, the final control output value is determined for the entire grade value of the conclusion part of each sentence by defuzzification operation such as calculation of the center of gravity. However, FIG.
In the conventional description format of expert rules shown in “Knowledge Base Construction Tool”, these plural conditions and conclusions have to be expressed in if-then format as they are.
If the rule content is small, there is no problem with this,
When a large-scale rule is used for actual operation in a plant, the creation and maintenance of the rule becomes difficult due to a decrease in readability (easiness of understanding the content).

Further, in the system configuration shown in FIG. 13, a dedicated terminal and a dedicated computer are required for rule editing and inference (rule) execution, which greatly expands the performance and greatly adversely affects integration with other plant monitoring and control systems. . Although a general-purpose computer is used in the system configuration of FIG. 14, the rule editing function and the inference execution function are integrated on a single computer. It cannot respond to requests for remote rule editing, a distributed processing mode in which rule editing processing and inference execution are divided, and the like.

Generally, in an expert system,
In many cases, a rule created by the rule editing function is converted into a certain internal expression format, and the rule execution computer (inference engine) interprets the rule to execute inference.
For example, in the “method of generating a meta-rule of an expert system” in FIG. 12, the rule contents are internally represented as matrix-format data. Although the internal representation of rules depends on each system, their readability is generally low and it is difficult for humans to directly understand the contents. In this case, in order to understand the rule contents, a dedicated rule editing I / F is required.
Is required. For this reason, the rule editing function and the rule execution function become inseparable, and the independence of each function decreases.

[0018]

Since the conventional expert system is configured as described above, it has a non-scalable system configuration built on a dedicated computer. However, there is a problem that the distribution arrangement cannot be performed.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems. The keyboard operation is reduced as much as possible, intuitive editing operability using a mouse, description of four arithmetic operations in a meta-rule, a rule table table are provided. It is an object of the present invention to provide an expert system that realizes various rule description expressions such as fuzzy rule description in a format. Also,
It is an object of the present invention to provide an expert system that realizes a system configuration with a high degree of freedom by separating the functions of an operation support system into a rule editing computer and a rule execution computer and increasing the independence of each.

[0020]

An expert system according to the present invention has a rule table format in which a GUI development library is placed on an OS and a GUI system, and a rule editing software is placed on the GUI development library. A rule editing computer that realizes each rule definition operation by a mouse operation in a fuzzy rule definition screen and a meta-rule definition screen by four arithmetic operations and rule processing.

[0021] In the expert system according to the second aspect of the present invention, the rule editing computer having a different OS and GUI system includes a GUI development library and rule editing software so as to absorb the difference between the respective OSs and GUI systems. Is what it is.

In the expert system according to the third aspect of the present invention, the function of the operation support system is divided into a rule editing computer and a rule execution computer, and an OS of the rule editing computer and a GUI development library and a rule editing computer are provided on a GUI system. The software is loaded, and the rule editing computer and the rule execution computer are connected to a network.

An expert system according to a fourth aspect of the present invention provides an expert system comprising: a rule editing computer and a rule execution computer;
/ F is a rule file in the C language format.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Embodiment 1 FIG. FIG. 1 shows a fuzzy rule definition screen diagram according to Embodiment 1 of the present invention.
1 is a fuzzy rule definition screen of the rule editor displayed on the display of the rule editing computer of the expert system, 2 is a list of condition parts of the fuzzy rule, 3 is a list of conclusion parts, and 4 is a fuzzy rule expressed in a rule table format. It is. The user defines a fuzzy rule by dragging and dropping a condition in the condition part list 2 and a conclusion in the conclusion part list 3 in the rule table 4 with a mouse. 5, 6 are regions for dropping the condition part, 7, 8 are regions for dropping the conclusion part, and 9, 10 are the grades (“slightly high”, “slightly open”) defined in the condition part and the conclusion part, respectively. , Etc.) are displayed.

Next, the operation will be described. First, Drag & Drop of the condition part list 2 will be described. The user drags an arbitrary item in the condition part list 2 and drops it to the area 5 or the area 6. When dropped into the area 5, it is automatically added to the end of the already defined condition part. When the item is dropped in the area 6, a dialog asking whether to add the item before or after the item at the drop position is displayed. After the user specifies the item, the item is added to the condition part. At the time of addition, the grade defined in the added condition item is automatically developed in the area 9.

In the example of FIG. 1, the condition part is created by the following operation. (1) Drag “water level 1” to area 5 from condition list 2
& To Drop (step ST1 _1). (2) Drag “Temperature 1” to Area 5 from Condition List 2
& Drop (step ST1 ₂ ). (3) Drag & Drop the “pump rotation speed” from the condition part list 2 to the area 5 (step ST1 ₃ ).

Next, Drag & Drop in the conclusion will be described. The user sets an arbitrary item in the conclusion part list 3 to Dr.
Ag and drop to the area 7 or the area 8. Area 7
Is automatically added to the end of the already defined conclusion part. If a drop is made in the area 8, a dialog asking whether to add the item before or after the item at the drop position is displayed, and after the user's specification, it is added to the conclusion part. At the time of addition, the grade defined in the added conclusion item is automatically added to the area 10
Expanded inside.

In the example of FIG. 1, the condition part is created by the following operation. (1) From the conclusion part list 3, "Valve 1" is Dra in area 7.
g & Drop (step ST1 _4). (2) From the conclusion list 3, “Valve 2” is set to Dra
g & Drop (step ST1 _5).

The association between the condition part list 2 and the conclusion part list 3 is performed by putting a mark in the rule table. For one grade in Conclusion List 3, OR of grades marked with ● in the same condition, and then AND between condition items is used as a conditional expression. The rule table in FIG. 1 represents the fuzzy rules shown in FIG.

The user marks a circle in the rule table 4 with a mouse operation. Move the mouse pointer to an arbitrary position in the rule table 4 where you want to add a mark, and click the 1 button of the mouse there.

As described above, according to the first embodiment, the definition of the condition part list 2 and the conclusion part list 3 is defined by pasting by drag & drop of the mouse, and the mark ● is moved and clicked by the mouse pointer. Thus, fuzzy rules can be defined by intuitive mouse operations that do not require keyboard operations. In addition, a plurality of if-then statements defined in one fuzzy rule can be compactly represented by a rule table format expression.

Embodiment 2 FIG. FIG. 3 shows a meta-rule definition screen according to the second embodiment of the present invention. In FIG. 3, reference numeral 11 denotes a meta-rule definition screen of a rule editor displayed on a display of a rule editing computer of the expert system. 12 is a rule variable list, and 13 is a fuzzy rule list. As shown in the rule variable list 12, an array type rule variable can also be used.
Each content in the fuzzy rule list 13 is created and edited on the fuzzy rule definition screen of the first embodiment. Reference numeral 14 denotes a logic component list, and the four arithmetic operations, conditional branches, and the like are described using these components. Also, there are "for" parts and "while" parts having functions similar to those of the C language for and while statements, and by using these, loop processing can be described. Reference numeral 15 denotes a meta rule editing area, in which lists 12, 13, and 14 are listed.
, A rule variable, a fuzzy rule, and a logic component are dragged and dropped, and a meta rule is described.

FIG. 4 is a diagram for explaining a meta-rule editing operation procedure using a rule editor.
The process of describing a meta-rule by an operation is shown. First, fuzzy rule 1 and fuzzy rule 2 from fuzzy rule list 13 and logic component list 14
Drag & Drop the if part (step ST4)
_1, step ST4 _2, Step ST4 _3). here,
Dropped if parts are conditional part, then part, els
e has three regions, each with a further part Dr
It has a structure that allows op.

Next, fuzzy rule 2, fuzzy rule 3, fuzzy rule 4,
Fuzzy rule 5 as well as Drag & Drop, the condition part to Drag & Drop logic components ">" (step ST4 _4, step ST4 _5, step ST
4 _6, step ST _7, step ST4 _8). The following parts including the part ">" =, +,-, x, /, <, <=,>,> =, ==,! = Are all binary operators (operators that require an rvalue and an lvalue), and have a structure that allows additional logic components to be dropped or numeric values input to the right and left sides after dropping. Hereafter, Drag & Dr of the parts in the order shown below
By inputting "op" and a numerical value, an arithmetic expression of the condition part is created.

[0035] (1) Parts ">" parts on the left-hand side value of "/" to Drag & Drop (step ST4 _9). (2) Drag & D the part “+” to the left side value of the part “/”
rop (step ST4 ₁₀ ). (3) components "+" to "variable 2" on the right-hand side value "variable 1" in the left-hand side value Drag & Drop to the (ST4 _11, ST
4 ₁₂ ). (4) The numerical value 2.0 is input to the right side value of the component “/”. (5) Enter the numerical value 4.0 as the right-hand side value of the component “>”.

As described above, the meta-rule can be defined by intuitive mouse operation except for numerical input. In the example, the case where a simple averaging process is described in the condition determination unit in the if statement by using the components “>”, “/”, and “+” is shown. A description of the process is also possible. With such a program-like flexible rule description capability, it is possible to absorb the average, aggregation, and other special operations that occur in each meta-rule in that rule.

Embodiment 3 FIG. 5 shows a system configuration of an expert system according to Embodiment 3 of the present invention. The function of the operation support system is divided into rule editing computers 17 to 19 and a rule execution computer 16, and the rule editing computers 17 to A GUI development library and rule editing software are loaded on the 19 OS and GUI system, and the rule editing computers 17 to 19 and the rule execution computer 1
6 are connected by a network 20. On these rule editing computers 17 to 19, a rule editor operates. In FIG. 5, as an example, the rule editing computers 17-1
9 are connected, but the number of rule editing computers is arbitrary in this system configuration. In the rule execution computer 16, an inference engine operates.

The network 20 is a general-purpose LAN.
22-a, 22-b, 22-c and 23 are rule files stored on the hard disks of the rule editing computers 17 to 19 and the rule execution computer 16, respectively. 22-
a, 22-b, and 22-c are temporary rule files transferred from the rule execution computer to the rule editing computer for rule editing. Reference numeral 23 denotes a rule file read by the inference engine from the plant bus 21 when executing the rule.

Next, the operation of the system shown in FIG. 5 will be described with reference to the flowchart of FIG. First, the rule editor issues a transfer request rule file to the inference engine (step ST6 _1). In response to this, the inference engine transfers the rule file (step ST6 _2). When the file transfer is completed, the rule editor displays the contents of the received rule file. The user edits the rule on the rule editor by the operation method described in the first and second embodiments. When the rule editing is complete, the rule file is transferred to the inference engine from the rules editor contrary to Step ST6 ₂ (step ST
6 _3). When the file transfer is completed, the inference engine analyzes the contents of the transferred file and executes the rules. All the file transfer processes are performed by TCP / IP communication on the general-purpose LAN 20.

As described above, in the third embodiment, the rule execution computer 16 and the rule editing computers 17 to 19 are used.
Can be configured to have a one-to-many correspondence, and one rule file can be edited from any rule editing computer. In addition, the connection by network communication using the general-purpose LAN 20 enables the rule execution computer 16 and the rule editing computers 17 to 19 to be installed and operated at a distance from each other. In addition, since a general-purpose LAN and a general-purpose protocol are used, a LAN with other plant monitoring equipment is used.
Sharing and interconnection can be easily realized.

Embodiment 4 FIG. FIG. 7 shows an example of a rule file as an I / F of a rule editor and an inference engine according to the fourth embodiment of the present invention. The rules created by the rule editor are converted into a readable text file in the C language format shown in FIG. The rule in FIG. 7 corresponds to the meta-rule in FIG. 3 described in the second embodiment. The I / F between the rule editor and the inference engine is only the rule file, and is exchanged by file transfer on the LAN. The inference engine reads this text file and executes the rules. Since the I / F between the rule editor and the inference engine is limited only to the rule file, even when the rule editing computers 17 to 19 are out of order, the rule editor cannot be used for some other reason, or when the rule editing computer is not installed. If the computer on the side of the inference engine has only a text file editing function, simple rule editing is possible due to the readability of the rule file. As a result, the separation and independence of the advanced rule editing function and the rule execution function using the GUI can be obtained.

Embodiment 5 FIG. FIG. 8 shows an S / W configuration of a rule editor according to Embodiment 5 of the present invention.
The rule editing GUI (rule editor) 81 is an S / W built on the GUI development library 82, and its implementation depends only on the API of the GUI development library 82. And the GUI development library 82 is UNI
X-Window, a GUI system on X / PC
Both (83) and Win32 (84) are supported, and the difference is hidden. Therefore, the same GUI rule editor can be realized on both the EWS and the PC.

Embodiment 6 FIG. FIG. 9 shows an S / W configuration of an inference engine according to Embodiment 6 of the present invention. As shown in FIG. 9, the inference engine 91 is an S / W built on UNIX 92. Including the contents described in the fifth embodiment, the expert system of the present invention
This is an S / W constructed on a general-purpose OS such as IX / Windows, and does not assume a specific H / W. In addition, a general-purpose protocol (TCP / I
P). For this reason, the expert system of the present invention can have a multi-vendor, multi-platform configuration independent of a specific vendor and a specific platform. Further, integration with other plant systems such as a supervisory control system is also easy.

FIG. 10 is an integrated diagram of a multi-platform business processing / plant monitoring and control system. In FIG. 10, reference numeral 24 denotes an EWS for monitoring and control connected to a plant bus, and 25 denotes a form processing and other business processing. It is a computer PC for calculation. 26 is EWS for rule execution
And the inference engine operates. And 24, 25
Activate the rule editor of the expert system with both. In the EWS 26 for rule execution, one EWS
Since the monitoring control function and the expert function coexist on the WS, the operator monitors the operating state of the plant while
Edit, execute, and tune expert rules. Further, even if the PC of the computer 25 is installed in a management office in a building separate from the plant equipment, a word processor, a spreadsheet S
In parallel with the office work by / W, rule editing and execution can be performed remotely via the local LAN.

[0045]

As described above, according to the first aspect of the present invention, the rule construction parts are stuck on the rules by operating the mouse, so that the rules can be created visually and intuitively. This makes it possible to perform editing and to create rules efficiently. Also, by describing the four arithmetic operations in the meta-rules, the averages and aggregations that occur locally in each meta-rule can be absorbed in each rule, and the fuzzy rules can be compactly described in the rule table format. is there.

According to the second aspect of the present invention, the OS, GU
The rule editing computer, which is different from the I system, is configured with a GUI development library and rule editing software so as to absorb the difference between each OS and GUI system. There is an effect that a system configuration using a multi-platform such as EWS can be realized.

According to the third aspect of the invention, the function of the operation support system is divided into a rule editing computer and a rule execution computer, and the OS of the rule editing computer and the GUI system are loaded with a GUI development library and rule editing software. , The rule editing computer and the rule execution computer are connected to each other via a network, so that the rule editing computer and the rule execution computer are distributed, the same rule is shared from a plurality of rule editing computers, and rules are remotely controlled via a network. This has the effect of enabling editing.

According to the fourth aspect of the present invention, since the I / F between the rule editing computer and the rule execution computer is constituted by the rule file in the C language format, the operation of the inference engine alone can be achieved. .

[Brief description of the drawings]

FIG. 1 is a diagram showing a fuzzy rule definition screen in an expert system according to Embodiment 1 of the present invention.

[Fig. 2] If- defined in one fuzzy rule
It is a figure showing the example of the then sentence.

FIG. 3 is a diagram showing a meta-rule definition screen in the expert system according to the second embodiment of the present invention.

FIG. 4 is a diagram showing an operation example by Drag & Drop in FIG. 3;

FIG. 5 is a system configuration diagram in an expert system according to a third embodiment of the present invention.

FIG. 6 is a flowchart illustrating the operation of the system configuration of FIG. 5;

FIG. 7 is a diagram showing an example of a rule file according to a fourth embodiment of the present invention.

FIG. 8 is an S / W configuration diagram of a rule editor according to a fifth embodiment of the present invention.

FIG. 9 is an S / W configuration diagram of an inference engine according to Embodiment 6 of the present invention.

FIG. 10 is a diagram when a system configuration according to a sixth embodiment of the present invention is a multi-platform configuration.

FIG. 11 is a diagram showing a fuzzy rule definition screen of a conventional expert system.

FIG. 12 is a meta-rule definition screen of a conventional expert system.

FIG. 13 is a system configuration diagram of a conventional expert system.

FIG. 14 is a system configuration diagram of another conventional expert system.

[Explanation of symbols]

1 fuzzy rule definition screen, 11 meta-rule definition screen, 16 rule execution computer, 17, 18, 19 rule editing computer, 20 general-purpose LAN (network), 8
2 GUI development library.

Claims (4)

[Claims] 1. A fuzzy rule definition screen in a rule table format and a meta-rule definition screen in a rule table format displayed on a display surface with a GUI development library on an OS and a GUI system and rule editing software on the GUI development library. An expert system equipped with a rule editing computer that realizes each rule definition operation by mouse operation. 2. A rule editing computer having a different OS and GUI system, wherein a GUI development library and rule editing software are loaded so as to absorb the difference between the respective OSs and GUI systems. Expert system as described. 3. The function of the operation support system is divided into a rule editing computer and a rule execution computer. An OS of the rule editing computer and a GUI system are loaded with a GUI development library and rule editing software. 3. The expert system according to claim 1, wherein the rule execution computer is connected to a network. 4. The expert system according to claim 3, wherein the I / F between the rule editing computer and the rule execution computer is a C-language rule file.