CN117235514A

CN117235514A - Elevator fault diagnosis system using cyclic neural network model

Info

Publication number: CN117235514A
Application number: CN202211404678.7A
Authority: CN
Inventors: 徐智慧
Original assignee: Hyundai Elevator Co Ltd
Current assignee: Hyundai Elevator Co Ltd
Priority date: 2022-06-15
Filing date: 2022-11-10
Publication date: 2023-12-15
Also published as: KR20230172091A; KR102702121B1

Abstract

The invention discloses an elevator fault diagnosis system using a cyclic neural network model, which is characterized by comprising: an elevator control panel for controlling the actions of the elevator and recording fault data related to faults in real time; the cloud receives and accumulates fault data recorded on an elevator control panel; a preprocessing unit that performs preprocessing for extracting main data for analyzing the cause of an elevator failure from failure data stored in the cloud, and arranging the main data in a time slot; a marking part for combining the data preprocessed by the preprocessing part with a fault reason label for representing the actual fault reason of the elevator to form a learning data set; a learning modeling unit that performs cyclic neural network learning using the learning data set, and generates a failure diagnosis model for estimating an actual failure cause of the elevator; and a storage management unit configured to store and manage the data preprocessed by the preprocessing unit, the learning data set marked by the marking unit, and the failure diagnosis model generated by the learning modeling unit, separately.

Description

Elevator fault diagnosis system using cyclic neural network model

Technical Field

The present invention relates to an elevator fault diagnosis system using a cyclic neural network model, and more particularly, to an elevator fault diagnosis system that performs machine learning using a cyclic neural network model, grasps an actual cause of a fault from a plurality of fault data continuously generated, and provides it to a maintainer, thereby supporting maintenance work.

Background

Elevator systems are included in various high-rise buildings constructed for residential, business, commercial, etc. purposes to enable passengers entering and exiting the building to move smoothly between floors.

In general, an elevator system is configured to include an elevator car that moves along a hoistway formed in a vertical direction inside a building in a state where passengers are seated, a machine unit that is configured by a motor unit that generates power for lifting and lowering the elevator car, a hoisting machine, and the like, and a control unit that performs control related to the operation of the elevator car.

On the other hand, maintenance of the elevator is achieved by: upon receiving an inspection request or a failure report for an elevator provided in a building, a maintenance person goes out to the site where the elevator is provided and diagnoses the state of the elevator, after which necessary measures are taken.

Typically, an elevator system is provided with sensing means that sense the operating state of the elevator system and, when a fault occurs, transmit a fault code indicating the type of fault to a maintenance center. When the elevator fails, the sensing device transmits a fault signal containing a fault code to a maintenance center, and the maintenance center receiving the fault information sends maintenance personnel to the site of the failed elevator to process the fault. The trouble code refers to a number or a combination of a number and an english letter outputted from an elevator control panel when an elevator malfunctions.

More specifically, the process of handling the elevator fault is observed, when the fault report of the elevator is received, a maintainer moving to the elevator site where the fault occurs connects the maintenance terminal to the elevator control panel, confirms the generated fault code, refers to the manual of the fault code, and takes measures according to the corresponding recovery instruction.

However, since the difficulty of failure analysis varies from one failure to another due to the characteristics of an elevator system to which various devices are organically connected, the time required for failure analysis and problem solving may vary depending on the skill level of the maintenance personnel.

In addition, in general, when an elevator fails, a plurality of fault codes are generated at the same time in most cases, not a single fault code, and therefore, it is difficult for unskilled maintenance personnel to accurately grasp the cause of an actual fault, which becomes an obstacle factor for rapidly handling the fault, and a main cause for increasing the downtime of the elevator.

That is, in the conventional elevator maintenance, there is a risk that, when a failure error analysis occurs due to insufficient skill of a maintenance person, there is a large difference in failure cause analysis and processing time depending on the skill of the worker, and if the worker cannot solve the problem, it is necessary to mobilize other manpower of the research institute, and thus there is a high possibility that additional manpower cost and time cost are consumed.

Disclosure of Invention

Technical problem to be solved

The operation data generated in various components (e.g., a hoisting machine motor, a brake, a door, an inverter, a control panel, etc.) constituting the elevator system has a characteristic of continuous generation. As described above, in order to accurately analyze a fault occurring in an elevator, it is necessary to confirm the content of data generated before and after the point in time when the fault data is generated, classify the range, and comprehensively analyze the data, because of the characteristics of the elevator system that continuously generates the operation data, but there is no clear standard for classifying the range.

Accordingly, an object of the present invention is to provide a fault diagnosis system capable of accurately and quickly grasping an actual cause of a fault from a plurality of fault data generated in succession by using a recurrent neural network model capable of learning time series data, and supporting maintenance work, thereby enabling a reduction in fault analysis errors regardless of the proficiency of a maintenance person and a reduction in time required for solving a problem.

Further, the present invention is directed to defining a new cause class indicating the cause of an actual failure based on learning of a continuous sequence of failure codes observed at the time of failure of an elevator, and providing the new cause class to a maintainer, thereby reducing the level of skill of the maintainer, reducing the possibility of occurrence of errors in failure analysis, and improving the stability of maintenance.

The technical problems of the present invention are not limited to the above-mentioned technical problems, and other technical problems not mentioned can be clearly understood by those skilled in the art from the following description.

Means for solving the problems

According to an aspect of the present invention for achieving the above object, there may be provided an elevator fault diagnosis system using a recurrent neural network model, comprising: an elevator control panel for controlling the actions of the elevator and recording fault data related to faults in real time; a cloud receiving and accumulating the fault data recorded on the elevator control panel; a preprocessing unit configured to perform preprocessing for extracting main data for analyzing a cause of a failure of the elevator from the failure data stored in the cloud, and arranging the main data in a time slot; a marking unit that combines the data preprocessed by the preprocessing unit with a failure cause tag indicating an actual failure cause of the elevator to form a learning data set; a learning modeling unit that performs cyclic neural network learning using the learning data set, and generates a fault diagnosis model for estimating an actual cause of a fault in the elevator; and a storage management unit configured to store and manage the data preprocessed by the preprocessing unit, the learning data set marked by the marking unit, and the failure diagnosis model generated by the learning modeling unit, separately.

The main data may be fault codes included in the fault data, and the preprocessing section may arrange the fault codes in time periods and perform grouping of codes located within a set similar time range into a group, and separate the fault codes into a continuous sequence.

The failure cause tag may be defined as a cause class indicating a failure cause of the elevator that actually occurs when the continuous sequence of the failure code is observed as a specific pattern, and the marking section may perform a marking operation of combining the continuous sequence of the failure code with the corresponding cause class with reference to past failure history data.

The fault diagnosis model may receive as input data pre-processed data of fault data of a specific elevator and output the corresponding cause class as a fault diagnosis result.

Repeated learning may be performed using data input during the fault diagnosis of the specific elevator and an output result according to the data as feed data, thereby updating the fault diagnosis model.

The fault diagnosis system according to an aspect of the present invention may further include a web server accessible through a user terminal and providing a user with a fault diagnosis result of the specific elevator selected by the user.

The web server may provide a web application including a service screen for the user to select the particular elevator and visually display the fault diagnosis results.

A two-way long and short term memory (LSTM: bidirectional Long Short Term Memory) may be used as the recurrent neural network model.

Effects of the invention

According to the present invention as described above, by applying the cyclic neural network model suitable for the characteristics of the elevator continuously generating the operation data, it is possible to derive an accurate diagnosis result for the fault occurring in the elevator and provide the result to the maintenance personnel using the advantage of deep learning, thereby having an effect of being able to reduce the time deviation required for the fault analysis and the problem solving regardless of the proficiency of the maintenance personnel.

In addition, the present invention has an effect of remarkably improving the accuracy of elevator fault analysis compared with the conventional method, because the present invention does not diagnose a fault with only one fault data, but finds out the actual cause of elevator fault by comprehensively analyzing a plurality of fault data occurring in time sequence.

In particular, according to the present invention as described above, it is possible to reduce the error analysis concerning the failure occurring in the elevator system, and also to provide a failure whose cause is unknown, which cannot be explained by the existing failure code, or a failure related to the elevator components, as a new cause class, thereby expanding the range of the failure analysis.

The effects of the present invention are not limited to the above-mentioned effects, and the effects not mentioned or other effects can be clearly understood by those skilled in the art from the following description.

Drawings

Fig. 1 is a schematic view showing the structure of an elevator fault diagnosis system according to the present invention.

Fig. 2 is a diagram showing an example of a marking operation between a failure code sequence and a cause class performed by the elevator failure diagnosis system according to the invention.

Description of the reference numerals

10: elevator control panel

20: cloud

30: pretreatment unit

40: storage management unit

50: marking part

60: learning modeling unit

70: network server

80: user terminal

Detailed Description

The objects, technical structures, and the operations and effects associated therewith of the present invention will be more clearly understood from the detailed description of the present invention based on the accompanying drawings in the specification of the present invention.

The terminology used in the description presented herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. For example, the terms "comprising … …" or "including" and the like used in this specification should not be construed as necessarily including the plurality of components or steps described in the present invention, as including a part of the components or a part of the steps therein, or as including additional components or steps. Furthermore, as used in this specification, the singular includes the plural unless the context clearly indicates otherwise.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The embodiments described below are provided to enable a person skilled in the art to easily understand the technical idea of the present invention, and should not be construed as limiting the present invention, and it is needless to say that the embodiments of the present invention can be applied to various applications to those skilled in the art.

Fig. 1 is a schematic view showing the structure of an elevator fault diagnosis system according to the present invention. Fig. 2 is a diagram showing an example of a marking operation between a failure code sequence and a cause class performed by the elevator failure diagnosis system according to the invention.

Referring to fig. 1, an elevator fault diagnosis system according to the present invention may include: an elevator control panel 10 that controls actions related to the operation of the elevator and records various operation data of the elevator including fault data related to a fault in real time; a cloud 20 that receives and accumulates fault data recorded in the elevator control panel 10; a preprocessing unit 30 that reads data from the cloud 20 and preprocesses the data into an input data format of a learning model; a storage management unit 40 that stores the data preprocessed by the preprocessing unit 30; a marking unit 50 that combines the preprocessed data stored in the storage management unit 40 with the failure cause tag to form a learning data set, and stores the learning data set in the storage management unit 40 again; a learning modeling unit 60 that acquires the labeled learning data set from the storage management unit 40, performs cyclic neural network learning to generate a learning model, and stores the learned model again in the storage management unit 40; and a web server 70 accessible through the user terminal 80 and providing an analysis result of a learning model having as input failure data generated by a specific elevator selected by a user as a failure diagnosis result.

The elevator control panel 10 can perform drive control related to the operation of the elevator and record parameters related to the actions of the various devices constituting the elevator and the control panel itself as operation data in real time.

For this purpose, the elevator control panel 10 may include a sensing device that periodically senses the operation state of the elevator, and collects signals related to the operation of the elevator as operation histories for the respective systems, respectively. The sensing means may receive data measured by a plurality of sensors that collect parameters related to the actions of the various devices that make up the elevator.

When the elevator control panel 10 detects a failure of a device included in the elevator, failure data including a failure code for identifying the failed device may be generated, and various operation data of the elevator including the failure data are transmitted from the elevator control panel 10 to the cloud 20 to be accumulated and stored.

In the cloud 20, various operation data recorded in the elevator control panel 10 can be collected at all times, and in particular, information including time information on occurrence of failure data related to failure of the elevator can be collected and accumulated.

The preprocessing unit 30 defines and extracts main data that affects analysis of the cause of occurrence of a fault from the raw data collected in the cloud 20, and preprocesses the main data into an input data format suitable for a learning model so as to be suitable for learning of a recurrent neural network. Here, "preprocessing into a form of input data suitable for the learning model" may refer to performing a function of converting into data of a form suitable for input to the learning model generated by the learning modeling section 60.

In the above description, "raw data" may be indicative of only fault data including a fault code, or may be indicative of the entire operation data of the elevator including the fault data. The failure data is input/output data generated when the elevator fails, and may include data exchanged between a main board and an inverter, a door inverter, and a console box (COP; car Operating Panel) board used in button operations, and the like.

The preprocessing performed by the preprocessing section 30 may include: extracting main data influencing analysis of failure occurrence reasons of the elevator from the original data; sorting the extracted main data based on the time of occurrence of the fault; and separating the ordered primary data into a sequence suitable for input into the learning model. As a preprocessing method that can be used at this time, general methods such as error filtering, data interpolation, over/under sampling, dimension reduction, and the like can be used.

As described below, the present invention is characterized in that an actual failure cause is deduced from a continuous sequence of a plurality of failure codes occurring simultaneously to define a new cause class, and in the above description, "main data having an influence on failure cause analysis" may refer to a failure code. That is, the preprocessing section 30 may extract the fault code as the main data from the fault data collected by the cloud 20.

Further, the preprocessing section 30 may sort the extracted trouble codes according to occurrence time and separate into sequences suitable for inputting the learning model. At this time, the preprocessing section 30 can separate the sequences by executing a packet (grouping) that groups the trouble codes that occur in the same period of time.

In the above, the "same period of time" does not mean to exactly coincide to hours/minutes/seconds, but may be interpreted as grouping data generated in relatively close periods of time into one group. For example, data generated in a time range of several seconds or several minutes may be grouped into one group, and setting of a time range as a grouping reference may be a matter of design of the present system, which may be changed.

The storage management section 40 may store the data preprocessed by the preprocessing section 30. Wherein the "preprocessed data" may be a set of fault codes arranged and grouped according to time periods as described above.

Further, the storage management section 40 may store and manage a learning data set marked by a marking section 50 described below, a learning model generated by the learning modeling section 60, and a model result (diagnosis result) output when the preprocessed data of the failure data generated at a specific elevator is input to the learning model.

As described below, the elevator fault diagnosis system according to the present invention may be divided into a "model learning process" and a "fault diagnosis process" to perform processes, the storage management section 40 may divide and manage learning data through the model learning process and input data through the fault diagnosis process, and periodically update data input during fault diagnosis and results of the learning model corresponding thereto to update and manage the learning model.

The marking section 50 may read the data preprocessed by the preprocessing section 30 from the storage management section 40 and combine it with the failure cause tag, thereby constituting a model-learned data set.

In the above description, the "failure cause tag" refers to a category indicating an actual elevator failure cause predicted when a specific pattern of the preprocessed data, that is, a specific pattern of the failure code sequence grouped in time periods is observed, and the present invention defines it as a new "cause category".

The "cause class" described in the present invention is a different concept from the previously defined "trouble code". If an existing fault code indicates the possibility of a fault of the individual devices constituting the elevator, the cause class is a specification of a new fault cause, undefined in the prior art, indicating in which part of the elevator the fault actually occurred when a continuous sequence of a plurality of fault codes was observed.

With reference to fig. 2, actions for performing a marking operation between a fault code sequence and a cause class with respect to the elevator fault diagnosis system according to the invention will be described in more detail.

Referring to fig. 2, first, at 11:02: 09-11: 02: the Fault code faults (Fault) a, fault B, fault F generated during the 10 time period may be grouped into a group, at 11:10:41 to 11:10: the fault codes fault D, fault E, fault B generated for the 42 time period may be grouped into another group. Such grouping operation is a state that has been performed by the preprocessing section 30.

In this case, a marking operation for combining the sequence of the fault codes grouped into one group in correspondence with the new cause class may be performed by the marking section 50. For example, the marking section 50 may combine the consecutive sequence of the fault codes generated in the order of the fault a→the fault b→the fault F with the Cause (Cause) a Cause class, and combine the consecutive sequence of the fault codes generated in the order of the fault d→the fault e→the fault B with the Cause B Cause class. Wherein the cause a and the cause B respectively prescribe the actual cause of the failure of the elevator.

At this time, the marking unit 50 may perform the association (marking) between the preprocessed data and the failure cause tag (cause type) with reference to the past failure history data stored in the cloud 20 or a failure server provided separately, and the data accumulated with the repeated learning by the learning modeling unit 60 described later may be used again as feed data (feed data).

As described above, it can be understood that the marking section 50 performs a function of mapping (mapping) the preprocessed data of the failure data observed in the failure state of the elevator with the actual failure cause, and the recurrent neural network learning can be performed using the learning data set constituted by the data marking.

The learning modeling section 60 may generate a "failure diagnosis model" by performing machine learning of a specific algorithm using the learning data set constituted by the labeling section 50.

The learning modeling section 60 may utilize a "recurrent neural network algorithm" that learns time information that is a main feature of input data, and as failure data is accumulated, additional learning may be periodically repeated and compared with an existing model to perform an operation of updating the model.

As described above, since the operation data of the elevator including the failure data has an attribute as time-series data that is continuously generated, it is difficult to grasp an accurate cause of the failure with only a single data, and the cause of the failure can be accurately analyzed only by comprehensively judging and analyzing other data generated before/after the point of time of the failure occurrence.

Therefore, the present invention selects a recurrent neural network model (RNN: recurrent Neural Network) as a learning model most suitable for deducing the cause of an actual failure by analyzing failure data generated in time series. The recurrent neural network algorithm is a model mainly applied to a data set having correlation between front/rear data as a neural network class modeling sequence data such as a time series.

In addition, the present invention may use a two-way long and short Term Memory (LSTM: bidirectional Long Short-Term Memory) as one of the recurrent neural network models in the course of executing the learning model generation algorithm. Due to the variability of the separate sequence lengths, a bi-directional LSTM is a state of the art LSTM that complements longer lengths with lower accuracy, and if the existing LSTM is only capable of processing forward data, in the case of a bi-directional rectangular LSTM, it may be a recurrent neural network algorithm capable of processing both forward and reverse bi-directional data.

More specifically, the learning modeling unit 60 may generate a learning model that automatically derives a cause class indicating an actual cause of a fault corresponding thereto when the preprocessed data of the newly generated fault data is input by repeatedly learning a learning data set composed of a combination of a sequence of fault codes generated in the same time period and the cause class corresponding to the sequence.

Therefore, as described below, when a user connected to the network server 70 through the user terminal 80 selects a specific elevator number, the learning modeling section 60 may receive the preprocessed data of the fault data occurring in the selected elevator as input data of the learning model to derive a result, and the result derived by the learning modeling section 60 may be stored in the storage management section 40 as a fault diagnosis result.

The web server 70 may receive input of a specific elevator number required by the user and provide the user with a learning model result as a failure diagnosis result of the corresponding elevator. The web server 70 may be accessed through the portable user terminal 80 and may provide a web application including a service screen for a user to select a specific elevator and may visually display the fault diagnosis result of the selected elevator. Wherein the "user" may be a maintenance person provided with the fault diagnosis service.

For reference, the fault code occurring in the elevator only indicates the possibility of a fault, and the fault code occurrence itself does not indicate that the elevator is actually faulty. Therefore, it is difficult to find the exact failure cause of the elevator only by the existing failure code, but the present invention has the following advantages: by the cyclic neural network learning, a fault diagnosis model is generated that deduces the actual cause of the fault from the sequence of continuously occurring fault codes, so that the actual cause of the fault of the elevator can be diagnosed with high accuracy.

On the other hand, the elevator fault diagnosis system using the recurrent neural network model according to the present invention may generally include a "model learning process" and a "fault diagnosis process" to perform actions. The model learning process is a process involving an operation of generating a failure diagnosis model, and the failure diagnosis process is a process involving an operation of diagnosing a cause of a failure actually occurring in an elevator using the generated learning model.

Hereinafter, the process actions of the elevator fault diagnosis system according to the present invention will be described, respectively.

Model learning process

First, the action of the "model learning process" of the elevator fault diagnosis system according to the present invention can be performed by the following process.

1. Records including fault data recorded in the elevator control panel 10 are accumulated in the cloud 20.

2. The preprocessing unit 30 reads data from the cloud 20, preprocesses the data into the form of input data of the learning model, and stores the data in the storage management unit 40.

3. The marking unit 50 combines the data preprocessed in response to the model input with the failure cause tag (cause type) to construct a learning data set, and stores the learning data set again in the storage management unit 40.

4. The learning modeling unit 60 obtains the labeled learning data set, learns the recurrent neural network model, generates a failure diagnosis model, and stores the failure diagnosis model again in the storage management unit 40.

Fault diagnosis process

Next, the action of the "failure diagnosis process" of the elevator failure diagnosis system according to the invention is performed by the following process.

1. The user (maintenance person) connects to the network server 70 through the user terminal 80 and selects an elevator to be subjected to failure analysis.

2. The preprocessing unit 30 reads the failure data of the selected elevator from the cloud 20, preprocesses the failure data into the input data format of the learning model, and stores the data in the storage management unit 40.

3. The learning modeling unit 60 inputs the preprocessed data stored in the storage management unit 40 as input data of the learning model, and stores the failure diagnosis result, which is a result derived from the learning model, in the storage management unit 40.

4. The web server 70 receives the failure diagnosis result from the storage management section 40 and displays it on the service screen of the user terminal 80, thereby providing the user with information indicating the actual cause of the failure of the elevator.

As described above, the elevator fault diagnosis system according to the present invention can generate a model for deducing the cause of an actual fault from a sequence of continuously generated fault codes through cyclic neural network learning, and when a maintenance worker inputs a desired fault analysis object elevator, provide the fault diagnosis result to the user terminal 80 to be able to confirm the fault diagnosis result on site, thereby supporting maintenance work.

According to the present invention as described above, by applying the recurrent neural network model suitable for the characteristics of the elevator in which the operation data continuously occurs, it is possible to derive an accurate diagnosis result regarding the failure occurring in the elevator by utilizing the advantage of the deep learning and provide the result to the maintenance personnel, thereby having an effect that the time deviation required for the failure analysis and the problem solving can be reduced regardless of the proficiency of the maintenance personnel.

In particular, according to the present invention as described above, not only can erroneous analysis concerning a fault occurring in an elevator system be reduced, but also there is an effect that a fault whose cause is unknown, which cannot be explained by an existing fault code, or a fault related to elevator components is provided as a new cause class, and the scope of the fault analysis is widened.

The elevator fault diagnosis system according to the present invention can be implemented by a computer device provided with one or more processors and storing one or more programs executed by the one or more processors. For this purpose, the elevator fault diagnosis system according to the invention can be implemented in the form of a program or software comprising one or more computer-executable instructions stored in a memory.

The present invention is not limited to the embodiments described, and various modifications and variations may be made without departing from the spirit and scope of the invention, as will be apparent to those of ordinary skill in the art. Therefore, such modifications and variations are intended to fall within the scope of the appended claims.

Claims

1. An elevator fault diagnosis system using a recurrent neural network model, comprising:

an elevator control panel for controlling the actions of the elevator and recording fault data related to faults in real time;

a cloud receiving and accumulating the fault data recorded on the elevator control panel;

a preprocessing unit configured to perform preprocessing for extracting main data for analyzing a cause of a failure of the elevator from the failure data stored in the cloud, and arranging the main data in a time slot;

a marking unit that combines the data preprocessed by the preprocessing unit with a failure cause tag indicating an actual failure cause of the elevator to form a learning data set;

a learning modeling unit that performs cyclic neural network learning using the learning data set, and generates a fault diagnosis model for estimating an actual cause of a fault in the elevator; and

and a storage management unit configured to store and manage the data preprocessed by the preprocessing unit, the learning data set marked by the marking unit, and the failure diagnosis model generated by the learning modeling unit, separately.

2. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 1, wherein,

the primary data is a fault code contained in the fault data,

the preprocessing section arranges the trouble codes in time periods and performs grouping that groups codes lying within a set similar time range into a group, while separating the trouble codes into a continuous sequence.

3. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 2, wherein,

the fault reason tag is specified as a cause class, which represents the fault cause of the elevator that actually occurs when the continuous sequence of fault codes is observed as a specific pattern,

the marking unit performs a marking operation for combining the continuous sequence of the fault codes with the corresponding cause class with reference to past fault history data.

4. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 3, wherein,

the fault diagnosis model receives preprocessing data of fault data of a specific elevator as input data and outputs the corresponding cause class as a fault diagnosis result.

5. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 4, wherein,

repeated learning is performed with data input during the fault diagnosis of the specific elevator and output results according to the data as feed data, thereby updating the fault diagnosis model.

6. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 5, wherein,

also included is a network server accessible through a user terminal and providing a user with a result of fault diagnosis of the particular elevator selected by the user.

7. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 6, wherein,

the web server provides a web application including a service screen for the user to select the particular elevator and visually display the fault diagnosis results.

8. The elevator failure diagnosis system using a recurrent neural network model as claimed in claim 1, wherein,

and using a two-way long-short term memory as the recurrent neural network model.