JPH0894357A - Method for automatically measuring position of marine vessel on-line - Google Patents

Abstract

PURPOSE: To stably measure the position of a marine vessel with high accuracy by expressing heaving, rolling, and pitching in a dynamics model expressing state variables as a linear system and combining the observation equations obtained by means of servo type inclined-angle sensors and a servo type acceleration sensor. CONSTITUTION: Three sensors of two servo type inclined-angle sensors S1 and S2 one servo type accelerometer S3 are positioned to point J on the deck 3 of a marine vessel immediately above the intersection 0 of the rolling axis X-X and pitching axis Y-Y of the vessel so that the measuring axis and sensitive axes of the sensors can intersect each other at right angles. Even when the sensors S1 , S2 , and S3 are collectively positioned to the T-point, each linear observation equation holds when the centrifugal force is contained in the term corresponding to noise. The dynamics models of heaving, rolling and pitching are prepared by introducing state variables composed of resultant waves of short-period waves and long-period waves to the equations and the outputs of the models are passed through a Kalman filter. Therefore, the attitude of the marine vessel can be detected with high accuracy regardless of the size of the vessel.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a ship attitude online automatic measuring method capable of accurately measuring the heaving, rolling and pitching of a ship online.

[0002]

2. Description of the Related Art Conventionally, in a seabed development or port construction, etc.
The shape of the sea floor is searched by collecting the reflected signal from the sea floor and analyzing the signal.At that time, the long period waves mainly generated offshore and the short period waves generated near the coast are detected. Due to the synthetic waves, the ship sways in the up / down, front / rear, and left / right directions, that is, heaving, rolling, and pitching.
Therefore, it is indispensable to accurately measure and correct these three factors in order to accurately search the sea floor.

From the above point of view, the present inventors previously installed four servo-type acceleration sensors appropriately on the ship and processed the outputs appearing on them in consideration of the wave characteristics,
We have already developed a ship attitude online automatic measurement method that consists of measuring heaving, rolling and pitching.

However, since this measurement method employs a method of calculating an approximate frequency of a measurement amount by using the maximum entropy method and then using these values as initial values to search an unknown parameter by an optimization method. , It takes a relatively long time to process information, and 1
Since it is performed independently for each data segment of 0 seconds, there is a problem that a waveform mismatch occurs at the connection point of the segments.

Furthermore, in order to solve these problems, the inventors continued to propose a two-sensor type automatic measuring method using a Kalman filter, but the measurement object is represented as a linear system by a dynamic model of state variable display. By combining the observation equation with the servo type acceleration sensor and applying the Kalman filter, it became possible to measure the vessel attitude with high accuracy online.

However, since a relatively large vessel was assumed to be applied to this method, the attitude was measured so that the effect of pitching could be ignored. In reality, however, the online attitude measurement for a medium-sized ship or a small ship was performed. That is, the pitching that occurs in these vessels can be accurately measured at the same time,
There is a demand for a measurement method that can correct the effect.

[0007]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention The on-line measuring method of the ship attitude proposed by the inventors is
In other words, there is no on-line attitude measurement of the ship attitude that is satisfactory for ships that cannot neglect the effects of pitching.

[0008]

According to the present invention, heaving rolling and pitching occurring in a ship are represented as a linear system by a dynamics model of state variable display.
And the dynamics model thereof, two servo-type tilt angle sensors, which are installed on the ship and obtain rolling information and pitching information, and heaving information 1
Combining observation equations by servo-type acceleration sensors,
An online ship attitude measurement method characterized by applying a Kalman filter.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described with reference to FIGS. 1 is the bow and 2 is the stern.

First, the arrangement of the tilt angle sensor and the acceleration sensor will be described.

FIG. 1 shows the arrangement of the servo type inclination angle sensors S ₁ and S ₂ and the servo type acceleration sensor S ₃ .

Here, in all three sensors, at a point J on the deck 3 just above the intersection 0 of the rolling axis XX and the pitching axis Y-Y, the measuring axis and the sensitive axis direction of each sensor are・ ・
Arrow S ₁ in FIG. 1, the S ₂ arrows, S ₃ arrows ...
They are installed so that they are perpendicular to each other.

The tilt angle sensor S ₁ supplies rolling information to S ₂
In order to obtain the pitching information, is mounted on the deck so that the measuring axis directions of the sensors are perpendicular to the rolling axis and the pitching axis, respectively.

Further, the acceleration sensor S ₃ is installed so that the sensing axis ZZ of the sensor is vertically upward with respect to the deck in order to obtain the heaving information.

In the present invention, the tilt angle sensors S ₁ and S ₂ and the acceleration sensor S ₃ are collectively arranged at the point J from a practical standpoint, but as will be described later, the tilt angle sensors S ₁ and S ₂ are respectively arranged. Even if it is shifted (translated) to an arbitrary point just above the rolling axis and just above the pitching axis, the same output as that at the point J is obtained.

Next, the output of the tilt angle sensor will be described. The tilt angle sensor originally measures a static tilt angle, and torque is applied by the servo mechanism so that the pendulum 4 always faces the main axis direction of the instrument. The tilt angle γ depends on the magnitude of the torque at this time. Is instructed. 5 is an instrument box.

A principle view of the tilt angle sensor is shown in FIG. When the inclined surface of the deck 3 is in a stationary state, the torque T _c applied to the pendulum of the tilt angle sensor by the servo mechanism is equal to the torque generated in the pendulum by gravity, the mass of the pendulum is M, the gravitational acceleration is g, and the pendulum's if the length from the fulcrum to the center of gravity and l _c

[0018]

[Equation 1]

(Since γ is small, sin γ≈γ)

[0020]

[Equation 2]

The tilt angle γ can be measured by means of the tilt angle sensor output z _c (t). It should be noted that the symbols in which "to" is drawn on "≈" and "-" represent approximately the same. Q-Q represents a horizontal plane.

In practice, the inclination angle γ is _calculated from the value of the current flowing through the servo mechanism, which is proportional to the torque T _c of the servo mechanism.
Is instructed.

However, when the inclination angle sensor is installed on the vessel, the pendulum 4 of the inclination angle sensors S ₁ and S ₂ is applied with accelerations of heaving, rolling and pitching (other than gravity acceleration) due to the movement of the vessel. Does not indicate the correct tilt angle and becomes a value that reflects their influence.

Hereinafter, the torque and output of the tilt angle sensors S ₁ and S ₂ arranged as shown in FIG. 1 will be obtained. As shown in FIG. 3, when the ship is at rest, the zero point is the origin, the rolling axis is the X axis, the pitching axis is the Y axis, and the vertically upward direction is the Z axis, the inclination installed to detect the rolling signal. The coordinates X _G , Y _G of the center of gravity G of the rigid pendulum in the angle sensor S ₁ ,
Z _G is

[0025]

[Equation 3]

[0026]

[Equation 4]

[0027]

[Equation 5]

[0028] However, θ, p, h, and φ are the rolling angle, the pitching angle, the heaving displacement, and the rotation angle of the pendulum from the instrument main axis direction, respectively, which are functions of time. Further, L represents the distance from the intersection 0 of the rolling axis and the pitching axis to the fulcrum of the pendulum, and l _c represents the distance from the fulcrum of the pendulum to the center of gravity G. Also, "1 black circle" attached to the head of the character or "'" attached to the right shoulder of the character both represent the first differentiation, and "2 black circles" attached to the head of the character or the right shoulder of the character “″” Represents the second differentiation.

The above equation, the kinetic energy K ₁ rigid pendulum inclination angle sensor S ₁ is

[0030]

[Equation 6]

It becomes However, I is the moment of inertia around the center of gravity of the pendulum, and M is the mass of the pendulum.

The potential energy U ₁ is

[0033]

[Equation 7]

It becomes

Therefore, these are calculated by the following Lagrange equation of motion.

[0036]

[Equation 8]

Substituting into, and φ, p, θ are small angles, cosφ
≈1, sinφ≈φ, cosp≈1, sinp≈p, cosθ≈1,
Note that the torque T ₁ (due to the servo mechanism) applied to the pendulum can be approximated as sin θ ≈ θ

[0038]

[Equation 9]

Is represented as

However, ζ ₁ is φ≈0 from the characteristics of the tilt angle sensor.
So, an approximation error corresponding section when this section is zero, also the _{A = l c + I / (} Ml c).

In reality, the heaving acceleration h ″ is about 5% of the gravitational acceleration g, the pitching angular velocity squared value p ² ′ is minute, and L (>> l _c ) is at most several meters. Considering the above, it can be approximated as g + h ″ − (L−1 _c ) p ² ′ ≈g. Further, since _{A = l c + I / (} Ml c) is sufficiently smaller than L, and eventually the torque T ₁ is

[0042]

[Equation 10]

Can be written as

Here, η ₁ is a collective expression of the approximation error terms. Output z ₁ (t) appearing on the tilt angle sensor S ₁
Since, in which the torque T ₁ as described above divided by Mgl _c,

[0045]

[Equation 11]

However, υ ₁ (t) is υ ₁ (t) = η ₁ / (Mgl
_c ), which is a noise equivalent term.

As described above, the output values of the tilt angle sensor S ₁ are the rolling angle θ (t) and the rolling angular acceleration θ ″ (t).
Is a linear observation equation.

Similarly, the output of the tilt angle sensor S ₂ installed for detecting the pitching information is

[0049]

[Equation 12]

A similar linear observation equation is obtained for the pitching angle p (t) and the pitching angular acceleration p ″ (t), where υ ₂ (t) is the tilt angle sensor S ₂
Is a noise equivalent term for.

Further, the tilt angle sensors S ₁ and S ₂ are not set at the J point,
Shifting to any point directly above the rolling axis and just above the pitching axis will only have a slight effect of centrifugal force on each sensor output.

Therefore, if the term relating to this centrifugal force is included in the term corresponding to noise, the output of S ₁ and S ₂ can be expressed by Equation 11,
The same linear observation equation as Equation 12 holds.

The output of the acceleration sensor will be described. The output z ₃ (t) that appears in the acceleration sensor S ₃ is a combination of the effects of heaving, rolling, and pitching, as described above, and α (t) is the heaving acceleration.

[0054]

[Equation 13]

It is expressed as follows.

However, υ ₃ (t) is a noise peculiar to the acceleration sensor.

As described above, the output of the acceleration sensor is affected by the rolling angle θ (t) and the pitching angle p (t) in addition to the heaving acceleration α (t), but the rolling angle and the pitching angle are separately described. Since online measurement is performed, by substituting those measured values, that is, estimated value θ (t) and estimated value p (t)

[0058]

[Equation 14]

A linear observation equation for the heaving acceleration α (t) is obtained. The chevron symbol (French Axan Silconflex) above the letters represents the estimated value.

When the acceleration sensor S ₃ is installed at a point other than point J, it is necessary to add the effects of rolling acceleration and pitching acceleration to the sensor output equation (13).

However, also in this case, since the rolling acceleration and the pitching acceleration are measured online at the same time as the rolling angle and the pitching angle, respectively, those measured values, that is, the estimated value θ ″ (t) and the estimated value p ″ (t) are rolled. By substituting the measured values of the angle and the pitching angle, that is, the estimated value θ (t) and the estimated value p (t) into the output formula of the acceleration sensor, a linear observation equation regarding the heaving acceleration α (t) is obtained.

Therefore, even if the sensor cannot be attached to the point J due to the structure of the ship, the state of the cargo, etc., the inclination angle sensors S ₁ and S ₂ are placed at the points directly above the rolling axis and the pitching axis, respectively. If attached (not necessarily on the deck), online measurement is possible even if the acceleration sensor S ₃ is installed at an arbitrary point.

However, if the sensors are collectively installed at point J,
This arrangement is adopted in the present invention because the hardware part of the system is compact and easy to handle.

Next, an outline of automatic measurement of the ship attitude will be described.

First, the dynamics expression of the ship attitude signal will be described.

Heaving, rolling, and pitching that occur in a ship are almost locally represented by a composite wave of two sine waves, a short period wave having a period of about 3 seconds and a long period wave having a period of about 6 to 10 seconds. It is known that So, for example, for heaving

[0067]

(Equation 15)

It can be modeled as follows. Here, _one of ω ₁ and ω ₂ corresponds to a long cycle and the other corresponds to a short cycle, but the amplitudes a ₁ and a _{2 of} the sine wave and the initial phases φ ₁ and φ ₂ are unknown parameters. χ (t) is the differential equation

[0069]

[Equation 16]

Since the following condition is satisfied, the state variable χ in the phase variable display
Introducing _n = d ^n-1 χ / dt ^n-1 (n = 1, ..., 4) gives the following linear dynamics equation for heaving:

[0071]

[Equation 17]

However, χ ≈ (χ ₁ , χ ₂ , χ ₃ , χ ₄ ) ^T

[0073]

(Equation 18)

On the other hand, since a bias generally occurs in rolling and pitching, the following model in which a bias is added to a short period wave and a long period wave can be considered.

[0075]

[Formula 19]

Of course, {a _i } {ω _i } {φ _i }, b are different for heaving, rolling and pitching. Number 1
9 is a differential equation

[0077]

[Equation 20]

Since the above condition is satisfied, if the state variables χ _n = d ^n-1 χ / dt ^n-1 (n = 1, .., 5) are introduced as in the case of heaving, rolling and pitching are followed. The dynamics equation is given by the equation (17).

However, at this time, A

[0080]

[Equation 21]

As described above, the heaving, rolling and pitching are modeled as a linear dynamics system, but they are not always expressed by a linear combination of two sinusoidal waves even though they are local, and they are time-dependent. Is 3-5
It may be represented by a first-order combination of a number of sine waves or more.

However, the dominant ones are the above-mentioned two short-cycle and long-cycle composite waves.

Therefore, in modeling, it is necessary to consider the non-dominant component. In the present invention, white noise with an average value of zero and an appropriate variance value σ ² is added to the dynamics of the equation (17) to obtain each signal. To give flexibility over time. That is, the following dynamics were adopted.

[0084]

[Equation 22]

[0085] However B is a constant matrix, when the heaving B = (0,0,1,0) ^T, if the rolling and pitching was B = (0,0,1,0,0) ^T. This is because the displacement is considered to change smoothly.

In practice, the contribution of non-dominant periodic waves is small in ships, and the above three signals can be locally sufficiently approximated by a composite wave of two long and short periodic waves.

Therefore, in the present invention, from the standpoint of on-line measurement, a method is adopted in which a non-dominant periodic wave is added to increase the model order, thereby adding flexibility to the signal waveform by adding white noise.

An automatic measuring method of rolling and pitching will be described. First, regarding the observation equations of rolling and pitching that occur in a ship, considering that the observation data is obtained at each sampling time in the equations 11 and 12 in the previous chapter, the above-mentioned state variable display is used.

[0089]

[Equation 23]

It is represented by

However, у _K = у (kΔT), χ _K = χ (kΔ
T) and υ _K means observation noise.

That is, in the case of rolling, υ _K = υ ₁ (k
ΔT), and in the case of pitching, ν _K = υ ₂ (kΔT). The observation matrix H is H = [1,0, -L / g, 0,0] for both rolling and pitching.

Similarly, if the dynamics are discretized,

[0094]

[Equation 24]

Is obtained. here,

[0096]

(Equation 25)

[0097]

(Equation 26)

Where L ⁻¹ represents the inverse Laplace transform.

The transition noise W _K has a mean value of zero and covariance.

[0100]

[Equation 27]

White Gaussian noise of

Thus, the measurement of rolling and pitching can be reduced to the state estimation problem of the linear discrete dynamic system given by the equations (23) and (24), and if the white Gaussian property of the observation noise ν _K is assumed, The following Kalman filter can be applied to the estimation.

[0103]

[Equation 28]

[0104]

[Equation 29]

[0105]

[Equation 30]

[0106]

[Equation 31]

[0107]

[Equation 32]

However,

[0109]

[Expression 33]

[0110]

[Equation 34]

Here, V is the variance of the observation noise. In addition to the noise peculiar to the sensor, the observation noise υ _K includes the influence of the engine vibration and the approximation error when deriving the equation (23).
It is necessary to use a value larger than the noise peculiar to the sensor.

Considering the actual state of the ocean, the periods of the long and short period wave components which are dominant in rolling and pitching are 6 to 10 seconds and 2 to 3 seconds, respectively, but these can also change with time. It is a thing. Therefore, in the present invention, within the range of these values, the angular frequencies ω ₁ , ω
_An appropriate number of ₂ candidates {(ω ⁱ ₁ , ω ⁱ ₂ ); (1 ≦ i ≦ M)} are provided, and the Kalman filter is applied to each candidate independently.

The expected value for the estimated value obtained by each candidate

[0114]

[Equation 35]

Is calculated and this is adopted as the final state estimated value at each time.

The estimated value χ ⁱ _{K / K} is the i-th candidate Ω _i =
State estimate χ calculated for (ω ⁱ ₁ , ω ⁱ ₂ )
_{K / K} and p ⁱ _K are posterior probabilities obtained by Bayes' theorem

[0117]

[Equation 36]

Is represented.

However, p (у _K / Ω _i , Y ^k-1 ) is the i-th candidate Ω _i = (ω ⁱ ₁ , ω ⁱ ₂ ) and Y ^K-1 ≡ {у _j ; j≤k-1. }
^Represents the conditional probability density function of y ^k for

[0120]

(37)

Here, the estimated values χ ⁱ _{k / k-1} and Λ ⁱ _k are i
The predicted values and variances of the equations (28) and (34) calculated for the th candidate are shown.

In the calculation of the posterior probability p ⁱ _k , the a priori probability may be, for example, p ⁱ ₋₁ = 1 / M (1 ≦ i ≦ M).

Further, in the measurement, in order to enhance the response to the fluctuation of the dominant periodic wave component, the lower limit ε of the posterior probability value is set to be sufficiently small.

That is, the posterior probability that p ^j _k <ε (jEΩ ₃ ) is forcibly set to p ^j _k = ε, and the other posterior probabilities are decreased by this probability increase according to the probability ratio. . In this way, even when the dominant long and short periodic waves fluctuate with time, this measurement system automatically determines the optimum one among the given candidates according to the changed dominant wave period. It has been selected, and online can be performed with high accuracy even for complicated attitude signal waveforms that combine fundamental waves.

An automatic heaving measurement method will be described.

The observation equation of heaving is given by the equation 14 as described in the previous section. Therefore,

[0127]

(38)

As a result, the observation equation of the same form as the equation 23 can be obtained for heaving.

However, H _k = [0,0, COS estimated value θ (t) × cos
Estimate p (t), 0] | t = k Δ T, it is _{υ K = υ 3 (kΔT)} . Since the dynamics of heaving is represented by the formula (22), heaving can be automatically measured by using exactly the same method as in rolling or pitching.

In the measuring system of the present invention, heaving,
The first, second and third components of the state estimation value χ ^o _{k / k} for rolling and pitching respectively correspond to the displacement, velocity and acceleration of each signal. Therefore, this measuring system has an advantage that not only the vessel attitude but also the velocity and acceleration of the attitude signal can be measured at the same time.

The operation will be described. First, as a simulation, the following heaving, rolling, and pitching displacement were considered.

[0132]

[Formula 39]

[0133]

[Formula 40]

[0134]

[Formula 41]

Here, {a _hi } ¹⁰ _{i = 1} = {0.15, 0.037
5, 0.0375, 0.0225, 0.0225, 0.045, 0.01125, 0.0112
5, 0.00675, 0.00675}, {ω _hi } ¹⁰ _{i = 1} = {0.8, 0.9,
0.7, 1.0, 0.6, 2.5, 2.25, 2.75, 2.0, 3.0},
{Φ _hi } ¹⁰ _{i = 1} = {5π / 18, π / 9, π / 2, π / 3, 4π / 9,
5π / 9, π / 6, 2π / 9, 7π / 18, 11π / 18}, {aθ _i } ¹⁰
_{i = 1} = {0.45, 0.1125, 0.1125, 0.0675, 0.0675, 1.5,
0.375, 0.375, 0.225, 0.225}, {ωθ _i } ¹⁰ _{i = 1} = {0.
8, 0.9, 0.7, 1.0, 0.6, 2.5, 2.75, 2.25, 3.0, 2.
0}, {φθ _i } = {π / 2, 7π / 18, 11π / 18, 2π / 9, 5
π / 9, π / 3, 5π / 18, π / 6, 4π / 9, π / 9}, {a _pi }
= {0.5, 0.125, 0.125, 0.0625, 0.0625, 2.0, 0.4,
0.4, 0.24, 0.24}, {ω _pi } = {0.8, 0.9, 0.7, 1.
0, 0.6, 2.25, 2.0, 2.5, 2.75, 1.75}, {φ _pi } ¹⁰
_{i = 1} = {4π / 9, π / 3, 5π / 9, 2π / 9, 2π / 3, π / 2, 11
π / 18, 7π / 18, 5π / 18, 13π / 18}. The rolling and pitching biases are bθ = 0.8u.
(T) −0.4u (t−30) (deg), b _p = −2.5u (t) −
A time-varying one such as 0.5u (t-30) (deg) was considered.

However, u (t) is a unit step function.

In any of the heaving, rolling, and pitching, the first five terms represent long-period wave components, and the remaining five terms represent short-period wave components. The dominant periodic waves are the first and sixth terms. This is the ingredient.

Regarding the amplitude of each component, the amplitude of the long-period wave component is set large in heaving and the amplitude of the short-period wave component is set large in rolling and pitching in consideration of the actual swaying condition of the ship.

When the output of the tilt angle sensor is created using these data, the values of l _c and A in the equation 9 related to the structure of the pendulum are l _c = 0.03 (m) and A = 0.04, respectively.
I said (m).

On the other hand, in the automatic posture measurement, the initial estimated value χ ⁱ _{o / -1} (1≤i≤M) used for the Kalman filter was set to 0 for heaving, rolling and pitching. Further, the a priori error covariance p ⁱ _{o / -1} is calculated for all i (1 ≦ i ≦
For M), p ⁱ _{o / -1} = 15I ₄ for heaving, and p ⁱ _{o / -1} = 60I ₅ for rolling and pitching (I _n represents an n-dimensional identity matrix).

This takes into consideration that the dimension of the rolling and pitching state variable vector is one larger than that of heaving, and that the rolling and pitching values are larger than those of heaving in the above unit system. Because I did.

Further, the dominant periodic wave candidates Ω _i = (ω ⁱ ₁ ,
In setting ω ⁱ ₂ ) (1 ≦ i ≦ M), the one that appears more predominantly among the long and short period waves was emphasized.

That is, regarding the period lengths of the long-period wave and the short-period wave, in heaving, nine periods were taken at regular intervals of 6 to 10 seconds, three periods at 2.0 seconds, 2.3 seconds, and 2.5 seconds, and rolling was performed. 6.5 seconds, 8.5 seconds, 10.5 with pitching, respectively
A total of 27 candidates were considered, 3 for each second and 9 for equally spaced 2-3 seconds. The lower limit ε of the posterior probability of these candidates is set to 10 ^-5 , the average value of noise added to the tilt angle sensor and the acceleration sensor is both zero, and the standard deviation is 1.0, respectively.
× 10 ^-4 (calculated from resolution 10 "etc.) and 0.02 (rated value 2g of 0.
1%).

Regarding the arrangement of the sensors, assuming that the simulation target is a medium or small vessel, L = 0.5 (m)
And On the other hand, regarding the variance value σ ² of white noise in dynamics, considering the magnitude of the non-dominant component wave, 1.0 × 10 ^-2 , 5.0 × 10 ^-3 , 1.0 × for heaving, rolling, and pitching, respectively. A value of 10 ^-3 was adopted.

Under the above conditions, the sampling time ΔT is set to 0.
Simulation was performed for 60 seconds for 2 _s .

FIG. 4 shows the output value of the tilt angle sensor S _{1 and} the output value based on the measured value of the rolling angle. Further, FIG. 5 shows the output value of the acceleration sensor S _{3 and} the output value based on the measured values of the heaving acceleration, the rolling angle and the pitching angle.

In each of FIGS. 4 and 5, the output waveform based on the measured value almost completely matches the true output waveform.

Exactly the same results are obtained for the other sensors S ₂ .

6 to 8 show the measured waveforms of the heaving displacement, rolling angle and pitching angle together with the simulation waveforms at this time.

Although the result of 20 to 40 seconds is shown in the figure,
Similar results were obtained in other sections. Further, FIGS. 6 and 7 also show measurement results of heaving and rolling by the two-sensor method previously proposed by the inventors.

As is clear from FIGS. 6 to 8, in the present measurement system, the measurement waveforms of heaving, rolling, and pitching are in good agreement with the simulation waveforms, and it is understood that highly accurate measurement is performed.

Further, in the previously proposed two-sensor method, the measurement of only heaving and rolling was considered on the assumption that it is mounted on a large ship, but the new measuring system proposed by the present invention uses two sensors for heaving and rolling. It can be seen that not only the measurement with almost the same accuracy as the method is performed (see FIGS. 6 and 7), but also the pitching can be performed with high accuracy (FIG. 8).

Measurement error rate of the heaving displacement and rolling angle by the two-sensor method when the noise series was changed 50 times and the simulation was performed ([effective value of measurement error / effective value of measured quantity] × 100 (%)) Average of 18.0%,
8.7%, but with this measurement system, heaving, rolling, and pitching are 17.4%, 10.5%, and
A result of 4.6% was obtained.

In this measurement system, compared with the two-sensor system,
Highly accurate measurements have been made with almost the same accuracy for heaving and slightly inferior for rolling.

The pitching measurement error rate is lower than that of the rolling angle because the absolute value of the pitching bias is larger than that of the rolling angle.

Furthermore, in this measurement system, the tilt angle sensor S ₁ ,
When the average value of the measurement error rates is calculated in the same manner as above, with the distance L between S ₂ and the 0 point set to 0.2 (m), L =
Similar to the case of 0.5 (m), it is 7.7% and 2.6% for rolling and pitching, respectively, and it can be seen that this measurement system is more effective when the tilt angle sensor is brought close to 0 point.

On the other hand, since the two-sensor method uses the acceleration sensor, the measurement accuracy cannot be improved even if the sensor is brought close to the zero point.

Further, in the two-sensor system, the pitching is negligible for the measurement. Therefore, if the vessel is pitched, the sensor output will be affected by the pitching.

The measurement result of the two-sensor method shown in the figure shows that the acceleration sensor is placed directly above the pitching axis (and above the rolling axis so that the pitching does not affect the sensor output).
The measurement accuracy is considerably deteriorated due to the influence of pitching at other places, which are arranged radially and symmetrically at an angle of 90 (deg).

Therefore, the measuring system of the present invention is not limited to medium-sized, small-sized and large-sized vessels, and when the tilt angle sensor is attached at a position near the intersection 0 of the rolling axis and the pitching axis, the rolling system is more effective than the two-sensor method. Highly accurate measurement can be performed.

Next, an example in which this measurement method is applied to experimental data collected in the Seto Inland Sea will be shown. A small boat was used for the measurement, and data was collected under the condition that the distance L from the intersection 0 of the rolling axis and the pitching axis to the three sensors was L = 0.66 (m). Dominant periodic wave candidates for rolling and pitching in automatic posture measurement Ω _i = (ω ⁱ ₁ , ω ⁱ ₂ )
(1 ≤ i ≤ M) used the same value as in the simulation, but regarding heaving, considering that the influence of short-period waves on fairly small vessels appears considerably, the period lengths of long-period waves and short-period waves were respectively We considered a total of 28 candidates, 7 equally spaced in 6-10 seconds and 4 equally spaced in 2-3 seconds.

The standard deviation of noise applied to the tilt angle sensor and the acceleration sensor was set to 5.0 × 10 ⁻⁴ and 0.2, respectively, in consideration of the sensor specifications and engine vibration.

A priori error covariance p ⁱ _{o / -1} (1 ≦ i ≦ M),
The same value as the simulation was used for the variance value σ ² of white noise in dynamics.

FIG. 9 shows the actual output and the calculated output based on the measured rolling angle for the tilt angle sensor S ₁ . Further, FIG. 10 shows the actual output of the acceleration sensor S ₃ and the calculated output based on the measured values of the heaving acceleration, rolling angle and pitching angle. Since the actual sensor output contains high-frequency components due to the effects of engine vibration, etc., there are some differences from the sensor output based on the measured values, but the sensor output based on the measured values still represents the actual sensor output almost accurately. You can see that

However, in the acceleration sensor output shown in FIG. 10, there is a portion that cannot be expressed by the calculation output waveform based on the measured value. However, this is because the acceleration sensor S ₃ is installed vertically upward from the deck, so engine vibration It is thought to be the result of being strongly influenced by the above.

11 to 13 show the measurement results of the heaving displacement, rolling angle and pitching angle by this measuring system.

In the measured values of the heaving displacement shown in FIG. 11, the influence of short-period waves with a period of about 2 seconds appears strongly, but in the section of 20 to 30 seconds, the period of 9 to 10 seconds, and 30 to 40 seconds. It is observed that long-period waves with a period of 8 to 9 seconds also appear in the second section.

Also, from FIGS. 12 and 13, the measured values of the rolling angle and the pitching angle mainly show short-period waves with a period of about 2 seconds, but the long-period wave component is not remarkable, and the simulation results are good. You can see that it corresponds.

From FIGS. 11 to 13, the amplitude of the heaving displacement is about 10 (cm), the bias of the rolling angle and its time-varying amplitude are about 2.5 (deg) and 3.0 (deg), and the bias of the pitching angle and that The amplitude of the time variation is 0.8 (de
g) and 2.0 (deg), both of which are reasonable in light of the scale of the vessel used for measurement and the conditions of the Seto Inland Sea.

[0170]

Since the present invention is constructed as described above,
Regardless of the scale of the ship, the ship attitude can be measured online with stability and high accuracy, and it is superior to the conventional ones in terms of wide application range, measurement accuracy, online measurement, and the like. For information processing, EWS (Toshiba AS4080)
However, by using a DSP board without using EWS or by slightly increasing the sampling time, an inexpensive personal computer can sufficiently support online measurement.

[Brief description of drawings]

FIG. 1 is a perspective view showing an arrangement of a servo type acceleration sensor and a tilt angle sensor.

FIG. 2 is an operation principle diagram of a tilt angle sensor installed on a ship.

FIG. 3 is a motion diagram of a pendulum of a tilt angle sensor S ₁ on a deck.

FIG. 4 is a correlation diagram showing an output value of an inclination angle sensor and an output value based on a measured value.

FIG. 5 is a correlation diagram showing the output value of the acceleration sensor and the output value based on the measured value.

FIG. 6 is a correlation diagram showing a comparison between a true value of a heaving displacement and a measured value by the proposed method.

FIG. 7 is a correlation diagram showing the comparison between the true value of the rolling angle and the measured value by the proposed method.

FIG. 8 is a correlation diagram showing a comparison between a true value of a pitching angle and a measured value by the proposed method.

FIG. 9 is a correlation diagram showing a comparison between the actual output value of the tilt angle sensor and the output value based on the measured value.

FIG. 10 is a correlation diagram showing a comparison between the actual output value of the acceleration sensor and the output value based on the measured value.

FIG. 11 is a correlation diagram showing measured values of heaving displacement.

FIG. 12 is a correlation diagram showing measured values of rolling angles.

FIG. 13 is a correlation diagram showing measured values of pitching angles.

[Explanation of symbols]

1 Bow 2 Stern 3 Deck 4 Pendulum 5 Instrument box J Reference point on deck L Distance from intersection of rolling axis and pitching axis to fulcrum of pendulum l _c Length from center of gravity of pendulum to fulcrum O Rolling axis and pitching axis Intersection point Q Horizontal plane S ₁ Rolling tilt angle sensor S ₂ Pitching tilt angle sensor S ₃ Heaving displacement sensor T _c Torque applied to the pendulum X Rolling axis Y Pitching axis Z Heaving measurement direction γ Pendulum swing angle

Claims (1)

[Claims] 1. Heaving / rolling and pitching occurring on a ship are represented by a dynamics model of a state variable display as a linear system, and two servos installed on the ship to obtain rolling information and pitching information are provided in the dynamics model. A vessel attitude online automatic measuring method, characterized in that a Kalman filter is applied by combining observation equations by a type inclination angle sensor and one servo type acceleration sensor for obtaining heaving information.