KR102651107B1 - Elevator abnormality detection system and method - Google Patents

Abstract

A system and method for detecting an elevator abnormality are disclosed. The elevator abnormality detection system according to the present invention is intended to provide a predictive maintenance service by precisely detecting abnormal symptoms and precursor symptoms of a failure in the elevator, and includes a sensor unit that measures physical parameters of the elevator related to the failure; A data collection unit that collects sensing data measured by the sensor unit; an abnormal pattern detection unit that detects abnormal patterns that deviate from normal patterns by determining whether the sensing data exceeds a preset threshold; an interval calculation unit that calculates an occurrence time difference between abnormal patterns occurring adjacent to each other as an interval time value; and an anomaly score calculation unit that determines the occurrence density of abnormal patterns based on the interval time value between abnormal patterns detected within a certain section and calculates an abnormality score based on the occurrence density.

Description

Elevator abnormality detection system and method {Elevator abnormality detection system and method}

The present invention relates to a system and method for detecting abnormalities in an elevator. More specifically, the present invention relates to a system and method for detecting abnormalities in an elevator. More specifically, the present invention relates to a system and method for detecting abnormalities in an elevator by using the density of occurrence of abnormal patterns that deviate from the normal pattern during elevator operation to precisely detect abnormal symptoms and precursor symptoms of a failure in the elevator. It relates to an elevator abnormality detection system and method that enables the provision of predictive maintenance services.

In general, various types of high-rise buildings built for residential, office, or commercial purposes are equipped with elevators to ensure smooth movement between floors of passengers entering and exiting the building.

An elevator is an elevator car that moves up and down along a vertically formed hoistway inside a building with passengers inside, a mechanical part consisting of a motor part and a traction machine that generates power to raise and lower the elevator car, and the operation of the elevator. It is configured to include a control unit that performs related control.

Meanwhile, elevator maintenance is usually done through regular inspections scheduled at a certain interval, or when an error code occurs, a maintenance engineer is dispatched to check the manual according to the error code and take action. .

However, in the event of an unexpected elevator breakdown, such as an elevator breakdown occurring before the regular inspection schedule, a maintenance engineer will be dispatched to the scene and passengers will not be able to use the elevator until maintenance is completed. Inconveniences arise.

Therefore, recently, the importance of predictive maintenance, which predicts the possibility of elevator failure in advance and performs maintenance before actual failure occurs, has emerged. A related conventional technology is to install a sensor in the elevator to set a threshold value. There is a known technology that predicts elevator failure and performs maintenance by detecting (counting) the number of events exceeding .

Known prior art infers that there is an abnormality in the operation of the elevator according to the number of occurrences of abnormal patterns that deviate from the normal pattern among the signals detected by sensors installed in the elevator. According to this prior art, misjudgment of elevator failure prediction is made. There is a risk that this will occur.

Specifically, even in an elevator that is operating normally, abnormal signals may occur due to the movement of passengers or the surrounding environment. In the prior art, even abnormal signals that can occur in an elevator in a normal state are collected to diagnose the condition of the elevator. As a result, it often occurred that the possibility of elevator failure was incorrectly predicted and maintenance technicians were dispatched incorrectly.

In order to reduce misjudgment of failure predictions and minimize erroneous dispatch of maintenance engineers when performing predictive maintenance of elevators, more improved technology is required to accurately diagnose the condition of the elevator and precisely predict the possibility of failure. .

The purpose of the present invention is to use machine learning or deep learning technology to detect numerous abnormal symptoms and precursor symptoms of failure in elevators, and to provide predictive maintenance services for elevators based on this.

More specifically, the present invention accurately diagnoses the condition of the elevator and predicts the possibility of failure more precisely, thereby reducing misjudgment of failure prediction and minimizing erroneous dispatch of maintenance technicians when performing predictive maintenance of the elevator. The purpose is to provide a system and method for detecting abnormalities in an elevator.

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

According to one aspect of the present invention for achieving the above object, a sensor unit that measures physical parameters of an elevator related to a failure; a data collection unit that collects sensing data measured by the sensor unit; an abnormal pattern detection unit that detects abnormal patterns that deviate from normal patterns by determining whether the sensing data exceeds a preset threshold; an interval calculation unit that calculates an occurrence time difference between abnormal patterns occurring adjacent to the abnormal patterns as an interval time value; and an abnormality score calculation unit that determines the occurrence density of the abnormal patterns based on the interval time value between the abnormal patterns detected within a certain section and calculates an abnormality score based on the occurrence density. An elevator abnormality detection system comprising: can be provided.

The abnormal pattern detection unit classifies the abnormal patterns into one or more levels according to the degree to which they exceed the threshold and provides an abnormality detection classification value according to the level, and the abnormality score calculation unit determines that the higher the abnormality detection classification value, the higher the abnormality score. It can be corrected with weights during calculation.

The interval calculation unit may calculate an interval time value between the abnormal patterns for each category classified by the abnormality detection classification value.

The elevator abnormality detection system according to one aspect of the present invention displays the abnormality detection classification value on the Y-axis and the interval time value on the It may further include a clustering unit for grouping the interval time values calculated for each anomaly detection classification value according to the degree of closeness, wherein the anomaly score calculation unit determines a cluster of interval time values for the grouped anomaly patterns to be the cluster center in the normal state. The distance from can be used as a weight coefficient for calculating an abnormality score.

The abnormal pattern detection unit further generates information regarding the number of occurrences of the abnormal patterns detected within the predetermined section, and the abnormality score calculation unit may correct the abnormality score with a weight when calculating the abnormality score as the number of occurrences increases.

An elevator abnormality detection system according to an aspect of the present invention relates to the density of occurrence of abnormal patterns and the number of occurrences estimated by performing density-based clustering on the abnormality detection classification value, the interval time value, and the interval time value. It may further include a learning unit that automatically calculates an abnormality score from the operation data of the elevator collected in real time and establishes a model to predict the probability of failure of the elevator by performing machine learning or deep learning using the information as feed data. You can.

Meanwhile, according to another aspect of the present invention for achieving the above object, abnormal patterns exceeding a preset threshold are detected in elevator operation data collected in relation to a failure, and abnormal patterns detected within a certain section are detected. An elevator abnormality detection system can be provided that estimates the probability of failure of the elevator by calculating an abnormality score based on the occurrence time interval between abnormal patterns that occur adjacent to each other.

Meanwhile, according to another aspect of the present invention for achieving the above object, an operation data collection step of collecting operation data of an elevator related to a breakdown; An abnormal pattern detection step of detecting an abnormal pattern that deviates from a normal pattern by determining whether the operation data of the elevator collected in the operation data collection step exceeds a preset threshold; An interval calculation step of calculating an occurrence time difference between abnormal patterns occurring adjacent to the abnormal patterns detected in the abnormal pattern detection step as an interval time value; A clustering step of performing density-based clustering (DBSCAN) to group the interval time values calculated in the interval calculation step according to the degree of closeness; And an abnormality score calculation step of performing density-based clustering on the interval time value and calculating an abnormality score of the elevator based on the density of occurrence of the estimated abnormality pattern.

In the abnormal pattern detection step, the abnormal patterns are classified into one or more levels according to the degree to which they exceed the threshold and an abnormality detection classification value is given according to the level level. In the abnormality score calculation step, the higher the abnormality detection classification value, the higher the abnormality detection classification value. When calculating abnormal scores, they can be corrected with weights.

In the interval calculation step, the interval time value between the abnormal patterns can be calculated for each category classified by the abnormality detection classification value.

In the abnormal pattern detection step, information regarding the number of occurrences of the abnormal patterns detected within the certain section is further generated, and in the abnormality score calculation step, as the number of occurrences increases, the abnormality score can be corrected with a weight when calculating the abnormality score.

An elevator anomaly detection method according to another aspect of the present invention is to determine the occurrence density and occurrence count of the estimated anomaly pattern by performing density-based clustering on the anomaly detection classification value, the interval time value, and the interval time value. By performing machine learning or deep learning using the relevant information as feed data, an abnormality score is automatically calculated from the elevator operation data collected in real time and a learning step is further performed to establish a model to predict the probability of failure of the elevator. It can be included.

In the anomaly score calculation step, density-based clustering is performed on the anomaly detection classification value and the interval time value, and information on the occurrence density and the number of occurrences of the anomaly pattern estimated is input, and according to the inference result of the model, the An ideal score can be calculated.

According to the present invention, by using machine learning or deep learning technology, numerous abnormal symptoms and precursor symptoms of elevator failure are detected, and based on this, preventive measures are taken in advance before an elevator malfunction actually occurs. Predictive maintenance services can be provided.

In addition, the present invention diagnoses the current state of the elevator based on the occurrence time interval between abnormal patterns that occur during operation of the elevator, that is, the density of occurrence of abnormal patterns, compared to the conventional technology that relies only on the number of occurrences of abnormal patterns. The possibility of failure can be predicted more precisely, thereby reducing misjudgment of failure predictions and minimizing the erroneous dispatch of maintenance technicians, as well as preventing failures of unknown cause that cannot be explained by existing error codes. This has the effect of expanding the prediction range for failures related to failures between elevator parts.

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

1 is a functional block diagram of an elevator abnormality detection system according to the present invention.
Figure 2 is a diagram for explaining a process included in the process of calculating an abnormality score of an elevator according to the present invention.
Figure 3 is a flowchart showing an elevator abnormality detection method according to the present invention.

Details regarding the purpose and technical configuration of the present invention and its operations and effects will be more clearly understood by the detailed description based on the drawings attached to the specification of the present invention.

The terminology used herein is merely used to describe specific embodiments and is not intended to limit the invention. For example, terms such as “consists of” or “comprises” used in the present specification should not necessarily be construed as including all of the various components or steps described in the invention, and only some of the components or steps may be included. It should be construed as not including them or as being able to further include additional components or steps. Additionally, as used herein, singular expressions include plural expressions unless the context clearly dictates otherwise.

Hereinafter, the present invention will be described in detail by explaining preferred embodiments of the present invention with reference to the accompanying drawings. The embodiments described below are provided so that those skilled in the art can easily understand the technical idea of the present invention, and should not be construed as limiting the present invention. The embodiments of the present invention are provided to those skilled in the art. It is natural that it can have various applications.

FIG. 1 is a functional block diagram of an elevator abnormality detection system according to the present invention, FIG. 2 is a diagram illustrating a process included in the abnormality score calculation process of an elevator according to the present invention, and FIG. 3 is a functional block diagram of an elevator abnormality detection system according to the present invention. This is a flowchart showing the abnormality detection method.

The elevator abnormality detection system and method according to the present invention detects abnormal symptoms and precursor symptoms of elevator failure using data acquired through sensors installed inside and outside the elevator, and various elevator abnormality symptoms collected in real time. By applying machine learning or deep learning technology to operation information, it is implemented to predict and respond to the possibility of elevator failure that may occur in the future.

To this end, as shown in FIG. 1, the elevator abnormality detection system according to the present invention includes a sensor unit 10, a data collection unit 20, an abnormal pattern detection unit 30, an interval calculation unit 40, and a clustering unit ( 50), a learning unit 60, and an abnormality score calculation unit 70.

The sensor unit 10 is for acquiring various operation information related to elevator failure, and can measure various physical parameters of the elevator used to predict the possibility of failure.

The sensor unit 10 includes a vibration sensor installed on or around the body of the elevator car, a noise sensor, an acceleration sensor for measuring the acceleration of the elevator car, a position sensor for detecting the position of the elevator car, and a sensor for opening and closing the elevator door. It may be comprised of one or more sensors that monitor the operating status of elevator devices or components and measure physical parameters related to the operation of the elevator, including motion sensors that measure related parameters.

The sensor unit 10 can be installed in various locations where elevator operation information is desired, and can be installed not only in the elevator car, but also in hoistways, landings, hoisting machines in machine rooms, and governors, etc. in places where it is desired to predict the possibility of failure in relation to the operation of the elevator. Applicable to all components of an elevator.

The data collection unit 20 collects data related to a failure of the elevator acquired by the sensor unit 10. The data collection unit 20 may receive data from the sensor unit 10 through a wireless network.

The abnormal pattern detection unit 30 detects an abnormal pattern that deviates from a normal pattern by determining whether the data collected by the data collection unit 20 exceeds a preset threshold.

At this time, the abnormal pattern detection unit 30 may set a reference level that divides the detected abnormal patterns into at least one level according to the degree to which they deviate from the threshold, and generate an 'abnormal detection classification value' by classifying the detected abnormal patterns according to the level.

That is, the abnormal pattern detection unit 30 can not only distinguish normal patterns from abnormal patterns, but also classify abnormal patterns into one or more levels according to the degree to which they exceed the threshold, and assign an index value (anomaly detection classification value) indicating the level. This determines the severity level according to the extent to which the abnormal pattern exceeds the threshold, and implements more precise elevator failure prediction by assigning different weights when calculating the abnormality score that is performed later depending on the severity level of the abnormal pattern. This is to do it.

The abnormal pattern detection unit 30 may label the classified data with an abnormality detection classification value according to the reference level and generate a time series pattern accordingly.

In one embodiment, referring to the graph shown at the top of FIG. 2, sensing data related to elevator operation is drawn over time. The data is measured by the sensor unit 10 and collected by the data collection unit 20.

The abnormal pattern detection unit 30 may classify the data collected by the data collection unit 20 into a plurality of categories according to reference levels. In this embodiment, the reference level is set to 0 to 3, where reference level 1 represents a threshold value, and reference levels 2 and 3 serve as standards for classifying abnormal patterns into one or more levels.

In this embodiment, sensing data that comes between reference levels 0 and 1, that is, sensing data below the threshold, may be considered a normal pattern. On the other hand, sensing data exceeding the reference level 1 can be considered as abnormal patterns, and in this case, the abnormal patterns can be classified into one or more levels again according to the reference level set higher.

The abnormal pattern detection unit 30 assigns '0' as an abnormality detection classification value to normal patterns and '1' as an abnormality detection classification value to abnormal patterns detected between reference levels 1 and 2, For abnormal patterns detected between reference levels 2 and 3, '2' may be assigned as an abnormality detection classification value. Although not shown in the drawing, abnormal patterns exceeding the standard level 3 may be assigned a higher value of '3' as an abnormality detection classification value.

The larger the anomaly detection classification value, the greater the degree to which the anomaly pattern deviates from the threshold, and this can serve as a weighting factor when calculating the anomaly score by the anomaly score calculation unit 70. In other words, the more the abnormality pattern deviates from the threshold, the more weight can be assigned to calculating the abnormality score.

Meanwhile, the embodiment shown in FIG. 2 is merely an illustration to aid understanding of the present invention, and the present invention is not necessarily limited or limited thereto. Of course, the standards can be simplified by further subdividing the standard levels or integrating multiple levels as needed.

The interval calculation unit 40 receives information about the abnormal patterns detected by the abnormal pattern detection unit 30 and calculates the occurrence time difference between the abnormal pattern and other abnormal patterns occurring adjacent to it as an 'interval time value'.

In addition, the interval calculation unit 40 can also calculate the interval time value between normal patterns. This is to clearly determine whether the elevator is in a normal operating state, and the purpose of the present invention is to determine the density of occurrence of abnormal patterns. Considering that the prediction is based on the possibility of elevator failure, the function of calculating the interval time between normal patterns is not necessarily required.

In the present invention, the interval calculation unit 40 can calculate the interval time value for each abnormality detection classification value. That is, the abnormal patterns detected by the abnormal pattern detection unit 30 are classified into categories called abnormality detection classification values, and the interval calculation unit 40 can calculate the interval time value for the abnormal patterns for each category. .

Referring to the embodiment shown in FIG. 2, the interval calculation unit 40 determines normal patterns to which 0 is assigned as an abnormality detection classification value, abnormal patterns to which 1 is assigned as an abnormality detection classification value, and an abnormality detection classification value. For abnormal patterns given a value of 2, the interval time value can be calculated for each category.

The clustering unit 50 2D plots the ‘anomaly detection classification value’ generated by the abnormal pattern detection unit 30 on the Y-axis and the ‘interval time value’ calculated by the interval calculation unit 40 on the X-axis. It is expressed as a graph, and the interval time value calculated for each anomaly detection classification value by performing density-based clustering (DBSCAN: Density-Based Spatial Clustering of Applications with Noise) on the anomaly detection classification value (Y) - interval time value (X) graph. are grouped according to degree of closeness. Accordingly, the interval time value data according to the anomaly detection classification value can be classified into similar clusters, and the distance from the cluster center in the normal state can be used as a weight coefficient when calculating the anomaly score to be performed later.

At the bottom of Figure 2, a 2D plotted graph is shown with the 'anomaly detection classification value' as the Y-axis and the 'interval time value' as the Since calculations are performed for each category, a graph in the form of FIG. 2 is derived.

The learning unit 60 determines the distance from the cluster center in the normal state (in density-based clustering, this is ' By performing machine learning or deep learning using feed data such as the 'epsilon radius', a model is created that automatically calculates anomaly scores from data collected and generated in real time and predicts the probability of elevator failure. can be established.

The abnormal score calculation unit 70 uses data collected and generated in real time by the above-described data collection unit 20, abnormal pattern detection unit 30, interval calculation unit 40, and clustering unit 50 as input variables. The abnormality score of the elevator can be calculated according to the inference results of the learning model according to the time series pattern, and the probability of elevator failure can be predicted from the calculated abnormality score.

More specifically, the abnormality score calculation unit 70 determines that the higher the abnormality detection classification value generated by the abnormal pattern detection unit 30 and the smaller the interval time value calculated by the interval calculation unit 40, that is, the higher the abnormality pattern detection unit 30 is. The more frequently it occurs, the more weight can be assigned to calculating the abnormality score. At this time, the interval time value may be a plurality of data clustered through density-based clustering performed by the clustering unit 50, and the distance from the center of the cluster in the normal state to the cluster of interval time values for abnormal patterns is greater than or equal to It can be used as a weighting coefficient in score calculation.

Based on the graph shown at the bottom of Figure 2, the higher the abnormality detection classification value on the Y-axis (higher severity level), and the smaller the occurrence time interval between abnormal patterns on the X-axis (higher occurrence density), the more likely the failure is. As the probability of occurrence is high, the abnormality score will increase.

As described above, the elevator abnormality detection system according to the present invention reflects the severity level of the abnormal pattern in calculating the abnormality score by classifying the sensing data collected during operation of the elevator into categories according to the degree to which they exceed the threshold, and determines the interval for each classified category. It is implemented to precisely detect elevator abnormalities by performing time value and density-based clustering and reflecting the density of occurrence of the estimated abnormal pattern in calculating the abnormality score.

Additionally, in addition to the above data, the anomaly score calculation unit 70 may comprehensively consider the 'number of occurrences' of the anomaly pattern as an additional input variable in the anomaly score calculation as before. Information regarding the number of occurrences of an abnormal pattern may be generated by the abnormal pattern detection unit 30, and as the number of occurrences of an abnormal pattern within a certain section increases, a weight in calculating an abnormality score may be assigned.

As described above, the data considered when calculating the anomaly score and the results (anomaly score) inferred by the anomaly score calculation unit 70 can be accumulated and stored, and the accumulated information is the feed data of the above-described learning unit 60. It can be used and learned repeatedly.

Unlike the prior art, which diagnosed or predicted elevator failure based solely on the number of occurrences of abnormal patterns, the present invention predicts elevator failure more precisely by complexly considering not only the number of occurrences of abnormal patterns, but also the severity level and occurrence density of abnormal patterns. can be expected.

In other words, conventionally, elevator failure was diagnosed or predicted using only one piece of data, but the present invention can have the effect of significantly improving the precision of elevator failure prediction compared to the existing method by calculating an abnormality score by comprehensively considering various data. As learning is repeated cumulatively, the precision of the learning model improves, making it possible to diagnose the condition of the elevator more accurately and precisely.

The elevator abnormality detection system according to the present invention sends an abnormality score calculated by the abnormality score calculation unit 70 and the risk of failure (probability of failure) according to the abnormality score to the maintenance person in charge of the cloud platform's dashboard or mobile to provide an alarm. can be provided, and maintenance personnel can plan and provide predictive maintenance services based on the information.

Hereinafter, the elevator abnormality detection method according to the present invention will be described with continued reference to FIGS. 1 to 3.

The elevator abnormality detection method according to the present invention is performed using the elevator abnormality detection system according to the present invention described above, and includes the steps of collecting elevator operation data related to a failure (S10); Detecting an abnormal pattern that deviates from the normal pattern from the collected elevator operation data (S20); Calculating an interval, which is the difference in occurrence time between adjacent abnormal patterns (S30); A step of performing density-based clustering and grouping the calculated intervals according to the degree of closeness (S40); And it may include calculating an abnormality score based on the occurrence density of abnormal patterns detected within a certain section (S50).

In step S10, the elevator operation data related to the failure may be acquired by the sensor unit 10 installed in the area where the failure of the elevator is to be predicted and collected by the data collection unit 20.

Step S20 may be performed by the abnormal pattern detection unit 30. At this time, the abnormal pattern detection unit 30 goes beyond distinguishing normal patterns from abnormal patterns and further classifies abnormal patterns into one or more levels according to the degree of deviation from the threshold, and generates an 'abnormal detection classification value' that represents the level as an index value. You can. Here, the anomaly detection classification value can be viewed as indicating the severity level of the abnormality pattern, and the higher the anomaly detection classification value, the more weight can be assigned when calculating the anomaly score in the future.

Additionally, in step S20, the abnormal pattern detection unit 30 may label data classified according to the reference level with an abnormality detection classification value and generate a time series pattern accordingly.

Step S30 may be performed by the interval calculation unit 40. At this time, the interval calculation unit 40 may calculate the 'interval (time value)' of the abnormal patterns for each abnormality detection classification value classified in step S20.

Step S40 may be performed by the clustering unit 50. The clustering unit 50 plots the abnormality detection classification value classified in step S20 on the Y-axis and the interval calculated in step S30 on the By performing density-based clustering (DBSCAN) on the interval ( The distance can be used as a weight coefficient when calculating an abnormality score.

Step S50 may be performed by the abnormality score calculation unit 70. The abnormality score calculation unit 70 may assign weight to the calculation of the abnormality score as the interval calculated in step S30 becomes smaller. At this time, the interval data can be clustered into a large number of data through density-based clustering, and the distance of the interval cluster for abnormal patterns from the center of the cluster in the normal state can be used as a weight coefficient for calculating the abnormality score.

Additionally, if the abnormality patterns are classified into multiple categories in step S20, the abnormality score calculation unit 70 may assign weight to the calculation of the abnormality score as the abnormality detection classification value increases.

Meanwhile, the elevator operation information collected in step S10, the abnormality detection classification value generated in step S20, the interval (time value) calculated in step S30, and the abnormality estimated by performing density-based clustering on the interval data in step S40. Information about the density of pattern occurrence and the abnormality score calculated in step S50 are accumulated and stored, and a model is created to automatically calculate the abnormality score from data collected and generated in real time and predict the probability of elevator failure. It can be used as learning data to establish.

In step S50, the abnormality score calculation unit 70 calculates the abnormality score of the elevator according to the inference result of the learning model according to the time series pattern using data collected and generated in real time as input variables, and the abnormality score of the elevator is calculated from the calculated abnormality score. Probabilities can be predicted.

In the above embodiment, the elevator abnormality detection method according to the present invention has been described by dividing it into a plurality of steps, but any one step may be combined with other steps and performed together, omitted, divided into detailed steps, or performed as described above. Of course, one or more steps that were not performed may be added and performed.

As described above, the present invention detects a failure or a precursor to an elevator failure based on the interval of how closely the abnormal pattern occurrence section exceeding the threshold is located among the parameters measured during operation of the elevator. The basic technical idea is to judge it as a significant signal for failure prediction only when a similar excess value signal appears within a predetermined time.

According to the present invention, machine learning or deep learning technology is used to detect numerous abnormal symptoms and precursor symptoms of elevator failure, and based on this, preventive measures are taken before an elevator malfunction actually occurs. We can provide proactive maintenance services.

In addition, the present invention diagnoses the current state of the elevator based on the occurrence time interval between abnormal patterns that occur during operation of the elevator, that is, the density of occurrence of abnormal patterns, compared to the conventional technology that relies only on the number of occurrences of abnormal patterns. The possibility of failure can be predicted more precisely, thereby reducing misjudgment of failure predictions and minimizing the erroneous dispatch of maintenance technicians, as well as preventing failures of unknown cause that cannot be explained by existing error codes. This has the effect of expanding the prediction range for failures related to failures between elevator parts.

The elevator abnormality detection method according to the present invention can be performed by a computer device having one or more processors and a memory that stores one or more programs executed by the one or more processors. To this end, the elevator abnormality detection system according to the present invention may be implemented in the form of a program or software including one or more computer-executable instructions and stored in memory.

The above-described data collection unit 20, abnormal pattern detection unit 30, interval calculation unit 40, clustering unit 50, learning unit 60, and abnormal score calculation unit 70 include an assembler for executing the program, It may be implemented in the form of a computer-readable code or program, such as a compiler or interpretation program, and the system may be configured to communicate with each other through a wired or wireless network.

The present invention is not limited to the described embodiments, and it is obvious to those skilled in the art that various modifications and changes can be made without departing from the spirit and scope of the present invention. Accordingly, such modifications or variations should be considered to fall within the scope of the claims of the present invention.

10: Sensor unit
20: Data collection unit
30: Abnormal pattern detection unit
40: Interval calculation unit
50: Clustering unit
60: Learning Department
70: Abnormal score calculation unit

Claims (13)

A sensor unit that measures physical parameters of the elevator related to failure;
a data collection unit that collects sensing data measured by the sensor unit;
Detects abnormal patterns that deviate from normal patterns by determining whether the sensing data exceeds a preset threshold, classifies the abnormal patterns into one or more levels according to the degree to which they deviate from the threshold, and assigns an abnormality detection classification value according to the level level. an abnormal pattern detection unit;
an interval calculation unit that calculates an occurrence time difference between abnormal patterns occurring adjacent to the abnormal patterns as an interval time value;
an abnormality score calculation unit that determines an occurrence density of the abnormal patterns based on an interval time value between the abnormal patterns detected within a certain section and calculates an abnormality score based on the occurrence density; and
A clustering unit that groups the interval time values calculated for each abnormality detection classification value according to the degree of closeness,
The interval calculation unit calculates an interval time value between the abnormality patterns for each category classified by the abnormality detection classification value,
The anomaly score calculation unit uses the distance of the cluster of interval time values for the grouped anomaly patterns from the center of the cluster in a normal state as a weight coefficient for calculating the anomaly score,
Elevator abnormality detection system.
In claim 1,
The anomaly score calculation unit is characterized in that the higher the anomaly detection classification value, the higher the anomaly score calculation, and correcting it with a weight when calculating the anomaly score.
Elevator abnormality detection system.
delete In claim 1,
The clustering unit plots the anomaly detection classification value on the Y-axis and the interval time value on the Characterized by grouping interval time values according to closeness,
Elevator abnormality detection system.
In claim 4,
The abnormal pattern detection unit further generates information regarding the number of occurrences of the abnormal patterns detected within the certain section,
The anomaly score calculation unit is characterized in that the more the number of occurrences, the more the anomaly score is calculated by correcting it with a weight.
Elevator abnormality detection system.
In claim 5,
Machine learning or deep learning is performed by performing density-based clustering on the anomaly detection classification value, the interval time value, and the interval time value, using the information on the occurrence density and the number of occurrences of the abnormal pattern estimated as feed data. By performing this, it further includes a learning unit that automatically calculates an abnormality score from the operation data of the elevator collected in real time and establishes a model for predicting the probability of failure of the elevator.
Elevator abnormality detection system.
An abnormal pattern that exceeds a preset threshold is detected from the elevator operation data collected in relation to a malfunction, and the abnormal pattern is detected based on the occurrence time interval between adjacent abnormal patterns for the abnormal patterns detected within a certain section. Determine the density of occurrence, calculate an abnormality score based on the density of occurrence,
Classifying the abnormal patterns into one or more levels according to the degree to which they exceed the threshold, and assigning an abnormality detection classification value according to the level level,
The interval time values calculated for each abnormality detection classification value are grouped according to the degree of closeness,
Calculate the interval time value between the abnormal patterns for each category classified by the abnormality detection classification value,
Estimating the probability of failure of the elevator by using the distance of the cluster of interval time values for the grouped abnormal patterns from the center of the cluster in a normal state as a weight coefficient for calculating the abnormality score,
Elevator abnormality detection system.
An operation data collection step of collecting elevator operation data related to a breakdown;
An abnormal pattern detection step of detecting an abnormal pattern that deviates from a normal pattern by determining whether the operation data of the elevator collected in the operation data collection step exceeds a preset threshold;
An interval calculation step of calculating an occurrence time difference between abnormal patterns occurring adjacent to the abnormal patterns detected in the abnormal pattern detection step as an interval time value;
A clustering step of performing density-based clustering (DBSCAN) to group the interval time values calculated in the interval calculation step according to the degree of closeness; and
An abnormality score calculation step of performing density-based clustering on the interval time value and calculating an abnormality score of the elevator based on the density of occurrence of the estimated abnormal pattern,
In the abnormal pattern detection step, the abnormal patterns are classified into one or more levels according to the degree to which they exceed the threshold, and an abnormality detection classification value is assigned according to the level level,
In the interval calculation step, the interval time value between the abnormality patterns is calculated for each category classified by the abnormality detection classification value,
In the anomaly score calculation step, the distance of the cluster of interval time values for the grouped anomaly patterns from the center of the cluster in the normal state is used as a weight coefficient for calculating the anomaly score,
Elevator abnormality detection method.
In claim 8,
In the anomaly score calculation step, the higher the anomaly detection classification value, the higher the anomaly score calculation, characterized in that correction is made with a weight when calculating the anomaly score.
Elevator abnormality detection method.
delete In claim 9,
In the abnormal pattern detection step, information regarding the number of occurrences of the abnormal patterns detected within a certain period is further generated,
In the anomaly score calculation step, the greater the number of occurrences, the more the anomaly score is calculated with a weight correction,
Elevator abnormality detection method.
In claim 11,
Machine learning or deep learning is performed by performing density-based clustering on the anomaly detection classification value, the interval time value, and the interval time value, using the information on the occurrence density and the number of occurrences of the abnormal pattern estimated as feed data. By performing this, it further includes a learning step of automatically calculating an abnormality score from the operation data of the elevator collected in real time and establishing a model for predicting the probability of failure of the elevator.
Elevator abnormality detection method.
In claim 12,
In the anomaly score calculation step, density-based clustering is performed on the anomaly detection classification value and the interval time value, and information on the occurrence density and the number of occurrences of the anomaly pattern estimated is input, and according to the inference result of the model, the Characterized by calculating an ideal score,
Elevator abnormality detection method.

Images