Multi-gateway big data comparative analysis based gatekeeper sub-health early warning method and system

Abstract

The application discloses a kind of gate ware sub-health early warning method and system based on multi-door big data contrast analysis, which comprises the following steps: collecting vehicle door equipment IO signal to extract features and carry out data cleaning;Data set segmentation and interval division, interval division is carried out to the segmented data set with time interval m And locate the same side vehicle door equipment opened or closed at the same time;Carry out multi-door horizontal comparison, identify each interval abnormal sample based on characteristic reference value, and carry out gate ware sub-health early warning.The application can make full use of gate ware data, realize gate ware sub-health early warning, and through multi-door horizontal comparison and analysis, reduce the sub-health diagnosis interference caused by the synchronous state change of multi-door system caused by working condition change, improve the accuracy of early warning.

Description

Technical Field

[0001] This invention relates to gate controllers, and more particularly to a method and system for early warning of sub-health conditions of gate controllers based on comparative analysis of multi-gate big data. Background Technology

[0002] Rail transit trains consist of multiple cars, each with several doors on both sides. The rail door system is classified as a Class A key core component of rail transit vehicles and is also the most frequently used vehicle equipment. During operation, the door system needs to open and close frequently, and coupled with the impact of large passenger flows, this leads to frequent door malfunctions. Statistics show that door malfunctions account for more than 30% of vehicle operation failures, seriously affecting normal vehicle operation. The opening and closing of each door in a rail vehicle is controlled by a door controller within the system, and the working status of the door controller can indirectly reflect the overall working status of the door. Therefore, monitoring the key parameters of the door controller can directly or indirectly monitor the working status of the door.

[0003] The existing door controller early warning mechanism is a single-door threshold judgment method. That is, the threshold judgment function is integrated into the door controller. When one or more signals exceed the preset threshold range, the door controller will report a fault. This single-door threshold judgment method does not effectively utilize the massive amount of data uploaded by the door controller for multi-door big data comparative analysis to provide early warning of sub-health of the door controller, nor can it analyze the historical health status of the door controller. Summary of the Invention

[0004] Purpose of the invention: The present invention aims to provide a method and system for early warning of sub-health of gate controllers based on multi-gate big data comparative analysis, which can make full use of data information to realize early warning of sub-health of gate controllers and improve the accuracy of early warning.

[0005] Technical solution: The present invention provides a sub-health early warning method for gate controllers based on multi-gate big data comparative analysis, comprising the following steps:

[0006] (1) Collect the IO signals of the door device, extract features, and perform data cleaning;

[0007] (2) Dataset segmentation and interval division: The segmented dataset is divided into intervals at time interval m, and the same-side door devices that open or close at the same time are located.

[0008] (3) Perform multi-gate horizontal comparison, identify abnormal samples in each interval based on feature benchmark values, and perform sub-health warning for gate controllers.

[0009] Preferably, the feature extraction in step (1) includes calculating the initial value S, the final value E, the number of transitions N, and the transition time T for each IO signal.

[0010] Preferably, the calculation process for the initial value S, the final value E, the number of transitions N, and the transition time T is as follows: For a set of IO signals [x1, x2, ..., x...] n If the sampling interval is m milliseconds, then the initial value of the IO signal is S = x1, and the final value is E = x2. n Number of transitions N = N up +N down , where N up N represents the number of rising edges of the I / O signal. down The number of falling edges of the I / O signal; the transition time T = len([x1, x2, ..., x...). first ])×m, where x first The len() function is used to calculate the amount of data in the list of values, indicating the point at which the IO signal first changes.

[0011] Preferably, the data cleaning in step (1) includes calculating the number of samples collected for each door device and deleting door devices with fewer than a signals collected on a day, where threshold a is the lower limit of the number of samples collected by the door controller per day.

[0012] Preferably, the dataset segmentation in step (2) includes segmenting the extracted feature set into an open signal dataset and a close signal dataset by marking the open and close doors with door opening and closing labels.

[0013] Preferably, the door opening / closing label marking includes determining the door state by comparing the initial and final values ​​of the door displacement signal. When the initial value of the door displacement signal is less than the final value, it is marked as an open door state, and when the initial value of the door displacement signal is greater than the final value, it is marked as a closed door state.

[0014] Preferably, after dividing the intervals, a second data cleaning is performed to calculate the number of samples in each interval and delete intervals with fewer than b groups of door devices in each time period, where the threshold b is the number of doors of a train that open simultaneously.

[0015] Preferably, step (3) includes the following steps:

[0016] (3.1) Iterate through and calculate the baseline value of the I / O signal characteristics for each interval;

[0017] (3.2) Compare each signal feature with its corresponding reference value. If it deviates from the reference value, output the abnormal door signal sample.

[0018] Preferably, the method for calculating the reference value of the IO signal characteristics in step (3.1) is as follows:

[0019] The reference value S of the initial value S of the signal base =mode([S1,S2,…,S) iThe reference value E of the signal's final value E) base =mode([E1,E2,…,E i The baseline value N for the number of signal transitions N. base =mode([N1,N2,…,N) i The reference T for the signal transition time T. base =median([T1,T2,…,T) i The `mode()` function calculates the mode of the list, and the `median()` function calculates the median value of the list. i E represents the initial value of the I / O signal for the i-th door. i N represents the last value of the IO signal for the i-th door. i T represents the number of I / O signal transitions for the i-th door. i This represents the transition time of the IO signal for the i-th door, where i represents the number of door devices within this interval.

[0020] Preferably, step (3.2) includes classifying the IO signal into three categories based on the number of transitions: the first category is a signal that has no transitions throughout the entire process, the second category is a signal with more than 5 transitions, and the third category is a signal with more than 1 transitions between 1 and 5 transitions.

[0021] For the first type of signal, if S≠S base E≠E base and N≠N base If any one of the three conditions is met, the output will be a door abnormality signal sample;

[0022] For the second type of signal, if S≠S base E≠E base N>N base +c and N < N base If any one of the four conditions is -c, the output will be a sample of abnormal door signal.

[0023] For the third type of signal, if S≠S base E≠E base N≠N base T > T base +d and T < T base If any one of the five conditions is c, the output will be a sample of abnormal door signal; where c and d are the threshold values.

[0024] Preferably, the sub-health warning of the door controller in step (3) includes statistical analysis of abnormal signal samples of each door device. If the number of abnormalities of a single door in a single day exceeds the threshold e times, the abnormal door device and signal are pushed to the message queue for maintenance personnel to analyze the historical health status of the door controller. If a door device has abnormalities for f consecutive days, a sub-health warning is issued.

[0025] The present invention discloses a sub-health early warning system for gate controllers based on multi-gate big data comparative analysis, comprising:

[0026] The data acquisition and feature extraction module is used to acquire the IO signals of the door device and perform feature extraction.

[0027] The dataset segmentation and interval partitioning module is used to segment the extracted features according to the door opening and closing state, divide the segmented dataset into intervals with a time interval m, and locate the same-side door devices that open or close at the same time.

[0028] The abnormal sample identification module is used to compare each signal feature with its corresponding reference value. If it deviates from the reference value, it outputs an abnormal signal sample of the car door.

[0029] The sub-health warning module for door controllers is used to push messages to the message queue when the number of abnormalities in a single door in a single day exceeds the threshold e. If a door device has abnormalities for f consecutive days, a sub-health warning will be issued.

[0030] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: it can make full use of gate controller data to realize gate controller sub-health early warning, and through multi-gate horizontal comparative analysis, it reduces the sub-health diagnosis interference caused by changes in the synchronous state of multi-gate systems due to changes in operating conditions, thereby improving the accuracy of early warning. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall scheme of the present invention;

[0032] Figure 2 This is a flowchart of the method of the present invention;

[0033] Figure 3 The waveform diagrams are shown for the eight groups of IO signals in this embodiment of the invention.

[0034] Figure 4 This is a schematic diagram of the feature array constructed in an embodiment of the present invention. Detailed Implementation

[0035] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0036] like Figure 1-2 As shown, the present invention provides a method for early warning of sub-health of gate controllers based on multi-gate big data comparative analysis, comprising the following steps:

[0037] (1) Data acquisition and feature extraction, specifically including:

[0038] (1.1) Acquire door equipment I / O signals. Collect I / O signal data for all 48 doors of train No. 283 on Hohhot Line 1 during the opening and closing times of all doors on March 18, 2023, including: door opening train line, door closing train line, zero-speed train line, door enable train line, inner door indicator light, door cut-off indicator light, buzzer, and outer door indicator light. Figure 3 The waveforms for each signal are shown. Door opening / closing is achieved using the door displacement signal. The initial and final values ​​of the displacement signal are compared to determine whether the door is open or closed. Specifically, when the initial value of the door displacement signal is less than the final value, it is marked as open (represented by the number 1); when the initial value is greater than the final value, it is marked as closed (represented by the number 0).

[0039] (1.2) Extract the features of each IO signal, including the initial value S, the final value E, the number of transitions N, and the transition time T. The calculation method is as follows: Assume the IO signal list is [x1, x2, ..., x n If the sampling interval is m milliseconds, then the initial value of the IO signal is S = x1, and the final value is E = x2. n Number of transitions N = N up +N down , where N up N represents the number of rising edges of the I / O signal, i.e., the number of times the signal changes from 0 to 1. down This represents the number of falling edges of the I / O signal, i.e., the number of times the signal changes from 1 to 0. The transition time T = len([x1, x2, ..., x...) first ])×m, where x first The `len()` function is used to calculate the number of data points in the list, representing the first transition point of the I / O signal. For example, if the I / O signal list is [0,0,0,1,1,1,0,0,0], with a sampling interval of 10 milliseconds, then the initial value of the I / O signal is S = 0, the last value is E = 0, the number of transitions is N = 2, and the transition time is T = 30ms.

[0040] (1.3) Create a feature dataset and perform data cleaning. Merge the sample data from each door device after feature extraction into an array with 8 columns and 21672 rows. The 8 columns contain the sampling time, door serial number, door open / close label, IO signal name, and the corresponding initial value S, final value E, number of transitions N, and transition time T. The feature array is as follows: Figure 4 As shown, the feature dataset is initially cleaned, the number of samples for each door device is calculated, and door devices with fewer than 10 door signals collected on a given day are deleted.

[0041] (2) Dataset segmentation and interval division: The segmented dataset is divided into intervals by time interval m, and the same-side door devices that open or close at the same time are located; m is the time required for one door opening and closing, and the value is 30-90 seconds.

[0042] The cleaned array was divided into two datasets using door open / closed labels: a door open signal dataset and a door close signal dataset. Door devices on the same side that opened or closed at the same time were located using interval division. The 24-hour period was divided into 1440 time intervals with a standard 1-minute interval. After interval division, the data underwent secondary cleaning, the number of samples in each interval was calculated, and intervals with fewer than 20 door devices were deleted. The remaining sample data from each interval was stored for subsequent cross-sectional comparison of multi-door signals.

[0043] (3) Perform multi-gate horizontal comparison, identify abnormal samples in each interval based on feature benchmark values, and provide sub-health warning for the gate controller. Specifically, this includes the following steps:

[0044] (3.1) Iterate through and calculate the reference value of each group of IO signal characteristics in each interval; the method for calculating the reference value of the IO signal characteristics in step (3.1) is as follows:

[0045] The reference value S of the initial value S of the signal base =mode([S1,S2,…,S) i The reference value E of the signal's final value E) base =mode([E1,E2,…,E i The baseline value N for the number of signal transitions N. base =mode([N1,N2,…,N) i The reference T for the signal transition time T. base =median([T1,T2,…,T) i The `mode()` function calculates the mode of the list, and the `median()` function calculates the median value of the list. i E represents the initial value of the I / O signal for the i-th door. i N represents the last value of the IO signal for the i-th door. i T represents the number of I / O signal transitions for the i-th door. i This represents the I / O signal transition time of the i-th door, where i represents the number of door devices in this section. Taking a certain section of a train line during door closing as an example, its characteristic baseline values ​​are as follows: initial value is 1, final value is 0, number of transitions is 1, and transition time is 2670ms.

[0046] (3.2) Based on the number of transitions, the characteristics of the eight IO signals are divided into three categories. The first category is signals that do not transition at all, including zero-speed train lines, door-enabled train lines, door cut-off indicator lights and outer door indicator lights. The second category is signals that transition multiple times, that is, signals with more than 5 transitions, including inner door indicator lights and buzzers. The third category is signals that transition once or less, that is, signals with between 1 and 5 transitions, including open-door train lines and closed-door train lines.

[0047] (3.3) For each type of IO signal, compare each signal feature in that type with its corresponding reference value. If the following criteria are met, output the abnormal door signal sample and store it:

[0048] For the first type of signal, if S≠S base E≠E base and N≠N base If any one of the three conditions is met, the output will be a door abnormality signal sample;

[0049] For the second type of signal, if S≠S base E≠E base N>N base +c and N < N base If any one of the four conditions is -c, the output will be a sample of abnormal door signal.

[0050] For the third type of signal, if S≠S base E≠E base N≠N base T > T base +d and T < T base If any one of the five conditions (c, d) is met, the output will be a door anomaly signal sample; where c and d are threshold values. In this embodiment, the two threshold values ​​are c = 1 and d = 100ms.

[0051] Table 1 shows the results of abnormal door signal identification for train No. 283 on Hohhot Line 1 on March 18, 2023. A total of 12 abnormal data were identified, including 6 inner door indicator lights, 1 door opening train line, and 5 buzzers. The abnormal features deviated from their baseline values ​​by 1 or more, which is consistent with the actual situation.

[0052] Table 1 Results of Door Signal Anomaly Identification

[0053]

[0054] (3.4) Gate controller sub-health warning. Specifically includes:

[0055] (a) Obtain abnormal signal samples for each door device from the database and group the abnormal samples using three fields: door SN, door opening / closing label, and IO signal name.

[0056] (b) Count the number of abnormal samples in each group. If the number of abnormalities in a single door exceeds the threshold of 10 times per day, push the abnormal door equipment and signal to the message queue for maintenance personnel to analyze the historical health status of the door controller.

[0057] (c) Set up a sub-health warning mechanism for door controllers. If a door device malfunctions for three consecutive days, a sub-health warning will be issued and maintenance personnel will be notified to carry out maintenance.

[0058] This invention also provides a gate controller sub-health early warning system based on multi-gate big data comparative analysis, comprising:

[0059] The data acquisition and feature extraction module is used to acquire the IO signals of the door device and perform feature extraction.

[0060] The dataset segmentation and interval partitioning module is used to segment the extracted features according to the door opening and closing state, divide the segmented dataset into intervals with a time interval m, and locate the same-side door devices that open or close at the same time.

[0061] The abnormal sample identification module is used to compare each signal feature with its corresponding reference value. If it deviates from the reference value, it outputs an abnormal signal sample of the car door.

[0062] The sub-health warning module for door controllers is used to push messages to the message queue when the number of abnormalities in a single door in a single day exceeds the threshold e. If a door device has abnormalities for f consecutive days, a sub-health warning will be issued.

Claims

1. A method for early warning of sub-health in gate controllers based on comparative analysis of multi-gate big data, characterized in that, Includes the following steps: (1) Collect the IO signals of the door device, extract features and clean the data; the feature extraction in step (1) includes calculating the initial value S, the last value E, the number of transitions N and the transition time T of each IO signal; (2) Data set segmentation and interval partitioning: the segmented data set is divided into intervals with time interval m and the same-side door devices that open or close at the same time are located. (3) Perform multi-gate horizontal comparison, identify abnormal samples in each interval based on feature benchmark values, and perform sub-health warning for the gate controller; Step (3) includes the following steps: (3.1) Iterate through and calculate the reference value of the IO signal characteristics for each interval; the reference value of the initial value S of the IO signal. The mode of the initial values ​​of the I / O signals of the same-side door devices that open or close at the same time, and the reference value of the last value E. The mode of the last value of the I / O signal of the multi-gate device, and the base value of the number of transitions N. The mode of the number of I / O signal transitions for this multi-gate device, and the base value of the transition time T. This is the median of the I / O signal transition times for the multi-gate device. (3.2) Compare each signal feature with its corresponding reference value. If it deviates from the reference value, output the abnormal door signal sample. The step (3.2) includes classifying the IO signals into three categories based on the number of transitions: the first category is signals that do not transition at all, the second category is signals with more than 5 transitions, and the third category is signals with between 1 and 5 transitions. For the first type of signal, if the following conditions are met , and If any one of the three conditions is met, the output will be a door abnormality signal sample; For the second type of signal, if the following conditions are met , , and If any one of the four conditions is met, the output will be a door abnormality signal sample; For the third type of signal, if the following conditions are met , , , and If any one of the five conditions is met, the output will be a door abnormality signal sample; where c and d are the threshold values.

2. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 1, characterized in that, The calculation process for the initial value S, the final value E, the number of transitions N, and the transition time T is as follows: For a set of IO signals If the sampling interval is m milliseconds, then the initial value of the IO signal is... The last value Number of jumps ,in The number of rising edges of the I / O signal. Number of falling edges of the I / O signal; transition time ,in The point where the IO signal first changes. The function is used to calculate the amount of data in a list of values.

3. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 1, characterized in that, The data cleaning in step (1) includes calculating the number of samples collected for each door device and deleting door devices with fewer than a signals collected on a day, where threshold a is the lower limit of the number of samples collected by the door controller per day.

4. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 1, characterized in that, The dataset segmentation in step (2) includes dividing the extracted feature set into an open signal dataset and a close signal dataset by labeling the open and close doors.

5. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 4, characterized in that, The door opening / closing label is used to determine the door's state by comparing the initial and final values ​​of the door displacement signal. When the initial value of the door displacement signal is less than the final value, it is marked as an open door, and when the initial value of the door displacement signal is greater than the final value, it is marked as a closed door.

6. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 1, characterized in that, After dividing the data into intervals, a second cleaning process is performed. The number of samples in each interval is calculated, and intervals with fewer than b groups of door devices in each time period are deleted. The threshold b is the number of doors that a train opens simultaneously.

7. The method for sub-health early warning of gate controllers based on multi-gate big data comparative analysis according to claim 1, characterized in that, In step (3), the sub-health warning of the door controller includes statistical analysis of abnormal signal samples of each door device. If the number of abnormalities of a single door in a single day exceeds the threshold e times, the abnormal door device and signal are pushed to the message queue for maintenance personnel to analyze the historical health status of the door controller. If a door device has abnormalities for f consecutive days, a sub-health warning is issued.

8. A gate controller sub-health early warning system based on multi-gate big data comparative analysis, used to implement the gate controller sub-health early warning method based on multi-gate big data comparative analysis as described in any one of claims 1-7, characterized in that, include: The data acquisition and feature extraction module is used to acquire the IO signals of the door device and perform feature extraction. The dataset segmentation and interval partitioning module is used to segment the extracted features according to the door opening and closing state, divide the segmented dataset into intervals with a time interval m, and locate the same-side door devices that open or close at the same time. The abnormal sample identification module is used to compare each signal feature with its corresponding reference value. If it deviates from the reference value, it outputs an abnormal signal sample of the car door. The sub-health warning module for door controllers is used to push messages to the message queue when the number of abnormalities in a single door in a single day exceeds the threshold e. If a door device has abnormalities for f consecutive days, a sub-health warning will be issued.

Images