JP5146750B2 - FT diagram creation program, FT diagram creation device, recording medium, and FT diagram creation method - Google Patents

Description

  The present invention relates to an FT diagram creation program, an FT diagram creation device, a recording medium, and an FT diagram creation method for creating an FT diagram based on trouble information such as an accident or failure occurring in a product.

  Conventionally, fault tree analysis (FTA / Fault Tree Analysis) is known as a technique for analyzing what kind of failure is possible in a system for realizing a certain process. A fault tree (hereinafter referred to as “FT” for short) is used in this FTA. What is an FT diagram as a cause of a failure in a system or process to be developed or evaluated? It is a figure showing whether an event is considered as data of a tree structure.

  As a method for creating this FT diagram, there is a conventional manual method. For example, there is a technique of creating an FT diagram in advance at the time of design or creating it based on raw data at the site. However, since the former is based on a designer's prediction, there is a problem that the contents are abstract and difficult to apply to the site. The latter has a problem that only partial information remains and lacks completeness. Further, when manually created, individual differences occur, and the reliability of the FT diagram also varies.

  On the other hand, quality assurance and troubleshooting have traditionally depended on knowledge (implicit knowledge) such as experience and intuition of skilled workers, but the amount of work of skilled workers is limited. Therefore, there is an urgent need in the field to formalize the knowledge of skilled workers so that anyone can work. Therefore, conventionally, formalization has been performed by recording and storing trouble cases and solutions. However, it is very difficult to search the same or similar cases from a large amount of accumulated cases and apply them to the problem to be solved.

  Therefore, there are conventionally various techniques for formalizing knowledge of experts and trouble cases. For example, the technique described in Patent Document 1 extracts keywords from case sentences, determines a cluster to which each case sentence belongs, and generates a problem solving tree characterized by a set of keywords based on the determination result. Then, the cluster to which each case sentence belongs is subdivided based on the language expression attribute of each case sentence belonging to each cluster of the problem solving tree. According to this technique, cases can be efficiently classified and a problem solving tree can be constructed, and cases including sentences that are syntactically and semantically similar to the search sentences can be searched.

  However, this technique is based on cluster classification, and does not focus on the causal structure of the cause and result of trouble cases, so it cannot be used for problem solving. In addition, there is a technique described in Patent Document 2 as a similar technique. The technique described in Patent Document 2 is a technique for extracting desired information from a vast amount of knowledge data stored in a computer, and is configured to fuse a plurality of knowledge data based on relevance. . This technology also shows relevance, does not focus on the causal structure, and therefore cannot be used for problem solving.

  If an FT diagram reflecting expert problem-solving techniques can be created without human intervention, it is effective in terms of cost and reliability, but there is no such thing yet.

JP 2001-188678 A

JP 2003-337933 A

  The present invention has been made in view of such problems, and an object of the present invention is to provide an FT diagram creation program and an FT diagram creation device capable of creating an FT diagram with low cost and high reliability. Another object of the present invention is to provide a recording medium and an FT diagram creation method.

In order to achieve the above object, the present invention analyzes topographic information of troubles such as accidents and failures occurring in an object and creates an FT diagram having a tree structure from the causes and results of the troubles. The FT diagram creation program includes a computer, a component subtree creation means, a function subtree creation means, a causal relation subtree creation means having at least an event extraction means and a causal relation extraction means, a node similarity It functions as an FT diagram creation means having a calculation means , a similar node connection means, a similar arc connection means, and a node arrangement determination means , and the part subtree creation means receives data relating to component parts of the object as parts. It is a means for creating a component subtree that represents each component node in a top-down tree structure as a node. The means for generating is a means for creating function subtrees in which data relating to functions in the object are respectively function nodes and each function node is represented in a top-down tree structure. The trouble information is structured by analysis, and all the events in the trouble are extracted based on the structured trouble information. The causal relation extracting means is the event extracted by the event extracting means. The causal relation subtree creating means uses each event extracted by the event extracting means as an event node, and each event node is extracted by the causal relation extracting means. Is a means for creating a causal relationship subtree expressed in a tree structure based on the Min trees, the function subtree and the causality subtree analyzed, a means for calculating a similarity of each node from the description contents of each node constituting each subtree, the similar node coupling means, said node If the similarity of the nodes calculated by the inter-similarity calculation means is greater than or equal to a certain threshold, it is determined that the nodes are the same node, and the component subtree, the functional subtree, and the causality This is a means for joining nodes of relational subtrees, and the similar arc joining means is a means for joining when two or more arcs exist between two nodes by determining that these arcs are the same arc. The node arrangement determining means is a means for determining a distance between the nodes based on the similarity between the nodes, and determining an optimal arrangement of the nodes based on the determined distance between the nodes. The diagram creation means is means for creating an FT diagram from the node combination by the similar node combination unit, the arc combination by the similar arc combination unit, and the optimum arrangement of each node determined by the node arrangement determination unit . Features.

  In the present invention, an FT diagram is automatically created by the following logic.

  One product usually consists of a plurality of parts, and each part functions normally by functioning as a whole.

  On the other hand, when one product fails, the failure occurrence location in the product is usually identified and the symptom is analyzed. If the symptom of the failure that finally occurred is investigated, it is a functional failure of one component or a plurality of components.

  Here, it is not necessary to have the skill of a skilled person to create a component sub-tree that is a top-down tree structure with the component parts of a product as nodes, and automatically using a computer if existing analysis techniques are used. Can be created automatically. Similarly, the function of each component can be easily specified from the specifications, etc., and the skill of an expert is required to create a functional subtree that is a top-down tree structure with the function in each component as a node. It is not necessary and can be automatically created by a computer using existing analysis techniques.

  On the other hand, various causes of failure and accidents in the product are conceivable, and it is necessary to have a high level of skill and intuition of a skilled person to specify the occurrence location. Usually, when a failure or accident occurs in a product, a report (example sentence) is created that describes the contents, location, and symptoms. If an FT diagram is created based on this example sentence, it is possible to create a highly reliable FT diagram reflecting the knowledge of an expert. However, when creating an FT diagram, the related cases are summarized. In addition, it is not easy to create a symptom, and it is necessary to create it manually by a skilled person.

  Therefore, in the present invention, the causal relationship between each event (unit describing physical phenomenon) that occurred in a failure or an accident is extracted from the example sentence, and the extracted causal relationship is created as described above. An FT diagram is created in association with the component subtree and the functional subtree.

  In other words, the present invention uses the dependency relationship between each component (node) and the hierarchical structure in the component sub-tree and the functional sub-tree, and applies the causal structure extracted from the case to these, thereby providing a concrete usable FT diagram. Can be created easily.

  Here, the “target object” is any object such as a product or its parts. Further, the “component” includes not only tangible objects but also intangible objects. For example, if the object is a “gas cylinder”, an intangible object such as an enclosed “gas” is also an object.

  The computer may be configured to realize a function of editing the created FT diagram.

  The part subtree creating means may acquire data related to the component parts of the object by reading a part configuration table by a reading means.

  The function subtree creating means may acquire data related to the function of the object by reading a specification with a reading means.

  You may make it acquire the said trouble information by reading the example sentence in which the information regarding a trouble was described by the reading means.

  As the “reading unit”, an optical reading unit such as a scanner is applicable.

  The FT diagram creation apparatus of the present invention is characterized by comprising the above FT diagram creation program.

  The recording medium of the present invention is characterized in that the FT diagram creation program is recorded.

  The recording medium may be a medium that can store a program mainly such as a floppy (registered trademark) disk or a CD-ROM that performs magnetic and optical storage.

In addition, the present invention analyzes topological information of troubles such as accidents and failures that occur in the object, generates a tree structure FT diagram from the cause and result of the trouble, creates a component subtree creation means, and creates a functional subtree. FT diagram creation means having at least a causal relation subtree creation means having at least an event extraction means and a causal relation extraction means, an internode similarity calculation means , a similar node connection means, a similar arc connection means, and a node arrangement determination means. a FT diagram creation method for creating with a computer to function, the data relating to the components of said object and part node respectively, the component parts of the component subtree that notation each part node in the top-down tree structure and the component parts tree creating step of creating, respectively function node data on the function of the object by tree creating means, The function subtree with notation function node in a top-down tree structure and function subtree preparing step of preparing the above functional parts tree creating means, and structuring the trouble information by analyzing the trouble information, the structured An event extracting step for extracting all the events in the trouble based on the trouble information extracted by the event extracting unit, and a causal relationship between the events extracted by the event extracting step by the causal relationship extracting unit an extraction step, each event extracted by the event extraction process to an event node, causality subtree the causality referred to the tree structure on the basis of a causal relationship to each event node extracted by the causality extraction step causation subtree creation step of creating the partial tree creating device, the component subtree, the functional subtree and the causal Analyzing the engagement subtree, and inter-node similarity calculation step of calculating by the inter-node similarity calculation means similarity of each node from the description contents of each node constituting each subtree, the inter-node similarity calculation step If the similarity of the nodes calculated by the above is equal to or greater than a certain threshold, it is determined that these nodes are the same node, and the parts subtree, the functional subtree, and the causal subtree The similar node combining process in which the similar nodes are combined by the similar node combining means, and when there are a plurality of arcs between two nodes, the arcs are determined to be the same arc and are combined by the similar arc combining means. Determine the distance between the nodes based on the arc coupling process and the similarity between the nodes, and determine the optimal arrangement of the nodes based on the determined distance between the nodes. An FT diagram is obtained from the node arrangement determination step determined by the means, the node combination by the similar node combination step, the arc combination by the similar arc connection step, and the optimum arrangement of each node determined by the node arrangement determination step . And an FT diagram creating step created by a diagram creating means .

  As described above, according to the present invention, a part subtree, a functional subtree, and a causal subtree are created and analyzed, and the similarity of the description contents of each node constituting each subtree is calculated and calculated. Based on the similarity, the component subtree, the functional subtree, and the causal subtree are integrated to create an FT diagram. As a result, an FT diagram that is specific and has high utility value can be created at low cost without depending on the knowledge of a skilled person.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is a functional block diagram for explaining a schematic configuration of an embodiment of the present invention, FIG. 2 is a functional block diagram showing details of FIG. 1, FIG. 3 is a flowchart showing a component part tree creation process, and FIG. FIG. 5 is a flowchart showing a causal relation subtree creation process, FIG. 6 is a flowchart showing an FT diagram creation process, FIG. 7 is a diagram showing an example of a part subtree, and FIG. 8 is a functional subtree. FIG. 9 is a diagram illustrating an example, FIG. 9 is an explanatory diagram for explaining a process of creating an FT diagram of the present invention by integrating a component subtree, a functional subtree, and a causal relationship subtree, and FIG. 10 is an FT created by the present invention. FIG. 11 is a diagram illustrating an example of a part configuration table, FIG. 12 is a diagram illustrating an example of a specification, and FIG. 13 is a diagram illustrating an example of a case sentence.

<Configuration>
FIG. 1 is a functional block diagram for explaining a schematic configuration of the present invention, and shows an embodiment of the present invention by a virtual functional block diagram. That is, FIG. 1 is a virtual functional block diagram showing that the computer C having the FT diagram creation program P creates and outputs an FT diagram by executing the FT diagram creation program P. .

  More specifically, as shown in FIGS. 1 and 2, the FT diagram creation program P of the present invention uses a computer C as a part subtree creation unit 110, a function subtree creation unit 210, and a causal relationship subtree creation unit 310. , A program that functions as the FT diagram creating means 400. The causal relationship subtree creation unit 310 includes at least an event extraction unit 312 and a causal relationship extraction unit 313, and the FT diagram creation unit 400 includes an inter-node similarity calculation unit 402.

  The computer C includes a CPU (C1), a main memory C2 using a RAM, an auxiliary storage device C3 such as a hard disk device, an input device C4 such as a keyboard / mouse or a scanner, and an output device C5 such as a display or a printer. .

  The FT diagram creation apparatus of the present invention is realized by a computer C that includes the FT diagram creation program P and executes the program P. The FT diagram creation method of the present invention is realized by executing the FT diagram creation program P. Therefore, in the following, only the FT diagram creation program of the present invention will be described, and description of the FT diagram creation device and the FT diagram creation method will be omitted.

  As shown in FIG. 1, the FT diagram creation program P of the present invention is generally configured as follows. First, the computer C is caused to function as the part subtree creating means 110 to create the part subtree 120 (see FIG. 7) from the part configuration table 100 (see FIG. 11). Next, the computer C is caused to function as the function subtree creating means 210 to create the function subtree 220 (see FIG. 8) from the specification 200 (see FIG. 12). Furthermore, the computer C is caused to function as the causal relation subtree creating means 310 to create a causal relation subtree (not shown) from the case sentence 300 (trouble information) (see FIG. 13). Then, the computer C is made to function as the FT diagram creating means 400 and is configured to create an FT diagram by integrating the component subtree 120, the functional subtree 220, and the causal relationship subtree (not shown). Yes. In the present embodiment, the FT diagram creation program P includes an editor 500 that can modify the created FT diagram.

  Each configuration will be described in detail with reference to FIG.

  The part subtree creation unit 110 includes a part configuration table analysis unit 111 and a part subtree construction unit 112, acquires electronic data from the part configuration table 100 by the input device C4, and creates the part subtree 120. Means.

  The part configuration table 100 is a document describing the part configuration of the target product or a document describing the part configuration of the part of the product, and an example thereof is shown in FIG. The part configuration table 100 includes a symbol column 101, a part name column 102, and a model / specification column 103 from the left in the drawing. Here, a part configuration table for a part “relay” of a product is shown.

  For example, the “relay” includes “electromagnet” (symbol: 1), “contact” (symbol: 2), “casing” (symbol: 4), and the like. In the symbol column 101, hierarchical symbols assigned to the respective components are entered. For example, the symbol “0” is assigned to the “relay” that is the highest part of the component configuration table, and the symbol “1” is assigned to the “electromagnet” that is the component of the “relay”.

  Furthermore, the symbol “1.1” is attached to the “coil” of the component part of “electromagnet” so that the hierarchical structure between the components can be clearly understood. Further, even when there are a plurality of component parts and a plurality of parts configuration tables are provided, the symbols are assigned so as not to overlap.

  Various methods can be applied as a method of acquiring electronic data from the component configuration table 100. For example, there are various methods such as a method of directly acquiring from the electronic data when the component configuration table 100 is created, and a method of acquiring the component configuration table 100 by optically reading the component configuration table 100 with a scanner or the like. The format of the parts configuration table is set so that it is easy to analyze and extract the hierarchical structure of the parts configuration and the dependency between the parts (for example, the part A is made up of parts B and C). Good.

  The part configuration table analysis unit 111 is a unit that understands and extracts the dependency relation regarding the configuration of the parts based on the configuration and structure of the part configuration table 100. If there is a more detailed part configuration table for the component, the component configuration is recursively decomposed to understand and extract the dependency. In the present embodiment, the component extracted by the component configuration table analysis unit 111 becomes a component node (N0, N01, N02,... In FIG. 7) that configures a component subtree 120, which will be described later. Thus, the hierarchical structure between the component nodes is extracted. Here, the symbols in the symbol column 101 are used to extract this dependency relationship. For example, the symbols given to the component names “coil” and “iron core” are the symbol: 1.1 and symbol: 1.2, respectively, and the symbol assigned to the component name “electromagnet” is “.” + Ascending order. A symbol with a number is used, and by extracting and analyzing this symbol, it is extracted that “coil” and “iron core” are dependent on “electromagnet”. Data relating to description knowledge relating to such charts is stored in advance in chart description knowledge 600 described later.

  The chart description knowledge 600 is a storage means such as a hard disk in which data relating to the chart such as the components of the functional block diagram and their legends, each item of the parts configuration table and the function definition table, the constituent elements and their meanings are stored. In this chart description knowledge 600, various types of component configuration tables and function definition tables are stored in advance as electronic data.

  The part subtree construction unit 112 creates a part subtree 120 by constructing the dependency between the component constituents extracted by the part configuration table analysis unit 111 in a tree structure, and generates data on the created part subtree 120. Save as internal data. The component components extracted and analyzed by the component configuration table analyzing unit 111 are defined as component nodes, and arcs (A1 to A16 in FIG. 7) between the component nodes are obtained from the dependency relationship between the component components. A part subtree 120 as shown in FIG.

  The function subtree creation means 210 includes a specification structure analysis means 211, a function block diagram understanding means 212, a function definition table understanding means 213, a function definition sentence understanding means 214, and a function structure tree construction means 215, and an input device C4. Is a means for acquiring electronic data from the specification 200 and creating a functional subtree 220.

  The specification document 200 is a design document describing the functional configuration of the target product and the parts constituting the product, and an example thereof is shown in FIG. The specification 200 is composed of a function block diagram 201, a definition 202 for each function, and an operation 203 for each function from the top as entries. The functional block diagram 201 is a diagram in which each component is functionally expressed as a block, and dependency relationships (signal flow, interlocking relationship, etc.) between the blocks are indicated by arrows.

  Each function definition 202 includes an item column and a definition column for each item. The name in the block is entered in the item column, and each configuration described in the item column is displayed in the definition column. An explanatory text or a definition text for the function is described. In the operation 203 of each function, the function / operation description for each configuration described in the item column is hierarchically described with a symbol for each item.

  As described above, in the specification 200, each component is described using a block diagram (functional block diagram 201), a function definition table (definition 202 for each function), and a function definition sentence (operation 203 for each function). ing. It should be noted that this specification sheet 200 may be created in a format that can be easily extracted and analyzed by each means to be described later, similarly to the component configuration table. Also, as a method for acquiring electronic data from the specification sheet 200, a method similar to the acquisition method in the component configuration table 100 can be applied.

  The specification structure analysis means 211 is a means for extracting a sentence / diagram / table which is an element of a functional subtree based on the chapter structure and diagram structure of the specification 200. This is a means for analyzing how the overall configuration of the specification 200 is. By this analysis, it is possible to grasp the overall configuration of the specification 200 such as “contains a sentence”, “contains a diagram”, and “contains a table”. Detailed analysis in each is performed in each means 212, 213, and 214 described later.

  The functional block diagram understanding unit 212 is a unit that interprets the arrangement of blocks, arrows, edges, and the like in the block diagram, and understands and extracts the dependency of each sub-function (contact opening / closing unit, housing function, connection unit, etc.). . The sub-function is a function exhibited by lower-order components in order to exhibit the main function. For example, when the product is a “relay”, the function of the entire “relay” is the main function, and the function of each component (eg, electromagnet, terminal, housing, etc.) constituting the “relay” is a sub-function. It is. When there are specifications, block diagrams, function definition tables, and texts that explain the details of sub-functions in detail, sub-functions are decomposed and understood recursively.

  The function definition table understanding unit 213 is a unit that interprets the items shown in the function definition table 202 and understands and extracts the contents of each sub-function and the dependency relationship between the sub-functions. When there are specifications, block diagrams, function definition tables, and texts that explain the details of sub-functions in detail, sub-functions are decomposed and understood recursively.

  The analysis / extraction in the specification structure analysis unit 211 and the function definition table understanding unit 213 is performed based on data relating to tables and diagrams accumulated in the diagram description knowledge 600 described above.

  The function definition sentence understanding means 214 is a means for interpreting the sentence shown in the function definition sentence 203 to understand and extract the definition / content of each subfunction and the dependency between the subfunctions. When there are specifications, block diagrams, function definition tables, and texts that explain the details of sub-functions in detail, sub-functions are decomposed and understood recursively. The analysis of the sentence shown in the function definition sentence 203 can be performed by various text analysis such as morphological analysis and dependency analysis. For example, as a tool for performing morphological analysis, there is “ChaSen” which is free software, and as a tool for performing dependency analysis, there is “CaboCha” which is free software.

  The grammar knowledge 700 is a storage means such as a hard disk in which grammar knowledge for sentence comprehension is accumulated. For example, dictionaries such as the above-mentioned teacup and Nanban dictionaries and Japanese vocabulary system (Iwanami Shoten) are stored.

  The function structure tree construction unit 215 constructs a dependency structure between the sub-functions extracted by each of the function block diagram understanding unit 212, the function definition table understanding unit 213, and the function definition sentence understanding unit 214 into a tree structure to generate internal data. Save to. The sub-functions extracted and analyzed by the understanding means 212, 213, and 214 are defined as function nodes (N1, N2, N3,... In FIG. 8), and arcs between the function nodes are determined from the dependency between the sub-functions. (B1 to B13 in FIG. 8) is obtained, and a functional subtree 220 as shown in FIG. 8 is created.

  The causal relation subtree creation means 310 includes a case sentence analysis means 311, an event extraction means 312, a causal relation extraction means 313, and a causal relation subtree construction means 314, and acquires electronic data from the case sentence 300 by the input device C 4. And a means for creating a causal relation subtree (not shown).

  The example sentence 300 is a document that describes information about various troubles that have occurred in the object such as defects, failures, and histories of the target product and parts of the product, and an example is shown in FIG. For example, this example sentence includes various documents such as an improvement report on a process defect, a customer trouble answer sheet, a failure history, a design change report, and a production work daily report. The example sentence 300 shown in FIG. 13 includes “details of returned product” 301, “your offer item” 302, “conclusion” 303, “analysis result” 304, and “occurrence mechanism” 305 as entry items. ing.

  Here, “Details of returned product” 301 describes information for identifying the target product in which the failure occurred, and “Your offer item” 302 describes information related to the failure received from the customer. The “conclusion” 303 describes the cause and result of the failure, that is, what kind of failure occurred for what reason, and the “analysis result” 304 describes the failure target product. The occurring symptom is described, and the “occurrence mechanism” 305 analyzes the cause of the occurrence of the failure and describes the mechanism. A method similar to the method in the component configuration table can also be applied to the method of acquiring electronic data from the case sentence 300.

  The case sentence analysis means 311 is a means for structuring the description content of the case sentence 300 based on the chapter structure and grammar analysis of the document. After analyzing the sentence structure, necessary tag information (morpheme / dependency information, chapter structure information, bullet information, etc.) is given based on the structure.

  The event extraction means 312 analyzes the structured document based on the assigned tag information and extracts events (units describing physical phenomena). In the event extraction, for example, a predicate and a set of independent words that modify the predicate are extracted with reference to data of event extraction knowledge 801 described later based on case frame analysis. For example, if an item is listed (including an ordered item), that item is treated as one event. Also, when expressed in natural language sentences, the dependency structure between vocabularies is analyzed by syntactic analysis, and as a set of independent words related to predicates-predicates (independent words related to multiple predicates for one predicate) Extract events.

  The event extraction knowledge 801 is storage means in which data related to the definition of an event is stored. For example, knowledge of predicates + vocabulary that modifies predicates, knowledge of using each item in bulleted items, and the like are accumulated.

  The causal relationship extraction unit 313 is a unit that determines the presence / absence of a causal relationship based on the assigned tag information and extracts the causal relationship / dependency between events. The presence / absence of a causal relationship between events is determined by a statistical model of causal relationships created in advance based on teacher data, and causal relationships and dependency relationships between events are extracted. For example, in the case of an event extracted from an itemized list, it is estimated that there is a causal relationship in order based on the chained structure of the itemized list and extracted.

  In addition, when a natural language sentence includes connection words such as “by”, “for”, and “with”, there is a causal effect between the event that exists immediately before the connection word and the event that immediately follows the connection word. Estimate and extract the relationship. Further, all the combinations of events extracted from cases are collated with a causal model, and it is estimated that a combination having a high degree of similarity with the model has a causal relationship and is extracted. The causal model is statistically learned by a statistical model (for example, GMM: Gaussian mixture distribution model) in which causal event pairs are provided as teacher data. The similarity with the model is obtained statistically by the likelihood between the model and the measurement object. As a known example of such a causal relationship extraction method, there is a method disclosed by the present inventors in “FIT2007 (6th Information Science and Technology Forum)“ Knowledge generation system in similar fields based on analogy ””. .

  The causal structure extraction knowledge 802 is a storage means such as a hard disk in which a causal relationship extraction criterion is stored as knowledge data. For example, the knowledge that the events before and after the connected vocabulary “by” are extracted, the statistical model of the causality that is learned by the teacher data, and the like are stored.

  The causal relation subtree construction means 314 is a means for constructing a causal relation / dependency relation between extracted events into a tree structure and storing it in internal data.

  The FT diagram creation unit 400 includes a tree structure data analysis unit 401, an internode similarity calculation unit 402, a similar node coupling unit 403, a similar arc coupling unit 404, a node arrangement determination unit 405, and an FT diagram construction unit 406.

  The tree structure data analysis unit 401 is a unit that analyzes the component subtree 120, the functional subtree 220, and the causal relationship subtree (not shown) and stores them in internal data in a common format.

  The inter-node similarity calculation unit 402 is a unit that calculates the distance between the nodes of the part subtree 120, the functional subtree 220, and the causal relationship subtree (not shown) based on the similarity of the contents described in the nodes. . The similarity between nodes is calculated by a cosine measure used in calculating a similarity between a search query vector and each document vector in a vector space model which is one of search models for performing a search. Details will be described later.

  The similar node combining means 403 is a means for determining that the similarity calculated by the inter-node similarity calculating means 402 is equal to or greater than a threshold value, and determining that these nodes are the same node and combining them. This threshold value needs to be set in advance by the user, but if it is configured to be changeable by the user, the same determination accuracy can be improved.

  When there are a plurality of arcs between two nodes, the similar arc coupling unit 404 determines that the arcs are the same, and couples the arcs. Referring to the example of the FT diagram shown in FIG. 10, there is no more than two arcs between two nodes. Therefore, when a plurality of arcs are formed between two nodes as a result of the combination by the similar node combining unit 403, the similar arc combining unit 404 consolidates this state into one arc.

  The node arrangement determining means 405 determines the distance between the nodes based on the similarity between the nodes, and determines the optimum arrangement of each node based on the distance between the nodes. The distance between nodes is given by, for example, the reciprocal of node similarity.

  The FT diagram construction means 406 is a means for reconfiguring the arrangement of each node and arc based on the tree network configuration constituted by nodes and arcs and the node arrangement determined by the distance between nodes. It is a means to create.

  The editor 500 is an application that can add / delete / update / modify nodes and arcs to the created FT diagram. The created FT diagram is displayed on the display of the computer C and can be modified by the user using a mouse or keyboard.

  In FIG. 2, the chart description knowledge 600, the grammar knowledge 700, the event extraction knowledge 801, and the causal structure extraction knowledge 802 are described separately, but these are described separately for convenience in drawing. In the embodiment, both are provided in the auxiliary storage device C3 shown in FIG.

<Processing operation>
The above is the description of the configuration of the present invention. Next, the processing operation of the present invention will be described with reference to FIGS.

  The outline of the processing operation in the present invention will be described as follows. First, the component subtree creation process (process 1) shown in FIG. 3, the functional subtree creation process (process 2) shown in FIG. 4, and the causal relation subtree creation process (process 3) shown in FIG. 5 are performed. Next, an FT diagram creation process (Process 4) is performed in which the part subtree, the functional subtree, and the causal relationship subtree created by these processes are integrated to create an FT diagram, and an FT diagram is created. The above is the outline of the processing operation of the present invention, but each processing operation will be described in detail below. Note that the processing 1, processing 2, and processing 3 are in no particular order. In addition, each process may be performed as a series when creating an FT diagram, or the process may be terminated in advance before creating the FT diagram.

[Process 1: Part subtree creation process]
As shown in FIG. 3, when the user operates a keyboard, a mouse, or the like of the computer C, data reading processing of the component configuration table 100 is started (S100). Here, as the original data of the parts structure table 100, electronic data when the parts structure table 100 is created is used. When this reading is completed, the component configuration table analysis unit 111 then accesses the diagram description knowledge 600 and analyzes the configuration / structure of the component configuration table 100 based on the data related to the diagram stored in the diagram description knowledge 600. Then, based on the analyzed result, the component parts that are the nodes of the component subtree 120 and the dependency relationship between the component parts are extracted (S102). In the present embodiment, since the symbols in the symbol column 101 are assigned hierarchically, this dependency relationship is extracted from the symbols in the symbol column 101.

  Thereafter, the component subtree construction unit 112 obtains each node and arc based on the extracted component and the dependency relationship between the components, and creates the component subtree 120 shown in FIG. 7 (S103). Data relating to the created part subtree 120 is stored as internal data (S104), and the part subtree creation process ends.

[Process 2: Function subtree creation process]
As shown in FIG. 4, first, a specification data reading process is performed (S200). This reading process is performed by the same method as the reading process (S100). Next, the structure / structure of the specification 200 is analyzed by the specification structure analysis means 211 (S201). In the present embodiment, since the specification document 200 includes a function block diagram 201, a function definition table 202, and a function definition sentence 203, a diagram, a table, and a sentence are extracted as a result of this analysis (S202). .

  Next, the functional block diagram understanding unit 212 interprets the inclusion relationship of the blocks in the functional block diagram and the arrangement of arrows, edges, etc., and understands the dependency relationship of each sub-function (contact opening / closing unit, housing function, connection unit, etc.). Extract (S203). Further, the function definition table understanding unit 213 interprets the items shown in the function definition table 202, and understands and extracts the contents of each subfunction and the dependency between the subfunctions (S204). For example, it analyzes whether or not other items are included in the sentence of each item, extracts the dependency, understands the meaning of the sentence of each item, and extracts the causal relationship. These extraction processes are performed by each means 212 and 213 based on data relating to the chart stored in the chart description knowledge 600 by accessing the chart description knowledge 600.

  In the function definition sentence understanding unit 214, the grammar knowledge 700 is accessed, and based on the data related to the grammar knowledge accumulated in the grammar knowledge 700, the hierarchical structure of each sentence structure (section structure) shown in the function definition sentence 203 The meaning of the sentence is analyzed, and the definition / content of each sub-function and the dependency between each sub-function are understood / extracted (S205). The subfunctions extracted by these extraction processes become nodes, and an arc is obtained from the dependency between the subfunctions.

  In the function subtree construction means 215, the function subtree 220 shown in FIG. 8 is obtained with the subfunctions as nodes and the dependencies between the subfunctions as arcs based on the extracted subfunctions and dependencies between the subfunctions. It creates (S206), stores the data related to the created function subtree 220 as internal data (S207), and ends the function branch tree creation process.

[Process 3: Causal Relation Subtree Creation Process]
As shown in FIG. 5, first, case data reading processing is performed (S300). This reading process is performed by the same method as the reading process (S100). Next, in the case sentence analysis means 311, the description content of the case sentence 300 is structured based on the chapter structure and grammatical analysis of the document, the sentence structure is text-analyzed, and necessary tag information is obtained based on the structure. (S301).

  Next, the event extraction unit 312 analyzes the case frame of the document structured based on the tag information given above, and extracts the event based on the data accumulated in the event extraction knowledge 801 (S302).

  In the causal relationship extraction means 313, the causal relationship extraction knowledge 802 is accessed, the presence or absence of the causal relationship is determined based on the knowledge data accumulated in the causal relationship extraction knowledge 802 and the tag information given above, and the causality between the events Relationships and dependency relationships are extracted (S303).

  The causal relation subtree construction means 314 creates a causal relation subtree using each event extracted by the extraction process as a node (event node) and a causal relation between each event as an arc (S304). Data relating to the causal relationship subtree is stored as internal data (S305), and the causal relationship subtree creation processing is terminated.

[Process 4: FT diagram creation process]
As shown in FIG. 6, the tree structure data analysis unit 401 analyzes the component subtree 120, the functional subtree 220, and the causal relationship subtree (not shown) created by the processes 1 to 3 and stored in the internal data. Then, the data is stored in the internal data in the common format (S401).

  Next, the inter-node similarity calculation means 402 calculates the distance between the nodes for all the nodes in the stored subtree based on the similarity of the contents described in the nodes (S401). FIG. 9 schematically shows the similarity calculation process. Here, the functional subtree 220 is shown in the upper row, the component subtree 120 is shown in the lower row, and each event extracted from the case sentence 300 is represented by R1, R10, R100,. For example, “A“ gas leak ”occurred and a“ contact arc ”occurred, resulting in a“ relay operation failure ”. Are extracted from the case sentence 300 regarding the failure as “gas leak” R100, “contact arc” R10, and “relay operation failure” R1, and the causal relationship between the events is indicated by arrows. .

  Further, a process of calculating similarity between each event R1, R10, R100 and the component node of the component subtree 120 and the function node of the functional subtree 220, and linking and combining similar events is schematically shown. Has been shown. Specifically, it is determined that the component node N042 and the event R100, the component node N02 and the event R10, and the function node N1 and the event R1 are similar to each other, and are combined.

  This inter-node similarity is calculated by a cosine measure as follows, for example.

The similarity between node A and node B is S, the vocabulary included in node A is W _A1 , W _A2 , W _A3 ,..., W _AN , and the vocabulary included in node B is W _B1 , W _B2 , W _B3 ,
.., W _BN , and a word vector representing the characteristics of the vocabulary W is V _W. The word vector V _W is given by enumerating vocabulary frequencies related to W obtained by dependency analysis of a large-scale document set. _A vector V _A representing the characteristics of the node A is a word vector sum V _WA1 + V _WA2 + V _WA3 +... + V _WAN . The same applies to the vector V _B representing the characteristics of the node B. Then, the similarity S is calculated by the cosine measure of vector _{V A} and _{V B.} If the similarity S = sim (a, b), the vector V _A is a, and the V _B is b, the similarity is calculated by the following equation.

<Formula>

  As a result of calculating the similarity, the similar node combining unit 403 checks whether the calculated similarity is equal to or higher than a threshold value, and nodes that are equal to or higher than the threshold value are combined (S402).

  After the above node combining process is completed, the similar arc combining means 404 checks whether or not there is an arc with the same From and To between nodes for all nodes. Are combined into one arc (S403). For example, when there are two or more arcs directly from node A to node B, one arc is left and the others are deleted.

  Next, the node arrangement determining means 405 calculates the distance between the nodes based on the similarity between the nodes (S404), and determines the optimum arrangement of each node based on the calculated distance between the nodes (S405). . The distance between the nodes is given by, for example, the reciprocal of the node similarity. The node arrangement is determined by the following algorithm.

The nodes are N ₁ , N ₂ , N ₃ ,... N _N , the coordinates of each node are N _n (x _n , y _n ), the distance D _nm between the nodes N _n and N _m = the reciprocal of the similarity between nodes The node coordinates that minimize the following expression are obtained by the steepest descent method.

なお、Σ（総和）は、全てのｍｎの組み合わせについて行う。ａｂｓ（）は絶対値を意味する。 <Formula>

Note that Σ (total) is performed for all mn combinations. abs () means an absolute value.

  After the above processing, the FT diagram constructing unit 406 creates an FT diagram in which nodes and arcs are optimally arranged (S406), displays the FT diagram on the display of the computer C (S407), and the processing ends. An example of the FT diagram created by this processing is shown in FIG. As described above, the causal relationship between each event in the case is created in association with the component subtree 120 and the functional subtree 220. Therefore, this FT diagram is an FT diagram that reflects the problem solving technique of the expert. It has become.

  When the created FT diagram is to be modified, the editor 500 is activated, and the user can add or modify a node or arc by operating a mouse or a keyboard.

The functional block diagram for demonstrating schematic structure of embodiment of this invention. The functional block diagram which shows the detail of FIG. The flowchart which shows the components subtree preparation process in embodiment of this invention. The flowchart which shows the function subtree preparation process in embodiment of this invention. The flowchart which shows the causal relationship subtree preparation process in embodiment of this invention. The flowchart which shows the FT figure creation process in embodiment of this invention. The figure which shows an example of a component partial tree. The figure which shows an example of a functional subtree. Explanatory drawing for demonstrating the process which unifies a component subtree, a functional subtree, and a causal relationship subtree, and produces the FT figure of this invention. The figure which shows an example of the FT figure created by this invention. The figure which shows an example of a components structure table. The figure which shows an example of a specification document. The figure which shows an example of a case sentence.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Parts structure table 110 Part subtree creation means 120 Part subtree 200 Specifications 210 Functional subtree creation means 220 Functional subtree 300 Case sentence 310 Causal relation subtree creation means 312 Event extraction means 400 FT diagram creation means 401 Similarity between nodes Degree calculation means 500 Editor C Computer P FT diagram creation program

Claims (8)

A FT diagram creation program that analyzes top-down trouble information such as accidents and breakdowns that occur on an object and creates a tree-structured FT diagram from the causes and results of the trouble,
The FT diagram creation program is
Computer, part subtree creation means, functional subtree creation means, causal relation subtree creation means having at least event extraction means and causal relation extraction means, internode similarity calculation means , similar node connection means, similar arc connection means And an FT diagram creating means having a node arrangement determining means ,
The part subtree creation means is
It is a means for creating a part subtree in which the data related to the component of the target object is a part node and each part node is represented in a top-down tree structure,
The above functional subtree creation means
It is a means for creating a functional sub-tree in which data related to functions in the object are function nodes, and each function node is represented in a top-down tree structure.
The event extraction means is:
It is a means for structuring the trouble information by analyzing the trouble information, and extracting all events in the trouble based on the structured trouble information,
The causal relationship extraction means is:
A means for extracting a causal relationship between events extracted by the event extracting means;
The above causal relationship subtree creation means is:
Each event extracted by the event extraction means is an event node, and a means for creating a causal relationship subtree that describes each event node in a tree structure based on the causal relationship extracted by the causal relationship extraction means,
The inter-node similarity calculation means includes:
Analyzing the part subtree, the functional subtree, and the causal subtree, and calculating the similarity of each node from the description contents of each node constituting each subtree;
The similar node coupling means is:
If the node similarity calculated by the inter-node similarity calculation means is greater than or equal to a certain threshold value, it is determined that the nodes are the same node, and the component subtree, the functional subtree, and A means for combining nodes of the causal relation subtree,
The similar arc combining means is means for determining and combining the arcs when the plurality of arcs are present between two nodes,
The node arrangement determining means is
A means for determining a distance between nodes based on a similarity between the nodes, and determining an optimal arrangement of each node based on the determined distance between the nodes;
The FT diagram creation means is:
The FT diagram is a means for creating an FT diagram from the combination of nodes by the similar node combination unit, the combination of arcs by the similar arc combination unit, and the optimum arrangement of each node determined by the node arrangement determination unit . Creation program.   The FT diagram creation program according to claim 1, wherein the computer is caused to realize a function of editing the created FT diagram.   The FT diagram creation program according to claim 1 or 2, wherein the part subtree creation means obtains data relating to the component parts of the object by reading a part configuration table by a reading means.   The FT diagram creation program according to any one of claims 1 to 3, wherein the functional subtree creation means acquires data related to the function of the object by reading a specification with a reading means.   The FT diagram creation program according to any one of claims 1 to 4, wherein the trouble information is acquired by reading a case sentence in which information related to the trouble is read by a reading unit.   An FT diagram creation apparatus comprising the FT diagram creation program according to claim 1.   6. A recording medium on which the FT diagram creation program according to claim 1 is recorded. Top-down analysis of trouble information such as accidents and failures occurring in the object, and FT diagram of the tree structure from the cause and result of the trouble , parts subtree creation means, functional subtree creation means, at least event extraction means And a causal relationship sub-tree creating means having a causal relation extracting means, an inter-node similarity calculating means , a similar node joining means, a similar arc joining means, and an FT diagram creating means having a node arrangement determining means. An FT diagram creation method to create,
A part subtree creating step of creating a part subtree that represents each component node in a top-down tree structure by means of the part subtree creating means , with the data relating to the constituent parts of the object as parts nodes,
A function subtree creating step of creating, by using the function subtree creating means, a function subtree representing each function node in a top-down tree structure, with data relating to functions in the object as function nodes.
Structuring the trouble information by analyzing the trouble information, an event extracting step of extracting all events in the trouble by the event extracting means based on the structured trouble information;
A causal relationship extracting step of extracting the causal relationship between the events extracted by the event extracting step by the causal relationship extracting means ;
The causal relation subtree creation means for creating a causal relation subtree in which each event extracted in the event extraction step is an event node and each event node is represented in a tree structure based on the causal relation extracted in the causal relation extraction step The causal subtree creation process created by
Analyze the part subtree, the functional subtree, and the causal subtree, and calculate the similarity of each node from the description contents of each node constituting each subtree by the internode similarity calculation means. Degree calculation step,
If the node similarity calculated by the inter-node similarity calculation step is equal to or greater than a certain threshold value, it is determined that the nodes are the same node, and the component subtree, the functional subtree, and A similar node combining step of combining nodes of the causal relation subtree by the similar node combining means;
When there are a plurality of arcs between two nodes, the arcs are determined to be the same arc and are combined by the similar arc combining means,
A node arrangement determining step of determining a distance between the nodes based on the similarity between the nodes, and determining an optimum arrangement of each node by the node arrangement determining means based on the determined distance between the nodes;
FT diagram creation step of creating an FT diagram by the FT diagram creation means from the node combination by the similar node combination step, the arc combination by the similar arc connection step, and the optimum arrangement of each node determined by the node arrangement determination step An FT diagram creation method characterized by comprising:

Images