JP7004603B2 - Platform door diagnostic equipment, diagnostic methods, and diagnostic programs - Google Patents

Description

The present invention relates to a platform door diagnostic device, a diagnostic method, and a diagnostic program.

As disclosed in Patent Document 1, an operation log relating to the opening / closing operation of the platform door is acquired from the platform door that is opened / closed when the platform door is opened / closed on the platform, and the acquired operation log is used. A determination device for determining the presence or absence of an abnormality in a platform door is known.

Japanese Unexamined Patent Publication No. 2008-94299

The determination device compares the operation log data itself with the data representing the normal operation of the platform door, and determines whether or not there is an abnormality depending on whether or not the two match. With such a simple judgment method, there is a limit to further optimizing the diagnosis of platform doors.

An object of the present invention is to provide a platform door diagnostic device, a diagnostic method, and a diagnostic program capable of performing soundness diagnosis more appropriately than before.

In order to achieve the above object, the platform door diagnostic apparatus according to the present invention is
A time-series data acquisition means for acquiring time-series data representing a physical quantity related to the control of the platform door or a physical quantity representing the state of the platform door from a platform door installed on the platform.
Using the time-series data acquired by the time-series data acquisition means, a feature value calculation means for calculating a feature value representing a variation characteristic of the time-series data, and a feature value calculation means.
The feature value calculated by the feature value calculation means is used as a frequency distribution of the feature value obtained when the platform door has an abnormality and a frequency distribution of the feature value obtained when the platform door is normal. A soundness determination means for determining the soundness of the platform door by comparing with a predetermined threshold value using and
An output means for outputting the determination result of the soundness determination means, and an output means.
To prepare for.

According to the above configuration, the soundness is determined by comparing the feature value calculated from the time-series data, not the time-series data itself, with a predetermined threshold value. The feature value represents the characteristic of the fluctuation of the time series data, and the threshold value is the frequency distribution of the feature value obtained when the platform door is abnormal and the feature value obtained when the platform door is normal. Since it is determined by using the frequency distribution, the soundness can be diagnosed more appropriately than before.

Conceptual diagram showing the configuration of the platform door system according to the first embodiment Conceptual diagram which shows the structure of the platform door diagnostic apparatus which concerns on Embodiment 1. Conceptual diagram showing the contents of the diagnostic table according to the first embodiment Conceptual diagram which shows the function of the platform door diagnostic apparatus which concerns on Embodiment 1. Flowchart of diagnostic processing according to the first embodiment Conceptual diagram showing the configuration of the threshold value determination system according to the first embodiment. Flow chart of threshold value determination process according to the first embodiment A graph showing an example of an abnormal frequency distribution and a normal frequency distribution according to the first embodiment. A graph showing another example of the abnormal frequency distribution and the normal frequency distribution according to the first embodiment. A graph showing an example of an abnormal frequency distribution and a normal frequency distribution according to the second embodiment. A graph showing another example of the abnormal frequency distribution and the normal frequency distribution according to the second embodiment. Conceptual diagram which shows the function of the platform door diagnostic apparatus which concerns on Embodiment 3. Conceptual diagram showing a list of diagnosis results according to the fourth embodiment

Hereinafter, embodiments 1 to 4 of the present invention will be described with reference to the drawings. In the figure, the same or corresponding parts are designated by the same reference numerals.

[Embodiment 1]
As shown in FIG. 1, the platform door system 400 according to the present embodiment includes N platform doors 100-1 to 100-N as diagnostic target devices and the entire platform doors 100-1 to 100-N. An integrated control unit 200 that controls control and a platform door diagnostic device 300 that diagnoses platform doors 100-1 to 100-N through the integrated control unit 200 are provided.

Each of the platform doors 100-1 to 100-N is arranged along the edge of the platform in order to separate the space on the platform from the space on the railroad track through which the railroad vehicle passes. Each of the platform doors 100-1 to 100-N is opened and closed as the boarding / alighting doors of the departing and arriving trains are opened and closed. In the following description, k is an arbitrary natural number from 1 to N.

The platform door 100-k includes a motor 120-k, a slide door 110-k that is opened and closed by a rotational force generated by the motor 120-k, and an individual control unit 130-k that controls the motor 120-k.

Each of the individual control units 130-k represents a physical quantity related to the control of the platform door 100-k or a physical quantity representing the state of the platform door 100-k (however, the physical quantity related to the control of the platform door 100-k is excluded). The time-series data of the species is output to the integrated control unit 200 in association with the platform door ID information that identifies the platform door 100-k.

The platform door diagnostic apparatus 300 acquires home door ID information and a plurality of types of time-series data from each of the individual control units 130-1 to 130-N through the integrated control unit 200. Then, the platform door diagnostic apparatus 300 diagnoses the soundness of each of the platform doors 100-1 to 100-N using the platform door ID information and a plurality of types of time-series data, and outputs the diagnosis result by display. do.

The platform door diagnostic device 300 is arranged in the station office, and the user can perform necessary maintenance on the platform doors 100-1 to 100-N according to the diagnosis result output by the platform door diagnostic device 300. Can be applied.

Hereinafter, the above-mentioned plurality of time-series data output by each of the individual control units 130-1 to 130-N will be specifically described.

The individual control unit 130-k uses the physical quantity related to the control of the motor 120-k as the physical quantity related to the control of the home door 100-k, specifically, the current and voltage supplied to the motor 120-k, and the motor 120-. Time-series data for each of the torque of k, the rotation speed of the motor 120-k, and the rotation angle are output. More specifically, the individual control unit 130-k uses the current supplied to the motor 120-k as a three-phase AC motor as a q-axis component which is a torque current component and a d-axis component which is a field current component. Vector control is performed separately for. Then, the individual control unit 130-k outputs the time series data related to the vector control.

Further, the individual control unit 130-k represents the physical quantity representing the state of the motor 120-k and the state of the portion of the platform door 100-k other than the motor 120-k as the physical quantity representing the state of the platform door 100-k. It also outputs time series data for each of the physical quantities.

Specifically, the individual control unit 130-k outputs the following first time series data to 19th time series data. The sampling period of each time-series data is sufficiently shorter than the period length during which the slide door 110-k opens and closes once.

The first time series data represents the U-phase current value supplied to the motor 120-k. The second time series data represents the current value of the V phase supplied to the motor 120-k. The third time series data represents the integrated value of the U-phase, V-phase, and W-phase currents supplied to the motor 120-k. The fourth time series data represents the ratio of the current value of the U phase supplied to the motor 120-k to the current value of the V phase.

The fifth time series data represents a q-phase current value which is a q-axis component of the current supplied to the motor 120-k. The sixth time series data represents a q-phase current error, which is an error from the command value of the q-phase current. The seventh time series data represents the ratio of the command value of the q-phase current to the q-phase current error. The eighth time series data represents the ratio between the command value of the d-phase current, which is the d-axis component of the current supplied to the motor 120-k, and the error from the command value of the d-phase current.

The ninth time series data represents a command value of the U-phase voltage supplied to the motor 120-k. The tenth time series data represents a command value of the voltage of the V phase supplied to the motor 120-k. The eleventh time series data represents a command value of the W phase voltage supplied to the motor 120-k. The twelfth time series data represents a command value of the ratio of the voltage of the U phase and the voltage of the V phase.

The thirteenth time series data represents a torque limit value which is an upper limit value of the torque of the motor 120-k. The individual control unit 130-k controls to change the torque limit value at any time in consideration of the load related to the motor 120-k. The 14th time series data is a command value of the torque of the motor 120-k.

The 15th time series data represents the rotation speed of the motor 120-k. The 16th time series data represents a command value of the rotation speed of the motor 120-k. The opening / closing speed of the slide door 110-k depends on the rotation speed of the motor 120-k. Therefore, the rotation speed of the motor 120-k represents the opening / closing speed of the slide door 110-k.

The 17th time series data represents a resolver value that depends on the rotation angle of the motor 120-k. The rotation angle of the motor 120-k corresponds to one-to-one with the feeding amount of the slide door 110-k. Therefore, the resolver value of the motor 120-k represents the feeding amount of the slide door 110-k.

The 18th time series data represents the temperature of the motor 120-k as a physical quantity representing the state of the motor 120-k. The 19th time series data accommodates the slide door 110-k or the slide door 110-k accompanying the operation of the motor 120-k as a physical quantity representing the state of the portion other than the motor 120-k in the platform door 100-k. Represents the amplitude of the vibration of the door pocket. The temperature can be detected by the temperature sensor, and the vibration can be detected by the vibration sensor.

Of the first time-series data to the 19th time-series data described above, the first time-series data to the 17th time-series data correspond to time-series data representing physical quantities related to the control of the home door 100-k. The 18th time-series data and the 19th time-series data correspond to time-series data representing physical quantities representing the states of platform doors 100-k.

Next, the configuration of the platform door diagnostic apparatus 300 for diagnosing the soundness using the above-mentioned first time-series data to 19th time-series data will be specifically described with reference to FIG.

As shown in FIG. 2, the platform door diagnostic device 300 includes a communication unit 310 connected to the communication line NE. The communication unit 310 is communicably connected to the integrated control unit 200 shown in FIG. 1 via the communication line NE, and the platform door ID information is obtained from the integrated control unit 200 from the first time series data to the 19th time series data described above. Get it separately.

Further, the platform door diagnostic device 300 includes a CPU (Central Processing Unit) 320 that performs an operation for diagnosing the soundness of the platform doors 100-1 to 100-N, and a main storage unit 330 as the main memory of the CPU 320.

The main storage unit 330 stores the first time-series data to the 19th time-series data acquired through the communication unit 310. The CPU 20 performs an operation for diagnosing the soundness of the platform doors 100-1 to 100-N using the first time-series data to the 19th time-series data.

Further, in the platform door diagnostic device 300, the user instructs the display unit 340 that displays the diagnosis result of the platform doors 100-1 to 100-N and the start of the diagnosis of the platform doors 100-1 to 100-N. The input unit 360 of the above is provided.

Further, the platform door diagnostic device 300 includes an auxiliary storage unit 350 that stores a diagnostic program 351 that defines a procedure for diagnosing the soundness of platform doors 100-1 to 100-N. The CPU 320 exhibits a function of diagnosing the soundness of platform doors 100-1 to 100-N by executing the diagnostic program 351.

In addition, the auxiliary storage unit 350 also stores a diagnostic table 352 used for diagnosing platform doors 100-1 to 100-N.

As shown in FIG. 3, the diagnostic table 352 specifies outlier removal processing for excluding outliers from the time-series data for each time-series data ID 352a that identifies the first time-series data to the 19th time-series data. Outlier removal processing Specific information 352b, Feature value calculation processing that specifies the feature value calculation process that calculates the feature value that represents the characteristics of the fluctuation of the time series data Specific information 352c, Minimum threshold value 352d for the feature value, and its features It is associated with the maximum threshold value 352e for the value.

Hereinafter, the functions expressed by the CPU 320 shown in FIG. 2 by executing the diagnostic program 351 will be specifically described with reference to FIG.

As shown in FIG. 4, the CPU 320 has a function of a time-series data acquisition unit 321 as a time-series data acquisition means for acquiring the first time-series data to the 19th time-series data. The time-series data acquisition unit 321 acquires the first time-series data to the 19th time-series data for each of the platform door ID information that identifies each of the platform doors 100-1 to 100-N.

Further, the CPU 320 is an outlier removing means for performing outlier removing processing for removing outliers for each of the first time-series data to the 19th time-series data acquired by the time-series data acquisition unit 321. It has the function of the value removing unit 322.

The outlier removing unit 322 refers to the diagnostic table 352 shown in FIG. 3, and for each of the first time-series data to the 19th time-series data, the outliers associated with the time-series data ID 352a of the time-series data. Value removal processing Outlier removal processing represented by the specific information 352b is executed. Specifically, the outlier removal unit 322 executes one of the following first outlier removal process to third outlier removal process specified by the outlier removal process specific information 352b.

In the first outlier removal process, the outlier removal unit 322 first obtains the median value and the median absolute deviation of the time-series data in order to specify the outliers, and the time-series data has a median of 0 and a center. Normalize so that the absolute deviation is 1. Next, the outlier removal unit 322 compares the values of the individual numerical data constituting the normalized time series data with the predetermined values, and takes a value larger than the predetermined value. Remove the data as outliers. After removing the outliers, the outlier removal unit 322 outputs the unnormalized time-series data performed to specify the outliers to the feature value calculation unit 323, which will be described later.

In the second outlier removal process, the outlier removal unit 322 first obtains the root mean square for the entire set of time series data. Next, the outlier removing unit 322 removes the individual numerical data constituting the time-series data having an absolute value whose difference from the root mean square is equal to or larger than a predetermined value, as outliers. After removing the outliers, the outlier removal unit 322 outputs the time-series data from which the outliers have been removed to the feature value calculation unit 323 described later.

In the third outlier removal process, the outlier removal unit 322 executes a grabs rejection test. Specifically, in the outlier removal unit 322, first, the value of (maximum value-mean value) / standard deviation, which is obtained from the maximum value, the average value, and the standard deviation of the time series data, is the value of the time series data. If it is larger than a predetermined value according to the number of data points, the maximum value is removed as an outlier. After removing the outliers, the outlier removal unit 322 outputs the time-series data from which the outliers have been removed to the feature value calculation unit 323 described later.

The diagnostic program 351 shown in FIG. 2 includes a subroutine that realizes the first outlier removal process, a subroutine that realizes the second outlier removal process, and a subroutine that realizes the third outlier removal process. The CPU 320 as the outlier removal unit 322 executes the subroutine specified by the outlier removal processing specific information 352b among those subroutines.

Further, the CPU 320 is a feature value calculation means for calculating feature values representing variations of the time-series data for each of the first time-series data to the 19th time-series data whose outliers have been removed by the outlier removal unit 322. It has the function of the feature value calculation unit 323 as.

The feature value calculation unit 323 refers to the diagnostic table 352 shown in FIG. 3, and for each of the first time series data to the 19th time series data, the feature associated with the time series data ID 352a of the time series data. Value calculation process The feature value calculation process represented by the specific information 352c is executed. Specifically, the feature value calculation unit 323 executes the following first feature value calculation process or second feature value calculation process as the feature value calculation process specified by the feature value calculation process specific information 352c.

In the first feature value calculation process, the feature value calculation unit 323 calculates the correlation coefficient between the time series data and the reference time series data prepared in advance for the time series data as the feature value. Here, the reference time series data represents an ideal opening / closing operation of the platform door 100-k, and although not shown, it is assumed that the auxiliary storage unit 350 shown in FIG. 2 is prepared in advance. The correlation coefficient is obtained by dividing the covariance between the time-series data and the reference time-series data by the product of the standard deviation of the time-series data and the standard deviation of the reference time-series data.

In the second feature value calculation process, the feature value calculation unit 323 Fourier transforms the time series data to obtain the power spectrum distribution, and calculates the peak frequency, which is the maximum value of the power spectrum distribution, as the feature value.

The diagnostic program 351 shown in FIG. 2 includes a subroutine that realizes the first feature value calculation process and a subroutine that realizes the second feature value calculation process. The CPU 320 as the feature value calculation unit 323 executes the subroutine specified by the feature value calculation processing specific information 352c among those subroutines.

Further, the CPU 320 uses the feature value calculated by the feature value calculation unit 323 to function as a soundness determination unit 324 as a soundness determination means for determining the soundness of the platform door 100-k representing the feature value. Has.

The soundness determination unit 324 refers to the diagnostic table 352 shown in FIG. 3, and has a minimum threshold value 352d and a maximum value associated with the time series data ID 352a of the time series data representing the feature values calculated by the feature value calculation unit 323. The threshold value 352e is specified. Then, the soundness determination unit 324 determines the soundness depending on whether or not the feature value calculated by the feature value calculation unit 323 falls within the range from the minimum threshold value 352d to the maximum threshold value 352e.

Further, the CPU 320 has a function of an output unit 325 as an output means for outputting a determination result of the soundness determination unit 324. The output unit 325 controls the display unit 340 shown in FIG. 2 to display the determination result of the soundness determination unit 324.

Hereinafter, a method for determining the above-mentioned minimum threshold value 352d and maximum threshold value 352e used as a criterion for soundness determination by the soundness determination unit 324 will be described. The minimum threshold value 352d and the maximum threshold value 352e are predetermined by the threshold value determination system 700 shown in FIG.

As shown in FIG. 6, the threshold determination system 700 is a physical quantity related to the control of the test platform door 500 from the test platform door 500 as a simulated device imitating the platform door 100-k and the test platform door 500, respectively. Alternatively, multiple types of time-series data representing the physical quantity representing the state of the test platform door 500 (however, the physical quantity related to the control of the test platform door 500 is excluded) are acquired multiple times, and the acquired time-series data is used. A threshold determination device 600 for determining a minimum threshold 352d and a maximum threshold 352e is provided.

Similar to the platform door 100-k, the test platform door 500 has a motor 520, a slide door 510 opened and closed by the motor 520, and an individual control unit 530 that controls the motor 520. The individual control unit 530 outputs the above-mentioned first time-series data to the 19th time-series data for the test platform door 500 to the threshold value determination device 600.

The threshold determination device 600 is used for each of the time-series data acquisition unit 610 that acquires the first time-series data to the 19th time-series data from the individual control unit 530 and the acquired first time-series data to the 19th time-series data. On the other hand, the deviation value removing unit 620 that performs any of the above-mentioned first deviation value removal processing to the third deviation value removal processing, and the first time-series data to the 19th time-series data from which the deviation values have been removed, respectively. The feature value calculation unit 630 for executing the first feature value calculation process or the second feature value calculation process described above is provided.

The condition of which of the first outlier removal process to the third outlier removal process is executed by the outlier removal unit 620, and the feature value calculation unit 630 performs the first feature value calculation process and the second feature value calculation process. The conditions for executing any of the above are the same as those specified in the diagnostic table 352 shown in FIG.

Further, the threshold value determining device 600 determines the minimum threshold value 352d and the maximum threshold value 352e for each of the first time series data to the 19th time series data by statistically analyzing the feature values calculated by the feature value calculation unit 630. A statistical analysis unit 640 is provided.

Hereinafter, the threshold value determination process performed by the threshold value determination device 600 will be specifically described with reference to FIG. 7. Note that FIG. 7 shows a process of determining a minimum threshold value 352d and a maximum threshold value 352e for one time series data from the first time series data to the 19th time series data. The threshold value determination process shown in FIG. 7 is performed for each of the first time-series data to the 19th time-series data.

As shown in FIG. 7, the time-series data acquisition unit 610 acquires time-series data from the test platform door 500 in which the abnormal state known by the user is expressed (step S21). In the following, the time series data acquired from the test platform door 500 in the abnormal state will be referred to as abnormal time series data.

Next, the outlier removing unit 620 removes the outliers from the abnormal time series data, and the feature value calculation unit 630 calculates the feature values of the abnormal time series data from which the outliers have been removed (step S22). In the following, the feature values for the abnormal time series data will be referred to as the abnormal feature values.

Next, the time-series data acquisition unit 610 determines whether or not the acquisition of the predetermined MA abnormal time-series data has been completed (step S23), and the acquisition of the abnormal time-series data for the MA times is still completed. If not (step S23; NO), the process returns to step S21. In this embodiment, MA refers to a value of 10,000 or more.

The abnormal time series data for one time is acquired in the process of opening and closing the slide door 510 once. The slide door 510 is opened and closed MA times in order to acquire the abnormal time series data for MA times. When the acquisition of the abnormal time series data for MA times is completed (step S23; YES), it means that a set of abnormal feature values of MA points is obtained. In the following, the frequency distribution represented by the set of abnormal feature values of the MA point will be referred to as an abnormal frequency distribution.

Next, when the acquisition of the abnormal time-series data for MA times is completed (step S23; YES), the time-series data acquisition unit 610 acquires the time-series data from the test platform door 500 adjusted to the normal state by the user. (Step S24). In the following, the time series data acquired from the test platform door 500 in the normal state will be referred to as normal time series data.

Next, the outlier removing unit 620 removes the outliers from the normal time series data, and the feature value calculation unit 630 calculates the feature values of the normal time series data from which the outliers have been removed (step S25). In the following, the feature values for the normal time series data will be referred to as normal feature values.

Next, the time-series data acquisition unit 610 determines whether or not the acquisition of the normal time-series data for the predetermined MB times has been completed (step S26), and the acquisition of the normal time-series data for the MB times is still completed. If not (step S26; NO), the process returns to step S24. In this embodiment, MB refers to a value of 10,000 or more. The value of MB may be the same as or different from the value of MA described above.

When the slide door 510 is opened and closed MB times and the acquisition of normal time series data for MB times is completed (step S26; YES), it means that a set of normal feature values of MB points is obtained. In the following, the frequency distribution represented by the set of normal feature values of the MB points will be referred to as a normal frequency distribution.

Next, when the acquisition of the normal time series data for MB times is completed (step S26; YES), the statistical analysis unit 640 determines the abnormal frequency distribution obtained in steps S21 to S23 as the abnormal frequency distribution measurement step. The minimum threshold 352d and the maximum threshold 352e are specified by using the normal frequency distribution obtained in steps S24 to S26 as the normal frequency distribution measurement step (step S27).

Hereinafter, the process performed by the statistical analysis unit 640 in step S27 as the threshold value determination step will be specifically described with reference to FIGS. 8A and 8B. First, the statistical analysis unit 640 obtains the average value μI of the abnormal frequency distribution and the average value μN of the normal frequency distribution.

With reference to FIG. 8A, the processing of the statistical analysis unit 640 when the average value μN of the normal frequency distribution is larger than the average value μI of the abnormal frequency distribution will be described. First, the statistical analysis unit 640 obtains a curve DBI that approximates the anomalous frequency distribution and a curve DBN that approximates the normal frequency distribution. Next, the statistical analysis unit 640 obtains the intersection of the curve DBI and the curve DBN, and determines the class value THA located at the intersection as the above-mentioned minimum threshold value 352d.

When an abnormal state similar to the known abnormal state developed in the test platform door 500 is expressed in the platform door 100-k, the probability that the feature value is smaller than the class value THA is higher in the feature value. It can be said that it is higher than the probability of taking a value higher than the class value THA. Therefore, it can be determined that the platform door 100-k tends to be abnormal because the feature value is smaller than the class value THA.

Further, the statistical analysis unit 640 defines the class value THB located at μN + 2σ as the above-mentioned maximum threshold value 352e when the standard deviation of the normal frequency distribution represented by the curve DBN is σ.

When the platform door 100-k is normal, it can be said that the probability that the feature value takes a value larger than the class value THB is sufficiently small. Therefore, it can be determined that the platform door 100-k has a tendency to be abnormal because the feature value is larger than the class value THB.

As shown in FIG. 8B, when the mean value μN of the normal frequency distribution is smaller than the mean value μI of the abnormal frequency distribution, the statistical analysis unit 640 determines the class value THA located at the intersection of the curve DBI and the curve DBN. It is defined as the above-mentioned maximum threshold value 352e.

When an abnormal state similar to the known abnormal state developed in the test platform door 500 is expressed in the platform door 100-k, the probability that the characteristic value becomes larger than the class value THA is higher than the characteristic value. It can be said that it is higher than the probability of taking a value less than or equal to the class value THA. Therefore, it can be determined that the platform door 100-k tends to be abnormal because the feature value is larger than the class value THA.

Further, the statistical analysis unit 640 defines the class value THB located at μN-2σ as the above-mentioned minimum threshold value 352d when the standard deviation of the normal frequency distribution represented by the curve DBN is σ.

When the platform door 100-k is normal, it can be said that the probability that the feature value takes a value smaller than the class value THB is sufficiently small. Therefore, it can be determined that the platform door 100-k tends to be abnormal because the feature value is smaller than the class value THB.

In addition, in FIGS. 8A and 8B, the class value THA located at the intersection differs depending on the abnormal state expressed in the test platform door 500. Therefore, the abnormal state that can be discriminated by the platform door diagnostic apparatus 300 described above can be changed depending on what kind of abnormal state is expressed in the test platform door 500 in step S21.

As described above, the minimum threshold value 352d and the maximum threshold value 352e for one time series data of the first time series data to the 19th time series data have been determined. By executing the threshold value determination process shown in FIG. 7 for all of the first time series data to the 19th time series data, the minimum threshold value 352d and the maximum threshold value 352e are set for each of the first time series data to the 19th time series data. It is determined.

Hereinafter, the description of the platform door diagnostic apparatus 300 will be returned to. The minimum threshold value 352d and the maximum threshold value 352e defined in the diagnostic table 352 shown in FIG. 3 are predetermined for each time series data as described above.

Next, the diagnostic processing executed by each unit 321 to 325 shown in FIG. 4 will be described with reference to FIG. Note that FIG. 5 shows a process of diagnosing one of the platform doors 100-1 to 100-N shown in FIG. The diagnostic process shown in FIG. 5 is executed for each of the platform doors 100-1 to 100-N.

As shown in FIG. 5, first, the time-series data acquisition unit 321 enters the first time-series data from the individual control unit 130-k of the platform door 100-k in operation through the integrated control unit 200 shown in FIG. Acquire the 19th time series data (step S1). Further, the time series data acquisition unit 321 substitutes 0 for the integer type variables j and F used in the following processing (step S2).

Next, the outlier removing unit 322 removes the outliers from the j + 1th time series data, and the feature value calculation unit 323 calculates the feature value (step S3). Specifically, the outlier removing unit 322 has the outliers specified in the diagnostic table 352 shown in FIG. 3 among the first outlier removing process to the third outlier removing process for the j + 1 time series data. Removal process Outliers are removed by the process specified by the specific information 352b. The feature value calculation unit 323 has the feature values defined in the diagnostic table 352 shown in FIG. 3 in the first feature value calculation process and the second feature value calculation process for the j + 1 time series data from which the outliers have been removed. Calculation process The feature value is calculated by the process specified by the specific information 352c.

Next, the soundness determination unit 324 refers to the minimum threshold value 352d and the maximum threshold value 352e defined for the j + 1 time series data in the diagnostic table 352 shown in FIG. 3, and is calculated by the feature value calculation unit 323. It is determined whether or not the feature value is the minimum threshold value 352d or more and the maximum threshold value 352e or less (step S4).

If the feature value is the minimum threshold value 352d or more and the maximum threshold value 352e or less (step S4; YES), the soundness determination unit 324 increments the value of the integer type variable j by 1 (step S6). On the other hand, when the feature value is less than the minimum threshold value 352d or exceeds the maximum threshold value 352e (step S4; NO), the soundness determination unit 324 assigns 1 to the integer type variable F (step S5). The process proceeds to step S6.

Next, the outlier removing unit 322 determines whether or not the value of the integer type variable j has reached 19 (step S7), and if it has not reached 19 yet (step S7; NO), the process returns to step S3. .. On the other hand, if the value of the integer variable j reaches 19 (step S7; YES), the output unit 325 determines whether or not the value of the integer variable F is 0 (step S8).

If the value of the integer type variable F is 0 (step S8; YES), the output unit 325 causes the display unit 340 shown in FIG. 2 to display the diagnosis result that the platform door 100-k is normal. On the other hand, in the output unit 325, if the value of the integer type variable F is not 0 (step S8; NO), at least one of the feature values of the first time series data to the 19th time series data is less than the minimum threshold value 352d. Since it indicates that the data was present or the maximum threshold value of 352e was exceeded, the diagnosis result indicating that the platform door 100-k has a tendency to be abnormal is displayed on the display unit 340 shown in FIG.

This completes the diagnosis of one of the platform doors 100-1 to 100-N shown in FIG. The diagnostic processing shown in FIG. 5 is executed for each of the platform doors 100-1 to 100-N, and the output unit 325 obtains the diagnostic results of the platform doors 100-1 to 100-N for each of the platform doors 100-1 to 100-1. ~ 100-N is displayed on the display unit 340 according to the identification platform door ID information.

As described above, according to the platform door diagnostic apparatus 300 according to the present embodiment, it is calculated from each of the first time series data to the 19th time series data, not from the first time series data to the 19th time series data itself. The soundness of the platform door 100-k is determined by comparing the determined feature values with the predetermined minimum thresholds 352d and maximum thresholds 352e.

The feature value represents the characteristic of the fluctuation of the time series data, and the minimum threshold 352d and the maximum threshold 352e are the abnormality frequency distribution obtained when there is an abnormality in the platform door 100-k and the platform door 100-k. Since it is determined by using the normal frequency distribution obtained when it is normal, the soundness can be diagnosed more appropriately than before.

Further, according to the platform door diagnostic apparatus 300, it is found that each of the platform doors 100-1 to 100-N tends to have an abnormality instead of detecting the occurrence of an abnormality after the fact as in the conventional case. Can be detected. Therefore, when the tendency of abnormality is detected, it is possible to prevent the failure of each of the platform doors 100-1 to 100-N by performing necessary maintenance.

Further, since the outliers are removed before the calculation of the feature values for each of the first time series data to the 19th time series data, the feature values that are not affected by the outliers can be calculated. Therefore, it is possible to prevent the diagnosis from becoming excessively strict, and it is possible to prevent misdiagnosis as having a tendency to be abnormal due to the presence of accidental outliers. This also contributes to the optimization of diagnosis.

Further, in the diagnostic table 352, outlier removal processing specific information 352b and feature value calculation processing specific information 352c are individually defined for each of the first time series data to the 19th time series data. Therefore, the outliers can be removed and the feature values can be calculated by a method suitable for each of the first time-series data to the 19th time-series data. This also contributes to the optimization of diagnosis.

Further, in the diagnostic table 352, the minimum threshold value 352d and the maximum threshold value 352e are individually set for each of the first time series data to the 19th time series data. Therefore, the abnormal state expressed in the test platform door 500 in determining the minimum threshold value 352d and the maximum threshold value 352e can be changed according to the time series data ID 352a. Thereby, the platform door 100-k can be diagnosed from the viewpoint of whether or not there is a tendency of any of a plurality of known abnormal conditions.

[Embodiment 2]
In the first embodiment, as shown in FIGS. 8A and 8B, the minimum threshold value 352d or the maximum threshold value 352e is determined based on the intersection of the curve DBI that approximates the abnormal frequency distribution and the curve DBN that approximates the normal frequency distribution. However, the minimum threshold 352d and the maximum threshold 352e may be set without using the intersection. Hereinafter, a specific example thereof will be described.

As shown in FIGS. 9A and 9B, in the present embodiment, the statistical analysis unit 640 has a class value PKI located at the peak value of the curve DBI that approximates the anomalous frequency distribution, and a peak of the curve DBN that approximates the normal frequency distribution. Find the class value PKN located in the value.

Then, as shown in FIG. 9A, the statistical analysis unit 640 determines the class value PKI as the minimum threshold value 352d and the class value PKN as the maximum threshold value 352e when the class value PKI is smaller than the class value PKN. Further, as shown in FIG. 9B, the statistical analysis unit 640 defines the class value PKI as the maximum threshold value 352e and the class value PKN as the minimum threshold value 352d when the class value PKI is larger than the class value PKN.

[Embodiment 3]
In the first embodiment, the outlier removal processing specific information 352b and the feature value calculation processing specific information 352c are predetermined in the diagnostic table 352 for each of the first time series data to the 19th time series data, but they are out of order. The user may be able to select a method for removing the value and a method for calculating the feature value. Hereinafter, a specific example thereof will be described.

In the present embodiment, the user uses the input unit 360 as the selection means shown in FIG. 2 to perform the first outlier removal process to the third outlier removal for each of the first time series data to the 19th time series data. Select which of the processes is to be executed by the outlier removing unit 322. This selection operation is performed in step S3 of FIG.

Further, the user uses the input unit 360 as the selection means to calculate the feature value of each of the first time-series data to the 19th time-series data, either the first feature value calculation process or the second feature value calculation process. Select whether to execute the unit 323. This selection operation is also performed in step S3 of FIG.

As shown in FIG. 10, in the present embodiment, the CPU 320 designates the outlier removal process to the third outlier removal process selected by the input unit 360 as the outlier removal unit 322, and the first outlier removal unit 322. It has the function of the designation unit 326 that designates the feature value calculation process and the second feature value calculation process selected by the input unit 360 to the feature value calculation unit 323.

In the present embodiment, in the diagnostic table 352, the minimum threshold value 352d and the maximum threshold value 352e to be referred to when the first feature value calculation process is executed for each of the first time series data to the 19th time series data, and the first 2. It is assumed that the minimum threshold value 352d and the maximum threshold value 352e to be referred to when the feature value calculation process is executed are defined.

When the feature value (hereinafter referred to as the first feature value) obtained by the first feature value calculation process is calculated for each of the first time series data to the 19th time series data, the soundness determination unit 324 is the first. One feature value is compared with the minimum threshold value 352d and the maximum threshold value 352e defined in the diagnostic table 352 for the first feature value. Further, when the feature value obtained by the second feature value calculation process (hereinafter referred to as the second feature value) is calculated, the soundness determination unit 324 is a diagnostic table for the second feature value and the second feature value. The minimum threshold 352d and the maximum threshold 352e defined in 352 are compared.

According to the present embodiment, the viewpoint of diagnosis of the platform door 100-k differs depending on whether the first feature value or the second feature value is calculated for each of the first time series data to the 19th time series data. Can be made. Therefore, the platform door 100-k can be diagnosed from various aspects.

[Embodiment 4]
In the first embodiment, as the diagnosis result of the platform door 100-k, it is presented to the user whether it is normal or has a tendency of abnormality, but the platform door diagnostic apparatus 300 has a tendency of abnormality in the platform door 100-k. If there is, further aspects of the presumed anomaly may be presented. Hereinafter, a specific example thereof will be described.

FIG. 11 illustrates a diagnosis result list 341 displayed on the display unit 340 in the present embodiment. As the diagnosis result list 341, the diagnosis result 341b is displayed for each of the platform door ID information 341a that identifies each of the platform doors 100-1 to 100-N. The diagnostic result 341b also includes a mode of presumed abnormality.

In order to enable such a diagnosis, in the present embodiment, in the diagnosis table 352 shown in FIG. 3, when the feature value deviates from the range from the minimum threshold value 352d to the maximum threshold value 352e for each time series data ID 352a. It is assumed that the mode of the presumed abnormality is registered in advance. The specific content of the estimated abnormality mode differs depending on the magnitude of the deviation of the feature value from the minimum threshold value 352d or the maximum threshold value 352e.

The soundness diagnosis unit 324 stores which of the first time-series data to the 19th time-series data is determined to be NO in step S4 in the diagnostic process shown in FIG. The soundness diagnosis unit 324 also stores the magnitude of deviation of the feature value of the time-series data determined to be NO in step S4 from the minimum threshold value 352d or the maximum threshold value 352e.

Then, the soundness diagnosis unit 324 creates the diagnosis result list 341 shown in FIG. 11 by collating the results of these diagnosis processes with the above-mentioned diagnosis table 352. The created diagnosis result list 341 is displayed on the display unit 340 by the output unit 325.

In FIG. 11, the diagnosis result 341b of the platform door 100-k in which the platform door ID information 341a is “MMM” is not only prone to abnormality, but also has a large resistance to opening and closing the sliding door 110-k. The mode of abnormality to that effect is displayed.

The mode of this abnormality is that the feature value is from the minimum threshold value 352d to the maximum threshold value for the 13th time series data representing the torque limit value of the motor 120-k or the 14th time series data representing the command value of the torque of the motor 120-k. It can be identified by the fact that it is out of the range up to 352e.

Further, if the magnitude of the deviation of the feature value of the 13th time series data or the 14th time series data from the minimum threshold value 352d or the maximum threshold value 352e exceeds a certain value, there is a possibility that false detection of a door stop may occur. It can also be identified that there is. The false detection of door contact means that the sliding door 110-k has a large resistance to opening and closing even though no passenger or luggage has hit the sliding door 110-k. It means that the individual control unit 130-k erroneously detects that a passenger or luggage has hit.

In addition, the soundness diagnosis unit 324 has at least one of the 9th time series data, the 10th time series data, the 11th time series data, and the 12th time series data representing the voltage supplied to the motor 120-k. From the feature value, it is possible to identify an abnormal mode such that the capacitance of the smoothing capacitor included in the power supply circuit that supplies voltage to the motor 120-k is insufficient.

Further, the soundness diagnosis unit 324 determines an abnormal mode such as rattling of the slide door 110-k and loosening of the screw according to the characteristic value of the 19th time series data showing the amplitude of the vibration accompanying the operation of the motor 120-k. Can be identified.

As described above, according to the present embodiment, the diagnosis result 341b of the platform door 100-k is estimated not only as information as to whether it is normal or prone to abnormality, but also when there is a tendency to be abnormal. The aspect of the anomaly to be performed is presented to the user.

Therefore, the user can easily determine what kind of maintenance should be performed, and can quickly start maintenance. Further, even if the platform door 100-k has a failure, the restoration work can be started quickly.

The embodiment of the present invention has been described above. The present invention is not limited to this, and the modifications described below are also possible.

In the first embodiment, it is described that the home door diagnostic apparatus 300 acquires all the data from the first time series data to the 19th time series data and diagnoses the soundness of all the data, but the first time series. Data-At least one data of the 19th time series data may be acquired, and the soundness of at least one data may be diagnosed.

In the first embodiment, the test platform door 500 as a simulated device is used to determine the minimum threshold value 352d and the maximum threshold value 352e, but the test platform door 500 is not always used to determine the minimum threshold value 352d and the maximum threshold value 352e. It does not have to be used. 1st time series data to 19 o'clock output by platform doors 100-1 to 100-N when it is known that platform doors 100-1 to 100-N as a device to be diagnosed are in a normal state or an abnormal state. The minimum threshold 352d and the maximum threshold 352e may be set using the series data.

In the above embodiments 1 and 2, the minimum threshold value 352d and the maximum threshold value 352e are determined by using the abnormal frequency distribution and the normal frequency distribution, but the minimum threshold value 352d and the maximum threshold value 352e are determined by using only the abnormal frequency distribution. May be good. Specifically, the class value PKI located at the peak of the curve DBI representing the anomalous frequency distribution may be set to the minimum threshold value 352d, and the class value located at the higher end point of the curve DBI may be set to the maximum threshold value 352e.

In the first embodiment, two threshold values, a minimum threshold value 352d and a maximum threshold value 352e, are used as criteria for determining soundness, but the number of threshold values used may be one or three. It may be the above. Further, when a plurality of threshold values are used, since three or more sections are defined with respect to the feature value by those threshold values, the soundness may be evaluated in a plurality of stages such as excellent, pass, and unacceptable. ..

In the fourth embodiment, the mode of the abnormality estimated when the feature value is out of the range from the minimum threshold value 352d to the maximum threshold value 352e is defined for each time series data ID 352a, but the feature value is from the minimum threshold value 352d to the maximum threshold value. The mode of the estimated anomaly may be determined by a combination of time series data outside the range up to 352e.

In the fourth embodiment, when the platform doors 100-1 to 100-N tend to be abnormal, the mode of the presumed abnormality is presented to the user, but further, the first time series data to the 19th time series data are presented. May be used to estimate a location where an abnormality has occurred or a location where an abnormality is likely to occur and present it to the user. Specifically, the 13th time-series data or the 14th time-series data representing the torque of the motor 120-k is collated with the 17th time-series data representing the feeding amount of the slide door 110-k, and the motor 120-k is collated. By specifying the amount of extension of the slide door 110-k when the torque of the slide door 110-k increases abnormally, it is possible to estimate the location where the abnormality occurs in the rail member that guides the slide door 110-k.

In the first embodiment, the first outlier removal process to the third outlier removal process are exemplified as the process specified by the outlier removal process specific information 352b, but the method of removing the outlier is not limited to these. Various known methods can be used. Further, as the process specified by the feature value calculation process specific information 352c, the first feature value calculation process and the second feature value calculation process are exemplified, but the method for calculating the feature value is not limited to these, and various known methods are known. Method can be used.

In the first embodiment, it is assumed that the platform door diagnostic device 300 is arranged at the station where the platform doors 100-1 to 100-N are installed, specifically, in the station office, but the platform door diagnostic device 300 The place where the platform screen doors are installed is not particularly limited. The platform door diagnostic device 300 may be arranged at a place other than the station office at the station, or may be arranged at a place other than the station. The platform door diagnostic device 300 may be located at the railway company or the management center of the platform door device. Platform doors of a plurality of stations may be diagnosed by one platform door diagnostic device 300 arranged in the management center.

By installing the diagnostic program 351 shown in FIG. 2 on a computer, the computer can function as a platform door diagnostic device 300. The diagnostic program 351 can be distributed through a communication line, or can be stored and distributed in a computer-readable recording medium such as an optical disk or a flash memory.

100-1 to 100-N ... Home door (device to be diagnosed), 110-1 to 110-N ... Slide door, 120-1 to 120-N ... Motor, 130-1 to 130-N ... Individual control unit, 200 ... Integrated control unit, 300 ... Home door diagnostic device, 310 ... Communication unit, 320 ... CPU, 321 ... Time series data acquisition unit (time series data acquisition means) 322 ... Outlier value removal unit (outlier value removal means) 323 ... Feature value calculation unit (feature value calculation means), 324 ... Soundness determination unit (health determination means), 325 ... Output unit (output means), 326 ... Designation unit, 330 ... Main storage unit, 340 ... Display unit, 341 ... Diagnosis result list, 341a ... Home door ID information, 341b ... Diagnosis result, 350 ... Auxiliary storage unit, 351 ... Diagnosis program, 352 ... Diagnosis table, 352a ... Time series data ID, 352b ... Outlier removal processing specific information, 352c ... Feature value calculation processing Specific information, 352d ... Minimum threshold, 352e ... Maximum threshold, 360 ... Input unit (selection means), 400 ... Home door system, 500 ... Test home door (simulation device), 510 ... Slide door, 520 ... motor, 530 ... individual control unit, 600 ... threshold determination device, 610 ... time series data acquisition unit, 620 ... outlier value removal unit, 630 ... feature value calculation unit, 640 ... statistical analysis unit, 700 ... threshold determination system, DBI, DBN ... curve, THA, THB, PKI, PKN ... class value.

Claims (8)

A time-series data acquisition means for acquiring time-series data representing a physical quantity related to the control of the platform door or a physical quantity representing the state of the platform door from a platform door installed on the platform.
Using the time-series data acquired by the time-series data acquisition means, a feature value calculation means for calculating a feature value representing a variation characteristic of the time-series data, and a feature value calculation means.
The feature value calculated by the feature value calculation means is used as a frequency distribution of the feature value obtained when the platform door has an abnormality and a frequency distribution of the feature value obtained when the platform door is normal. A soundness determination means for determining the soundness of the platform door by comparing with a predetermined threshold value using and
An output means for outputting the determination result of the soundness determination means, and an output means.
A platform door diagnostic device equipped with. The feature value calculating means has a first feature value as the feature value of the time series data and a second feature value of the time series data that represents a feature different from the feature represented by the first feature value. It has a function to calculate the feature value and
When the soundness determination means calculates the first feature value by the feature value calculation means, the first feature value is compared with the first threshold value which is the threshold value for the first feature value. When the second feature value is calculated by the feature value calculating means, the second feature value is compared with the second threshold value which is the threshold value for the second feature value.
The platform door diagnostic device according to claim 1. A selection means for the user to select whether to have the feature value calculation means calculate the first feature value or the second feature value.
2. The platform door diagnostic apparatus according to claim 2. The time-series data acquisition means acquires a plurality of types of time-series data representing the physical quantities that are different from each other.
The feature value calculation means calculates the feature value for each of the plurality of types of the time series data acquired by the time series data acquisition means.
The soundness determining means compares each of the feature values calculated by the feature value calculating means with the threshold value determined for the feature value.
The platform door diagnostic apparatus according to any one of claims 1 to 3. The time-series data acquisition means acquires the time-series data from each of the plurality of platform doors installed on the platform.
The feature value calculation means calculates the feature value for each platform door using the time-series data acquired from each of the plurality of platform doors.
The soundness determination means determines the soundness for each platform door using the feature value calculated for each platform door.
The output means outputs the determination result of the soundness determination means for each platform door.
The platform door diagnostic apparatus according to any one of claims 1 to 4. A known abnormal state is expressed in a diagnosis target device or a simulation device that imitates the diagnosis target device, and the diagnosis target device or the simulation device in the abnormal state is used to display the diagnosis target device or the simulation device. Time-series data representing the physical quantity related to the control of the device or the physical quantity representing the state of the diagnosis target device or the simulation device is acquired multiple times, and each of the acquired time-series data is used to characterize the fluctuation of the time-series data. Anomalous frequency distribution measurement step for obtaining an abnormal frequency distribution, which is a frequency distribution of the feature value, by calculating a feature value representing
Using the abnormal frequency distribution, a threshold value determination step for determining a threshold value for determining the soundness of the diagnostic target device for the feature value, and a threshold value determination step.
The time series data representing the physical quantity related to the control of the diagnosis target device or the physical quantity representing the state of the diagnosis target is acquired from the operation target device, and the acquired time series data is used to obtain the time series. A diagnostic step for diagnosing the soundness of the device to be diagnosed based on a comparison between the characteristic value calculated by calculating the characteristic value representing the characteristic of data fluctuation and the threshold value determined in the threshold determination step. ,
Diagnostic methods, including. Before the threshold determination step,
From the diagnosis target device or the simulation device in a normal state, the time-series data representing the physical quantity related to the control of the diagnosis target device or the simulation device or the physical quantity representing the state of the diagnosis target device or the simulation device is generated a plurality of times. Normal frequency distribution measurement to obtain the normal frequency distribution, which is the frequency distribution of the feature values, by calculating the feature values that represent the characteristics of the fluctuation of the time series data by using each of the acquired time series data. Step,
Including
In the threshold value determination step, the intersection of the curve represented by the normal frequency distribution and the curve represented by the abnormal frequency distribution is obtained, and the intersection is used to determine the threshold value.
The diagnostic method according to claim 6. On the computer
A time-series data acquisition function that acquires time-series data representing the physical quantity related to the control of the platform door or the physical quantity representing the state of the platform door from the platform door installed on the platform.
Using the time-series data acquired by the time-series data acquisition function, a feature value calculation function for calculating a feature value representing a variation characteristic of the time-series data, and a feature value calculation function.
The feature value calculated by the feature value calculation function is used as a frequency distribution of the feature value obtained when the platform door has an abnormality and a frequency distribution of the feature value obtained when the platform door is normal. The soundness determination function for determining the soundness of the platform door by comparing with a predetermined threshold value using
A diagnostic program that realizes.

Images