JPH1165651A - Preventive maintenance device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily construct a preventive maintenance system by providing an information retrieval means for making an access via a network to technical information, library, and publications which have been made into electronic form, and converting information obtained by the information retrieval means into ruled knowledge. SOLUTION: Knowledge related to the object of preventive maintenance is retrieved from technical information made into electronic form inside or outside an office by an information retrieving means 10. This obtained text is converted into an 'IF THEN rule' by a knowledge converting means 20. Then, an inference engine 42 of an inferring means 40 detects abnormality based on a rule stored in the knowledge base 41, and presumes the cause of the abnormality. The inferred result and the inferring process are displayed on a motoring screen by an information displaying means 30, and are recorded at the same time as the history of diagnosis in a real machine data base. Also, an information display means 30 displays the text obtained by the information retrieving means 10 and the rule converted by the knowledge converting means 20 on the monitoring screen. Thus, an operator can confirm the displayed technical information, rule, inferred result, and inferring process, and evaluate the validity.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a preventive maintenance apparatus for measuring or observing the state of equipment such as a plant, a power generation facility, and a communication facility, detecting an abnormality of a target equipment based on the result, and diagnosing the cause. About.

[0002]

2. Description of the Related Art As conventional techniques, an expert system (ES), a fault tree analysis (FTA) and the like are known. ES is a method in which the knowledge of an expert regarding an abnormality detection method and an abnormality diagnosis method is previously set as a rule, and diagnosis is performed based on the ruled knowledge. Therefore, a knowledge engineer who specializes in making rules into knowledge and an expert who is familiar with the subject repeatedly interview to build a knowledge base. Similarly, FTA
A knowledge engineer organizes the knowledge of the subject-specific experts and builds a knowledge base that branches in a tree shape.

[0003] Japanese Patent Application Laid-Open No. 1-2229331 discloses a knowledge-type information processing means for inferring a countermeasure for a state in which a target device is detected using knowledge stored in a knowledge base.

[0004]

Such a preventive maintenance system requires a great deal of effort to build a knowledge base. In addition, since it is necessary to formulate knowledge specific to the target, if the target of the preventive maintenance is different, it is necessary to newly construct a preventive maintenance system.

[0005] Further, depending on the target, there are few specialists who have knowledge on preventive maintenance, or it is difficult to obtain specialized knowledge for some reason. In such a case, the preventive maintenance system does not function sufficiently because a sufficient knowledge base cannot be constructed.

[0006]

In order to solve the above-mentioned problems, a preventive maintenance device according to the present invention uses computerized technical information,
It has information searching means that can access books and publications via a network, and has knowledge converting means for converting the information obtained by the information searching means into rule-based knowledge.

According to a first aspect of the present invention, there is provided a knowledge conversion unit for converting information relating to an abnormality of an equipment obtained by the information search unit into an abnormality detection rule, an abnormality cause estimation rule, or both of them, and the obtained rule. And an inference means for estimating at least one of the abnormality of the equipment and the cause of the abnormality based on the measurement data.

[0008] A second invention is characterized by having display means for displaying one or both of the information obtained by the information search means and the rules obtained by the knowledge conversion means.

According to a third aspect of the present invention, there is provided a preventive maintenance accuracy evaluation means for evaluating diagnosis accuracy based on a diagnosis result obtained by the inference means and an actual state of a device, and a new information search means when the diagnosis accuracy is low. It is characterized by having rule changing means for obtaining information and changing rules in the knowledge base according to the obtained information.

A fourth invention is characterized in that the requirements of the first to third inventions are provided in a diesel engine power generation system, and abnormality is detected or diagnosed.

A fifth aspect of the present invention is a knowledge base constructing apparatus having information search means, knowledge conversion means, and display means for displaying information or knowledge obtained by them.

[0012]

FIG. 1 shows a first embodiment of the present invention.

In this embodiment, an abnormality of an emergency diesel engine generator is detected, and when an abnormality occurs, its cause is estimated. The present invention provides an information retrieval unit 10, a knowledge conversion unit 2,
0, information display means 30, and inference means 40. The inference means 40 includes a knowledge base 41 and an inference engine 42.

The information retrieving means 10 is a means for retrieving knowledge on the subject of preventive maintenance from electronic technical information inside and outside the company. The in-house site contains technical reports, technical specifications for products, technical manuals, and actual machine data and maintenance and inspection data for emergency diesel engine generators. External sites include electronic publications, electronic libraries, electronic technical information centers, and the like. In order to obtain information from such internal and external sites, enter keywords related to the symptoms or causes of abnormalities, and obtain sentences that include the keywords from the digitized technical information and can be used as rules I do. In the knowledge conversion means 20, the sentence obtained by the information search means 10 is referred to as "IF THE
N rules ". At this time, since the rules are corrected while referring to the information of the actual device database, rules suitable for the actual device can be created. The inference means 40 includes a knowledge base 41
The inference engine 42 stores the "IF THEN rule" created by the above means or the "IF THEN rule" directly input by an engineer. The inference engine 42 detects an abnormality based on the rules stored in the knowledge base 41,
Estimate the cause of the abnormality. The result of the inference and the inference process are displayed on the monitor screen by the information display means 30, and are simultaneously recorded as a diagnosis history in the actual machine database. From the inference result and the inference process, a keyword to be input to the information retrieval means 10 can be selected. The information display means 30
Can display, on a monitor screen, a sentence obtained by the information search means 10 and a rule converted by the knowledge conversion means 20. Therefore, the worker can check the displayed technical information, rules, inference results, and inference process, and can evaluate the validity thereof.

Next, the above example will be described in more detail.

FIG. 2 shows details of the information retrieval means 10. The information search means 10 includes a search engine 11, a keyword input unit 12, a keyword database 13, a search history management unit 14, a search format unit 15, and another search engine interface unit 16.

The keyword input section is a section for inputting a keyword for searching for information. The keyword is a free keyword, and a Japanese or English word can be input. Alternatively, a keyword registered in advance in the keyword database may be referred to and selected and input. In addition, in the event of an abnormality, the inference engine 42 displays the inference result and the inference process on the information display means 30, and a part thereof can be used as a keyword.

The search engine 11 can search for technical information arranged hierarchically. That is, in the same way as when searching for books in an actual library, a library (site) is searched based on a keyword, a bookshelf (field) in a related field is searched, and a book (technical information) is searched. If a book (technical information) is found, the information is searched for from the sentences and figures in the contents. Also in this case, if there is a table of contents, a chapter related to a keyword is specified in advance so that information is found from an actual book, and only sentences included in the chapter are searched. Alternatively, if there is an index, the number of pages is specified from the index, and only the sentences included in the page are searched. By performing a hierarchical search in this way, it is possible to reduce the search time. Such a search can proceed automatically, and the search ends when all the accessible sites registered in advance have been searched. However, past search histories are stored in the search history management unit 14, and when searching with the same keyword, sites whose contents have not been updated since the previous search are excluded from the search range. Further, it is possible for an operator to perform a search while manually determining a search range. Further, by inputting some keywords and taking an AND condition or an OR condition, it is possible to search efficiently.

FIG. 3 shows the details of the knowledge conversion means 20. The knowledge conversion means 20 includes a conversion database 21, a conversion engine 22, and a knowledge debugger 23. The conversion database 21 stores a keyword for confirming that the retrieved information can be converted into a rule, a keyword used for conversion to a rule, and a rule for conversion to a rule. As an example of the information registered in the conversion database 21, there are the following keywords.

1) Words indicating that the sentence is related to preventive maintenance "abnormal""normal""failure""cause""inspection""standard"
"Check" etc. 2) Words indicating conditions "-If", "-If", "-If", "-When", etc. 3) Words related to quantification "-times", "-%", "within" 4) Words indicating measured values and their characteristic amounts, such as "engine speed", "generator voltage", "field voltage", "minimum value", and "rated value". Also stores information on the table format used to obtain the information summarized in the table format. For example, if the title is "Table 1 Inspection items of generator and cause of abnormality", it is described as "Two columns of table, the first column describes inspection items of generator, and the second column describes the cause of abnormality". Such a rule is registered in the conversion database 21. The use of information in which rules are determined in advance in this way makes it very easy to convert rules.

In the conversion engine 22, the conversion database 2
The information obtained based on the information of No. 1 is converted into a rule.
Since the information retrieval means 10 also retrieves information in the actual machine database, it is possible to create rules using the latest actual machine information.

As an example of creating a rule by the conversion engine 22, an operation for converting a sentence "AVR is abnormal when the maximum value of the generator voltage is greater than 1.5 times the rated value" is converted into a rule. explain.

Since there is a word indicating the condition of “-case”, the part written before the word is the antecedent part (“IF
~ "), And the part after it is the consequent (" THE
N) ”). That is, Rule 1: “The maximum value of the IF generator voltage is 1.5
THENAVR greater than twice is abnormal. "
It is possible to make inferences using the rules described in the language, but in order to express the "maximum value of the generator voltage" and "1.5 times the rated value" with symbols and numerical values, they are represented by "Vmax" and "Vr", respectively.
X1.5 "rule 1.2:" THEN AVR with IF Vmax greater than Vr x 1.5 is abnormal. " Further, when “greater than” is converted to “>”, rule 1.3: “IF (Vmax> Vr × 1.5) THE
"N AVR is abnormal." If the symbol "Vmax" described as a rule matches the symbol "Vmax" to which a numerical value is substituted, the determination becomes very easy. in this way,
When converting a sentence into rules, it can be converted into rules in several stages. Therefore, in this example, a rule format suitable for the inference engine 42 is determined in advance, and converted in accordance with the format. Also,
When converting tabular information into rules, conversion rules are registered in the conversion database as described above, and rules are created according to the rules.

The knowledge ruled by the conversion engine 22 is input to the knowledge debugger 23. Knowledge Debugger 23
Then, it is checked whether there is any contradiction between the already constructed knowledge base 41 and the newly added knowledge. If the created rule does not contradict the already constructed knowledge base 41, it is stored in the knowledge base 41. If there is any inconsistency in the rules, the created rule and the rules in the knowledge base 41 that are inconsistent with the rule are displayed by the information display means 30 without being added to the knowledge base 41. For example, in the knowledge base 41, Rule 2: "IF The maximum value of the generator voltage is 1.
8 times or less, THENAVR is normal. " Therefore, in such a case, the information is not stored in the knowledge base 41 and the information display means 3 is not stored.
0 indicates "contradicted" on the monitor screen, and at the same time, conflicting rules 1 and 2 are displayed. If there is no contradiction, it is added to the knowledge base 41 as a new rule.

When the knowledge conversion means 20 obtains a text that is difficult to convert or a chart data recorded as image data, the obtained information is displayed on a monitor screen without being converted into rules. At that time, the message "Conversion difficult" or "Conversion impossible" is also displayed.

Next, the information display means 30 will be described. As described above, the information display means 30 automatically displays information when it cannot be converted into a rule or when the converted rule contradicts an existing rule. Even in cases other than such cases, information and rules specified by the operator can be arbitrarily displayed, and if necessary, sentences before and after the search can be displayed at the same time. Therefore, even if it is information that cannot be converted into a rule only by the retrieved text, it may be possible to convert the information into a rule by showing the preceding and following text to the operator. Further, the rules converted by the knowledge conversion means 20 can be displayed, and the validity can be verified by comparing with the text before conversion.

In the example described above, the information display means 30
Although an example in which information is searched for using the above is shown, when there is an information search means unique to an information site, information may be acquired by effectively utilizing the search means. However, when using the site-specific information search means, it is necessary to input a keyword and obtain the searched information for each site. In this case, an external search engine is used through the other search engine interface unit 16. The other search engine interface unit 16 provides an interface for effectively using another search engine. That is, a site that uses a unique search engine is registered in the other search engine interface unit 16, and an automatic processing program for inputting a keyword of the registered site and obtaining information is registered. Therefore, an operator can automatically search for information on a number of sites by only once inputting a keyword without being aware of other search engines.

FIG. 4 shows a second embodiment of the present invention.

This embodiment has a preventive maintenance accuracy evaluation means 50 and a rule changing means 60 in addition to the example shown in FIG. In this example, the diagnostic accuracy is self-evaluated and spontaneously learned by these means to improve the preventive maintenance accuracy.

FIG. 5 shows the algorithm of this embodiment. The preventive maintenance accuracy evaluation means 50 evaluates the preventive maintenance accuracy by comparing the determination result recorded in the actual machine database with the actual state of the apparatus via the information search means. If there is a misdiagnosis, pick up the misdiagnosed diagnosis process and rules used for the diagnosis. Here, this rule is assumed to be rule A.
Since the cause of the misdiagnosis is rule A, the keyword related to rule A is transmitted to the information search means 10, and the search history management unit 14 determines whether new information can be obtained. If new information is available, search the information and create a new rule. When a new rule (rule B) is created from a keyword related to rule A, rule A is first deleted from knowledge base 41, and rule B is input to knowledge debugger 23 to determine whether there is any contradiction with other rules. I do. If there is no contradiction, the rule B is added to the knowledge base 41. By these operations, rule A is changed to rule B.
The misdiagnosed data is determined again by the changed knowledge base 41, and the preventive maintenance accuracy is evaluated. If the determination is correct, the process ends. If the determination is not correct, the same process is repeated from the step of diagnosing the misdiagnosis again, the step of picking up the rule used for the diagnosis, until there is no erroneous diagnosis. However, if new information cannot be obtained or a new rule cannot be created, the process ends. If there are a plurality of rules that have caused a misdiagnosis, these processes are performed for all the rules.

[0031]

As described above, even if there is no detailed knowledge about the target of preventive maintenance, it is possible to easily construct a preventive maintenance system by searching for and obtaining digitized information. Also, as shown in the second example,
Self-evaluation of the preventive maintenance accuracy, self-learning by means of improving the rules can improve the preventive maintenance accuracy.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a first embodiment of the present invention.

FIG. 2 is a diagram showing details of an information search unit.

FIG. 3 is a diagram showing details of an information conversion unit.

FIG. 4 is a diagram showing a second embodiment of the present invention.

FIG. 5 is a diagram illustrating a learning algorithm according to a second embodiment of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Preventive maintenance device, 10 ... Information search means, 20 ... Knowledge conversion means, 30 ... Information display means, 40 ... Inference means, 41 ...
Knowledge base, 42 ... inference engine.

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Masanori Kanaga 5-2-1 Omikacho, Hitachi City, Ibaraki Prefecture Inside the Omika Plant, Hitachi, Ltd. (72) Inventor Satoshi Goto 5-chome, Omikamachi, Hitachi City, Ibaraki Prefecture No. 1 Inside the Omika Plant of Hitachi, Ltd.

Claims (5)

[Claims] 1. An information search means capable of accessing digitized technical information, books, and publications via a network, wherein information on an abnormality of equipment obtained by the information search means is detected by an abnormality detection rule and / or A preventive maintenance device having knowledge conversion means for converting to an abnormality cause estimation rule, and inference means for detecting an abnormality of equipment and / or estimating an abnormality cause based on the rule and measurement data. 2. A preventive maintenance device according to claim 1, further comprising an information display device for displaying the obtained information and / or rules converted by the knowledge conversion means. 3. The apparatus according to claim 1, further comprising preventive maintenance accuracy evaluation means for evaluating diagnosis accuracy based on a diagnosis result by said inference means and an actual state of the device, and said information retrieval when the diagnosis accuracy is low. A preventive maintenance device characterized in that it has a rule changing means for obtaining new information by means and changing rules in the knowledge base according to the obtained information. 4. A diesel engine power generation system comprising the preventive maintenance device according to claim 1, 2 or 3, wherein a measured state quantity of the system is input to the preventive maintenance device, and abnormality detection and ( Or) A diesel engine power generation system characterized by performing abnormality diagnosis. 5. An information search means capable of accessing digitized technical information, books and publications via a network, a knowledge conversion means for converting the obtained information into knowledge, and obtained by the information search means. A knowledge base construction apparatus comprising display means for displaying information and / or knowledge converted by the knowledge conversion means.