JPH09189548A - Attitude angle sensor for train - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve an attitude angle sensor which is portable, is self-supporting, and can accurately measure an attitude angle over the movement range of a train. SOLUTION: A sensor has an inclination meter used as an acceleration sensor and an angular velocity sensor and adds a signal obtained by multiplying a previous system output by a negative sign to the output of the inclination meter and multiplies it by a feedback coefficient 1/T. Since the value is expressed by the navigation coordinate system, the value is converted to the value of the machine axis coordinate system, is added to the output of the angular velocity sensor and is subjected to integral processing, and then is outputted as an attitude angle. The feedback coefficient is changed so that the output of the inclination meter can be increased at a region where the frequency is low relatively for the sampling frequency and the output of the angular velocity can be increased at a frequency region which is higher than it, thus measuring attitude angle with less error in a wide frequency range.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an attitude angle sensor, and more particularly to a sensor for measuring an attitude angle of train movement.

[0002]

2. Description of the Related Art Attitude angle (roll angle, pitch angle, yaw angle)
A typical sensor for measuring is described. 1. Angular velocity sensor A single or multiple sensors that detect angular velocity are used to perform attitude angle calculation processing and output the relative change values of the roll angle, pitch angle, and yaw angle. 2. Inclinometer An accelerometer that detects gravity or one equivalent to it is used to measure the roll angle and pitch angle by obtaining the projected component of the gravity acceleration vector. 3. Vertical gyro This is to detect the posture of an object to which the vertical gyro is attached by taking out the signal from the outside by utilizing the property that the restoring force due to the rotation moment always points in the same direction. Since a drift error occurs when the free gyro state is continued for a long time, a level and an acceleration detection switch are provided to switch the operation. Often used for general dynamic posture measurement. The output is only the roll angle and the pitch angle, and the yaw angle is not output. 4. Inertial measurement device It is composed of a triaxial acceleration sensor, a triaxial angular velocity sensor, and a data processing unit, and can measure not only the roll angle, pitch angle, yaw angle, but also velocity and position. There are two methods, platform method and strapdown method,
The former creates a virtual platform by calculation, and the latter creates a mechanical platform in space, which requires a mechanical control mechanism. Regarding the former, the specific internal processing has not been clarified, but the basic processing procedure is considered to be represented by the block chart shown in FIG.

[0003]

The above-mentioned prior art has the following problems. 1. Angular velocity sensor Since only relative attitude changes can be determined, the initial reference attitude must either rely on other sensors or some other method. Moreover, since the posture is calculated after spatially integrating the angular velocity over time, the error diverges with respect to time. In addition, since one axis has a relation with other axes, the error occurrence in one axis affects all the other axes, which promotes error divergence with respect to time. If only a rate gyro is used to obtain sufficient accuracy, it will be extremely expensive. 2. Inclinometer It is possible to obtain the roll angle and pitch angle with respect to the horizontal plane.However, since the gravity acceleration vector is referenced as the posture reference, it is extremely significant if acceleration other than gravity acceleration, that is, acceleration / deceleration or centrifugal acceleration during turning, occurs. A big error occurs. The pendulum hanging in the train
It is stable vertically downward when stopped, but its behavior can be understood from the fact that it does not point vertically downward during acceleration / deceleration or turning. 3. Vertical gyro When a motion acceleration or a centrifugal acceleration is applied for a long time, the operation as a free gyro becomes long, and a drift error occurs remarkably. In the train movement,
This is a big problem because acceleration may be applied for a considerably long time such as generation of motion acceleration due to acceleration or deceleration or generation of centrifugal acceleration due to traveling on a curve. In addition, since it uses a precise mechanical mechanism, it requires regular maintenance, is vulnerable to vibrations and shocks, and requires a static settling time of about 1 hour to stabilize the rotation, making it highly portable. Is difficult. 4. Inertial measurement devices There are standardized devices for aircraft, but train movement and aircraft movement differ in all respects, and cannot be used as is. Moreover, the one optimized for trains has not been reported, and a general-purpose one will be used. In what is called a general-purpose inertial measurement device,
In many cases, input of speed is required, and independence cannot be maintained in some cases. From the above problems, an attitude angle sensor having the following characteristics is required. 1. Accurate measurement can be performed under the environment of train running, and no breakdown occurs. 2. It is possible to use sensors in which the cost balance of the constituent sensors is almost equal. 3. Does not require regular maintenance of mechanical mechanism. 4. The attitude angle accuracy is not limited by time or is sufficiently small. 5. Attitude angle accuracy does not significantly deteriorate in the range of train movement. 6. It can be started in a short time after turning on the power, and can reproduce sufficient accuracy. 7. It is portable and maintains independence without the need for information from trains.

[0004]

The angular velocity sensor and the inclinometer have mutually opposite properties in the frequency domain. By combining these two sensors in an optimal way, the disadvantages can be compensated and the advantages can be extended. The attitude angle sensor of the present invention uses an inclinometer using a single or a plurality of acceleration sensors and a single or a plurality of angular velocity sensors, and outputs the output signal in the measurement of the train inclination angle (roll angle and pitch angle). In the electrical processing, the inclinometer is in charge of a region having a relatively low frequency with respect to the sampling frequency in digital signal processing, and the angular velocity sensor is in charge of a frequency region higher than this, thereby eliminating an error. Here, the frequency that serves as the delimiter of the frequency range in charge of the inclinometer and the angular velocity sensor will be referred to as a crossover frequency or a region sharing boundary frequency for convenience. This enables measurement in a wide frequency range that could not be realized with an angular velocity sensor or inclinometer alone, does not require periodic maintenance like a vertical gyro, has few failures, maintains independence, and is highly accurate. An attitude angle sensor can be realized.

[0005]

First, the validity of the principle of the present invention will be proved. Assuming time domain data r (t) and time differential data r '(t) of this data, r (t) and r'
Consider two sensors that separately sense (t). Here, inclinometer and r'for sensing r (t)
An angular velocity sensor shall be used for sensing (t),
Assuming that the respective sensing data are x (t) and y (t), the errors ε _x (t) and ε _y (t) have the following relationship. x (t) = r (t) + ε _x (t)… (1) y (t) = r ′ (t) + ε _y (t)… (2) r ′ (t) = d (r (t)) / Dt (3) When the above equations (1) and (2) are Laplace transformed and considered in the frequency domain, (4) and (5) are obtained from their linearity. X (s) = R (s) + E _x (s) (4) Y (s) = s R (s) + E _y (s) (5) E _x (s) and E _y (s) are unknown However, by considering the frequency characteristics of the sensor, it is possible to know its properties in general.
You can find the error elimination method. Now, the frequency characteristic of the error is assumed as follows from the characteristic of the sensor. E _x (s) = k _x s (6) E _y (s) = k _y / s (7) This is because the error increases at X (s) and at Y (s) as the frequency increases. This shows that the error is reduced. When E _y (s) is inversely Laplace transformed, it becomes a step function, which corresponds to a bias in the time domain. Now, as a low pass filter, a first-order lag system, 1 /
(1 + T _s ), first-order lead system as a high-pass filter, T _s
Considering / (1 + T _s ), each of them is applied to both sides of (4) and (5), and when added to each side, the following expression is obtained as the output C (s).

[0006]

[Equation 1] Inverse Laplace transform of this gives the following equation.

[0007]

[Equation 2] (Δ (t): delta function u (t): step function) Although only the fourth term is a time transient term and describes the transient phenomenon of the filter, it converges to 0 after a sufficient time. An open loop block chart showing this processing is shown in FIG. However, in this method, when the disturbance E (s) occurs, the influence directly appears on the output. In an actual system, the error of the other axis corresponds to the disturbance given to the current axis. Then, the block chart as shown in FIG. 3 is obtained by converting into the closed loop processing. The crossover frequency f _c and the feedback time constant T of this system are expressed by the following equation with T _s as a sampling period. T = (1 + πf _c T _s ) / (2πf _c T _s ) ... (10) FIG. 4 shows a processing chart when this processing is applied to a three-dimensional posture angle calculation. Where vector θ: inclinometer output,
Vector φ: angular velocity sensor output, vector ψ: system output. The output of the inclinometer is summed with the previous system output multiplied by a negative sign and then multiplied by the feedback factor 1 / T. Further, since this angle change amount is described in the navigation coordinate system, it is converted into a value in the machine axis coordinate system.
At this time, the previous posture is referred to. The angle change amount coordinate-converted to the machine axis system is added to the angle change amount detected by the angular velocity sensor, and integration processing by a quaternion is performed to obtain a system output. The above assumes that no error is included in the low-frequency component of the inclinometer, but in running the train,
An error may occur in the inclinometer even in a very low frequency region due to motion acceleration generated during acceleration / deceleration, centrifugal acceleration generated during turning, and the like. To avoid this,
These motion states may be detected by some method, and the ratio of the inclinometer output and the angular velocity sensor output in the frequency domain may be changed continuously or discretely. Specifically, the feedback coefficient 1 / T shown in FIG. 4 may be reduced continuously or discretely to reduce the dependency of the inclinometer. In order to detect the motion acceleration due to acceleration / deceleration, it is sufficient to prepare two time-series inclinometer data before and after sampling and determine whether or not these two data are zero-crossed. This is because, as shown in FIG. 5, the acceleration / deceleration of the train is almost constant acceleration, and therefore a phenomenon in which an offset component larger than a vibrational component is generated in the output of the inclinometer during acceleration / deceleration is used. . To determine the binary zero crossings before and after, it suffices to check whether the sign of the product of the two values is positive or negative. When the sign of the product of the front and rear binary values is positive, it is determined that the vehicle is in the accelerated state, and when it is negative, it is determined that the vehicle is in the constant velocity state. To detect the centrifugal acceleration during turning, a judgment threshold value is set for the output value of the system roll angle, and when the absolute value becomes larger than a certain angle, it may be judged that the curve is passing. When a train passes through a curve at high speed, the track is provided with a cant angle in the roll direction. Since this inclination is designed to be almost proportional to the centrifugal acceleration, it has an extremely strong correlation between the roll angular centrifugal accelerations and can be used as a judgment condition.

A recurrence formula is required to convert such processing into software. When describing the recurrence formula,
The following data series and functions are defined. (Data series) θ _n : Inclinometer output angle φ _n : Angular velocity sensor output angular velocity ψ _n : System output attitude angle dt: Sampling time C _n : Coordinate conversion matrix α _n : Feedback coefficient (= 1 / T) β _n : Navigation System feedback amount γ _n : Machine system feedback amount (Note that θ _n , φ _n , ψ _n , α _n , β _n , and γ _n are vectors.) (Function) C = F (dμ): Machine system angle A function that updates the coordinate transformation matrix C with the increment dμ. ν = G (C): A function for obtaining the posture angle μ from the coordinate conversion matrix C. dμ = H (dλ, ψ): Aircraft system angle increment dλ and attitude angle ψ
A function to obtain the navigation system angular increment dμ from. (Note that μ, ν, ψ, and λ are vectors.) According to the block chart of the description of the analog transfer function shown in FIG. become that way. β _n = α _n (θ _n −ψ _{n -1} ) (11) γ _n = H (β _n , ψ _{n -1} ) (12) C _n = F (γ _n + φ _n dt) (13) ψ _n = G (C _n ) ... (14) (where β _n , α _n , θ _n , ψ _n-1 , γ _n , φ _n , ψ _n
Is a vector. ) In order to reduce the influence of acceleration / deceleration acceleration and centrifugal acceleration, the vector α _n may be changed in real time according to the respective judgment conditions.

[0009]

EXAMPLE FIG. 6 shows an example of the present invention. As the angular velocity sensor, an optical fiber gyro 1 having good starting characteristics and resistant to vibration is used. As the acceleration sensor, a servo type acceleration sensor 2 having a small crosstalk and a good resolution will be used. Further, as an anti-aliasing filter at the time of A / D conversion, a passive filter having a good phase characteristic (or an active filter having substantially the same property) is used. The personal computer 7 having the RS232C and the 16-bit A / D converter built therein is used for the process related to the data acquisition and the sensor fusion. First, the angular velocity information sensed by the optical fiber gyro 1 is passed through the RS232C serial interface 4, and the acceleration information sensed by the servo type acceleration sensor 2 is passed through the LPF (low pass filter) 3 to obtain a 16-bit A / A signal. It is taken into the arithmetic processing unit 6 of the personal computer 7 via the D converter 5. Here, processing is performed according to FIG. 4 and equations (11) to (14), and the posture angle is calculated every 30 ms. The crossover frequency f _c is the resolution of the angular velocity sensor of 0.000.
An optimum value of about 1 mHz is set for 5 degrees / second. In equation (10), the unset term crossover frequency f
_{Although c} exists, the optimum crossover frequency is expressed by the following equation, where the angular velocity sensor resolution is Ω _s and the system attitude angle required accuracy is θ _s . f _c = Ω _s / (2πθ _s ) ... (15) Usually, the resolution of the angular velocity sensor is often determined first, so the crossover frequency is determined using this formula as a guide. The crossover frequency may be too high with this formula depending on the motion state of the object, but this formula is used in the train application. Generally, a roll angle of more than 10 degrees and a pitch angle of more than 2 degrees do not occur as train motion conditions for safe train operation. According to the experiment. Assuming that the resolution of the angular velocity sensor is in the range of 0.01 to 0.001 degree / sec, the crossover frequency f _c is 5 m.
It has been found that it is preferable to set it between Hz and 0.5 mHz.

[0010]

Although the angular velocity sensor and the inclinometer themselves are stronger than the vertical gyro, their applications are limited in the train motion measurement due to the frequency dependence of the error. By the sensor fusion method according to the present invention,
It became possible to measure the train attitude angle in a wide frequency range that could not be realized by the angular velocity sensor or the inclinometer alone. In addition, it does not require regular maintenance like a vertical gyro and has few mechanical parts, so there are few failures. Furthermore, no information from the train is needed. It can be handled as an independent sensor and is highly portable. Furthermore, good accuracy can be obtained by performing signal processing optimized for train motion.

[Brief description of the drawings]

FIG. 1 is a block chart showing data processing by a conventional inertial measurement device.

FIG. 2 is a block chart showing data processing by an open loop type sensor in which an inclinometer and an angular velocity sensor are combined.

FIG. 3 is a block chart showing data processing by a closed loop type sensor in which an inclinometer and an angular velocity sensor are combined.

FIG. 4 is a block chart showing data processing by the attitude angle sensor of the present invention.

FIG. 5 is a diagram showing the output of the inclinometer during acceleration / deceleration in a horizontal state.

FIG. 6 is a block chart showing an embodiment of the present invention.

[Explanation of symbols]

1 Optical fiber gyro 5 A / D converter 2 Servo type acceleration sensor 6 Arithmetic processing unit 3 LPF (low pass filter) 7 Personal computer 4 Serial interface

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Yumitsu Yoshida Inoue Hiromichi, Chiyoda-ku, Tokyo 1-6-5 Marunouchi East Japan Railway Company (72) Inoue Miyazaki 5-1-1 Hidakacho, Hitachi City, Ibaraki Prefecture No. 1 Hitachi Cable Co., Ltd. Hidaka Factory (72) Inventor Futoshi Mugo 4-6 Kanda Surugadai, Chiyoda-ku, Tokyo Hitachi Systems Co., Ltd.

Claims (4)

[Claims] 1. An inclinometer using a single or a plurality of acceleration sensors, and a single or a plurality of angular velocity sensors, wherein output signals of the inclinometer and the angular velocity sensor are electrically processed to obtain a posture angle (roll angle, In a train attitude angle sensor that measures pitch angle and yaw angle, the train inclination angle (roll angle and pitch angle) during running is relatively low compared to the system sampling frequency in digital signal processing. A train attitude angle sensor characterized in that an inclinometer is in charge of a region, and an angular velocity sensor is in charge of a higher frequency region. 2. The train attitude angle sensor according to claim 1, which detects a direct current acceleration generated in the output of the inclinometer at the time of train acceleration or deceleration, and uses the inclinometer in the measurement of the pitch angle according to this value. An attitude angle sensor for a train, characterized in that the frequency domain sharing ratio of the angular velocity sensor is changed continuously or discretely. 3. The train attitude angle sensor according to claim 1, wherein the sensor itself detects a roll angle generated when passing through a train curve, and the roll angle is measured according to this value.
A train attitude angle sensor characterized by continuously or discretely changing the frequency domain sharing ratio of the inclinometer and the angular velocity sensor. 4. The train attitude angle sensor according to claim 1, wherein the signal filtering process is performed by a feedback digital filter.