A method and device for controlling temperature drift of a fiber optic gyroscope

Abstract

The invention discloses a method and device for controlling temperature drift of an optical fiber gyroscope, and relates to the technical field of optical fiber gyroscopes. The method comprises the following steps: obtaining a temperature control box temperature T1, an inner wall temperature T2 and an outer wall temperature T3 of an optical fiber gyroscope, and a drift output y in real time, obtaining a learning sample {T1, T2, T3, y}, inputting an adaptive reinforcement learning algorithm to obtain a predicted drift output value, calculating a difference between the predicted drift output value and an actual drift output value y to obtain a temperature drift control value of the optical fiber gyroscope, subtracting the temperature drift control value from the actual drift output value y to achieve temperature drift control; the device comprises an optical fiber gyroscope temperature drift control module, constructing a learning sample from the actual temperature control box temperature, the inner wall temperature and the outer wall temperature of the optical fiber gyroscope, and the drift output value, inputting the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value, calculating a difference between the predicted drift output value and the actual drift output value to obtain the temperature drift control value, and achieving improvement of optical fiber gyroscope drift with low time cost.

Description

Technical Field

[0001] The present application relates to the technical field of fiber optic gyroscopes, and in particular to a method and device for controlling temperature drift of a fiber optic gyroscope. Background Art

[0002] The fiber optic gyroscope is an optical instrument for accurately positioning moving objects. It has a simple shape, small size, and is wear-resistant and impact-resistant. It is widely used in many scenarios such as sea, land and air traffic control, GPS satellite navigation, and underground natural resource surveys.

[0003] Among them, the focus in the civil field is on the application of medium and low precision fiber optic gyroscopes. The main application scenarios are: automatic navigation, positioning and orientation, and vehicle control of ground vehicles; attitude control of agricultural aircraft, sowing and spraying of pesticides; positioning tools and rescue tools for finding damaged power lines, pipelines and communication optical (electric) cables in underground engineering maintenance; positioning and path surveying for geodetic surveying, mineral exploration, oil exploration, oil drilling guidance, tunnel construction, etc., as well as the use of fiber optic gyroscope rotation angle and linear displacement to achieve dam inclinometer, etc. It can be seen that the application scenarios of fiber optic gyroscopes have a trend of continuous expansion and extension.

[0004] In real applications, the fiber optic gyroscope will deviate due to changes in the surrounding environment and its own temperature, which affects the measurement accuracy of the fiber optic gyroscope. In order to improve its accuracy, temperature adjustment and control are required. In specific research, two technical methods are mainly used to control this: one is to improve the fiber winding method to suppress temperature drift; the other is to build a model and use an algorithm to compensate for temperature. For the first method, the improvement cost is high and it is not applicable to existing finished products; so the second method is mainly used in reality.

[0005] The search formula is TACD_ALL:(fiber optic gyroscope AND temperature AND drift AND compensation AND construction AND sample AND (learning OR adaptation OR regularization OR weighting)), and a relatively close existing technical solution is obtained.

[0006] The authorization announcement number is CN101408427B, and the name is a distributed hierarchical temperature error compensation method for fiber optic gyroscope, hereinafter referred to as comparative document 1. The steps are as follows: (1) Take the five optical components of optical fiber ring, light source, Y waveguide, detector, and coupler as temperature monitoring objects, and determine the number and distribution form of temperature monitoring points of each optical component; (2) Perform adaptive recursive least squares processing on the monitoring values ​​of each temperature monitoring point to obtain the temperature output value of the fiber optic gyroscope; (3) Use the temperature output value of the fiber optic gyroscope and the real-time output data of the fiber optic gyroscope to establish a neural network model to obtain the temperature drift compensation value; (4) Subtract the temperature drift compensation value from the real-time output data of the fiber optic gyroscope, that is, the fiber optic gyroscope is temperature compensated.

[0007] The authorization announcement number is CN102095419B, and the name is Fiber Optic Gyroscope Temperature Drift Modeling and Error Compensation Method, hereinafter referred to as Comparative Document 2. It includes the following steps: obtaining learning samples and denoising the samples; establishing a fiber optic gyroscope temperature drift error compensation model based on temperature gradient; establishing a parameter formula based on the temperature change rate to determine the relationship between the temperature change rate and the various parameters of the fiber optic gyroscope temperature drift error compensation model; subtracting the temperature drift compensation value obtained by using the fiber optic gyroscope temperature drift error compensation model from the real-time output data of the fiber optic gyroscope, that is, temperature compensation is performed on the fiber optic gyroscope.

[0008] The authorization announcement number is CN101840529B, and the name is "Optical Fiber Gyroscope Random Drift Modeling Method Based on Local Variable Integrated Neural Network", hereinafter referred to as Comparative Document 3. The method includes the following steps: (1) establishing an integrated neural network model; (2) obtaining learning samples; (3) training the integrated neural network; and (4) modeling the random drift of the optical fiber gyroscope.

[0009] Combining the above three patent documents and existing technical solutions, the inventors analyzed and found that:

[0010] Comparative Document 1 overcomes the deficiencies of the prior art and can systematically and comprehensively reflect the temperature field inside the fiber optic gyroscope.

[0011] Comparative Document 2 not only realizes the modeling of the temperature drift of the fiber optic gyroscope from a mechanism perspective, but also greatly improves the simplicity and accuracy of the modeling.

[0012] Compared with document 3, this method first uses the FLP / wavelet filtering method to reduce the noise of the original output signal of the fiber optic gyroscope, forming a data preprocessing system with good denoising effect and wide application range; secondly, the local variable integrated neural network is applied to the modeling, and the integrated neural network overfits the training samples, overcoming the shortcomings of low prediction accuracy and poor stability of the sub-neural network. The introduction of the temperature-based local variable weight method enables the network to adjust the integrated weight matrix according to the change of temperature, ensuring that the best network modeling accuracy can be obtained at different temperatures, improving the generalization ability of the network, and realizing the dynamic modeling of the zero drift of the fiber optic gyroscope.

[0013] In order to obtain a new technical route, the inventor analyzed and considered as follows:

[0014] Although there are many existing temperature drift error compensation model algorithms, they are mainly concentrated in BP neural network and RBF neural network. These algorithms have many parameter settings, high order, long sample training time, limited generalization of constructed models, long calculation delay, and often overfitting. In short, the compensation effect is not satisfactory.

[0015] In fact, there is no exact model for the temperature drift compensation algorithm of fiber optic gyroscopes so far. Not only is it impossible to determine the specific temperature-related quantities that cause temperature drift errors, but the relationship between these related quantities cannot be accurately described. Therefore, in the case that such a model is difficult to accurately construct, relying on the rapid development of big data and machine learning, it is very necessary to invent a new control method to achieve more precise control, thereby improving the drift problem of fiber optic gyroscopes and further improving the measurement accuracy of fiber optic gyroscopes.

[0016] Existing technical problems and considerations:

[0017] How to solve the technical problem of improving fiber optic gyroscope drift. Summary of the invention

[0018] The technical problem to be solved by the present application is to provide a method and device for controlling the temperature drift of a fiber optic gyroscope, so as to solve the technical problem of improving the drift of the fiber optic gyroscope.

[0019] In order to solve the above technical problems, the technical solution adopted in the present application is: a method for controlling temperature drift of a fiber optic gyroscope comprises the following steps: adjusting the temperature T1 of a temperature control box between -40°C and 60°C, obtaining in real time the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope, and then obtaining a learning sample {T1, T2, T3, y}, inputting the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value of the fiber optic gyroscope, subtracting the predicted drift output value of the fiber optic gyroscope from the actual drift output value y to obtain the temperature drift control value of the fiber optic gyroscope, subtracting the temperature drift control value from the actual drift output value y of the fiber optic gyroscope, and realizing temperature drift control.

[0020] A further technical solution is to set the weights of the temperature control box temperature T1, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope to 1 / 3 in the initial state, and start adaptive reinforcement learning.

[0021] A further technical solution is that the adaptive reinforcement learning algorithm is an adaptive self-reinforced random learning algorithm, which can update weights while learning without any prior knowledge, and combine multiple weak learning effects in a weighted manner, and reinforce learning multiple times in a cycle, so as to enhance the prediction accuracy and achieve the optimal solution.

[0022] A fiber optic gyroscope temperature drift control device comprises a fiber optic gyroscope temperature drift control module, which is a program module, and is used for adjusting the temperature T1 of a temperature control box between -40°C and 60°C, and obtaining in real time the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope, and then obtaining a learning sample {T1, T2, T3, y}, and inputting the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value of the fiber optic gyroscope, and obtaining a temperature drift control value of the fiber optic gyroscope by subtracting the predicted drift output value of the fiber optic gyroscope from an actual drift output value y, and subtracting the temperature drift control value from the actual drift output value y of the fiber optic gyroscope to achieve temperature drift control.

[0023] A further technical solution is that it also includes a fiber optic gyroscope, a temperature control box, a host computer, a first temperature sensor for obtaining the temperature of the temperature control box, a second temperature sensor for obtaining the inner wall temperature of the fiber optic gyroscope, and a third temperature sensor for obtaining the outer wall temperature of the fiber optic gyroscope. The host computer is separately connected to and communicates with the temperature control box, the fiber optic gyroscope, the first temperature sensor, the second temperature sensor, and the third temperature sensor.

[0024] A temperature drift control device for an optical fiber gyroscope comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above corresponding steps are realized.

[0025] A fiber optic gyroscope temperature drift control device includes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above corresponding steps are implemented.

[0026] The beneficial effects of adopting the above technical solution are:

[0027] First, a method for controlling temperature drift of a fiber optic gyroscope comprises the following steps: adjusting the temperature T1 of a temperature control box between -40°C and 60°C, obtaining in real time the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope, and then obtaining a learning sample {T1, T2, T3, y}, inputting the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value of the fiber optic gyroscope, calculating the difference between the predicted drift output value of the fiber optic gyroscope and the actual drift output value y to obtain a temperature drift control value of the fiber optic gyroscope, subtracting the temperature drift control value from the actual drift output value y of the fiber optic gyroscope, and realizing temperature drift control. This technical solution constructs a learning sample from the actual temperature of the temperature control box, the inner and outer wall temperatures of the fiber optic gyroscope, and the drift output value, inputs the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value, and calculates the difference between the predicted drift output value and the actual drift output value to obtain a temperature drift control value, thereby improving the drift of the fiber optic gyroscope with low time cost.

[0028] Second, a fiber optic gyroscope temperature drift control device includes a fiber optic gyroscope temperature drift control module, which is a program module used to adjust the temperature T1 of the temperature control box between -40°C and 60°C, and obtain the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope in real time, and then obtain the learning sample {T1, T2, T3, y}, input the learning sample into the adaptive reinforcement learning algorithm to obtain the predicted drift output value of the fiber optic gyroscope, and obtain the temperature drift control value of the fiber optic gyroscope by taking the difference between the predicted drift output value of the fiber optic gyroscope and the actual drift output value y, and subtract the temperature drift control value from the actual drift output value y of the fiber optic gyroscope to achieve temperature drift control. This technical solution, through the fiber optic gyroscope temperature drift control module, etc., can improve the drift of the fiber optic gyroscope with low time cost.

[0029] Please refer to the detailed description of the specific implementation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is the principle block diagram of the experimental device for this application;

[0031] Figure 2 This is the principle block diagram of the adaptive self-enhanced random learning algorithm;

[0032] Figure 3 This is a flow chart of temperature drift control of the fiber optic gyroscope of this application;

[0033] Figure 4 This is a curve diagram of the temperature drift control error value in this application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present application and its application or use. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0036] Embodiment 1:

[0037] The present application discloses a method for controlling temperature drift of a fiber optic gyroscope, comprising the following steps:

[0038] Fiber optic gyroscope temperature drift control steps:

[0039] The temperature T1 of the temperature control box is adjusted between -40°C and 60°C, and the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope are obtained in real time.

[0040] The learning samples {T1, T2, T3, y} are obtained. In the initial state, the weights of the temperature control box temperature T1, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope are all set to 1 / 3, and adaptive reinforcement learning is started.

[0041] The learning samples are input into the adaptive reinforcement learning algorithm to obtain the predicted drift output value of the fiber optic gyroscope and update it.

[0042] The predicted drift output value of the fiber optic gyroscope is subtracted from the actual drift output value y to obtain the temperature drift control value of the fiber optic gyroscope, and the temperature drift control value is subtracted from the actual drift output value y of the fiber optic gyroscope to achieve temperature drift control.

[0043] Embodiment 2:

[0044] The present application discloses a fiber optic gyroscope temperature drift control device, which includes a fiber optic gyroscope temperature drift control module. The fiber optic gyroscope temperature drift control module is a program module.

[0045] The fiber optic gyroscope temperature drift control module is used to adjust the temperature T1 of the temperature control box between -40°C and 60°C, obtain the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope in real time, and then obtain the learning sample {T1, T2, T3, y}, input the learning sample into the adaptive reinforcement learning algorithm to obtain the predicted drift output value of the fiber optic gyroscope, calculate the difference between the predicted drift output value of the fiber optic gyroscope and the actual drift output value y to obtain the temperature drift control value of the fiber optic gyroscope, subtract the temperature drift control value from the actual drift output value y of the fiber optic gyroscope, and realize temperature drift control.

[0046] Embodiment 3:

[0047] On the basis of Example 2, the present application discloses a fiber optic gyroscope temperature drift control device including a fiber optic gyroscope, a temperature control box, a host computer, a first temperature sensor for obtaining the temperature of the temperature control box, a second temperature sensor for obtaining the inner wall temperature of the fiber optic gyroscope, and a third temperature sensor for obtaining the outer wall temperature of the fiber optic gyroscope, as well as a fiber optic gyroscope temperature drift control module, wherein the host computer is separately connected to and communicates with the temperature control box, the fiber optic gyroscope, the first temperature sensor, the second temperature sensor, and the third temperature sensor.

[0048] The fiber optic gyroscope temperature drift control module is used for the host computer to control the temperature T1 of the temperature control box to be adjusted between -40°C and 60°C. The host computer obtains the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope in real time, and then obtains the learning sample {T1, T2, T3, y}, inputs the learning sample into the adaptive reinforcement learning algorithm to obtain the predicted drift output value of the fiber optic gyroscope, calculates the difference between the predicted drift output value of the fiber optic gyroscope and the actual drift output value y to obtain the temperature drift control value of the fiber optic gyroscope, subtracts the temperature drift control value from the actual drift output value y of the fiber optic gyroscope, and realizes temperature drift control.

[0049] Embodiment 4:

[0050] The present application discloses a temperature drift control device for a fiber optic gyroscope, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of Example 1 are implemented.

[0051] Embodiment 5:

[0052] The present application discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in Embodiment 1 are implemented.

[0053] brief:

[0054] The present application discloses a novel method for controlling temperature drift of fiber optic gyroscope, which comprises the following steps: setting the temperature of the inner and outer walls of the fiber optic gyroscope and the temperature control box as the temperature monitoring point, and placing the fiber optic gyroscope in a temperature control box with a variable temperature of -40°C to 60°C for operation; collecting the temperature of all monitoring points and the drift output of the fiber optic gyroscope at regular intervals, and constructing an initial learning sample and a prediction model; transmitting the learning sample to the machine learning algorithm, and updating the existing learning results immediately when new data comes in to form a new sample set, and calculating the output value of the fiber optic gyroscope through the constructed compensation model by using the obtained sample data; calculating the difference between the calculated prediction value and the actual output value to obtain the temperature drift control value of the fiber optic gyroscope, and then feeding back the control value to the fiber optic gyroscope to implement temperature drift control. The present application is simple in design and has few parameters, but can effectively control the temperature drift of the fiber optic gyroscope and improve its measurement accuracy, and has practical value for the engineering of medium and low precision fiber optic gyroscopes.

[0055] Technical Contribution:

[0056] This application adopts the online learning, adaptive, and self-enhancement mode of a machine learning algorithm to provide a method for controlling the temperature drift of a fiber optic gyroscope, which can help improve the measurement accuracy of the fiber optic gyroscope.

[0057] 1. A novel method for controlling temperature drift of a fiber optic gyroscope, characterized by comprising the following steps:

[0058] Step 1 Start the fiber optic gyroscope

[0059] The fiber optic gyroscope is placed in a temperature control box for operation, and the temperature monitoring points of the fiber optic gyroscope are set. The temperature of the temperature control box can be adjusted between [-40℃, 60℃], and the temperature monitoring points include the temperature of the temperature control box, the inner wall temperature of the fiber optic gyroscope, and the outer wall temperature of the fiber optic gyroscope.

[0060] Step 2: Construct learning samples and control models

[0061] At regular intervals, the temperature of the temperature control box T1, the temperatures of the inner and outer walls of the fiber optic gyroscope T2 and T3, and the drift output y of the fiber optic gyroscope mentioned in step 1 are collected to construct learning samples {T1, T2, T3, y}, and the prediction output model is set to f(t).

[0062] Step 3 Update learning samples and predict output values ​​through machine learning

[0063] The learning samples are transferred to the machine learning algorithm, and the existing learning results are updated immediately when new data comes in to form a new sample set. The obtained sample data is used to calculate the output temperature value of the fiber optic gyroscope through the constructed compensation model.

[0064] Step 4: Get the temperature drift control value of the fiber optic gyroscope

[0065] The temperature drift control value of the fiber optic gyroscope can be obtained by subtracting the predicted value obtained in step 3 from the actual output value. The temperature drift control of the fiber optic gyroscope can be achieved by subtracting this control value from the actual output value of the fiber optic gyroscope.

[0066] 2. The three monitoring points are evenly distributed in space.

[0067] 3. The weights of the learning samples T1, T2, and T3 are ω = (ω1, ω2, ω3). In the initial state, let The initial output prediction value f(t) is obtained by formula 1.

[0068]

[0069] Note: In Formula 1, t represents time in seconds.

[0070] T1, T2, T3 represent temperature, unit is ℃.

[0071] After that, samples will be collected and updated regularly. Each sample will first enter several learning machines. Learning, its weights will be adjusted adaptively, and the weight θ of each independent learner is calculated by formula 2.

[0072]

[0073] After the learner is trained, the sample weight ω is adjusted by formula 3.

[0074]

[0075] The predicted FOG output value is calculated by Equation 4.

[0076]

[0077] Note: In Formula 4, f(t) represents the output of the fiber optic gyroscope at the current acquisition time, and f(t-1) represents the output of the fiber optic gyroscope at the previous acquisition time.

[0078] The temperature drift control value of the fiber optic gyroscope is the difference between the calculated predicted value and the actual output value, as shown in formula 5, T i That is, the temperature drift control value of the fiber optic gyroscope.

[0079] Ti =f(t i )-y i Formula 5

[0080] 4. The machine learning algorithm is characterized in that the online learning algorithm used is specifically called adaptive self-reinforced random learning. This algorithm can update weights while learning without any prior knowledge, and combine multiple weak learning effects in a weighted manner, and strengthen learning multiple times in a cycle, so as to enhance the prediction accuracy and achieve the optimal solution.

[0081] Technical contribution description:

[0082] The present application adopts an adaptive self-enhanced random learning mode, aiming to provide a method for controlling the temperature drift of a fiber optic gyroscope.

[0083] The technical solution of this application is described as follows.

[0084] A method for controlling temperature drift of an optical fiber gyroscope, the method comprising the following steps:

[0085] Step 1. Place the fiber optic gyroscope in a temperature control box with a variable temperature of -40℃ to 60℃, and set the temperature monitoring points of the fiber optic gyroscope in advance. Three monitoring points are selected in this application, namely the temperature of the temperature control box, the inner wall temperature of the fiber optic gyroscope, and the outer wall temperature of the fiber optic gyroscope, and these three monitoring points need to be evenly distributed in space.

[0086] Step 2. Collect the temperature of the temperature control box T1, the inner wall temperature T2, the outer wall temperature T3 of the fiber optic gyroscope, and the real-time output value y of the fiber optic gyroscope mentioned in step 1 at regular intervals, and construct a learning sample set {T1, T2, T3, y} to predict the output value f(t).

[0087] Among them, ω = (ω1, ω2, ω3) is called the weight of the sample {T1, T2, T3}. When learning for the first time, The initial output prediction value f(t) is obtained by formula 1

[0088]

[0089] Step 3. Transfer the learning samples to the learning algorithm, and update the existing learning results immediately when new data comes in to form a new learning sample set; use the machine to calculate the output prediction value of the fiber optic gyroscope using the obtained sample data.

[0090] Each time the sample is updated, the weights will be adaptively adjusted. The weights θ of each independent learner are calculated by formula 2

[0091]

[0092] Then the sample weight ω will also be updated, which is calculated by formula 3

[0093]

[0094] After each round of weight update, the predicted FOG output value is calculated by Equation 4.

[0095]

[0096] Step 4. Difference the predicted value calculated in step 3 with the actual output value to obtain the temperature drift control value of the fiber optic gyroscope. As shown in formula 5, T i That is, the temperature drift control value of the fiber optic gyroscope.

[0097] T i =f(t i )-y i Formula 5

[0098] Note:

[0099] 1) In formula 1, t represents time in seconds; T1, T2, T3 represent temperature in °C.

[0100] 2) In formula 4, f(t) represents the output of the fiber optic gyroscope at the current acquisition time, and f(t-1) represents the output of the fiber optic gyroscope at the previous acquisition time.

[0101] Technical solution description:

[0102] like Figure 1 As shown in the figure, it is a schematic diagram of the experimental device of this application. The temperature drift control method adopted in this example is: firstly, the temperature monitoring point of the fiber optic gyroscope is set, the fiber optic gyroscope is fixed on a turntable, and then placed in a temperature control box with a variable temperature of -40℃ to 60℃ for operation.

[0103] The small single-axis electric rate position turntable model used in this embodiment is SST300, the temperature control box is an ETOMA test box that can quickly change high and low temperatures, the fiber optic gyroscope is a medium and low precision single-axis fiber optic gyroscope FOGS170M, and the three selected monitoring points are the temperature of the temperature control box, the inner wall temperature of the fiber optic gyroscope, and the outer wall temperature of the fiber optic gyroscope, and these three monitoring points need to be evenly distributed in space.

[0104] Next, we will proceed to specific data collection and calculation. The temperature of the temperature control box, the inner wall temperature of the fiber optic gyroscope, and the outer wall temperature of the fiber optic gyroscope are set to T1, T2, and T3 respectively, and the output value of the fiber optic gyroscope is set to y, thereby constructing the learning sample {T1, T2, T3, y}, and setting the output model to f(t).

[0105] The learning samples are transferred to the learning algorithm, and the existing learning results are updated immediately when new data comes in to form new learning samples.

[0106] like Figure 2 As shown in the figure, the specific learning algorithm process is described in detail as follows:

[0107] (1) Since there is no prior data, the weights of samples T1, T2, and T3 are all taken as 1 / 3, and the initial output prediction value f(t) is determined according to Formula 1.

[0108]

[0109] (2) After the start of operation, the sample data will be updated according to the algorithm. Each updated data is mainly regulated by weights. The weight θ of each independent learner is calculated by Formula 2.

[0110]

[0111] (3) Then the sample weight ω will also be updated through the strong learner, which is calculated by Formula 3.

[0112]

[0113] (4) After each round of weight update, the predicted FOG output value is calculated by Equation 4.

[0114]

[0115] (5) The difference between the predicted value and the actual output value is calculated to obtain the temperature drift control value of the fiber optic gyroscope. As shown in formula 5, Ti is the temperature drift control value of the fiber optic gyroscope.

[0116] T i =f(t i )-y i Formula 5

[0117] Finally, the obtained control value is fed back to the fiber optic gyroscope to control its temperature drift.

[0118] like Figure 3 As shown, such a cyclic execution constitutes a method for controlling temperature drift of a fiber optic gyroscope.

[0119] In this embodiment, the data collection speed is 5 seconds / time, the temperature changes from -40°C to 60°C, and the temperature change rate is random, mainly distributed in [5°C / min, 15°C / min].

[0120] like Figure 4As shown in the figure, the error result after 10 minutes of operation, the horizontal axis is the time value, and the vertical axis is the output control error of the fiber optic gyroscope. Among them, the output value of the fiber optic gyroscope fluctuates during the operation of the turntable and the temperature change. Overall, the machine learning algorithm used in this application is simple and easy, but the error can be controlled within ±0.1° / h, which is an economical and feasible solution for improving the accuracy of low-end fiber optic gyroscopes.

[0121] The fiber optic gyroscope temperature drift control method involved in this embodiment mainly uses an adaptive self-enhanced machine learning algorithm, which can effectively control temperature drift. The algorithm is simple and low-cost, and is suitable for use in the engineering of mid- and low-end fiber optic gyroscopes.

[0122] After this application was run internally for a period of time, the on-site technicians reported that the application was beneficial in the following ways:

[0123] Compared with the current existing technology, the substantial progress of this application is reflected in:

[0124] The model constructed in this application has a simple design concept, few parameters, and the algorithm is mainly based on linear calculation, with low time cost and no increase in order.

[0125] The temperature compensation model in the present application changes with the change of weights, thereby avoiding the immutability of the existing model and increasing the flexibility and adaptability of the model.

[0126] The machine learning algorithm used in this application is adaptive self-enhanced random learning, which can prevent the model from being affected by noise in the training samples, significantly reduce the overfitting phenomenon that often occurs in the existing optical gyroscope temperature compensation model, and enhance the robustness of the model.

[0127] Experiments have shown that the temperature control model constructed by the adaptive self-enhanced random online learning method is simple to calculate and fast to run, and the compensation accuracy achieved is significantly improved.

[0128] The application method is easy to implement in engineering and can improve the adaptability of the fiber optic gyroscope to ambient temperature and its accuracy in practical applications.

Claims

1. A method for controlling temperature drift of a fiber optic gyroscope, characterized in that: The method comprises the following steps: adjusting the temperature T1 of the temperature control box between -40°C and 60°C, obtaining the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope in real time, and then obtaining the learning sample {T1, T2, T3, y}, inputting the learning sample into the adaptive reinforcement learning algorithm to obtain the predicted drift output value of the fiber optic gyroscope, calculating the difference between the predicted drift output value of the fiber optic gyroscope and the actual drift output value y to obtain the temperature drift control value of the fiber optic gyroscope, subtracting the temperature drift control value from the actual drift output value y of the fiber optic gyroscope, and realizing temperature drift control; in the initial state, setting the weights of the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope to 1 / 3, and starting the adaptive reinforcement learning; The adaptive reinforcement learning algorithm is an adaptive self-reinforced random learning algorithm. Without any prior knowledge, the algorithm can update weights while learning, and combine multiple weak learning effects in a weighted manner, and strengthen learning multiple times in a cycle to enhance prediction accuracy and achieve optimal solution. The weights of the learning samples T1, T2, and T3 are ω = (ω1, ω2, ω3). In the initial state, let The initial output prediction value f(t) is obtained by formula 1; In formula 1, f(t) is the initial predicted output value; t represents time, in seconds; T1 is the temperature of the temperature control box, T2 is the inner wall temperature of the fiber optic gyroscope, and T3 is the outer wall temperature, in °C; After that, samples will be collected and updated regularly. Each sample will first enter several learning machines. Learning, the sample weights will be adaptively adjusted, and the weight of each independent learner is calculated by Equation 2; In formula 2, θ j is the weight of each independent learner; After the learner is trained, the sample weight ω will also be updated and adjusted by formula 3; In formula 3, ω i is the sample weight; The predicted FOG output value is calculated by Equation 4; In formula 4, f(t) represents the output of the fiber optic gyroscope at the current acquisition time, and f(t-1) represents the output of the fiber optic gyroscope at the previous acquisition time; The temperature drift control value of the fiber optic gyroscope is the difference between the calculated predicted value and the actual output value, as shown in Equation 5; T i = f(t i )-y i formula 5 In formula 5, T i That is, the temperature drift control value of the fiber optic gyroscope.

2. A fiber optic gyroscope temperature drift control device, used in the fiber optic gyroscope temperature drift control method according to claim 1, characterized in that: The invention comprises a fiber optic gyroscope temperature drift control module, which is a program module, and is used for adjusting the temperature T1 of a temperature control box between -40°C and 60°C, and obtaining the temperature T1 of the temperature control box, the inner wall temperature T2 and the outer wall temperature T3 of the fiber optic gyroscope when the fiber optic gyroscope is in the temperature control box, and the drift output y of the fiber optic gyroscope in real time, and then obtaining a learning sample {T1, T2, T3, y}, inputting the learning sample into an adaptive reinforcement learning algorithm to obtain a predicted drift output value of the fiber optic gyroscope, and comparing the predicted drift output value of the fiber optic gyroscope with the predicted drift output value of the fiber optic gyroscope. The temperature drift control value of the fiber optic gyroscope is obtained by taking the difference of the actual drift output value y, and the temperature drift control value is subtracted from the actual drift output value y of the fiber optic gyroscope to realize temperature drift control; it also includes a fiber optic gyroscope, a temperature control box, a host computer, a first temperature sensor for obtaining the temperature of the temperature control box, a second temperature sensor for obtaining the inner wall temperature of the fiber optic gyroscope, and a third temperature sensor for obtaining the outer wall temperature of the fiber optic gyroscope, and the host computer is separately connected to and communicates with the temperature control box, the fiber optic gyroscope, the first temperature sensor, the second temperature sensor, and the third temperature sensor.

3. A temperature drift control device for a fiber optic gyroscope, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the corresponding steps in the method for controlling temperature drift of an optical fiber gyroscope according to claim 1 are implemented.

4. A fiber optic gyroscope temperature drift control device, comprising a computer readable storage medium, wherein the computer readable storage medium stores a computer program, characterized in that: When the computer program is executed by the processor, the corresponding steps in the method for controlling temperature drift of an optical fiber gyroscope according to claim 1 are implemented.

Images