CN101408427A - Distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope - Google Patents

Abstract

An optical fibre gyroscope distributed layered temperature error compensation method comprises the following steps: (1) five optical components of a fiber-optic ring, a light source, a Y wave guide, a detector and a coupler are taken as temperature monitoring objects for determining the number of the temperature monitoring points of each optical component and the distributing form; (2) self-adapting recurrence and least squares treatment are carried out on monitoring value of each temperature monitoring point for obtaining temperature output value of the optical fibre gyroscope; (3) a neural network model is built for obtaining temperature drift compensation value by utilizing the temperature output value and the real time outputting data of the optical fibre gyroscope; (4) the real time outputting data of the optical fibre gyroscope minus the temperature drift compensation value means temperature compensation of the optical fibre gyroscope. The method can overcome the disadvantages of the prior art and reflect the temperature field inside the optical fibre gyroscope systematically and in an all-round way, and has significance in the performance research and improvement under the condition of the temperature environment of the optical fibre gyroscope.

Description

A kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope

Technical field

The present invention relates to a kind of temperature compensation method for optical fiber gyroscope.

Background technology

The temperature on fiber gyroscope especially performance of high-precision optical fiber gyro instrument has a significant impact, and temperature variation derives from fibre optic gyroscope self-heating and variation of ambient temperature.Because temperature variant not stationarity of fibre optic gyroscope and time variation, the time series common as if employing etc. then make the accuracy of model relatively poor based on the technology of parameter identification, and compensation effect is desirable not to the utmost.And, the temperature model of most of fibre optic gyroscopes only limits to the static temperature model or the simplification dynamic temperature model that temperature monitoring is set up is carried out in the part, and it far can not satisfy the stable and consistance of performance under the temperature variation that in the engineering application high-precision optical fiber gyro instrument is proposed.

The fibre optic gyroscope temperature model of being set up at present, the subject matter that exists is: system is not carried out in the temperature field of fibre optic gyroscope comprehensively analyze, all optical components are lacked analysis, the selection principle that does not have the design temperature monitoring point, emphasis is only paid close attention to fiber optic loop, has ignored other important steps, and the model that causes setting up does not embody the real temperature field of optical fibre gyro, when system applies, fiber optic gyroscope performance is deterioration along with exterior temperature change.

Patent publication No. CN101013035A, denomination of invention " a kind of optical fibre gyro that carries out temperature compensation based on neural network ", in a kind of optical fibre gyro that carries out temperature compensation based on neural network is disclosed, the temperature drift compensation mode that this gyro seemingly adopts Learning Algorithm to carry out, the correction temperature compensation coefficient that obtains through this temperature compensation is to provide parameter for the output of optical fibre gyro high precision.4 sensors have been installed in this application altogether, have been installed in respectively on light source, Y waveguide and the fiber optic loop, wherein the interior outside of fiber optic loop is respectively installed one.Collecting temperature value by these four sensors is carried out temperature compensation based on neural network to above-mentioned four road real time temperatures, emphasis in the literary composition is how based on the temperature compensation of neural network, comprise the extraction of the weights coefficient of neural network, the foundation of model of temperature compensation, do not describe definite principle of number of sensors in detail, and just consider light source, Y waveguide and fiber optic loop in this application, do not consider the temperature of detector and coupling mechanism, also do not have to introduce and how the temperature of gathering is handled.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope is provided, this method can system comprehensively reflect fibre optic gyroscope temperature inside field.

Technical solution of the present invention is: a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope, and step is as follows:

(1) with fiber optic loop, light source, Y waveguide, detector, five optical components of coupling mechanism as the temperature monitoring object, determine the temperature monitoring number of spots and the distribution form of each optical component;

(2) monitor value to described each temperature monitoring point carries out the processing of ADAPTIVE RECURSIVE least square, obtains the temperature output valve of fibre optic gyroscope;

(3) utilize the temperature output valve of described fibre optic gyroscope and the real time output data of fibre optic gyroscope, set up neural network model, obtain the temperature drift compensation value;

(4) real time output data with fibre optic gyroscope deducts described temperature drift compensation value, promptly fibre optic gyroscope has been carried out temperature compensation.

The quantity of the temperature monitoring point of the fiber optic loop of described step (1) determines that formula is:

$n=\mathrm{\Δ}\·{(L\·D\·{\left(\frac{\mathrm{\λ}}{\mathrm{\α}}\right)}^2)}^{1/6}$

Wherein, n is a temperature monitoring point number;

L is a fiber optic loop length;

D fiber optic loop diameter;

λ is the heat-conduction coefficient of fiber optic loop framework material;

α is the thermal expansivity of fiber optic loop framework material;

Δ is an empirical parameter, usually value between 0.85～5.35.

When described light source was Er-Doped superfluorescent fiber source, the quantity of its temperature monitoring point determined that formula is:

$n={\mathrm{\Δ}}^{\′}\·\frac{\stackrel{\‾}{\mathrm{\λ}}}{{\mathrm{\λ}}_{\mathrm{pump}}}\·P_{\mathrm{pump}}$

Wherein, λ is the Er-doped fiber mean wavelength;

λ _PumpBe pumping wavelength;

P _PumpBe pump power;

Δ ' be empirical parameter, value between 50～80 usually.

Distribution form in the described step (1) generally adopts symmetry uniform.

The quantity of described Y waveguide, detector, coupling mechanism temperature monitoring point is rule of thumb determined, generally gets 1, is placed on the centre of the shell of Y waveguide, detector, coupling mechanism.

The described neural network model of setting up is chosen three layers of BP neural network N _1,5,1, promptly input layer and output layer are all got a node, and hidden layer is got 5 nodes, and the excitation function of latent node is selected formula:

$f\left(x\right)=\frac{1}{1+e^{-2x}}$

The weights of BP neural network are adjusted formula:

$w_{\mathrm{ij}}(t+1)=w_{\mathrm{ij}}\left(t\right)-0.5\frac{\∂J\left(t\right)}{\∂w_{\mathrm{ij}}\left(t\right)}+0.1[w_{\mathrm{ij}}\left(t\right)-w_{\mathrm{ij}}(t-1)]$

In the formula,

Be the error objective function, Be the actual output of BP network after t weights adjustment.

The present invention compared with prior art beneficial effect is:

(1) the present invention passes through fiber optic loop, light source, Y waveguide, detector, five optical components of coupling mechanism as the temperature monitoring object, and determine the quantity and the distribution mode of each temperature monitoring point, and to the laggard trip temperature compensation of the Temperature Treatment of gathering, this method can system comprehensively reflect fibre optic gyroscope temperature inside field, improves the precision of fibre optic gyroscope under the temperature variation condition.

(2) the present invention adopts the ADAPTIVE RECURSIVE least square to handle the temperature acquisition value of five big optical components, has guaranteed the true and comprehensive of neural network model.

Description of drawings

Fig. 1 is the inventive method process flow diagram;

Fig. 2 is fibre optic gyroscope schematic diagram of the present invention and temperature sensor schematic layout pattern;

Fig. 3 is a fiber optic loop temperature monitoring point distribution plan of the present invention;

Fig. 4 is a light-source temperature of the present invention monitoring point distribution plan;

Fig. 5 is a Y waveguide temperature monitoring point distribution plan of the present invention;

Fig. 6 is a detector temperature of the present invention monitoring point distribution plan;

Fig. 7 is a coupling mechanism temperature monitoring point distribution plan of the present invention;

Fig. 8 is a BP learning method process flow diagram of the present invention;

Fig. 9 is BP network curve of output of the present invention and the contrast of actual curve of output;

Figure 10 is a curve of output after the fibre optic gyroscope temperature compensation of the present invention.

Embodiment

Below in conjunction with specific embodiment, introduce the inventive method in detail.

As shown in Figure 1, 2, the step of a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope of the present invention is as follows:

(1) with fiber optic loop, light source, Y waveguide, detector, five optical components of coupling mechanism as the temperature monitoring object, determine the temperature monitoring number of spots and the distribution form of each optical component;

Utilize formula (1) can obtain the quantity of the required temperature monitoring point of fiber optic loop:

$n=\mathrm{\Δ}\·{(L\·D\·{\left(\frac{\mathrm{\λ}}{\mathrm{\α}}\right)}^2)}^{1/6}---\left(1\right)$

Wherein, n is a temperature monitoring point number;

L is fiber optic loop length (m);

D fiber optic loop diameter (m);

λ is the heat-conduction coefficient (W/ (mK)) of fiber optic loop framework material;

α is the thermal expansivity (1/K) of fiber optic loop framework material;

Δ be empirical parameter (individual/

), value between 0.85～5.35 usually.

The length of fiber optic loop temperature field and fiber optic loop is directly related with the product of fiber optic loop diameter, i.e. n ∝ (LD) ^1/6, and the fiber optic loop framework material is very big to the heat radiation and the thermal expansion influence of fiber optic loop, what of temperature monitoring point are the performance of heat radiation and expansion determine, and heat radiation and normally interaction of expansion, influence each other, and are definite $n\∝{\left(\frac{\mathrm{\λ}}{\mathrm{\α}}\right)}^{1/3},$ Therefore, $n={(L\·D\·{\left(\frac{\mathrm{\λ}}{\mathrm{\α}}\right)}^2)}^{1/6},$ In order better to reflect the development technology of fibre optic gyroscope, need to increase an empirical parameter Δ usually and adjust, to obtain better temperature field monitoring effect.Finally, the computing formula of fiber optic loop temperature monitoring number of spots as shown in Equation (1).

Temperature monitoring number of spots by fiber optic loop in the calculating present embodiment is 10, as shown in Figure 3, and outer uniform 4 of fiber optic loop, uniform 4 of fiber optic loop internal layer, each one up and down of fiber optic loop.

Light source described in the present invention can adopt SLD, also can adopt Er-Doped superfluorescent fiber source, generally gets 1 for SLD temperature monitoring point, utilizes formula (2) to obtain for the required temperature monitoring point of Er-Doped superfluorescent fiber source:

$n={\mathrm{\Δ}}^{\′}\·\frac{\stackrel{\‾}{\mathrm{\λ}}}{{\mathrm{\λ}}_{\mathrm{pump}}}\·P_{\mathrm{pump}}---\left(2\right)$

Wherein, λ is Er-doped fiber mean wavelength (nm);

λ _PumpBe pumping wavelength (nm);

P _PumpBe pump power (W);

Δ ' be empirical parameter (individual/W), value between 50～80 usually.

The temperature field of Er-Doped superfluorescent fiber source is mainly relevant with Er-doped fiber mean wavelength, pumping wavelength, pump power.The proportional relation of the product of the quantity of temperature monitoring point and Er-doped fiber mean wavelength and pump power, i.e. n ∝ λ P _Pump, and with the pumping wavelength relation of being inversely proportional to, i.e. n ∝ 1/ λ _Pump,, need to increase an empirical parameter Δ ' adjust, usually to obtain better temperature field monitoring effect for the better development technology of reflection fibre optic gyroscope.Finally, the computing formula of Er-Doped superfluorescent fiber source temperature monitoring number of spots as shown in Equation (2).

Temperature monitoring number of spots by Er-Doped superfluorescent fiber source in the calculating present embodiment is 4, and as shown in Figure 4, pump laser is placed 1, and it is uniform to be 120 degree by circumference around fiber optic coils.

The quantity of described Y waveguide, detector, coupling mechanism temperature monitoring point is rule of thumb determined, generally gets 1, is placed on the centre of the shell of Y waveguide, detector, coupling mechanism, respectively shown in Fig. 5～7.

(2) monitor value to described each temperature monitoring point carries out the processing of ADAPTIVE RECURSIVE least square, obtains the temperature output valve of fibre optic gyroscope;

After having determined the temperature monitoring point of above-mentioned five big optical components, the temperature output valve of fiber optic loop is weighted processing, can utilizes formula (3) to obtain:

T _f＝0.13·(T ₁+T ₃+T ₅+T ₈)+0.095·(T ₂+T ₄+T ₆+T ₇)+0.055·(T ₉+T ₁₀) (3)

The temperature output valve of light source can utilize formula (4) to obtain:

T _S＝0.12·(T ₁+T ₂+T ₃)+0.65·T ₄ (4)

The temperature output valve of Y waveguide is T _Y, the temperature output valve of detector is T _p, the temperature output valve T of coupling mechanism _c

Five temperature values that obtain are above carried out the ADAPTIVE RECURSIVE least square handle, utilize formula (5) to obtain the temperature value of fibre optic gyroscope:

T＝0.4·T _f+0.25·T _S+0.1·T _Y+0.2·T _p+0.05·T _c (5)

(3) utilize the temperature output valve of described fibre optic gyroscope and the real time output data of fibre optic gyroscope, set up the BP neural network model, obtain the temperature drift compensation value; Process is as follows:

The first step is selected the high precision high-low temperature test chamber, (gets-20～+ 40 ℃ in this example) in given temperature range, repeatedly carries out the test of high-precision optical fiber gyro instrument once electrification, obtains many group test samples.

Second step, input sample set T=-20: 0.05: 40, output sample collection O=f (T).

In the 3rd step, choose three layers of BP neural network N _1,5,1, promptly input layer and output layer are all got a node, and hidden layer is got 5 nodes.The excitation function of latent node is selected the asymmetric Sigmoid function shown in the formula (6):

$f\left(x\right)=\frac{1}{1+e^{-2x}}---\left(6\right)$

In the 4th step, get and decide weights adjustment formula:

$w_{\mathrm{ij}}(t+1)=w_{\mathrm{ij}}\left(t\right)-0.5\frac{\∂J\left(t\right)}{\∂w_{\mathrm{ij}}\left(t\right)}+0.1[w_{\mathrm{ij}}\left(t\right)-w_{\mathrm{ij}}(t-1)]---\left(7\right)$

In the formula,

Be the error objective function.(

Be the actual output of BP network after t weights adjustment)

In the 5th step, carry out the BP neural net model establishing by the flow process of Fig. 8.The initial value of weights is set at random, and through 500 training, error objective function J (t) almost no longer reduces.As shown in Figure 9, BP model curve of output can approach the gyrostatic curve of output of actual fiber in good condition.

The 6th step, will set up in the actual output of BP network model substitution fibre optic gyroscope, obtain the temperature drift compensation value of fibre optic gyroscope.

The 7th step, the real time output data of fibre optic gyroscope is deducted described temperature drift compensation value, promptly fibre optic gyroscope has been carried out temperature compensation, result after the compensation as shown in figure 10, as can be seen from the figure, zero of compensation back optical fibre gyro (0.005 °/h) zero stable partially (0.06 °/h) nearly 10 times of raising before the compensation of stability partially.

The unspecified part of the present invention belongs to general knowledge as well known to those skilled in the art.

Claims (6)

1, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope is characterized in that step is as follows:

(1) with fiber optic loop, light source, Y waveguide, detector, five optical components of coupling mechanism as the temperature monitoring object, determine the temperature monitoring number of spots and the distribution form of each optical component;

(2) monitor value to described each temperature monitoring point carries out the processing of ADAPTIVE RECURSIVE least square, obtains the temperature output valve of fibre optic gyroscope;

(3) utilize the temperature output valve of described fibre optic gyroscope and the real time output data of fibre optic gyroscope, set up neural network model, obtain the temperature drift compensation value;

(4) real time output data with fibre optic gyroscope deducts described temperature drift compensation value, promptly fibre optic gyroscope has been carried out temperature compensation.

2, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope according to claim 1 is characterized in that: the quantity of the temperature monitoring point of the fiber optic loop of described step (1) determines that formula is:

$n=\mathrm{\Δ}\·{(L\·D\·{\left(\frac{\mathrm{\λ}}{\mathrm{\α}}\right)}^2)}^{1/6}$

Wherein, n is a temperature monitoring point number;

L is a fiber optic loop length;

D fiber optic loop diameter;

λ is the heat-conduction coefficient of fiber optic loop framework material;

α is the thermal expansivity of fiber optic loop framework material;

Δ is an empirical parameter, usually value between 0.85～5.35.

3, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope according to claim 1 is characterized in that: when described light source was Er-Doped superfluorescent fiber source, the quantity of its temperature monitoring point determined that formula is:

$n={\mathrm{\Δ}}^{\′}\·\frac{\stackrel{\‾}{\mathrm{\λ}}}{{\mathrm{\λ}}_{\mathrm{pump}}}\·P_{\mathrm{pump}}$

Wherein, λ is the Er-doped fiber mean wavelength;

λ _PumpBe pumping wavelength;

P _PumpBe pump power;

Δ is an empirical parameter, usually value between 50～80.

4, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope according to claim 1 is characterized in that: the distribution form in the described step (1) adopts symmetry uniform.

5, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope according to claim 1, it is characterized in that: the quantity of described Y waveguide, detector, coupling mechanism temperature monitoring point is rule of thumb determined, generally get 1, be placed on the centre of the shell of Y waveguide, detector, coupling mechanism.

6, a kind of distributed layer-dividing grade temperature error compensating method of optical fiber gyroscope according to claim 1 is characterized in that: the described neural network model of setting up is chosen three layers of BP neural network N _1,5,1, promptly input layer and output layer are all got a node, and hidden layer is got 5 nodes, and the excitation function of latent node is selected formula:

$f\left(x\right)=\frac{1}{1+e^{-2x}}$

The weights of BP neural network are adjusted formula:

$w_{\mathrm{ij}}(t+1)=w_{\mathrm{ij}}\left(t\right)-0.5\frac{\∂J\left(t\right)}{\∂w_{\mathrm{ij}}\left(t\right)}+0.1[w_{\mathrm{ij}}\left(t\right)-w_{\mathrm{ij}}(t-1)]$

In the formula, Be the error objective function, Be the actual output of BP network after t weights adjustment.

Images