JP2000251126A - Analyzing device for proper maintenance operation - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make an early recovery from a fault and to prevent the fault from occurring again by analyzing proper maintenance operation meeting the conditions at the time of fault occurrence according to a past maintenance operation history and a past fault occurrence history. SOLUTION: By referring to fault information recorded in a fault information recording part 11, a totalization part 13 for the number of faults before maintenance operation and a totalization part 14 for the number of faults after maintenance operation sum up faults occurring in certain periods before the maintenance operation as to pieces of maintenance operation information recorded in a maintenance operation information recording part 12 and record the totalization results in the maintenance operation recording part 12. A maintenance operation classification part 15 refers to the maintenance operation information in the maintenance operation information recording part 12 to classify maintenance operation into proper operation and improper operation. A maintenance operation information learning part 16 fractionizes fault occurrence conditions and learn maintenance operation information meeting individual fault occurrence conditions according to the maintenance operation information classified by the classification part 15.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic transaction apparatus such as an automatic teller machine used in a financial institution such as a bank, or an automated ticket vending machine or an automatic ticket gate used in a transportation facility. The present invention relates to a proper maintenance work analyzer for analyzing and instructing proper maintenance work in a device.

[0002]

2. Description of the Related Art For example, in an automated teller machine or an automated machine such as a ticket vending machine or an automatic ticket gate, a normal operation is maintained and an early response to a failure is performed.
It is important to perform proper maintenance work.

[0003] Specific examples of the troubles that occur in the above-mentioned automated equipment include jams of bills and tickets, malfunctions of a transporting path for carrying bills and tickets, poor removal of bills and tickets, bills and boarding. Abnormal transport of tickets, cutting of journal paper, etc.

Conventionally, this maintenance work has been performed in accordance with conditions at the time of occurrence of a failure based on a maintenance manual prepared by a developer or maintenance staff of the automation equipment.

Here, specific examples of the fault occurrence conditions that occur in the above-mentioned automated equipment include the date and time of occurrence, the installation location of the generated equipment, the location of occurrence, the content of the fault, the degree of the fault, and the like.

Further, specific examples of the maintenance work for the above-mentioned automated equipment include adjustment of gaps and mounting positions, replacement of parts and units, cleaning, and the like.

[0007]

However, with the recent high performance and multifunctionalization of the automation equipment, the configuration of the automation equipment has become very complicated. As a result, the contents of the troubles that have occurred have become complicated, and there have been many troubles that cannot be anticipated in advance. For example, maintenance manuals created manually cannot handle the troubles.

Conventionally, such complicated failures have been dealt with on a case-by-case basis based on the experience and intuition of maintenance personnel. For this reason, there has been a problem that the contents of the work are different depending on the maintenance staff even for the same fault, and it is difficult to understand the appropriate maintenance work, and it takes time to deal with the fault.

[0009] Further, since maintenance is performed without knowing what kind of work is appropriate, there is a problem that a failure cannot be completely recovered, and the failure cannot be prevented from recurring.

Therefore, the present invention analyzes the appropriate maintenance work according to the conditions at the time of the failure based on the past maintenance work history and the failure occurrence history, and thereby makes it possible to recover the failure early and prevent the recurrence. An object of the present invention is to provide a maintenance work analyzer.

[0011]

A proper maintenance work analyzing apparatus according to the present invention relates to a trouble information recording means for recording trouble information indicating the contents of a trouble occurrence in a target device, and a maintenance work for a past trouble occurrence in the target device. Maintenance work information recording means for recording maintenance work information, and the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means, each maintenance work from the time of maintenance work In addition, the number of fault occurrence conditions in the previous specific period is totaled, and the total number of fault occurrences before maintenance work recording the totaled result in the maintenance work information recording unit; the fault information recorded in the fault information recording means; Based on the maintenance work information recorded in the maintenance work information recording means, a failure occurrence condition in a specific period after the maintenance work for each maintenance work. The number of post-maintenance failure occurrences totaling means, and the number of pre-maintenance failure occurrences totaled by the pre-maintenance failure occurrence totalizing means, and the maintenance work Maintenance work classifying means for classifying the maintenance work information in the maintenance work information recording means in which the number of post-maintenance trouble occurrences totalized by the post-failure occurrence number totaling means is recorded into proper maintenance work and maintenance work which is not so; And maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the maintenance work classification means.

Further, the proper maintenance work analyzing apparatus of the present invention comprises:
Failure information recording means for recording failure information indicating the content of the failure occurrence in the target equipment; maintenance work information recording means for recording maintenance work information relating to maintenance work for a past failure occurrence in the target equipment; and the failure information recording means. Based on the recorded failure information and the maintenance work information recorded in the maintenance work information recording means, the number of failure occurrence conditions in a specific period prior to the maintenance work is totaled for each maintenance work, and the totaled result is described above. The maintenance work information recording unit records the number of trouble occurrences before maintenance work, and the maintenance work information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means. Of the number of fault occurrence conditions during a specific period after the maintenance work, and the totaling result is recorded in the maintenance work information recording unit. The number of trouble occurrences after the work, the number of failures before the maintenance work counted by the number of troubles before the maintenance work, and the number of failures after the maintenance work, totaled by the number of troubles after the maintenance work are recorded. A maintenance work classifying means for classifying the maintenance work information stored in the maintenance work information recording means into a proper maintenance work and a maintenance work which is not proper; and a maintenance work based on the maintenance work information classified by the maintenance work classification means. A maintenance work information learning means for learning knowledge about the maintenance work information learning means, and a maintenance work prediction means for predicting maintenance work according to the failure occurrence condition for designated maintenance work information, using the learning result of the maintenance work information learning means; Using maintenance work information obtained from the maintenance work classification means, and a prediction result evaluation means for evaluating reliability of a prediction result by the maintenance work prediction means. It is Bei.

[0013] The proper maintenance work analyzer of the present invention comprises:
Failure information recording means for recording failure information indicating the content of the failure occurrence in the target equipment; maintenance work information recording means for recording maintenance work information relating to maintenance work for a past failure occurrence in the target equipment; and the failure information recording means. Based on the recorded failure information and the maintenance work information recorded in the maintenance work information recording means, the number of failure occurrence conditions in a specific period prior to the maintenance work is totaled for each maintenance work, and the totaled result is described above. The maintenance work information recording unit records the number of trouble occurrences before maintenance work, and the maintenance work information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means. Of the number of fault occurrence conditions during a specific period after the maintenance work, and the totaling result is recorded in the maintenance work information recording unit. The number of trouble occurrences after the work, the number of failures before the maintenance work counted by the number of troubles before the maintenance work, and the number of failures after the maintenance work, totaled by the number of troubles after the maintenance work are recorded. A maintenance work classifying means for classifying the maintenance work information stored in the maintenance work information recording means into a proper maintenance work and a maintenance work which is not proper; and a maintenance work based on the maintenance work information classified by the maintenance work classification means. Maintenance work information learning means for learning knowledge about the maintenance work information learning means, and maintenance work prediction means for predicting maintenance work according to the failure occurrence condition for designated maintenance work information using learning results obtained by the maintenance work information learning means; Using the prediction result by the maintenance work prediction means,
A maintenance work instructing means for instructing a maintenance work according to the failure occurrence condition for a newly generated failure.

Further, the proper maintenance work analyzing apparatus according to the present invention comprises a failure information recording means for recording failure information indicating the content of a failure occurrence in a target device, and maintenance work information relating to a maintenance work for a past failure occurrence in the target device. The maintenance work information recording means to be recorded, and the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means specify each maintenance work before the maintenance work. Means for counting the number of fault occurrence conditions during a period, and the counting result is recorded in the maintenance work information recording unit; and the failure information and the maintenance work information records recorded in the failure information recording means. Based on the maintenance work information recorded in the means, the number of fault occurrence conditions for each maintenance work in a specific period after the maintenance work is tabulated. A post-maintenance failure occurrence count means for recording the totaled result in the maintenance work information recording unit; and a pre-maintenance failure occurrence count and the post-maintenance failure count counted by the pre-maintenance failure count totalization means. Maintenance work classifying means for classifying maintenance work information in the maintenance work information recording means, in which the number of trouble occurrences after the maintenance work totaled by the number of cases totaling means is recorded, into proper maintenance work and maintenance work which is not proper; Maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the work classification means, and failure occurrence conditions for designated maintenance work information using the learning result by the maintenance work information learning means. Maintenance work prediction means for predicting maintenance work according to the A maintenance work instructing means for instructing a maintenance work in accordance with a harm occurrence condition, and a prediction result evaluation means for evaluating reliability of a prediction result by the maintenance work prediction means using maintenance work information obtained from the maintenance work classification means. Is provided.

According to the present invention, it is possible to learn about maintenance work for each fault occurrence condition that may occur during maintenance work. It is possible to specifically grasp whether or not to perform.

In addition, the maintenance work at the time of occurrence of an unknown failure can be predicted using the learning result, and the prediction result can be evaluated using the actual failure information and maintenance work information. It becomes possible to grasp.

Further, since maintenance work at the time of occurrence of an unknown failure can be predicted using the learning result, and appropriate maintenance work can be instructed for the failure that has occurred using the prediction result, a new failure can be indicated. Therefore, it is possible to perform appropriate maintenance work with uniform quality regardless of the experience and intuition of maintenance personnel, and it is possible to support early failure recovery and prevention of failure recurrence.

[0018]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 schematically shows a configuration of a proper maintenance work analyzer according to an embodiment of the present invention. FIG.
, The failure information recording unit 11 records failure information of a target device that performs maintenance work such as an automatic transaction device. A specific example of the failure information is, for example, an error log. The maintenance work information recording unit 12 records information on maintenance work (past maintenance work information) for a failure that has occurred in a target device. As a specific example of the maintenance work information, there is, for example, a maintenance slip recorded by the maintenance staff during the maintenance work.

The pre-maintenance trouble occurrence count totaling unit 13 refers to the trouble information recorded in the trouble information recording unit 11 and, for each maintenance work information recorded in the maintenance work information recording unit 12, The number of failures that occurred during the previous fixed period is totaled, and the totaling result is recorded in the maintenance work information recording unit 12. Specifically, the fixed period takes a value such as the past month, for example.

The post-maintenance-work-failure-counting unit 14 refers to the trouble information recorded in the trouble information recording unit 11 and, for each maintenance work information recorded in the maintenance work information recording unit 12, The number of failures that occurred during a certain period later is counted, and the counting result is recorded in the maintenance work information recording unit 12. Specifically, the certain period takes a value, for example, up to one month in the past.

The maintenance work classifying section 15 is a maintenance work information recording section 1 in which the results of the totalization by the pre-maintenance work failure occurrence totaling section 13 and the post-maintenance work failure occurrence totaling section 14 are recorded.
By referring to the maintenance work information of No. 2, the maintenance work is classified into an appropriate work and a work that is not appropriate according to the condition of the aggregation result.

The maintenance work information learning section 16 learns from the maintenance work information classified by the maintenance work classification section 15 subdivisions of fault occurrence conditions and maintenance work information corresponding to individual fault occurrence conditions.

The maintenance work prediction unit 17 predicts maintenance work according to the failure occurrence condition using the learning result of the maintenance work information by the maintenance work information learning unit 16.

The prediction result evaluator 18 includes the maintenance work predictor 1
The reliability of the prediction result is evaluated using the prediction result of No. 7 and the classified maintenance work information obtained from the maintenance work classifying unit 15.

The maintenance work information instructing section 19 gives an instruction regarding the maintenance work to the maintenance staff using the predicted result of the maintenance work according to the failure occurrence condition predicted by the maintenance work prediction section 17.

FIG. 2 shows an example of the failure information recorded in the failure information recording section 11 in a table format. As the failure information, the date and time of occurrence, the installation location of the device, the machine ID (device identification information), the content of the failure, and the like are recorded.

FIG. 3 shows an example of the maintenance work information recorded in the maintenance work information recording section 12 in a table format.
As maintenance work information, the date and time of occurrence, the installation location of the device, the machine ID, the content of the fault, the cause, the location of the occurrence, the content of the work, etc. are recorded. Of these, items other than the work content will be referred to herein as fault occurrence conditions.

FIG. 4 shows the number of trouble occurrences before maintenance work 1
3 and a part of the maintenance work information recorded in the maintenance work information recording unit 12 after being counted by the trouble occurrence number totaling unit 14 after the maintenance work. For each maintenance operation, the same aircraft,
The number of failures with the same content is tabulated.

FIG. 5 shows the result of the maintenance work classification unit 15 classifying the maintenance work into proper maintenance work and maintenance work that is not proper. Here, as an easy-to-understand example, a case in which the failure has not recurred for a certain period after the maintenance work is taken as the appropriate action, but other criteria may be used for this part.

FIG. 6 shows a decision tree as a learning result by the maintenance work information learning unit 16. This allows
For example, a proper maintenance operation such as "execute adjustment when the cause is balance (unbalance)" is obtained.

FIG. 7 shows an example in which a decision tree is generated without using the maintenance work information learning unit 16. As described above, a decision tree can be created even if attributes are blindly selected without learning, but the created tree is large in size and inefficient.

Here, a description will be given of a mathematical expression for obtaining the gain ratio of the information amount used for learning the decision tree representing the proper maintenance work from the maintenance work information classified as the proper maintenance work.

The following equation (1) is an equation for obtaining the information amount of the maintenance work information to be learned. When the set is T,
| T | represents the size of the set. In the case of this embodiment,
When T is a set of appropriate maintenance work information, | set of maintenance work information | = 8.

K is the number of classes, and freq (Cj
, T) is the number of cases in the set T whose class is Cj. Specific examples of the class include, for example, work contents such as adjustment and replacement. In the case of the present embodiment, there are two classes, “adjustment” and “exchange”, k = 2, freq (set of adjustment and maintenance work information) = 5, freq (set of exchange and maintenance work information) = 3 It is.

[0036]

(Equation 1)

The following equation (2) is an equation for calculating the information amount when the maintenance work information is divided by the attribute X. Here, n represents the number of attribute values of the attribute X, and Ti represents a set in which the attribute value of the attribute X is i. In the case of the present embodiment, for example,
When the attribute X is “cause”, the number of attribute values is “wear”, “deformation”, and “balance”, and n = 3.

[0038]

(Equation 2)

The following equation (3) represents the gain of the information amount obtained by the division. It is obtained by subtracting the information amount after division from the information amount before division.

[0040]

(Equation 3)

The following Expression 4 represents the information amount of the division itself by the attribute X.

[0042]

(Equation 4)

The following equation (5) represents the gain ratio of the information amount obtained by the division. This is obtained by normalizing the difference in the number of attribute values of the divided items with respect to the gain of the above equation (3). When learning the decision tree from the maintenance work information, an efficient decision tree is created by preferentially selecting one having a high gain ratio of this information amount.

[0044]

(Equation 5)

FIG. 8 is a prediction example of the maintenance work by the maintenance work information learning unit 16. When a new failure occurs and the failure occurrence condition is, for example, "the cause is wear and the place of occurrence is the transport belt", the proper maintenance work can be predicted as "adjustment" by tracing the decision tree. .

FIG. 9 shows an example of the evaluation result by the prediction result evaluation section 18. For each maintenance work performed after the learning, a proper maintenance work is predicted based on the learning result and compared with the work actually performed by the maintenance staff.
Here, only whether or not the forecast and the results match is dealt with. However, whether the maintenance work used for the evaluation was actually appropriate is determined from the new failure information, and each maintenance work information is (matched, appropriate). , (Mismatch, appropriate), (match, inappropriate), (mismatch, inappropriate).

FIG. 10 is an example of a display screen of a maintenance work instruction by the maintenance work instruction section 19. Regarding the newly generated failure, the maintenance staff will be responsible for the cause = wear,
Is input, and the maintenance work instructing unit 19 gives an instruction to execute the adjustment.

Next, the number of trouble occurrences before maintenance work counting section 13
Will be described with reference to the flowchart shown in FIG.

First, at step S 1, one piece of maintenance work information is read from the maintenance work information recording unit 12. Next, in step S2, out of the failure information recorded in the failure information recording unit 11, a failure occurred during a certain period before the work date and time of the read maintenance work information, and the same device as the read maintenance work is used. ,
Search for faults with the same content. It should be noted that a value such as the past month is set in advance for the certain period.

Next, in step S3, the number of fault occurrences is totaled from the search result in step S2, and in step S4, the totaling result is written in the maintenance work information in the maintenance work information recording unit 12.

Next, in step S5, it is checked whether or not all the maintenance work information has been processed. If unprocessed maintenance work information remains, the process returns to step S1 to continue the processing and perform all processing. If so, it ends.

Next, the post-maintenance trouble occurrence number counting section 14
Will be described with reference to the flowchart shown in FIG.

First, in step S11, one piece of maintenance work information is read from the maintenance work information recording unit 12. Next, in step S12, of the fault information recorded in the fault information recording unit 11, a fault that has occurred during a certain period after the work date and time of the read maintenance work information, and is the same device as the read maintenance work. , Search for a fault having the same content. Note that, for the certain period, a value such as one month after the maintenance work is set in advance.

Next, in step S13, step S12
The number of failure occurrences is counted from the search result of step S14, and step S14
Then, the totaling result is written in the maintenance work information in the maintenance work information recording unit 12.

Next, in step S15, it is checked whether or not all the maintenance work information has been processed. If unprocessed maintenance work information remains, the process returns to step S11 to continue the processing and complete the processing. If so, it ends.

Next, the processing of the maintenance work classifying unit 15 will be described with reference to the flowchart shown in FIG.

First, in step S21, one piece of maintenance work information recorded in the maintenance work information recording unit 12 is read. In this case, before the execution of step S21, the number of failure occurrences before maintenance work counting section 13 and the number of failure occurrences after maintenance work counting section 1 are set.
It is assumed that the totaling of the number of trouble occurrences by No. 4 has been completed.

Next, in step S22, a comparison is made between the number of trouble occurrences before work and the number of troubles after work in the maintenance work information. As a result of this comparison, if the number of post-work failure occurrences is smaller than the number of pre-work trouble occurrences, step S2
At 3, the appropriate work flag is set in the maintenance work information.

In this case, the criteria for determining the appropriate maintenance work is a case where the number of failures after the work is smaller than the number of failures before the work. Is considered to be “0”, for example, and several methods are conceivable, and the criterion for determining appropriate maintenance work is not limited to the present embodiment.

Next, in step S24, it is checked whether or not the above processing has been performed for all maintenance work information.
If not, step S21 and subsequent steps are continued. If all steps have been processed, the process ends.

Next, the processing of the maintenance work information learning unit 16 will be described with reference to the flowcharts shown in FIGS. The processing of the maintenance work information learning unit 16 includes two routines, a main routine shown in FIG. 14 and a decision tree learning routine shown in FIG.

First, in the main routine of FIG. 14, all the maintenance work information is read from the maintenance work classification unit 15 in step S31. Next, in step S32, the decision tree learning routine is called with all the maintenance work information as arguments.

Next, in the decision tree learning routine of FIG.
First, in step S33, one of the attributes of the maintenance work information is set.
Choose one. Specific examples of the attributes include date and time, location, machine ID, fault content, cause, and the like. Next, step S
At 34, a gain ratio of the information amount to the selected attribute is determined. The information amount gain ratio is a criterion representing the efficiency of dividing a tree in a decision tree. Although the gain ratio of the information amount is used in the present embodiment, the gain itself may be used, and the method of obtaining the reference is not limited to the present embodiment.

Next, in step S35, it is checked whether or not all the attributes have been processed. If there is an attribute that is not to be processed, the process returns to step S33 to execute again. If all the processes have been completed, the process returns to step S36. Proceed to step S34
The attribute that maximizes the gain ratio of the amount of information obtained in step is selected.

Next, in step S37, one attribute value belonging to the selected attribute is selected. Next, step S38
Then, it is checked whether the maintenance work information having the selected attribute values all belong to the same class. Here, the class indicates a group to which the maintenance work information belongs, and the content of the class differs depending on the content to be learned. Specific examples of the class include proper work contents and distinction between proper work and unsuitable work.

If the result of the check in step S38 is that the class does not belong to a single class, the decision tree learning routine is recursively called in step S39 with the subset of the maintenance work information having the selected attribute value as an argument. If the result of the check in step S38 is that the class belongs to a single class, the class is set in step S40.

Next, in step S41, it is checked whether all attribute values have been processed. As a result of this check, if there is an unprocessed attribute value, step S37
When all the attribute values have been processed, the process returns to the parent routine that called the decision tree learning routine.

Next, the processing of the maintenance work prediction section 17 will be described with reference to the flowcharts shown in FIGS. The processing of the maintenance work predicting unit 17 includes a main routine shown in FIG. 16 and a decision tree search routine shown in FIG.
Consists of two routines.

First, in the main routine of FIG. 16, the learning result is read from the maintenance work information learning section 16 in step S51. Next, in step S52, a failure occurrence condition is read. The fault occurrence condition is given from the prediction result evaluation unit 18 or the maintenance work instruction unit 19.

Next, in step S53, a decision tree learning search routine is called using the root node of the learning result and the fault occurrence condition as arguments. Here, a node represents a node of the decision tree. As a specific example, “cause” or “adjustment” in FIG. 6 corresponds to a node. The root node represents a root node of the decision tree. As a specific example, FIG.
Is the root node.

Next, in the decision tree search routine of FIG.
First, in step S54, an attribute value corresponding to the attribute of the node is selected from the failure occurrence conditions. For example, in the case of the example of FIG. 3, if the node is “cause”, the attribute value selected from the failure occurrence condition is “wear”. Next, step S5
At 5, a subnode corresponding to the attribute value is selected. For example, in FIG. 6, the subnodes corresponding to the "cause" node are "adjustment", "replacement", and "occurrence location".

Next, in step S56, it is checked whether the class has been reached. As a result of this check, if the class has not been reached, the decision tree search routine is recursively called in step S57 using the subnode and the fault occurrence condition as arguments. When the class is reached as a result of the above check, the value of the class is set as the maintenance work prediction in step S58.

Next, the processing of the prediction result evaluation section 18 will be described with reference to the flowchart shown in FIG.

First, in step S 61, one piece of maintenance work information that has been classified as appropriate is read from the maintenance work classification unit 15. Here, the maintenance work information to be read can be arbitrarily both the same maintenance work information used for learning or another maintenance work information not used for learning, depending on the evaluation purpose at the time of implementation. .

Next, in step S62, the fault occurrence condition in the maintenance work information is passed to the maintenance work prediction routine of FIG. Next, in step S63, a prediction result is received from the maintenance work prediction routine.

Next, in step S64, step S63
It is checked whether or not the prediction of the appropriate maintenance work obtained in step 1 and the actually performed maintenance work obtained from the maintenance work information match. If the result of this check is that they match, a correct answer graph is set in the maintenance work information in step S65.

Next, in step S66, it is checked whether or not all maintenance work has been processed. If any unprocessed maintenance work remains, the flow returns to step S61 to continue the processing.
If the result of the check in step S64 is that they do not match, step S65 is jumped to step S66.
Proceed to.

If the result of the check in step S66 is that all the processes have been completed, then in step S67, the correct rate of prediction is calculated. Specifically, the prediction correct rate can be obtained from (the number of correct answers / the total number of maintenance work to be evaluated).

Next, the processing of the maintenance work instruction unit 19 will be described with reference to the flowchart shown in FIG.

First, in step S71, a fault occurrence condition of a newly generated fault is read. The failure occurrence condition is input, for example, by a user such as a maintenance person when maintenance work occurs. Next, in step S72, the input fault occurrence condition is passed to the maintenance work prediction routine in FIG. Next, in step S73, a prediction result is received from the maintenance work prediction routine. Next, in step S74, the prediction result is displayed.

[0081]

As described above in detail, according to the proper maintenance work analyzing apparatus of the present invention, it is possible to learn about maintenance work for each fault occurrence condition that can occur during the maintenance work. Thus, it is possible to specifically grasp what maintenance work should be performed according to the failure occurrence condition.

Further, according to the proper maintenance work analyzing apparatus of the present invention, the maintenance work at the time of occurrence of an unknown failure is predicted using the learning result, and the prediction result is evaluated using the actual failure information and the maintenance work information. Therefore, it is possible to grasp the reliability of the prediction result.

Further, according to the proper maintenance work analyzing apparatus of the present invention, the maintenance work at the time of occurrence of an unknown failure is predicted using the learning result, and an appropriate maintenance work for the generated failure is performed using the prediction result. Because it is possible to give instructions, it is possible to perform appropriate maintenance work with uniform quality regardless of the experience and intuition of maintenance personnel, even for newly occurring failures, and to promptly recover from failures and reoccurrence of failures. Can help prevent it.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing a configuration of a proper maintenance work analyzer according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of fault information recorded in a fault information recording unit.

FIG. 3 is a diagram showing an example of maintenance work information recorded in a maintenance work information recording unit.

FIG. 4 is a diagram showing an example of maintenance work information recorded in a maintenance work information recording unit after performing the tallying.

FIG. 5 is a diagram showing an example of a classification result by a proper maintenance work classifying unit.

FIG. 6 is a diagram illustrating an example of a decision tree as a learning result by a maintenance work information learning unit.

FIG. 7 is a diagram illustrating an example in which a decision tree is generated without using a maintenance work information learning unit.

FIG. 8 is a diagram illustrating an example of maintenance work prediction by a maintenance work prediction unit.

FIG. 9 is a diagram illustrating an example of an evaluation result by a prediction result evaluation unit.

FIG. 10 is a diagram showing an example of a display screen of a maintenance work instruction by a maintenance work instruction unit.

FIG. 11 is a flowchart illustrating processing of a failure occurrence count totaling unit before maintenance work.

FIG. 12 is a flowchart for explaining the processing of a post-maintenance work failure count counting unit;

FIG. 13 is a flowchart illustrating processing of a maintenance work classifying unit.

FIG. 14 is a flowchart illustrating processing of a maintenance work information learning unit.

FIG. 15 is a flowchart illustrating processing of a maintenance work information learning unit.

FIG. 16 is a flowchart illustrating processing of a maintenance work prediction unit.

FIG. 17 is a flowchart illustrating processing of a maintenance work prediction unit.

FIG. 18 is a flowchart illustrating processing of a prediction result evaluation unit.

FIG. 19 is a flowchart illustrating processing of a maintenance work instruction unit.

[Explanation of symbols]

11: Failure information recording unit (failure information recording unit) 12: Maintenance work information recording unit (maintenance work information recording unit) 13: Number of trouble occurrences before maintenance work (unit of failure occurrences before maintenance work) 14: Maintenance work Post-failure occurrence counting section (maintenance post-operation failure counting section) 15: maintenance work classification section (maintenance work classification section) 16: maintenance work information learning section (maintenance work information learning section) 17: maintenance work prediction section (maintenance) 18: Prediction result evaluation unit (Prediction result evaluation unit) 19: Maintenance work instruction unit (Maintenance operation instruction unit)

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 3E040 BA07 BA13 BA20 CA07 CA11 DA10 5B055 BB03 GG01 GG05 GG10 LL01 NA18 NB03 NB05 NC08 ND01 ND09 ND10

Claims (9)

[Claims] 1. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; With the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means,
Pre-maintenance work failure count totaling means for summing up the number of fault occurrence conditions during a specific period prior to the maintenance work for each maintenance work, and recording the tally result in the maintenance work information recording unit; With the fault information recorded in the means and the maintenance work information recorded in the maintenance work information recording means,
A post-maintenance work trouble occurrence number totaling means for summing up the number of fault occurrence conditions in a specific period after the maintenance work for each maintenance work and recording the summed result in the maintenance work information recording unit; Maintenance work information in the maintenance work information recording means in which the number of failures before maintenance work totaled by the number of trouble occurrences totalization means and the number of failure occurrences after maintenance work totaled by the number of failure occurrence after maintenance work totalization means are recorded. Maintenance work classifying means for classifying the maintenance work into appropriate maintenance work and maintenance work which is not appropriate; and maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the maintenance work classification means. A proper maintenance work analyzer, comprising: 2. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; With the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means,
Pre-maintenance work failure count totaling means for summing up the number of fault occurrence conditions during a specific period prior to the maintenance work for each maintenance work, and recording the tally result in the maintenance work information recording unit; With the fault information recorded in the means and the maintenance work information recorded in the maintenance work information recording means,
A post-maintenance work trouble occurrence number totaling means for summing up the number of fault occurrence conditions in a specific period after the maintenance work for each maintenance work and recording the summed result in the maintenance work information recording unit; Maintenance work information in the maintenance work information recording means in which the number of failures before maintenance work totaled by the number of trouble occurrences totalization means and the number of failure occurrences after maintenance work totaled by the number of failure occurrence after maintenance work totalization means are recorded. Maintenance work classifying means for classifying the maintenance work into appropriate maintenance work and maintenance work which is not appropriate; maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the maintenance work classification means; A maintenance work predictor that predicts maintenance work according to the failure occurrence condition for designated maintenance work information using a learning result of the maintenance work information learning means. And a predictive result evaluating means for evaluating the reliability of the prediction result by the maintenance work predicting means using the maintenance work information obtained from the maintenance work classifying means, . 3. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; With the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means,
Pre-maintenance work failure count totaling means for summing up the number of fault occurrence conditions during a specific period prior to the maintenance work for each maintenance work, and recording the tally result in the maintenance work information recording unit; With the fault information recorded in the means and the maintenance work information recorded in the maintenance work information recording means,
A post-maintenance work trouble occurrence number totaling means for summing up the number of fault occurrence conditions in a specific period after the maintenance work for each maintenance work and recording the summed result in the maintenance work information recording unit; Maintenance work information in the maintenance work information recording means in which the number of failures before maintenance work totaled by the number of trouble occurrences totalization means and the number of failure occurrences after maintenance work totaled by the number of failure occurrence after maintenance work totalization means are recorded. Maintenance work classifying means for classifying the maintenance work into appropriate maintenance work and maintenance work which is not appropriate; maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the maintenance work classification means; A maintenance work predictor that predicts maintenance work according to the failure occurrence condition for designated maintenance work information using a learning result of the maintenance work information learning means. Proper maintenance characterized by comprising: maintenance work instructing means for instructing a maintenance work according to the failure occurrence condition for a newly generated failure using a prediction result by the maintenance work prediction means. Work analyzer. 4. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; With the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means,
Pre-maintenance work failure count totaling means for summing up the number of fault occurrence conditions during a specific period prior to the maintenance work for each maintenance work, and recording the tally result in the maintenance work information recording unit; With the fault information recorded in the means and the maintenance work information recorded in the maintenance work information recording means,
A post-maintenance work trouble occurrence number totaling means for summing up the number of fault occurrence conditions in a specific period after the maintenance work for each maintenance work and recording the summed result in the maintenance work information recording unit; Maintenance work information in the maintenance work information recording means in which the number of failures before maintenance work totaled by the number of trouble occurrences totalization means and the number of failure occurrences after maintenance work totaled by the number of failure occurrence after maintenance work totalization means are recorded. Maintenance work classifying means for classifying the maintenance work into appropriate maintenance work and maintenance work which is not appropriate; maintenance work information learning means for learning knowledge about maintenance work based on the maintenance work information classified by the maintenance work classification means; A maintenance work predictor who predicts maintenance work according to the failure occurrence condition for designated maintenance work information using learning results obtained by the maintenance work information learning means. A maintenance work instructing means for instructing a maintenance work according to a failure occurrence condition for a newly generated failure using a prediction result by the maintenance work prediction means; and a maintenance work obtained from the maintenance work classification means. A proper maintenance work analysis device, comprising: prediction result evaluation means for evaluating reliability of a prediction result by the maintenance work prediction means using information. 5. The maintenance work information learning means according to claim 1, wherein said maintenance work information learning means learns what maintenance work is appropriate for each fault occurrence condition that may occur during the maintenance work. The proper maintenance work analyzer according to any one of the above. 6. The maintenance work information learning means according to claim 1, wherein said maintenance work information learning means learns what maintenance work is improper for each failure occurrence condition that may occur during maintenance work. 5. The proper maintenance work analyzer according to any one of the above items 4 to 4. 7. The maintenance work information learning means may include a failure occurrence condition when maintenance work is appropriate, a failure occurrence condition when maintenance work is inappropriate, and a failure occurrence condition and maintenance work when maintenance work is inappropriate. 5. The method according to claim 1, wherein knowledge about the difference between the two is learned.
The proper maintenance work analyzer according to any one of the above. 8. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; Maintenance work information learning means for learning knowledge about maintenance work based on the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means; and a learning result obtained by the maintenance work information learning means. A proper maintenance work analyzer, comprising: output means for outputting. 9. A failure information recording means for recording failure information indicating the content of a failure occurrence in a target device; a maintenance work information recording means for recording maintenance work information on maintenance work for a past failure occurrence in the target device; Output means for outputting appropriate maintenance work content for a symmetrical device in which a failure has occurred based on the failure information recorded in the failure information recording means and the maintenance work information recorded in the maintenance work information recording means. A proper maintenance work analyzer.