JPH03251924A - Knowledge editing/displaying device for expert system - Google Patents

Abstract

PURPOSE:To easily construct the knowledge by displaying the logic relation of knowledge in a logic diagram and editing automatically the knowledge when the knowledge is constructed in an expert system. CONSTITUTION:An interface 700 performs the output of the display information and the input of the manual selection information. Then the interface 700 stores the display information in a display data table 706, the logic symbols in a logic symbol store table 704, the rules in a rule pattern table 705 for each logic symbol, the item names and the coordinates in an item name/coordinate table 709, the assurance under diagnosis in an interim result store table 708, and the information for reference to the table 708 in an interim result store table map 707 respectively. Based on these information, a logic diagram display means 703 displays the causal relation of the knowledge for study of the factors of the plant abnormality in a logic diagram. This diagram is inputted to a screen of a display device 300 in an image as it is for construction of a knoledge base 102. When the plant information is inputted, an inference engine 101 interprets the knowledge based on the base 102 and executes an inference. A QA processing means 104 responds to an operator 400. In such a constitution, the knowledge is easily constructed.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention is a method for editing diagnostic knowledge and diagnosis progress for diagnosing the causes of plant abnormalities using an expert system, which is a field of knowledge engineering. This invention relates to a knowledge editing/display device in an expert system that displays information.

(Conventional technology) For example, in order to operate a thermal power plant with high reliability, advanced monitoring capabilities and abnormality diagnosis functions are required. We provide the operational know-how of experienced power generators and specialized knowledge regarding plant characteristics to the computer used in the project.
It is designed to make inferences and judgments about the plant.

The general configuration of the expert system is shown in FIG.

In FIG. 10, diagnostic knowledge is filed in a knowledge base 102. A knowledge base is a collection of fragmented knowledge, or rules, and each rule is if ~, then
It is written in the file as a text image in the form of ~. Further, inference is executed by the inference engine 101 interpreting the rules in the knowledge base 102.

Editing the knowledge base in conventional expert systems is done through the knowledge editing function, and in practice, if~,・t
hen ~ rules in a normal programming editor,
For example, it is edited as a text image using a screen editor.

Editing of knowledge is performed by modifying and adding rules as text images as described above, but in reality, it is necessary to check for mistakes in rule descriptions, inconsistencies in knowledge, etc.

In conventional expert systems, an inference engine actually temporarily executes inference, and a check is performed using error messages issued by the inference engine.

That is, editing a knowledge base in a conventional expert system is performed by first performing coding similar to normal programming and then depacking.

Therefore, conventionally, in order to edit a knowledge base, it is necessary to have both knowledge necessary for diagnosis and knowledge for depacking the knowledge base. On the other hand, as a method of expressing knowledge of plant diagnosis, there is a logic diagram that expresses knowledge in a sequence, as shown in Figure 2.

An example of a logic diagram of diagnostic knowledge is shown.

In FIG. 2, each block shows a phenomenon of a plant including a degree of certainty. Confidence is a numerical value that quantifies how confident a human (expert to be diagnosed) is about the corresponding phenomenon, and its range is, for example, O
It is expressed as a numerical value between O and 1, where is a complete negation and l is a complete affirmation.

In addition, the logical relationships in the diagram correspond to conventional rules. For example, in Figure 2, from the right, "The confidence level of cause 1 is the AND of the confidence level of intermediate phenomenon 1 and the confidence level of question item 3. The rule ``determined by the relationship'' and the rule ``the confidence level of intermediate phenomenon 1 is determined by the OR relationship between the confidence level of question item 1 and the confidence level of question item 2'' are described.

In general, there are many definitions for calculating confidence in AND and OR relationships, but here, AND is +min.
, OR are defined as wax operations.

In conventional expert systems, such rules are coded as text images, and even when knowledge is expressed in logic diagrams, text images are used to input rules.

(Problem to be solved by the invention) As described above, knowledge bases in expert systems have traditionally been coded as text images, so knowledge construction and editing are side-by-side, leading to errors in editing and time-consuming research. There is a problem that it takes

Furthermore, the progress of diagnosis at the time of depacking requires visual tracking of the coding list in the knowledge base, which is difficult to understand intuitively.

The present invention was made in consideration of the above problems, and is a rational expert system in which knowledge rules are constructed and edited as a logic diagram using graphic images, and the progress of inference can also be seen from the logic diagram. The purpose is to provide a knowledge editing and display device.

[Structure of the invention]

(Means and effects for solving the problems) The present invention includes a data input section for inputting plant data from a plant, a display device for displaying data, a knowledge base for storing knowledge for diagnosis, A QA processing means that requests data other than plant data through a display device when making a diagnosis, and an inference engine that interprets knowledge from a knowledge base and performs a diagnosis using data from the data input section and the QA processing means. A knowledge editing display device for an expert system comprising: a man-machine interface means for outputting display data to the display device and inputting manually selected data from the display device; and a display device for storing data displayed on the display device. A data table, a logical symbol storage table that stores logical symbols for display, a rule pattern table that stores rule templates for each logical symbol, an item name coordinate table that stores item names and their coordinates, and confidence during diagnosis. an intermediate result storage table that stores the degree, an intermediate result storage table map that stores data for referring to the intermediate result storage table, item names input via the man-machine interface, their coordinates on the screen, Create a diagnostic rule based on manually selected data such as logical relationships and the rule template from the rule pattern table above, save it in the knowledge base above, and save the item name and its coordinates in the item name coordinate table. , furthermore, a graphic/rule conversion means for storing display data in a display data table and further creating an intermediate result storage table for referring to the intermediate result storage table, the logical symbol storage table, the display data table, A logic diagram display means is provided for displaying a logic diagram on a display device via the man-machine interface means by referring to an intermediate result storage table and an intermediate result storage table map, and is equipped with a logic diagram display means for displaying a logic diagram on a display device through the above-mentioned man-machine interface means. The relationship is displayed in a logic diagram, and the knowledge base can be constructed by inputting the image as is on the screen of the display device.

(Example) An embodiment of the present invention is shown in FIG.

FIG. 1 shows the configuration of a knowledge base editing and display device system in an expert system according to an embodiment of the present invention, in which the expert system is used for plant diagnosis. The cause of plant abnormality is determined based on this, and in the process of inference, missing information is determined in advance.
The system memorizes the necessary information and continues inference by asking the operator questions and receiving answers.

That is, in FIG. 1, there is a data input unit 50o that takes in plant data from the plant 600, an inference engine 1o1 that interprets knowledge and executes inference, a knowledge base 102 for inference, and information that is missing when executing inference. It is comprised of a QA processing means 103 that notifies the operator 400 and receives answers (data) from the operator 400, and a display device [300] that displays the information.

In addition, the knowledge base editing part includes a logical symbol storage table 7 that stores graphic images of logical symbols representing knowledge.
04, a man-machine interface means 701 for inputting and displaying knowledge in the form of a logic diagram image, a graphic/rule conversion means 702 for generating rules from logic diagram information (item names, logic symbols), and a system necessary for generating rules. It consists of a rule pattern table 705 that stores rule templates.

Furthermore, the part that displays knowledge includes a knowledge result storage table 708 that stores intermediate results of inference from the inference engine;
The intermediate result storage table map 707 is that map.
, a display data table 706 that stores data for display, and an item name coordinate tape 709.
It consists of a logic diagram display means 703 that displays a logic diagram using these tables.

An example of screen operations for editing knowledge using the configuration shown in FIG. 1 will be described with reference to FIG. 3.

In Figure 3, functions are selected by clicking the mouse. First, enter the name of the question item to investigate the plant status required for diagnosis in the desired left column (multiple).
, select a question item that has a causal relationship, and click on t=+, tb, tb, -, which represents the depth of inference (number of steps), and the result is AND.

A logical relationship specification screen 1000 such as OR is opened.

Select a logical symbol connecting the selected question items (AND is selected in FIG. 3).

Also, the name of the intermediate phenomenon caused by this logical symbol,
Alternatively, by inputting the name of the cause item, a logic diagram representing one causal relationship is completed on the screen, and knowledge is thereby inputted on the screen.

Next, we will explain how to create a knowledge base for the above operation side.

First, through the man-machine interface means 701, the name logical symbols of the question items and intermediate phenomena are converted to the graph ink/rule conversion means 70 along with the corresponding coordinates on the screen.
Send to 2. Furthermore, in order to display the logical symbol on the screen, a graphic image of the AND symbol is extracted from the logical symbol storage table 704 and displayed on the screen.

FIG. 4 shows an example of the format of the logical symbol storage table 704.

The graphic/rule conversion means 702 creates a corresponding text image rule from the input knowledge, ie, intermediate phenomenon name, question item, and logical symbol. The steps to create it are as follows.

That is, a rule template is stored in advance in the rule pattern table 705 for each logical symbol. FIG. 5 shows an example of AND-related rules.

In FIG. 5, intermediate phenomenon names and question item names are entered in the higher order items or lower order items 1 to n on the right side of each item, respectively.

If the logical relationship is AND as in the case of FIG. 3, the graphic/rule converting means 702 brings the AND template, replaces the name of the intermediate phenomenon with the upper item, and replaces the lower item with the question item name. If you enter the desired rule, you will get the desired rule.

The reason why you can tell that the upper item is the intermediate phenomenon and the lower item is the question name is because
This is from the coordinates sent from the man-machine interface 701. In other words, the items on the left side of the logical relationship are the subordinate items.
The items on the right side are determined to be the top items. The number of subordinate items (n in FIG. 5) can be obtained by counting the coordinates of the name sent from the man-machine interface means 701 to the left of the logical relationship.

You only need to have as many rule templates as there are sub-items, and rules created in this way can be used as knowledge base 1.
Saved in 02.

Examples of the formats of the display data table 706, item name coordinate table 709, intermediate result storage table map 707, and intermediate result storage table 708 created by the graphic/rule conversion means 702 are shown in FIGS. 6 and 7, respectively. , 8 and 9.

The item name coordinate data table 709 and the display data table 706 are tables in which the logic diagram display means 703 stores data for displaying the logic diagram.
The intermediate result storage table map 707 is a map that indicates where in the intermediate result storage table 708 a desired degree of certainty is located when the logic diagram display means 703 displays the degree of certainty for each item.

Logic diagram display means 703 displays data table 7 for display.
The item names and logical relationships are taken out from 06, and the graphic images corresponding to the logical relationships are taken out from the logical symbol storage table 704, and these are sent to the man-machine interface means 701 to display a logic diagram on the display device 300.

Also, with reference to the intermediate result storage table map 707,
A desired degree of certainty is retrieved from the intermediate result storage table 708, and the degree of certainty is displayed at a desired position, that is, at a coordinate position corresponding to the name, by information such as color change. Here, the coordinate position can be determined by referring to the item name coordinate data table 709.

〔Effect of the invention〕

As explained above, according to the present invention, when constructing knowledge in an expert system, the logical relationships of knowledge are graphically displayed using a logic diagram, and knowledge can be edited automatically by changing the graphic image. Knowledge can be easily constructed without using troublesome procedures such as modifying the information using an editor.

Furthermore, when making a diagnosis using the logic engine, the logical relationships of the knowledge used for inference are similarly displayed graphically, making it easy to read the progress of the diagnosis.

Furthermore, since the probability of occurrence of the knowledge used in the inference is calculated and displayed in different colors depending on its size, it is easy to determine which knowledge was used in the diagnosis.

[Brief explanation of drawings]

FIG. 1 is a system diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of diagnostic knowledge in the present invention, FIG. 3 is a diagram showing an example of a display screen in the present invention, and FIG. 4 is a diagram showing an example of the diagnostic knowledge in the present invention. Figure 5 showing the format of the logical symbol storage table in the invention
The figure shows an example of the AND-related rule template of diagnostic knowledge in the present invention, Figure 6 shows the format of the display database in the present invention, and Figure 7 shows the format of the item name coordinate data table in the present invention. FIG. 8 is a diagram showing the format of the intermediate result storage table map in the present invention, and FIG. 9 is a diagram showing the format of the intermediate result storage table in the present invention.
FIG. 10 is a system diagram showing the general configuration of a conventional expert system. 100... Expert system 101... Inference engine 102... Knowledge base 103... Knowledge, editing means (screen editor)
104...QA processing means 300...Display device 400...Operator 500...Data input section 600...Plant 700...Knowledge base editing/display device 701...Man-machine interface means 702...・Graphic/rule conversion means 703 ・Logic diagram display means 704 ・Logic symbol storage table 705 ・Rule pattern table 706 ・Display data table 707 ・Intermediate result storage table map 708 ・Intermediate result storage table 709 ...Item name Zacho table (8733) Agent Patent attorney Yoshiaki Inomata (and 1 others)
name) 00 Figure 1 Figure 2 Figure 06 (Figure Figure Figure Figure Figure Figure

Claims (1)

[Claims] A data input section for inputting plant data from the plant, a display device for displaying data, a knowledge base for storing knowledge for diagnosis, and a display device for requesting data other than plant data when diagnosing. In a knowledge editing/display device in an expert system comprising a QA processing means and an inference engine that interprets knowledge of a knowledge base and performs diagnosis using data from the data input section and the QA processing means, the display device includes: A man-machine interface means for outputting display data and inputting manually selected data from the display device, a display data table for storing data to be displayed on the display device, and logic for storing logical symbols for display. A symbol storage table, a rule pattern table that stores rule templates for each logic symbol, an item name coordinate table that stores item names and their coordinates, an intermediate result storage table that stores confidence levels during diagnosis, and the above intermediate results. An intermediate result save table map that saves data for referring to the saved table, manually selected data such as item names input through the man-machine interface, their coordinates on the screen, logical relationships, etc., and the rule pattern table above. Create a diagnostic rule based on the rule template from and save it in the knowledge base, save the item name and its coordinates in the item name coordinate table, save the display data in the display data table, Further, the graphic/rule conversion means for creating the intermediate result storage table map, the logical symbol storage table, the display data table, the intermediate result storage table, and the intermediate result storage table map are referred to, and the man-machine interface is converted to the means. 1. A knowledge editing/display device for an expert system, comprising a logic diagram display means for displaying a logic diagram on a display device via a logic diagram.