Tree generation device and tree generation method

Abstract

To integrate a plurality of kinds of causal models generated from separate information sources into one tree structure.SOLUTION: A tree generation device 100 comprises a model coupler 25 which uses a top event as a root node of a mixed tree 31 and generates a mixed tree 31 including a plurality of causal models, and an output device 32 which displays the mixed tree 31. When a combination of a notation of a first node in a first causal model and a notation of a second node in a second causal model among a plurality of causal models is registered in an attribute DB 22 as a notation variability, the model coupler 25 couples the first causal model and the second causal model by inserting an adapter node for converting the notation variability between the first node and the second node and connecting the nodes by a link.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a tree generation device and a tree generation method. [Background technology]

[0002] It is often the case that past cases are referred to in order to investigate the causes of defects that occurred in those past cases. In such cases, it is inefficient for humans to read the documents of many past cases one by one. Therefore, a system has been proposed that converts the documents of past cases into a data structure that is easy for humans to read. Patent document 1 describes "a case acquisition means for acquiring cases consisting of attributes and attribute values ​​to form a case base, a rule generation means for generating a decision tree from the case base, and a consulting means for determining a conclusion attribute value by referring to the generated decision tree." [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2005-339529 A Summary of the Invention [Problem to be solved by the invention]

[0004] Here, causal models that show the causal relationships of defects may be generated independently by different people from various perspectives. For example, a university professor generates a mathematical model from a formula that shows a physical event (physical phenomenon). A mathematical model here is a causal model based on a formula. On the other hand, a field worker generates an empirical model from actual examples that he or she has experienced. An empirical model is a causal model based on experience.

[0005] However, since the terms used in multiple types of causal models are created by people from different fields, there is often no uniformity in the way basic terms are written. For example, the defect report from which the empirical model was created may state that "a part slips" from the viewpoint of a field worker, but the mathematical expression representing the physical phenomenon from which the mathematical model was created may state the phenomenon of "a part slips" in a different technical term, such as "the part moves at an excessively high speed."

[0006] In this way, when trying to connect different models with no standardized notation, the names of the nodes are all different, so it is impossible to specify which cause node should be connected to which result node. Therefore, it is difficult to construct a tree structure (fault tree) in which the result node is the root (tree root) and the cause node can be found by gradually tracing the branches and leaves of the tree from the result node. In addition, Patent Document 1 provides a mechanism for generating a tree structure called a decision tree. However, conventional techniques such as Patent Document 1 only generate a tree structure from one perspective, and do not provide a means for integrating multiple types of causal models into one tree structure.

[0007] Therefore, a main object of the present invention is to integrate multiple types of causal models generated from different information sources into one tree structure. [Means for solving the problem]

[0008] In order to solve the above problems, the tree generating device of the present invention has the following features. The present invention includes an input unit for receiving an input of a top event to be analyzed; a model combiner that generates a causal tree including a plurality of causal models, the causal tree being the root node of the causal tree based on the top event; an output unit that outputs the causal tree generated by the model combiner to a presentation device; The model combiner comprises: A plurality of causal models, each of which is generated from a different information source, are read from a model DB as a causal model in which nodes in which a causal relationship exists are connected by a link; The method is characterized in that, by referring to a multiple notation DB that matches combinations of notation variations of nodes between the multiple causal models read, if a combination of the notation of a first node in a first causal model among the multiple causal models read and the notation of a second node in a second causal model is registered in the multiple notation DB as the notation variation, an adapter node for converting the notation variation is inserted between the first node and the second node and connected with a link, thereby combining the first causal model and the second causal model and generating the causal tree. Other means will be described later. Effect of the Invention

[0009] The present invention allows multiple types of causal models generated from different information sources to be integrated into a single tree structure. [Brief description of the drawings]

[0010] [Figure 1] FIG. 2 is a configuration diagram of a tree generation device according to the present embodiment. [Diagram 2] 4 is a flowchart showing the process of the tree generating device according to the present embodiment. [Diagram 3] FIG. 2 is a hardware configuration diagram of a tree generation device according to the present embodiment. [Figure 4] FIG. 11 is an explanatory diagram of a fault tree manually created from an empirical model DB as a comparative example. [Diagram 5] FIG. 2 is an explanatory diagram showing an example of the mixed tree of FIG. 1 according to the present embodiment. [Figure 6] 2 is an explanatory diagram showing a detailed example of the mixed tree of FIG. 1 according to the present embodiment; FIG. [Figure 7] 11 is an explanatory diagram showing a process in which an apex event combiner according to the present embodiment creates a formula tree; FIG. [Figure 8] FIG. 2 is an explanatory diagram showing an example of a synonym DB according to the embodiment; [Figure 9] 9 is an explanatory diagram showing an example of a mathematical expression tree created using the synonym DB of FIG. 8 according to the present embodiment. FIG. [Figure 10] FIG. 2 is an explanatory diagram showing an example of an attribute DB according to the present embodiment; [Figure 11] FIG. 1 is an explanatory diagram showing a mathematical model and an empirical model having different levels of abstraction according to the present embodiment. [Figure 12] 12 is an explanatory diagram showing an example of processing in which an abstraction level converter creates an adapter node with reference to the attribute DB in FIG. 10 for the data structure shown in FIG. 11 according to this embodiment. FIG. [Figure 13] FIG. 2 is an explanatory diagram showing an example of a product structure DB according to the present embodiment. [Figure 14] FIG. 1 is an explanatory diagram showing a mathematical model and an empirical model having different levels of abstraction according to the present embodiment. [Figure 15] 15 is an explanatory diagram showing an example of a process in which a structure converter creates an adapter node by referring to the product structure DB in FIG. 13 for the data structure shown in FIG. 14 according to this embodiment. FIG. [Figure 16] 14 is an explanatory diagram showing an example of a product structure DB different from that in FIG. 13 according to the present embodiment. FIG. [Figure 17] FIG. 2 is a screen diagram showing a first example of a screen on which an output device according to the present embodiment displays a mixed tree. [Figure 18] FIG. 11 is a screen diagram showing a second example of a screen on which the output device according to the embodiment displays a mixed tree. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

[0012] FIG. 1 is a diagram showing the configuration of a tree generating device 100. As shown in FIG. The tree generating device 100 has, as processors, an input unit 14, a top event combiner 15, a tree generator 20 (an abstraction level converter 21, a structure converter 23, and a model combiner 25), and an output unit 32. The storage unit of the tree generating device 100 stores a synonym DB 11, an empirical model DB 13, a mathematical expression model DB 12, a tree generator 20 (an attribute DB 22 and a product structure DB 24), a mathematical expression tree 16, and a mixed tree 31. Alternatively, at least one of the various DBs (the synonym DB 11, the empirical model DB 13, the mathematical expression model DB 12, the attribute DB 22, and the product structure DB 24) may be present outside the tree generating device 100, and the tree generating device 100 may appropriately refer to these various DBs when generating the mixed tree 31. Details of each component of tree generating device 100 in FIG. 1 will be made clear in the description of the flowchart in FIG.

[0013] Below, we will briefly explain "models" and "trees" (see Figure 5 for details). In both models and trees, the upper node indicates the result of an event, and the lower node connected to the upper node by a link corresponds to the cause of the event. In addition, the formula tree 16, the mixed tree 31, and other trees have the top event input via the input unit 14 as the root node (top node). On the other hand, the causal models stored in the empirical model DB13 and the mathematical model DB12 do not have top events as root nodes, but are prepared in advance as partial materials (a collection of nodes with links between them) to be integrated into the tree.

[0014] Causal models, which are formalized mathematical expressions, are stored in advance as mathematical expressions models in the mathematical expression model DB 12. The mathematical expressions used in the mathematical expression models can be extracted from those described in technical documents such as research reports, engineering handbooks, and specifications. Causal models that formalize information on defects that have occurred in the past are stored in advance as empirical models in the empirical model DB 13. Defect information used in the empirical model may be extracted from reports by on-site workers who have dealt with defects, or may be extracted from defect information provided by experts, for example. Therefore, the mathematical model DB 12 and the empirical model DB 13 store a plurality of (separate) causal models generated from mutually different information sources as causal models that connect nodes in which causal relationships exist with links.

[0015] The input unit 14 accepts input of a top event to be analyzed. The top event combiner 15 creates a mathematical expression tree 16 by using the top event input via the input unit 14 as a root node and combining a mathematical expression model below it. The tree generator 20 creates a mixed tree 31 by combining an empirical model below the mathematical expression tree 16. The output unit 32 displays the mixed tree 31 by outputting the mixed tree 31 to a presentation device (such as a monitor device or a display device).

[0016] Here, in each data (top event, mathematical expression model, empirical model) that is the material of one mixed tree 31, there may be spelling variations in the terms used in the nodes due to differences in viewpoints such as different data creators. Therefore, the tree generating device 100 provides means for resolving spelling variations, such as the following (Resolution Means 1) to (Resolution Means 3). (Resolution means 1) The top event combiner 15 resolves the spelling variations between the top event and the mathematical model by referring to the synonym DB 11. For this purpose, the synonym DB 11 has registered in advance combinations of synonyms that associate terms that have the same meaning but multiple spellings with each other (see FIG. 8 for details).

[0017] (Resolution Means 2) The abstraction level converter 21 resolves the variation in notation of events (attributes) that constitute nodes between the mathematical model and the empirical model by referring to the attribute DB 22. For this purpose, the attribute DB 22 preregisters combinations of attributes that associate multiple expressions with different abstraction levels (broad and narrow expressions) for one attribute (see FIG. 10 for details).

[0018] (Resolution Means 3) The structure converter 23 resolves the variation in notation of part names (structures) constituting nodes between the mathematical model and the empirical model by referring to the product structure DB 24. For this reason, the attribute DB 22 has registered in advance combinations of part names that associate multiple structural expressions with different levels of abstraction (part names in a broad sense and part names in a narrow sense) for one part (see FIG. 13 for details). In other words, the attribute DB 22 and the product structure DB 24 each associate combinations of notation variations of nodes among a plurality of causal models.

[0019] The model combiner 25 creates a mixed tree 31 by combining the mathematical expression tree 16 (mathematical expression model) in which the notation variations have been resolved by (Resolution Means 2) and (Resolution Means 3) with an empirical model. That is, the model combiner 25 generates a mixed tree 31 in which the top event is set as the root node of the mixed tree 31 and which includes multiple causal models. Here, when a combination of the notation of a first node in a first causal model among multiple causal models and the notation of a second node in a second causal model is registered in the attribute DB22 as a notation variation, the model combiner 25 combines the first causal model and the second causal model by inserting an adapter node for converting the notation variation between the first node and the second node and connecting them with a link.

[0020] FIG. 2 is a flowchart showing the processing of the tree generating device 100. The tree generating device 100 prepares various models (empirical model DB13, mathematical model DB12) that are components of the mixed tree 31, and various DBs (synonym DB11, attribute DB22, product structure DB24) that are referenced to connect the various models, by data input from the user, etc. (S11).

[0021] The input unit 14 receives an input of a top event (S12), as exemplified below. When a malfunction occurs, a field worker can input the malfunction event as a top event (e.g., part=belt, event=breakage) into the input device 14, and use the output mixed tree 31 to identify the cause of the malfunction. When designing a new part, the designer can input, as a top event, an event that is undesirable for the new part to occur in the system, into the input device 14, and use the output mixed tree 31 to identify design consideration items (to be reflected in the design requirements). When purchasing parts, the purchaser inputs the name of the purchased part as the top event (e.g., part=belt, event=*) into the input device 14, and the output mixed tree 31 can be used to select the purchased item and identify the inspection items at the time of receipt. Event=* is a wildcard input when no event is specified. The mixed tree 31 includes nodes indicating belt-related failure causes such as wear, slippage, and creep deformation.

[0022] The top event combiner 15 refers to the synonym DB 11 and creates a mathematical expression tree 16 by combining the top event of S12 with a mathematical expression model retrieved from the mathematical expression model DB 12 (S13). The mathematical expression model retrieved in S13 is a mathematical expression model that includes the top event as a node, or a mathematical expression model that includes a synonym of the top event as a node. The model combiner 25 searches the empirical model DB 13 for an empirical model to be connected to the mathematical expression tree 16 (S14). The empirical model searched for in S14 is an empirical model including a node of a mathematical expression model in the mathematical expression tree 16 (e.g., a terminal leaf node that does not include a subordinate node from itself) or a node that has a notation variation relationship with the node of the mathematical expression model.

[0023] The abstraction level converter 21 determines whether there is a notation variation due to a difference in abstraction level between the node of the mathematical formula model searched in S14 and the node of the empirical model (S21). If the result in S21 is Yes, the abstraction level converter 21 refers to the attribute DB 22 and generates an adapter node for converting (absorbing) the abstraction level (S22). The structure converter 23 determines whether there is a notation variation due to a difference in structural expression between the node of the mathematical formula model searched in S14 and the node of the empirical model (S23). If the answer is Yes in S23, the structure converter 23 refers to the product structure DB 24 and generates an adapter node for converting (absorbing) the difference in structural expression (S24).

[0024] The model combiner 25 creates a mixed tree 31 that connects the mathematical expression tree 16 and the empirical model (S25). If an adapter node generated in S22 or S24 exists at the point where the mathematical expression tree 16 and the empirical model are connected in S25, the adapter node is used. The output unit 32 displays the mixed tree 31 of S25 (S31).

[0025] FIG. 3 is a diagram showing the hardware configuration of the tree generating device 100. As shown in FIG. The tree generating device 100 is configured as a computer 900 having a CPU 901 , a RAM 902 , a ROM 903 , a HDD 904 , a communication I / F 905 , an input / output I / F 906 , and a media I / F 907 . The communication I / F 905 is connected to an external communication device 915. The input / output I / F 906 is connected to an input / output device 916. The media I / F 907 reads and writes data from a recording medium 917. Furthermore, the CPU 901 executes a program (also called an application or its abbreviation, an app) loaded into the RAM 902, thereby improving and controlling each processing unit. This program can also be distributed via a communication line, or recorded on a recording medium 917 such as a CD-ROM and distributed.

[0026] FIG. 4 is an explanatory diagram of a fault tree 200 that is manually created from the empirical model DB 13 as a comparative example. The fault tree 200 is configured with the occurred fault event “slip” as the root node, and multiple first layer nodes (e.g., small friction coefficient) that are the direct cause of the “slip” and multiple second layer nodes (e.g., design error, foreign matter contamination) that are the causes of the first layer nodes are connected to it. In addition, multiple lower nodes are connected to one upper node, and this connection relationship may be an "AND relationship" in which an event at the upper node occurs when all of the events at the multiple lower nodes are satisfied, or an "OR relationship" in which an event at the upper node occurs when any of the events at the multiple lower nodes are satisfied, or it may be a mixture of AND and OR relationships.

[0027] A manually constructed fault tree 200 is prone to the following problems because it is constructed from ambiguous knowledge such as human memory. -It is not possible to extract 100% of terminal events (events at the lowest hierarchical nodes). · There are events at different levels within the same hierarchy, making it difficult to check for omissions. There is no logical development, it is not possible to understand causation, and it is difficult to draw out the bottom-line events from experts.

[0028] FIG. 5 is an explanatory diagram showing an example of the mixed tree 31 of FIG. In a fault tree 200, which is an example of the mixed tree 31, a top event node 211 which is a root node, a mathematical model 212, and an empirical model 213 are connected in this order from the top of the tree. The link 212a between the top event node 211 and the first layer node of the mathematical model 212 connects the first layer node (large input torque, small frictional holding torque) obtained by expanding the mathematical expression “input torque Ti>frictional holding torque Th” indicating fastener slippage to the top event node 211 of “fastener slippage” input to the input device 14.

[0029] Based on the inference that the cause of the left-hand side variable shown in the upper node being inappropriate (too large or too small) is the right-hand side variable shown in the lower node being inappropriate, each link in the mathematical model 212 is formed as follows. Link 212b between the first layer node and the second layer node of the mathematical model 212 connects the second layer node (friction coefficient μ small, binding force N small on the right side of the equation) obtained by expanding the equation "friction holding torque Th = friction coefficient μ × binding force N" to the first layer node (friction holding torque Th on the left side of the equation). The link 212c between the second layer node and the third layer node of the mathematical model 212 connects the third layer node (small deformation amount x and small elastic coefficient K on the right side of the equation) obtained by expanding the mathematical equation "binding force N = elastic coefficient K × deformation amount x" to the second layer node (binding force N on the left side of the equation). The empirical model 213 is a node of an end event obtained from the experience of an expert, etc. In this way, there is a high possibility that there will be omissions when only one case is used, but by aggregating and integrating the causal models of multiple cases, it is possible to generate a fault tree 200 that is highly comprehensive as collective intelligence.

[0030] FIG. 6 is an explanatory diagram showing a detailed example of the mixed tree 31 of FIG. In the explanations up to now, for ease of understanding, the event of a component called a DC plug generating heat has been described in a single node such as "DC plug generating heat." In the explanations that follow, the component name (e.g. DC plug) and the event name that occurs in that component (e.g. heat generation) will be separated and explained as a single node [component name, event name]. The fault tree in Figure 6 has the following components connected: Top event node = [DC plug, thermal deformation] · Nodes in the mathematical model = [DC plug, heat generation], [electrode, dead short], [electrode, rare short]. Note that intermediate events such as heat generation include all factors leading up to the top event, including intermediate factors. Empirical model nodes = [insulation site, mechanical bending]. Note that end events such as mechanical bending are factors that lead to intermediate events.

[0031] FIG. 7 is an explanatory diagram showing the process (S13) in which the apex event coupler 15 creates the formula tree 16. The top event coupler 15 sets the top event node 221 input in S12 as the root node, and searches for a mathematical expression model that includes the root node as a node from the mathematical expression model DB 12. As the search result this time, the top event coupler 15 acquires from the mathematical expression model DB 12 two mathematical expression models 222 and 223 that include a node with the same name as the root node [part A, event A]. The top event coupler 15 creates a mathematical expression tree 224 by connecting the mathematical expression model 222 and the mathematical expression model 223 as lower nodes of the top event node 221. Note that a part 224A of the mathematical expression tree 224 that is lower than the root node is a part extracted from the mathematical expression models 222 and 223, and a node of the mathematical expression models 222 and 223 that has the same name as the root node is integrated into the root node.

[0032] Hereinafter, the details of resolving the notational variation between the top event and the mathematical expression model (resolving means 1) will be described with reference to FIG. 8 and FIG. FIG. 8 is an explanatory diagram showing an example of the synonym DB 11. As shown in FIG. The synonym DB11 indicates by a link connection that "part A: slippage" and "part A: excessive speed" are in a synonymous relationship (substitution OR relationship). In the mathematical model in the mathematical model DB 12, an event name is assigned to the excessive / insufficient state of a variable. For example, for the variable "speed", an excessive speed state is called "slip". In the thesaurus DB 11, such paraphrases are registered as prior knowledge.

[0033] Fig. 9 is an explanatory diagram showing an example of the formula tree 16 created using the synonym DB 11 of Fig. 8. The fault tree 230 has the following components connected together. Top event node 231 = [Part A: slippage] node 231A, [Part A: excessive speed] node 231B. Nodes of the first mathematical expression model 232 = [Part A: external force inappropriate] node 232A, [Part A: friction force inappropriate] node 232B, and so on. Nodes of the second formula model 233 = [Part B1: Inappropriate friction coefficient] node 233A, [Part B2: Inappropriate normal force] node 233B, etc.

[0034] Here, the [part A: slippage] node 231A input from the input unit 14 is an expression that is not written in a mathematical formula, and therefore cannot be connected as it is to the [part A: unsuitable external force] node 232A of the first mathematical formula model 232. However, the top event coupler 15 detects that a connection is possible by using a similar word registered in the synonym DB 11 of FIG. Therefore, the top event coupler 15 creates a new [part A: excessive speed] node 231B, which is a synonym for the [part A: slippage] node 231A, as an adapter node for connection. Then, the top event coupler 15 connects the [part A: excessive speed] node 231B, which is defined in advance by a formula, to the [part A: external force unsuitable] node 232A and the [part A: friction force unsuitable] node 232B by inserting an adapter node as a lower node of the [part A: slippage] node 231A. This allows the top event coupler 15 to create the fault tree 230 by inserting an adapter node, even if there is a notation variation between the top event and the formula model.

[0035] In this way, when a combination of the notation of the top event and the notation of the third node in one of the multiple causal models is registered in the attribute DB 22 as a notation variation, the model combiner 25 combines the top event with the causal model by inserting an adapter node for converting the notation variation (in Figure 9, [Part A: excessive speed] node 231B) between the root node of the top event and the third node (in Figure 9, [Part A: inadequate external force] node 232A).

[0036] Hereinafter, with reference to Figs. 10 to 12, a method (resolving means 2) for resolving the notation variations of events (attributes) constituting nodes between the mathematical expression model and the empirical model will be described in detail. FIG. 10 is an explanatory diagram showing an example of the attribute DB 22. As shown in FIG. In the attribute DB22, the event "dust contamination" is previously associated with another attribute "dirt contamination" with a different level of abstraction (here, a higher level of abstraction), and with an even higher level of abstraction, "foreign matter contamination." Also, the event "parts falling" is associated with "foreign matter contamination." The information in the attribute DB22 is referenced to absorb the difference in the level of abstraction between the mathematical model and the empirical model.

[0037] FIG. 11 is an explanatory diagram showing a mathematical model 311 and an empirical model 312 that are different in level of abstraction. The [Part B = power supply: Event B = foreign matter contamination] node in the mathematical model 311 and the [Part C = power supply: Event C = dust contamination] node in the empirical model 312 are the same parts, but there is a variation in the notation of the events, and they cannot be connected as is.

[0038] FIG. 12 is an explanatory diagram showing an example of the process (S22) in which the abstraction level converter 21 creates an adapter node for the data structure shown in FIG. 11 by referring to the attribute DB 22 in FIG. The abstraction level converter 21 acquires from the attribute DB 22 in Fig. 10 that foreign matter contamination and dust contamination are information with different abstraction levels. Then, the abstraction level converter 21 creates an adapter node 313 in which foreign matter contamination is changed to dust contamination, and inserts it between the mathematical model 311 and the empirical model 312. When displaying the mixed tree 31, the output unit 32 highlights the adapter node 313 as shown in FIG. 12, thereby enabling the user to understand that the spelling variations have been automatically resolved.

[0039] Hereinafter, with reference to Figs. 13 to 15, the details of resolving the notation variation of the part names (structures) constituting the nodes between the mathematical expression model and the empirical model (resolving means 3) will be described. FIG. 13 is an explanatory diagram showing an example of the product structure DB 24. As shown in FIG. The product structure DB 24 is a data structure that, when an event such as "loosening" occurs, associates in advance notations of different levels of abstraction with the part name that is the subject of the event. For example, when a flat head screw loosens, information is registered in the product structure DB 24 to convert between different ways of perceiving the subject of the event, such as whether the loosening is an event of a "flat head screw," a more abstract event that applies to "screws" in general, or an even more abstract event of a "fastening part." The information in the product structure DB 24 is referenced to absorb the difference in abstraction between the mathematical model and the empirical model.

[0040] FIG. 14 is an explanatory diagram showing a mathematical model 301 and an empirical model 302 that have different levels of abstraction. The [Part B = fastener: event B = loosening] node in the mathematical model 301 and the [Part C = screw: event C = loosening] node in the empirical model 302 have the same event, but there is a variation in the notation of the parts, and they cannot be connected as they are.

[0041] FIG. 15 is an explanatory diagram showing an example of the process (S24) in which the structure converter 23 creates an adapter node for the data structure shown in FIG. 14 by referring to the product structure DB 24 in FIG. The structure converter 23 acquires from the product structure DB 24 of Fig. 10 that fastening parts and screws are information with different levels of abstraction. Then, the structure converter 23 creates an adapter node 303 in which the fastening part is changed to a screw, and inserts it between the mathematical model 301 and the empirical model 302. When displaying the mixed tree 31, the output unit 32 highlights the adapter node 303 as shown in FIG. 15, thereby enabling the user to understand that the spelling variations have been automatically resolved.

[0042] 10-15, the attribute DB 22 associates combinations of notations with different levels of abstraction for notations indicating occurring events or parts. Then, the model combiner 25 inserts an adapter node (adapter node 303 in FIG. 15) for changing the level of abstraction of the first node as an adapter node between the first node (the [Part B=fastening part: Event B=loosening] node in FIG. 15) and the second node (the [Part C=screw: Event C=loosening] node in FIG. 15).

[0043] FIG. 16 is an explanatory diagram showing an example of a product structure DB 24 different from that of FIG. The product structure DB 24 in FIG. 13 has been described as having a tree structure in which a higher level of abstraction (fastening part) is an upper node and a lower level of abstraction (flat head screw) is a lower node for the same part. On the other hand, the product structure DB 24 in Fig. 16 has a tree structure in which the notation with the widest scope of application is the upper node for the object from which the business originates. For example, the product structure DB 24 starts with the product unit (circuit board) with the widest scope of application, and associates notations with gradually narrower scope of application in the order of unit unit (DC plug) → unit component unit (electrode) → unit part unit (insulating part).

[0044] 16, the attribute DB 22 associates a combination of notations indicating a component where an event has occurred with a combination of notations having different scopes of application of the event. Then, the model combiner 25 inserts an adapter node (e.g., an electrode node) that changes the scope of application of the component of the first node between the first node (e.g., a board node in the mathematical model) and the second node (e.g., an electrode node in the empirical model) as an adapter node. This makes it possible to absorb (convert) differences in how people perceive an event when one user interprets it as "a problem with the electrode" and another as "a problem with the board."

[0045] FIG. 17 is a screen diagram showing a first example of a screen on which the output unit 32 displays the mixed tree 31 (S31). The output unit 32 displays on the screen the mixed tree 400 created from the top event [Substrate: Break]. Here, each node in the mixed tree 400 is also shown with its own score, such as "component = circuit: event = loose [score = 4095]." This score indicates the importance of the node calculated from the occurrence frequency of each cause in the fault tree, and the higher the score, the more important the node.

[0046] The output unit 32 may sort and display the nodes so that nodes with higher scores are displayed at the top. This allows the user to efficiently identify the cause of a problem and identify items to be examined by examining nodes with higher scores first. Alternatively, the model combiner 25 may remove nodes with low scores (nodes with scores lower than a predetermined threshold or nodes with a lower rank than a predetermined rank in the same hierarchy) from the mixed tree 31 (tree branch pruning). This allows nodes with weak causal relationships to be eliminated, making it possible to create a practical tree.

[0047] FIG. 18 is a screen diagram showing a second example of the screen on which the output unit 32 displays the mixed tree 31 (S31). When displaying the mixed tree 31 in which the causal model 413 is developed from the top event 412 [small belt transmission force], the output unit 32 may display it in conjunction with a preregistered function tree 411. For this reason, for the function [power transmission] at the end of the function tree 411, the event [small belt transmission force] that inhibits the function is defined as the top event. This allows the user to understand the cause of the problem from a functional standpoint.

[0048] The tree generating device 100 of the present embodiment described above can link causal models via an adapter node that converts spelling variations, even when the causal models cannot be linked as is because there are spelling variations between causal models from different information sources. For this reason, as a database for previously associating spelling variations between different models, for example, an attribute DB 22 that associates differences in the abstraction levels between mathematical expressions and defect information, and a product structure DB 24 that associates differences in how the creators of defect information perceive the subject in which an event has occurred are referenced. This makes it possible to integrate multiple types of causal models generated from different information sources into a single mixed tree 31, thereby enabling efficient cause analysis (FTA: Fault Tree Analysis) from a wide range of information sources, including malfunctions that have occurred at other companies.

[0049] Furthermore, the present invention is not limited to the above-described embodiments, and various other applications and modifications are possible without departing from the gist of the present invention as described in the claims. For example, the above-described embodiments are detailed and specific descriptions of the configuration of the tree generating device 100 in order to easily explain the present invention, and are not necessarily limited to those including all of the components described. It is also possible to replace a part of the configuration of one embodiment with a component of another embodiment. It is also possible to add a component of another embodiment to the configuration of one embodiment. It is also possible to add, replace, or delete other components to, part of the configuration of each embodiment.

[0050] In addition, the above-mentioned configurations, functions, processing units, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. As the hardware, a broad processor device such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit) may be used. Furthermore, each component of the tree generating device 100 according to the above-described embodiment may be implemented in any hardware as long as the respective hardware can transmit and receive information to each other via a network. Furthermore, the processing executed by a certain processing unit may be realized by a single piece of hardware, or may be realized by distributed processing by multiple pieces of hardware. [Explanation of symbols]

[0051] 11 Synonym DB 13 Empirical Model DB (Model DB) 12 Formula model DB (Model DB) 14 Input Device 15 Apex Event Coupler 16 Formula Trees 20 Tree Generator 21 Abstraction Level Converters 22 Attribute DB (multiple notation DB) 23 Structural Converter 24 Product structure DB (multiple notation DB) 25 Model Coupler 31 Mixed Trees (Causal Trees) 32 Output device 100 Tree Generator

Claims

1. an input unit for receiving an input of a top event to be analyzed; a model combiner that generates a causal tree including a plurality of causal models, the causal tree being the root node of the causal tree based on the top event; an output unit that outputs the causal tree generated by the model combiner to a presentation device; The model combiner comprises: reads, from a model DB, the plurality of causal models generated from mutually different information sources as causal models that connect nodes in which causal relationships exist with links; and referring to a multiple notation DB that associates combinations of notation variations of nodes among the read causal models, when a combination of a notation of a first node in a first causal model and a notation of a second node in a second causal model among the read causal models is registered in the multiple notation DB as the notation variation, an adapter node for converting the notation variation is inserted between the first node and the second node, and a link is formed to connect the first causal model and the second causal model, thereby generating the causal tree. Tree generator.

2. Each node in the causal model is composed of a combination of an event that occurs and the part where that event occurs. The multiple notation DB associates combinations of notations having different levels of abstraction with each other for notations indicating occurring events or components, The model combiner is characterized in that the adapter node, which changes the abstraction level of the first node, is inserted between the first node and the second node.

2. The tree generating apparatus of claim 1.

3. Each node in the causal model is composed of a combination of an event that occurs and the part where that event occurs. The multiple notation DB associates combinations of notations indicating components in which an event has occurred, each of which has a different scope of application of the event, The model combiner is characterized in that the adapter node, which changes the application range of the part of the first node, is inserted between the first node and the second node.

2. The tree generating apparatus of claim 1.

4. When a combination of a notation of the top event and a notation of a third node in any one of the plurality of causal models is registered in the plurality of notation DB as the notation variation, the model combiner combines the top event and the causal model by inserting the adapter node for converting the notation variation between the root node of the top event and the third node.

2. The tree generating apparatus of claim 1.

5. An input device; A model combiner; and an output unit, an input step of accepting an input of a top event to be analyzed using the input device; a generating step of generating a causal tree using the model combiner, the causal tree including a plurality of causal models and the top event being a root node of the causal tree; and outputting the causal tree to a presentation device using the output device; The generating step uses the model combiner to reads, from a model DB, the plurality of causal models generated from mutually different information sources as causal models that connect nodes in which causal relationships exist with links; and referring to a multiple notation DB that associates combinations of notation variations of nodes among the read causal models, when a combination of a notation of a first node in a first causal model and a notation of a second node in a second causal model among the read causal models is registered in the multiple notation DB as the notation variation, an adapter node for converting the notation variation is inserted between the first node and the second node, and a link is formed to connect the first causal model and the second causal model, thereby generating the causal tree. Tree generation methods.

Images