JP4648693B2 - Anomaly detection apparatus and anomaly detection method - Google Patents

Description

  The present invention relates to an abnormality detection apparatus and an abnormality detection method, and more particularly to an abnormality detection apparatus and an abnormality detection method that are effective for detecting an abnormality of a damper provided in a vehicle carriage.

  The vehicle is vibrated during its travel. For example, vibration is generated in a railway vehicle due to an irregular track or the like. In order to reduce this vibration, a spring such as an air spring or a coil spring is used for the bogie of the railway vehicle, and a damper is arranged in parallel with the spring for the damping. As this damper, an oil damper is often used. However, in the case of an oil damper, when an internal seal deterioration, oil deterioration, leakage, or the like occurs, the damping force decreases with time. Therefore, it is necessary to replace the damper at an appropriate time.

  Therefore, in the past, preventive maintenance has been performed by periodically replacing the damper. However, since the damper whose damping force is not reduced, that is, a normal damper is also replaced at the time of periodic replacement, the maintenance cost becomes expensive. In addition, it is not possible to replace a damper in which an abnormality has occurred at an early stage by periodic replacement.

  On the other hand, the use of an MT system (Mahalanobis Taguchi system) has recently attracted attention as a method for detecting product abnormalities. An anomaly detection system using an MT system generates a plurality of vector data based on characteristic quantities acquired by a sensor or the like from each of a plurality of products in a normal state, and forms a unit space by the plurality of vector data. The abnormality detection system calculates the Mahalanobis distance by substituting vector data acquired from the abnormality detection target product into the Mahalanobis distance arithmetic expression derived in the unit space, and compares the Mahalanobis distance with a predetermined threshold value. To detect product abnormalities.

As an example using such an MT system, a technique described in Non-Patent Document 1 below is known.
Koichi Onishi and two others, “Construction of an abnormality diagnosis system for race vehicles by telemetering”, 10th Quality Engineering Research Conference, 2002, p. 286-289.

  However, in the abnormality detection system using the MT system as described above, there are many errors in detecting abnormality or normality of the vehicle damper.

  Therefore, an object of the present invention is to provide an abnormality detection device and an abnormality detection method that can determine the state of a damper provided on a carriage of a vehicle with high accuracy.

  The inventor of the present application has a large variation in the number of false detections of the abnormality detection system using the conventional MT system for the abnormality detection of the vehicle damper with respect to the normal damper and any Malahalanobis distance obtained from the abnormal damper. I found out that. That is, it has been found that the state determination of the damper based on a simple threshold may not be able to separate the Mahalanobis distance between the normal damper and the abnormal damper. The inventors of the present application have found that this is caused by a variation in the Mahalanobis distance obtained from a physical quantity such as the acceleration of a damper in a vehicle traveling on various tracks such as a joint of a track.

  Therefore, an abnormality detection device according to one aspect of the present invention includes (a) a sensor that measures a physical quantity related to a damper provided in a vehicle carriage, and (b) a time series of numerical values based on the physical quantity acquired by the sensor. A processing unit that applies a plurality of bandpass filters having different bandpass characteristics and generates data based on the output of each of the plurality of bandpass filters; and (c) a unit space based on a plurality of data for a normal damper. And (d) a Mahalanobis for data generated for the abnormality detection target damper by using a Mahalanobis distance arithmetic expression based on the parameters stored in the storage means. A distance calculating means for calculating the distance; and (e) a plurality of data based on a plurality of data for the damper to be detected for abnormality. And of calculating a ratio of Mahalanobis distance becomes a predetermined distance or more of the Mahalanobis distance, the ratio is in case exceeds a predetermined ratio, an abnormality detection object damper abnormality determining means, the.

  An abnormality detection method according to another aspect of the present invention includes: (a) a first step of measuring a physical quantity related to a damper provided in a vehicle carriage by a sensor; and (b) a processing unit acquired by the sensor. A second step of applying a plurality of bandpass filters having different bandpass characteristics to a time series of numerical values based on the obtained physical quantities, and generating data based on outputs of the plurality of bandpass filters, (c) distance The calculation means calculates a Mahalanobis distance for the data generated for the abnormality detection target damper by using a Mahalanobis distance arithmetic expression derived from a unit space based on a plurality of data for a normal damper. A step, and Calculating a ratio of the distance or more to become the Mahalanobis distance, the ratio contains a fourth step of determining an abnormality when the abnormality detection target of the damper exceeds a predetermined ratio.

  According to the present invention, a plurality of Mahalanobis distances are calculated by using a plurality of time series of numerical values. This time series of numerical values is based on, for example, the acceleration or displacement of the damper. Since the state of the damper is determined by the ratio of the Mahalanobis distance greater than or equal to a predetermined distance among the plurality of Mahalanobis distances, according to the present invention, even if the Mahalanobis distance varies, the state of the damper, that is, normal and abnormal Can be discriminated with high accuracy.

In the abnormality detection apparatus and the abnormality detection method of the present invention described above, the predetermined ratio is abnormal from the ratio of the Mahalanobis distance that is equal to or greater than the predetermined distance among a plurality of Mahalanobis distances to a damper that is known to be normal. It is preferable that the ratio is a predetermined ratio in a numerical range up to a ratio of the Mahalanobis distance that is equal to or greater than a predetermined distance among a plurality of Mahalanobis distances with respect to a known damper.

  Further, in the abnormality detection device and abnormality detection method of the present invention, the processing means is a predetermined cycle of the center frequency in the band of the corresponding bandpass filter from the time series of the numerical values included in the output from each of the plurality of bandpass filters. It is preferable to derive an effective value of minutes and include the effective value in the data.

  ADVANTAGE OF THE INVENTION According to this invention, the abnormality detection apparatus and abnormality detection method which can distinguish the state of the damper provided in the trolley | bogie of a vehicle with high precision are provided.

  DESCRIPTION OF EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals.

  FIG. 1 is a diagram showing a configuration of an abnormality detection apparatus according to an embodiment of the present invention. An abnormality detection apparatus 10 shown in FIG. 1 is particularly suitable for a bogie of a railway vehicle, and includes a sensor 12, a signal conversion unit 14, a numerical conversion unit 16, a numerical processing unit 18, an MD calculation unit 20, a storage unit 22, and An abnormality determination unit 24 is provided.

  The sensor 12 is a sensor that measures a physical quantity related to a damper provided on a bogie of a railway vehicle. For example, the sensor 12 measures a physical quantity relating to an operation such as a displacement of the damper as a physical quantity relating to the damper. In the present embodiment, an acceleration sensor is employed as the sensor 12. Here, the bogie of the railway vehicle will be described. FIG. 2 is a diagram illustrating a bogie in a railway vehicle. 2A is a plan view showing the carriage 30 as viewed from above, and FIG. 2B is a side view of the carriage 30.

  As shown in FIG. 2, the carriage 30 supports the bottom surface of the body (base frame) 26 of the railway vehicle from below. Specifically, the carriage 30 has a carriage frame 32. The carriage frame 32 has a pair of side beams 32a and a pair of horizontal beams 32b. The pair of side beams 32a extends in the vehicle traveling direction on both sides in the vehicle width direction. A lateral beam 32b is connected to the approximate center in the longitudinal direction of the side beams 32a.

  A pillow spring (air spring) 34 is provided at the center of each side beam 32a in the longitudinal direction. The pillow spring 34 reduces vibration in the vertical direction of the vehicle body 26 and is connected to the bottom surface of the vehicle body 26.

  A left and right dynamic damper 36 extending in the vehicle width direction is attached to each of the horizontal beams 32b. These left and right dynamic dampers 36 reduce vibrations in the width direction of the vehicle body 26, and are connected to one end of each of the left and right dynamic dampers 36 and a center pin (traction device) 28 of the vehicle body 26.

  Axial springs 38 are provided at both ends of the side beam 32a in the vehicle traveling direction. The shaft spring 38 relieves vibrations in the vertical direction, and one end of the shaft spring 38 is connected to the shaft box 40. The axle box 40 rotatably supports a wheel shaft 42 extending in the vehicle width direction.

  Further, the carriage 30 has a shaft damper 44 arranged in parallel with each shaft spring 38. One end of the shaft damper 44 is connected to the shaft box 40, and the other end is connected to the shaft spring 38. The above-described sensor 12 is attached to each shaft damper 44.

  Returning to FIG. 1, the abnormality detection device 10 will be further described below. In the abnormality detection device 10, the same processing is performed on the acceleration measured by the sensor 12. Therefore, hereinafter, the configuration of the abnormality detection apparatus 10 will be described by focusing on the processing for the output from one sensor 12.

  The sensor 12 outputs the acceleration of the corresponding shaft damper 44 as a signal to the signal conversion unit 14. The signal conversion unit 14 includes an amplifier and the like, amplifies the signal from the sensor 12, and outputs the amplified signal to the numerical value conversion unit 16.

  The numerical value conversion unit 16 includes an A / D converter, discretizes each signal from the signal conversion unit 14, and outputs a numerical value obtained by the discretization to the numerical value processing unit 18.

  The numerical processing unit 18 performs numerical processing on the time series of numerical values within a predetermined time output from the numerical conversion unit 16. Specifically, the numerical processing unit 18 applies a plurality of bandpass filters having different pass bandwidths to a time series of numerical values. As the plurality of bandpass filters, for example, 20 bandpass filters that pass between 1 and 20 Hz in 1 Hz units can be used. Note that the bandwidth and the number of bandpass filters can be arbitrarily set.

  Further, the numerical processing unit 18 calculates an effective value for obtaining an effective value (RMS: Root Mean Square) for each output of the bandpass filter, that is, each time series of numerical values that have undergone bandpass filter processing by a plurality of bandpass filters. Execute the process. In this effective value calculation process, the effective value corresponding to each time in the time series is a plurality of numerical values including the above numerical values corresponding to the time in the time series of numerical values, and a predetermined frequency of the center frequency of the bandpass filter is determined. Calculated from the number of cycles. For example, the effective value at each time is calculated from numerical values for 10 cycles. Therefore, a time series of effective values corresponding to the number of bandpass filters is calculated through the bandpass filter process and the effective value calculation process. The numerical processing unit 18 outputs data including these effective values to the MD calculation unit 20.

The MD calculation unit 20 calculates the Mahalanobis distance (MD) from the data output by the numerical processing unit 18. Specifically, MD calculation unit 20 calculates the Mahalanobis distance D ² by the Mahalanobis distance calculating formula of the following formula (1).

  The Mahalanobis distance arithmetic expression prescribed in the equation (1) is formed by using a plurality of data for an axis damper known to be normal as data for creating a unit space and using the data for creating the unit space. It is derived in unit space.

  In the formula (1), i and j are integer indexes, and n represents an effective value used for the Mahalanobis distance, that is, the number of items of the feature amount. X indicates an item selected from effective values (features) included in the above data for the axis damper to be detected for abnormality. It should be noted that all effective values included in the above data may be selected as items, and effective values effective for discrimination from abnormal values to normal or normal may be selected from the effective values by effectiveness analysis. In equation (1), m is the average value of the items, σ is the standard deviation of the items, and both are derived from the data for creating the unit space. Further, a indicates each element in the inverse matrix of the correlation matrix derived from the data for creating the unit space.

  Parameters such as “a”, “m”, “σ”, and “n” in the Mahalanobis distance arithmetic expression defined in the equation (1) are stored in the storage unit 22.

  In the present embodiment, it is assumed that the shaft dampers whose damping force is 80% or more of the specified value are normal, the number of normal shaft dampers used is more than the number used for the bogie, and the bogies of these shaft dampers are used. Data for unit space creation is acquired by making a difference in the combination, the attachment position, the traveling speed of the vehicle, the traveling section, and the like. As a result, a unit space in which variation is allowed is generated, so that it is possible to form a unit space according to the axis damper that should be determined to be normal.

  The abnormality determination unit 24 detects an abnormality of the shaft damper 44 by using a plurality of MDs calculated for the shaft damper 44 that is an abnormality detection target. That is, the abnormality determination unit 24 detects an abnormality of the shaft damper using a plurality of MDs calculated from a plurality of time series of the above numerical values with respect to the shaft damper 44 to be detected.

  Specifically, the abnormality determination unit 24 calculates the proportion of MDs that are equal to or greater than a predetermined distance among the plurality of MDs calculated for the axis damper 44 that is the abnormality detection target. As this predetermined distance, a distance such as “4” used for separation of abnormality and normality in the MT system is used.

  The abnormality determination unit 24 determines that the abnormality detection target damper is abnormal when the calculated ratio exceeds a predetermined ratio. This predetermined ratio is the ratio of the MD that is equal to or greater than the predetermined distance among the plurality of MDs with respect to the shaft damper that is known to be normal, and the above among the plurality of MDs with respect to the damper that is known to be abnormal. It is a ratio between the ratio of MD that is equal to or greater than a predetermined distance, and is predetermined. In the present embodiment, when the rate at which MD is less than 4 is 80% or more, it is determined that the damper for abnormality detection is normal, and when the rate at which MD is less than 4 is less than 80% Therefore, it is determined that the abnormality detection target damper is abnormal.

  Hereinafter, the operation of the abnormality detection apparatus 10 will be described. In addition, the abnormality detection method according to the embodiment of the present invention will be described. FIG. 3 is a flowchart of the abnormality detection method according to the embodiment of the present invention.

  As shown in FIG. 3, in this abnormality detection method, first, the sensor 12 attached to the shaft damper 44 measures the acceleration of the shaft damper 44 (step S1). The sensor 12 outputs this acceleration as a signal to the signal conversion unit 14.

  Upon receiving the signal from the sensor 12, the signal conversion unit 14 applies signal conversion such as amplification to the signal based on the acceleration measured by the sensor 12 to the signal (step S2). The signal conversion unit 14 outputs a signal obtained by signal conversion to the numerical value conversion unit 16.

  Next, the numerical conversion unit 16 discretizes the signal from the signal conversion unit 14 (step S3), and outputs the numerical value obtained by the discretization to the numerical processing unit 18.

  Next, the numerical processing unit 18 applies the above-described bandpass filter processing (step S4) and effective value calculation processing (step S5) to the time series of the numerical values from the numerical processing unit 18. Through these processes, the numerical processing unit 18 generates time series of effective values for the number of bandpass filters, and outputs data including these time series of effective values to the MD calculation unit 20.

  Next, the MD calculation unit 20 calculates the Mahalanobis distance from the data from the numerical processing unit 18 (step S6). As described above, the item (effective value) used for calculating the Mahalanobis distance may be an item selected by effectiveness analysis or the like.

  Next, the abnormality determination unit 24 obtains a ratio of MDs of a predetermined distance or more among a plurality of MDs with respect to the axis damper 44 to be detected, and if the ratio exceeds a predetermined ratio, the axis damper 44 is abnormally detected. (Step S7).

  As described above, the abnormality detection apparatus 10 determines the state of the shaft damper 44, that is, normal and abnormal, based on a ratio exceeding a predetermined distance threshold among a plurality of MDs with respect to the shaft damper 44 to be detected. . Therefore, even if it is difficult to determine the state of the shaft damper by comparing the Mahalanobis distance with the threshold value due to variations in the Mahalanobis distance with respect to the normal shaft damper and the Mahalanobis distance with respect to the abnormal shaft damper, the shaft damper This state, that is, normal and abnormal can be determined with high accuracy. As a result, unnecessary replacement of the shaft damper can be avoided and an abnormality of the shaft damper can be detected at an early stage.

  The present invention is not limited to the above-described embodiment, and various modifications can be made. For example, the sensor 12 provided in the shaft damper 44 may be a displacement sensor (potentiometer) that measures the displacement of the shaft damper 44.

  Further, the abnormality detection device 10 can also detect an abnormality in the left-right motion damper 36 or the yaw damper by providing the sensor 12 in the left-right motion damper 36 or the yaw damper.

  In the above embodiment, the state of the abnormality detection target is determined by comparing one predetermined distance with the MD. However, the abnormality level of the abnormality detection target is determined by comparing a plurality of different predetermined distances with the MD. May be extracted in stages.

FIG. 1 is a diagram showing a configuration of an abnormality detection apparatus according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a bogie in a railway vehicle. FIG. 3 is a flowchart of the abnormality detection method according to the embodiment of the present invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Abnormality detection apparatus, 12 ... Sensor, 14 ... Signal conversion part, 16 ... Numerical value conversion part, 18 ... Numerical value processing part, 20 ... MD calculation part, 22 ... Memory | storage part, 24 ... Abnormality determination part, 30 ... Dolly, 44 ... shaft damper.

Claims (8)

A sensor for measuring a physical quantity related to a damper provided on a carriage of the vehicle;
Processing means for applying a plurality of bandpass filters having different bandpass characteristics to a time series of numerical values based on physical quantities acquired by the sensor, and generating data based on outputs of the plurality of bandpass filters;
Storage means for storing parameters of a Mahalanobis distance arithmetic expression derived from a unit space based on a plurality of the data for a normal damper;
Distance calculation means for calculating the Mahalanobis distance for the data generated for the abnormality detection target damper by using a Mahalanobis distance arithmetic expression based on the parameters stored in the storage means;
A ratio of the Mahalanobis distance that is equal to or greater than a predetermined distance among the plurality of Mahalanobis distances based on the plurality of data with respect to the damper of the abnormality detection target is calculated, and when the ratio exceeds a predetermined ratio, the abnormality detection target Determination means for determining that the damper is abnormal;
An abnormality detection device comprising: The predetermined ratio is a ratio of a plurality of Mahalanobis distances to dampers known to be abnormal from a ratio of Mahalanobis distances that are equal to or greater than the predetermined distance among the plurality of Mahalanobis distances to dampers known to be normal. The abnormality detection device according to claim 1, wherein the abnormality detection device has a predetermined ratio in a numerical range up to a ratio of the Mahalanobis distance that is equal to or greater than the predetermined distance.   The abnormality detection device according to claim 1, wherein the physical quantity is acceleration.   The processing means derives an effective value for a predetermined cycle of a center frequency in a band of a corresponding bandpass filter from a time series of numerical values included in an output from each of the plurality of bandpass filters, and the effective value is calculated as the effective value. The abnormality detection device according to any one of claims 1 to 3, which is included in data. A first step of measuring a physical quantity related to a damper provided in a vehicle carriage by a sensor;
A processing unit applies a plurality of bandpass filters having different bandpass characteristics to a time series of numerical values based on a physical quantity acquired by the sensor, and generates data based on outputs of the plurality of bandpass filters. And the steps
The distance calculation means calculates a Mahalanobis distance for the data generated for the abnormality detection target damper by using a Mahalanobis distance arithmetic expression derived from a unit space based on a plurality of the data for a normal damper. 3 steps,
When the determination unit calculates a ratio of the Mahalanobis distance that is equal to or greater than a predetermined distance among the plurality of Mahalanobis distances based on the plurality of data with respect to the abnormality detection target damper, and the ratio exceeds a predetermined ratio, A fourth step of determining that the abnormality detection target damper is abnormal;
An abnormality detection method including: The predetermined ratio is a ratio of a plurality of Mahalanobis distances to dampers known to be abnormal from a ratio of Mahalanobis distances that are equal to or greater than the predetermined distance among the plurality of Mahalanobis distances to dampers known to be normal. The abnormality detection method according to claim 5, wherein the abnormality detection method is a ratio set in advance in a numerical range up to a ratio of Mahalanobis distance that is equal to or greater than the predetermined distance.   The abnormality detection method according to claim 5, wherein the physical quantity is acceleration.   In the second step, the processing means derives an effective value for a predetermined cycle of the center frequency in the band of the corresponding bandpass filter from a time series of numerical values included in the output from each of the plurality of bandpass filters. The abnormality detection method according to claim 5, wherein the effective value is included in the data.

Images