JP3293897B2 - Maintenance support equipment in an expert system for abnormality diagnosis for aircraft - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a maintenance support system in an abnormality diagnosis expert system for an aircraft, and more particularly to a maintenance support system using past cases.

[0002]

2. Description of the Related Art Conventionally, certainty representing uncertainty in executing an inference process is determined by directly inquiring of an expert or by appropriately modifying the obtained information. Further, no case database for storing past failure cases has been constructed at all, and the quality and number of data are not sufficient to be used uniformly.

[0003]

Conventionally, the certainty of rules has been determined by experts in each field based on subjective judgments, and the numerical value of certainty fluctuates due to individual differences among experts. There is a problem that appropriate and reasonable certainty has not been determined. In addition, rules are often created by the experts themselves, and experts who are unfamiliar with creating and editing rules require a great deal of effort, and the content of the rules created differs for each expert, There is a problem that it is difficult to unify the rule contents and it is difficult to maintain and manage the rules. In addition, the learning function for updating the predetermined certainty while executing the inference processing is insufficient. In other words, for a case with a high failure rate, the certainty is increased,
There is a problem that the function of sequentially improving reliability of the certainty by reducing the certainty of the case with a low failure occurrence rate is not provided.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned problems, to determine an objective and reasonable certainty that does not depend on the subjectivity of an expert, and to update the certainty based on a new case. An object of the present invention is to provide a maintenance support device in an abnormality diagnosis expert system for a possible aircraft.

[0005]

The present invention relates to (a) a failure case data input / output device 14, which includes the date and time when a failure occurred, the date and time when the failure was discovered, and the name of the device where the failure occurred on a display screen. And the device number, the type of defect that caused the failure, the reproduction category indicating the reproducibility, the failure state and the content of the failure defect, the details of the measures taken to repair the failure,
The date and time of completion of the treatment, the treatment time of the treatment, the work unit code indicating the contents of the treatment work, the number of the part used for the treatment, the cumulative time the part was used, the part failure status, and the treatment A failure case data input / output device 14 for displaying a failure case unit which is a worker who has performed and inputting failure case data using this display screen; and (b) a case database means 11, which is provided before the flight of the aircraft. The contents of the maintenance work work card, which describes the standard work procedures for regular inspections after the flight and for a certain period of time, constitute the case data. Each attribute that has a consequent part and a certainty factor, and is a condition part and a consequent part, includes a work unit code name, a department name, an aircraft ID number, an occurrence date and time, a flight time, a defect finding time, The number of people and time required for maintenance, the elapsed maintenance time,
The attributes of the work unit code name, the description of the failure, and the failure status of the machine have a handling example that has been dealt with by adjustment, replacement and removal / attachment, while referring to the dictionary for data unification. The contents of input data of cases by the failure case data input / output device 14 are unified and filed, and a case database is provided. This case database has a hierarchical structure for each system of devices including a hydraulic system and a brake system. Among the attributes of the divided hierarchical case database and the case database, the attributes used for troubleshooting and the order of use thereof are almost constant, and the attributes, the number of cases, and the ID code for searching the cases are networked in advance. And a case network, wherein the certainty is determined by the consequent part as to the total number of past causes of failure in the condition part of each attribute. Examples are the number of the case database means for storing the fault case data in the inference of rules format inference means 1 to infer (c) case database means 11
2. At the time of the periodic inspection, the weight of the search item of the newly generated failure case is set while referring to the weight definition file of the data item, and the past is referred to while referring to the hierarchical case database and the case network. The similarity with each failure case is calculated, and the calculated similarity exceeds a predetermined threshold and selects the failure case data similar to the unstored failure case, and the attribute of each attribute of the selected similar failure case data is selected. Inside, based on the basic data that is the attribute to be the search condition,
Searching and performing case matching processing, narrowing down fault cases obtained by searching using a pointing device and displaying maintenance support information, and based on the number of cases in the consequent part and the number of applicable cases in the condition part regarding search results And an inference means for calculating the certainty of the rule.

[0006]

According to the present invention, a failure case similar to an unstored failure case is retrieved from the case database of the case database means and objectively determined based on the number of cases in the consequent part and the number of cases in the condition part. It is possible to determine a reliable unified certainty, and to update the reliability to a more reliable certainty according to a newly occurring failure case.

Further, since the contents described in the maintenance work card constitute the case data as it is, every time a new failure case is input during the maintenance work, the certainty is automatically updated, and human labor is reduced. Can be reduced.

Further, since the case database itself has a condition part and a consequent part and is configured in a rule format that can execute inference as it is, there is no need to newly create a conventional IF-THEN format rule. In addition, it is possible to solve various problems associated with rule creation, such as difficulty in creating rules, securing uniformity of rules, and labor for maintaining and managing rules.

[0009]

FIG. 1 is a configuration diagram of a maintenance support device according to an embodiment of the present invention. The maintenance support device 10 includes a case database 11 for storing a large number of failure cases, an inference engine 12 that executes inference processing based on the case database 11, and management that is involved in a dialog between the user 15 and the inference engine 12. A device 13 and a case data input / output device 14 for a user 15 to edit the contents of the case database 11
It is composed of The inference engine 12 starts inference processing of the case database 11 in response to a request from the management device 13 according to a user's instruction, outputs the processing result to the management device 13, and notifies the user 15. In this way, by preparing user interfaces such as the management device 13 and the case data input / output device 14, the user 15 can easily operate the database of abnormal cases without being conscious of it.

FIG. 2 is a flowchart showing functions of the maintenance support device 10 according to one embodiment of the present invention. When an abnormality or failure occurs in the target mechanical device, it is recognized that a problem has occurred in step a1, and in step a2, the problem that has occurred is analyzed, and the contents of the defect and maintenance data are analyzed. Register in the case database 11 to be stored. Next, in step a3, a past failure case similar to the failure case including the generated problem is searched, and in the next step a5, some obtained failure cases are displayed as solution candidates. Next, step a6
In, a failure case that becomes a solution is selected from among the solution candidates, and the content of the failure defect or the treatment is displayed or output. Next, in step a7, a document such as a failure repair manual is searched with reference to the output failure content and action content, and a more specific failure content is examined.

On the other hand, if an appropriate solution is not found in the displayed solution candidates in step a5, it is determined that the new failure case has not occurred in the past, and in step a8, the failure content of the case is determined. Enter data such as maintenance information for At this time, step a9
There is also a case where a new failure case is input while being appropriately corrected via the. With respect to the data such as the maintenance information input in step a8, the corrected maintenance record and data are input and stored in the case database 11 in step a4, and the network is expanded for the next high-speed search.

FIG. 3 is a flowchart showing functions of the case data input / output device 14 shown in FIG. First, starting from step b1, in step b2, it is determined whether the input is an online input in which a new failure case is sequentially transmitted or an offline input in which batch processing is performed for each predetermined number. In step b3, an abnormal case database or a difference case database is determined. To quote. The difference case database is case data newly input from the time of the previous batch processing, and is extracted as difference case data when the case data is not stored in a separate file. When a new failure case is input online, the process proceeds to step b4 to input a newly generated case. In step b8, the difference case database 1 is created and added, and the process proceeds to step b6. On the other hand, when a new failure case is input off-line, the process proceeds to step b5, in which difference case data is extracted from the data, and the process proceeds to step b5.
In step 8, the difference case database 1 is created and added, and the process proceeds to step b6. In step b6,
Relationship definition database of step b7 and step b
By referring to the difference case database 1 of No. 8, only the data to be inferred defined in the relation definition database is extracted from the items input for each case. The relational database is a file that defines attributes that are a condition part and a consequent part.

Next, in step b9, step b
While referring to the ten data unification dictionaries, the contents of the input data of cases having individual differences are unified, the difference case database 2 is created in step b11, and the process ends in step b12.

FIG. 4 is an example of a display screen of the case data input / output device 14. The date and time when the failure occurred, the date and time when the failure was discovered, the name and number of the device where the failure occurred, the type of defect that caused the failure, and its reproducibility so that the data content of the failure case can be grasped at a glance on the screen Reproduced category indicating the failure status, details of the failure status and failure, the content of the action performed to repair the failure, the date and time of the completion of the action, the time required for the action, the operation unit code indicating the content of the action, The number of the part used for this purpose, the cumulative time the part has been used, the failure state of the part, the operator who performed the treatment, and the like are displayed in characters or symbols. Using this display screen, data is input while moving the cursor for each item.

FIG. 5 is a flowchart showing the operation of the inference engine 12 shown in FIG. First, step c
Starting from Step 1, in Step c2, Step c3
Quoting the hierarchical case database or case network of, and starting a search based on basic data, which is an attribute that is always a search condition among the attributes of case data,
Case collation processing is performed, and the collation result is displayed on the screen in step c4. The hierarchical case database is, for example, a case database divided into a hierarchical structure for each system of devices such as a hydraulic system and a brake system, and the processing is speeded up by narrowing a search range. In the case network, since the attributes used for searching for a failure among the attributes of the case database and the order of use thereof are almost constant, the attributes, the attribute values, the number of cases, and the ID code for searching the cases are networked in advance. , In order to speed up the inference processing.

Next, at step c5, a search item, that is, an attribute name is selected from the displayed collation list by a pointing device such as a mouse at step c6, and then at step c7, cases are narrowed down. A list of examples narrowed down in step c8 is displayed on the screen.

Next, in step c9, it is determined whether or not the search has been completed.
And step c7 is repeated. When the search is completed, the process proceeds to step c10, where the searched case is presented, and is displayed on the screen in step c11.
In step c13, necessary information such as maintenance support information is presented, and displayed on the screen in step c13.
Ends with In this way, it is possible to search for past failure cases similar to new failure cases while narrowing down search cases in an interactive manner on the screen.

Next, a procedure for searching for a fault will be described by taking an aircraft as an example. FIG. 6 is a flowchart showing a procedure for searching for a failure in an aircraft, and FIG. 7 is an example of the structure of a case database. First, in the case of a periodic inspection performed before or after a flight or at a predetermined time such as a flight for a certain period of time, starting from step d1, in step d2 standard work procedures are arranged and described in units of failure cases as shown in FIG. An inspection is performed based on the completed maintenance work card to check for any defects. Next, in step d3, if there is no defect, the process proceeds to step d13 to terminate the inspection. If a defect is found, the process proceeds to step d8, and after identifying the defective portion, in step d9, ,
By inferring a case database storing past trouble cases, the cause of the trouble is investigated and specified.
The case database is divided into networks for each attribute name as shown in FIG. 7 and networked. As shown in FIG. 8, a condition part, a consequent part, and certainty are stored for one attribute name. ing. For example, paying attention to the attribute name WCODE, when the cause of the failure is the oxygen sensor, the number of cases of the consequent part dealt with by the adjustment is five times for the total number of past condition parts of seven times, that is, the reliability is 5/7. And
The case dealt with by replacement has a certainty of 1/7, and the case dealt with by removal or attachment has a certainty of 1/7. Here, if a new failure case is dealt with by replacing the oxygen sensor, the total number of past cases is changed to eight, and the reliability of the replacement is updated to 2/8. In this way, past failure cases are integrated, and a more reliable certainty is obtained.

Next, in step d10 of FIG. 6, it is determined whether the level of the defect is the component level or the machine level. If the level is the component level, the process proceeds to step d11 where the corresponding component is inspected. If is the machine level, the process proceeds to step d12, where the maintenance process is actually performed, and the process ends in step d13.

On the other hand, if maintenance is to be performed at the time of occurrence of a malfunction, the process proceeds from step d4 to step d5, where inspection of the aircraft is started. At step d6, it is determined whether or not the location of the malfunction can be specified. If not, the process proceeds to step d7, and the worker actually performs an inspection work, tries to identify a defective portion, and proceeds to step d8. If the location where the failure has occurred can be identified, step d9
After that, the procedure becomes the same as that of the periodic inspection.

FIG. 9 is a flowchart for automatically searching the case database. First, step e1
In step e2, the weight of the search item of the newly generated failure case is set while referring to the weight definition file of the data item of step e3, and then, in step e4, the hierarchical case database or the case of step e5 is set. While referring to the network, the similarity with each past failure case is calculated, and in step e6, the case where the calculated similarity exceeds a predetermined threshold is selected. Then, in step e7, each item of the selected similar case is extracted and displayed on the screen in step e8, and in step e9, text data or a work drawing representing a work procedure related to the similar case is represented. Necessary information such as maintenance support information such as image data is extracted and displayed on the screen in step e10.

As described above, the cause of a newly generated failure case can be quickly searched for while referring to the certainty based on the past failure cases stored in the case database.

[0023]

As explained in detail above, according to the present invention,
Investigate the cause of failure because it can calculate unified objective certainty that does not depend on the subjectivity of experts and can update it to more reliable certainty based on newly occurring failure cases Can be done easily and quickly. Further, since the update of the certainty is performed every time a new failure case is input, it is possible to reduce human labor.

Further, since the case database can be inferred and executed as a rule as it is, the conventional I
There is no need to create rules in the F-THEN format, and it is possible to solve various problems associated with rule creation, such as difficulty in rule creation, ensuring rule uniformity, and labor for rule maintenance management.

In particular, according to the present invention, the failure case data input / output device 14 displays failure cases in units of failure cases unique to an aircraft on a display screen thereof and inputs failure case data. The attribute that is a condition part and a consequent part for each inspection work item is an attribute unique to the aircraft shown in FIG. 7 described above. The decision can be reasonably and accurately updated, and by having a hierarchical case database and a case network, the rational certainty is further improved.

In addition, aircraft require regular inspections before flight, after flight and for a certain period of time, and the present invention is advantageously implemented for such periodic inspections, and the determination of certainty is rationally determined. Achieved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a maintenance support device 10 according to an embodiment of the present invention.

FIG. 2 is a flowchart showing functions of a maintenance support device 10 according to one embodiment of the present invention.

FIG. 3 is a flowchart showing functions of the case data input / output device 14 shown in FIG.

FIG. 4 is an example of a display screen of the case data input / output device 14;

FIG. 5 is a flowchart showing the operation of the inference engine 12 shown in FIG. 1;

FIG. 6 is a flowchart showing a procedure for searching for a failure in an aircraft.

FIG. 7 is an example of the structure of a case database 11;

FIG. 8 is an example of a rule of each attribute of the case database 11 shown in FIG. 7;

FIG. 9 is a flowchart for automatically searching the case database 11;

[Explanation of symbols]

 Reference Signs List 10 maintenance support device 11 case database 12 inference engine 13 management device 14 case data input / output device 15 user

──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Susumu Hashidera 1-1, Kawasaki-cho, Akashi-shi, Hyogo Kawasaki Heavy Industries, Ltd. Inside the Akashi Plant (72) Inventor, Norimasa Sakakawa 1, Kawasaki-cho, Kakamigahara-shi, Gifu Prefecture Kawasaki Heavy Industries (72) Inventor, Yukihiro Yamada 1 in Kawasaki-cho, Kakamigahara-shi, Gifu Prefecture Kawasaki Heavy Industries, Ltd. Inside the Gifu Plant (72) Inventor Tatsuya Sugioka 1 Kawasaki-cho, Kakamigahara City, Gifu Prefecture Kawasaki Heavy Industries, Ltd. Gifu Plant (56) References JP-A-2-200048 (JP, A) JP-A-3-154846 (JP) , A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 9/44 550 G01M 17/00 G01M 19/00

Claims (1)

(57) [Claims] (A) A failure case data input / output device 14, which includes a date and time when a failure occurred, a date and time when a failure was discovered, a device name and a device number where the failure occurred, and a failure caused by the display screen. Defect type, reproducibility category indicating its reproducibility, failure state and failure defect content, details of treatment performed to repair the failure, date and time of completion of the treatment completed, treatment time taken for the treatment, and details of the treatment work The work unit code, the number of the part used for the treatment, the cumulative time the part has been used, the part failure status, and the failure case unit that is the worker who performed the treatment are displayed. A failure case data input / output device 14 for inputting data; and (b) the case database means 11, which is a standard product for a regular inspection of a flight before, after, and for a certain period of time. Procedure is the description of the maintenance work work card that has been described, constitute a case data, each attribute that has been divided into different inspection work items of the work card,
Each attribute, which has a condition part, a consequence part and a certainty factor, was required for the work part code name, department name, aircraft ID number, date and time of occurrence, flight time, defect discovery time, and maintenance. The attributes of the work unit code, the failure details, and the failure status of the machine are adjusted, replaced, and removed. It has a handling case dealt with by mounting, quoting the dictionary for data unification, unifying the contents of the input data of the case by the failure case data input / output device 14, and has a case database. For each system of the system including the hydraulic system and the brake system, a hierarchical case database divided into a hierarchical structure, and among the attributes of the case database, Attributes and their use order used in impaired inquiry is substantially constant, attributes,
A case network in which the number of cases and the case search ID code are pre-networked, and the certainty is the number of the coping cases of the consequent part with respect to all past cases of the failure cause of the condition part of each attribute, Case database means for storing failure case data in an inferable rule format; and (c) inference means 1 for inferring the case database means 11
2. In the periodic inspection, the weight of the search item of the newly generated failure case is set while referring to the weight definition file of the data item, and the past is referred to while referring to the hierarchical case database and the case network. Calculate the similarity to each failure case, select the failure case data that is similar to the unstored failure case when the calculated similarity exceeds a predetermined threshold, and select the attribute of each attribute of the selected similar failure case data. Based on the basic data, which is the attribute used as a search condition, search and case matching processing are performed, and the failure cases obtained by the search are narrowed down by a pointing device to display maintenance support information, and the search results are displayed. And inference means for calculating the certainty of the rule based on the number of cases in the consequent part and the number of applicable cases in the condition part. Maintenance support device of the diagnostic expert system.