JPH0242535A - Method and device for formation of knowledge base and trouble diagnostic method using the knowledge base - Google Patents

Abstract

PURPOSE:To simplify the formation of a fault tree by separating with hierarchy the fault tree into a duplex tree group and an original fault tree group and forming a knowledge base based on the fault tree. CONSTITUTION:An original fault tree group is produced on a CRT screen with use of an input/output device connected to a fault tree input device 1. Then the tree group is evolved into a hierarchical knowledge base 3 and an on-line mapping table 4 by a fault tree data analyzing device 2. A mapping process mechanism 7 secures the connection between the knowledge data base variable stored in the table 4 and the address of a process data table 8 to produce the addressing process data and supplies this data to an on-line inference mechanism 5. The mechanism 5 performs an inference process based on the addressing process data supplied from the mechanism 7 and the base 3. An inference result output device 6 shows the inference result to an observer in an optimum format.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention provides a knowledge base suitable for diagnosing plant failures, which is constructed by layering failure causes and events resulting in failures. - A method and apparatus for constructing a tree-based system and a method for diagnosing a fault based on the knowledge base.

In particular, when a group of trees constituting the fault tree overlaps within the fault tree, the group of trees is separated and made common, and the fault tree is separated and made common. tree group (hereinafter, f
Fi? (abbreviated as U-tree group) and other tree groups (abbreviated as U-tree group).
(hereinafter abbreviated as the original fault tree group) and hierarchization (
The present invention relates to a method and apparatus for constructing a knowledge base based on the separated/layered fault tree (hereinafter abbreviated as separation/layering) and a method for diagnosing a fault based on the knowledge base.

(Prior Art) As shown in Japanese Patent Application Laid-Open No. 62-75720, a conventional fault tree for diagnosing plant failures has a hierarchical structure consisting of events representing the cause of failure and events representing the results of the failure. are integrated, and even if there are overlapping pieces of knowledge, the overlapping pieces of knowledge have not been shared.

(Problem to be Solved by the Invention) In the above-mentioned conventional technology, even if there is a group of overlapping trees in the fault tree, each group of overlapping trees is not separated/hierarchized, so the fault tree becomes large. As the scale increases, the following problems arise.

1. Because the knowledge base expands freely into two-dimensional space, it becomes difficult to understand as it becomes larger.

2. Since duplicate tree groups are stored separately, the usage efficiency of the storage device as a hardware decreases and the device becomes larger.

3. Since overlapping tree groups are inferred separately, many inference processes are required to obtain the same inference result, which reduces the inference speed.

Regarding the fault tree construction system,
Knowledge construction tools using personal computers have recently been developed, but they are limited to constructing offline fault trees where processing speed is not an issue.

An object of the present invention is to solve the above-mentioned problems and provide a small-sized knowledge base construction device that can easily construct a large-scale fault tree and perform high-speed inference.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the present invention provides a knowledge base construction system that constructs a knowledge base based on fault trees, in which a group of trees constituted by a plurality of events is If there are duplicates in the fault tree, the fault tree is divided into a group of duplicate trees and a group of original fault trees by separating the group of duplicate trees from the fault tree. The feature is that a knowledge base is constructed based on the separated/layered fault tree.

Furthermore, when performing inference, the inference result obtained by executing the group of duplicate trees is copied to the event that constitutes the group of original fault trees and to which the group of duplicate trees is linked. The points are distinctive.

(Function) According to the above configuration, within the fault tree! 1!
It is possible to separate the existing tree group from the fault tree and make it common, and to create a structure in which the fault tree is further layered into a duplicate tree group and an original fault tree group.

Therefore, even when displaying a fault tree on a display means, a group of duplicate trees can be displayed in common, making it easy to understand the contents of the entire fault tree and simplifying the construction of the fault tree. be able to do it.

Furthermore, since redundant knowledge processing for a group of redundant trees can be avoided, high-speed inference is possible.

Furthermore, since redundant storage of duplicate tree groups can be avoided, storage capacity can be saved and the device can be downsized.

(Embodiment) FIG. 1 is a block diagram of a fault diagnosis system to which the fault tree construction system of the present invention is applied. This system uses a fault tree consisting of causes and effects as a representation of the knowledge base.

A knowledge base using a fault tree is input using a fault tree input device 1 consisting of input/output devices such as a CRT display, a keyboard, and a mouse. The fault tree input device 1 is capable of creating the base of a fault tree on the CR7 screen.

The fault tree created by the fault tree input device 1 is expanded into a hierarchical knowledge base and an online mapping table by the fault tree data analysis device 2.

The hierarchical knowledge base 3 consists of meta-rules for controlling rules, a group of rules for controlling inference, and a group of frames for storing and controlling intermediate assumptions, inference results, and the like.

These three types of knowledge databases are highly independent (object-oriented) and are used to reach a final conclusion by communicating messages with each other.

The mapping processing mechanism 7 generates addressing process data by associating addresses between the knowledge database variables stored in the online mapping table 4 and the process data table 8, and turns this on and off. - Supplied to the line inference mechanism 5.

In other words, the target plant usually has a large number of observation points, and these observation points a>+ are distinguished by a predetermined code. However, since the code in the system indicating the observation point required by the system to perform inference and the observation point code do not necessarily match,
In order to supply plant data to the online reasoning mechanism 5, it is necessary to link these codes. The mapping processing mechanism 7 performs this connection.

The mapping processing mechanism 7 is disclosed in Japanese Patent Application No. 1983.
-22695 describes in more detail.

The online inference mechanism 5 executes inference processing using a known appropriate method based on the addressing reference process data supplied from the mapping processing mechanism 7 and the hierarchical knowledge base 3.

The inference result output device 6 uses an output device (not shown) such as a CRT or a printer to present the inference result in an optimal format to a viewer a such as a driver a.

FIG. 2 shows the structure of a general fault tree, which has a hierarchical structure in which causes and effects of failures are layered.

For example, the event “low pressure casing thermal deformation” occurs from the event “condenser vacuum abnormality, gland steam pressure abnormality” or “low pressure exhaust chamber temperature high”, and the event “rotor alignment failure” occurs from the event “high pressure casing thermal deformation”. This indicates that this occurs from "thermal deformation of the low-pressure casing" or "thermal deformation of the low-pressure casing."

Here, the higher-level event "Gland steam pressure abnormality" whose lower-order events are "Low control oil pressure" and "Gland steam regulator abnormality" is the event "High-pressure casing thermal deformation" as shown in indications A and H in the figure. ” and “low pressure casing thermal deformation”
It is the cause (sub-event). T like this! IV
In general, the larger the scale of the knowledge base, the higher the proportion of the fragmented knowledge in the knowledge base, causing the above-mentioned problems.

Therefore, in the present invention, in order to solve these problems,
A fault tree having overlapping fragmentary knowledge (A and B in the same figure) as shown in FIG. 2 is converted into a common fault tree as shown in FIG. At the same time, the molecules are lit/hierarchized into a group of upper trees, which are the original fault trees, and a group of lower trees, which are shared as a group of overlapping trees, and a knowledge base is constructed based on the separated/hierarchical fault trees. do.

In addition, the number (1) attached after each event name in Figure 3
) to (12) are identification codes to distinguish between events with the same name in the same tree group. For example, two "grand steam pressure anomalies" are assigned different numbers. ((i), by assigning (8), the two can be reliably distinguished.

Also, minute i! When linking a group of it/hierarchical trees to a predetermined event of the original fault tree group, the degree of contribution to abnormal spread generally differs depending on the event (connection point) targeted for linking.

For example, in Fig. 3, the degree of influence of the abnormality inferred from the accident "Low control oil pressure (11)" or "Gland steam regulator abnormality (12)" to the higher level event is "High pressure casing thermal deformation (4)" The case where it contributes to “grand steam pressure abnormality (6)” which is a subordinate event of “low pressure casing thermal deformation (5)” is different from the case where it contributes to “grand steam pressure abnormality (6)” which is a subordinate event of “low pressure casing thermal deformation (5)” There are many things.

Therefore, in this system, it is possible to define an anomaly degree transfer coefficient β, which will be explained in detail later with reference to FIG. 8, as a coefficient indicating the degree of spread of an abnormality in a lower tree group to a connection point in an upper tree group. As a result, in this system, even when the original fault tree group existing in the same space is divided into #i/hierarchical layers and then linked and inferred, inference with high probability is possible.

Below (using the fault diagnosis system shown in Fig. 1),
The method for constructing the tree will be explained along the flowchart of FIG.

In step S1, an original fault tree group (upper tree group) shown in FIG. 3 is created on the CR1 screen using input/output devices such as a keyboard and a mouse connected to the fault tree input device 1.

At this time, the events (connection points) which are the lowest event of the upper tree group and linked with the lower event which is the duplicate tree group, namely "Grand Steam Pressure Abnormality (6)" and "1 Grand Steam Pressure Abnormality (8)" The events described below are treated as dummy events without creating a frame, which will be explained later with reference to FIG. 8, and for other events, frames for each event are also created.

In step S2, the duplicate tree group (lower tree group) shown in FIG. 3 is displayed on the CRT screen in the same way as step S1.
, and also create frames for each event.

In step S3, the fault tree data analysis device 2 develops the fault tree into meta rules, rules, and frames.

The method for expanding the fault tree will be described in detail below.

Meta rules are those that control a group of rules.
This is knowledge for determining which rule group is to be the subject of inference. FIG. 6 is a diagram showing the contents of the metal rule file when the separated/hierarchical fault tree shown in FIG. 3 is expanded, and the inference operation is activated by a command from 5TART.

The execution part of the meta rule (below THEN) is composed of a rule group and a priority.

In FIG. 6, "on a rule group" refers to a rule group to be activated, and the order in which these rule groups are activated is determined by priority. In addition, priority indicates that the higher the number, the higher the level. In the example in Figure 6, "Ground Steam Pressure Abnormality" on the rule group is activated first, and then "Turbine" on the rule group is activated. This indicates that "large vibration" is activated.

Here, activation specifically means executing a rule group registered on the indicated rule group.

These rule groups correspond to the separation/hierarchical fault tree shown in FIG. 3, and are determined by the separation method. Furthermore, the priority is determined by the level of the hierarchical structure. Then, by configuring in this way, the minute i! Inference with It/layered fault trees is performed from a group of lower trees.

Next, each separated/hierarchical tree group is
A method of developing a rule group as shown in the figure will be explained.

For example, as shown in Figure 3, separation/layering faults and
In the rule group, "large turbine vibration" is composed of a plurality of subordinate events such as "rotor alignment failure" and "high pressure casing thermal deformation."

In addition, in FIG. 7, in order to simplify the explanation, "large turbine vibration" on the rule group includes rule 11 rule 2, 2
Only one rule is listed, and other rules are omitted.

Here, the rule number file shown in Figure 7 is a logical system that performs the following operations at the time of inference.■ Search for the unexpanded lowest level event, and if there are multiple lowest level events, , select an arbitrary event.

(2) Check whether or not a frame whose frame name is the selected lowest event name and identification code can be executed, and if it has not been executed, activate the frame J to make it executable.

■The lowest event selected this time is marked as expanded.

■Investigate whether there are any undeveloped events, and if there are, return to ■,
If there is no rule, rule expansion ends.

Here, before executing the rule group developed in this way,
The frame group name that is the scope and target of the rule must be activated as described below with respect to FIG. Therefore, the activation frame group is defined before the rule group.

To summarize the above, one fault tree has the 7th
As shown in the figure, it is expanded to a rule file with the following structure.

(a) Above the rule group, above the Nitry group (b) Above the activation frame group, above the Nitry group (c) Events within the rule minitory group In Fig. 8, convert the separation/layering fault tree in Fig. 3 to a frame file. An expanded example is shown.

A frame file is composed of a frame group name representing each tree group name and a frame name representing an event existing within the tree group. Each frame is divided into an input slot Y, which stores the degree of abnormality of an event estimated from a group of lower-level events through combine function processing, an abnormality transfer coefficient slot, β, which stores the degree of propagation of an abnormality to a higher-order event, and an old calculation based on the process amount. The abnormality inference result slot α is determined by the maximum value of the abnormality degree and the input slot Y.

For example, the event "control oil pressure paper" in the lower tree group is expanded to the frame "control oil pressure paper (11)" as the first frame (frame 1) belonging to "ground steam pressure abnormality" on the frame group. Here, "(11)" is the number given to the "control oil standard" among the year codes determined within the lower tree group.

Furthermore, the event “Grand Steam Regulator Abnormality” in the lower tree group is treated as “Grand Steam Regulator Abnormality (12)” as the second frame (frame 2) belonging to “Grand Steam Pressure Abnormality” on the frame group. expanded into a frame.

In addition, the topmost event "grand steam pressure abnormality" in the lower tree group is the third frame (frame 3) of the frames belonging to "grand steam pressure abnormality" on the frame group, and the frame "grand steam pressure abnormality (lO)" will be expanded to.

Here, Frame 1 “Control Oil Pressure Paper (11)” and Frame 2 “Ground Steam, M Regulator Abnormality (12)” are the lowest level events, and since there is no transmission of the degree of abnormality from the lower level events,
Only combine function processing is defined.

In step S4, processing means related to slot copy processing, which is a feature of the present invention, is added to each frame of the topmost event of the duplicate tree group.

Slot copy processing is when performing fault diagnosis (inference) using the knowledge base constructed by this system, the inference result obtained by executing the group of duplicate trees is copied to the source to which the group of duplicate trees is connected. This refers to the process of copying the lowest event in the fault tree.

For example, in FIG. 8, the “ground steam pressure abnormality (lO
) “ is the upper tree group, the original fault tree “
Since this is an event linked to "Grand Steam Pressure Abnormality (6)'" and "Grand Steam Pressure Abnormality (8)", two slots are created to link the abnormality degree of "Grand Steam Pressure Abnormality (10)" to these events. Copy processing has been added.

Note that the inference method using the slot copying process added to the frame in this way will be explained in detail later.

In step S5, the meta rules, rules and frames developed in this way are stored in the hierarchical knowledge base 3 as a knowledge base.

In addition, in FIG. 8, in order to simplify the explanation, only two frames are shown in "large turbine vibration" on the frame group, and the other frames are omitted.

In this way, in the present invention, even if there is a group of heavily protected trees in a fault tree, only one of the duplicate tree groups is created and shared, so the storage capacity of the storage device can be reduced. can do.

Furthermore, according to the present invention, even when inference is performed, inference is performed only once for a group of overlapping trees, and the inference results are copied to other events that are connection points of overlapping trees li by slot copy processing. Therefore, the time required for inference can be shortened.

Next, we will show two examples of how to display on a CRT when confirming and creating connection relationships of separated/hierarchical fault trees on a display device, or in an actual system. It's not limited.

FIGS. 4(a) and 4(b) respectively show fault trees that are actually separated/hierarchized and displayed on the display device, and FIG. 4(a) shows the fault tree shown in FIG. The original fault tree group (upper tree group) when separated and hierarchized, and FIG.

Events that serve as connection points for linking each tree group (in this example, "ground steam pressure abnormality") are marked with different colors and other display methods, and the connection relationship between both tree groups is marked. It has been revealed. For example, "Ground Steam Pressure Abnormality", which is a higher level event in the duplicate tree group, is connected to two "Ground Steam Pressure Abnormal" events in the original fault tree group, so the display color (hatched in the figure) (presence or absence) is displayed as different from other events in the duplicate tree group.

Similarly, among the events in the original fault tree group, the event "Ground Steam Pressure Abnormality", which is the connection point with the duplicate tree group, is displayed in a different color from the others, making it easy for the operator to understand. That's how it is.

Note that the above-mentioned discrimination from other events is not limited to discrimination based on color, but may also be based on a difference in brightness, the presence or absence of blinking, or the like.

FIG. 5 shows the relationship between the separated/hierarchical tree groups shown in FIG. 3 in a table format on a display device. In this way, in this embodiment, by displaying the relationship between a certain tree group and the upper tree group to which it is connected in a table format on the display device, it is possible to clearly indicate the connection relationship between the tree groups. .

For example, in Fig. 5, the right column of the yli elephant "Grand Steam Pressure Abnormality" displayed in the upper column of the lower tree group shows:
The events “High Pressure Casing Thermal Deformation” and “Low Pressure Casing Thermal Deformation” are displayed as the four corresponding upper trees, and the above event “Ground Steam Pressure Abnormality” has the upper tree that is the target of the link. It turns out that there are two.

As explained in FIGS. 4 and 5, the connection relationships of separated/hierarchical fault trees are defined on the display device.
Because confirmation is possible, faults with duplicate trees
The creation of large tree-based knowledge bases is possible on the display device.

Next, specific inference processing according to the present invention will be explained.

In Fig. 3, the lowest event in the lower tree group is “control oil hole (ll)-” and “grand steam regulator abnormality (12
)” means that there is no elephant in the lower order!1 and the abnormality degree inferred from this is 0, so the control oil holes (ll)/Y and the ground steam regulator abnormality (12)/Y are both 0. become(
The expression control oil hole (11)/Y is the frame “
means slot Y of the control oil hole (11).

In addition, the abnormality degree transfer coefficient β in the lower tree group is defined at the time of creating the knowledge base, and in this example, all the abnormality degree transfer coefficients β are set to 80% (one's own abnormality is 100%).
If it is assumed that 80% of the degree of influence on the abnormality level of the upper level event is defined. Furthermore, it is assumed that the input slot α that stores the degree of abnormality of an event determined by the process amount stores the calculation result of the definition formula as shown below prior to inference processing.

Control oil hole (11)/α-0,3 Gland steam regulator abnormality (12)/α-0.5 Gland steam pressure abnormality (10)/α-0,1 Here, according to the rules, the four frame When frame 1 of “vapor pressure abnormality (10)” is activated, frame “control oil hole (1
1) “operates as follows.

16 Determine the degree of abnormality of the self-event: Substitute the maximum value (MAX) of control oil hole (11)/α and control oil hole (II)/Y to control oil hole (11)/α. That is, from Y-0, α-0, 3, α-MAX
(0,0,3) -0,32, Calculate the degree of transmission abnormality A to the upper level event: control Uraso paper (11)/α and control oil hole (I
I)/β. (From α-0,3, β-0,8,
A-0, -24) 3. Inference abnormality degree calculation of “grand steam pressure abnormality (10)” by combine function processing: Transfer abnormality degree A, upper event “grand steam pressure abnormality (lO
)” abnormal inference result Y (ground steam pressure abnormality (to)
/Y), the inference abnormality degree is calculated using the following combine function.

Combine function CF (x 1.X 2 ) −1(I
Xt) X(I

x2) -1-(1-A) x (1-grand steam 2-pressure car °
Hetsu (10)/Y)-1-(1-0,24)X(1-0
)-0,24 When frame 2 is activated according to the following rule, the frame "Grand Steam Regulator Abnormality (12)" similarly operates as follows.

1. Determine the degree of abnormality of the own event Ni Grand Steam Regulator Abnormality (12)/α-MAX (Grand Steam Regulator Abnormality (12)/α, Grand Steam Regulator Abnormality (12)/Y) -MAX (0,5 , O) - 〇, 5 2. Calculate the transmission abnormality degree A to the upper level event: A - (Ground steam: AfiW difference 5 (12) / a ) X (Grand steam regulator abnormality (12) / β) - 0 ,5XO,8-0,
4 3. Inference abnormality degree calculation by combine function processing Nigerland steam pressure abnormality (1G)/Y-1-(1-A)×(1-
Gland steam pressure abnormality (10)/Y) -1-(1-0
, 4)

1. Determine the degree of abnormality of the self-event Nigrand steam pressure abnormality (In)/α-MAX
10)/Y) -MAX (0,1,0゜544)-0
,544 2. Calculate the degree of transmission abnormality A to the higher-level event: Combine function processing is not executed because it is the highest-level event.

3. Slot copying process: When the abnormality degree of the lower tree group is calculated in this way,
This value is copied to "grand steam pressure abnormality (6)" and "grand steam pressure abnormality (8)" by the slot copy process.

Here, the degree of influence of this degree of abnormality on the higher level event is 0.8 when it contributes to the “grand steam pressure abnormality (8)” which is a lower level event of “high pressure casing thermal deformation (4)”. 0 if it contributes to "Gland Steam Pressure Abnormality (6)" which is a subordinate event of "Low Pressure Casing Thermal Deformation (5)".
5, the abnormality inference result slot α for each event is as follows.

(a) Large turbine vibration/abnormal gland steam pressure (6)/
α-Gland steam pressure abnormality/Gland steam pressure abnormality (1
G)/α×β(-0,8) -0,544X0. 8"
0. 44 (b) Large turbine vibration/Gland steam pressure abnormality (8)/
α-Gland steam pressure abnormality/Gland steam pressure abnormality (1
0)/α×β(-0,5) -0,544×0.5 out of 0
．． 27 Here, the process that actually implements the lit/layered fault tree is the slot copy process.

As described above, according to the present invention, even when inference is performed, inference for a group of duplicate trees is performed only once, and the inference results are copied by slot copy processing for other events that are connection points of a group of duplicate trees. Therefore, the time required for inference can be shortened.

Next, the overall flow of this inference process will be explained using FIG. 9. Note that Fl, F2, etc. shown in the frame group correspond to frame 1, frame 2,...Φ, respectively, explained using FIG. F3 in "large vibration" corresponds to the event "Condenser vacuum abnormality (7)", and F4 corresponds to the event "Gland steam pressure abnormality (8)".

In addition, F5 corresponds to "low pressure exhaust chamber temperature high (9)", F6 corresponds to "low pressure casing thermal deformation (5)", and F7 corresponds to "high pressure casing thermal deformation (4)".

In the figure, meta-rules control the entire group, and the meta-rules activate rule groups one after another. The activated rule group activates a frame group representing the entity.

The activated frame Fi performs data communication between frames to continue inference.

In this embodiment, when the frame group "Ground Steam Pressure Abnormality" is activated, Fl and F2 are activated, and then F3 is activated.Furthermore, information on F3 is transmitted through data communication to the frame group " Fl, F4 with “large turbine vibration”
After that, each frame is sequentially activated in the same way.

In this way, according to this embodiment, minutes! /Inference is realized by expanding the hierarchical fault tree into meta-φ rules, rule groups, and frame groups.

(Effects of the Invention) As is clear from the above description, according to the present invention, fault trees can be constructed, displayed, and stored in a hierarchical manner, so that the following effects can be achieved.

(1) Even in a large or complex fault tree, overlapping events can be displayed in a hierarchical manner, making it easy to understand the overall content and making fault tree construction simple and easy. .

(2) High-speed inference is possible because redundant knowledge processing for redundant events in the fault tree can be avoided.

(3) Since redundant storage of duplicate events in the fault tree can be avoided, storage capacity can be saved and the device can be made smaller.

(4) When linking the duplicate tree group to the original fault tree group, the degree of abnormality spread is determined based on the abnormality degree transfer coefficient, so inference with high probability is possible.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a fault diagnosis system to which the present invention is applied. FIG. 2 is a diagram showing a non-layered fault tree. FIG. 3 is a diagram showing a layered fault tree. FIG. 4 is an example of a method for representing a layered fault tree. FIG. 5 shows another embodiment of the method of representing a layered fault tree. FIG. 6 is a diagram showing the contents of the meta rule. FIG. 7 is a diagram showing the contents of the rules. FIG. 8 is a diagram showing the contents of the frame. FIG. 9 is a diagram showing an outline of the inference processing. FIG. 10 is a flowchart showing a method of constructing a knowledge base. DESCRIPTION OF SYMBOLS 1... Fault tree input device, 2... Fault tree data analysis device, 3... Hierarchical knowledge base, 4... Online φ mapping table, 5
...Online inference mechanism, 6.Inference result output device, 7.Mapping processing mechanism

Claims (8)

[Claims] (1) In a knowledge base construction system that creates a fault tree consisting of events that cause a system failure and events that result in a failure, and constructs a knowledge base based on the fault tree, means for inputting a tree; display means for displaying a fault tree; means for separating a group of trees that exist redundantly in the fault tree from the fault tree and making them common; and the group of duplicate trees. means for creating a separate/hierarchical fault tree by hierarchizing the original fault tree group that is the other tree group; a first storage means for storing the duplicate tree group; and the original fault tree group A knowledge base construction device comprising: second storage means for storing. (2) The overlapping tree group is configured to copy the inference result obtained by executing the overlapping tree group to an event that constitutes the original fault tree group and to which the overlapping tree group is linked. The knowledge base construction device according to claim 1, further comprising means. (3) Each of the events constituting the separated/hierarchical fault tree is set with an anomaly degree transmission coefficient that indicates the degree to which the anomaly spreads to higher-level events. The knowledge base construction device according to item 1 or 2. (4) A claim characterized in that, among the events constituting the separated/hierarchical fault tree, events having at least a group of overlapping trees as subordinate events are given different identification codes. The knowledge base construction device according to any one of Items 1 to 3. (5) The display means distinguishes at least one of the topmost event of the group of duplicate trees and an event forming the original fault tree group to which the group of duplicate trees is linked, from other events. The knowledge base construction device according to any one of claims 1 to 4, characterized in that the knowledge base construction device displays the information using a display method according to the present invention. (6) The separated/layered fault tree is
Claim 1 further comprising means for expanding into rules, rule groups, and frame groups.
6. The knowledge base construction device according to any one of items 5 to 5. (7) A knowledge base construction method in which a fault tree is created by linking events that cause a system failure and events that result in a failure, and a knowledge base is constructed based on the fault tree. inputting a tree, separating a group of trees that exist redundantly in the fault tree from the fault tree and making them common, and forming a hierarchy between the group of duplicate trees and the original fault tree group that is a group of other trees. to create a separated/hierarchical fault tree, and the group of duplicate trees is configured to combine the inference results obtained by executing the group of duplicate trees with the events constituting the group of original fault trees and the group of duplicate trees. A knowledge base construction method characterized by copying information to linked events. (8) In a knowledge-based fault diagnosis method using a fault tree configured by linking events that cause a system failure and events that result in a failure, A fault diagnosis method using a knowledge base, characterized in that the inference result obtained is copied to an event that constitutes an original fault tree group that is a tree group other than a duplicate tree group, and to which the duplicate tree group is linked. .