Fiber-optic gyroscope drift compensation method and system fusing lms adaptive filtering

Abstract

The application provides a fiber-optic gyroscope drift compensation method and system fusing LMS adaptive filtering, and relates to the technical field of LMS adaptive filtering, wherein the method comprises the following steps: synchronously collecting multi-point temperature data, temperature change rate, axial and radial gradient information and original output signals of a gyroscope shell, separating the original signals into temperature drift components and non-temperature noise components through variational mode decomposition and filtering out the noise; then, a non-linear mapping relationship is established by combining the temperature field data and the drift components to generate a drift prediction value, the residual error of the prediction value and an actual drift component is input into an LMS adaptive filter for dynamic error compensation; finally, the compensation result and the prediction value are superimposed to output a corrected gyroscope signal. The application improves the drift compensation precision and stability of the fiber-optic gyroscope under a complex variable temperature environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of LMS adaptive filtering, and in particular to a fiber-optic gyroscope drift compensation method and system fusing LMS adaptive filtering. BACKGROUND

[0002] Under complex variable-temperature conditions, such as rapid start-stop of aerospace equipment or sudden temperature difference of field operation equipment, the output signal of the fiber-optic gyroscope is easily affected by non-uniform distribution of the temperature field, resulting in increased drift error. The traditional single-point temperature compensation method is difficult to cope with temperature gradient changes and dynamic thermal shock, and there is an urgent need for a high-precision compensation technology that can combine multi-dimensional temperature field characteristics and dynamically correct drift.

[0003] The current more advanced scheme adopts a temperature drift compensation method based on support vector regression, which measures single-point temperature data of the gyroscope shell and its historical change trend, establishes a static mapping model of temperature and drift error. This method introduces the temperature change rate as an auxiliary feature, uses a kernel function to fit the nonlinear relationship between temperature and drift, and optimizes the model parameters in the calibration stage to improve the compensation accuracy.

[0004] This scheme relies on single-point temperature data and has limited ability to represent the spatial distribution characteristics of the temperature field. When there is an axial or radial temperature gradient, the compensation residual still has large fluctuations. The static model parameters are difficult to adapt to the dynamic nonlinear error in the rapid temperature change process, resulting in obvious compensation lag, especially in the temperature mutation stage, the error correction effect decreases. SUMMARY

[0005] The present application provides a fiber-optic gyroscope drift compensation method and system fusing LMS adaptive filtering, to solve the problems of low drift compensation precision and poor stability of the fiber-optic gyroscope under complex variable-temperature environments in the prior art.

[0006] To solve the above technical problems, in a first aspect, the present application provides a fiber-optic gyroscope drift compensation method fusing LMS adaptive filtering, comprising:

[0007] Collecting multi-point temperature measurement data, temperature change rate data, axial and radial gradient information of the fiber-optic gyroscope, and original output signal of the gyroscope;

[0008] Performing variational mode decomposition on the original output signal of the gyroscope to generate a temperature drift component and a non-temperature noise component, and filtering out the non-temperature noise component;

[0009] Based on the multi-point temperature measurement data, the temperature change rate data, the axial and radial gradient information, and the temperature drift component, a nonlinear mapping relationship is established;

[0010] Based on the nonlinear mapping relationship, a drift prediction value is generated, the drift prediction value is associated with the temperature drift component, and the association result is input into an LMS adaptive filter for dynamic error compensation.

[0011] The compensation result is superimposed with the drift prediction value to generate a compensated gyro output signal.

[0012] Optionally, the association of the drift prediction value with the temperature drift component and the input of the association result into the LMS adaptive filter for dynamic error compensation comprises:

[0013] The drift prediction value is subtracted from the temperature drift component to generate a residual sequence;

[0014] The residual sequence is input into the LMS adaptive filter, and based on a statistical feature set of the residual sequence, an adjustment factor of the LMS adaptive filter is dynamically updated;

[0015] An approximation operation is performed on the residual sequence by using the updated adjustment factor to generate a target compensation amount, and the target compensation amount is a compensation result.

[0016] Optionally, the adjustment factor of the LMS adaptive filter comprises an update step factor, a direction memory factor, and a memory depth factor;

[0017] Based on the statistical feature set of the residual sequence, the adjustment factor of the LMS adaptive filter is dynamically updated, comprising:

[0018] The residual sequence is time-domain segmented by the LMS adaptive filter to obtain a plurality of residual sub-sequences;

[0019] For each time-domain segment, a statistical feature set is extracted from the corresponding residual sub-sequence, and the statistical feature set comprises a fluctuation intensity value, a change trend value, and a correlation decay rate;

[0020] When the fluctuation intensity value is greater than a preset first threshold value, the update step factor of the corresponding time-domain segment is increased;

[0021] When the change trend value is greater than a preset second threshold value, the direction memory factor of the corresponding time-domain segment is activated;

[0022] According to the correlation decay rate, the memory depth factor of the corresponding time-domain segment is dynamically adjusted.

[0023] Optionally, the approximation operation on the residual sequence by using the updated adjustment factor to generate a target compensation amount comprises:

[0024] With the updated adjustment factor, each data point in the residual sequence is traversed in chronological order, and the initial compensation amount corresponding to each data point is calculated;

[0025] The initial compensation increment at the current time is recursively superimposed on the initial compensation amount at the previous time to generate an intermediate compensation amount.

[0026] The intermediate compensation amount is subjected to boundary constraint processing to generate a target compensation amount.

[0027] Optionally, the nonlinear mapping relationship is established based on the multi-point temperature measurement data, the temperature rate of change data, the axial-radial gradient information, and the temperature drift component, including:

[0028] The axial-radial gradient information is subjected to spatial interpolation processing to generate temperature gradient distribution data.

[0029] The multi-point temperature measurement data, the temperature rate of change data, and the temperature gradient distribution data are aligned according to the time stamp to construct multi-dimensional temperature field data, including a temperature measurement data matrix, a temperature rate of change vector, and a temperature gradient tensor.

[0030] The spatial coupling term of the temperature measurement data matrix and the temperature gradient tensor is calculated, and the dynamic coupling term of the temperature rate of change vector and the temperature gradient tensor is calculated.

[0031] The spatial coupling term and the dynamic coupling term are fused according to a preset proportion to generate an interaction factor, and an enhanced temperature field feature is generated based on the interaction factor.

[0032] The enhanced temperature field feature is taken as a mapping source, and the temperature drift component corresponding to the time stamp is taken as a mapping target, and a nonlinear mapping relationship from the mapping source to the mapping target is established through an iterative optimization process.

[0033] Optionally, the variational modal decomposition of the gyro original output signal generates a temperature drift component and a non-temperature noise component, including:

[0034] The variational modal decomposition operation is performed on the gyro original output signal to obtain a plurality of intrinsic modal components.

[0035] For each intrinsic modal component, a correlation coefficient of the intrinsic modal component and the temperature rate of change data is calculated.

[0036] From all the intrinsic modal components, the key intrinsic modal components whose correlation coefficients exceed a preset correlation threshold are screened out, all the key intrinsic modal components are combined into a temperature drift component, and the remaining intrinsic modal components are classified as non-temperature noise components.

[0037] Optionally, the superimposing the compensation result with the drift prediction value to generate a compensated gyro output signal comprises:

[0038] Performing time domain smoothing processing on the compensation result to generate a stable compensation amount;

[0039] Performing algebraic addition of the stable compensation amount and the drift prediction value to generate a modified drift amount;

[0040] Performing amplitude limiting processing on the modified drift amount;

[0041] Removing the amplitude limiting processed modified drift amount from the gyro original output signal to obtain a compensated gyro output signal.

[0042] In a second aspect, the present application provides a fiber-optic gyroscope drift compensation system fused with LMS adaptive filtering, comprising:

[0043] A collection module configured to collect multi-point temperature measurement data, temperature rate of change data, axial and radial gradient information of a fiber-optic gyroscope, and a gyro original output signal;

[0044] A decomposition module configured to perform variational mode decomposition on the gyro original output signal to generate a temperature drift component and a non-temperature noise component, and filter out the non-temperature noise component;

[0045] An establishment module configured to establish a nonlinear mapping relationship based on the multi-point temperature measurement data, the temperature rate of change data, the axial and radial gradient information, and the temperature drift component;

[0046] A generation module configured to generate a drift prediction value based on the nonlinear mapping relationship, associate the drift prediction value with the temperature drift component, and input an association result into an LMS adaptive filter to perform dynamic error compensation;

[0047] A superimposition module configured to superimpose a compensation result with the drift prediction value to generate a compensated gyro output signal.

[0048] In a third aspect, the present application provides an electronic device, comprising:

[0049] A memory configured to store a computer program;

[0050] A processor configured to execute the computer program to implement the steps of the fiber-optic gyroscope drift compensation method fused with LMS adaptive filtering according to the first aspect.

[0051] In a fourth aspect, the present application provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and the computer program, when executed by a processor, can implement the steps of the fiber-optic gyroscope drift compensation method based on fusion LMS adaptive filtering according to the first aspect.

[0052] In the present application, a fiber-optic gyroscope drift compensation method based on fusion LMS adaptive filtering is provided, which comprises the following steps: collecting multi-point temperature measurement data, temperature rate of change data, axial and radial gradient information of the fiber-optic gyroscope, and original output signals of the gyroscope; performing variational mode decomposition on the original output signals of the gyroscope to generate temperature drift components and non-temperature noise components, and filtering out the non-temperature noise components; establishing a non-linear mapping relationship based on the multi-point temperature measurement data, the temperature rate of change data, the axial and radial gradient information, and the temperature drift components; generating a drift prediction value based on the non-linear mapping relationship, correlating the drift prediction value with the temperature drift components, and inputting a correlation result into an LMS adaptive filter for dynamic error compensation; and superimposing a compensation result with the drift prediction value to generate a compensated gyroscope output signal.

[0053] The technical scheme provided by the present application has the following beneficial effects:

[0054] In the present application, by collecting multi-point temperature measurement data, temperature rate of change data, axial and radial gradient information, and original output signals of the gyroscope, original data reflecting spatial distribution characteristics and dynamic change characteristics of a temperature field are comprehensively obtained, thereby establishing a data basis for subsequent accurate compensation. Signal components strongly related to temperature changes are effectively separated, and the influence of non-temperature interference factors such as vibration on compensation accuracy is excluded. The complex non-linear relationship between the spatial gradient of the temperature field and the drift is accurately described, and the prediction accuracy of the model is improved. Residual errors not fitted by the model are corrected in real time to adapt to the dynamic compensation demand in the case of rapid temperature changes. The combination of static prediction and dynamic compensation is realized, and the stability of the output signal is ensured.

[0055] Further, in the present application, the residual sequence of the drift prediction value and the actual temperature drift component is calculated, the LMS adaptive filter is used to analyze the statistical characteristics of the residual to dynamically adjust the filter parameters, and the accurate compensation amount is generated by approximation operation on the residual based on the updated parameters, thereby realizing intelligent tracking and real-time correction of the model fitting errors.

[0056] Moreover, the technical effect is embodied in the ability to adaptively eliminate dynamic errors caused by rapid temperature changes, effectively suppress compensation lag, and improve the stability and reliability of the gyroscope output in a complex variable temperature environment.

[0057] These and other aspects of the present application will become more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the accompanying drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0059] Figure 1 A flow chart of a fiber optic gyroscope drift compensation method fusing LMS adaptive filtering provided by an embodiment of the present application;

[0060] Figure 2 A specific implementation schematic diagram of a fiber optic gyroscope drift compensation method fusing LMS adaptive filtering provided by an embodiment of the present application;

[0061] Figure 3 Another specific implementation schematic diagram of a fiber optic gyroscope drift compensation method fusing LMS adaptive filtering provided by an embodiment of the present application;

[0062] Figure 4 A structural schematic diagram of a fiber optic gyroscope drift compensation system fusing LMS adaptive filtering provided by an embodiment of the present application. DETAILED DESCRIPTION

[0063] The existing fiber optic gyroscope temperature drift compensation method mainly relies on single-point temperature measurement data, predicts the drift error through a static model and corrects it. This method is still applicable in the scene where the temperature distribution is uniform and changes slowly, but in actual application, especially in the environment where the temperature changes rapidly or there is obvious temperature gradient, the single-point temperature data is difficult to fully reflect the real thermal state of the gyroscope. The static model cannot dynamically adjust the compensation parameters, resulting in compensation lag when the temperature suddenly changes, increasing the residual error, and affecting the stability and precision of the gyroscope output.

[0064] In view of the above problems, the present application proposes a fiber optic gyroscope drift compensation method fusing LMS adaptive filtering. This method constructs an input feature that can represent the spatial distribution characteristics of the temperature field through multi-point temperature measurement data, temperature change rate and axial and radial gradient information, and extracts pure temperature drift components combined with signal decomposition technology. On this basis, a nonlinear mapping model is used to predict the drift trend, and adaptive filtering technology is used to correct the prediction residual in real time. This method not only can more comprehensively capture the dynamic change characteristics of the temperature field, but also can automatically adjust the compensation parameters when the temperature fluctuates rapidly, improve the precision and stability of the gyroscope output in complex variable temperature environment, and solve the compensation deficiency problem caused by relying on single-point data and static model in the prior art.

[0065] For those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0066] The core of the present application is to provide a kind of fusion LMS adaptive filtering fiber optic gyroscope drift compensation method, the flowchart of a specific embodiment of the present application is as shown in Figure 1 The method comprises:

[0067] Step 101: collect the multi-point temperature measurement data of the fiber optic gyroscope, the temperature rate data, the axial radial gradient information and the original output signal of the gyroscope.

[0068] In step 101, the multi-point temperature measurement data refers to the spatial temperature distribution information obtained by arranging temperature sensors at different positions of the gyroscope shell, which is used to characterize the non-uniform characteristics of the temperature field. The temperature rate data reflects the speed of temperature change of each temperature measuring point with time, which is used to capture the dynamic thermal shock effect. The axial and radial gradient information describes the steepness of temperature change along the axial and radial directions of the gyroscope, which quantifies the spatial difference of the temperature field. The original output signal of the gyroscope is the gyroscope angular velocity measurement value without compensation processing, which contains the true angular velocity and temperature drift error.

[0069] In the embodiments of the present application, a plurality of temperature sensors are arranged on the surface of the gyroscope shell according to a specific topological structure, and the sensor positions cover the key temperature change regions in the axial and radial directions; each sensor synchronously collects temperature data at a fixed sampling frequency, and records the corresponding time stamp; the temperature rate of each point is obtained by differentiating the temperature difference of adjacent sampling periods; the temperature gradients in the axial and radial directions are calculated based on the temperature data of adjacent sensors; the gyroscope angular velocity output signal is synchronously acquired through a high-speed data interface; finally, the time-aligned multi-point temperature data, temperature rate, gradient information and original gyroscope signal are packaged into a unified data frame.

[0070] For example, a certain vehicle-mounted fiber optic gyroscope measures the temperatures of four temperature sensors arranged on the surface of the shell as 25.0℃, 26.8℃, 28.5℃ and 30.2℃ respectively during the starting stage, calculates the temperature rates as 0.2℃ / s, 0.23℃ / s, 0.25℃ / s and 0.27℃ / s, the axial gradient as 0.5℃ / cm, the radial gradient as 0.6℃ / cm, and the synchronously collected original output of the gyroscope as 15.3° / h. The gradient value is calculated by dividing the temperature difference between adjacent sensors by the interval, such as the axial gradient=(30.2-25.0) / 10.4=0.5℃ / cm.

[0071] Step 102: performing variational mode decomposition on the gyro original output signal to generate a temperature drift component and a non-temperature noise component, and filtering out the non-temperature noise component.

[0072] In step 102, the variational mode decomposition is an adaptive signal decomposition method that separates a complex signal into components with different vibration characteristics. The temperature drift component refers to a signal component related to temperature change, and its change trend is highly synchronized with temperature data. The non-temperature noise component contains interference signals such as mechanical vibration that are not related to temperature.

[0073] In the embodiments of the present application, the gyro original output signal is processed by variational mode decomposition, and after a preset number of decomposition layers, a plurality of intrinsic mode components are obtained by iterative optimization. The correlation coefficients of each component and the temperature change rate data are calculated, and the key components with correlation coefficients exceeding a threshold value are selected. The key components are superimposed to form a temperature drift component, and the remaining components are classified as non-temperature noise. Digital filtering technology is used to filter out the non-temperature noise component and retain the pure temperature-related signal.

[0074] For example, a gyro signal of 15.3° / h is decomposed into 5 components with amplitudes of 6.2, 4.1, 2.8, 1.5, and 0.7° / h. The correlation coefficients of each component and the temperature change rate are calculated to be 0.62, 0.88, 0.91, 0.65, and 0.58. The 4.1 and 2.8 components with correlation coefficients greater than 0.8 are selected and combined into a temperature drift component of 6.9° / h, and the remaining components are combined into a noise component of 8.4° / h. The correlation coefficient calculation formula is wherein represents the correlation coefficient, represents the observation value of the i-th component signal (unit: ° / h), represents the average value of the component signal, represents the observation value of the i-th temperature change rate, represents the average value of the temperature change rate, represents the observation sample number. Step 103: establishing a non-linear mapping relationship based on the multi-point temperature measurement data, the temperature change rate data, the axial-radial gradient information, and the temperature drift component.

[0075] In step 103, the non-linear mapping relationship is a mathematical model that describes the complex relationship between the multi-dimensional temperature field characteristics and the temperature drift component.

[0076]

[0077] ​​In the embodiments of the present application, first, the axial and radial gradient information is subjected to spatial interpolation processing to generate continuous gradient distribution data; then, the temperature measurement data, the rate of change data and the gradient data are aligned in time sequence to construct multi-dimensional input features containing spatial characteristics and dynamic characteristics; then, the spatial coupling term of the temperature measurement matrix and the gradient tensor and the dynamic coupling term of the rate of change vector and the gradient tensor are calculated; finally, the two types of coupling terms are fused according to a preset proportion to generate enhanced features, and a nonlinear mapping relationship with the temperature drift component is established through iterative optimization.

[0078] For example, the temperature data [25.0, 26.8, 28.5, 30.2] °C, the rate of change [0.2, 0.23, 0.25, 0.27] °C / s and the gradient data [[0.5, 0.6], [0.55, 0.65]] °C / cm are constructed to construct multi-dimensional features, the spatial coupling term 2.85 and the dynamic coupling term 0.145 are calculated, and the mapping relationship with the 6.9° / h drift component is established after fusion according to the 7:3 proportion.

[0079] Step 104: generating a drift prediction value based on the nonlinear mapping relationship, associating the drift prediction value with the temperature drift component, and inputting the association result into a least mean square adaptive filter (LMS) for dynamic error compensation.

[0080] In step 104, the drift prediction value is a temperature drift estimate calculated based on multi-dimensional temperature field data through a nonlinear mapping relationship, reflecting the expected gyro output drift size under the current temperature field condition, and the prediction value is used as a reference compensation amount for subsequent dynamic error correction. The LMS adaptive filter is a filter that can automatically adjust parameters, used for dynamic tracking of error changes.

[0081] In the embodiments of the present application, first, the difference between the drift prediction value and the actual temperature drift component is calculated to obtain a residual sequence; then, the residual sequence is subjected to time domain segmentation, and statistical features such as fluctuation intensity and change trend of each segment are extracted; the step factor, memory factor and other parameters of the filter are dynamically adjusted according to the characteristic values; finally, the residual sequence is filtered using the adjusted parameters to output a real-time updated compensation amount.

[0082] For example, the residual of the prediction value 2.1° / h and the actual value 6.9° / h is 4.8° / h, the fluctuation intensity is 0.05 after analysis, the step factor is adjusted to 0.015, and the compensation amount sequence 0.0054, 0.0099, 0.01341, 0.01657, 0.01891° / h is generated.

[0083] Step 105: superimposing the compensation result and the drift prediction value to generate a compensated gyro output signal.

[0084] In step 105, the compensated gyro output signal is the modified result of removing the temperature drift error from the original output, which eliminates the measurement deviation caused by temperature through superimposing static prediction compensation and dynamic residual compensation, and finally outputs an accurate signal closer to the true angular velocity.

[0085] In the embodiments of the present application, the compensation amount output by the LMS filter is algebraically added to the prediction value to obtain a modified value; then the modified value is amplitude-limited to ensure that it does not exceed the physical range of the gyro; and finally the modified value is subtracted from the original output signal to obtain the accurate compensated gyro output.

[0086] For example, the compensation amount 0.01891° / h is added to the prediction value 2.1° / h to obtain 2.11891° / h, which is limited to 2.12° / h and then subtracted from 15.3° / h, and finally an output of 13.18° / h is obtained.

[0087] This method effectively suppresses the gyro drift error in a complex variable temperature environment by combining multi-dimensional temperature field feature extraction, signal decomposition, and dynamic compensation, improves the stability and reliability of the output signal, and solves the problem of insufficient compensation of traditional methods in the case of rapid temperature change and gradient.

[0088] In order to solve the problem of insufficient adaptability of traditional temperature drift compensation methods in dynamic variable temperature environments, in some embodiments, step 104: the drift prediction value is associated with the temperature drift component, and the association result is input into an LMS adaptive filter for dynamic error compensation, as shown in Figure 2 , which includes:

[0089] Step 201: Subtract the drift prediction value from the temperature drift component to generate a residual sequence.

[0090] In step 201, the residual sequence is the difference sequence of the drift prediction value and the actual temperature drift component, which reflects the residual error that is not fitted by the nonlinear mapping model, and this sequence contains dynamic error characteristics when the temperature changes rapidly. The residual sequence is the association result.

[0091] In the embodiments of the present application, the drift prediction value calculated based on the temperature field data is subtracted from the actual temperature drift component obtained by signal decomposition at each time point to generate a residual data sequence containing time sequence characteristics, which provides error input for subsequent adaptive filtering.

[0092] Step 202: Input the residual sequence into an LMS adaptive filter, and dynamically update the adjustment factor of the LMS adaptive filter based on the statistical characteristics set of the residual sequence.

[0093] In step 202, the statistical feature set includes three feature indexes of fluctuation intensity, change trend and correlation attenuation extracted from the residual sequence, wherein the fluctuation intensity represents the intensity of error change, the change trend reflects the evolution direction of error, and the correlation attenuation describes the duration of error impact. The adjustment factor includes a step parameter for controlling the convergence speed, a memory parameter for keeping the consistency of compensation direction, and a length parameter for determining the range of historical data used.

[0094] In the embodiments of the present application, the residual sequence is first divided into several period sub-sequences, and three statistics of fluctuation range, change slope and autocorrelation characteristics are calculated for each sub-sequence. When the fluctuation exceeds the set standard, the step parameter is increased to speed up the response. When a sustained one-way change is detected, the memory parameter is enhanced to keep the compensation direction. The number of historical data used is dynamically adjusted according to the duration of error impact.

[0095] Step 203: using the updated adjustment factor, performing approximation operation on the residual sequence to generate a target compensation amount, wherein the target compensation amount is the compensation result.

[0096] In step 203, the approximation operation refers to the process of step-by-step correction of the residual sequence using the adjusted filter parameter. The target compensation amount is the final compensation value output after dynamic adjustment.

[0097] In the embodiments of the present application, each data point in the residual sequence is processed in time sequence. The basic compensation amount is calculated according to the current step parameter, the historical compensation amount is weighted and fused in combination with the memory parameter, and the target compensation amount matching the real-time error is finally output within the limited historical data range.

[0098] The following is a specific example:

[0099] In the temperature compensation process of vehicle-mounted fiber-optic gyroscope, based on the obtained drift prediction value 2.1 ° / h and the actual temperature drift component 6.9 ° / h, the difference 4.8 ° / h between the two is first calculated as the initial residual, and then the residual sequence is segmented to obtain 5 data points 4.8, 4.7, 4.5, 4.3, and 4.1. The statistical characteristics are calculated for the first segment residual 4.8, 4.7, and 4.5, wherein the fluctuation intensity is obtained by the standard deviation formula σ = sqrt[(4.8-4.667)²+(4.7-4.667)²+(4.5-4.667)²] / 3 = 0.152 ° / h, the change trend is calculated by the linear fitting slope b = -0.15 ° / h / point, and the correlation decay rate is calculated by the autocorrelation function as 0.92. Since the fluctuation intensity 0.152 exceeds the preset threshold 0.1, the step factor of the LMS filter is adjusted from 0.01 to 0.015; the absolute value of the change trend -0.15 exceeds the threshold 0.1, the direction memory factor is activated and set to 0.9; and the memory depth is set to 5 data points according to the correlation decay rate 0.92. The residual sequence is approximated using the adjusted parameters, and the first data point 4.8 ° / h is scaled by the step factor 0.015 to obtain the compensation increment 0.072 ° / h, and the previous historical compensation amount is 0.072 ° / h, so the output is directly 0.072 ° / h; the second data point 4.7 ° / h calculates the increment 0.0705 ° / h, and the previous compensation amount 0.072 ° / h and the memory factor 0.9 are combined to obtain the new compensation amount 0.9 x 0.072 + 0.0705 = 0.1353 ° / h; after processing all the residuals in turn, the compensation amount sequence [0.072, 0.1353, 0.1938, 0.2475, 0.2967] ° / h is generated. The sequence is superimposed with the prediction value 2.1 ° / h to obtain the correction value sequence [2.172, 2.2353, 2.2938, 2.3475, 2.3967] ° / h, which is limited to ensure that it does not exceed 2.4 ° / h, and finally the compensated output [13.128, 13.0647, 13.0062, 12.9525, 12.9033] ° / h is obtained from the original output 15.3 ° / h, completing the whole dynamic error compensation process. In the fluctuation intensity calculation, 4.667 is the average of the three residuals, the change trend slope wherein is the residual value, is the time point, and are the mean values, indicates the number of data points.

[0100] In the embodiments of the present application, the compensation parameters are adaptively adjusted by dynamically tracking the residual characteristics, so that the system can quickly respond to the error changes caused by temperature mutations, while maintaining the stability of the compensation process. The lag and mismatch problems of the traditional fixed parameter compensation method under variable temperature conditions are effectively solved, and the compensation effect under complex environments is improved.

[0101] To further improve the parameter adjustment accuracy of the LMS adaptive filter under dynamic variable temperature environment, in some embodiments, step 202: the adjustment factors of the LMS adaptive filter include an update step factor, a direction memory factor, and a memory depth factor.

[0102] Based on the statistical feature set of the residual sequence, the adjustment factors of the LMS adaptive filter are dynamically updated, such as Figure 3 As shown in the figure, it includes:

[0103] Step 301: Time domain segmentation of the residual sequence by the LMS adaptive filter to obtain a plurality of residual sub-sequences.

[0104] In step 301, the residual sub-sequence is a short period data segment divided by fixed time window from continuous residual data, each sub-sequence contains residual values of a plurality of continuous sampling points, used to reflect the error characteristics in local period.

[0105] In the embodiments of the present application, the entire residual sequence is divided by sliding according to the preset time window length, and part of the data is overlapped between adjacent sub-sequences to ensure continuity, obtaining a plurality of sub-sequence segments covering different time periods, providing analysis units for subsequent feature extraction.

[0106] Step 302: For each time domain segmentation, extract a statistical feature set from the corresponding residual sub-sequence, the statistical feature set includes fluctuation intensity value, change trend value and correlation decay rate.

[0107] In step 302, the fluctuation intensity value represents the degree of change of the residual in the sub-sequence, which is obtained by calculating the dispersion of the data points. The change trend value reflects the evolution direction of the residual, which is represented by fitting the slope. The correlation decay rate describes the duration of the residual influence, which is obtained by calculating the autocorrelation characteristics.

[0108] In the embodiments of the present application, the standard deviation of each residual sub-sequence is calculated as the fluctuation intensity, the least square method is used to fit the slope of the straight line as the change trend, and the correlation decline rate of adjacent data points is calculated by the autocorrelation function as the decay rate, forming three key indicators describing the residual characteristics of the time period.

[0109] Step 303: When the fluctuation intensity value is greater than a preset first threshold, the update step factor of the corresponding time domain segmentation is increased.

[0110] In step 303, the preset first threshold is a critical value for judging whether the residual fluctuation is intense. When the fluctuation intensity exceeds the value, it indicates that the temperature change is intense, and the step size factor needs to be increased to speed up the compensation response. Example: Set the first threshold to 0.1° / h. When the calculated fluctuation intensity of a certain residual sub-sequence is 0.15° / h (calculated by the standard deviation formula), the step size factor adjustment is triggered.

[0111] In the embodiments of the present application, when the fluctuation intensity of a certain sub-sequence exceeds the set threshold, the step size factor is gradually increased according to the preset adjustment rule, so that the filter can track the intense error change faster, and an upper limit is set to prevent excessive adjustment from causing oscillation. The specific implementation process is as follows: first, calculate the proportion of the fluctuation intensity exceeding the threshold, then find the corresponding adjustment multiple in the preset step size adjustment coefficient table according to the proportion, and finally multiply the current updated step size factor by the multiple to obtain the new step size factor value; for example, the current fluctuation intensity value is 0.05, which exceeds the first threshold 0.03, the exceeding proportion is 66%, and the adjustment multiple obtained by looking up the table is 1.2. The original step size factor 0.01 is adjusted to 0.01x1.2=0.012.

[0112] Step 304: When the change trend value is greater than a preset second threshold, the direction memory factor corresponding to the time domain segment is activated.

[0113] In step 304, the preset second threshold is a boundary value for determining whether the residual change trend is. Exceeding the value indicates that the error has a continuous one-way change trend, and the direction memory factor needs to be activated to keep the compensation direction stable. Example: Set the second threshold to 0.1° / h / point. When the change trend slope of the fitted residual sub-sequence is -0.15° / h / point, the direction memory factor is activated to 0.9.

[0114] In the embodiments of the present application, when the change trend of the sub-sequence is detected to exceed the threshold continuously, the direction memory function is activated, so that the current compensation direction inherits the main direction of the historical trend, reducing unnecessary direction adjustment.

[0115] Step 305: According to the correlation decay rate, dynamically adjust the memory depth factor corresponding to the time domain segment.

[0116] In the embodiments of the present application, the memory depth is dynamically adjusted according to the correlation decay rate. When the decay is fast, the amount of historical data used is reduced to improve the response speed, and when the decay is slow, the amount of historical data used is increased to ensure stability.

[0117] The following is a specific example:

[0118] In the temperature compensation process of the vehicle-mounted fiber-optic gyroscope, based on the residual sequence [4.8, 4.7, 4.5, 4.3, 4.1] ° / h obtained in the foregoing embodiment, first, time domain segmentation is performed with 3 data points as a window to obtain two overlapping subsequences [4.8, 4.7, 4.5] ° / h and [4.7, 4.5, 4.3] ° / h. For the first subsequence [4.8, 4.7, 4.5] ° / h, the average value is calculated as 4.667 ° / h, the fluctuation intensity is obtained as 0.152 ° / h by the formula σ = sqrt[(4.8-4.667)²+(4.7-4.667)²+(4.5-4.667)²] / 3, where 4.667 is the average value of the three residuals, and the change trend is fitted by the least square method to obtain the slope The calculation result is -0.15 ° / h / point, where is the residual value, is the time point sequence number, and are the average values, respectively, the correlation decay rate is obtained as 0.92 by calculating the autocorrelation function value of adjacent data points. Since the fluctuation intensity 0.152 exceeds the preset first threshold 0.1, the step length factor is increased from 0.01 to 0.015; the absolute value of the change trend -0.15 exceeds the second threshold 0.1, the direction memory factor is activated and set as 0.9; according to the decay rate 0.92 being in the interval of 0.9-1.0, the memory depth is set as 5 data points. The second subsequence [4.7, 4.5, 4.3] ° / h is calculated to obtain the fluctuation intensity 0.2 ° / h, the change trend -0.2 ° / h / point, and the decay rate 0.88, and accordingly the step length factor is increased to 0.018, the direction memory factor is kept as 0.9, and the memory depth is adjusted to 4 points.

[0119] In the embodiments of the present application, the filter parameters are adaptively adjusted by dynamically analyzing the residual characteristics, so that the system can quickly respond to the error change caused by temperature mutation and maintain the stability of the compensation process, effectively solving the adaptability problem of the fixed parameter filter under complex variable temperature conditions, and improving the compensation accuracy and reliability.

[0120] In order to further improve the accuracy and stability of the generated compensation quantity, in some embodiments, the step 203 of generating the target compensation quantity by using the updated adjustment factor on the residual sequence comprises:

[0121] Step 401: using the updated adjustment factor, traversing each data point in the residual sequence in time sequence, and calculating the initial compensation quantity corresponding to each data point.

[0122] In step 401, the initial compensation quantity refers to the basic compensation value directly calculated according to the current residual data point and the adjustment factor, which reflects the preliminary correction quantity of the error at this moment.

[0123] In the embodiment of the present application, the data points in the residual sequence are processed one by one in chronological order, each data point is multiplied by the updated step factor to obtain an initial compensation amount reflecting the current error size, which provides a basic input for subsequent recursive superposition.

[0124] Step 402: recursively superimpose the initial compensation increment of the current time and the initial compensation amount of the previous time to generate an intermediate compensation amount.

[0125] In step 402, the intermediate compensation amount is a transition compensation value generated by combining the current compensation increment and the historical compensation amount, which considers the current error characteristics and maintains the continuity of compensation.

[0126] In the embodiment of the present application, the initial compensation amount calculated at the current time and the compensation result at the previous time are weighted and summed according to the weight determined by the direction memory factor, so that the compensation amount responds to the latest error change and inherits the historical compensation trend, avoiding sudden changes in the compensation process.

[0127] Step 403: boundary constraint processing is performed on the intermediate compensation amount to generate a target compensation amount.

[0128] In step 403, boundary constraint processing is a process of checking the rationality of the compensation amount to ensure that the compensation amount is within the range allowed by the physical characteristics of the gyroscope.

[0129] In the embodiment of the present application, it is checked whether the intermediate compensation amount exceeds the preset upper and lower limit values, and when it exceeds, the truncation processing is performed according to the boundary value, and the mutation amplitude of the compensation amount is limited according to the change rate to ensure the safety and stability of the output compensation amount.

[0130] The following is a specific example:

[0131] In the temperature compensation process of the vehicle-mounted fiber-optic gyroscope, based on the obtained residual sequence [4.8, 4.7, 4.5, 4.3, 4.1] ° / h and the adjustment parameter step factor 0.015 and the direction memory factor 0.9, the first residual point 4.8 ° / h is processed first, and the initial compensation amount is obtained by multiplying the step factor, that is, 4.8*0.015=0.072 ° / h; then the second residual point 4.7 ° / h is processed, and the initial compensation amount is calculated as 4.7*0.015=0.0705 ° / h; the previous compensation amount 0.072 ° / h is weighted and superimposed according to the memory factor 0.9, that is, 0.9*0.072+0.0705=0.1353 ° / h, which is taken as the new intermediate compensation amount; the third residual point 4.5 ° / h is continuously processed, the initial compensation amount is 4.5*0.015=0.0675 ° / h, and the previous compensation amount 0.1353 ° / h is superimposed to obtain 0.9*0.1353+0.0675=0.1938 ° / h; the complete sequence is processed in this way to obtain the compensation amounts 0.072, 0.1353, 0.1938, 0.2475, 0.2967 ° / h. In the boundary constraint processing stage, the upper limit of the single-step compensation amount is set to 0.3 ° / h, and it is checked that all the intermediate compensation amounts do not exceed the limit, so the target compensation amount is directly output.

[0132] In the embodiments of the present application, the step factor is dynamically adjusted to ensure the compensation response speed, the direction memory factor is used to maintain the compensation continuity, the boundary constraint is combined to ensure the rationality of the output, so that the generated compensation amount can quickly track the error change and maintain stable output, and the problems of compensation lag and overshoot when the temperature changes rapidly are effectively solved.

[0133] In order to further improve the accuracy of the temperature drift prediction model, in some embodiments, step 103: based on the multi-point temperature measurement data, the temperature change rate data, the axial-radial gradient information and the temperature drift component, a nonlinear mapping relationship is established, including:

[0134] Step 501: performing spatial interpolation processing on the axial-radial gradient information to generate temperature gradient distribution data.

[0135] In step 501, the temperature gradient distribution data is continuous gradient field data generated by performing spatial interpolation processing on the discrete axial-radial gradient information, reflecting the change trend of the temperature in space.

[0136] In the embodiments of the present application, the bilinear interpolation method is used to expand the gradient information obtained by the limited measurement points into continuous distribution data covering the entire gyroscope surface, so as to provide spatial gradient information for constructing a complete temperature field feature.

[0137] Step 502: Aligning the multi-point temperature measurement data, the temperature rate of change data and the temperature gradient distribution data according to timestamps to construct multi-dimensional temperature field data, the multi-dimensional temperature field data including a temperature measurement data matrix, a temperature rate of change vector and a temperature gradient tensor.

[0138] In step 502, the multi-dimensional temperature field data is a comprehensive feature set integrating temperature spatial distribution, time rate of change and gradient field, wherein the temperature measurement data matrix records temperature values at different positions, the temperature rate of change vector represents the rate of change, and the temperature gradient tensor describes the spatial change relationship.

[0139] In the embodiment of the present application, the interpolated gradient data is aligned with the original temperature measurement data and the rate of change data according to a unified timestamp to construct a multi-dimensional feature matrix containing spatial distribution characteristics and dynamic change characteristics, thereby laying a foundation for subsequent feature fusion.

[0140] Step 503: Calculating a spatial coupling term of the temperature measurement data matrix and the temperature gradient tensor, and calculating a dynamic coupling term of the temperature rate of change vector and the temperature gradient tensor.

[0141] In step 503, the spatial coupling term represents the cooperative change relationship between the temperature value and the gradient field, and the dynamic coupling term represents the influence degree of the temperature rate of change on the gradient evolution. The dynamic coupling term is a quantitative index representing the dynamic interaction between the temperature rate of change and the temperature gradient field, which is obtained by calculating the dot product of the temperature rate of change vector and the temperature gradient tensor, and reflects the transient response characteristics of the gradient field when the temperature changes rapidly. The numerical value of the dynamic coupling term represents the degree of influence of the temperature change on the spatial heat distribution. For example, when the temperature rises rapidly, the numerical value of the coupling term increases, indicating that the temperature gradient field is dynamically adjusting.

[0142] In the embodiment of the present application, the spatial coupling strength of the temperature measurement value and the gradient tensor is calculated by matrix point multiplication, and the dynamic interaction strength of the temperature rate of change and the gradient tensor is calculated by vector dot product, thereby extracting the correlation between the spatial and dynamic characteristics in the temperature field.

[0143] Step 504: Fusing the spatial coupling term and the dynamic coupling term according to a preset proportion to generate an interaction factor, and generating an enhanced temperature field feature based on the interaction factor.

[0144] In step 504, the interaction factor is a composite index obtained by weighted fusion of the spatial and dynamic coupling characteristics. The enhanced temperature field feature is a high-dimensional feature vector fused with the interaction factor.

[0145] In the embodiment of the present application, the two types of coupling terms are linearly combined into an interaction factor according to a preset importance proportion, and the factor is spliced with the original temperature field feature to form an enhanced feature set containing deep correlation characteristics.

[0146] Step 505: Establishing a nonlinear mapping relationship from the enhanced temperature field feature to the temperature drift component with the corresponding timestamp as the mapping target through an iterative optimization process.

[0147] In the embodiments of the present application, the enhanced feature is taken as the input and the actual drift component is taken as the output, the prediction model is trained by using the iterative optimization algorithm, the prediction error is minimized by continuously adjusting the model parameters, and finally a high-precision nonlinear mapping relationship is established.

[0148] The following is a specific example:

[0149] In the start-up stage of the vehicle-mounted fiber-optic gyroscope, based on the collected four temperature measurement point data 25.0℃, 26.8℃, 28.5℃, 30.2℃ and the calculated axial gradient 0.5℃ / cm and radial gradient 0.6℃ / cm, first, the gradient data is processed by bilinear interpolation to generate a temperature gradient distribution matrix containing 9 nodes, wherein the gradient value of the newly added interpolation point is obtained by linear calculation of the gradient of the adjacent measurement points. The interpolated gradient matrix is aligned with the original temperature data 25.0℃, 26.8℃, 28.5℃, 30.2℃ and the change rate 0.2℃ / s, 0.23℃ / s, 0.25℃ / s, 0.27℃ / s according to the millisecond timestamp, to construct a 4×4 temperature measurement matrix, a 4-dimensional change rate vector and a 3×3 gradient tensor. When calculating the spatial coupling term, the sum of the product of the corresponding elements of the temperature measurement matrix and the gradient tensor is obtained as 2.85, and the calculation formula is wherein is the temperature matrix element, is the gradient tensor element; when calculating the dynamic coupling term, the dot product of the change rate vector and the main diagonal element of the gradient tensor is obtained as 0.145, and the calculation formula is wherein is the change rate element, is the gradient main diagonal element. The spatial coupling term 2.85 and the dynamic coupling term 0.145 are fused according to the weight of 7:3 to obtain the interaction factor 2.85×0.7+0.145×0.3=2.0085, and the factor is combined with the original temperature field feature to generate a 12-dimensional enhanced feature. With the enhanced feature as the input and the 6.9° / h temperature drift component as the output, the gradient descent algorithm is used for 300 iterations of training, and finally a nonlinear mapping relationship is established, which has a prediction error of less than 0.05° / h on the test set, providing an accurate drift prediction benchmark for subsequent dynamic compensation.

[0150] In the embodiments of the present application, by constructing a fusion space gradient and a dynamically changing multi-dimensional temperature field feature, and extracting the deep coupling relationship thereof, the prediction model established can more comprehensively represent the nonlinear relationship between the temperature field and the drift, and improve the accuracy and robustness of the drift prediction under a complex variable temperature environment.

[0151] To further improve the extraction accuracy of the temperature drift component, in some embodiments, step 102: the variational mode decomposition is performed on the gyro original output signal to generate a temperature drift component and a non-temperature noise component, comprising:

[0152] Step 601: performing a variational mode decomposition operation on the gyro original output signal to obtain a plurality of intrinsic mode components.

[0153] In step 601, the intrinsic mode component refers to a signal component with different vibration characteristics separated from the original signal by the variational mode decomposition method, and each component contains specific frequency range and amplitude characteristics.

[0154] In the embodiments of the present application, the gyro original output signal is input into the variational mode decomposition algorithm, and the complex signal is adaptively decomposed into a plurality of intrinsic mode components arranged from high to low in frequency by a preset decomposition layer number and a constraint condition, thereby providing a basis for subsequent temperature-related component screening.

[0155] Step 602: calculating, for each intrinsic mode component, a correlation coefficient of the intrinsic mode component and the temperature change rate data.

[0156] In step 602, the correlation coefficient is an index for measuring the degree of linear correlation between the intrinsic mode component and the temperature change rate data, and the numerical range is between -1 and 1, and the greater the absolute value, the stronger the correlation.

[0157] In the embodiments of the present application, for each intrinsic mode component, the Pearson correlation coefficient of the intrinsic mode component and the temperature change rate data is calculated, and the correlation degree between the two is quantified by analyzing the synchronicity of the component signal and the temperature change trend.

[0158] Step 603: screening a key intrinsic mode component with a correlation coefficient exceeding a preset correlation threshold from all intrinsic mode components, merging all key intrinsic mode components into a temperature drift component, and classifying the remaining intrinsic mode components as non-temperature noise components.

[0159] In step 603, the key intrinsic mode component refers to a signal component related to temperature change, and the correlation coefficient thereof exceeds the preset threshold.

[0160] In the embodiments of the present application, a reasonable correlation threshold is set, components with correlation coefficients exceeding the threshold are screened out as temperature drift components for merging, and components with low correlation are removed, so as to effectively separate the temperature drift signal from noise.

[0161] The following is a specific example:

[0162] In the starting process of the vehicle-mounted fiber-optic gyroscope, based on the collected gyroscope original output signal 15.3° / h, the number of layers of variational mode decomposition is first set to 5 layers, and the signal is decomposed into 5 intrinsic mode components through iterative optimization, and the amplitudes thereof are 6.2° / h, 4.1° / h, 2.8° / h, 1.5° / h and 0.7° / h. For each component, the temperature change rate data 0.2℃ / s, 0.23℃ / s, 0.25℃ / s, 0.27℃ / s collected synchronously are combined, and the correlation coefficient formula is used to calculate the correlation. The calculated correlation coefficients of the components are 0.62, 0.88, 0.91, 0.65 and 0.58 in turn. The correlation threshold is set to 0.8, the two components corresponding to the correlation coefficients 0.88 and 0.91, i.e., 4.1° / h and 2.8° / h, are screened out, the amplitudes thereof are added, i.e., 4.1+2.8=6.9° / h, as the temperature drift component, and the remaining three components, i.e., 6.2° / h, 1.5° / h and 0.7° / h, are combined as 8.4° / h as the non-temperature noise component and removed.

[0163] In the embodiments of the present application, the signal component closely related to the temperature change can be accurately separated by combining variational mode decomposition and correlation analysis, the non-temperature interference such as vibration can be effectively filtered out, a pure input signal is provided for subsequent temperature drift compensation, and the extraction precision of the temperature drift component under complex working conditions is improved.

[0164] In order to further improve the stability and reliability of the gyroscope output signal, in some embodiments, the step 105 of superimposing the compensation result and the drift prediction value to generate a compensated gyroscope output signal comprises:

[0165] Step 701: performing time domain smoothing processing on the compensation result to generate a stable compensation amount.

[0166] In step 701, the stable compensation amount refers to the compensation result after smoothing filtering processing, and the compensation amount change is made more gentle by suppressing high-frequency fluctuations.

[0167] In the embodiments of the present application, the sliding average filtering method is used to process the original compensation result, the average value of adjacent multiple compensation amounts is taken to eliminate random fluctuations, a smooth and stable compensation amount sequence is generated, and sudden changes in the compensation process are avoided.

[0168] Step 702: algebraically add the stable compensation amount to the drift prediction value to generate a modified drift amount.

[0169] In step 702, the modified drift amount is a comprehensive compensation value obtained by directly adding the stable compensation amount to the drift prediction value, which contains both static prediction and dynamic compensation.

[0170] In the embodiments of the present application, the smoothed compensation amount and the drift prediction value at the corresponding time are added one by one according to the time point to form the final compensation amount considering both the temperature field characteristic modeling result and the real-time error tracking result.

[0171] Step 703: amplitude limiting processing is performed on the modified drift amount.

[0172] In step 703, the amplitude limiting processing is a process of checking the reasonableness of the modified drift amount, ensuring that the compensation amount is within the range allowed by the physical characteristics of the gyroscope.

[0173] In the embodiments of the present application, upper and lower threshold values of the compensation amount are set, and when the modified drift amount exceeds the threshold value, it is forcibly limited within the boundary value to prevent abnormal output caused by excessive compensation.

[0174] Step 704: remove the amplitude limiting processed modified drift amount from the original output signal of the gyroscope to obtain the compensated gyroscope output signal.

[0175] In the embodiments of the present application, the amplitude limiting processed modified drift amount is correspondingly deducted from the synchronously collected original gyroscope output signal to obtain the accurate output signal after eliminating the influence of temperature drift.

[0176] The following is a specific example:

[0177] In the starting process of the vehicle-mounted fiber-optic gyroscope, based on the obtained compensation quantity sequence 0.0054, 0.0099, 0.01341, 0.01657, 0.01891° / h and the drift prediction value 2.1° / h, first, the compensation quantity is processed by three-point sliding average, the first point 0.0054° / h remains unchanged, the second point takes the average of the first three compensation quantities (0.0054+0.0099+0.01341) / 3=0.00957° / h, the third point (0.0099+0.01341+0.01657) / 3=0.01329° / h, the fourth point (0.01341+0.01657+0.01891) / 3=0.01630° / h, and the last point 0.01891° / h remains unchanged, to obtain the smoothed stable compensation quantity sequence [0.0054, 0.00957, 0.01329, 0.01630, 0.01891]° / h. Add these stable compensation quantities to the prediction value 2.1° / h point by point to obtain the modified drift quantity sequence [2.1054, 2.10957, 2.11329, 2.11630, 2.11891]° / h. Set the maximum compensation quantity allowed by the system to be 2.12° / h, check that none of the modified values exceeds the limit, so the modified values are directly used for final compensation. Subtract these modified values from the original gyroscope output 15.3° / h in turn to obtain the compensated output signal [13.1946, 13.19043, 13.18671, 13.1837, 13.18109]° / h. In the sliding average calculation formula, the smoothed compensation quantity is the arithmetic average of the adjacent three original compensation quantities, which ensures the smoothness of the compensation quantity change; the modified drift quantity calculation uses simple algebraic addition of the compensation quantity and the prediction value; in the amplitude limiting process, it is checked whether each modified value exceeds the preset upper limit of 2.12° / h. Through this series of processing, the final output of the gyroscope effectively eliminates the influence of temperature drift, while ensuring the smoothness of the output change, so that the gyroscope can still maintain stable and reliable measurement performance during the temperature change in the starting stage.

[0178] In the embodiments of the present application, the compensation stability is improved by smoothing processing, and the safety is ensured by amplitude limiting protection. The finally generated compensated signal not only effectively eliminates the influence of temperature drift, but also maintains the smoothness of the output change, thereby improving the measurement reliability of the gyroscope in complex temperature environment.

[0179] Figure 4 A specific implementation structure diagram of a fiber-optic gyroscope drift compensation system combined with LMS adaptive filtering is provided for the embodiments of the present application, referring to Figure 4 The system can include:

[0180] The acquisition module 41 is configured to acquire multi-point temperature measurement data, temperature change rate data, axial and radial gradient information of the fiber-optic gyroscope, and an original output signal of the gyroscope.

[0181] a decomposition module 42 configured to perform a variational mode decomposition on the gyro original output signal to generate a temperature drift component and a non-temperature noise component, and filter out the non-temperature noise component.

[0182] a building module 43 configured to build a non-linear mapping relationship based on the multi-point temperature measurement data, the temperature change rate data, the axial-radial gradient information, and the temperature drift component.

[0183] a generating module 44 configured to generate a drift prediction value based on the non-linear mapping relationship, associate the drift prediction value with the temperature drift component, and input an association result into an LMS adaptive filter for dynamic error compensation.

[0184] a superimposing module 45 configured to superimpose a compensation result with the drift prediction value to generate a compensated gyro output signal.

[0185] The fusion LMS adaptive filtering fiber optic gyroscope drift compensation system according to the embodiments of the present application is used to implement the aforementioned fusion LMS adaptive filtering fiber optic gyroscope drift compensation method, and thus the specific implementation manners of the fusion LMS adaptive filtering fiber optic gyroscope drift compensation system can be seen from the aforementioned embodiment part of the fusion LMS adaptive filtering fiber optic gyroscope drift compensation method, and the specific implementation manners can be referred to the description of the corresponding embodiment part, which will not be described herein again.

[0186] The present application further provides an electronic device, which comprises a memory configured to store a computer program, and a processor configured to execute the computer program to implement the steps of any one of the aforementioned fusion LMS adaptive filtering fiber optic gyroscope drift compensation methods.

[0187] The present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of any one of the aforementioned fusion LMS adaptive filtering fiber optic gyroscope drift compensation methods.

[0188] In an exemplary embodiment, the aforementioned computer readable storage medium can include, but is not limited to, a U disk, a read-only memory, a random access memory, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0189] The embodiments of the present application further provide a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps in any one of the aforementioned fusion LMS adaptive filtering fiber optic gyroscope drift compensation method embodiments.

[0190] Those skilled in the art will further realize that the mechanisms of the various examples described herein are capable of being implemented using computer software, firmware, hardware, or combinations of them, and that the various examples give the necessary control signals and data information to a responsible application-specific computer or network component to cause the computer or network component to implement various aspects of the mechanisms effectively. The computer software referred to herein can be stored in main memory and / or secondary memory associated with the computer hardware and implemented using the processor(s) of the computer(s). Suitable computer software includes, but is not limited to, source code, object code, and interpreted code. The software implementation can be in a dedicated language, or can be spread out in several different languages.

[0191] The above provides a kind of fusion LMS adaptive filtering's optical fiber gyro drift compensation method and system provided in the application.The principle and implementation of the present application are described in the text by applying specific examples, the above example is only used to help understand the method and its core idea of the present application.It should be pointed out that, for the ordinary skilled in the art, without departing from the principle of the present application, the present application can be improved and modified in several ways, and these improvements and modifications also fall within the scope of the present application.

Claims

1. A method for drift compensation of a fiber-optic gyroscope by fusion LMS adaptive filtering, characterized in that, The method comprises the following steps: Collecting multi-point temperature measurement data, temperature rate of change data, axial and radial gradient information of the fiber-optic gyroscope, and original output signals of the gyroscope; Performing variational mode decomposition on the original output signals of the gyroscope to generate temperature drift components and non-temperature noise components, and filtering out the non-temperature noise components; Based on the multi-point temperature measurement data, the temperature rate of change data, the axial and radial gradient information, and the temperature drift components, a nonlinear mapping relationship is established; Based on the nonlinear mapping relationship, a drift prediction value is generated, the drift prediction value is associated with the temperature drift components, and the associated result is input into an LMS adaptive filter for dynamic error compensation; The compensation result is superimposed with the drift prediction value to generate a compensated gyroscope output signal; The association of the drift prediction value with the temperature drift components and the input of the associated result into the LMS adaptive filter for dynamic error compensation comprises: The drift prediction value is subtracted from the temperature drift components to generate a residual sequence; The residual sequence is input into the LMS adaptive filter, and based on a statistical feature set of the residual sequence, an adjustment factor of the LMS adaptive filter is dynamically updated; An updated adjustment factor is used to perform approximation operation on the residual sequence to generate a target compensation amount, which is the compensation result.

2. The method of claim 1, wherein, The adjustment factor of the LMS adaptive filter comprises an update step factor, a direction memory factor, and a memory depth factor; Based on the statistical feature set of the residual sequence, the adjustment factor of the LMS adaptive filter is dynamically updated, which comprises: The LMS adaptive filter is used to perform time domain segmentation on the residual sequence to obtain a plurality of residual sub-sequences; For each time domain segment, a statistical feature set is extracted from the corresponding residual sub-sequence, the statistical feature set comprising a fluctuation intensity value, a change trend value, and a correlation decay rate; When the fluctuation intensity value is greater than a preset first threshold value, the update step factor of the corresponding time domain segment is increased; When the change trend value is greater than a preset second threshold value, the direction memory factor of the corresponding time domain segment is activated; According to the correlation decay rate, the memory depth factor of the corresponding time domain segment is dynamically adjusted.

3. The method of claim 1, wherein, The use of the updated adjustment factor to perform approximation operation on the residual sequence to generate a target compensation amount comprises: Using the updated adjustment factor, each data point in the residual sequence is traversed in time sequence, and an initial compensation amount corresponding to each data point is calculated; The initial compensation increment at the current time is recursively superimposed with the initial compensation amount at the previous time to generate an intermediate compensation amount; The intermediate compensation amount is subjected to boundary constraint processing to generate a target compensation amount.

4. The method of claim 1, wherein, The establishment of the nonlinear mapping relationship based on the multi-point temperature measurement data, the temperature rate of change data, the axial and radial gradient information, and the temperature drift components comprises: The axial and radial gradient information is subjected to spatial interpolation processing to generate temperature gradient distribution data; aligning the multi-point temperature measurement data, the temperature rate of change data and the temperature gradient distribution data according to time stamps to construct multi-dimensional temperature field data, the multi-dimensional temperature field data comprising a temperature measurement data matrix, a temperature rate of change vector and a temperature gradient tensor; calculating a spatial coupling term of the temperature measurement data matrix and the temperature gradient tensor, and calculating a dynamic coupling term of the temperature rate of change vector and the temperature gradient tensor; fusing the spatial coupling term and the dynamic coupling term according to a preset proportion to generate an interaction factor, and generating an enhanced temperature field feature based on the interaction factor; taking the enhanced temperature field feature as a mapping source, and taking the temperature drift component of the corresponding time stamp as a mapping target, and establishing a nonlinear mapping relationship from the mapping source to the mapping target through an iterative optimization process.

5. The method of claim 1, wherein, The variational modal decomposition of the gyro original output signal to generate the temperature drift component and the non-temperature noise component comprises: performing variational modal decomposition on the gyro original output signal to obtain a plurality of intrinsic modal components; for each intrinsic modal component, calculating a correlation coefficient of the intrinsic modal component and the temperature rate of change data; from all the intrinsic modal components, screening out key intrinsic modal components whose correlation coefficients exceed a preset correlation threshold, combining all the key intrinsic modal components into a temperature drift component, and classifying the remaining intrinsic modal components as non-temperature noise components.

6. The method of claim 1, wherein, The superposition of the compensation result and the drift prediction value to generate a compensated gyro output signal comprises: performing time domain smoothing processing on the compensation result to generate a stable compensation amount; algebraically adding the stable compensation amount and the drift prediction value to generate a modified drift amount; performing amplitude limiting processing on the modified drift amount; removing the amplitude limiting processed modified drift amount from the gyro original output signal to obtain a compensated gyro output signal.

7. A fiber optic gyroscope drift compensation system with fused LMS adaptive filtering, characterized by, It comprises: a collection module for collecting multi-point temperature measurement data, temperature rate of change data, axial and radial gradient information and a gyro original output signal of an optical fiber gyro; a decomposition module for performing variational modal decomposition on the gyro original output signal to generate a temperature drift component and a non-temperature noise component, and filtering out the non-temperature noise component; an establishment module for establishing a nonlinear mapping relationship based on the multi-point temperature measurement data, the temperature rate of change data, the axial and radial gradient information and the temperature drift component; a generation module for generating a drift prediction value based on the nonlinear mapping relationship, associating the drift prediction value with the temperature drift component, and inputting the association result into an LMS adaptive filter for dynamic error compensation; a superposition module for superimposing a compensation result and the drift prediction value to generate a compensated gyro output signal; The association of the drift prediction value and the temperature drift component, and the input of the association result into an LMS adaptive filter for dynamic error compensation comprises: subtracting the drift prediction value from the temperature drift component to generate a residual sequence; inputting the residual sequence into an LMS adaptive filter, dynamically updating an adjustment factor of the LMS adaptive filter based on a set of statistical features of the residual sequence; performing an approximation operation on the residual sequence by using the updated adjustment factor to generate a target compensation quantity, the target compensation quantity being a compensation result.

8. An electronic device, comprising: The method comprises the steps of: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the fiber-optic gyroscope drift compensation method with fused LMS adaptive filtering according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer program is stored in the computer-readable storage medium and can be executed by the processor to implement the fiber-optic gyroscope drift compensation method with fused LMS adaptive filtering according to any one of claims 1 to 6.

Images