JP2019018979A - Elevator system - Google Patents

Abstract

To provide an elevator system for improving a maintenance efficiency by memorizing and learning failure information and a series of operations as maintenance work addressing the failure for accumulation of knowledge for failure countermeasures and when the failure happens by inferring a series of adequate operations from the leaning result when a failure occurs.SOLUTION: An elevator system has: an elevator controller for controlling an elevator apparatus; a maintenance terminal for providing a maintenance work environment for maintenance workers; operation history storage means for storing a series of operation histories for operating the maintenance terminal; failure information analysis means for analyzing failure information and outputting an analysis result; learning processing means for outputting a learning result with the operation histories and the analysis result as teacher data; and inference means for inferring an adequate work procedure corresponding to new failure information on the basis of the leaning result.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an elevator system for inferring appropriate maintenance work corresponding to an arbitrary failure or an operation procedure for countermeasures against the failure.

  Elevators in recent years have been digitized, and operate in a combination of a mechanical system such as a hoisting machine or a car door, an electrical system such as a control controller, and software that operates on an arithmetic device. When a problem occurs in such an elevator, if only the mechanical system is the cause, the countermeasure is completed with simple operations such as replacement and repair of the corresponding parts.

  As a prior art for coping with mechanical problems, although not related to an elevator, there is Patent Document 1. This summary includes the following: “Appropriate maintenance work that enables early recovery of failures and prevention of recurrence by analyzing appropriate maintenance work according to the conditions at the time of failure based on past maintenance work history and failure occurrence history. As a means to solve the problem of “providing an analysis device”, “learning maintenance work according to failure occurrence conditions during maintenance work from past maintenance work information, and using that learning result, maintenance work when a new failure occurs. And predicting the prediction result using actual failure information and maintenance work information, and supporting the appropriate maintenance work for the failure occurring in the target device using the prediction result. Yes. Further, in paragraph 0046 of the same document, “FIG. 9 is an example of an evaluation result by the prediction result evaluation unit 18. For each maintenance work performed after learning, the appropriate maintenance work is predicted based on the learning result. Compared to the work actually performed by maintenance personnel, this section deals only with whether the predictions and results match, but it is also new whether the maintenance work used for the evaluation was actually appropriate. It is also possible to evaluate the maintenance work information by dividing it into four categories: (match, appropriate), (mismatch, appropriate), (match, inappropriate), and (mismatch, inappropriate). It is described.

  That is, Patent Document 1 classifies mechanical faults, classifies work performed on the faults, associates the classified mechanical faults with the classified work on a one-to-one basis, There is a description of an appropriate maintenance work analysis device that evaluates a combination divided into a plurality of categories.

JP 2000-251126 A

  In recent elevators that have been digitized, when a malfunction is caused by a combination of mechanical, electrical, and software factors, different countermeasures are required for each combination of malfunctions. In an actual elevator, there are a large number of combinations of defects, so depending on the misunderstanding or skill level of the maintenance staff, inappropriate countermeasures may be implemented. There is also a risk of end.

  In addition, in order to infer an appropriate work procedure when a malfunction occurs in an elevator, all knowledge and experience of the elevator's mechanical system, electrical system, and software are required for each model. It is difficult for all maintenance personnel to acquire for various models. Therefore, there is a demand for a system that can infer an appropriate work procedure by reusing the knowledge and work results of skilled maintenance personnel.

  As shown in FIG. 9 and paragraph 0046 of the same document, the maintenance work of Patent Document 1 evaluates the replacement / adjustment work when a mechanical failure occurs. The appropriate work for is presented. That is, what can be realized by Patent Document 1 is limited to computerization of manuals and rule books for mechanical system replacement / adjustment work. For this reason, the technique disclosed in this document cannot infer a work procedure corresponding to a failure / failure of an elevator system caused by a composite element of failure of an electric system or software.

  In order to deal with problems involving software, the value of a specific data area of the program is changed, and the behavior of the elevator that accompanies the change is confirmed while the program is repeatedly executed and paused. The work history is important, but the technique of Patent Document 1 cannot cope with such work, and it is inefficient when a maintenance person who has little knowledge and experience of mechanical, electrical, and software performs the work. .

  The present invention has been made to solve such problems, and its technical problem is that when a complex failure occurs in an elevator, it is not affected by the knowledge and experience of maintenance personnel. Inferring a series of work procedures necessary for recovery from a failure based on failure information, presenting it to maintenance personnel, and providing it to the controller of the elevator so that the failure can be recovered appropriately. It is.

  In order to solve the above problems, in the elevator system according to the present invention, an elevator controller that controls elevator equipment, a maintenance terminal that provides a maintenance work environment to maintenance personnel, and a series of operation histories that operated the maintenance terminal are stored. Operation history storage means, failure information analysis means for analyzing failure information and outputting an analysis result, learning processing means for outputting a learning result using the operation history and the analysis result as teacher data, and based on the learning result And an inference means for inferring an appropriate work procedure corresponding to the new defect information.

  According to the present invention, by analyzing a newly generated defect using a learned network, it is possible to infer an appropriate countermeasure procedure reflecting the knowledge of skilled maintenance personnel, and the maintenance staff or the controller can follow the countermeasures accordingly. By implementing, you can recover from the failure early.

The figure explaining the whole structure of the elevator system of Example 1. FIG. Learning flow of maintenance operation in the present invention Inference flow of maintenance operation in the present invention Configuration example of elevator system of embodiment 2 Configuration example of elevator system of embodiment 3 Configuration example of elevator system of embodiment 4 Configuration example of elevator system of embodiment 5 Configuration example of elevator system of embodiment 6 Example of defect notification form learning / inference configuration in Example 7 Example of network for learning / inference of defect report form in Example 7 Example of error log learning / inference configuration of Example 8 Example of error log learning / inference network in Example 8 Example 9 of rationality check configuration for maintaining a safe state in Example 9 Process flow of rationality check of embodiment 9

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

  First, the outline | summary of the elevator system of Example 1 is demonstrated using FIG. Note that some or all of the operation history storage unit 50, the defect information analysis unit 60, the learning processing unit 70, and the operation reasoning unit 80, which will be described later, are not necessarily provided with dedicated hardware, but a main storage device such as a semiconductor memory. It may be realized by processing a program stored in the memory or data stored in an auxiliary storage device such as a hard disk with an arithmetic device such as a CPU.

  In FIG. 1, reference numeral 100 a denotes an elevator device such as a hoisting machine that lifts and lowers a car of the elevator 100 and an opening / closing device that opens and closes a car door. Reference numeral 10 denotes an elevator controller that controls the elevator apparatus 100a.

  A maintenance terminal 30 provides a maintenance work environment for the elevator 100 to the maintenance staff 20 and is connected to the elevator controller 10 by wire or wirelessly. The maintenance terminal 30 may be a dedicated terminal specialized for maintenance work, or may be a general-purpose product such as a personal computer or a smartphone that incorporates maintenance work software.

  Between the elevator controller 10 and the maintenance terminal 30, the control data 12 required for control of the elevator apparatus 100a is transmitted / received according to a predetermined communication protocol. In accordance with the input control data 12, the elevator controller 10 notifies the maintenance terminal 30 of the state of each elevator device 100a or causes each of the elevator devices 100a to perform a desired operation.

  Reference numeral 40 denotes a defect information storage unit that stores a defect that has occurred in the elevator 100, and stores past defect information. When a new defect occurs in the elevator 100, the defect information related to the new defect is maintained. It is added by any one of the staff member 20, the elevator controller 10, and the maintenance terminal 30. The defect information storage unit 40 can be realized in various forms, and examples thereof include a database recorded in a storage medium such as a hard disk.

  The defect information stored in the defect information storage unit 40 can take various data formats. For example, text data equivalent to a document expressed in a natural language such as a bug report, program level text data such as an error log, or binary data such as a core dump generated when a program malfunctions However, the data format is not limited to these.

  Reference numeral 50 denotes an operation history storage unit that stores, as an operation history 32, a procedure in which the maintenance staff 20 operates the maintenance terminal 30 in order to cope with a problem. In order to improve the quality as teacher data, it is desirable to store as many operation histories 32 as possible in the operation history storage unit 50.

  A defect information analysis unit 60 analyzes the defect data 42 input from the defect information storage unit 40 and outputs it as an analysis result 62. At the same time, if new defect information is included, new defect data 44 is generated. Is output as For example, when the data format of the defect data 42 is a natural language, lexical phrases such as nouns, verbs, and adjectives extracted therefrom are output as the analysis result 62. Similarly, if the defect data 42 is an error log, an error number or an error tag is output as an analysis result 62, and if the defect data 42 is a core dump, a program name, function name, variable name, and address are output as an analysis result 62. To do.

  A learning processing unit 70 performs learning processing using the operation history data 52 input from the operation history storage unit 50 and the analysis result 62 input from the defect information analysis unit 60 as teacher data. Output. An example of the learning process performed here is deep learning. When deep learning is used, the learning result 72 is generally output as a learned network. Details of the learning processing unit 70 will be described in detail in the eighth and ninth embodiments.

  Reference numeral 80 denotes an operation inference unit that uses the learning result 72 input from the learning processing unit 70 to infer an appropriate operation procedure for eliminating a new defect event expressed as the defect data 44, as operation procedure data 82. Output. Although FIG. 1 illustrates a configuration in which the operation procedure data 82 is input to both the maintenance terminal 30 and the elevator controller 10, a configuration in which the operation procedure data 82 is input to either one may be employed.

  The operation procedure data 82 input to the maintenance terminal 30 is displayed on a display device provided in the maintenance terminal 30. As a result, the maintenance staff 20 can know an operation necessary for recovery from a new defect only by following the operation procedure displayed on the maintenance terminal 30, so even if an unskilled maintenance staff is in charge. Therefore, it is possible to implement the same countermeasures that reflect the knowledge of skilled maintenance personnel.

Note that if the problem is not resolved even according to the operation procedure data 82, the maintenance staff 20 performs a recovery work based on his / her own knowledge and experience. In this case, since the operation procedure is registered in the operation history storage unit 50 and reflected in a new learning process, the subsequent inference accuracy is further increased.
<Modification 1>
Next, a modification of the first embodiment will be described. In addition, duplication description is abbreviate | omitted in common with Example 1. FIG.

  In the first embodiment described above, the maintenance terminal 30 displays the input operation procedure data 82 on the display device, and the maintenance staff 20 works according to the display. However, in this modification, the operation procedure data 82 is displayed. The input maintenance terminal 30 generates the control data 12 without the involvement of the maintenance staff 20 and controls the elevator 100 via the elevator controller 10. That is, this modification can cope with a problem even in an environment where the maintenance staff 20 is absent.

  Specifically, when new defect information is registered in the defect information storage unit 40 by the elevator controller 10, the operation reasoning unit 80 uses the learning result 72 to eliminate the defect event corresponding to the new defect data 44. An appropriate operation procedure is inferred, operation procedure data 82 is generated, and the generated operation procedure data 82 is passed to the maintenance terminal 30. The maintenance terminal 30 generates control data 12 based on the received operation procedure data 82 and controls the elevator 100 via the elevator controller 10.

By adopting such a configuration, when a problem occurs, even if the maintenance staff 20 is not present, the problem can be automatically dealt with, and the time required to resolve the problem can be shortened. In addition, since it is not necessary for the maintenance staff 20 to be dispatched, the workload can be reduced.
<Modification 2>
Next, another modification of the first embodiment will be described. In addition, duplication description is abbreviate | omitted in common with the modification 1.

  In the first modification described above, the maintenance terminal 30 to which the operation procedure data 82 is input generates the control data 12 without the involvement of the maintenance staff 20 and controls the elevator 100 via the elevator controller 10. In this modification, the elevator controller 10 to which the operation procedure data 82 is input controls the elevator 100 without the involvement of the maintenance staff 20. In other words, this modification can cope with a problem even in an environment where the maintenance terminal 30 is not connected.

  Specifically, when new defect information is registered in the defect information storage unit 40 by the elevator controller 10 or the like, the operation reasoning unit 80 uses the learning result 72 to indicate a defect event corresponding to the new defect data 44. An appropriate operation procedure to be eliminated is inferred, operation procedure data 82 is generated, and the generated operation procedure data 82 is passed to the elevator controller 10. The elevator controller 10 controls the elevator 100 based on the received operation procedure data 82.

By adopting such a configuration, when a problem occurs, it becomes possible to automatically cope with the problem with only the elevator 100, and the work of the maintenance staff 20 can be omitted.
<Learning and reasoning>
Next, a learning flow until the learning processing unit 70 outputs the learning result 72 and an inference flow until the operation inference unit 80 outputs the operation procedure data 82 will be described with reference to FIGS. 2A and 2B.

  First, the learning flow will be described with reference to FIG. 2A. When the skilled maintenance worker 20 performs an operation on the maintenance terminal 30 to eliminate the failure event (S1), the operation history storage unit 50 receives an operation history corresponding to the operation of the maintenance worker 20 from the maintenance terminal 30. 32 is received and stored (S2). The defect information analysis unit 60 analyzes the defect data 42 input from the defect information storage unit 40 and obtains an analysis result 62 (S3).

  The learning processing unit 70 acquires the operation history data 52 from the operation history storage unit 50 (S4), and acquires the analysis result 62 such as lexical analysis from the defect information analysis unit 60 (S5). Then, learning such as deep learning is performed using the operation history data 52 and the analysis result 62 as teacher data (S6). By this learning process, a learning result 72 such as a learned network can be obtained.

  Next, the inference flow will be described with reference to FIG. 2B. The operation reasoning unit 80 acquires new defect data 44 from the defect information analysis unit 60 (S10), and acquires a learning result 72 from the learning processing unit 70 (S11). Then, by inputting the new defect data 44 to the learning result 72 (learned network), a series of operation procedures suitable for resolving the defect of the new defect data 44 is inferred (S12). By such an inference process, when a new defect occurs, it is possible to infer operation procedure data 82 for eliminating it.

  According to the present embodiment described above, a new trouble event is newly created by learning the relationship between the information related to the trouble event and a series of work histories performed in the past by the maintenance staff to eliminate the trouble event. When it occurs, it is possible to infer a work procedure corresponding to the new failure event. As a result, the know-how of skilled maintenance personnel is accumulated, and even a non-skilled worker can solve the trouble phenomenon while effectively utilizing the know-how of the skilled worker.

  The outline | summary of the elevator system of Example 2 is demonstrated using FIG. In addition, duplication description is abbreviate | omitted in common with Example 1. FIG.

  In FIG. 3, reference numeral 200 denotes a center that is usually installed in a building different from the building where the elevator 100 is installed. As shown here, in the elevator system of the present embodiment, one elevator system is configured by connecting the elevator 100 and the maintenance terminal 30 via the communication path 110, and connecting the maintenance terminal 30 and the center 200 via the communication path 120. doing. The elevator 100 includes an elevator device 100a and an elevator controller 10, the maintenance terminal 30 includes an operation history storage unit 50, the center 200 includes a defect information storage unit 40, a defect information analysis unit 60, A learning processing unit 70 and an operation reasoning unit 80 are included.

  Thus, by connecting the elevator 100 and the maintenance terminal 30 via the communication path 110, the maintenance staff 20 can operate the elevator 100 via the maintenance terminal 30. In addition, by providing the operation history storage unit 50 in the maintenance terminal 30, the operation history 32 can be stored by the maintenance terminal 30 alone even when the maintenance terminal 30 and the center 200 are not connected.

  On the other hand, by connecting the maintenance terminal 30 and the center 200 through the communication path 120, the maintenance staff 20 transmits the operation history data 52 stored in the operation history storage unit 50 of the maintenance terminal 30 to the learning processing unit 70 of the center 200. It becomes possible. Thereby, when the communication path 120 cannot always be connected, after completing maintenance work in a plurality of buildings, the communication path 120 can be connected and the learning process can be performed collectively.

  In operations during periodic inspection / maintenance visits, maintenance personnel 20 carry maintenance terminals 30 to the site, perform maintenance work, and after returning to center 200, the accumulated work history is often uploaded. Therefore, the structure of the elevator system of the present embodiment is suitable for a general business form.

  According to the configuration of the present embodiment described above, when a malfunction occurs, after the maintenance staff returns to the center 200, the malfunction data 42 is acquired on the center 200 side, and the operation reasoning unit 80 acquires the experienced maintenance staff. It is possible to infer a series of operation procedures reflecting the knowledge and confirm the operation procedure corresponding to the defect.

  The outline | summary of the elevator system of Example 3 is demonstrated using FIG. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above. As shown here, in the elevator system of the present embodiment, the elevator 100 includes the elevator apparatus 100a and the elevator controller 10, and the center 200 includes the defect information storage unit 40, the operation history storage unit 50, and the defect information. An analysis unit 60, a learning processing unit 70, and an operation reasoning unit 80 are included.

  Thus, by connecting the elevator 100 and the maintenance terminal 30 via the communication path 110, the maintenance staff 20 can operate the elevator 100 via the maintenance terminal 30.

  Further, by connecting the maintenance terminal 30 and the center 200 via the communication path 120, the maintenance staff 20 can store the operation history 32 of the maintenance terminal 30 in the operation history storage unit 50 of the center 200.

  In the present embodiment, since the communication path 120 that connects the maintenance terminal 30 and the center 200 is always connected when the maintenance worker 20 is working, the operation history 32 can be sent to the operation history storage unit 50 in real time. There is an advantage that the processing in the learning processing unit 70 can be started early. The maintenance terminal 30 is a form suitable for using a device such as a smartphone that can be always connected. By connecting the elevator 100 and the maintenance terminal 30 via the communication path 110, the maintenance staff 20 can operate the elevator 100.

  The outline | summary of the elevator system of Example 4 is demonstrated using FIG. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above. As shown here, in this embodiment, one elevator system is configured by connecting the elevator 100 and the maintenance terminal 30 via the communication path 110. Moreover, the elevator 100 has the elevator apparatus 100a and the elevator controller 10, and the maintenance terminal 30 includes a defect information storage unit 40, an operation history storage unit 50, a defect information analysis unit 60, a learning processing unit 70, An operation reasoning unit 80 is included. That is, the maintenance terminal 30 of the present embodiment is configured to have all the information and functions necessary for inferring an appropriate operation procedure for eliminating a new malfunction event.

  There are various types of elevators 100 in terms of the maximum speed, the presence or absence of a machine room, etc., and the phenomenon and tendency of failure may vary greatly depending on the type. The operation procedure corresponding to these failures is different for each type. Often different. For this reason, in this embodiment, the maintenance terminal 30 is prepared for each type, and maintenance operation is performed for each type, whereby knowledge for each type is accumulated, and an efficient operation procedure can be inferred for each type.

  The outline | summary of the elevator system of Example 5 is demonstrated using FIG. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above. As shown here, in this embodiment, one elevator system is configured by connecting the elevator 100 and the center 200 through a communication path 130 that is a dedicated connection line. Moreover, the elevator 100 has the maintenance terminal 30 which incorporated the elevator apparatus 100a, the elevator controller 10, and the operation history memory | storage part 50, and the center 200 learned the defect information storage part 40, the defect information analysis part 60, and learning. A processing unit 70 and an operation reasoning unit 80 are included. That is, the maintenance terminal 30 according to the present embodiment is not carried by maintenance personnel, but is built in the elevator 100 and is always connected to the center 200 via the dedicated communication path 130. It is.

  In the present embodiment, since the maintenance terminal 30 is incorporated in the elevator 100, the maintenance staff 20 can complete the maintenance work only by the elevator 100, and the operation history 32 at the time of the maintenance work is stored in the operation history storage unit 50. Can be remembered.

  In this way, the operation history 32 stored in the operation history storage unit 50 can be learned by the learning processing unit 70 by connecting the elevator 100 and the center 200 through the communication path 130.

  Note that many of the online maintenance services for the elevator 100 are often provided via a dedicated line connecting the elevator 100 and the center 200 as in the communication path 130 of this embodiment. Since such a dedicated line generally has a narrow line bandwidth, the operation history 32 stored in the operation history storage unit 50 is stored in the center 200 at a time zone such as midnight when there is little communication such as operation data and abnormal messages. It is desirable to send.

  According to the configuration of the present embodiment described above, since the maintenance terminal 30 is incorporated in the elevator 100, the maintenance staff 20 does not need to carry the maintenance terminal 30 at the site, and the burden on the maintenance staff 20 is reduced. Can be planned.

  The outline | summary of the elevator system of Example 6 is demonstrated using FIG. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above. As shown here, in the elevator system of the present embodiment, the elevator 100 includes an elevator device 100a, an elevator controller 10, a maintenance terminal 30, a defect information storage unit 40, an operation history storage unit 50, and a defect information analysis. Unit 60, learning processing unit 70, and operation reasoning unit 80. That is, this is a form in which all information and functions necessary for inferring an appropriate operation procedure for eliminating a new malfunction event are integrated in the elevator 100.

  According to the configuration of the present embodiment, work can be performed using the maintenance terminal 30 incorporated in the elevator 100 as in the fifth embodiment, so that the maintenance staff 20 does not need to carry the maintenance terminal 30. The burden on the maintenance staff 20 can be reduced.

  Furthermore, when a new malfunction occurs in the state where learning of the maintenance operation in the learning processing unit 70 has progressed using the operation history 32 by the maintenance staff 20 as teacher data, the malfunction information 150 generated by the elevator controller 10 is displayed. On the other hand, by executing a series of learning and inference, a series of operation procedure data 82 suitable for the defect information 150 can be inferred seamlessly in the elevator 100, and the inferred operation procedure data 82 is transmitted to the elevator controller 10. Thus, it is possible to realize a self-healing elevator system in which appropriate processing for the failure is performed simultaneously with the occurrence of the failure.

  The elevator system of Example 7 is demonstrated using FIG. 8A and FIG. 8B. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above.

  The elevator system of the present embodiment shown in FIG. 8A describes the defect information analysis unit 60 of the first embodiment in more detail, and the defect data 42a provided from the defect information storage unit 40 is a natural language such as Japanese. In this embodiment, the natural language analysis process in the natural language analysis unit 60a, which is a kind of the defect information analysis unit 60, and the learning process in the learning processing unit 70 are described. . In FIG. 8A, the same configuration as in FIG. 1 is omitted except for a part.

  In FIG. 8A, defect report form 40a is digital data such as electronic document data created by word processor software or electronic form data of spreadsheet software, and is read from defect information storage unit 40 (not shown).

  The natural language analysis unit 60a performs lexical analysis processing, which is a kind of natural language analysis, on the defect communication form 40a input as the defect data 42a. By the lexical analysis processing here, the defect communication form 40a is decomposed into lexical information 62a of nouns, verbs, and adjectives.

  On the other hand, the operation history data 52 read from the operation history storage unit 50 is a combination of operation data accompanied by operations such as vertical movement and variable data for changing various parameters. In addition, information corresponding to the order and execution interval is also added to the operation history data 52.

  As shown in FIG. 8B, the learning processing unit 70 learns the deep network 72a that receives the lexical information 62a and outputs the operation history data 52. When the deep learning in the learning processing unit 70 is completed, the deep network 72a is updated to the learned deep network.

  Thereafter, when the defect communication form 40a for the newly generated defect is input, the operation reasoning unit 80 infers a series of operation procedures for the defect communication form 40a using the updated deep network 72a, and the operation procedure data It outputs as 82.

  The operation procedure data 82 output from the operation reasoning unit 80 may be displayed on the maintenance terminal 30 as in the first embodiment, or may be automatically executed on the maintenance terminal 30 as in the first modification. It may be automatically executed by the elevator controller 10 as in the second modification.

  According to the elevator system of the present embodiment described above, it can be understood if read by using the deep network 72a reflecting the knowledge of skilled maintenance personnel, but it is difficult to determine how to actually deal with it. Moreover, it is possible to easily derive the appropriate operation procedure data 82 based on the trouble report slip 40a written in a natural language. Thereby, even if it is a non-skilled maintenance worker etc., the recovery operation | work from a malfunction can be performed according to the appropriate operation procedure data 82 in which the knowledge of the skilled maintenance worker was reflected.

  The elevator system of Example 8 is demonstrated using FIG. 9A and FIG. 9B. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above.

  The elevator system of the present embodiment shown in FIG. 9A describes the defect information analysis unit 60 of the first embodiment in more detail, and the defect data 42a provided from the defect information storage unit 40 is executed by the elevator controller 10. When the control program to be output is the error log 40b output when a failure occurs, an error log analysis process in the error log analysis unit 60b, which is a kind of the defect information analysis unit 60, and a learning process in the learning processing unit 70 will be described. This is an example. In FIG. 9A, the same configuration as in FIG. 1 is omitted except for a part thereof.

  In FIG. 9A, the error log 40b is text data in a predetermined format, as illustrated in FIG. 9A, and is read from the defect information storage unit 40 (not shown).

  The error log analysis unit 60b performs a text filter process according to the format of the error log 40b in order to extract necessary data from the error log 40b input as the defect data 42b. By the text filter processing here, the error log 40b is decomposed into text information 62b such as an error number and a warning message.

  On the other hand, the operation history data 52 read from the storage in the operation history storage unit 50 is a combination of operation data accompanied by operations such as up and down movement and variable data for changing various parameters.

  As shown in FIG. 9B, the learning processing unit 70 learns a deep network 72b that receives text information 62b such as an error number and a warning message as input and outputs operation history data 52 as output. When the deep learning in the learning processing unit 70 is completed, the deep network 72b is updated to the learned deep network.

  Thereafter, when the newly generated error log 40b is input, the operation reasoning unit 80 uses the updated deep network 72b to infer a series of operation procedures for the error log 40b, and outputs them as operation procedure data 82. .

  As shown in FIG. 9A, when the failure data 42b is expressed by an error log 40b that is difficult for humans to understand, a normal maintenance person cannot understand a specific failure event just by looking at the error log 40b. Although the operation procedure cannot be predicted to solve the problem, the use of the elevator system according to the present embodiment makes it possible to easily obtain useful operation procedure data 82 in which the knowledge of skilled maintenance personnel is reflected in the error log 40b.

  The elevator system of Example 9 is demonstrated using FIG. 10, FIG. In addition, duplication description is abbreviate | omitted in common with the Example mentioned above.

  As shown in FIG. 1 of the first embodiment, when the operation procedure data 82 output from the operation inference unit 80 is directly input to the elevator controller 10 or the maintenance terminal 30, the operation procedure data 82 is not suitable for the current state of the elevator 100. There is a case. This is because the state of the elevator may be different at the time of occurrence of the malfunction and at present.

  For example, if the operation procedure data 82 includes an operation for raising and lowering the car, if the current elevator 100 is in a state in which the car door is opened, raising and lowering is not safe, and a new problem may occur. is expected. On the other hand, if the current elevator 100 is in a state in which the car door is closed, raising and lowering is a safe operation and no problem occurs.

  Therefore, in the elevator system of the present embodiment, it is confirmed whether the operation procedure data 82 generated by the operation reasoning unit 80 is reasonable with respect to the current elevator state, and if it is reasonable, the elevator controller 10 or the like. If it is not reasonable, we decided to search for a way to eliminate it. In the case of the previous example, the car door is closed before the lifting operation.

  In FIG. 10, 90 is a rationality check unit, and whether the input operation procedure data 82 is reasonable for the current elevator 100, that is, when the elevator 100 is operated according to the operation procedure data 82. This is to confirm whether or not a new defect occurs.

  If the rationality check unit 90 can confirm the safety, the operation procedure data 82 is passed to the elevator controller 10 and executed as it is. On the other hand, when it is determined that it is not safe, the operation procedure data 82 and the failure prediction data 92 representing the predicted failure are passed to the failure avoidance procedure search unit 94. The defect avoidance procedure search unit 94 is connected to a defect avoidance procedure storage device 96 that stores a defect avoidance procedure (rule), searches the defect avoidance procedure storage device 96 for a procedure for solving the predicted defect, and performs a search. By adding the trouble avoidance procedure that has been performed to the operation procedure data 82, new operation procedure data 82a can be generated, registered in the operation history storage unit 50, and used to relearn the learning processing unit 70.

  By adopting such a configuration, even when the current situation of the elevator 100 is different from the situation in which the maintenance staff 20 has worked, by re-learning the operation procedure to which a safety ensuring procedure has been added, safer operation procedure data can be obtained. 82 can be inferred.

  In FIG. 10, the operation procedure data 82 a is input to the operation history storage unit 50, but the operation procedure data 82 a may be input to the elevator controller 10.

  FIG. 11 is a diagram for explaining the processing flow of this embodiment. First, the operation reasoning unit 80 infers and generates operation procedure data 82 corresponding to a defect (S20).

  Next, the rationality check unit 90 checks whether a series of operations included in the operation procedure data 82 is safe (S21). If it is safe, the operation procedure data 82 is transmitted to the elevator controller 10 as it is, and the elevator controller 10 executes a process according to the operation procedure data 82 (S22). When a failure is predicted, the rationality check unit 90 generates failure prediction data 92 indicating the predicted failure (S23).

  Then, based on the defect prediction data 92, the defect avoidance procedure search unit 94 searches for a defect avoidance method predicted from the defect avoidance procedure storage device 96 (S24). Then, the searched avoidance method and the operation procedure data 82 are connected to generate new operation procedure data 82a (S25). New operation procedure data 82a is stored in the operation history storage unit 50. The learning processing unit 70 corrects the learning result 72 in consideration of the new operation procedure data 82a stored in the operation history storage unit 50 (S27). By the above operation, an environment in which a series of work in a safer state can be inferred can be constructed.

  According to the configuration of the present embodiment, it is confirmed whether or not there is a problem in the operation procedure data 82 generated by the operation reasoning unit 80. If there is a problem, a learning process is performed to correct the learning result 72 so as to eliminate the problem. Therefore, it is possible to infer a safer operation procedure for dealing with a failure event.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Moreover, it is possible to add, delete, and replace other configurations with respect to the configuration of each embodiment.

DESCRIPTION OF SYMBOLS 10 Elevator controller 12 Control data 20 Maintenance staff 30 Maintenance terminal 32 Operation history 40 Defect information storage unit 40a Defect communication slip 40b Error logs 42, 42a, 42b, 44, 44a, 44b Defect data 50 Operation history storage unit 52 Operation history data 60 Defect information analysis unit 60a Natural language analysis unit 60b Error log analysis unit 62 Analysis result 62a Lexical information 62b Text information 70 Learning processing unit 72 Learning result 72a, 72b Deep network 80 Operation reasoning unit 82, 82a Operation procedure data 90 Rationality check unit 92 failure prediction data 94 failure avoidance procedure search unit 96 failure avoidance procedure storage device 100 elevator 100a elevator equipment 110, 120, 130 communication path 150 failure information 200 center

Claims (12)

An elevator controller for controlling elevator equipment;
A maintenance terminal that provides maintenance work environment to maintenance personnel;
Operation history storage means for storing a series of operation histories for operating the maintenance terminal;
Defect information analysis means for analyzing defect information and outputting analysis results;
Learning processing means for outputting a learning result using the operation history and the analysis result as teacher data;
An inference means for inferring an appropriate work procedure corresponding to new defect information based on the learning result;
An elevator system comprising: The elevator system according to claim 1,
The operation system inferred by the inference means is displayed on the maintenance terminal. The elevator system according to claim 1,
An elevator system characterized in that the operation procedure inferred by the inference means is executed by the maintenance terminal. The elevator system according to claim 1,
An elevator system in which the operation procedure inferred by the inference means is executed by an elevator controller. The elevator system according to claim 1,
The defect information is written in natural language,
The defect information analyzing means performs lexical analysis processing. The elevator system according to claim 1,
The defect information is a program error log,
An elevator system characterized in that the defect information analysis means performs log analysis processing. In the elevator system as described in any one of Claims 1-6,
Rationality check means for judging the rationality of the operation procedure inferred by the inference means;
If there is a problem in the operation procedure inferred by the inference means, it further has a failure avoidance procedure search means for searching for an avoidance procedure of the problem,
The said learning process means corrects the said learning result by using the said avoidance procedure as teacher data, The elevator system characterized by the above-mentioned. In the elevator system as described in any one of Claims 1-6,
The elevator controller is provided in the elevator,
The operation history storage means is provided in the maintenance terminal,
The defect information analysis means, the learning processing means, and the inference means are provided in a center,
While connecting the elevator and the maintenance terminal through a first communication path,
An elevator system, wherein the maintenance terminal and the center are connected by a second communication path. In the elevator system as described in any one of Claims 1-6,
The elevator controller is provided in the elevator,
The operation history storage means, the defect information analysis means, the learning processing means, and the inference means are provided in a center,
While connecting the elevator and the maintenance terminal through a first communication path,
An elevator system, wherein the maintenance terminal and the center are connected by a second communication path. In the elevator system as described in any one of Claims 1-6,
The elevator controller is provided in the elevator,
The operation history storage means, the defect information analysis means, the learning processing means, and the inference means are provided in the maintenance terminal,
An elevator system in which the elevator and the maintenance terminal are connected by a communication path. In the elevator system as described in any one of Claims 1-6,
The elevator controller, the maintenance terminal, and the operation history storage means are provided in an elevator,
The defect information analysis means, the learning processing means, and the inference means are provided in a center,
An elevator system in which the elevator and the center are connected by a communication path. In the elevator system as described in any one of Claims 1-6,
An elevator system, wherein the elevator controller, the maintenance terminal, the operation history storage means, the failure information analysis means, the learning processing means, and the inference means are provided in an elevator.

Images