KR102347874B1 - using Non-contact Information Detection Sensor - Google Patents

Abstract

The present invention relates to a balise predictive diagnostic system and, more specifically, to a balise predictive diagnostic system which enables cable real-time state monitoring and C interface monitoring between an LEU and a balise and has a function of applying a big data analysis system to an ATP ground device remote monitoring technology.

Description

Balinese predictive diagnosis system and method using the same {using Non-contact Information Detection Sensor}

The present invention relates to a Balinese predictive diagnosis system, and more specifically, through data collection and frequency analysis using a non-contact sensor, it is possible to monitor the real-time status of the Bali cable and monitor the 'C' interface between LEU and Bali, and analyze big data. It relates to a ballistic predictive diagnostic system with a function of applying the system to the ATP ground device remote monitoring technology.

ATP (Automatic Train Protection) is a system that maintains safe train operation by detecting trains, maintaining distance between preceding and following trains, interlocking routes and limiting speed, etc., and corresponds to European standard ERTMS/ETCS Level 1.

The ground device is the Balise Group and the Track Change Control Unit (LEU), which exchange operation information between trains through the Balis installed at the boundary of the blocked section to secure the minimum braking distance, shorten the driving time, and Executes the train protection function according to capacity increase and train collision.

The track change control unit is a ground device of the ETCS Level 1 system, and it is a facility that controls the transmission of appropriate telegrams to the ballis by receiving the current status information of the hall, departure signal, and blocking signal or operation information of the crossing control device. It is classified into a centralized type installed in the machine room in the station premises and a distributed type that is accommodated in a trackside fixture box.

LEU consists of motherboard (MB), power board (PB), signal input board (LDB or DIB), balisse driving board (BDB), and fault detection board (LMB). Power is supplied to each board through the power board, and the Balis drive board collects the information of the signal input board and transmits the telegram information corresponding to the signal display condition to the Balis, and the failure detection board monitors whether the LEU is operating normally. do.

According to Korean Patent Registration No. 10-2105321, when there is a train on the track or when there is no train, the loop antenna installed under the ballis not only monitors the ballis, but also the electrical equipment technical support system can manage the failure of each device in an integrated way. It provides an integrated monitoring system for balis, and receives variable information collected from the ATP function monitoring unit through the loop transmitting antenna and loop receiving antenna, so that it is possible to determine and manage the presence of failure of the balis more efficiently than the existing system.

However, for the monitoring of ATP volleys, a technology has been developed that can remotely check whether the balis is operating normally by installing a loop antenna under the balis, but only limited information about the normal operation of the balis It is only possible to check whether it is working or not.

Therefore, although the technology to detect volley errors through the ground device and on-board detection device has been developed, it is necessary to develop a system for remote monitoring of the valise and LEU constituting the ATP ground device in real time on the ground.

In addition, due to the difficulty in distinguishing between valise failure, LEU failure, and connection cable failure, efficient maintenance in terms of time and cost according to periodic inspection activities is required.

In addition, there was a need to apply real-time remote monitoring technology capable of predicting failures, breaking away from the existing on-site inspection method in preparation for increased ATP installation routes and equipment aging.

The present invention has been made to solve the above problems, and the present invention provides real-time monitoring of LEU operation status, real-time monitoring of interface status between LEU and Bali, interworking with electrical equipment technical support system (indoor monitoring facility), and big data It aims to provide a Balinese predictive diagnosis system that can perform preventive maintenance through analysis.

In addition, the present invention in detail monitors the signal information at the terminal block and inspection point and transmits it to the control unit to monitor the signal information between the signal machine and the LEU, collect the operation state of the ballis through the LMB board and the fault condition through the DIB board and send it to the control unit It transmits, monitors the operating status of various boards in real time, and transmits the ballistic information to the control unit without the effect of impedance using the non-contact sensor, and monitors the real-time status of the ballistic cable through data collection and frequency analysis using the non-contact sensor. The purpose of this is to provide a Balinese predictive diagnosis system that can monitor the 'C' interface between the LEU and Balis.

In addition, by applying a trend analysis system for failures and failures of field facilities (electronic linkage, line switcher, TLDS, etc.) purpose is to

Another object of the present invention is to provide a ballistic predictive diagnostic system having a scientific diagnostic technique for checking the overall progress of all elements of the maintenance personnel's bales, LEU, and cables.

In order to solve the above problems, the present invention is a track change control unit (LEU, LEU, Lineside Electronic Unit); Ballis, which provides location information, trajectory and signal information with an on-board device; a cable connected from the LEU to the ballis; and a non-contact information detection sensor that is installed on the cable and does not affect impedance. A control unit that receives a signal sampled through the non-contact sensor; including, monitors the ballistic cable state using the non-contact sensor and transmits it to the control unit, and analyzes the amplitude of the signal detected by the non-contact sensor to 'C1 It is determined whether the 'interface signal is appropriate or not, and the amplitude of the signal detected by the non-contact sensor is analyzed to determine whether the 'C6' interface signal is appropriate.

The operation state of the ballis is transmitted to the control unit through the LMB board, and it is determined whether the DIB board is faulty and transmitted to the control unit.

The above 'C1' interface characteristics are C1 signal bit PRESENCE detection (alarm display when C1 signal is not detected), C1 signal amplitude C1_BITS_AMPLITUDE measurement, (alarm display when C1 signal amplitude exceeds the set range), C1 signal bit rate C1_BIT_RATE measurement (setting) It is collected through analysis of C1_BIT_CONTINUITY (341,1023 bits) signal of continuity and periodicity of C1 telegram (display of alarm when condition is not met).

The above 'C6' interface characteristics determine the presence of LEU 8.82KHz carrier signal (alarm display when there is no signal), amplitude of LEU 8.82KHz carrier signal (alarm display when amplitude exceeds settable range), and frequency of LEU 8.82KHz carrier signal. It is collected through measurement (indicates an alarm when the frequency exceeds the settable range).

The present invention relates to a method using a baliss predictive diagnosis system, comprising the steps of: using a non-contact sensor to monitor the state of a valley cable without the influence of impedance (disconnection, short circuit, crosstalk) and transmitting it to a control unit; determining whether the 'C1' interface signal is appropriate by analyzing the amplitude of the signal detected by the non-contact sensor; and determining whether the 'C6' interface signal is appropriate by analyzing the amplitude of the signal detected by the non-contact sensor.

Pre-processing by finding and pre-processing the telegram (movement authority (packet 12) and linking (packet 5), etc.) information transmitted to the balis, 'C1' and 'C6' characteristic signals of data include

As the 'C1' and 'C6' characteristic signal classification technology, the input signal is primarily divided into a normal state and an abnormal state, and the abnormal state is secondarily classified by abnormal cause, and the method of classifying the abnormal state includes descriptive statistics Classification of supervised learning and clustering of unsupervised learning are used based on decision trees and machine learning based on .

Calculate the real-time exponential moving average value (EMA) of the telegram (movement authority (packet 12), linking (packet 5), etc.) information transmitted to the balis, 'C1' and 'C6' characteristic signals, and based on the exponential moving average value It determines the allowable sensitivity indicating the measurement error range by the sensitivity and pre-processes after determining the allowable lower limit value, which is the minimum range for judging a sudden signal change.

In the present invention as described above, since the frequency at which the signal of the ballistic cable is actually measured and the FSK frequency collected using the non-contact sensor are the same, the state of the ballistic cable without the influence of impedance using the non-contact sensor (disconnection, short circuit, crosstalk) can be monitored.

In addition, the present invention can help the prompt maintenance implementation by expressing a facility history report such as a similar case display and past measures when a failure occurs.

In addition, it is possible to provide basic data necessary for parts operation and device introduction by organizing the operation history and abnormal occurrence details for each facility.

In addition, the present invention uses a non-contact sensor to monitor the state of the ballless cable (disconnection, short circuit, crosstalk) without the influence of impedance, and it is possible to develop an ATP predictive diagnostic device that converts the After Service to the Before Service.

In addition, the present invention can apply the big data analysis system to the ATP ground device remote monitoring technology, or the real-time remote monitoring technology capable of predicting failures outside the existing on-site inspection method in preparation for increased ATP installation routes and equipment aging.

In addition, the present invention enables economical preventive maintenance using the LEU operating state (failure), the LEU-valis interface state, and abnormal diagnosis/predictive diagnosis for the valise board.

In addition, the present invention learns the experience of field maintenance experts in the data analysis device, and supports a quick response by providing the primary judgment on the input signal and the data based thereon.

1 is a view showing the overall configuration according to an embodiment of the present invention.
2 is a view showing a detailed configuration and a visualization module according to an embodiment of the present invention.
3 is a view showing a sensor installation configuration according to an embodiment of the present invention.
4 is a diagram for comparing detected waveforms according to another embodiment of the present invention.
5 is a view showing the configuration of the ballless drive unit communication according to another embodiment of the present invention.
6 is a view showing a visualization screen according to another embodiment of the present invention.
7 is a diagram illustrating a remote monitoring facility communication procedure according to another embodiment of the present invention.
8 is a view showing a sigmoid function according to another embodiment of the present invention.
9 is a diagram illustrating management using an EMA or the like by comparing an observed value and a predicted value using the ballistic prediction diagnosis system according to another embodiment of the present invention.

In order to fully understand the present invention, preferred embodiments of the present invention will be described with reference to the accompanying drawings. Embodiments of the present invention may be modified in various forms, and the scope of the present invention should not be construed as being limited to the embodiments described in detail below. This example is provided to more completely explain the present invention to those of ordinary skill in the art. Accordingly, the shape of elements in the drawings may be exaggerated to emphasize a clearer description. It should be noted that in each drawing, the same member is shown with the same reference numerals in some cases. In addition, detailed descriptions of well-known functions and configurations determined to unnecessarily obscure the gist of the present invention will be omitted.

As shown in FIGS. 1 to 3, the present invention receives information on the current state of the hall, departure signal, and occlusion signal or operation information of the crossing control device as a ground device, and transmits an appropriate telegram to the ballis. control unit (LEU, Lineside Electronic Unit; 20); Ballis (30) for providing location information, trajectory and signal information to an on-board device; a cable connected from the LEU to the ballis; and a non-contact information detection sensor (41) that is installed on the cable and does not affect impedance. a control unit (10; ATP function monitoring device) receiving a signal sampled through the non-contact sensor; visualization module 50 and the like.

Therefore, by using the non-contact sensor, the state of the ballistic cable (disconnection, short circuit, crosstalk) is monitored without the influence of impedance and transmitted to the control unit, and the amplitude of the signal detected by the non-contact sensor is analyzed to determine the appropriateness of the 'C1' interface signal. It is determined whether or not the 'C6' interface signal is appropriate by analyzing the amplitude of the signal detected by the non-contact sensor.

The above 'C1' interface characteristics are C1 signal bit PRESENCE detection (alarm display when C1 signal is not detected), C1 signal amplitude C1_BITS_AMPLITUDE measurement, (alarm display when C1 signal amplitude exceeds the set range), C1 signal bit rate C1_BIT_RATE measurement (setting) It is collected through analysis of C1_BIT_CONTINUITY (341,1023 bits) signal of continuity and periodicity of C1 telegram (display of alarm when condition is not met).

The 'C1' interface is used to transmit a telegram from the LEU to the Balis.

In this case, the signal is polarity independent, which means that it does not affect the bit stream received over the two-input cable exchange.

The transmission signal also becomes a baseband signal in an electrical conductor. Electrical data of the electrical conductor is shown in Table 1 below.

In order to define the return loss at the 'C' interface in Table 2 above, the return loss is the current impedance at the 'C1' interface of 120Ω and the existing impedance at the 'C6' interface of 170Ω, the reflected power at the LEU and the LEU It is defined as the ratio of the incident power to the output port (expressed in dB). In this case, the reflected power does not include a direct signal intentionally transmitted from the LEU.

The above 'C6' interface characteristics determine the presence of LEU 8.82KHz carrier signal (alarm display when there is no signal), amplitude of LEU 8.82KHz carrier signal (alarm display when amplitude exceeds settable range), and frequency of LEU 8.82KHz carrier signal. It is collected through measurement (indicates an alarm when the frequency exceeds the settable range).

The 'C6' interface is the auxiliary energy input of the balis, which powers the uplink serial interface input circuitry from the LEU to the balis, using the 'C6' interface. When implementing the 'C4' interface, the 'C6' interface is used as a carrier for the occlusion signal.

In the physical transmission medium, the signal is polarity independent, meaning that the exchange of two input cables does not affect the interface function.

The electrical characteristics of the 'C' interface are: 'C6' interface link impedance amplitude detection (alarm display when amplitude changes), cable disconnection, short circuit (short), or impedance imbalance due to a cable problem between LEU and valise. It is obtained by impedance amplitude detection using a non-contact sensor at the terminal (LEU).

The purpose of this measurement is to prevent crosstalk caused by cable disconnection, short circuit (short), poor connection, and damage.

The test results are as shown in FIGS. 4 (b) and (c), and it was confirmed that there is a significant effect on the frequency of 8.82KHz of the 'C6' interface.

Therefore, looking at the 'C6' interface configuration, the power supply of the balis is absolutely necessary because the vehicle antenna passes in a relatively short time, and the biasing must be continuously applied while the LEU is connected, and consists of 1 pair and 2 cores. It is desirable to set the signal level to 22.0 +1/-2 Vpp and the signal configuration to be 8.82 ± 0.1 kHz sine wave (measured when using the reference log 170Ω).

That is, it was confirmed that the contact method could not be used to collect the characteristics of the 'C1' and 'C6' interfaces and telegram data.

However, when the characteristics of the 'C1' and 'C6' interfaces were extracted using the non-contact sensor, it was confirmed that there was no attenuation or distortion of the signal.

At this time, the return loss (at the LEU connector) with a resistive load of 170Ω is lower than 4 dB, and the frequency is sine wave at 8.82 kHz ±0.1 kHz. it's a signal

In addition, it is preferable that the second harmonic signal component (resistive load 170Ω) in the LEU is -20dBm or less, and the high frequency signal component (120Ω load impedance) in the LEU is -40dBm or less between 0.1 and 1MHz.

The LEU uses the 'C6' interface to supply auxiliary power using the Up-Link 'C1' input circuit of Balis.

It is also used to bias the telegram switch blocking 'C4' output circuit and to verify the intervalley connection in the LEU.

As shown in FIG. 5 , the telegram transmission according to the present invention is in a balis driving board (BDB), and the balis driving unit (BDU) is a line code encoder (Line Code Encoder; 31), a filter (32) and It is composed of a line transformer (33).

The data input of the Balinese drive unit (BDU) comes from an Encoder Logic Unit (ELU) and is coded as "0" = "00" and "1" = "10".

The line code encoder 31 applies a DBPL (Differential Bi-Phase Level) code format to a Balinese driving unit (BDU) input signal.

As shown in Table 3, the descriptive statistics used in the present invention determine the mean, median, mode, and variance for determining the Central Tendency of LEU, etc. Standard deviation and quartiles are used for this purpose, and frequency analysis, trend analysis, causal analysis, and correlation analysis of abnormal signals are performed.

In addition, as shown in Table 3 below, in the big data analysis of the present invention, preprocessing (eg, expert experience data, etc.) occupies 70% or more of the entire process.

Data preprocessing used in the present invention is a process for finding feature values in data and finding insights and values. Data confirmation/missing value processing/outlier processing/data distribution transformation/data unit transformation use the

For example, it is possible to find and pre-process the telegram (movement authority (packet 12) and linking (packet 5), etc.) information received from Bali, and the feature value of the data according to the 'C1' and 'C6' characteristic signals. have.

Therefore, it is necessary to organize, transform, and process the missing values of the above-mentioned raw data of the Balinese so that the characteristics of the data are well expressed, and to organize them in a state that is easy for analysis and visualization.

On the other hand, as a 'C1' and 'C6' characteristic signal classification technology, the input signal is firstly classified into a normal state and an abnormal state, and the abnormal state is secondarily classified by the cause of the abnormality.

In the present invention, a method for classifying an abnormal state uses a decision tree based on descriptive statistics, classification of supervised learning based on machine learning, and clustering of unsupervised learning. For example, you can use SVM, KMeans, etc.

As a diagnostic technique for the characteristic signal, exploratory data analysis is performed to find trends and patterns (transitions) of data.

In addition, plots, graphs, and summary statistics are used, and distribution of all variables/time-series data are plotted/variable transformation/scatterplot matrix, etc. are used to determine the correspondence of variables. It is a method to systematically check data by generating summary statistics of all variables.

In the present invention, technologies such as Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and XGBoost are selectively applied as predictive diagnostic technologies.

Looking at the model generation technology according to the present invention, a model is a method of determining an input signal, a method of determining with a decision tree according to the characteristics of data for an input signal, and a method of determining using machine learning by mixing two methods. use.

The decision tree determination method of the present invention includes a method of determining using the correlation and causality of detected values and a method of using a machine learning algorithm. There are pros and cons to both techniques, so they are used interchangeably.

2 and 6, the visualization module is the most useful technique in big data analysis. It is used in the entire process of understanding/analysis of data, checking the results of analysis to the user, and implementing the system. It is a very important factor for intuitively understanding large amounts of data.

The technology applied to the data analysis device of the present invention is collection/storage/processing/analysis/expression/utilization/management for each data processing step, and is divided into structured and unstructured data types.

For example, in case of failure of the above-mentioned ballis, text messages in unstructured data such as facility history reports such as display of similar cases and past actions, monitoring data, etc. By parsing, the consistency check and causal relationship to check the consistency with the detection data are identified.

In one embodiment, the present invention learns the experiences of field maintenance experts in the data analysis device, and supports a quick response by providing the primary judgment on the input signal and the data based thereon.

As one embodiment, as shown in FIG. 8, the experience of field maintenance experts for balancing errors from the ground equipment and on-board detection equipment starts with a small value and approaches a certain maximum value over time.

At this time, using the sigmoid function (used mainly for the output layer), if the ReLU function (Rectified Linear Unit) is used to solve the gradient vanishing problem of sigmoid and tanh, the experience of experts is sufficient. There is a problem that cannot be reflected.

That is, in the field of balis predictive diagnosis systems that require constant and accurate expert experience, inaccurate experiences due to gradient loss may not be included.

In particular, C1 signal bit PRESENCE detection (alarm is displayed when C1 signal is not detected), C1 signal amplitude C1_BITS_AMPLITUDE measurement, (alarm is displayed when C1 signal amplitude exceeds the set range), C1 signal bit rate C1_BIT_RATE measurement (when exceeding the settable range) Alarm display), C1 telegram continuity and periodicity signal C1_BIT_CONTINUITY (341,1023 bits) analysis (alarm display when condition is not met), 'C1' interface characteristics collected through analysis, LEU 8.82KHz carrier signal presence (when there is no signal) Alarm display), the amplitude of the LEU 8.82KHz carrier signal (alarm is displayed when the amplitude exceeds the settable range), and the frequency of the LEU 8.82KHz carrier signal (alarm is displayed when the frequency exceeds the settable range). The experience data of experts determined through the C6' interface characteristics, etc. can be processed with the above-described sigmoid function to sufficiently reflect the experience value.

On the other hand, when a failure occurs, it is also necessary to visualize the analysis result of the input signal at the time of the failure, the trend result for the same case, and the frequency analysis.

In the event of a failure, it can help to promptly implement maintenance by displaying similar cases and facility history reports such as past actions.

In addition, it is possible to provide basic data necessary for parts operation and device introduction by organizing the operation history and abnormal occurrence details for each facility.

As an embodiment of the present invention, as shown in FIG. 9 , the various methods described above can be used to reduce the difference between the predicted value and the observed value (eg, near the red arrow).

In addition, as an embodiment of the present invention, the real-time exponential moving average value (EMA) of telegram (moving authority (packet 12), linking (packet 5)) information transmitted to Balis, 'C1', 'C6' characteristic signals, etc. is calculated. And, based on the exponential moving average value, an allowable sensitivity indicating a measurement error range by sensitivity is determined, and an allowable lower limit value, which is the minimum range for determining a sudden signal change, is determined (pre-processing), and then an analysis process such as the above-described machine learning may proceed.

At this time, when the value of the telegram (movement authority (packet 12), linking (packet 5)) information, 'C1' , 'C6' characteristic signal, etc. transmitted to the ballis is greater than the allowable sensitivity, the alarm stage is set to "normal" ", and when it is between the permissible sensitivity and the permissible lower limit, the gas sensor value stays between the permissible sensitivity and the permissible lower limit continuously to prevent malfunction due to the sensor sensitivity. can

10: control unit
20: LEU
30 : Ballis
31: line code encoder
32 : filter
33: line transformer
41: non-contact sensor

Claims (8)

LEU that controls to transmit a telegram to Bali by receiving information on the current status of the field, departure signal, and occlusion signal or operation information of the crossing control device from the ground device;
Ballis, which provides location information, trajectory and signal information with an on-board device;
a cable connected from the LEU to the ballis;
a non-contact sensor that is installed on the cable and does not affect impedance;
Including; a control unit receiving the sampled signal through the non-contact sensor;
Using the non-contact sensor to monitor the cable state without the influence of impedance and transmit it to the control unit,
The control unit analyzes the continuity and periodicity of the telegram or the amplitude and frequency of the LEU signal through the 'C' interface used to transmit the telegram between the LEU and the Balis,
The control unit detects the Bali signal bit (PRESENCE) through the 'C' interface, measures the Bali signal amplitude (BITS_AMPLITUDE) and the Bali signal bit rate (BIT_RATE) to display an alarm when the setting range is exceeded, and A ballistic predictive diagnosis system, characterized in that it analyzes the continuity and periodicity signal (BIT_CONTINUITY) of the telegram and displays an alarm when conditions are not met. The method according to claim 1,
Balless predictive diagnosis system, characterized in that it transmits to the control unit through a failure detection board (LMB) that monitors whether the LEU is operating normally, and also determines whether there is a failure through the signal input board (DIB) and transmits it to the control unit. delete The method according to claim 1,
The control unit collects the presence of the 8.82KHz carrier signal of the LEU, the amplitude of the 8.82KHz carrier signal of the LEU, and the frequency of the 8.82KHz carrier signal of the LEU through the 'C' interface. . In the method using the Balinese predictive diagnosis system of claim 1,
using a non-contact sensor to monitor a cable state without the influence of impedance and transmit it to a control unit;
determining, by the controller, the appropriateness of the 'C' interface signal by analyzing the amplitude of the signal detected by the non-contact sensor;
The control unit analyzes the amplitude of the signal detected by the non-contact sensor to determine whether the 'C' interface signal is appropriate;
Analyze the continuity and periodicity of the telegram or the amplitude and frequency of the LEU signal through the 'C' interface used to transmit the telegram between the LEU and the Balis by analyzing the amplitude of the signal detected by the non-contact sensor,
The control unit detects the Bali signal bit (PRESENCE) through the 'C' interface, measures the Bali signal amplitude (BITS_AMPLITUDE) and the Bali signal bit rate (BIT_RATE) to display an alarm when the setting range is exceeded, and A method using a Balinese predictive diagnosis system, characterized in that by analyzing the continuity and periodicity signal (BIT_CONTINUITY) of a gram and displaying an alarm when conditions are not met. 6. The method of claim 5,
and pre-processing, by the controller, finding and preprocessing, by the controller, a feature value of data according to the 'C' interface characteristic signal and the telegram information transmitted to the balis. 7. The method of claim 6,
How the control unit first classifies the input signal as the 'C' interface characteristic signal classification technology, the normal state and the abnormal state, secondly classifies the abnormal state by abnormal cause, and classifies the abnormal state by abnormal cause A method using a Vallis predictive diagnosis system, characterized in that it uses a decision tree based on descriptive statistics, classification of supervised learning based on machine learning, and clustering of unsupervised learning. 7. The method of claim 6,
A method using a ballistic prediction diagnosis system, characterized in that the real-time exponential moving average (EMA) of the 'C' interface characteristic signal transmitted to the ballis is calculated and pre-processed.

Images