Laser Gyro Temperature Compensation Method Based on Gaussian Process Regression and Distributed Temperature Measurement

Abstract

The present invention relates to the technical field of laser gyroscopes, and specifically relates to a temperature compensation method for laser gyroscopes based on Gaussian process regression and distributed temperature measurement, which can be used for temperature compensation of laser gyroscopes in complex temperature-changing environments, and is divided into the following steps: S1: Configuration of distributed temperature measurement for laser gyroscopes; S2: Construction of a temperature compensation model for laser gyroscopes; S3: SOC real-time prediction of temperature drift and temperature compensation; S4: SOC corrects the angular velocity output of the laser gyroscope at time k to obtain the corrected angular velocity output at time k. The present invention combines a distributed temperature measurement scheme with Gaussian process regression, and the accuracy of the angular velocity output of the laser gyroscope after temperature compensation is significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of laser gyroscopes, and particularly to a temperature compensation method for a laser gyroscope based on Gaussian process regression and distributed temperature measurement, which can be used for temperature compensation of a laser gyroscope in a complex temperature-changing environment. Background Art

[0002] A laser gyroscope is an angular velocity sensor based on the Sagnac effect and can be used for angular velocity measurement in the fields of navigation, guidance, and control. A high-precision laser gyroscope can measure tiny angular velocities such as the earth's rotation and hull deformation and is one of the very important sensors in the measurement field. Since a laser gyroscope is an optical instrument based on the Sagnac effect, there are components such as an optical path and a laser inside. These components will all undergo geometric deformation and physical property changes with the change of the external temperature, resulting in changes in the optical path and ultimately causing errors in the measurement of the laser gyroscope. This error caused by temperature change is called temperature drift.

[0003] Researchers have adopted various methods from both the device level and the algorithm level to suppress the temperature drift problem of a laser gyroscope in a complex temperature-changing environment. From the device level, even after using ultra-low expansion coefficient quartz glass and optimizing the internal structure of the gyroscope, there is still an obvious temperature drift. Further optimizing the material and structure not only has a long research period but also a high R & D cost. From the algorithm level, traditional temperature compensation methods usually use methods such as linear regression, support vector machines, and neural networks for temperature compensation. By constructing a temperature compensation model, the prediction of temperature drift is realized. However, in the actual application process, people have found that the temperature drift prediction accuracy of the above methods is relatively low, and the generalization of temperature compensation is also relatively low. The temperature drift problem of a laser gyroscope in a complex temperature-changing environment has not been effectively solved, which greatly limits the application scenarios of the laser gyroscope. Summary of the Invention

[0004] The present invention proposes a temperature compensation method for a laser gyroscope based on Gaussian process regression and distributed temperature measurement, which can compensate for the temperature drift of a laser gyroscope in a complex temperature-changing environment with high precision and can effectively solve the problems existing in the prior art.

[0005] The present invention is realized through the following technical solutions. A temperature compensation method for a laser gyroscope based on Gaussian process regression and distributed temperature measurement, the method is divided into the following steps:

[0006] S1: Configuration of distributed temperature measurement of the laser gyroscope;

[0007] The laser gyroscope used in S1.1 includes a box body, a system on chip (SOC), and a cavity; the cavity is fixed in the box body through a mounting hole, and optical components required for laser propagation and photoelectric signal detection, such as an anode, a cathode, a reflector, a path length control mirror, a semi-transmissive and semi-reflective mirror, a prism, a diaphragm, and a photodiode, are installed in the cavity; the system on chip is located on the inner wall of the box body, and the system on chip (SOC) is used to collect the output signal of the laser gyroscope, process and calculate the signal, and finally output the measured angular velocity output outward;

[0008] In S1.2, temperature measuring resistors are arranged near the cathode, left anode, right anode, diaphragm of the laser gyroscope and near the mounting boss of the box body, and the temperature measuring resistors are PT1000 temperature measuring resistors. Arranging the temperature measuring resistors is used to measure the temperatures near the cathode, anode, diaphragm and the mounting boss of the box body to characterize the global temperature characteristics of the laser gyroscope.

[0009] In S1.3, the temperature measuring resistors at 5 points in S1.2 are formed into a temperature measuring bridge and input into the SOC, and the SOC measures the temperatures of each point in real time. The SOC sends the angular velocity output of the laser gyroscope and the temperatures of each temperature point to the computer through a data transmission cable;

[0010] S2: Construct a temperature compensation model for the laser gyroscope

[0011] In S2.1, the laser gyroscope is placed in a temperature rapid change test chamber, the laser gyroscope is installed and fixed in the temperature rapid change test chamber, and the power supply cable and data transmission cable of the laser gyroscope are respectively connected to the power supply and the computer outside the temperature rapid change test chamber;

[0012] In S2.2, set the temperature change program of the temperature rapid change test chamber:

[0013] 1) Turn on the machine and keep the temperature at 20°C for 2 hours;

[0014] 2) Cool down to -40°C at a temperature change rate of 0.5°C / min and keep it for 2 hours;

[0015] 3) Heat up to 60°C at a temperature change rate of 0.5°C / min and keep it for 2 hours;

[0016] 4) Cool down to 20°C at a temperature change rate of 0.5°C / min and keep it for 2 hours;

[0017] 5) Cool down to -40°C at a temperature change rate of 1°C / min and keep it for 2 hours;

[0018] 6) Heat up to 60°C at a temperature change rate of 1°C / min and keep it for 2 hours;

[0019] 7) Cool down to 20°C at a temperature change rate of 1°C / min and keep it for 2 hours;

[0020] 8) Cool down to -40°C at a temperature change rate of 1.5°C / min and hold for 2 hours;

[0021] 9) Heat up to 60°C at a temperature change rate of 1.5°C / min and hold for 2 hours;

[0022] 10) Cool down to 20°C at a temperature change rate of 1.5°C / min and hold for 2 hours;

[0023] 11) Cool down to -40°C at a temperature change rate of 2°C / min and hold for 2 hours;

[0024] 12) Heat up to 60°C at a temperature change rate of 2°C / min and hold for 2 hours;

[0025] 13) Cool down to 20°C at a temperature change rate of 2°C / min and hold for 2 hours, then turn off the machine;

[0026] S2.3 Power on the ring laser gyroscope, and transmit the output data to the computer in real time and save it; the temperature change test chamber starts the temperature change program of S2.2. From the first step of power on to the 13th step of power off, it takes a total of 143,600 seconds, denoted as N seconds; after the temperature change program ends, the ring laser gyroscope is powered off.

[0027] The collected and saved data includes:

[0028] 1) The angular velocity output data ω of the ring laser gyroscope per second i , where the subscript i represents the moment, i = 1, 2, 3…, N;

[0029] 2) The real-time temperature of 5 temperature measurement points of the ring laser gyroscope per second

[0030] 3) The status data s of the ring laser gyroscope per second i , including status parameters such as the light intensity, dither frequency, and control voltage of the ring laser gyroscope;

[0031] S2.4 Construct a temperature compensation model for the ring laser gyroscope based on Gaussian process regression (S. T. Ounpraseuth, "Gaussian Processes for Machine Learning," J. Amer. Statist. Assoc., pp. 429 - 429, Mar. 2008, doi: 10.1198 / jasa.2008.s219.); specifically as follows:

[0032] S2.4.1 Construct the input data X and response data Y of the temperature compensation model training set:

[0033] The response data is the temperature drift output data of the ring laser gyroscope per second: Where is the average angular velocity output by the ring laser gyroscope at room temperature, which is calculated from the angular velocity data collected during the 20°C heat preservation stage in S2.2;

[0034] The input data is the real-time temperature of 5 temperature measurement points of the ring laser gyroscope per second as well as the square terms, time gradient terms, and spatial gradient terms of the 5 temperatures. The specific forms are as follows:

[0035]

[0036] Among them,

[0037]

[0038] And it is defined that

[0039] Let x i represent the temperature and its square term, time gradient term, and spatial gradient term at the i-th moment, that is:

[0040] Therefore, the input data X can be expressed as:

[0041]

[0042] S2.4.2 Calculate the Gaussian process regression kernel matrix

[0043] Calculate the Gaussian process regression kernel matrix K of the input data X as follows:

[0044] K = K(X, X)

[0045] Among them, K(X, X) represents calculating the Gaussian process regression kernel matrix of the input data X:

[0046]

[0047] K rc = k(x r , x c ); (r = 1, 2, 3…, N, c = 1, 2, 3…, N) represents the kernel function of the input data X. The subscripts r and c represent the r-th row and c-th column of K(X, X) respectively; the kernel function can map low-dimensional data to high-dimensional space, thereby improving the nonlinear fitting ability of the model. The expression of the kernel function is:

[0048]

[0049] Among them, σ 2 is the amplitude hyperparameter of the kernel function, which is used to control the variance of the output; l represents the length scale hyperparameter of the kernel function, which is used to control the influence of the input distance on the covariance; based on experience, it is selected as

[0050] σ 2 = 0.0001, l = 1;

[0051] S2.4.3 Store the calculated Gaussian process regression kernel matrix K, input data X, and response data Y into

[0052] the SOC of the ring laser gyro;

[0053] S3: The SOC predicts the temperature drift in real time and performs temperature compensation

[0054] S3.1 Denote the actual usage time as the k-th moment. The SOC collects the angular velocity output ω of the ring laser gyro at the k-th moment k and the temperature observation data T at the k-th moment k :

[0055]

[0056] S3.2 Construct the temperature observation vector Z k :

[0057]

[0058] S3.3 The SOC calculates the covariance vector between the temperature observation vector Z k and the input data X of the training set:

[0059]

[0060] S3.4 The SOC calculates the Gaussian process regression prediction mean μ at the k-th moment k :

[0061] μ k = K ZX T K -1 Y

[0062] where the prediction mean μ k represents the estimate of the output; K ZX T represents the transpose of the covariance vector K ZX ; K -1 represents the inverse matrix of the kernel matrix K;

[0063] Store the calculated Gaussian process regression prediction mean μ at the k-th moment k as the temperature drift of the ring laser gyro at the k-th moment;

[0064] S4 The SOC corrects the angular velocity output of the ring laser gyro at the k-th moment to obtain the corrected angular velocity output at the k-th moment

[0065]

[0066] After S4 is executed, k = k + 1, and S3 to S4 are looped to execute, so that the angular velocity output of the ring laser gyroscope after temperature compensation at the latest moment can be obtained in real time.

[0067] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0068] 1. Based on the working principle and basic characteristics of the ring laser gyroscope, the present invention designs a distributed temperature measurement scheme, and can obtain observation data that can comprehensively reflect the working state and temperature characteristics of the ring laser gyroscope;

[0069] 2. The present invention performs temperature compensation based on Gaussian process regression. Compared with the traditional linear regression model, it can effectively process the non-linear relationship between the output of the ring laser gyroscope and temperature, and is applicable to various types of function approximation; 3. Compared with the traditional support vector machine temperature compensation method, the temperature compensation method based on Gaussian process regression does not need to specify a complex model structure in advance, which reduces the burden of parameter adjustment;

[0070] 4. The traditional neural network temperature compensation method has a very high requirement for the amount of training data. Compared with the traditional neural network temperature compensation method, the temperature compensation method based on Gaussian process regression can achieve high generalization using small sample training data, saving training costs and shortening the production cycle;

[0071] 5. The present invention combines the distributed temperature measurement scheme and Gaussian process regression, and the accuracy of the angular velocity output of the ring laser gyroscope after temperature compensation is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 : Schematic diagram of the structure of the ring laser gyroscope of the present invention;

[0073] Figure 2 : Layout diagram of the distributed temperature measurement scheme of the ring laser gyroscope of the present invention;

[0074] Figure 3 : Implementation process of the present invention;

[0075] Figure 4 : Comparison diagram of various temperature compensation schemes. DETAILED DESCRIPTION OF THE INVENTION

[0076] To describe in detail the technical solutions disclosed in the present invention, the following further elaborates with specific embodiments. The specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0077] Figure 3 The implementation process of the present invention is divided into the following steps:

[0078] S1: Distributed temperature measurement configuration for laser gyroscopes;

[0079] S2: Construct a temperature compensation model for laser gyroscopes;

[0080] S3: SOC predicts temperature drift in real time and performs temperature compensation

[0081] S4 SOC corrects the angular velocity output of the laser gyroscope at time k Obtain the corrected angular velocity output at time k

[0082]

[0083] After each execution of S4, k = k + 1, and return to S3 to start over.

[0084] This embodiment conducted a test. The angular velocity output of the laser gyroscope is as Figure 4 shown by the thin solid line in the figure. After performing real-time temperature compensation, The superscript M is the test duration of this embodiment, as Figure 4 shown by the thick solid line in the figure. The thin solid line in the figure is the angular velocity output without temperature compensation. It can be seen that it fluctuates violently under temperature changes. The dotted line, the dash-dotted line, and the dashed line are different traditional temperature compensation methods respectively. It can be seen from the figure that their trends are not as stable as the thick solid line of the present invention. The output zero-bias instability of the laser gyroscope under different temperature compensation methods is statistically shown in Table 1.

[0085] Table 1 Comparison of zero-bias instability of temperature compensation by different methods

[0086]

[0087] As can be seen from Table 1, the zero-bias instability of the method proposed in the present invention is the smallest, indicating that the temperature compensation accuracy of this method is the highest, verifying the effectiveness and superiority of this method, and reflecting the beneficial effects of the present invention.

Claims

1. A laser gyro temperature compensation method based on Gaussian process regression and distributed temperature measurement, characterized in that: The method is divided into the following steps: S1: Laser gyro distributed temperature measurement configuration: The laser gyro used in S1.1 includes a box body, a system on chip SOC, and a cavity; the cavity is fixed in the box body through a mounting hole, and the cavity is equipped with an anode, a cathode, a reflector, a path length control mirror, a semi-transparent and semi-reflective mirror, a prism, an aperture, and a photodiode. The optical components required for laser propagation and photoelectric signal detection; the system on chip is located on the inner wall of the box body, and the system on chip SOC is used to collect the output signal of the laser gyro, process and calculate the signal, and finally output the measured angular velocity output to the outside; S1.2 Place temperature measuring resistors near the cathode, left anode, right anode, aperture and the mounting boss of the laser gyro to measure the temperature of the cathode, anode, aperture and the mounting boss of the box to characterize the global temperature characteristics of the laser gyro; S1.3 forms a temperature measuring bridge with the temperature measuring resistors at the five points in S1.2 and inputs it into the SOC, which measures the temperature of each point in real time; the SOC sends the angular velocity output of the laser gyro and the temperature of each temperature point to the computer through the data transmission cable; S2: Constructing laser gyro temperature compensation model S2.1 Place the laser gyro in the temperature rapid change test chamber, install and fix the laser gyro in the temperature rapid change test chamber, and connect the power cord and data transmission cable of the laser gyro to the power supply and computer outside the temperature rapid change test chamber respectively; S2.2 Set the temperature change program of the temperature rapid change test chamber: 1) Turn on the machine and keep it at 20℃ for 2 hours; 2) Cool down to -40°C at a temperature ramp rate of 0.5°C / min and keep warm for 2 hours; 3) Raise the temperature to 60°C at a temperature ramp rate of 0.5°C / min and keep warm for 2 hours; 4) Cool down to 20°C at a temperature change rate of 0.5°C / min and keep warm for 2 hours; 5) Cool down to -40°C at a temperature ramp rate of 1°C / min and keep warm for 2 hours; 6) Raise the temperature to 60°C at a temperature ramp rate of 1°C / min and keep warm for 2 hours; 7) Cool down to 20°C at a temperature change rate of 1°C / min and keep warm for 2 hours; 8) Cool down to -40°C at a temperature ramp rate of 1.5°C / min and keep warm for 2 hours; 9) Raise the temperature to 60°C at a temperature ramp rate of 1.5°C / min and keep warm for 2 hours; 10) Cool down to 20°C at a temperature ramp rate of 1.5°C / min and keep warm for 2 hours; 11) Cool down to -40°C at a temperature ramp rate of 2°C / min and keep warm for 2 hours; 12) Raise the temperature to 60°C at a temperature ramp rate of 2°C / min and keep warm for 2 hours; 13) Cool down to 20°C at a temperature change rate of 2°C / min, keep warm for 2 hours, and then turn off the machine; S2.3 The laser gyro is powered on, and the output data is transmitted to the computer in real time and saved; the temperature rapid change test chamber starts the temperature change program of S2.2, from the first step of power on to the thirteenth step of power off, which takes a total of 143600 seconds, recorded as N seconds; After the temperature change program is completed, the laser gyro is shut down and powered off; The collected and saved data include: 1) The laser gyro outputs angular velocity data per second ω i , subscript i represents the time, i=1,2,3…,N; 2) Real-time temperature of the laser gyro at 5 temperature measurement points per second 3) Laser gyro status data per second s i , including the laser gyro's light intensity, jitter frequency, and control voltage state parameters; S2.4 builds a laser gyro temperature compensation model based on Gaussian process regression; the details are as follows: S2.4.1 Construct the input data X and response data Y of the temperature compensation model training set: The response data is the temperature drift output data of the laser gyro per second: in is the average angular velocity output by the laser gyro at room temperature, which is calculated from the angular velocity data collected during the 20℃ insulation stage in 1) S2.2; The input data is the real-time temperature of the laser gyro's five temperature measurement points per second. And 5 temperature square terms, time gradient terms, and space gradient terms. The specific forms are as follows: in, And define Use x i represents the temperature at the i-th moment and its square term, time gradient term and space gradient term, that is: Therefore, the input data X can be expressed as: S2.4.2 Calculation of Gaussian Process Regression Kernel Matrix The Gaussian process regression kernel matrix K for the input data X is calculated as follows: K=K(X,X) Among them, K(X,X) represents the Gaussian process regression kernel matrix for calculating the input data X: K rc = k(x r ,x c );(r=1,2,3…,N,c=1,2,3…,N) represents the kernel function of the input data X, and the subscripts r and c represent the rth row and cth column of K(X,X) respectively; the kernel function can map low-dimensional data to high-dimensional space. Thereby improving the nonlinear fitting ability of the model, the expression of the kernel function is: where σ 2 is the amplitude hyperparameter of the kernel function, which is used to control the variance of the output; l represents the length scale hyperparameter of the kernel function, which is used to control the effect of input distance on covariance; S2.4.3 storing the calculated Gaussian process regression kernel matrix K, input data X and response data Y into the SOC of the laser gyro; S3: SOC real-time prediction of temperature drift and temperature compensation S3.1 The actual use time is recorded as time k, and SOC collects the laser gyro angular velocity output ω at time k k And the temperature observation data T at time k k : S3.2 Constructing the temperature observation vector Z k : S3.3 SOC calculation temperature observation vector Z k The covariance vector between the training set input data X is: S3.4 Gaussian process regression prediction mean μ for SOC calculation at time k k : μ k =K ZX T K -1 Y Among them, the predicted mean μ k Represents an estimate of the output; K ZX T Denotes the covariance vector K ZX The transpose of K -1 represents the inverse matrix of the kernel matrix K; The Gaussian process at time k is regressed to predict the mean μ k is the temperature drift of the laser gyro at time k; S4 SOC corrects the angular velocity output of the laser gyro at time k Get the corrected angular velocity output at time k After S4 is executed, k=k+1, and S3 to S4 are executed in a loop, so that the angular velocity output of the laser gyro after temperature compensation at the latest moment can be obtained in real time.

2. The laser gyro temperature compensation method based on Gaussian process regression and distributed temperature measurement according to claim 1, characterized in that: In S1.2, the temperature measuring resistor is a PT1000 temperature measuring resistor.

3. The laser gyro temperature compensation method based on Gaussian process regression and distributed temperature measurement according to claim 1, characterized in that: In S2.4.2, the magnitude hyperparameter σ of the kernel function 2 =0.0001, and the length scale hyperparameter of the kernel function is l=1.

Images