JP2015194831A - Fault phenomenon information analysis device and fault phenomenon information analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a fault phenomenon information analysis device and method capable of referring to the past instance or related information, and expecting operation efficiency improvement.SOLUTION: The fault phenomenon information analysis device includes means including: a function for analyzing a morpheme from input first and second phenomenon texts; a function for extracting first and second phenomenon representations on the basis of a phenomenon representation dictionary from the analyzed morpheme, and for analyzing a relation between the extracted phenomenon representations; a function for creating first and second phenomenon graphs from the relation between the extracted phenomenon representations, and for analyzing the extracted morpheme of a phenomenon text; a function for calculating a distance between the first phenomenon graph and the second phenomenon graph on the basis of similarity between the first and second phenomenon graphs; and a function for, when the calculated distance is equal to or less than a predetermined reference value, presenting a second phenomenon text recorded in a fault phenomenon information recording part and fault phenomenon information including countermeasure information associated with the second phenomenon graph to a user side.

Description

The present invention relates to an apparatus and method for analyzing failure phenomenon information recorded in maintenance work such as network equipment, for example, maintenance work record (log).
More specifically, the present invention relates to a maintenance work log analysis visualization device having a function of visualizing analysis results.

  Equipment maintenance services are indispensable to enable customers to continue their work. And in order to provide customers with quick and accurate maintenance service, if the service base is set up as close to the customer as possible and an abnormality occurs in the equipment and a failure report is received from an alarm device or customer, the cause of the abnormality Therefore, it is required to quickly perform maintenance work such as movement.

Information systems that support such maintenance services are now widely used.
The operator on the maintenance service side accepts calls such as phone calls and emails from customers, listens to explanations from customers about phenomena that occur on equipment, records the contents of the interviews in a database, and contacts customer engineers to instruct countermeasures. To do.
The customer engineer notifies the customer of the countermeasure method based on the contents recorded in the database and the maintenance manual, or prepares a replacement part and visits the site (customer side) to carry out trouble countermeasure work.

  Increasing the efficiency of maintenance services and reducing costs by making these series of business processes more efficient with information systems is a management issue.

  As background art for such a problem, there are JP-A-2004-13594 (Patent Document 1), JP-A-2003-271757 (Patent Document 2), and JP-A-2005-148867 (Patent Document 3).

  Patent Document 1 discloses a method of automating the selection of a failed part by an operator by voice recognition, and further exchanging information such as user information, availability and availability, and estimated arrival time with a customer engineer.

  Further, Patent Document 2 collects support points of a plurality of vendors in the center server for efficiency, and searches for a maintenance service database case pattern using a word indicating a phenomenon / state selected from a list as a key. A method for notifying a user of a coping method corresponding to the case pattern is disclosed.

  Further, Patent Document 3 discloses techniques for supporting maintenance of plant facilities, such as customer authentication, voice input, prediction and notification of spilling abnormalities, and inventory management of replacement parts.

JP 2004-13594 A JP 2003-271757 A JP 2005-148867 A

In an information system that supports maintenance services, the operator who is in charge of maintenance service generally asks the customer about the phenomenon that occurred on the equipment when an abnormality occurs in the equipment, and the response contents of the inquiry as a text database It is only accumulated in.
That is, visualization and analysis of the text contents have not been sufficiently performed, and failure phenomenon information has not been used for analysis of failure tendency of devices.

  In addition, it is difficult for a customer engineer in charge of equipment maintenance work to search for similar cases in the past for cases where multiple phenomena occur in characteristic patterns. It was not made.

Patent Document 1 discloses a method of searching for a failure case from a character string obtained by converting a user's utterance voice into text. However, the detailed method is not disclosed.
In general, it is not obvious how to extract a matching failure case from a sentence expressing a failure phenomenon.

Patent Document 2 discloses a method for allowing an input person to select a failure phenomenon or state from a list. According to this method, the search hit rate with past cases can be increased. However, in this method, the user or the substitute operator has to search the list for items that match the phenomenon, which is not easy.
A method of comprehensively listing all possible phenomena in advance can be considered, but at present, the problem is that the selection cost increases as the number of items increases, and the possibility of search omission increases due to incorrect selection. Yes, practically difficult.

  Patent Document 3 discloses a method of narrowing down the cause of failure by repeating a YES / NO question to a plant manager a plurality of times. However, with this method, the number of questions increases according to the total number of causes of failure, and the administrator must determine for each question whether or not it matches the current phenomenon, which is a heavy burden. In addition, there remains a technical problem on how to create a question to be presented to the administrator.

Many equipment abnormalities are accompanied by some typical occurrence patterns. For example, even if there are two phenomena, phenomenon 1 and phenomenon 2, they occur at the same time, whether they occur alternately, periodically, or by a specific operation. The occurrence pattern such as
However, even if the set of phenomena that occur in different failure cases is the same, the countermeasures are often different for each occurrence pattern of the phenomenon.
Further, even if an attempt is made to search using a phenomenon expression pattern as text as a key, the text expression of the event is optional, and the detection rate is extremely lowered by a simple method such as direct search.

  Although it is conceivable to maintain the detection rate by applying existing natural language processing technologies such as fuzzy search and similar document search, it does not go into the special structure of phenomenon occurrence pattern, and there is a problem in accuracy Conceivable.

As described above, in the conventional system, since it is not possible to easily search for cases focusing on the occurrence pattern, there is a problem that the search accuracy of past case data is lowered.
In addition, the customer engineer finally confirms whether the search result is appropriate, but if the occurrence pattern of the phenomenon is expressed in a sentence, the correctness of the search result cannot be intuitively determined, and it takes time to confirm. There was a problem.
Furthermore, when analyzing past logs based on the occurrence pattern of the phenomenon, there is a problem that it is difficult for an analyst to describe the occurrence pattern or to grasp the past occurrence pattern.

  Therefore, the present invention provides a failure phenomenon information analysis apparatus and method that can refer to past cases and related information and can be expected to improve work efficiency.

  In addition, it is possible to intuitively grasp typical occurrence patterns for phenomena, and feed back the results of analysis of occurrence patterns that occur frequently throughout maintenance operations to the supply chain such as manufacturing, sales, operation, and maintenance. Disclosed is a failure phenomenon information analysis apparatus and method that can lead to maintenance and improvement.

In order to solve the above problem, the present invention has a function of analyzing a morpheme from an inputted first phenomenon text,
A function to extract a phenomenon expression from the analyzed morpheme based on the development expression dictionary and analyze a relationship between the extracted phenomenon expressions;
A function for creating a first phenomenon graph from the relationship between the extracted phenomenon expressions, and a failure development information recording unit that records failure phenomenon information including the inputted second phenomenon text are searched, and the second phenomenon text is searched. The ability to extract,
A function of analyzing the extracted morpheme of the second phenomenon text,
A function of extracting a phenomenon expression from the analyzed morpheme based on a development expression dictionary, a function of creating a second development graph from the extracted search development text, and a failure phenomenon including the second phenomenon text and the second phenomenon graph The ability to manage information,
A function for calculating the distance between the first phenomenon graph and the second development graph based on the similarity between the first and second phenomenon graphs, and the distance obtained by the calculation is a predetermined reference value. A function for presenting to the user the failure phenomenon information including the second phenomenon text recorded in the failure phenomenon information recording unit and the countermeasure information associated with the second phenomenon graph when the value is equal to or less than the threshold; Have.

For example, the failure phenomenon information analyzer of the present invention is
Means for maintaining a phenomenon expression dictionary;
Means for extracting a phenomenon expression defined in the phenomenon expression dictionary from a first failure phenomenon text input by a user;
A relationship extracting means for extracting a dependency relationship between the phenomenon expressions from the first failure phenomenon text as a relationship between the phenomenon expressions;
Means for creating a first phenomenon graph in which the phenomenon expression is a node and the relationship between the phenomenon expressions is an arc;
Means for managing a failure phenomenon information table including one or more records including a second failure phenomenon text in the field;
Means for creating one or more second phenomenon graphs from one or more second failure phenomenon texts by means similar to the first phenomenon graph and the first phenomenon graph creation;
Inter-score score calculating means for calculating each distance between the first phenomenon graph and one or more second phenomenon graphs based on the similarity of the graphs;
And means for presenting the contents of the record of the failure phenomenon information table associated with the second phenomenon graph whose calculated distance is equal to or less than a predetermined threshold to the user or the second user. To do.

Further, the recording means of the failure phenomenon information analysis apparatus of the present invention further includes:
A recording unit that records a relation expression dictionary defining a relationship between the phenomenon expressions;
The first and second phenomenon expression relationship extraction units of the arithmetic device further include:
Labeling the relation between the phenomenon expressions with a specific relation name obtained by referring to the relation expression dictionary from the conjunction expression included between the dependency relations between the phenomenon expressions;
The inter-graph score calculation unit further calculates the inter-graph score in consideration of a label attached to the relationship between the phenomenon expressions.

Further, the failure phenomenon information presenting means of the failure phenomenon information analyzing apparatus of the present invention further comprises:
It has a means for presenting the phenomenon first graph and the second graph to the user or the second user via the output device.

The failure phenomenon information analysis apparatus of the present invention further includes:
Means for managing a failure phenomenon information table including at least one record including the failure phenomenon text and a countermeasure method thereof;
Means for creating one or more phenomenon graphs from the one or more failure phenomenon texts in the same manner as the first and second phenomenon graphs;
A table for calculating the distance between one or more of the phenomenon graphs based on the similarity and recording fault phenomenon information of the recording means associated with the phenomenon graph for each of the distances that are equal to or less than a predetermined threshold A means for obtaining a statistical value of an arbitrary column of the record is included.

The arithmetic unit of the failure phenomenon information analysis apparatus of the present invention further includes:
An extraction unit for extracting a phenomenon modification clause for modifying the phenomenon expression from the failure phenomenon information;
The means for creating the first and second phenomenon graphs is a means for creating a phenomenon graph in which the phenomenon expression and the phenomenon modification clause are nodes and the relationship between the phenomenon expressions is an arc.
It is characterized by that.

Moreover, the failure phenomenon information analyzer of the present invention is
A failure in which an input device, an output device, an arithmetic device, and a storage device are connected to each other, a failure phenomenon text input from the input device is received, and a phenomenon graph created based on the text is output and displayed on the output device A phenomenon information analysis system,
The storage device
Means for holding a phenomenon expression dictionary in which a morpheme and a phenomenon expression of an obstacle phenomenon text input from the input device are associated; means for storing failure phenomenon information including a second obstacle phenomenon text;
The arithmetic unit is:
Management means for managing means for storing fault phenomenon information, phenomenon creating means for creating a phenomenon graph, distance calculating means for calculating the distance between the phenomenon graphs,
The phenomenon graph creating means includes:
Receiving the first failure phenomenon text input from the input means, replacing the plurality of morphemes described in the text with a plurality of first expression phenomena based on the phenomenon expression dictionary, and extracting the expression phenomenon Expression development extraction unit,
An inter-phenomenon relationship extraction unit for extracting a dependency relationship between the plurality of first expression phenomena;
A phenomenon graph creating unit for creating a first phenomenon graph in which the first expression phenomenon is a node and a dependency relationship between the first expression phenomena is an arc connecting nodes of the expression phenomenon;
The second failure phenomenon of the means for storing the failure phenomenon information is extracted, the failure phenomenon is defined as a node, and the dependency relationship between the second expression phenomena is an arc connecting the nodes of the expression phenomenon. A second phenomenon graph creation unit for creating a second phenomenon graph and storing the failure information as a second phenomenon graph in the means for storing the failure information;
The distance calculation means includes
Based on the first phenomenon graph and the second graph, a calculation unit that calculates each distance between the two based on the similarity between the graphs,
When the calculated distance is equal to or less than a predetermined threshold, a presentation unit that outputs failure information of the means for storing the failure phenomenon information associated with the second phenomenon graph to the output device side,
It is characterized by having.

Moreover, the failure phenomenon information analysis method of the present invention includes:
A phenomenon expression dictionary, a failure phenomenon information database management program, and a search program are stored, and failure phenomenon information input from an input device is analyzed using an arithmetic device that operates based on the phenomenon expression dictionary, the failure phenomenon information database management program, and the search program. A failure phenomenon information analysis method,
Extracting a phenomenon expression defined in the phenomenon expression dictionary from a first failure phenomenon text input by a user when receiving a failure phenomenon information database control command;
A relationship extracting means for extracting a dependency relationship between the phenomenon expressions from the first failure phenomenon text as a relationship between the phenomenon expressions;
Creating a first phenomenon graph in which the phenomenon expression is a node and the relationship between the phenomenon expressions is an arc;
Managing a failure phenomenon information table including one or more records including a second failure phenomenon text in the field;
Creating one or more second phenomenon graphs from the one or more second failure phenomenon texts by means similar to the first phenomenon graph;
A graph-to-graph score calculation step of calculating each distance between the first phenomenon graph and the one or more second phenomenon graphs based on the similarity of the graphs;
Including the step of presenting to the user or the second user the contents of the record of the failure phenomenon information table associated with the second phenomenon graph whose calculated distance is equal to or less than a predetermined threshold value. To do.

The failure phenomenon information analysis method of the present invention includes:
Means for managing a failure phenomenon information table including one or more records including failure phenomenon texts and countermeasures;
Means for similarly creating one or more phenomenon graphs from one or more of the failure phenomenon texts;
A method for analyzing the failure phenomenon table in a failure phenomenon information analyzer,
When receiving an analysis instruction, the distance between one or more of the phenomenon graphs is calculated based on the similarity, and the failure phenomenon information associated with the phenomenon graph for each of the distances that are equal to or less than a predetermined threshold The method includes a step of obtaining a statistical value of an arbitrary column of the record of the table.

The failure phenomenon information analysis method of the present invention includes:
A failure in which an input device, an output device, an arithmetic device, and a storage device are connected to each other, a failure phenomenon text input from the input device is received, and a phenomenon graph created based on the text is output and displayed on the output device A phenomenon information analysis system,
The storage device
Means for holding a phenomenon expression dictionary in which a morpheme and a phenomenon expression of an obstacle phenomenon text input from the input device are associated; means for storing failure phenomenon information including a second obstacle phenomenon text;
The arithmetic unit is:
When the failure phenomenon information database control command is received, the first failure phenomenon text input from the input means is received, and a plurality of morphemes described in the text are converted into a plurality of first expressions based on the phenomenon expression dictionary. Replacing the phenomenon and extracting the expression phenomenon,
Extracting a dependency relationship between the plurality of first expression phenomena;
Creating a first phenomenon graph in which the first expression phenomenon is a node and the dependency relationship between the first expression phenomena is an arc connecting nodes of the expression phenomenon;
The second failure phenomenon of the means for storing the failure phenomenon information is extracted, the failure phenomenon is defined as a node, and the dependency relationship between the second expression phenomena is an arc connecting the nodes of the expression phenomenon. Creating a second phenomenon graph and storing it as a second phenomenon graph in the means for storing the failure information;
Calculating each distance between the two based on the similarity between the graphs based on the first phenomenon graph and the second graph;
If the calculated distance is less than or equal to a predetermined threshold, outputting the fault information of the means for storing the fault phenomenon information associated with the second phenomenon graph to the output device side;
It is characterized by having.

  According to the present invention, a user, for example, a customer engineer can refer to past cases and related information for a complex phenomenon, so that improvement in work efficiency can be expected.

In addition, a representative generation pattern for a phenomenon can be intuitively grasped by an illustration. Furthermore, the maintenance and improvement of quality can be expected by feeding back the results of analysis of occurrence patterns that occur frequently throughout the entire maintenance work to the supply chain such as manufacturing, sales, operation and maintenance.
Problems, configurations, and effects other than those described above will become apparent from the following description of embodiments.

FIG. 1 is a block diagram showing a configuration example of the failure phenomenon information analysis apparatus of the present invention. FIG. 2 is a diagram illustrating a configuration example of the failure phenomenon information table 31. FIG. 3 is a flowchart for explaining the operation of the failure phenomenon information database management program 21. FIG. 4 is a flowchart for explaining the operation of the phenomenon graph creation routine 24. FIG. 5 is a diagram showing an example of intermediate data created by the phenomenon graph creation routine 24. FIG. 6 is a diagram illustrating an example of the phenomenon expression dictionary 32. FIG. 7 is a flowchart for explaining the operation of the search program 22. FIG. 8 is a flowchart for explaining the operation of the distance calculation routine 25. FIG. 9 is a diagram showing the contents of the phenomenon expression distance matrix 33. FIG. 10 is a diagram illustrating the contents of the inter-phenomenon relationship distance matrix 34. FIG. 11 is a diagram illustrating a specific example of distance calculation. FIG. 12 is a diagram illustrating an output example of search results. FIG. 13 is a flowchart for explaining the operation of the analysis program 23. FIG. 14 is a functional block diagram showing an example of an arithmetic unit constituting the failure phenomenon information analysis apparatus of the present invention.

  Hereinafter, examples will be described with reference to the drawings. Although the present invention is mainly based on failure phenomenon information analysis, in this example, an example including a configuration for presenting (outputting), for example, visualizing (displaying) the analysis result to a user (one or more) will be described. To do.

FIG. 1 is a block diagram showing a configuration example of a failure phenomenon information analysis apparatus according to the present invention.
In FIG. 1, the failure phenomenon information analysis apparatus 1 includes an arithmetic device 11, a storage device 12, an input device 13, and an output device 14.

  The arithmetic device 11 can be a general CPU (Central Processing Unit), the storage device 12 can be a hard disk or DRAM (Dynamic Random Access Memory), the input device 13 can be a keyboard, and the output device 14 can be a display or a printer.

  The present invention is not limited to this configuration. For example, a network input / output device may be provided instead of the input device 13 and the output device 14, and data input / output may be performed via an external device such as a client terminal via the network. Alternatively, some or all of the data in the storage device 12 may be provided in an external device, and data may be read and written via a network.

The storage device 12 stores a failure phenomenon information database management program 21, a search program 22, an analysis program 23, a phenomenon graph creation routine 24, and a distance calculation routine 25 as a program group. The failure phenomenon information database management program 21 incorporates an existing relational database management system. This program may be provided separately.
In addition, the storage device 12 stores a failure phenomenon information table 31, a phenomenon expression dictionary 32, a phenomenon expression distance matrix 33, and a phenomenon expression relation distance matrix 34.
The program group is read into the arithmetic unit 11 and executed. Details thereof will be described later.

The failure phenomenon information table 31 is assumed to be a relational database management system table generally used in this embodiment, but the present invention is not limited to this.
The contents of each program, each table, dictionary, and matrix will be described below.

FIG. 2 is a diagram showing the contents of the failure phenomenon information table 31.
In the figure, the failure phenomenon information table 31 includes, as fields, a phenomenon text 311, a phenomenon graph 312, a countermeasure 313, an occurrence time 314, an occurrence count 315, and the like.

  The phenomenon text 311 is a field indicating text indicating a phenomenon that has occurred in the maintenance work in the past, and is, for example, “error sound after printing error,“ 12A34B56 ”,...

  The phenomenon graph 312 is a field that is automatically generated from the phenomenon text 311 by a method that will be described later. For example, as shown in the figure, “normal” is associated with “paper feed NOT”, “EC: 12A34B56”, and “sound pea”. This is a field for storing the graph expressed as above.

  The countermeasure 313 is a field for storing text, drawings, links, etc. describing the countermeasure against the phenomenon of the phenomenon text, and indicates, for example, “Check, clean, replace,...

  The occurrence time 314 and the occurrence count 315 are fields for storing the occurrence date “2013/9/5,...” And the occurrence count “1,. These fields may be provided as necessary. Further, other fields than those shown may be included.

FIG. 3 is a flowchart for explaining the operation of the failure phenomenon information database management program 21.
In the figure, a failure phenomenon information database management program 21 is a computer database management system based on a relation model, and is executed by the arithmetic unit 11 when the failure phenomenon information analysis apparatus 1 is started, for example. This activation may be performed by the user at an arbitrary timing. Then, it has a function of creating a phenomenon graph and storing the phenomenon graph in the failure phenomenon information table 31.

  First, in step S211, the input device 13 shown in FIG. 1 waits for a user to input a database control command. Here, the database control command is a command corresponding to addition (INSERT), deletion (DELETE), or update (UPDATE) of a record with respect to the failure phenomenon information table 31 in the relational database.

  When the input of the control command is confirmed, in step S212, the command is executed using the existing relational database management system.

  Next, in step S213, whether or not the phenomenon text 311 in the failure phenomenon information table 31 is added or changed is determined by executing the instruction.

  If the determination result shows that there is an additional change (Y), the process proceeds to step S214, and the phenomenon graph creation routine 24 of FIG. Then, a phenomenon graph is created by a phenomenon graph creation routine 24 (see FIG. 4) described later, and the phenomenon creation graph is stored in the field of the phenomenon graph 312 of the failure phenomenon information table 31.

  FIG. 4 is a flowchart for explaining the operation of the phenomenon graph creation routine 24. FIG. 5 is a diagram showing an example of intermediate data created by the phenomenon graph creation routine 24. FIG. 6 is a diagram showing an example of the phenomenon expression dictionary 32.

Hereinafter, the operation of the phenomenon graph creation routine 24 will be described with reference to these drawings.
4, first, in step S241, morpheme (surface word) analysis and dependency analysis are performed on the phenomenon text (see 311 in FIG. 2) targeted by this routine.

Here, for the morphological analysis and dependency analysis, it is possible to use publicly available software such as JUMAN and KNP, for example. JUMAN is a morphological analysis system, and KNP is a dependency analyzer based on JUMAN (a system that performs syntactic / case / anaphoric analysis of Japanese sentences).
A dictionary necessary for morphological analysis is prepared in the phenomenon expression dictionary 32 shown in FIG. Alternatively, it may be used in combination with an arbitrary dictionary.

  The phenomenon expression dictionary 32 has fields of a word expression 321 and a phenomenon expression 322 as shown in FIG.

In the field of the word notation 321, for example, paper feed, printing error, printing error,... 12A34B56, etc. are described. Also, the phenomenon notation 322 describes paper feed, print error, print error... 12A34B56, etc. corresponding to the word notation 321.
The character string described in the word notation 321 is incorporated into the morpheme analysis software as a user dictionary, and is extracted preferentially as a morpheme.

  Also, the morpheme word notation 321 corresponding to the output phenomenon expression is replaced with the corresponding phenomenon expression (phenomenon expression) 322. As a result, it is possible to correctly analyze a phenomenon text including an expression peculiar to the maintenance work, and to absorb notation fluctuation by an operator.

  For example, as shown in FIG. 5, the results 52a and 52b of this processing are output for the phenomenon texts 51a and 51b. Phenomenon text results 52a and 52b indicate that the dependency relationship of each morpheme string is expressed using ruled line characters, and that what is described in the upper part depends on what is indicated in the lower part.

  That is, in FIG. 5, the development text 51a is “paper not fed error“ 12A34B56 ”is displayed and a beeping sound every 1 second”, and the development text 51b is “error sound after printing error,“ 12A34B56 ””. In this case, the dependency relationship of each morpheme sequence output as the result 52a of this processing is “paper feed ... not displayed, EC: 12A34B56 ... is displayed, at intervals of 1 second ... beep”, processing As a result 52b, “printing error... After ... error sound EC12A34B56.

  Next, in step S242, phenomenon expression extraction is performed. That is, a morpheme that matches the phenomenon notation (phenomenon expression) 322 of the phenomenon expression dictionary 32 is defined as a phenomenon expression.

  Next, in step S243, phenomenon modification clause extraction is performed. A morpheme group applied to the phenomenon expression 322 is defined as a phenomenon modification clause. Also, rule-based semantic analysis is performed.

  For example, as shown in the phenomenon graph 53a of FIG. 5, for the dependency relationship “not fed” in the phenomenon expression 322, a phenomenon modification clause “NOT” is added after “paper feeding”. For the dependency relationship “EC: 12A34B56... Is displayed” in the phenomenon expression 322, a rule that the phenomenon modification clause is omitted is used. Although it is difficult to implement semantic analysis that can handle all expressions, it seems that the number of necessary and effective rules is not so large, so it is desirable to add rules as needed while operating the system .

  Next, in step S244, the relationship between the phenomenon expressions is extracted. For example, as shown in the output result 53a, the parallel relationship of two phenomena representing the dependency relationship (EC: 12A34B56 and paper feed [NOT], EC: 12A34B56 and sound: Pee [1 second interval]) Examine the conjunctions that exist between them.

  In the case of hierarchical relationships, relationships such as logical order / paradox, conditions, time series before / after, etc. are extracted by a dictionary according to the notation of conjunctions existing between them. Because there are not many types of conjunctions, dictionaries can be incorporated into programs.

  Next, in step S245, a graph is created. For example, as shown in FIG. 5, the result phenomenon graphs 53a and 53b of this processing are created for the processing results 52a and 52b.

That is, a phenomenon expression (paper feed [NOT], EC: 12A34B56, sound: bee [1 second interval], EC: 12A34B56 and sound: bee [1 second interval]) is a node, and the relationship between these phenomenon expressions is shown. An effective graph can be created by using arcs that connect nodes representing each phenomenon (EC: 12A34B56 and paper feed [NOT], and EC: 12A34B56 and sound: Pee [1 second interval]).
Further, as indicated by 53a and 53b in FIG. 5, a “normal” node is added, and an “unknown” arc is connected from a “normal” node to a node that is not designated as a transition destination of any arc.
Also, if the transition rule such as A → B is satisfied if A → B and B → C as in the parallel relation, a “parallel” arc of A → C is applied to the arc of A → B and B → C. to add.

  As described above, the failure phenomenon information database management program 21 automatically creates a phenomenon graph corresponding to the addition / change of the failure phenomenon information table 31 while using the phenomenon graph creation routine 24 internally, and both are stored in the failure phenomenon information table 31. Can be stored.

  In the present embodiment, the processing time at the time of searching is reduced by creating a phenomenon graph in advance before searching for a search text (failure phenomenon text) described later, but the present invention is not limited to this configuration. Even if a phenomenon graph of the failure phenomenon information table 31 is created at the time of retrieval, the same result as in this embodiment can be obtained.

FIG. 7 is a flowchart illustrating the operation of the search program 22.
The search program 22 is activated at the timing when the user searches. This activation may be executed when the maintenance work log analysis / visualization apparatus 1 is activated.

  In the figure, first, in step S221, input of a search text (failure phenomenon text) from the input device 13 is awaited.

  When the search text input is confirmed, in step S222, a phenomenon graph is created using the phenomenon graph creation routine 24 (see FIG. 4) already described from the input text.

Next, in step S223, distance calculation is performed using the distance calculation routine 25. The method for calculating the distance will be described later.
Next, in step S224, the search result is output.

FIG. 8 is a flowchart for explaining the operation of the distance calculation routine 25 used in step S223 of FIG.
The distance calculation routine 25 calculates the distance between the two input phenomenon graphs α and β using [Equation 1].

  This distance is defined by a phenomenon expression distance matrix 33 as shown in FIG.

FIG. 9 is a diagram showing the contents of the phenomenon expression distance matrix 33.
In the figure, in the matrix 33, the phenomenon expression of the row item 331 and the column item 332 (none, paper feed, printing error,..., EC: 12A34B56, EC: 12A34B57,...) Is shown, and each value ( 0, 5, 10, 100, etc.) represent the distance between the row and column phenomenon representations.

  This table may be created manually, or the best parameters may be machine-learned from past cases. The method may be executed by well-known means and is not the essence of the present invention.

The above [Equation 1]
R _A (α _Aj , _k , β _{Ax (j)} , _{x (k)} ) is the arc connecting from the jth node to the kth node of the graph α and the x (j) th of the corresponding graph β The distance from the node to the arc connected to the x (k) -th node is shown.

  This distance is defined by a matrix 33 indicating the phenomenon expression distance. For example, in the table of the matrix 33, two phenomenon graphs α and β (see 53a and 53b in FIG. 5) are respectively “none”, “paper feed”, “printing error”,..., “EC: 12A34B”, It has a phenomenon expression (node) such as “EC: 12A34B”, and the distance between the nodes is indicated by “0”, “5”, “100”, and the like.

FIG. 10 shows the contents of the inter-phenomenon expression relationship distance matrix 34.
In the figure, in the table of the relationship matrix 34 between phenomenon expressions, the row item 341 and the column item 342 represent the relationship of the phenomenon expression, and each value (0, 1, 5, 2, etc.) of the table is the row item 341 and the column. The distance between the phenomenon expression relations of the item 342 is represented. This table may be created manually, or the best parameters may be machine-learned from past cases.

  For example, in the table of the matrix 33, two phenomenon graphs α and β (see 53 a and 53 b in FIG. 5) are “None”, “Unknown”, “Parallel”, “ The distance between the arcs is indicated by “0”, “1”, “5”, “2”, and the like.

FIG. 11 is a diagram showing a specific example of distance calculation in the distance calculation routine 25.
In the distance calculation, the distances of the phenomenon graphs 53a and 53b in FIG. 5 are calculated using the phenomenon expression distance matrix 33 and the phenomenon expression relationship distance matrix 34 shown in FIGS.

For example, the distance between nodes (“none”, “paper feed”, “printing error”,..., “EC: 12A34B”, “EC: 12A34B”, etc.), arc (“none”, “unknown”, “ Distance between “parallel”, “after”, etc.) for all pairs, and the resulting sums R _A (0, 10, 0, 10), R _N (0, 0, 1, 2, 0: 0) ) Is the distance.

  As described above, the failure phenomenon information database management program 21 creates a phenomenon graph (first), the search program 22 creates a phenomenon graph (second), and the distance between these phenomenon graphs is represented by the graph. By calculating on the basis of similarity and presenting the result, for example, to the user (customer engineer) side, the user can refer to past cases and related information for multiple phenomena , Work efficiency can be improved.

  FIG. 12 is a diagram for explaining an example in which the result of distance calculation based on the search program 22 is displayed on the screen of the display unit 141.

  In the figure, based on the search program 22, a record with a distance equal to or less than a predetermined threshold value N or a record with the top M shortest distances is extracted, and the search result is output to the display unit 141 of the output device 14. The threshold value N (relative value of the distance between the node and arc) and the top M shortest distances are arbitrarily set by the user.

  In the upper part of the screen of the display unit 141, a phenomenon text input by the user (12A34B56 is displayed without a paper feed, a beep) and a phenomenon graph created therefrom are displayed. In the lower part of the screen, “Phenomenon graph”, “Phenomenon text”, and “Countermeasure” of each item obtained as a search result of the failure phenomenon information table 31 are displayed.

  By displaying the phenomenon graph illustrating the phenomenon text entered by the user and the phenomenon graph of each item in the search result side by side, it becomes easier to select the appropriate item from the items obtained as the search result, and countermeasure information Can be easily obtained.

  As described above, the search program 22 uses the phenomenon graph creation routine 24 and the distance calculation routine 25 to extract the record of the failure phenomenon information table 31 that matches the phenomenon text input by the user, and matches the phenomenon. Appropriate measures can be presented to the user in an easy-to-understand manner.

FIG. 13 is a flowchart illustrating the operation of the analysis program 23.
The analysis program 23 is activated at the timing when the user performs analysis. This activation may be performed when the failure phenomenon information analysis apparatus 1 is activated.

  In the figure, first, in step S231, input of a database command by the user from the input device 13 is awaited. The database command is a general command including the above-described database control command, and includes an analysis command such as SQL aggregate calculation (GROUP BY).

  When the database command input is confirmed in step S232, in step S232, if the command is an analysis command, the distance between the phenomenon graphs of the records in the failure phenomenon information table 31 is calculated according to the content of the analysis command.

  For example, in the case of an instruction for totaling those whose distances are equal to or less than the threshold value P, the distances between the phenomenon graphs of all records are calculated so that clusters of those whose distances are equal to or smaller than the threshold value P can be created.

  Next, in step S234, the instruction is executed. The instruction is executed by sending it as an SQL instruction to a relational database management system (not shown). Here, in the case of an analysis command, the relational database management system can refer to the calculated distance between the phenomenon graphs as necessary. SQL is a database language (query language) for manipulating and defining data in a relational database management system, and is standardized as a standard by ANSI (American National Standards Institute) and ISO (International Organization for Standardization).

  Next, when the phenomenon text is added or changed by instruction execution, a phenomenon graph is created using the phenomenon graph creation routine 24 already described, and stored in the <phenomenon graph> of the failure phenomenon information table 31.

  Thus, the analysis program 23 can analyze the contents of the failure phenomenon information table 31 using the distance of the phenomenon graph while using the phenomenon graph creation routine 24 and the distance calculation routine 25 internally.

  For example, as an example of analysis using the distance between graphs, those having a close distance between graphs are clustered, and the average response time of the same phenomenon can be obtained by obtaining the average response time for each cluster. . In other words, according to such analysis, a distance between one or more phenomenon graphs is calculated based on similarity, and a failure information phenomenon associated with the phenomenon graph for each of the distances that are equal to or less than a predetermined threshold A statistical value of an arbitrary column of the record of the table 31 can be obtained.

  FIG. 14 is a functional block diagram illustrating an example of the arithmetic device 11, and the above-described processing will be described in relation to the input device 13 and the output device 14 using programs and data stored in the storage device 12. It is a functional block diagram.

  In the figure, the arithmetic unit 11 includes a phenomenon expression extraction unit 241, a phenomenon expression relationship extraction unit 242, a phenomenon graph creation unit 243, an inter-graph distance calculation unit 231, an inter-graph distance score calculation unit 251, a created graph presentation unit 221, Have

  The phenomenon expression extracting unit 241 constitutes first and second phenomenon expression extracting means, the inter-phenomenon expression relationship extracting unit 242 constitutes first and second phenomenon expression extracting means, and the phenomenon graph creating unit 243 is the first. The second phenomenon graph creation means, the inter-graph distance calculation unit 231 and the inter-graph distance score calculation unit 251 constitute the first and second inter-phenomenon graph distance / score calculation means, and the created graph presentation unit 221 Constitutes one or more fault phenomenon information presenting means including the second fault phenomenon text.

  The phenomenon expression extraction unit 241 of the first and second phenomenon expression extraction means (1) receives the first failure phenomenon text input from the input device 13, and (2) analyzes the morpheme of the first phenomenon text. (3) A first phenomenon expression defined in the phenomenon expression dictionary of the storage device 12 is extracted from the analyzed morpheme.

  (4) Analyzing the morpheme of the second phenomenon text recorded in the failure phenomenon information recording unit 31 of the storage unit 12, and from the analyzed morpheme, a second defined by the phenomenon expression dictionary of the storage device 12 Extract phenomenon expression.

  The phenomenon expression extraction unit 241 of the first and second phenomenon expression extraction means may have the same configuration or different configurations as in this embodiment.

  The inter-phenomenon relation extraction unit 242 of the first and second inter-phenomenon expression extraction means (3) is a dependency between a plurality of phenomenon expressions in the first phenomenon text extracted by the first phenomenon expression extraction unit 241. The relationship is extracted as a relationship between the first phenomenon expressions.

  Similarly, (5) a dependency relationship between a plurality of phenomenon expressions in the second phenomenon text extracted by the phenomenon expression extracting unit is extracted as a relationship between second phenomenon expressions.

  The phenomenon expression relationship extracting unit 242 of the first and second phenomenon expression relationship extracting means may be the same or different from each other as in this embodiment.

  The first and second phenomenon graph creating means 243 (6) from the first phenomenon expression relationship extracted by the phenomenon expression relationship extracting unit 242, the phenomenon expression is used as a node, and the relationship between the phenomenon expressions is expressed as each phenomenon expression. A first phenomenon graph (phenomenon graph α) is created as an arc connecting the nodes.

  Similarly, (7) from the second phenomenon expression relationship extracted by the phenomenon expression relationship extracting unit 242, the phenomenon expression is a node, and the phenomenon expression relationship is an arc connecting nodes of each phenomenon expression. A second phenomenon graph is created.

  The intergraph distance calculator 231 of the first and second intergraph graph distance / score calculating means calculates the similarity between the first and second phenomenon graphs by the intergraph distance score calculator 251 as a score. The inter-graph distance calculation unit 231 calculates the distance between the first and second phenomenon graphs based on the inter-graph distance score by the inter-graph distance score calculation unit 251.

  The creation phenomenon graph presentation unit 221 of the failure phenomenon information presentation unit presents (1) one or more pieces of failure phenomenon information including the second phenomenon text of the failure information.

  When the distance calculated by the inter-graph distance score calculation unit 231 is equal to or less than a threshold value that is a predetermined reference value, the failure phenomenon information presentation unit is recorded in the table of the failure phenomenon information recording unit of the storage device 12, and the second The second failure content (second development text, second phenomenon graph, countermeasure, occurrence time, number of occurrences, etc.) recorded (recorded) in association with this phenomenon graph is extracted and displayed on the output device 14 Displayed on the unit 141.

  The storage device 12 includes a recording unit that stores the phenomenon expression dictionary and the relation expression dictionary described above, a program recording unit that stores each program, a recording unit that records fault phenomenon information, and a recording unit that records a distance matrix.

According to the embodiment described above, when a user inputs a phenomenon as text, the user can instantaneously refer to a record in the failure phenomenon information database having a similar phenomenon occurrence pattern.
Furthermore, by displaying the first phenomenon graph and the second phenomenon graph on the display unit of the output device and presenting them to the user or the second user, the user inputs the text inputted as the phenomenon and the search result as The occurrence pattern of each phenomenon of the obtained fault information can be confirmed with a graph, and it can be easily confirmed whether or not the search result is valid.

  In addition, a fault phenomenon information database including at least one record including a fault phenomenon text and a countermeasure method thereof is managed, and one or more phenomenon graphs are similarly created from the one or more fault phenomenon texts. The distance between the phenomenon graphs is calculated based on the similarity, and a statistical value of an arbitrary column of the record of the failure phenomenon information database associated with the second phenomenon graph for each of the distances that are equal to or less than a predetermined threshold is calculated. As a result, the user can perform an analysis for obtaining statistical information for each occurrence pattern of the phenomenon on the failure phenomenon information database.

In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described.
Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment.
Each of the above-described configurations, functions, processing units, processing means, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit.
In addition, each of the above-described configurations, functions, and the like may be realized by software by the arithmetic device interpreting and executing a program that realizes each function. Information such as programs and tables for realizing each function can be stored in a recording device such as a memory or a hard disk.
Further, the control lines and information lines indicate what is considered necessary for the explanation, and not all the control lines and information lines on the product are necessarily shown. Actually, it may be considered that almost all the components are connected to each other.

  In the field where the phenomena and countermeasures of failures that occur in maintenance work are used as a case database, search for phenomena described in text, visualize and understand phenomena, and statistical analysis based on the phenomenon structure In the case of performing maintenance, it is possible to give more accurate results, so that it is possible to promote quality improvement and cost reduction of maintenance work.

1 Failure phenomenon information analyzer (maintenance work log analysis and visualization device)
11 arithmetic device 12 storage device 13 input device 14 output device 21 fault phenomenon information database management program 22 search program 23 analysis program 24 phenomenon graph creation routine 25 distance calculation routine 31 fault phenomenon information table 32 phenomenon expression dictionary 33 phenomenon expression distance matrix 34 phenomenon Representation relationship distance matrix

Claims (9)

Means for maintaining a phenomenon expression dictionary;
Means for extracting a phenomenon expression defined in the phenomenon expression dictionary from a first failure phenomenon text input by a user;
A relationship extracting means for extracting a dependency relationship between the phenomenon expressions from the first failure phenomenon text as a relationship between the phenomenon expressions;
Means for creating a first phenomenon graph in which the phenomenon expression is a node and the relationship between the phenomenon expressions is an arc;
Means for managing a failure phenomenon information table including one or more records including a second failure phenomenon text in the field;
Means for creating one or more second phenomenon graphs from one or more of the second failure phenomenon texts by means similar to the first phenomenon graph;
Inter-score score calculating means for calculating each distance between the first phenomenon graph and one or more second phenomenon graphs based on the similarity of the graphs;
And means for presenting the contents of the record of the failure phenomenon information table associated with the second phenomenon graph whose calculated distance is equal to or less than a predetermined threshold to the user or the second user. Failure phenomenon information analyzer. The apparatus according to claim 1, further comprising means for holding a relational expression dictionary,
The relation extraction means further labels the relation between the phenomenon expressions with a specific relation name obtained by referring to the relation expression dictionary from a conjunction expression included between the dependency relations between the phenomenon expressions,
The inter-graph score calculation means further calculates in consideration of a label attached to the relationship between the phenomenon expressions. The apparatus of claim 1, comprising:
The failure phenomenon information analysis apparatus further comprises means for presenting the first graph and the second graph to the user or the second user. Means for managing a failure phenomenon information table including one or more records including failure phenomenon texts and countermeasures;
Means for similarly creating one or more phenomenon graphs from one or more of the failure phenomenon texts;
A distance between one or more of the phenomenon graphs is calculated based on the similarity, and an arbitrary column of a record of the failure phenomenon information table associated with the phenomenon graph is associated with each other whose distance is equal to or less than a predetermined threshold. A failure phenomenon information analysis apparatus comprising means for obtaining a statistical value. An apparatus according to claim 1 to claim 4, wherein
Means for extracting a phenomenon modification clause for modifying the phenomenon expression from the obstacle phenomenon information,
The failure phenomenon information analysis apparatus characterized in that the means for creating the phenomenon graph is a means for creating a phenomenon graph in which a phenomenon expression and a phenomenon modification clause are nodes and the relation between the phenomenon expressions is an arc. A failure in which an input device, an output device, an arithmetic device, and a storage device are connected to each other, a failure phenomenon text input from the input device is received, and a phenomenon graph created based on the text is output and displayed on the output device A phenomenon information analysis system,
The storage device
Means for holding a phenomenon expression dictionary in which a morpheme and a phenomenon expression of an obstacle phenomenon text input from the input device are associated; means for storing failure phenomenon information including a second obstacle phenomenon text;
The arithmetic unit is:
Management means for managing means for storing fault phenomenon information, phenomenon creating means for creating a phenomenon graph, distance calculating means for calculating the distance between the phenomenon graphs,
The phenomenon graph creating means includes:
Receiving the first obstacle phenomenon text input from the input means, replacing a plurality of morphemes described in the text with a plurality of first expression phenomena based on the phenomenon expression dictionary, and extracting the expression phenomenon Expression development extraction unit,
An inter-phenomenon relationship extraction unit for extracting a dependency relationship between the plurality of first expression phenomena;
A phenomenon graph creating unit for creating a first phenomenon graph in which the first expression phenomenon is a node and a dependency relationship between the first expression phenomena is an arc connecting nodes of the expression phenomenon;
The second failure phenomenon of the means for storing the failure phenomenon information is extracted, the failure phenomenon is defined as a node, and the dependency relationship between the second expression phenomena is an arc connecting the nodes of the expression phenomenon. A second phenomenon graph creation unit for creating a second phenomenon graph and storing the failure information as a second phenomenon graph in the means for storing the failure information;
The distance calculation means includes
Based on the first phenomenon graph and the second graph, a calculation unit that calculates each distance between the two based on the similarity between the graphs,
When the calculated distance is equal to or less than a predetermined threshold, a presentation unit that outputs failure information of the means for storing the failure phenomenon information associated with the second phenomenon graph to the output device side,
A failure phenomenon information analysis apparatus characterized by comprising: A phenomenon expression dictionary, a failure phenomenon information database management program, and a search program are stored, and failure phenomenon information input from an input device is analyzed using an arithmetic device that operates based on the phenomenon expression dictionary, the failure phenomenon information database management program, and the search program. A failure phenomenon information analysis method,
Extracting a phenomenon expression defined in the phenomenon expression dictionary from a first failure phenomenon text input by a user when receiving a failure phenomenon information database control command;
A relationship extracting means for extracting a dependency relationship between the phenomenon expressions from the first failure phenomenon text as a relationship between the phenomenon expressions;
Creating a first phenomenon graph in which the phenomenon expression is a node and the relationship between the phenomenon expressions is an arc;
Managing a failure phenomenon information table including one or more records including a second failure phenomenon text in the field;
Creating one or more second phenomenon graphs from the one or more second failure phenomenon texts by means similar to the first phenomenon graph;
A graph-to-graph score calculation step of calculating each distance between the first phenomenon graph and the one or more second phenomenon graphs based on the similarity of the graphs;
Including the step of presenting to the user or the second user the contents of the record of the failure phenomenon information table associated with the second phenomenon graph whose calculated distance is equal to or less than a predetermined threshold value. To analyze failure phenomenon information. Means for managing a failure phenomenon information table including one or more records including failure phenomenon texts and countermeasures;
Means for similarly creating one or more phenomenon graphs from one or more of the failure phenomenon texts;
A method for analyzing the failure phenomenon table in a failure phenomenon information analyzer,
When receiving an analysis instruction, the distance between one or more of the phenomenon graphs is calculated based on the similarity, and the failure phenomenon information associated with the phenomenon graph for each of the distances that are equal to or less than a predetermined threshold A failure phenomenon information analysis method comprising the step of obtaining a statistical value of an arbitrary column of a table record. A failure in which an input device, an output device, an arithmetic device, and a storage device are connected to each other, a failure phenomenon text input from the input device is received, and a phenomenon graph created based on the text is output and displayed on the output device A phenomenon information analysis system,
The storage device
Means for holding a phenomenon expression dictionary in which a morpheme and a phenomenon expression of an obstacle phenomenon text input from the input device are associated; means for storing failure phenomenon information including a second obstacle phenomenon text;
The arithmetic unit is:
When the failure phenomenon information database control command is received, the first failure phenomenon text input from the input means is received, and a plurality of morphemes described in the text are converted into a plurality of first expressions based on the phenomenon expression dictionary. Replacing the phenomenon and extracting the expression phenomenon,
Extracting a dependency relationship between the plurality of first expression phenomena;
Creating a first phenomenon graph in which the first expression phenomenon is a node and the dependency relationship between the first expression phenomena is an arc connecting nodes of the expression phenomenon;
The second failure phenomenon of the means for storing the failure phenomenon information is extracted, the failure phenomenon is defined as a node, and the dependency relationship between the second expression phenomena is an arc connecting the nodes of the expression phenomenon. Creating a second phenomenon graph and storing it as a second phenomenon graph in the means for storing the failure information;
Calculating each distance between both based on the similarity between the graphs based on the first phenomenon graph and the second graph;
If the calculated distance is less than or equal to a predetermined threshold, outputting the fault information of the means for storing the fault phenomenon information associated with the second phenomenon graph to the output device side;
A failure phenomenon information analysis method characterized by comprising:

Images