KR102301011B1 - System for failure prediction of elevator door - Google Patents

Abstract

The present invention relates to a failure prediction system for an elevator door, comprising a plurality of sensors installed in the elevator door, and a door failure prediction server connected via an elevator control panel and a gateway, wherein the door failure prediction server includes the plurality of sensors and a data collection unit that collects data related to the failure of the elevator door in real time from the elevator control panel, a preprocessor that preprocesses the collected data and extracts features to configure a learning data set, and guides with the learning data set It is characterized in that it is configured to include a learning and model generation unit for generating a failure prediction model by performing learning and unsupervised learning, and a failure prediction unit for determining the possibility of failure of the door through the failure prediction model.

Description

Elevator door failure prediction system {SYSTEM FOR FAILURE PREDICTION OF ELEVATOR DOOR}

The present invention relates to a failure prediction technology of an elevator door, and more particularly, an elevator that collects data related to door failure to generate a predictive model and predicts and responds to failure of an elevator door in advance through the failure prediction model It relates to a door failure prediction system.

2. Description of the Related Art In general, various types of high-rise buildings constructed for residential, business, and commercial purposes are provided with an elevator device for smooth inter-floor movement of passengers entering and exiting the building.

The elevator device is equipped with an elevator car that moves passengers while moving up and down along a hoistway formed in a vertical direction inside a building while a passenger is on board, and a motor unit and a hoisting machine that generate a predetermined power. It is configured to include a machine unit for moving the car to the corresponding floor according to the passenger's button operation, and an elevator control unit for controlling the machine unit according to the passenger's button operation while controlling the elevator car so that the elevator car can be operated smoothly and stably.

On the other hand, when a failure occurs in the elevator device, even if the failure is diagnosed remotely, elevator maintenance personnel can only go to the site and repair the failure after the failure occurs, so passengers cannot use the elevator during the repair time. There was a problem being

In particular, in a situation where there are many high-rise buildings like these days, even a short downtime can cause great inconvenience to users, so it is very important to predict elevator failure. Since more than 50% of elevator failures occur in the door, predicting door failure is more important than any other part.

Recently, due to the development of machine learning and deep learning technology, predictive maintenance technology, which was difficult in the past, is being introduced in various fields, but the application of elevators is still insufficient. In particular, the application of machine learning and deep learning requires a lot of sensor measurement data, especially in the event of a failure. Therefore, new technologies are needed to solve these problems.

Korea Patent Publication No. 10-0179886 (1998.11.28) relates to an elevator failure detection device, and it expands the types of failures that can be automatically detected by statistically analyzing the instantaneous values of elevator state variables and their histories. And, by analyzing the history of anomalies, it is possible to predict elevator failures in advance and prevent failures in advance so that passengers can safely and comfortably use the elevators.

Korean Patent Application Laid-Open No. 10-2017-0078943 (2017.07.10) relates to a failure prediction system of an elevator, and includes a vibration meter installed in a certain place of the elevator to measure vibration, and a vibration meter connected to the vibration meter to detect or A simple configuration consisting of a controller having a built-in program, circuit board or communication module to be transmitted, and a display device connected by the controller to convert the frequency measured by the vibration meter into data and display it visually or sound an alarm By understanding the operating status of the elevator by data analysis and data analysis, it is possible to quickly process the inspection of the main components of the device or replacement of parts.

Korean Patent Publication No. 10-0179886 (January 28, 1998) Korean Patent Publication No. 10-2017-0078943 (2017.07.10)

An embodiment of the present invention is to provide a failure prediction system of an elevator door that collects data related to door failure to generate a predictive model and predicts and responds to failure of the elevator door in advance through the failure prediction model.

An embodiment of the present invention generates a failure prediction model by learning through a variety of various data collected through sensors and a control panel and determines the possibility of failure of the door through the generated failure prediction model, so that the failure prediction is flexible for various situations. It is intended to provide an elevator door failure prediction system that can improve the prediction performance by performing

An embodiment of the present invention is to provide a failure prediction system of an elevator door that configures a supervised learning data set by including a label in actual failure data and generates a predictive model using a DNN (Deep Neural Network) algorithm.

An embodiment of the present invention configures a training data set that does not include a label and uses an LSTM (Long-Short Term Memory) autoencoder anomaly detection algorithm to generate a predictive model. Elevator door failure prediction system would like to provide

In embodiments, the elevator door failure prediction system includes a plurality of sensors installed in the elevator door and a door failure prediction server connected to the elevator control panel and the gateway via a gateway, wherein the door failure prediction server includes the plurality of sensors And a data collection unit for collecting data related to the failure of the elevator door from the elevator control panel in real time, a preprocessing unit for preprocessing the collected data and extracting features to configure a learning data set, supervised learning with the learning data set and a learning and model generation unit that performs unsupervised learning to generate a failure prediction model, and a failure prediction unit that determines a failure possibility of a door through the failure prediction model.

The data collection unit may collect sensing data including the speed of the elevator door through the plurality of sensors and operation data and user manipulation data related to opening and closing of the elevator door through the elevator control panel.

The pre-processing unit removes the data generated by the user's action through the pre-processing of the collected data, and configures a learning data set including the speed, acceleration, and torque pattern of the door by making one cycle from the start of the door operation to the completion of the operation. can

The pre-processing unit removes data generated by the user's operation through the pre-processing of the collected data, and then analyzes the operation sequence of the door in consideration of the user's button operation, safety edge, and multi-beam state to configure a learning data set. .

The learning and model generation unit may combine the failure state and steady state labels with the learning data set to configure a supervised learning data set, and may generate a predictive model by supervised learning the supervised learning data set using a DNN algorithm.

The learning and model generation unit labels the learning data set through mapping between the failures automatically collected due to door inverter errors and actual failures caused by foreign matter caught or poor connection to configure the supervised learning data set. have.

The training and model generator may generate a predictive model by unsupervised learning on the training data set using an LSTM autoencoder anomaly detection algorithm.

The disclosed technology may have the following effects. However, this does not mean that a specific embodiment should include all of the following effects or only the following effects, so the scope of the disclosed technology should not be construed as being limited thereby.

The elevator door failure prediction system according to an embodiment of the present invention collects data related to door failure to generate a predictive model, predicts elevator door failure through the failure prediction model in advance, and enables preparation for this. , it is possible to greatly reduce the inconvenience of users by minimizing downtime due to the failure of the elevator door, and it is possible to reduce the occurrence of malfunction of the elevator door.

Elevator door failure prediction system according to an embodiment of the present invention is flexible for various situations in which elevator door failure may occur by learning using various data collected through sensors and control panel and generating a failure prediction model. Failure prediction can be performed to improve prediction performance.

Elevator door failure prediction system according to an embodiment of the present invention by including a label in the actual failure data to construct a supervised learning data set and generate a predictive model using a DNN (Deep Neural Network) algorithm, Data labeling in supervised learning can be facilitated by mapping component failures with easily observable and automatically diagnosed failures.

Elevator door failure prediction system according to an embodiment of the present invention configures a training data set that does not include a label, and predicts a model using a Long-Short Term Memory (LSTM) autoencoder anomaly detection algorithm. By generating, it is possible to generate a failure prediction model with only a large number of normal data and a small number of failure data.

1 is a view showing a system for predicting failure of an elevator door according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining a label mapping process of a training data set in the training and model generation unit of FIG. 1 .
FIG. 3 is a diagram for explaining a DNN algorithm used for supervised learning in the learning and model generation unit of FIG. 1 .
FIG. 4 is a diagram for explaining an anomaly detection algorithm used for unsupervised learning in the learning and model generation unit of FIG. 1 .
FIG. 5 is a diagram for explaining an LSTM autoencoder anomaly detection algorithm used for unsupervised learning in the learning and model generation unit of FIG. 4 .
6 is a flowchart for explaining a failure prediction process in the failure prediction system of the elevator door in FIG. 1 .

Since the description of the present invention is merely an embodiment for structural or functional description, the scope of the present invention should not be construed as being limited by the embodiment described in the text. That is, since the embodiment may have various changes and may have various forms, it should be understood that the scope of the present invention includes equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only such effects, it should not be understood that the scope of the present invention is limited thereby.

On the other hand, the meaning of the terms described in the present application should be understood as follows.

Terms such as “first” and “second” are for distinguishing one component from another, and the scope of rights should not be limited by these terms. For example, a first component may be termed a second component, and similarly, a second component may also be termed a first component.

When a component is referred to as being “connected” to another component, it may be directly connected to the other component, but it should be understood that other components may exist in between. On the other hand, when it is mentioned that a certain element is "directly connected" to another element, it should be understood that the other element does not exist in the middle. Meanwhile, other expressions describing the relationship between elements, that is, “between” and “immediately between” or “neighboring to” and “directly adjacent to”, etc., should be interpreted similarly.

The singular expression is to be understood to include the plural expression unless the context clearly dictates otherwise, and terms such as "comprises" or "have" refer to the embodied feature, number, step, action, component, part or these It is intended to indicate that a combination exists, and it should be understood that it does not preclude the possibility of the existence or addition of one or more other features or numbers, steps, operations, components, parts, or combinations thereof.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Terms defined in general used in the dictionary should be interpreted as being consistent with the meaning in the context of the related art, and cannot be interpreted as having an ideal or excessively formal meaning unless explicitly defined in the present application.

1 is a view showing a failure prediction system of an elevator door according to an embodiment of the present invention.

1, the elevator door failure prediction system 100 predicts door failure using data collected through a plurality of sensors 110 and elevator control panel 120 installed at various points including elevator car doors and platform doors. While managing and learning in the server 130, it is implemented so that an abnormal operation of an elevator door or a failure of a part can be predicted in advance, and then appropriately responded accordingly.

To this end, in the elevator door failure prediction system 100 according to an embodiment, the sensors 110 and the elevator control panel 120 installed in the field, such as the elevator car door and platform door, collect, manage, and learn data. It is configured in connection with the door failure prediction server 130 that performs a function of predicting and reporting a failure through the

Here, the door failure prediction server 130 includes a data collection unit 131 , a data storage unit 132 , a preprocessor 133 , a learning and model generation unit 134 , and a failure prediction unit 135 , etc. is composed

The data collection unit 131 receives and receives various measurement data from a plurality of sensors 110 installed in the elevator door via a gateway, and receives and receives data related to the operation of the elevator door from the elevator control panel 120 . It performs the receiving function.

In one embodiment, the data collection unit 131 collects at least one of real-time change data measured by the sensor 110 and door opening/closing related motion data and user manipulation data transmitted from the control panel 120 as door failure related data. can do. Here, the door failure related data may be used to generate a failure prediction model. What kind of data is properly collected in creating a failure prediction model is the most important. In the prior art, door failure was diagnosed or predicted using only one data, but here, a failure can be predicted by using a plurality of data complexly.

The data storage unit 132 performs a function of storing and providing measurement data from each sensor 110 collected through the data collection unit 131 and the operation and user operation data of the elevator door from the control panel 120 . do.

The pre-processing unit 133 performs a function of pre-processing the data stored in the data storage unit 132 to configure the learning data set.

In an embodiment, the pre-processing unit 133 may perform pre-processing on data stored in the data storage unit 132 and extract features. For the preprocessing performed here, general methods such as error filtering, data interpolation, over/under sampling, and dimensionality reduction may be applied. The preprocessor 133 may configure a learning data set by extracting features through preprocessing. In an embodiment, the preprocessor 133 may configure a learning data set with data actually sensed by the sensor 110 , and may configure a learning data set with a door operation sequence of the control panel 120 . Here, the pre-processing unit 133 may pre-process the sensing data at that time from the start of the operation of the elevator door to the completion of the operation as one cycle to configure the learning data set for the normal state and the abnormal state. For example, it is important to properly filter such data because there are cases in which doors are incompletely opened and closed due to, for example, a user's operation of a button on a platform hall or a button in an elevator car or an interrupt. The preprocessor 133 may configure a learning data set for a normal state and an abnormal state by appropriately filtering the sensed data.

In addition, the preprocessor 133 may configure the learning data set by analyzing the user's button manipulation and the door operation sequence. Here, the pre-processing unit 133 removes the door sequence by the user's button operation or the elevator safety edge multi-beam state is not a factor that affects the failure, and removes it from the normal sequence and actual foreign material jamming or failure of parts. A learning data set can be constructed by extracting only the abnormal sequence of the door caused by the

The learning and model generating unit 134 performs a function of generating a predictive model by driving a specific algorithm for predicting failure of an elevator door. In an embodiment, the learning and model generating unit 134 may perform learning on the learning data set formed by analyzing the learning data set formed from the sensing data and the operation sequence of the door through the preprocessing unit 133 . . Here, the learning and model generator 134 may perform supervised learning and unsupervised learning.

In general, supervised learning is a method of learning in a state in which a label for data, that is, an explicit correct answer is given, and learning is carried out in the form of (data, label). Unsupervised learning is a method of learning in a state where no explicit correct answer is given, that is, a label for data, and learning is carried out in the form of (data (date)). Unsupervised learning is used to discover hidden features or structures in data.

In one embodiment, the learning and model generation unit 134 performs supervised learning by setting the failure of the elevator door as a label in the training data set configured through preprocessing in the preprocessing unit 133, and using only the training data set Perform unsupervised learning. In this case, supervised learning uses a deep neural network (DNN) algorithm, and unsupervised learning uses an anomaly detection algorithm.

The learning and model generator 134 may perform data labeling for supervised learning. Elevator door failures can be broadly classified into hall door failures and car door failures. In order to construct a learning data set for supervised learning, the state of the data (normal/failure) must be labeled and data must exist in case of a certain degree of failure. And in order to automatically configure the learning data set, such failure data must also be automatically collected. In fact, the items to be set as failure prediction targets are failures for individual parts, so it is not easy to observe failures on individual parts. Therefore, it is possible to automatically collect a large number of failure data through mapping between an output variable that is difficult to observe, which is set as an actual failure prediction target, and a failure code that is easy to observe and can be automatically collected. In other words, data that can be observed automatically and easily without the observation of failures in individual parts by mapping the failure of the door inverter, communication, switch, command error, etc. By labeling with , supervised learning can be performed. At this time, since the model to be learned is not a failure diagnosis model but a failure prediction model, the corresponding label is added to the data before the occurrence of the failure for a certain period of time, not when the failure actually occurred. In other words, failures that can be automatically collected are mapped to failures of the same individual component and labeled in the data.

The learning and model generating unit 134 may configure a learning data set in which failure/steady state labels are combined, and may perform supervised learning with this learning data set. The learning and model generator 134 may perform unsupervised learning with a training data set that does not include a label. Here, supervised learning and unsupervised learning of the learning and model generating unit 134 will be described later.

The learning and model generation unit 134 may generate a door failure prediction model through supervised learning and unsupervised learning.

The failure prediction unit 135 performs a function of performing actual failure prediction based on the prediction model generated by the learning and model generation unit 134 . In one embodiment, the failure prediction unit 135 predicts the failure probability of the elevator door with real-time sensing data after the prediction model is generated by the learning and model generation unit 134 . The failure prediction unit 135 notifies the user when a failure probability greater than or equal to a predetermined probability is predicted so that the user can prepare for failure.

FIG. 2 is a diagram for explaining a label mapping process of a training data set in the training and model generation unit of FIG. 1 .

Referring to FIG. 2 , the learning and model generation unit 134 may perform supervised learning by setting whether the elevator door malfunctions as a label in the collected data. To this end, a failure of a specific part of the elevator door is randomly generated to cause a measurable system failure (steps S210 and S220). In fact, since it is difficult to secure failure data in the field, failures of individual parts that are actually failure prediction targets in elevator doors are arbitrarily generated.

After arbitrarily generating a failure for the corresponding item of the elevator door, the code is collected in the failure state and the failure code for the collected specific component failure is mapped (steps S230 and S240). By mapping the door inverter, communication, switch, command error, etc., which are easily observed automatically and easily in the failure state, with the corresponding failure, the data collected in the failure state of the door can be labeled.

FIG. 3 is a diagram for explaining a DNN algorithm used for supervised learning in the learning and model generation unit of FIG. 1 .

Referring to FIG. 3 , the learning and model generation unit 134 may generate a predictive model by performing supervised learning using a DNN algorithm on a training data set including a label.

DNN (Deep Neural Network) is an algorithm used in deep learning, and it is an artificial neural network (ANN) consisting of hidden layers between an input layer and an output layer. The DNN algorithm can model complex non-linear relationships like general artificial neural networks. DNN can be trained with standard error backpropagation algorithm. In this case, the weights may be updated using Equation 1 below.

[Equation 1]

Here, η is the learning rate, and C is the cost function. The choice of the cost function is determined by factors such as the type of learning and the activation function. For example, when supervised learning is performed on a multi-class classification problem, the activation function and the cost function are generally determined as the softmax function and the cross-entropy function, respectively. The Softmax function and cross entropy are defined by Equations 2 and 3 below.

[Equation 2]

Here, Pj represents the class probability, and Xj and Xk represent the total input to unit j and the total input to unit k, respectively.

[Equation 3]

Here, dj represents the target probability for the output unit j, and Pj is the probability output for j after the corresponding activation function is applied.

In one embodiment, the training and model generation unit 134 combines the failure state and the steady state label with the preprocessed data to form a training data set, and weights ( weight) and the offset. Here, the input layer consists of several input nodes (x1, ..., xn), and the preprocessed data and labels constituting the training data set are input to the input node, and the output nodes (y1, y1, …,ym), the probability of door failure is output.

FIG. 4 is a diagram for explaining an anomaly detection algorithm used for unsupervised learning in the learning and model generation unit of FIG. 1 .

Referring to FIG. 4 , the learning and model generation unit 134 generates a predictive model by performing unsupervised learning using an anomaly detection algorithm on a training data set composed only of data pre-processed and extracted through the pre-processing unit 133 . can do.

Anomaly detection algorithm learns a model that can detect anomalies by using data input over time through time series analysis and anomaly detection. The algorithm used to generate these models is the LSTM autoencoder anomaly detection algorithm.

In an embodiment, the learning and model generation unit 134 learns a model capable of detecting an abnormal state of an elevator door by performing an abnormality detection algorithm using an autoencoder.

First, a model is created assuming that the existing input is normal data. That is, an encoder/decoder model is created with learning data composed of a set of input features of the door. After encoding and decoding the real-time sensing data with the model created in this way, if it has a loss value less than the original data and a certain value, it is judged as normal.

Here, LSTM is applied for time series analysis. In other words, similar to autoencoder anomaly detection, an autoencoder is created with normal data. Thereafter, the restoration loss value is compared with respect to the real-time sensed data, and if it exceeds a certain level, it is determined as abnormal data.

FIG. 5 is a diagram for explaining an LSTM autoencoder anomaly detection algorithm used for unsupervised learning in the learning and model generation unit of FIG. 4 .

Referring to FIG. 5 , the learning and model generation unit 134 may apply LSTM for time series analysis while performing an autoencoder anomaly detection algorithm for unsupervised learning.

LSTM (Long Shot-Term Memory) is a type of RNN, and is an algorithm mainly used when applying time-series data.

Like the autoencoder anomaly detection algorithm shown in FIG. 4, the LSTM autoencoder anomaly detection algorithm assumes the existing input as normal data and encodes and decodes it to generate the autoencoder. Then, by comparing the restoration loss value with respect to the real-time sensed data, if the loss value is above a certain level, it is determined as abnormal data.

6 is a flowchart for explaining a failure prediction process in the failure prediction system of the elevator door in FIG. 1 .

In Figure 6, the elevator door failure prediction system 100 according to an embodiment collects data related to the failure of the elevator door (step S610).

More specifically, in the data collection unit 131 of the door failure prediction server 130, a plurality of sensors 110 and elevator control panel 120 attached to the site where the elevator door is installed through the gateway are related to elevator door failure. data can be collected. In an embodiment, the type of collected data may include sensing data such as door speed, opening/closing related operation data and user manipulation data. The data collected by the data collection unit 131 may be stored in the data storage unit 132 which is a storage (RAW) and may be provided for generation of a failure prediction model.

The elevator door failure prediction system 100 pre-processes the collected data and extracts features to configure a learning data set (step S620).

In one embodiment, the pre-processing unit 133 of the door failure prediction server 130 removes the data generated by the user's operation through the pre-processing of the collected data, and then sets the operation start to the operation completion of the door as one cycle and then It is possible to construct a training data set consisting of the speed, acceleration, and torque patterns of The preprocessor 133 may also configure the learning data set by analyzing the operation sequence of the door in consideration of the user's button operation, the elevator safety edge, and the state of the multi-beam from the collected data.

The failure prediction system 100 of the elevator door generates a failure prediction model by performing supervised learning and unsupervised learning with the learning data set (step S630).

In an embodiment, the learning and model generation unit 134 of the door failure prediction server 130 may generate a predictive model through a failure prediction algorithm based on the preprocessed data. The learning and model generator 134 may configure a supervised learning data set by including a label in the training data set configured through preprocessing. Here, the label may be configured through mapping between a failure that is automatically collected due to a door inverter error and an actual failure site caused by a foreign material caught or a poor connection. The learning and model generating unit 134 may generate a predictive model using a DNN algorithm using the supervised learning data set. The training and model generator 134 may generate a predictive model using an LSTM autoencoder anomaly detection algorithm with a training data set that does not include a label.

The failure prediction system 100 of the elevator door determines the failure possibility of the door through the failure prediction model (step S640).

In an embodiment, the failure prediction unit 135 of the door failure prediction server 130 may predict the failure of the elevator door in real time with the failure prediction model generated by the learning and model generation unit 134 . Here, the failure prediction unit 135 can predict the failure probability of individual parts of the door with real-time data measured after the prediction model is generated, and when the failure probability greater than or equal to the predetermined probability is predicted, the failure prediction unit 135 informs the user. You can prepare for failure. Accordingly, it is possible to pre-recognize and respond to abnormal operation of the elevator door or failure of parts.

Although the above has been described with reference to preferred embodiments of the present invention, those skilled in the art can variously modify and change the present invention within the scope without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that it can be done.

100: Elevator door failure prediction system
110: sensor 120: control panel
130: door failure prediction server
131: data collection unit 132: data storage unit
133: preprocessor 134: training and model generation unit
135: failure prediction unit

Claims (7)

A plurality of sensors installed in the elevator door and a door failure prediction server connected via the elevator control panel and gateway,
The door failure prediction server
a data collection unit for collecting data related to the failure of the elevator door in real time from the plurality of sensors and the elevator control panel;
a pre-processing unit for pre-processing the collected data and extracting features to construct a learning data set;
a learning and model generation unit for generating a failure prediction model by performing supervised learning and unsupervised learning with the learning data set; and
It is configured to include a failure prediction unit for determining the failure possibility of the door through the failure prediction model,
The learning and model generation unit
A supervised learning data set is constructed by combining a failure state and a steady state label with the learning data set, and the supervised learning data set is supervised using a DNN algorithm to generate a predictive model,
Elevator door, characterized in that by labeling the learning data set through the mapping between the failures automatically collected due to door inverter errors and the actual failures caused by foreign matter caught or poor connection, to configure the supervised learning data set Failure Prediction System.
According to claim 1, wherein the data collection unit
Elevator door failure prediction system, characterized in that the sensing data including the speed of the elevator door through the plurality of sensors and the operation data and user operation data related to the opening and closing of the elevator door are collected through the elevator control panel.
According to claim 1, wherein the pre-processing unit
After removing the data generated by the user's motion through the pre-processing of the collected data, the learning data set including the speed, acceleration, and torque pattern of the door is configured by making one cycle from the start of the door operation to the completion of the operation. Elevator door failure prediction system.
According to claim 1, wherein the pre-processing unit
Elevator, characterized in that after removing the data generated by the user's operation through the pre-processing of the collected data, the operation sequence of the door is analyzed in consideration of the user's button operation, the safety edge, and the state of the multi-beam to configure the learning data set. Door failure prediction system.
delete delete The method of claim 1, wherein the learning and model generation unit
Elevator door failure prediction system, characterized in that the prediction model is generated by unsupervised learning using the LSTM autoencoder anomaly detection algorithm on the training data set.

Images