KR20240056865A - A method of predicting the life cycle by diagnosing the state of the drive motor - Google Patents

Abstract

The present invention is a way to solve the problem of a rapid increase in the number of breakdowns and maintenance cases over a certain period of time due to long-term use and aging of the drive motors used in electric entrance doors of electric trains, etc. It relates to a life cycle prediction method through the condition diagnosis of a drive motor that collects voltage data and then uses it to diagnose the condition according to the aging of the machinery and prepare an appropriate life cycle.

Description

Life cycle prediction method by diagnosing the state of the drive motor {A method of predicting the life cycle by diagnosing the state of the drive motor}

The present invention is a way to solve the problem of a rapid increase in the number of breakdowns and maintenance cases over a certain period of time due to long-term use and aging of the drive motors used in electric entrance doors of electric trains, etc. It relates to a life cycle prediction method through the condition diagnosis of a drive motor that collects voltage data and then uses it to diagnose the condition according to the aging of the machinery and prepare an appropriate life cycle.

In general, railway vehicles are a means of transportation that runs on tracks, such as subways or trains. Each vehicle maintains a certain distance and is equipped with doors on both sides to allow passengers to get on and off when the vehicle stops at a station. For this reason, the driver in the vehicle cab operates the door opening or closing, or the door opens and closes automatically.

At this time, the entire door is not opened or closed, but the train controller and monitoring system has the history information, so when the station is stopped, the door is opened and closed by the door control unit in the corresponding direction according to the history information. A signal is transmitted, and the door control device opens and closes the door in the corresponding direction accordingly.

There are 8 sets of pocket sliding electric doors installed in the above-described railway vehicles per vehicle, and they consist of a door panel, a door operator, a door control unit (DCU), and an emergency handle.

In addition, the door operates by receiving signals generated by the Train Control and Monitoring System (TCMS), and the door closing and locked status signals are received by the door control device and the It is transmitted to the comprehensive control device.

Related to this technology, the technology of an electric door failure prediction device is proposed in conventional patent No. 2235728, and its composition, as shown in Figure 1, includes a data collection module 10, a preprocessing module 20, and an analysis prediction module ( 30), and a decision support module 40.

In addition, the data collection module 10 is configured to collect on-line data (current data) about the current state of the electric door and off-line data (past data) about the past state of the electric door. It includes a collection unit 11 and an off-line data collection unit 12.

In addition, the preprocessing module 20 receives on-line data and off-line data from the data collection module 10, and preprocesses the on-line data and off-line data respectively to form on-line data and off-line data. It is configured to extract and store characteristic factors of each line data and includes an off-line data pre-processing unit 21, a database unit 22, and an on-line data pre-processing unit 23.

And, the analysis prediction module 30 consists of a data learning unit 31 and a diagnosis prediction unit 32, and the data learning unit includes a diagnosis model unit 311 and a prediction model unit 312, The diagnosis and prediction unit includes a diagnosis unit 321 and a prediction unit 322.

In addition, the decision support module 40 receives and outputs the prediction results input from the prediction unit 322 of the analysis prediction module 30, thereby providing maintenance information for maintenance and operation management of the maintenance door. The structure that supports decision-making includes a maintenance support department (41) and an operation management support department (42).

However, the on-line data of the data collection module 10 as described above consists of at least one of the motor current value output from the motor of the door control device, event record, and encoder data of the encoder that measures the position of the motor, and the amount of data is It has the disadvantage of being large and making work difficult.

In addition, there is a disadvantage in that maintenance costs increase as the recommended durability life of the mechanical device cannot be known because there is little basis for diagnosing the condition of the door and setting an appropriate durability life according to the aging of the door.

The purpose of the present invention to improve the conventional problems described above is to set an appropriate replacement cycle for the door system according to the aging of the door, to enable systematic maintenance of the door, and to reduce maintenance costs. It provides a life cycle prediction method by diagnosing the condition of the drive motor, which not only predicts the failure of electric train electric doors in advance, but also predicts the remaining lifespan, and predicts demand for parts while improving the performance of electric gates. I'm doing it.

In order to achieve the above object, the present invention collects only the operating current corresponding to the acceleration section, constant speed section, and deceleration section among the operating current data of the electric door,

After normalizing only specific data among the collected data, main components are selected,

Analyze data trends by applying Mahalanobis distance to the selected principal components.

It provides a life cycle prediction method through condition diagnosis of the drive motor that compares the trend of the analyzed data with the predetermined trend data for each life cycle and then performs life cycle and condition diagnosis.

In addition, the data collection of the present invention provides a life cycle prediction method through condition diagnosis of the drive motor to minimize errors by relatively collecting the current signal through the DCU control signal and the voltage signal according to the PWM control signal.

In addition, the data normalization of the present invention provides a life cycle prediction method through condition diagnosis of a drive motor that performs Z-score normalization and PCA analysis to display the characteristics of the collected data in time series.

In addition, the life cycle of the present invention provides a life cycle prediction method through condition diagnosis of a drive motor calculated by comparing the trend data for each life cycle calculated by data measured by a new door.

Continuing, the data of the present invention is transmitted to the server in real time by a wireless communication unit connected to the DCU or is used to diagnose the condition of the drive motor in a configuration in which one or more of the data measured from the vehicle received for maintenance is used. Provides a life cycle prediction method through

In addition, the present invention provides a life cycle prediction method through condition diagnosis of a drive motor that measures the distribution and relative distance through the Mahalanobis distance to calculate the distance between two points.

As described above, according to the present invention, it is possible to set an appropriate replacement cycle for the door system according to the aging of the door, enable systematic maintenance of the door, reduce maintenance costs, and prevent the failure of electric train doors. In addition to predicting the remaining lifespan, it is possible to predict parts demand while improving the performance of electric doors.

Figure 1 is a block diagram showing a conventional door failure prediction device.
Figure 2 is a diagram showing the installation state of the condition diagnosis system according to the present invention.
Figure 3 is a block diagram showing the condition diagnosis system of the present invention.
Figure 4 is a flowchart showing a life cycle prediction method according to the present invention.

Hereinafter, embodiments of the present invention will be described in detail based on the attached drawings.

Figure 2 is an installation state diagram of the condition diagnosis system according to the present invention, Figure 3 is a block diagram showing the condition diagnosis system of the present invention, and Figure 4 is a flow chart showing the life cycle prediction method according to the present invention.

The condition diagnosis system (1) according to the invention includes a data collection unit (100) that is connected to the DCU for controlling the doors of the electric train (500) and collects electrical signal data consisting of current or voltage signals, and data that preprocesses the collected data. It consists of a preprocessing unit 200, a data analysis unit 300 that analyzes the preprocessed data, and a life cycle setting unit 400 that sets the life cycle through the analyzed data.

At this time, it is made up of a combination of various mechanical devices to drive the door to the electric train 500, and when an electric signal is supplied to the drive motor, the bolt connected to the drive motor rotates, and when the bolt rotates, the nut connected to the door, etc. When a ball bearing is connected to the bolt as a medium, the nut is moved in the longitudinal direction of the bolt when the bolt is rotated by the drive motor to open and close the door.

That is, the data collection unit 100 of the present invention is connected to the drive motor and is installed to measure the input and output of the operation signal transmitted to the drive motor at a certain period of time more than a certain number of times and convert it into data, and the drive motor of the electric train 500 It is connected to the PWM, which controls the input voltage signal, and the DCU, which controls the output current signal.

At this time, the data collection unit 100 is connected to a data filtering unit 150 to selectively collect data, and the data collection unit 100 is connected to the server 190 through the network 180 or during maintenance work. It may be installed to be directly collected by an operator and input directly into the server 190, and a data storage unit 130 is further connected to the data collection unit 100.

In addition, the data collection unit 100 collects data on current or voltage according to the opening and closing of the door through a collection means (not shown) such as an oscilloscope connected to the DCU.

In addition, the data collection unit 100 is connected to a data preprocessing unit 200 that preprocesses the collected data.

At this time, the data pre-processing unit 200 is installed to remove outliers in the measured data and extract data features to increase the accuracy of analysis.

In addition, the data pre-processing unit 200 includes a data analysis unit 300 that uses the collected data to diagnose the condition according to the period of use, and a life cycle setting unit 400 that sets the life cycle based on the analyzed data.

At this time, the data analysis unit 300 is installed to set algorithms (machine learning, deep learning) for data analysis and classification tasks.

In addition, the life cycle setting unit 400 is installed to diagnose the status by visualizing the data analysis and comparing it with the visualization for each usage time pre-entered by the program, and to set the life cycle through this.

In other words, the trend of the analyzed data is compared with the predetermined trend data for each life cycle and is then installed to perform life cycle and condition diagnosis.

In addition, the data collection unit 100 is installed to collect only the operating current corresponding to the acceleration section, constant speed section, and deceleration section among the operating current data of the electric door.

In addition, the data pre-processing unit 200 is installed to normalize only specific data among the collected data and then select the main component.

In addition, the data analysis unit 300 is installed to analyze the trend of data by applying the Mahalanobis distance to the selected principal components.

In addition, the life cycle setting unit 400 is installed to output the life cycle after comparing the trend of the analyzed data with the trend data for each predetermined exchange cycle.

The operation of the present invention configured as described above will be described.

As shown in Figures 2 to 4, in the case of general electric train electric entrance doors, the number of breakdowns and maintenance cases rapidly increases due to long-term use and aging, but it has become difficult to diagnose the condition according to the life cycle and set a replacement cycle according to the service life, but the present invention After collecting current data according to the opening and closing of the electric door, it is possible to diagnose the aging condition and determine the appropriate life cycle.

In addition, the present invention allows not only the age of the door to be known through condition diagnosis, but also the remaining lifespan to be known.

In addition, time series data features are extracted using the current signal acquired during the operation of the door of the present invention, Z-Score normalization and PCA analysis that are less sensitive to outliers are performed so that the extracted data is reflected at the same scale, and multivariate By applying the Mahalanobis distance to the location information of the data to express it as a relative distance to the overall distribution, it is possible to diagnose the condition of each life cycle and know the durability period.

In other words, the door of the present invention is opened and closed by an electrical signal from the DCU, and when the door is opened and closed, a data signal is measured by a sensing device such as an oscilloscope, and various mechanical devices (bolts, nuts, ball bearings) to drive the door are used. It was found that there was a difference in the data collected depending on the degree of wear.

In addition, the present invention does not use all the data of the DCU, but operates at a constant speed after the ① acceleration section and acceleration section for the door operation at the beginning of opening or closing accompanied by a change in current as shown in Table 1 below during the opening and closing operation. A filtering step is performed to selectively utilize data from the ② constant speed section, which performs a cushioning function to protect the passengers and the impact of the mechanical device before opening or immersion is completed.

[Table 1]

At this time, as shown in [Table 2], it was confirmed through collecting data for each section and life cycle according to opening and closing that the graph of the above section differs for each life cycle, and data for each life cycle was extracted from a plurality of doors. Save.

[Table 2]

In addition, the present invention secures data from a plurality of new vehicles when securing door data and secures data from a plurality of doors for each life cycle.

Continuing, a feature extraction step is performed to process and transform the collected data using formulas so that the characteristics of each data are clearly displayed.

At this time, the feature extraction method extracts time series data features such as Peak, Mean, Root mean square, Root, and Standard deviation using DCU current data that controls the door.

In addition, data such as the above are normalized so that the differences in the result values of each feature are large, so that the features are reflected on the same scale.

At this time, the normalization was performed using Z-core normalization, which is less sensitive to outliers.

In addition, in order to solve the problem of data density being distributed due to an increase in the number of features when extracting features as described above, five or less main components that maintain the variance of the raw data are used so that the total variance is about 95%.

Using the time series data extracted as above, time series data characteristics are derived, and the data extracted according to the service life and condition of each door is expressed as an overall distribution and relative distance, and the Mahalanobis distance is calculated to enable life cycle and condition diagnosis. Apply.

At this time, outliers below the minimum and above the maximum are removed using a boxplot of the Mahalanobis distance, as shown in [Table 3].

[Table 3]

Continuing, as shown in [Table 4], principal component analysis was performed using door data according to new product and usage cycle, the number of principal components and Eigen vector were extracted to explain more than 95% of the total variance, and the collected data was projected. do.

[Table 4]

And, as shown in [Table 5], the covariance matrix and average are calculated using the door data projected onto the Eigen vector, and the Mahalanobis distance of the data for each aging is calculated through this.

[Table 5]

As described above, the present invention allows for diagnosis of life cycle and life cycle status by inputting the data value of the currently used door in the same manner as above by comparing it with the data value of each life cycle and state for each life cycle pre-stored in the server. do.

At this time, the life cycle of the door is set considering the operating environment, maintenance conditions, etc., and the Mahalanobis distance is visualized to send a warning signal when the life cycle arrives. The life cycle and use state are normal. , warning, and alarm, and notify the administrator in case of an alarm.

The data collection of the present invention minimizes errors by relatively collecting the current signal through the DCU control signal and the voltage signal according to the PWM control signal.

In other words, the current value of the DCU control signal of the present invention is transmitted as a signal value determined according to each condition regardless of the state of the mechanical device. In the present invention, the voltage value is measured using the PWM control signal to determine abnormalities in the mechanical device. , the data value at this time is not reflected and a signal for maintenance, etc. is generated.

Continuing, the data of the present invention is transmitted in real time to the server by a wireless communication unit connected to the DCU or is used as one or more of the data measured from the vehicle received for maintenance, and the data can be extracted at a remote location or on-site. It becomes possible.

100...Data Collection Department
200...Data pre-processing unit
300...Data Analysis Department
400...Life cycle setting department

Claims (8)

Collects electrical signal data transmitted to the drive motor,
The collected data is converted into time series, and only specific data among the time series data are normalized and selected as main components.
Analyze data trends by applying Mahalanobis distance to the selected principal components.
A life cycle prediction method through condition diagnosis of the drive motor that compares the trend of the analyzed data with the predetermined trend data for each life cycle and then sets the replacement life cycle. The method of claim 1, wherein the electrical signal data is electrical signal data that measures the input voltage supplied to the driving motor and the output current generated during operation of the driving motor, respectively,
A life cycle prediction method through condition diagnosis of a drive motor, characterized in that the collection of the data minimizes errors by relatively collecting the current signal through the DCU control signal and the voltage signal according to the PWM control signal. The life cycle prediction method according to claim 1, wherein the data normalization involves performing Z-score normalization and PCA analysis to display the characteristics of the collected data in time series. The life cycle prediction method according to claim 1, wherein the life cycle is calculated by comparing it with data for each life cycle calculated from data measured by new products and doors for each service life. method. The driving method of claim 1, wherein the data is transmitted in real time to a server via a network by a wireless communication unit connected to the DCU or is used as one or more of data measured from a vehicle received for maintenance. Life cycle prediction method through motor condition diagnosis. The lifespan of the drive motor according to claim 1, wherein the electrical signal collects only the operating current or operating voltage corresponding to the acceleration section, constant speed section, and deceleration section among the operating current data of the electric door. Cycle prediction method. The life cycle prediction method according to claim 3, wherein the time series is performed using five main components consisting of Peak, Mean, RMSE, Root, and ST. The life cycle prediction method according to claim 1, wherein the life cycle output distinguishes the state of the device into normal, warning, and alarm, and notifies the manager in case of an alarm.

Images