JP7628028B2 - Track state estimation method, track state estimation device, and vehicle - Google Patents

Description

The present invention relates to a track state estimation method, a track state estimation device, and a vehicle.

One of the causes of derailment of vehicles running on rails is irregularity of the rails relative to their original position (hereafter referred to as "track irregularity"). As track irregularity directly affects derailment, it is important to properly inspect and maintain track irregularity in order to reduce the risk of derailment.

Inspection of track displacement (hereinafter also referred to as "track inspection") is carried out, for example, using a hand-pushed track inspection device. However, when the rail to be inspected becomes long, it becomes difficult to inspect it frequently. Furthermore, track inspection using a hand-pushed track inspection device is a static inspection carried out under no-load conditions, which differs from the conditions during a derailment. For this reason, it is desirable to be able to carry out dynamic inspections frequently while the vehicle is running.

One method for frequently performing dynamic inspections while a vehicle is in motion is to use an operating vehicle to perform track inspections. For example, Patent Document 1 discloses a track inspection device that uses a vehicle, in which a detector unit made up of various detectors is attached to the bogie frame of the vehicle, and five types of track displacement are measured based on the inertial sine wave method while the vehicle is in motion.

Patent No. 3411861

However, while the technology described in Patent Document 1 can measure five types of track displacement using a detector unit, the detection mechanism using the detector unit is complex and expensive. In addition, because a large detector unit needs to be mounted, the measurement position may be far from the wheels depending on the structure of the vehicle, and there is a possibility that deformation of the rail when the vehicle is loaded may not be fully detected.

The present invention has been made in consideration of the above problems, and the object of the present invention is to provide a track condition estimation method, a track condition estimation device, and a vehicle that can easily and accurately measure tracks using an operating vehicle.

In order to solve the above problem, according to one aspect of the present invention, a track condition estimation method is provided that includes an inclination angle estimation step of estimating the horizontal inclination angle and the longitudinal inclination angle of the wheel set based on the three-axis acceleration and the three-axis angular velocity measured by an inertial measurement unit installed on the axle box of the vehicle, and a track condition calculation step of calculating track condition information representing the track condition based on the horizontal inclination angle and the longitudinal inclination angle of the wheel set.

In the tilt angle estimation step, the translational acceleration components during vehicle travel are extracted by filtering from the three-axis acceleration measured by the inertial measurement unit, and the error covariance matrix of the observation model of the Kalman filter is set based on the extracted translational acceleration components. The horizontal tilt angle and longitudinal tilt angle of the wheel set are estimated from the three-axis acceleration and three-axis angular velocity using a Kalman filter formulated from a state space model composed of the observation model in which the error covariance matrix is set. In the track state calculation step, the planarity deviation is calculated as track state information based on the horizontal tilt angle of the wheel set.

The translational acceleration components may also be extracted using band-pass filtering to remove frequency components that are predominant in track displacement from the lateral acceleration and longitudinal acceleration among the three-axis accelerations.

The error covariance matrix may be set by adding values obtained by multiplying the extracted translational acceleration components by a predetermined coefficient to the diagonal elements.

In addition, in order to solve the above problem, according to another aspect of the present invention, a track condition estimation device is provided that includes an inclination angle estimation unit that estimates the horizontal inclination angle and the longitudinal inclination angle of the wheel set based on the three-axis acceleration and the three-axis angular velocity measured by an inertial measurement unit installed on the axle box of the vehicle, and a track condition calculation unit that calculates track condition information that represents the track condition based on the horizontal inclination angle and the longitudinal inclination angle of the wheel set.

Furthermore, in order to solve the above problem, according to another aspect of the present invention, a vehicle is provided that includes an inertial measurement unit installed on an axle box of the vehicle, a track state estimation device having an inclination angle estimation unit that estimates the horizontal inclination angle and the longitudinal inclination angle of the wheel set based on the three-axis acceleration and the three-axis angular velocity measured by the inertial measurement unit, and a track state calculation unit that calculates track state information that represents the track state based on the horizontal inclination angle and the longitudinal inclination angle of the wheel set.

As described above, the present invention allows for easy and accurate track inspection using an operating vehicle.

1 is a schematic diagram showing a schematic configuration of a vehicle according to an embodiment of the present invention; FIG. 1 is a schematic diagram for explaining a planarity displacement. FIG. 2 is an explanatory diagram for explaining measurement values obtained by an inertial measurement unit. FIG. 2 is a schematic cross-sectional view of a wheelset for explaining values measured by an inertial measurement unit. 13 is a graph showing an example of calculation of an index _Rx representing translational acceleration. 13 is a graph showing an example of calculation of an index _Ry representing translational acceleration. FIG. 2 is a functional block diagram showing a configuration of a track state estimation device according to the embodiment. 4 is a flowchart illustrating an example of a track state estimation method according to the embodiment. 13 is a graph showing the estimation result of planarity deviation in the example.

The preferred embodiment of the present invention will be described in detail below with reference to the attached drawings. Note that in this specification and the drawings, components having substantially the same functional configuration are designated by the same reference numerals to avoid redundant description.

[1. Vehicle configuration]
First, a schematic configuration of a vehicle 1 according to an embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a schematic diagram showing a schematic configuration of a vehicle 1 according to the present embodiment.

The vehicle 1 according to this embodiment is configured to be able to run on a pair of extended rails 5. As shown in FIG. 1, the vehicle 1 is configured with a car body section 10 consisting of a body structure which is the main structural part of the vehicle, and a bogie section 20 which supports the car body section 10 and has a running mechanism. In the following, the direction in which the vehicle travels (front-rear direction) is defined as the X direction, the width direction (left-right direction) as the Y direction, and the height direction (up-down direction) as the Z direction.

Four wheel pairs are arranged in the vehicle travel direction on the bogie section 20 of the vehicle 1 shown in Figure 1. Each wheel pair is formed by connecting two wheels 23 that are paired in the width direction by a wheel set 25. An axle box 27, which is the bearing part of the wheel set 25, is provided on both ends of each wheel set 25. The axle box 27 is connected to the bogie frame 21 by a coil spring 29. Note that in Figure 1, the axle box 27 and coil spring 29 are shown for only one wheel 23 to show the wheel set 25, but all of the wheels 23 are equipped with the axle box 27 and coil spring 29.

The vehicle 1 according to this embodiment is equipped with an inertial measurement unit (IMU) 30 on the axle box 27 of at least one wheel 23. The inertial measurement unit 30 is a device that detects three-dimensional inertial motion (translational motion and rotational motion in three orthogonal axial directions). The inertial measurement unit 30 detects translational motion using an acceleration sensor and detects rotational motion using an angular velocity sensor. Measurement values by the inertial measurement unit 30 are output to a track condition estimation device (track condition estimation device 100 in FIG. 3) that estimates the track condition of the rail 5 on which the vehicle 1 runs.

The vehicle 1 according to this embodiment is also equipped with a position detection device (not shown) that detects the absolute position of the vehicle 1 on the rails 5. The position detection device may be, for example, an RFID (Radio Frequency Identification) reader. In this case, the absolute position of the vehicle 1 can be detected by reading an RFID tag that is set in advance at the middle position in the width direction of the pair of rails 5 with the RFID reader. Note that the method of detecting the absolute position of the vehicle 1 is not limited to this example, and well-known technology may be used. For example, position information obtained using radio waves from an artificial satellite may be used as the absolute position of the vehicle 1.

2. Estimation of orbital conditions
The track conditions of a rail include track deviation, alignment deviation, elevation deviation, level deviation, and gauge deviation. In the following, a case where track deviation, which contributes greatly to derailment, is estimated as a track condition, will be described. Track deviation, as shown in FIG. 2, refers to a twisted state with respect to the plane of the track, and is expressed by a level difference CL(x) between the left and right rails 5 at two points (e.g., points P1 and P3, and points P2 and P4) that are spaced apart by a predetermined distance d in the direction of vehicle travel. Specifically, since the fixed wheelbase of a general vehicle is about 5 m, track deviation Tw(x) is expressed by the following formula (1).

Tw(x)=CL(x)-CL(x-5)...(1)

Here, assuming that there is no floating of the wheels 23, the vertical displacement of the wheel set 25 can be considered to be equal to the vertical displacement of the track. Therefore, by estimating the roll angle θx of the wheel set 25 based on the measurement value of the inertial measurement unit 30 installed on the axle box 27, the level difference CL(x) can be estimated from the following formula (2). In formula (2), L represents the standard gauge length.

CL(x)=Ltanθx...(2)

In this embodiment, the track displacement immediately below the wheels is measured using only the measurement values of the inertial measurement unit 30 installed on the axle box 27. By installing the inertial measurement unit 30 on the axle box 27, the track displacement immediately below the wheels can be measured, but it also becomes easier to measure noise such as impulsive vibrations caused by vehicle travel. For this reason, the track displacement estimated from the measurement values of the inertial measurement unit 30 is prone to errors due to acceleration components (hereinafter also referred to as "translational acceleration components"), including impulsive vibration responses caused by vehicle travel, which are measured by the inertial measurement unit 30 translating in the measurement axis direction. Therefore, in this embodiment, a method is presented for estimating planarity displacement by reducing the influence of errors due to the translational acceleration components during vehicle travel contained in the measurement values of the inertial measurement unit 30.

The following describes the details of the track state estimation method for estimating runway deviation as a track state, as well as the track state estimation device and track state estimation method that execute the track state estimation process based on the track state estimation method.

[2-1. Orbit state estimation method]
First, a track state estimation method for estimating track states according to this embodiment will be described with reference to Figs. 3 to 6. Fig. 3 is an explanatory diagram for explaining measured values by the inertial measurement unit 30. Fig. 4 is a schematic cross-sectional view of a wheelset for explaining measured values by the inertial measurement unit 30. Fig. 5 is a graph showing an example of calculation of an index _Rx representing translational acceleration. Fig. 6 is a graph showing an example of calculation of an index _Ry representing translational acceleration.

In the following description, the measured values of the three-axis acceleration at time k measured by the inertial measurement unit 30 are a _xk , a _yk , and a _zk , and the measured values of the three-axis angular velocities are ω _xk , ω _yk , and ω _zk . As shown in Fig. 3 , the roll angle at time k is θ _k , the pitch angle is φ _k , and the yaw angle is ψ _k .

In this embodiment, a Kalman filter is used to estimate the roll angle θx of the wheel set 25 from the measurement values of the inertial measurement unit 30. The Kalman filter is one of the data assimilation methods, and is a method that sequentially estimates state quantities that minimize the error with the observation quantities using Bayesian estimation. The roll angle estimate value and pitch angle estimate value obtained from the measurement values of the three-axis acceleration are used as the observation quantities, and the Kalman filter can be formulated from a time update equation based on the integration of the roll angular velocity and pitch angular velocity in absolute coordinates.

The state vector x _k is defined by the following equation (3).

The observation vector _yk can be expressed as in the following formula (4) based on the relationship of the projection of the gravitational acceleration g in the three axial directions obtained by extracting the low-frequency components of the measured values _axk , _ayk , and _azk of the three-axis acceleration shown in Fig. 4. In the observation vector _yk , _θ0k is an estimated value of the roll angle at time k, and _φ0k is an estimated value of the pitch angle at time k.

The gravitational acceleration projection components of the three-axis acceleration measurement values a _xk , a _yk , and a _zk are low-frequency acceleration components, for example, components in the range of 0 to 1 Hz. The tilt angle estimation unit 110 can extract the gravitational acceleration projection components from the three-axis acceleration measurement values a _xk , a _yk , and a _zk using, for example, a low-pass filter.

The time update equation is derived from the integral of the angle in the absolute coordinate system, and is a nonlinear relational equation for the state vector. Therefore, in this embodiment, an extended Kalman filter that performs linearization around the state vector x _k−1 is used. In this case, the system equation is expressed by the following equation (5), where _{v k} is the process noise.

Furthermore, when the observation noise is w _k and the observation matrix is H _k , the observation equation is expressed by the following formula (7).

Using a Kalman filter based on the state space model expressed by the above equations (5) to (7), the roll angle θ _k and the pitch angle φ _k can be estimated from the measured values a _xk , a _yk , and a _zk of the three-axis accelerations and the measured values ω _xk , ω _yk , and ω _zk of the three-axis angular velocities.

Here, in this embodiment, in order to reduce the influence of errors due to the translational acceleration components when the vehicle is traveling that are included in the measurement values of the inertial measurement unit 30, the tilt angle estimator 110 extracts the translational acceleration components from the measurement values a _xk , a _yk , a _zk of the three-axis acceleration when the vehicle is traveling, which are error factors, by filtering processing, and sets the observation error covariance matrix of the Kalman filter based on the extracted translational acceleration components.

The inventors of the present application have investigated error factors in angle estimation and found that the impact response that occurs when the wheel flange contacts the rail at a sharp curve portion is a factor in error in the angle estimation. In the above formula (5) for calculating the angle from the integral of the angular velocity and the above formula (7) for calculating the angle from the low-frequency component of acceleration (i.e., the projected component of gravitational acceleration), the process noise _vk and the observation noise _wk are assumed to be Gaussian noise with a mean value of zero. This noise setting represents the error components of each relational expression. The process noise _vk means the measurement error of the angular velocity data and the integration error. The observation noise _wk means the measurement error of the acceleration data and the estimation error of the projected component of gravitational acceleration extracted from the acceleration.

When the wheel flange comes into contact with the rail in a sharp curve, both the angular velocity measurement value and the acceleration measurement value may show an impulsive response, but the impulsive response is characterized by being prominent in all frequency bands. Here, the above formula (5) is valid regardless of the presence or absence of an impulsive response, but the observation vector y _k in the above formula (7) is an angle calculated from the gravitational acceleration projection component, and becomes an error factor when translational acceleration acts on the inertial measurement unit 30. That is, in such a traveling section, it is considered that the actual response can be appropriately expressed by reducing the reliability of the above formula (7) (in other words, increasing the reliability of the above formula (5)) by setting the observation noise w _k to a large value.

Therefore, in this embodiment, the accuracy of the angle estimation value is improved by setting the observation noise _wk according to the magnitude of the translational acceleration acting on the inertial measurement unit 30. Specifically, indices _Rx and _Ry representing the translational acceleration are introduced, and the covariance matrix R of the observation noise _wk according to the magnitude of the translational acceleration component is sequentially set.

The calculation accuracy of the roll angle θ _0k and pitch angle φ _0k of the observation vector y _k is governed by the lateral acceleration a _yk and the longitudinal acceleration a _xk , respectively, and errors may occur in both cases due to translational acceleration. For the lateral acceleration a _yk , the shock response when passing through a sharp curve appears as translational acceleration, and for the longitudinal acceleration a _xk , the acceleration/deceleration required to keep the vehicle speed constant appears as translational acceleration. Therefore, indexes R _x and R _y representing the respective translational accelerations are defined.

The index R _x is a value obtained by extracting only high-frequency components from the measured value a _x of the longitudinal acceleration using a band-pass filter, squaring the extracted value, and then taking a moving average for a predetermined section (number of moving average samples). The index R _y is a value obtained by extracting only high-frequency components from the measured value a _y of the lateral acceleration using a band-pass filter, squaring the extracted value, and then taking a moving average for a predetermined section (number of moving average samples). Since the indexes R _x and R _y are indexes for correcting low-frequency responses (i.e., the tilt angle estimated from the gravitational acceleration projection component), the estimated value is stabilized by squaring and taking a moving average to remove high frequencies. For example, the high-frequency components to be extracted may be frequency components of 1 to 7 Hz, and the number of moving average samples may be 250 sample times.

The shock response or the response due to acceleration/deceleration is a forced response acting from the outside, and can be assumed to be dominant in all frequency components. On the other hand, it is considered that components due to angle changes are dominant in the low frequency band, for example, less than 1 Hz, and measurement noise is dominant in the high frequency band, for example, above 7 Hz. For this reason, in this embodiment, a bandpass filter is used to remove the frequency components in these bands.

In detail, assuming that the vehicle speed is almost constant, the average vehicle speed in the measurement data to be filtered is V _ave [m/s], and the maximum spatial frequency of the track irregularity is F _space [cycle/m], the cutoff frequency F _time [Hz] of the time history waveform can be expressed by the following equation (8).

F _time = F _space × V _ave ... (8)

The track irregularity has the following characteristics: the maximum spatial frequency of the track irregularity F _space =1, and the vehicle speed V _ave =1. Therefore, in estimating the track irregularity, the cutoff frequency F _time that is the boundary between the low frequency band and the high frequency band is set to 1 Hz.

The high-frequency cutoff frequency F _{time_h} can be expressed as an empirical formula having generality as shown in the following formula (9).

F _{time_h} = 7 x F _space x V _ave ... (9)

The moving average sample number s of the indices R _x and R _y can be expressed as an empirical formula by the following formula (10) using the sampling frequency fs [Hz] of the measurement value.

s=fs/(2×F _time ) (10)

For example, when the sampling frequency fs=500 Hz and the cutoff frequency F _time =1 Hz, the number of moving average samples s is 250 sample times.

Fig. 5 shows an example of an index _Rx representing translational acceleration, and Fig. 6 shows an example of an index _Ry representing translational acceleration. In Fig. 5, positions where the value of the index _Rx shown in a dashed line frame is large indicate that the vehicle is being influenced by acceleration/deceleration to keep the vehicle speed constant. Also, in Fig. 6, positions where the value of the index _Ry shown in a dashed line frame is large indicate that the vehicle is being influenced by an impact when passing through a sharp curve.

Next, the calculated indices R _x and R _y are used to sequentially set the observation noise w _k according to the magnitude of the translational acceleration component. The error covariance matrix R(k) may be set by adding a value obtained by multiplying the extracted translational acceleration component by a predetermined coefficient to the diagonal components. Specifically, the variance-covariance matrix R(k) of the observation noise w _k at time k is set as shown in the following formula (11) using indices R _y (k) and R _x (k) representing the translational acceleration components, and coefficients K _y and K _x for the indices R _y (k) and R _x (k).

In addition, the process noise Q may be set to be constant regardless of time, for example, as shown in the following equation (12).

The coefficients K _y and K _x are values for setting the noise variance of the observation equation expressed by the above formula (7) from the indices R _y (k) and R _x (k), and cannot be derived theoretically. Therefore, the coefficients K _y and K _x are set to values that maximize the estimation accuracy of the planarity deviation as optimal values. The process noise Q is also appropriately set based on the value that maximizes the estimation accuracy of the planarity deviation.

In this manner, in this embodiment, the variance-covariance matrix R(k) of the observation noise w _k at time k is set as in the above formula (11), and the observation noise w _k of the observation equation shown in the above formula (7) is sequentially set according to the magnitude of the translational acceleration component. This makes it possible to reduce the influence of errors due to the translational acceleration component during vehicle travel, which is included in the measurement value by providing the inertial measurement unit 30 on the axle box 27. As a result, the observation vector y _k at time k, that is, the estimated value θ _0k of the roll angle and the estimated value φ _0k of the pitch angle, can be obtained with high accuracy. If the roll angle θ _k and the pitch angle φ _k are estimated by the Kalman filter based on the state space model expressed by formulas (5) to (7) and in which the variance-covariance matrix R(k) of the observation noise w _k is set by formula (11), the planarity displacement Tw(x) can be calculated using the above formulas (1) and (2).

[2-2. Orbit state estimation device]
The configuration of a track state estimation device 100 that estimates the track state based on the above-mentioned track state estimation method will be described with reference to Fig. 7. Fig. 7 is a functional block diagram showing the configuration of the track state estimation device 100 according to this embodiment. As shown in Fig. 7, the track state estimation device 100 has an inclination angle estimation unit 110 and a track state calculation unit 120.

The inclination angle estimation unit 110 estimates the horizontal inclination angle (roll angle θ k ) and the longitudinal inclination angle (pitch angle φ _k ) of the wheel set 25 based on the three-axis accelerations a _xk , a _yk , a _zk and the three-axis angular velocities ω _xk , ω _yk , ω _zk measured by the inertial measurement unit 30 installed on the axle box ₂₇ of the vehicle 1.

First, the tilt angle estimator 110 extracts gravitational acceleration projection components and translational acceleration components from the three-axis acceleration measurement values a _xk , a _yk , and a _zk . The tilt angle estimator 110 also calculates indices R x and R _y representing translational acceleration from the translational acceleration components extracted from the three-axis acceleration measurement values a _xk , a _yk , and _{a zk} , and sets the variance-covariance matrix R(k) of the observation noise w _k at time k using the above formula (11). The tilt angle estimator 110 then calculates the observation vector y _k (the roll angle estimated value θ _0k and the pitch angle estimated value φ _0k ) using the set variance-covariance matrix R(k) in the above formula (7). Furthermore, the inclination angle estimation unit 110 calculates the roll angle θ k and the pitch angle φ _{k by a Kalman filter based on the state space model expressed by equations (5) to (7) and in which the variance-covariance matrix R(k) of the observation noise w k is set by equation (11). The inclination angle estimation unit 110 outputs the calculated roll angle θ k and pitch angle φ k of the} wheel set 25 to the track state calculation unit 120.

The track state calculation unit 120 calculates track state information representing the track state based on the horizontal inclination angle (roll angle θ _k ) and the longitudinal inclination angle (pitch angle φ _k ) of the wheel set 25. In this embodiment, a runway deviation is calculated as the track state information. When a runway deviation is obtained as the track state information, the track state calculation unit 120 calculates a runway deviation Tw(x) using the above formulas (1) and (2) based on the roll angle θ _k of the wheel set 25 calculated by the inclination angle estimation unit 110. The track state calculation unit 120 outputs the calculated runway deviation Tw(x) to, for example, the output device 200. The output device 200 may be, for example, a display device such as a display.

The track state estimation device 100 according to this embodiment is configured with various processors, such as a CPU (Central Processing Unit) and a DSP (Digital Signal Processor), and the above functions can be realized by operating the processors according to a predetermined program.

The track state estimation device 100 according to this embodiment may be installed on the vehicle 1 or may be installed at a separate location. The measurement values a _xk , a _yk , a _zk of the three-axis acceleration and the measurement values ω _xk , ω _yk , ω _zk of the three-axis angular velocity obtained by the inertial measurement unit 30 are recorded in a storage device (not shown) in association with the absolute position of the vehicle 1 on the rail 5 detected by the position detection device. The track state estimation device 100 can estimate the track state by acquiring the measurement values a _yk , a _zk of the three-axis acceleration, the measurement values ω _xk , ω _yk , ω _zk of the three-axis angular velocity, and the absolute position of the vehicle 1 recorded in the storage device.

[2-3. Orbit state estimation method]
A track state estimation method implemented by the track state estimation device 100 according to this embodiment will be described with reference to Fig. 8. Fig. 8 is a flowchart showing an example of the track state estimation method according to this embodiment.

(S10, S20: Acquisition of 3-axis acceleration and 3-axis angular velocity)
First, the three-axis accelerations a _xk , a _yk , a _zk and the three-axis angular velocities ω _xk , ω _yk , ω _zk are measured by the inertial measurement unit 30 installed on the axle box 27 of the vehicle 1 (S10, S20). The measured values of the three-axis accelerations and the three-axis angular velocities a _xk , a _yk , a _zk , ω _xk , ω _yk , ω _zk measured by the inertial measurement unit 30 are output to the inclination angle estimation unit 110 of the track state estimation device 100.

(S30: Extraction of gravitational acceleration projection component)
Next, the tilt angle estimation unit 110 extracts gravitational acceleration projection components from the three-axis acceleration measurement values a _xk , a _yk , and a _zk (S30). The gravitational acceleration projection components of the three-axis acceleration measurement values a _xk , a _yk , and a _zk are acceleration low-frequency components, for example, components of 0 to 1 Hz. The tilt angle estimation unit 110 can extract the gravitational acceleration projection components from the three-axis acceleration measurement values a _xk , a _yk , and a _zk by, for example, a low-pass filter.

(S40: Extraction of translational acceleration component)
The inclination angle estimator 110 also extracts translational acceleration components from the measured values a _xk , a _yk , and a _zk of the three-axis acceleration (S40). The translational acceleration components of the measured values a _xk , a _yk , and a _zk of the three-axis acceleration are high-frequency acceleration components, for example, components of 1 to 7 Hz. The inclination angle estimator 110 can extract the translational acceleration components from the measured values a _xk , a _yk , and a _zk of the three-axis acceleration, for example, by using a band-pass filter. As described above, it has been found that, with respect to planarity displacement, the translational acceleration components, which are acceleration components caused by impacts that predominate when the vehicle 1 is traveling around a curve, greatly affect the estimation accuracy of the estimated value θ _k of the roll angle and the estimated value φ _k of the pitch angle. Therefore, the inclination angle estimator 110 extracts the translational acceleration components and calculates the indices R _x and R _y that represent the translational acceleration.

(S50: Estimation of tilt angle)
Next, the inclination angle estimator 110 estimates the inclination angle θ (here, the roll angle θx) of the wheel set 25 (S50). The roll angle θx of the wheel set 25 can be calculated using a Kalman filter based on the state space model expressed by the above formulas (5) to (7). Here, the inclination angle estimator 110 sets the variance-covariance matrix R(k) of the observation noise w _k at time k from the above formula (11) using the indexes R _x and R _y calculated in step S40. Thereby, the observation noise w _k of the observation equation shown in the above formula (7) is sequentially set according to the magnitude of the translational acceleration component. As a result, the inertial measurement unit 30 is provided on the axle box 27, and the influence of errors due to the translational acceleration component during vehicle travel, which is included in the measurement value, is reduced, and the observation vector y _k (the estimated value θ _0k of the roll angle and the estimated value φ _0k of the pitch angle) can be obtained with high accuracy.

The inclination angle estimation unit 110 calculates the roll angle θ k and the pitch angle φ _{k by a Kalman filter based on a state space model represented by equations (5) to (7) and in which the variance-covariance matrix R(k) of the observation noise w k is set by equation (11). The inclination angle estimation unit 110 then outputs the calculated roll angle θ k and pitch angle φ k of the} wheel set 25 to the track state calculation unit 120.

(S60: Estimation of track state information)
Thereafter, the track condition calculation unit 120 calculates track condition information representing the track condition based on the horizontal tilt angle (roll angle θ _k ) and longitudinal tilt angle (pitch angle φ _k ) of the wheel set 25 (S60). In this embodiment, a runway deviation is calculated as the track condition information. When a runway deviation is obtained as the track condition information, the track condition calculation unit 120 calculates a runway deviation Tw(x) using the above formulas (1) and (2) based on the roll angle θ _k of the wheel set 25 calculated in step S50. The track condition calculation unit 120 outputs the calculated runway deviation Tw(x) to the output device 200.

3. Summary
The vehicle according to one embodiment of the present invention, and the track condition estimation device and the track condition estimation method for estimating the turntable deviation as the track condition have been described above. According to this embodiment, an inertial measurement unit is installed on the axle box of the vehicle, and the horizontal inclination angle and the longitudinal inclination angle of the wheel set are estimated based on the three-axis acceleration and the three-axis angular velocity measured by the inertial measurement unit. The horizontal inclination angle and the longitudinal inclination angle of the wheel set can be estimated by a Kalman filter. In this case, in order to reduce the influence of errors due to the translational acceleration component during vehicle running contained in the measurement value of the inertial measurement unit, the translational acceleration component that is an error factor is extracted from the measurement value of the three-axis acceleration during vehicle running by filtering processing, and the observation error covariance matrix of the Kalman filter is sequentially set based on the extracted translational acceleration component. This makes it possible to accurately estimate the horizontal inclination angle and the longitudinal inclination angle of the wheel set. By estimating the horizontal inclination angle and the longitudinal inclination angle of the wheel set with high accuracy, the estimation accuracy of the turntable deviation can be improved. In addition, the three-axis acceleration and the three-axis angular velocity can be measured by running the vehicle, so that the measurement can be easily performed.

To verify the effectiveness of the track state estimation method according to the above embodiment, the estimated values of track roughness using the track state estimation method were evaluated. The vehicle used for the verification was a vehicle with the configuration shown in FIG. 1, and an inertial measurement unit was installed on each of the axle boxes of the pair of wheels at the front of the vehicle in the direction of travel (X direction). As track conditions, the running section was set to a curved section (R = 100 m) with a total length of 200 m and a part of which had a track roughness of up to about 40 mm.

In this verification, the planarity measurement values measured by a hand-pushed track inspection device were taken as true values (True). As an example, the planarity measurement values were estimated using the track state estimation method according to the above embodiment. At this time, for the coefficients K _y and K _x in the above formula (11), values that would provide the highest accuracy in estimating planarity deviation were searched for. As a result, both the coefficients K _y and K _x were 1.0×10 ⁻⁶ , and the variance-covariance matrix R was set based on these coefficients K _y and K _x . The process noise Q was taken as given by the above formula (12).

As a comparative example, the observation noise _wk in the observation equation shown in the above formula (7) is set as a fixed value to calculate the observation vector _yk , and the flatness deviation is calculated. That is, in the comparative example, the flatness deviation is estimated in a state where the acceleration component (translational acceleration component) caused by the impact that is prominent when the vehicle is traveling on a curve remains as noise. In the comparative example, the variance-covariance matrix R and the process noise Q are as follows.

Figure 9 shows the planarity deviations estimated in the example and the comparative example. As shown in Figure 9, in the comparative example, the estimated planarity deviation deviated from the true value, but in the example, the estimated planarity deviation was almost the same as the true value. This shows that by applying the track state estimation method according to the above embodiment, the influence of errors due to translational acceleration, including the influence of shock response during vehicle travel, can be reduced, and the estimation accuracy of planarity deviation can be improved.

The above describes in detail preferred embodiments of the present invention with reference to the attached drawings, but the present invention is not limited to such examples. It is clear that a person with ordinary knowledge in the technical field to which the present invention pertains can conceive of various modified or revised examples within the scope of the technical ideas described in the claims, and it is understood that these also naturally fall within the technical scope of the present invention.

For example, in the above embodiment, the acceleration low-frequency components, which are the gravitational acceleration projection components of the three-axis acceleration measurement values a _xk , a _yk , and a _zk, are set to 0 to 1 Hz, and the acceleration high-frequency components, which are the translational acceleration components of the three-axis acceleration measurement values a _xk , a _yk , and a _zk, are set to 1 to 7 Hz, but the present invention is not limited to such an example. The frequency bands of the acceleration low-frequency components and the acceleration high-frequency components are set assuming a low-speed vehicle used, for example, to transport heavy materials in a steelworks. The frequency bands of the acceleration low-frequency components and the acceleration high-frequency components may be appropriately set according to the vehicle's running speed, for example, based on the above formulas (8) and (9). The moving average sample number s of the indexes R _x and R _y may also be set based on the above formula (10).

REFERENCE SIGNS LIST 1 vehicle 5 rail 10 vehicle body section 20 bogie section 21 bogie frame 23 wheel 25 wheel set 27 axle box 29 coil spring 30 inertial measurement device 100 track state estimation device 110 tilt angle estimation section 120 track state calculation section 200 output device

Claims (5)

an inclination angle estimating step of estimating a horizontal inclination angle and a longitudinal inclination angle of a wheel set based on a three-axis acceleration and a three-axis angular velocity measured by an inertial measurement device installed on an axle box of the vehicle;
a track condition calculation step of calculating track condition information representing a track condition based on the horizontal inclination angle and the longitudinal inclination angle of the wheel set;
Including,
In the tilt angle estimation step,
extracting a translational acceleration component during vehicle travel from the three-axis acceleration measured by the inertial measurement unit through a filtering process;
setting an error covariance matrix of an observation model of a Kalman filter based on the extracted translational acceleration component;
estimating a horizontal inclination angle and a longitudinal inclination angle of the wheel set from the three-axis acceleration and the three-axis angular velocity using a Kalman filter formulated from a state space model configured by an observation model in which the error covariance matrix is set;
A track condition estimating method , wherein in the track condition calculating step, a turn error is calculated as the track condition information based on a horizontal tilt angle of the wheelset . 2. The orbit state estimating method according to claim 1 , wherein the translational acceleration components are extracted by band-pass filtering to remove frequency components in which track irregularity predominates from the lateral acceleration and the longitudinal acceleration among the three-axis accelerations. 3. The orbit state estimation method according to claim 1 , wherein the error covariance matrix is set by adding values obtained by multiplying the extracted translational acceleration components by predetermined coefficients to diagonal elements. an inclination angle estimation unit that estimates a horizontal inclination angle and a longitudinal inclination angle of a wheel set based on a three-axis acceleration and a three-axis angular velocity measured by an inertial measurement device installed on an axle box of the vehicle;
a track condition calculation unit that calculates track condition information representing a track condition based on the horizontal inclination angle and the longitudinal inclination angle of the wheel set;
Equipped with
The tilt angle estimation unit is
extracting a translational acceleration component during vehicle travel from the three-axis acceleration measured by the inertial measurement unit through a filtering process;
setting an error covariance matrix of an observation model of a Kalman filter based on the extracted translational acceleration component;
estimating a horizontal inclination angle and a longitudinal inclination angle of the wheel set from the three-axis acceleration and the three-axis angular velocity using a Kalman filter formulated from a state space model configured by an observation model in which the error covariance matrix is set;
The track state estimation device, wherein the track state calculation unit calculates a turn error as the track state information based on a horizontal tilt angle of the wheelset . An inertial measurement unit installed on an axle box of the vehicle;
a tilt angle estimation unit that estimates a horizontal tilt angle and a longitudinal tilt angle of a wheel set based on the three-axis acceleration and the three-axis angular velocity measured by the inertial measurement unit; and
a track condition estimating device having a track condition calculating unit that calculates track condition information representing a track condition based on the horizontal inclination angle and the longitudinal inclination angle of the wheelset;
Equipped with
The orbit state estimation device comprises:
The tilt angle estimation unit
extracting a translational acceleration component during vehicle travel from the three-axis acceleration measured by the inertial measurement unit through a filtering process;
setting an error covariance matrix of an observation model of a Kalman filter based on the extracted translational acceleration component;
estimating a horizontal inclination angle and a longitudinal inclination angle of the wheel set from the three-axis acceleration and the three-axis angular velocity using a Kalman filter formulated from a state space model configured by an observation model in which the error covariance matrix is set;
a track condition calculation unit calculating a turnability deviation as the track condition information based on a horizontal inclination angle of the wheelset .

Images