Filtering optimization method and system for optical fiber gyroscope of inertial measurement unit

Abstract

The invention provides a filtering optimization method and system for an inertial measurement unit optical fiber gyroscope. A three-dimensional vibration signal of an optical fiber gyroscope of an inertial measurement unit is collected, and vibration mode characteristic data is generated through short-time Fourier transform. Dynamically adjusting the magnetic field intensity of the gyroscope according to the data, and converting environment vibration energy into current data based on the adjusted magnetic field intensity; extracting first and second noise features from the current data and the original angular velocity data respectively by combining the original angular velocity data output by the gyroscope and adopting a dual-channel convolutional neural network, and outputting preliminary noise reduction data after fusion; and performing variational mode decomposition on the preliminary noise reduction data, and screening effective mode components according to a preset kurtosis rule to obtain target angular velocity data for realizing filtering optimization of the gyroscope. According to the technical scheme provided by the invention, the measurement precision and the filtering optimization effect of the optical fiber gyroscope of the inertial measurement unit are improved.

Description

Technical Field

[0001] The present application relates to the field of filtering optimization technology, and in particular to a filtering optimization method and system for an inertial measurement unit fiber optic gyroscope. Background Art

[0002] Inertial measurement unit (IMU) fiber optic gyroscopes (FOGs), whether for attitude control of drones, motion sensing of autonomous vehicles, or orbital adjustment of spacecraft, all rely on the high-precision angular velocity data output by the gyroscopes. However, during operation, gyroscopes are susceptible to factors such as deformation of internal vibrating components, external electromagnetic interference, and ambient temperature changes. This can cause noise in multiple frequency bands to be mixed into the original measurement data. This noise can reduce data accuracy and, in turn, affect the reliability of subsequent control decisions. Therefore, filtering optimization methods are urgently needed to eliminate this noise interference and improve data stability and accuracy.

[0003] Currently, a common solution to this noise problem is to use an adaptive filtering method based on fixed parameters. This method uses a preset noise model, determines the filtering parameters based on the initial statistical characteristics, and processes the raw data output by the gyroscope in real time. By continuously adjusting the filtering strength, it attempts to suppress noise and obtain a purer measurement signal.

[0004] However, this existing solution has obvious flaws: its preset noise model is difficult to adapt to the complex and changeable actual environment. When the noise frequency band changes dynamically with factors such as vibration intensity and electromagnetic interference level, the adaptive filtering with fixed parameters cannot adjust the suppression strength of noise in different frequency bands in time, resulting in the noise in some frequency bands not being effectively filtered out, and the effective signal components may be excessively weakened, ultimately affecting the authenticity and integrity of the filtered data. Summary of the Invention

[0005] The present application provides a filtering optimization method and system for an inertial measurement unit fiber optic gyroscope, which is used to solve the problems of poor filtering optimization effect caused by insufficient model adaptability, untimely parameter adjustment, and unbalanced filtering effect in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a filtering optimization method for an inertial measurement unit fiber optic gyroscope, comprising: Collecting three-dimensional vibration signals from an inertial measurement unit fiber optic gyroscope, performing short-time Fourier transform on the three-dimensional vibration signals, and generating vibration modal characteristic data; Dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength; Based on the current data and raw angular velocity data output by an inertial measurement unit fiber optic gyroscope, a dual-channel convolutional neural network is used to extract a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, and the first noise feature and the second noise feature are fused to output preliminary noise reduction data; Variational modal decomposition is performed on the preliminary denoised data, and effective modal components are screened from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data. The target angular velocity data is used to achieve filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0007] Optionally, based on the current data and in combination with raw angular velocity data output by an inertial measurement unit fiber optic gyroscope, a dual-channel convolutional neural network is used to extract a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, and the first noise feature and the second noise feature are fused to output preliminary noise reduction data, including: Transmitting the current data into a first processing path in the dual-channel convolutional neural network, and transmitting the raw angular velocity data into a second processing path in the dual-channel convolutional neural network, wherein the first processing path and the second processing path are run simultaneously; Based on the first processing path, performing multiple feature extractions on the current data to obtain multiple first candidate features; Based on the second processing path, performing multiple feature extractions on the original angular velocity data to obtain multiple second candidate features; Extracting a first noise feature with the highest correlation with the noise-related feature from the plurality of first candidate features, and extracting a second noise feature with the highest correlation with the noise-related feature from the plurality of second candidate features; The first noise feature and the second noise feature are weighted according to their noise characterization capabilities and combined to generate a comprehensive noise feature. Based on the comprehensive noise feature, a noise transfer function including environmental parameters is constructed. Based on the noise transfer function, preliminary noise reduction data is generated to achieve dynamic suppression of multi-band noise.

[0008] Optionally, assigning weights to the first noise feature and the second noise feature according to their noise characterization capabilities and merging the features to generate a comprehensive noise feature, constructing a noise transfer function including environmental parameters based on the comprehensive noise feature, and generating preliminary noise reduction data based on the noise transfer function to achieve dynamic multi-band noise suppression includes: generating a first noise characteristic value corresponding to the first noise characteristic and a second noise characteristic value corresponding to the second noise characteristic, and assigning weights to the first noise characteristic value and the second noise characteristic value respectively based on a noise characterization capability; Calculating the first noise characteristic value and the second noise characteristic value with corresponding weights respectively to obtain corresponding comprehensive noise characteristic values, and combining the comprehensive noise characteristic values ​​of all preset dimensions to form a comprehensive noise feature; collecting temperature and humidity parameters of the current environment as environmental parameters, and constructing a noise transfer function including noise suppression coefficients corresponding to different frequency bands based on the environmental parameters and the comprehensive noise characteristics; Based on the noise transfer function, corresponding noise suppression coefficients are applied to the respective frequency bands to generate preliminary noise reduction data.

[0009] Optionally, performing variational modal decomposition on the preliminary denoised data, screening effective modal components from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, wherein the target angular velocity data is used to implement filtering optimization of an inertial measurement unit fiber optic gyroscope, includes: Perform variational mode decomposition on the preliminary denoised data to obtain multiple data segments; Calculating a waveform distribution characteristic value of each of the data segments; Matching each waveform distribution characteristic value with a characteristic threshold range in a preset kurtosis rule, and screening data segments whose waveform distribution characteristic values ​​are within the characteristic threshold range as valid modal components; Each of the effective modal components is combined to obtain target angular velocity data.

[0010] Optionally, dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength, includes: Dividing the vibration modal characteristic data into preset frequency bands, and extracting vibration frequency data and corresponding amplitude data based on each divided frequency band; Calculating, based on the corresponding amplitude data, a power ratio of the vibration frequency data in the vibration modal characteristic data; Determining an adjustment direction and a value of the magnetic field strength based on the power ratio, and adjusting the magnetic field strength based on the adjustment direction and the value; Based on the adjusted magnetic field strength, the vibration sensing component inside the fiber optic gyroscope of the inertial measurement unit is controlled to produce deformation. During the deformation process, it is converted into a corresponding initial voltage signal according to a predetermined conversion rule. Based on the initial voltage signal, current data is generated through processing by a signal conversion component.

[0011] Optionally, determining an adjustment direction and a value of the magnetic field strength based on the power proportion, and adjusting the magnetic field strength based on the adjustment direction and the value, includes: Presetting corresponding upper and lower limits for the vibration frequency data of each frequency band, and calculating, based on the magnitude of the vibration frequency data, a difference between the vibration frequency data and the upper or lower limit to obtain a first deviation value for each frequency band; Based on the power proportion of each frequency band, assigning a corresponding weight coefficient to each frequency band; Calculating a second deviation value by multiplying the first deviation value of each frequency band by the weight coefficient and adding the results together, and determining an adjustment direction according to the positive or negative value of the second deviation value; determining an adjustment value of the magnetic field strength based on a product of the second deviation value and a preset adjustment coefficient; Based on the adjustment direction and the adjustment value, the initial magnetic field strength value is calculated to obtain the adjusted magnetic field strength.

[0012] Optionally, the converting into a corresponding initial voltage signal according to a predetermined conversion rule, and generating current data based on the initial voltage signal through processing by a signal conversion component, includes: Determining a change in magnetic flux within the vibration sensing component based on the deformation, wherein the change in magnetic flux and the deformation amount satisfy a preset functional relationship; Based on the change in the magnetic flux, an induced electromotive force is generated in the coil; The induced electromotive force is converted into an initial voltage signal by a rectifier according to a predetermined conversion rule, and a high-frequency fluctuation component is filtered out by a filter device based on the initial voltage signal to obtain a stable target voltage signal; The target voltage signal is converted into current data by a current conversion device.

[0013] In a second aspect, an embodiment of the present application provides a filtering optimization system for an inertial measurement unit fiber optic gyroscope, comprising: An acquisition module is used to acquire a three-dimensional vibration signal from an inertial measurement unit fiber optic gyroscope, perform short-time Fourier transform on the three-dimensional vibration signal, and generate vibration modal characteristic data; an adjustment module, configured to dynamically adjust the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and convert the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength; a fusion module, configured to extract, based on the current data and raw angular velocity data output by an inertial measurement unit fiber optic gyroscope, a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, by employing a dual-channel convolutional neural network, and fuse the first noise feature with the second noise feature to output preliminary noise reduction data; An optimization module is configured to perform variational modal decomposition on the preliminary denoised data, and to filter effective modal components from the preliminary denoised data after the variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, wherein the target angular velocity data is used to implement filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0014] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a filtering optimization method for an inertial measurement unit fiber optic gyroscope as described in any one of the first aspects.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a filtering optimization method for an inertial measurement unit fiber optic gyroscope as described in any one of the first aspects.

[0016] The present application provides a filtering optimization method for an inertial measurement unit (IMU) fiber optic gyroscope (FOG), comprising: collecting a three-dimensional vibration signal from the IMU fiber optic gyroscope, performing a short-time Fourier transform on the three-dimensional vibration signal, and generating vibration modal characteristic data; dynamically adjusting the magnetic field strength of the IMU fiber optic gyroscope according to the vibration modal characteristic data, and converting the ambient vibration energy of the IMU fiber optic gyroscope into current data based on the adjusted magnetic field strength; extracting a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, by using a dual-channel convolutional neural network based on the current data and in combination with raw angular velocity data output by the IMU fiber optic gyroscope, and fusing the first noise feature and the second noise feature to output preliminary denoised data; performing variational modal decomposition on the preliminary denoised data, and screening effective modal components from the preliminary denoised data after the variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, wherein the target angular velocity data is used to implement filtering optimization of the IMU fiber optic gyroscope.

[0017] The present application has the following advantages: by collecting the three-dimensional vibration signal of the fiber optic gyroscope of the inertial measurement unit and performing a short-time Fourier transform to generate vibration modal characteristic data, the original vibration signal can be converted into a characteristic form that can be used for subsequent processing; by adjusting the magnetic field intensity according to the vibration modal characteristic data and converting the vibration energy into current data, the conversion of vibration energy into electrical signals can be achieved; by extracting and fusing noise features from the current data and the original angular velocity data through a dual-channel convolutional neural network to output preliminary noise reduction data, the noise features can be identified by combining multi-source data; by performing variational modal decomposition on the preliminary noise reduction data and screening the effective components to obtain the target angular velocity data, the data can be further optimized to achieve filtering optimization.

[0018] Furthermore, the present application respectively transmits the current data and the original angular velocity data into the two paths of the dual-channel convolutional neural network for parallel processing, extracts the candidate features and screens out the features with the highest correlation with noise, assigns weights according to the characterization capabilities and merges them to generate comprehensive noise features, and constructs a transfer function containing frequency band noise suppression coefficients in combination with environmental parameters, and applies this function to generate preliminary noise reduction data; therefore, the present application can realize the coordinated processing of multi-source data, accurately capture noise features, and specifically suppress multi-band noise, thereby improving the quality of preliminary noise reduction data.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flowchart of a filtering optimization method for an inertial measurement unit fiber optic gyroscope provided in an embodiment of the present application; Figure 2 A schematic structural diagram of a filter optimization system for an inertial measurement unit fiber optic gyroscope provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution 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.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0025] In order to solve the problems of poor filtering optimization effects caused by insufficient model adaptability, untimely parameter adjustment, and unbalanced filtering effects in the prior art, an embodiment of the present application provides a filtering optimization method for an inertial measurement unit fiber optic gyroscope. The method adopts the following concepts: by converting continuous vibration signals into feature data, effective extraction and analysis of vibration information are achieved, providing a basis for subsequent processing; dynamically adjusting the magnetic field based on vibration characteristics and converting vibration energy into current data broadens the data source for noise analysis; using a dual-channel convolutional neural network to fuse multi-source noise features, the noise interference in the data is preliminarily reduced; through variational mode decomposition and effective component screening, the data quality is further improved, and the target angular velocity data finally obtained can effectively optimize the filtering effect of the gyroscope and improve measurement accuracy and reliability.

[0026] Figure 1 A flowchart of a filtering optimization method for an inertial measurement unit fiber optic gyroscope provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes: S11. Collect three-dimensional vibration signals from the fiber optic gyroscope of the inertial measurement unit, perform short-time Fourier transform on the three-dimensional vibration signals, and generate vibration modal characteristic data.

[0027] Among them, the three-dimensional vibration signal refers to the recorded data of the vibration of the fiber optic gyroscope of the inertial measurement unit in the X, Y, and Z directions, including the strength and changes of the vibration; the short-time Fourier transform is a method of dividing the signal into multiple short time periods and analyzing the frequency components contained in the signal in each time period separately; the vibration modal characteristic data is obtained after the short-time Fourier transform, which can reflect the vibration characteristics at different times and frequencies.

[0028] In an embodiment of the present application, first, the vibration data of the fiber optic gyroscope of the inertial measurement unit in the X, Y, and Z directions is collected through the sensing component. For example, when the gyroscope vibrates during the operation of the device, the change in the magnitude of the vibration in each direction over time is recorded. Secondly, the collected three-dimensional vibration signal is divided into multiple segments according to a fixed short time length, and a short-time Fourier transform is applied to each segment to obtain the strength of the different frequency components in each segment. Finally, the analysis results of all the segments are integrated to form vibration modal characteristic data, which contains both time information and frequency information.

[0029] S12. Dynamically adjust the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and convert the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength.

[0030] Among them, vibration modal characteristic data is data that can reflect the characteristics of vibration in time and frequency; magnetic field intensity refers to the strength of the magnetic field inside the fiber optic gyroscope of the inertial measurement unit; current data is a record that can reflect the changes in the strength of the electrical signal, which is converted from vibration energy.

[0031] In the embodiments of the present application, the strength of the magnetic field within the fiber optic gyroscope (FOG) of the inertial measurement unit (IMU) is first adjusted based on the vibration intensity and frequency changes reflected in the vibration modal characteristic data. Secondly, in the adjusted magnetic field, the environmental vibrations to which the gyroscope is subjected cause internal components to move. This movement drives the relevant components to cut through the magnetic field, converting the vibration energy into an electrical signal. Finally, the generated electrical signal is converted into current data that reflects its intensity changes, and the magnitude of this data varies with the vibration energy.

[0032] S13. Based on the current data and the original angular velocity data output by the fiber optic gyroscope of the inertial measurement unit, a dual-channel convolutional neural network is used to extract a first noise feature and a second noise feature from the current data and the original angular velocity data, respectively, and the first noise feature and the second noise feature are fused to output preliminary noise reduction data.

[0033] Among them, the current data is an electrical signal record converted from vibration energy; the raw angular velocity data is directly measured by the fiber optic gyroscope of the inertial measurement unit, and reflects the raw data of rotation speed; the dual-channel convolutional neural network is a computing model that can simultaneously process two different data; the first noise feature and the second noise feature are extracted from the current data and the raw angular velocity data respectively, and can reflect the information of noise characteristics; the preliminary noise reduction data is the data obtained by fusing the two noise features, with reduced noise.

[0034] In this embodiment, current data and raw angular velocity data are first fed into two processing components of a dual-channel convolutional neural network, each operating simultaneously. Second, in the first processing path, the current data is analyzed multiple times to identify information that characterizes noise, namely the first noise signature. In the second processing path, similar processing is performed on the raw angular velocity data to obtain the second noise signature. Finally, the first and second noise signatures are combined and computationally adjusted to reduce the noise in the raw data, resulting in preliminary noise-reduced data.

[0035] S14. Perform variational modal decomposition on the preliminary denoised data, and screen effective modal components from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data. The target angular velocity data is used to achieve filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0036] Among them, the preliminary denoised data is the data with reduced noise after preliminary processing; variational mode decomposition is a method of decomposing data into multiple different parts, each with its own characteristics; the preset kurtosis rule is a standard for judging whether a data part is valid, based on the distribution of extreme values ​​in the data; the effective modal component refers to the data part that meets the preset kurtosis rule; the target angular velocity data is the final data obtained after screening and combination, which can be used to optimize the gyroscope filtering effect.

[0037] In this embodiment, the variational mode decomposition method is first applied to the preliminary noise reduction data, dividing it into multiple distinct components, each reflecting different fluctuation characteristics in the data. Next, each decomposed component is checked according to a preset kurtosis rule to determine whether it is valid. Finally, all valid components are combined to form the target angular velocity data.

[0038] For example, during the operation of the fiber optic gyroscope (FOG) on device A's inertial measurement unit, vibration data is first collected in the horizontal left-right, horizontal front-back, and vertical directions. This data is then divided into 0.5-second segments, and a short-time Fourier transform (SFT) is performed on each segment to obtain vibration modal signature data that reflects the vibration conditions at different frequencies within each time period. The gyroscope's internal magnetic field strength is then dynamically adjusted based on this data, for example, by increasing the magnetic field as the vibration frequency increases. The vibration energy during device operation is converted into electrical signals through the movement of internal components in the magnetic field, generating corresponding current data. The current data and the measured raw angular velocity data are then fed into two separate parts of a computational model. The model extracts the noise signature caused by vibration from the current data and the noise signature caused by interference from the raw angular velocity data. These two signatures are combined to process the raw data to produce preliminary noise-reduced data. Finally, the preliminary noise-reduced data is decomposed into five distinct components. After checking against pre-set rules, three of these components meet the requirements. These three components are then combined to obtain the target angular velocity data that can be used to optimize the gyroscope's filtering effect.

[0039] By executing S11 to S14, the embodiment of the present application realizes the effective extraction and analysis of vibration information by converting continuous vibration signals into characteristic data, providing a basis for subsequent processing; dynamically adjusts the magnetic field based on the vibration characteristics and converts vibration energy into current data, broadening the data source for noise analysis; utilizes a dual-channel convolutional neural network to fuse multi-source noise characteristics, preliminarily reducing noise interference in the data; and further improves data quality through variational mode decomposition and effective component screening. The target angular velocity data finally obtained can effectively optimize the filtering effect of the gyroscope and improve measurement accuracy and reliability.

[0040] In one possible embodiment, S13, based on the current data and in combination with the raw angular velocity data output by the inertial measurement unit fiber optic gyroscope, a dual-channel convolutional neural network is used to extract a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, and the first noise feature and the second noise feature are fused to output preliminary noise reduction data, including: Step 131: The current data is transmitted to the first processing path in the dual-channel convolutional neural network, and the raw angular velocity data is transmitted to the second processing path in the dual-channel convolutional neural network. The first processing path and the second processing path run simultaneously.

[0041] Among them, the current data is an electrical signal record converted from the vibration energy of the fiber optic gyroscope of the inertial measurement unit, which can reflect the changes in vibration energy; the raw angular velocity data is directly measured by the gyroscope and reflects the original information of the rotation speed; the dual-channel convolutional neural network is a computing model that can process two different data at the same time. It contains two independent processing parts, called the first processing path and the second processing path. The first processing path is used to process current data, and the second processing path is used to process raw angular velocity data. The two paths can run simultaneously to speed up the processing speed.

[0042] In an embodiment of the present application, the current data is input into the first processing path of the dual-channel convolutional neural network, and the raw angular velocity data is input into the second processing path of the network. The two paths start processing the data they receive at the same time. For example, in the gyroscope data processing of device A, the current data enters the first processing path and the raw angular velocity data enters the second processing path. The two paths start the analysis program at the same time and perform data processing without affecting each other.

[0043] Step 132: Based on the first processing path, perform multiple feature extractions on the current data to obtain multiple first candidate features.

[0044] Among them, the first processing path is the part of the dual-channel convolutional neural network specifically used to process current data; multiple feature extraction refers to analyzing the current data in multiple different ways, and each analysis can extract information that reflects the characteristics of a certain aspect of the data; the first candidate feature is the multiple information that can reflect the different characteristics of the current data obtained after multiple feature extractions.

[0045] In an embodiment of the present application, in a first processing path, a first feature extraction is performed on the current data, such as analyzing the fluctuation amplitude of the data over time to obtain a first candidate feature that reflects the size of the fluctuation; then a second extraction is performed to analyze the frequency change of the fluctuation to obtain a first candidate feature that reflects the frequency characteristics; and then a third extraction is performed to analyze the duration of the fluctuation to obtain a first candidate feature that reflects the duration. Through such multiple different analyses, multiple first candidate features are obtained. For example, the current data of device A is extracted three times to obtain three candidate features, namely, fluctuation amplitude, frequency change, and duration.

[0046] Step 133: Based on the second processing path, perform multiple feature extractions on the original angular velocity data to obtain multiple second candidate features.

[0047] Among them, the second processing path is the part of the dual-channel convolutional neural network specifically used to process the raw angular velocity data; multiple feature extractions refer to analyzing the raw angular velocity data in multiple different ways, and each analysis can extract information that reflects the characteristics of a certain aspect of the data; the second candidate features are multiple pieces of information that can reflect the different characteristics of the raw angular velocity data obtained after multiple feature extractions.

[0048] In an embodiment of the present application, in the second processing path, the first feature extraction is performed on the original angular velocity data, such as analyzing the numerical range of the rotation speed in the data to obtain a second candidate feature that reflects the magnitude of the speed; then a second extraction is performed to analyze the difference in speed changes at adjacent moments to obtain a second candidate feature that reflects the speed change rate; and then a third extraction is performed to analyze the stability of the speed over a long period of time to obtain a second candidate feature that reflects stability. Through such multiple different analyses, multiple second candidate features are obtained. For example, the original angular velocity data of device A is extracted three times to obtain three candidate features of speed magnitude, speed of change, and stability.

[0049] Step 134: extracting a first noise feature having the highest correlation with the noise-related feature from the plurality of first candidate features, and extracting a second noise feature having the highest correlation with the noise-related feature from the plurality of second candidate features.

[0050] Among them, multiple first candidate features are multiple information reflecting different characteristics extracted from current data; multiple second candidate features are multiple information reflecting different characteristics extracted from original angular velocity data; noise-related features refer to information that can reflect noise characteristics, such as irregular fluctuations, sudden numerical changes, etc.; correlation refers to the degree of similarity between candidate features and noise-related features; the first noise feature is the information with the highest correlation with the noise-related feature selected from multiple first candidate features; the second noise feature is the information with the highest correlation with the noise-related feature selected from multiple second candidate features.

[0051] In an embodiment of the present application, the reference characteristics of the noise-related features are first determined, such as irregular small fluctuations, sudden numerical jumps, etc.; then the similarity between each first candidate feature and these reference characteristics, that is, the correlation degree, is calculated. For example, the correlation degrees of the three first candidate features are 0.7, 0.9, and 0.6, respectively, and the feature corresponding to the highest correlation degree of 0.9 is selected as the first noise feature; at the same time, the correlation degree of each second candidate feature and the reference characteristics is calculated. For example, the correlation degrees of the three second candidate features are 0.8, 0.6, and 0.95, respectively, and the feature corresponding to the highest correlation degree of 0.95 is selected as the second noise feature.

[0052] Step 135: Assign weights to the first noise feature and the second noise feature according to their noise characterization capabilities and merge them to generate a comprehensive noise feature. Based on the comprehensive noise feature, construct a noise transfer function containing environmental parameters. Based on the noise transfer function, generate preliminary noise reduction data to achieve dynamic suppression of multi-band noise.

[0053] Among them, the first noise feature and the second noise feature are the information with the highest correlation with noise, screened from the current data and the original angular velocity data respectively; the characterization capability refers to the clarity of the feature's description of the noise characteristics; the weight is the importance value assigned to each feature based on the characterization capability; the comprehensive noise feature is the information that more comprehensively reflects the noise characteristics obtained by merging the two noise features according to the weight; environmental parameters include information about the surrounding environment such as temperature and humidity; the noise transfer function is a calculation method that includes environmental parameters and noise suppression rules for different frequency bands; the preliminary noise reduction data is the data with reduced noise after processing; and multi-band noise dynamic suppression refers to the targeted real-time suppression of noise of different frequencies.

[0054] In an embodiment of the present application, weights are assigned to the noise characterization capabilities of the first noise feature and the second noise feature. For example, the first noise feature has a stronger characterization capability and is assigned a weight of 0.6, and the second noise feature is assigned a weight of 0.4. The two features are multiplied by the corresponding weights and then added to obtain a comprehensive noise feature. For example, the value of the first noise feature is 25, and the value of the second noise feature is 20. The calculation of the comprehensive noise feature is 25×0.6+20×0.4=15+8=23. Parameters such as temperature and humidity of the current environment are collected, and a noise transfer function is constructed in combination with the comprehensive noise feature. The function specifies the suppression method of noise of different frequencies. The original data is processed according to this function to generate preliminary noise reduction data to achieve dynamic suppression of noise in different frequency bands.

[0055] For example, when device A is operating, the current data generated by the gyroscope is passed to the first processing path of the dual-channel computing model, and the measured raw angular velocity data is passed to the second processing path of the model. Both paths start processing simultaneously. In the first processing path, feature extraction is performed on the current data three times, and candidate features representing fluctuation amplitude, frequency change, and duration are obtained in sequence. The correlations between these features and noise-related features are calculated to be 0.72, 0.89, and 0.65, respectively. The feature corresponding to 0.89 is selected as the first noise feature. In the second processing path, feature extraction is performed on the raw angular velocity data three times, and candidate features representing velocity magnitude, speed of change, and stability are obtained in sequence. The correlations between these features and noise-related features are calculated to be 0.85, 0.71, and 0.93, respectively. The feature corresponding to 0.93 is selected as the second noise feature. Based on the characterization capability, weights of 0.6 and 0.4 are assigned to the first and second noise features. If the value of the first noise feature is 30 and the second is 25, the comprehensive noise feature is calculated as 30×0.6+25×0.4=18+10=28. Combined with the ambient temperature of 26°C and the humidity of 55%, a noise transfer function is constructed. The raw data is processed according to the function to generate preliminary noise reduction data.

[0056] By executing steps 131 to 135, the embodiment of the present application processes two channels of data in parallel through dual channels, thereby improving the efficiency of feature extraction; multiple feature extractions comprehensively capture the potential characteristics in the data, providing rich materials for screening noise features; selecting the noise features with the highest correlation ensures the accuracy of noise information; allocating weights according to characterization capabilities and constructing transfer functions in combination with environmental parameters achieves targeted dynamic suppression of noise in different frequency bands, and the final generated preliminary noise reduction data effectively reduces noise interference, laying a good foundation for subsequent data optimization.

[0057] In one possible embodiment, step 135 assigns weights to the first noise feature and the second noise feature according to their noise characterization capabilities and combines them to generate a comprehensive noise feature, constructs a noise transfer function including environmental parameters based on the comprehensive noise feature, and generates preliminary noise reduction data based on the noise transfer function to achieve dynamic multi-band noise suppression, including: a1. Generate a first noise characteristic value corresponding to the first noise characteristic and a second noise characteristic value corresponding to the second noise characteristic, and assign weights to the first noise characteristic value and the second noise characteristic value respectively based on noise characterization capability.

[0058] Among them, the first noise eigenvalue is a numerical value indicating the strength of the first noise feature, which is used to quantify the first noise feature; the second noise eigenvalue is a numerical value indicating the strength of the second noise feature, which is used to quantify the second noise feature; the characterization ability of noise refers to the clarity of the eigenvalue in describing the noise characteristics; the weight is the importance value assigned to the eigenvalue based on the characterization ability. The stronger the characterization ability, the greater the weight.

[0059] In the embodiment of the present application, the first and second noise features are first converted into corresponding first and second noise characteristic values. For example, 20 represents the fluctuation strength of the first noise feature, and 15 represents the sudden change strength of the second noise feature. Next, weights are assigned based on the characterization capability. If the first eigenvalue describes the noise more clearly, a weight of 0.6 is assigned, and a weight of 0.4 is assigned to the second eigenvalue. For example, in device A, the first eigenvalue better reflects the noise pattern, so it has a higher weight.

[0060] a2. Calculate the first noise characteristic value and the second noise characteristic value with the corresponding weights respectively to obtain the corresponding comprehensive noise characteristic value, and combine the comprehensive noise characteristic values ​​of all preset dimensions to form a comprehensive noise feature.

[0061] Among them, the comprehensive noise characteristic value is the product of the first noise characteristic value and the corresponding weight plus the product of the second noise characteristic value and the corresponding weight; the preset dimensions are pre-set to describe different aspects of noise (such as amplitude, frequency, duration, etc.); the comprehensive noise feature is the information formed by combining the comprehensive noise characteristic values ​​of all preset dimensions, which can comprehensively reflect the characteristics of the noise.

[0062] In this embodiment of the present application, the comprehensive noise eigenvalue for a single preset dimension is first calculated. For example, the first eigenvalue is 25 × 0.6 = 15, and the second eigenvalue is 20 × 0.4 = 8. The sum of 23 is the comprehensive value for that dimension. The above calculation is then repeated for each preset dimension to obtain multiple comprehensive values. Finally, these comprehensive values ​​are combined to form a comprehensive noise signature, such as the comprehensive value of the amplitude, frequency, and duration dimensions in device A, to obtain comprehensive noise information.

[0063] a3. Collect the temperature and humidity parameters of the current environment as environmental parameters. Based on the environmental parameters and the comprehensive noise characteristics, construct a noise transfer function that includes noise suppression coefficients corresponding to different frequency bands.

[0064] Among them, environmental parameters are the temperature and humidity data of the current environment, reflecting the impact of the environment on noise. Different frequency bands are noise categories divided by frequency range, including low frequency 0-100Hz, medium frequency 100-1000Hz, and high frequency above 1000Hz. The noise suppression coefficient is a numerical value corresponding to each frequency band, which is used to determine the degree of noise reduction in that frequency band. Each frequency band corresponds to a coefficient, and the larger the coefficient, the stronger the reduction. The noise transfer function is an expression that describes the relationship between noise frequency, environmental parameters, comprehensive noise characteristics, and suppression coefficient. The expression is G(f,T,H,F)=k(f)×g(T,H,F), where G(f,T,H,F) is the noise transfer function, f is the noise frequency, T is the temperature, H is the humidity, F is the comprehensive noise characteristics, k(f) is the suppression coefficient corresponding to frequency f, and g(T,H,F) is a correction function based on environmental parameters and comprehensive noise characteristics.

[0065] In this embodiment, the ambient temperature and humidity are first collected, for example, for device A, where the ambient temperature is 25°C and the humidity is 60%. Next, the suppression coefficients for low, medium, and high frequencies are calculated, combining these environmental parameters with the comprehensive noise characteristics. For example, the values ​​are 0.3 for low frequency, 0.5 for medium frequency, and 0.7 for high frequency. Finally, a noise transfer function is constructed, which maps these frequency bands to the coefficients. The function outputs the suppression coefficient for the corresponding frequency based on the input parameters.

[0066] a4. Based on the noise transfer function, apply the corresponding noise suppression coefficient to each frequency band to generate preliminary noise reduction data.

[0067] The noise transfer function is an expression containing frequency bands and corresponding suppression coefficients; frequency bands are noise categories divided by frequency (low frequency, medium frequency, high frequency); the noise suppression coefficient is the value of attenuating the noise in the corresponding frequency band; and the preliminary noise reduction data is the noise-reduced data obtained by applying the suppression coefficient to the original data.

[0068] In this embodiment, the frequency bands of low-frequency, mid-frequency, and high-frequency noise in the original data are first determined based on the noise transfer function. Next, corresponding suppression coefficients are applied to the noise in each frequency band, such as using a coefficient of 0.2 for low-frequency, 0.4 for mid-frequency, and 0.6 for high-frequency. Finally, the processed data from each frequency band is integrated to generate preliminary noise reduction data, such as the combined processed low-frequency, mid-frequency, and high-frequency data from device A, to obtain the noise reduction result.

[0069] For example, in device A, the first noise signature is converted to 25, and the second noise signature is converted to 20. After evaluation, weights of 0.6 and 0.4 are assigned to each. The integrated noise signature values ​​for each preset dimension are calculated: 25 × 0.6 + 20 × 0.4 = 15 + 8 = 23 for the amplitude dimension, and 18 × 0.6 + 12 × 0.4 = 10.8 + 4.8 = 15.6 for the frequency dimension. These values ​​are combined to form the integrated noise signature. The data is collected at an ambient temperature of 26°C and a humidity of 55%. Combined with the integrated noise signature, the suppression coefficients for low frequencies (0-100Hz), mid frequencies (100-1000Hz), and high frequencies (above 1000Hz) are calculated as 0.2, 0.4, and 0.6, respectively. A noise transfer function is then constructed. Finally, the corresponding coefficients are applied to each frequency band of the original data, attenuating low frequencies by 20%, mid frequencies by 40%, and high frequencies by 60%. This is then combined to generate preliminary noise reduction data.

[0070] By executing steps a1 to a4, the embodiment of the present application achieves differentiated processing of noise features by quantifying the noise features and assigning weights, allowing features that better reflect the noise characteristics to play a greater role; the comprehensive noise features formed by integrating multi-dimensional eigenvalues ​​comprehensively capture various aspects of noise information; the noise transfer function constructed in combination with environmental parameters clarifies the relationship between different frequency bands and corresponding suppression coefficients, and achieves targeted suppression of noise in each frequency band; the preliminary noise reduction data finally generated effectively reduces noise interference, lays the foundation for subsequent data optimization, and improves the accuracy and reliability of overall data processing.

[0071] In one possible embodiment, S14, performing variational modal decomposition on the preliminary denoised data, screening effective modal components from the preliminary denoised data after the variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, the target angular velocity data being used to implement filtering optimization of an inertial measurement unit fiber optic gyroscope, including: Step 141: Perform variational mode decomposition on the preliminary denoised data to obtain multiple data segments.

[0072] Among them, the preliminary denoised data is the data whose noise has been reduced after preliminary processing; variational mode decomposition is a method of splitting data into multiple parts, each of which has its own unique fluctuation characteristics; data segments are multiple independent data parts obtained through variational mode decomposition, each of which can reflect a specific fluctuation situation in the original data.

[0073] In an embodiment of the present application, the variational mode decomposition method is used to process the preliminary noise reduction data. This method splits the data into multiple independent data segments according to the different fluctuation characteristics in the data. For example, in device A, after processing the preliminary noise reduction data, 5 data segments are obtained, each segment corresponding to fluctuations of different frequencies or amplitudes in the original data.

[0074] Step 142: Calculate the waveform distribution characteristic value of each data segment.

[0075] Among them, the data segments are multiple independent data parts obtained after variational mode decomposition; the waveform distribution eigenvalue is a numerical value used to describe the waveform distribution in the data segment, which can reflect whether the waveform is concentrated or dispersed.

[0076] In an embodiment of the present application, the distribution state of the waveform of each data segment is analyzed, and the waveform distribution characteristic value reflecting this distribution state is obtained by calculation. For example, in device A, the five decomposed data segments are calculated, and their waveform distribution characteristic values ​​are obtained as 2.3, 1.8, 3.5, 2.7, and 1.5, respectively. These values ​​reflect the degree of concentration or dispersion of the waveform of each segment.

[0077] Step 143 : Match each waveform distribution characteristic value with a characteristic threshold range in a preset kurtosis rule, and select data segments whose waveform distribution characteristic values ​​are within the characteristic threshold range as valid modal components.

[0078] Among them, the waveform distribution eigenvalue is a numerical value that describes the waveform distribution of the data segment; the preset kurtosis rule is a pre-set standard for judging whether the data segment is valid; the characteristic threshold range is the interval of valid waveform distribution eigenvalues ​​specified in the preset kurtosis rule; the effective modal component refers to the data segment whose waveform distribution eigenvalue is within the characteristic threshold range, that is, the data part that meets the preset standard.

[0079] In an embodiment of the present application, the characteristic threshold range in the preset kurtosis rule is first determined, for example, this range is set to 1.6 to 3.0, and then the waveform distribution characteristic value of each data segment is compared with this range. If the characteristic value is within this range, the corresponding segment is selected as a valid modal component. For example, in device A, the characteristic values ​​of the five data segments are 2.3, 1.8, 3.5, 2.7, and 1.5, respectively, of which 2.3, 1.8, and 2.7 are between 1.6 and 3.0, so these three segments are selected as valid modal components.

[0080] Step 144: Combine each effective modal component to obtain target angular velocity data.

[0081] Among them, the effective modal component is the data segment that meets the preset standards after screening; the target angular velocity data is the final data obtained after combining all the effective modal components together, which is used to reflect the angular velocity of the fiber optic gyroscope of the inertial measurement unit.

[0082] In an embodiment of the present application, all the filtered effective modal components are collected and combined according to their time sequence in the original data or the correlation between the fluctuations to form a complete data sequence, that is, the target angular velocity data. For example, in device A, the three filtered effective modal components are combined according to their order in the preliminary noise reduction data to obtain the target data that can reflect the angular velocity of the gyroscope.

[0083] For example, in device A, variational modal decomposition is first performed on the preliminary noise reduction data to obtain five data segments; then, each segment is analyzed and calculated to obtain their waveform distribution eigenvalues, which are 2.1, 1.9, 3.2, 2.5, and 1.6, respectively; then, according to the characteristic threshold range of 1.7 to 3.1 in the preset kurtosis rule, these eigenvalues ​​are compared one by one, and it is found that 2.1, 1.9, and 2.5 are within this range and are determined to be valid modal components; finally, these three valid modal components are combined according to their time sequence in the original data to obtain the target angular velocity data.

[0084] By executing steps 141 to 144, the embodiment of the present application realizes a detailed analysis of the data by splitting the preliminary noise reduction data into multiple data segments, making the fluctuation characteristics of each segment clearer; calculating the waveform distribution characteristic value provides a quantitative basis for judging the validity of the data, making the screening process more objective; screening out valid modal components according to preset rules ensures that the retained data partially meets the standards; combining the valid components into target angular velocity data further improves the reliability and accuracy of the data, and provides high-quality results for the filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0085] In a possible embodiment, S12, dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength, includes: Step 121 : Divide the vibration modal characteristic data into preset frequency bands, and extract vibration frequency data and corresponding amplitude data based on each divided frequency band.

[0086] Among them, vibration modal characteristic data is data that can reflect the characteristics of vibration at different times and frequencies; the preset frequency band range is multiple pre-set frequency intervals, such as low frequency of 0-100Hz, medium frequency of 100-1000Hz, and high frequency of above 1000Hz, which are used to divide vibration frequencies; vibration frequency data is the specific frequency value of vibration in each frequency band; amplitude data is the strength data of the corresponding vibration frequency.

[0087] In the embodiment of the present application, first, the vibration modal feature data is divided into different parts according to the preset frequency band range. For example, in device A, it is divided into three ranges: 0-100Hz, 100-1000Hz, and above 1000Hz. Secondly, in each divided frequency band, the vibration frequency value contained in the frequency band and the corresponding strength data are extracted. For example, in the low frequency band, a frequency of 50Hz and an amplitude of 0.8 are extracted, and in the mid-frequency band, a frequency of 500Hz and an amplitude of 1.2 are extracted.

[0088] Step 122: Calculate the power ratio of the vibration frequency data in the vibration modal characteristic data based on the corresponding amplitude data.

[0089] Among them, amplitude data is a numerical value reflecting the strength of the vibration frequency; vibration frequency data is the specific frequency value of each frequency band; power proportion refers to the proportion of the vibration power of a certain frequency band in the total power of all frequency bands, and the power is calculated by the square of the amplitude data.

[0090] In the embodiment of the present application, first, the vibration power of each frequency band is calculated. The calculation method is to square the amplitude data of the frequency band. For example, if the amplitude of a certain frequency band is 1.2, its power is 1.2 multiplied by 1.2, which is equal to 1.44. Secondly, the power of all frequency bands is added together to obtain the total power. For example, if the power of three frequency bands is 0.64, 1.44, and 0.36 respectively, the total power is 0.64 plus 1.44 plus 0.36, which is equal to 2.44. Finally, the power proportion of each frequency band is divided by the total power to obtain the power proportion. For example, the proportion of the mid-frequency band is 1.44 divided by 2.44, which is approximately equal to 0.59.

[0091] Step 123: Determine the adjustment direction and value of the magnetic field strength based on the power ratio, and adjust the magnetic field strength based on the adjustment direction and value.

[0092] Among them, the power proportion is the proportion of the vibration power of a certain frequency band in the total power; the adjustment direction of the magnetic field strength refers to strengthening or weakening the magnetic field; the adjustment value is the specific size of the change in magnetic field strength; and the magnetic field strength is the strength of the magnetic field inside the gyroscope.

[0093] In the embodiment of this application, first, a power ratio threshold is set, such as 50%. If the power ratio of a certain frequency band exceeds this threshold, it indicates that the vibration impact of this frequency band is greater, and the magnetic field strength needs to be increased. Otherwise, it needs to be weakened. For example, in device A, the power ratio of the mid-frequency band is 59%, which exceeds the threshold and requires a stronger magnetic field. Secondly, the adjustment value is determined based on the difference between the power ratio and the threshold. The larger the difference, the greater the adjustment. For example, a difference of 9% corresponds to an adjustment value of 0.2T, and the original magnetic field strength of 1.0T is finally adjusted to 1.2T.

[0094] Step 124: Based on the adjusted magnetic field strength, the vibration sensing component inside the fiber optic gyroscope of the inertial measurement unit is controlled to produce deformation. During the deformation process, the vibration sensing component is converted into a corresponding initial voltage signal according to a predetermined conversion rule. Based on the initial voltage signal, current data is generated through processing by the signal conversion component.

[0095] Among them, the adjusted magnetic field strength is the strength of the magnetic field after adjustment; the vibration sensing component is the component inside the gyroscope that can change its shape due to vibration; the deformation is the change in the shape of the component; the initial voltage signal is the electrical signal generated during the deformation process; the signal conversion component is a device that converts the voltage signal into a current signal; and the current data is a record reflecting the strength of the electrical signal.

[0096] In the embodiments of the present application, first, the adjusted magnetic field strength causes the vibration-sensing component inside the gyroscope to change shape. The stronger the magnetic field and the more intense the vibration, the greater the deformation. For example, after the magnetic field in device A is enhanced, the component deformation amplitude changes from 0.3mm to 0.5mm. Secondly, during the deformation process, an initial voltage signal is generated according to a predetermined rule (e.g., the greater the deformation, the higher the voltage). For example, a 2V voltage signal is generated corresponding to a deformation of 0.5mm. Finally, the signal conversion component converts the initial voltage signal into current data, such as converting a 2V voltage into a 0.5A current, generating the corresponding current data.

[0097] For example, in device A, the vibration modal signature data is first divided into low frequencies (0-100Hz), medium frequencies (100-1000Hz), and high frequencies (above 1000Hz). This data is extracted for low frequencies (50Hz) (amplitude 0.8), medium frequencies (500Hz) (amplitude 1.2), and high frequencies (2000Hz) (amplitude 0.6). The power in each frequency band is calculated: 0.8 times 0.8 = 0.64 for the low frequency, 1.2 times 1.2 = 1.44 for the medium frequency, and 0.6 times 0.6 = 0.36 for the high frequency. The total power is 0.64 + 1.44 + 0.36 = 2.44. The power contribution of the medium frequency band is 1.44 divided by 2.44, which is approximately 0.59. Because this contribution exceeds the preset 50% threshold, the magnetic field strength is increased by 0.2T from 1.0T to 1.2T. The enhanced magnetic field causes the vibration sensing component to deform to 0.5mm, generating a 2V initial voltage signal, which is then processed by the signal conversion component to generate 0.5A current data.

[0098] By executing steps 121 to 124, the embodiment of the present application can accurately locate vibration conditions of different frequencies by dividing the vibration data according to preset frequency bands; calculating the power ratio can clarify the degree of influence of vibration in each frequency band, providing a basis for adjusting the magnetic field; adjusting the magnetic field strength according to the power ratio can specifically deal with the influence of major vibrations; converting deformation into current data realizes the effective capture of vibration energy, provides reliable data for subsequent analysis, and helps to improve the stability and measurement accuracy of the gyroscope.

[0099] In a possible embodiment, step 123, determining an adjustment direction and value of the magnetic field strength based on the power ratio, and adjusting the magnetic field strength based on the adjustment direction and value, includes: b1. Preset corresponding upper and lower limits for the vibration frequency data of each frequency band, and calculate the difference between the vibration frequency data and the upper or lower limit based on the magnitude of the vibration frequency data to obtain a first deviation value for each frequency band.

[0100] Among them, frequency band is the vibration category divided by frequency range, such as low frequency, medium frequency, and high frequency; vibration frequency data is the specific frequency value of vibration in each frequency band; upper limit value and lower limit value are the highest value and lowest value of the reasonable frequency range pre-set for each frequency band; the first deviation value is the difference between the vibration frequency data and the upper limit value or lower limit value. When the vibration frequency data is greater than the upper limit value, the first deviation value is the vibration frequency data minus the upper limit value. When it is less than the lower limit value, it is the lower limit value minus the vibration frequency data. If it is between the upper and lower limits, the first deviation value is 0.

[0101] In this embodiment of the present application, corresponding upper and lower limits are set for the vibration frequency data of each frequency band. For example, in device A, the upper limit of the low frequency band (0-100Hz) is set to 90Hz and the lower limit is set to 10Hz, while the upper limit of the mid-frequency band (100-1000Hz) is set to 900Hz and the lower limit is set to 200Hz. The vibration frequency data of each frequency band is compared with the upper and lower limits, and a first deviation value is calculated. If the low frequency band frequency is 100Hz (greater than the upper limit of 90Hz), the deviation value is 100-90=10; if the mid-frequency band frequency is 150Hz (less than the lower limit of 200Hz), the deviation value is 200-150=50; if the frequency is within the range, the deviation value is 0.

[0102] b2. Based on the power proportion of each frequency band, assign a corresponding weight coefficient to each frequency band.

[0103] Among them, the power share is the proportion of the vibration power of a certain frequency band in the total power; the weight coefficient is a numerical value indicating the importance assigned to each frequency band based on the power share. The larger the power share, the larger the weight coefficient, which is used to reflect the influence of the frequency band in the overall adjustment.

[0104] In this embodiment, the power contribution of each frequency band is determined. For example, in device A, the power contributions of low frequency, medium frequency, and high frequency are 10%, 70%, and 20%, respectively. Weight coefficients are assigned based on the power contribution, with higher weights being assigned to higher weights. For example, 0.1 is assigned to low frequency, 0.7 to medium frequency, and 0.2 to high frequency, giving greater weight to frequency bands with greater influence in subsequent calculations.

[0105] b3. Calculate the accumulated result of multiplying the first deviation value of each frequency band by the weight coefficient to obtain a second deviation value, and determine the adjustment direction according to the positive or negative value of the second deviation value.

[0106] Among them, the first deviation value is the difference between the vibration frequency data and the upper and lower limits; the weight coefficient is a numerical value indicating the importance of the frequency band; the second deviation value is the sum of the first deviation value of each frequency band multiplied by the corresponding weight coefficient; the adjustment direction refers to strengthening or weakening the magnetic field. When the second deviation value is positive, the adjustment direction is strengthening, and when it is negative, it is weakening.

[0107] In this embodiment, the product of the first deviation value and the corresponding weight coefficient for each frequency band is calculated. For example, the low-frequency deviation value for device A is 10 × 0.1 = 1, the mid-frequency deviation value is 50 × 0.7 = 35, and the high-frequency deviation value is 0 × 0.2 = 0. Adding these products yields the second deviation value of 1 + 35 + 0 = 36. Since this value is positive, the adjustment direction is determined to be enhancement.

[0108] b4. Determine an adjustment value of the magnetic field strength based on the product of the second deviation value and the preset adjustment coefficient.

[0109] Among them, the second deviation value is the sum of the deviation values ​​of each frequency band and the product of the weight; the preset adjustment coefficient is a pre-set fixed value used to convert the second deviation value into the magnetic field adjustment amplitude; the adjustment value is the specific size of the change in magnetic field intensity, which is obtained by multiplying the second deviation value by the preset adjustment coefficient.

[0110] In the embodiment of the present application, a preset adjustment coefficient (such as 0.01T) is determined. For example, the preset coefficient in device A is 0.01T. The product of the second deviation value and the preset coefficient is calculated, that is, 36×0.01T=0.36T, and the adjustment value is 0.36T.

[0111] b5. Based on the adjustment direction and the adjustment value, the initial magnetic field strength value is calculated to obtain the adjusted magnetic field strength.

[0112] Among them, the adjustment direction is to strengthen or weaken the magnetic field; the adjustment value is the specific size of the change in magnetic field strength; the initial magnetic field strength is the strength of the magnetic field before adjustment; the adjusted magnetic field strength is the result of the initial magnetic field strength calculated according to the adjustment direction and value. When it is strengthened, it is the initial value plus the adjustment value, and when it is weakened, it is the initial value minus the adjustment value.

[0113] In the embodiment of the present application, the initial magnetic field strength (e.g., 1.0 T) is specified. For example, the initial strength of device A is 1.0 T. Based on the adjustment direction (enhancement) and the value (0.36 T), the adjusted magnetic field strength is calculated as 1.0 T + 0.36 T = 1.36 T.

[0114] For example, in device A, the low frequency band (0-100Hz) is set to an upper limit of 80Hz and a lower limit of 20Hz; the mid-frequency band (100-1000Hz) is set to an upper limit of 800Hz and a lower limit of 300Hz; and the high frequency band (1000Hz and above) is set to an upper limit of 1500Hz and a lower limit of 1200Hz. For a low-frequency vibration frequency of 90Hz (greater than the upper limit), the first deviation value is 90-80 = 10; for a mid-frequency of 250Hz (less than the lower limit), the first deviation value is 300-250 = 50; and for a high-frequency of 1300Hz (within the range), the deviation value is 0. Based on power proportions of 10%, 70%, and 20%, weights of 0.1, 0.7, and 0.2 are assigned. The calculated value for the low frequency is 10 × 0.1 = 1, the mid-frequency is 50 × 0.7 = 35, and the high frequency is 0 × 0.2 = 0. The total of 36 is the second deviation value (positive, indicating an enhancement). The preset adjustment coefficient is 0.01T, and the adjustment value = 36×0.01=0.36T; the initial magnetic field strength is 1.0T, and after adjustment it is 1.0+0.36=1.36T.

[0115] By executing b1 to b5, the embodiment of the present application sets upper and lower limits to quantify the deviation of the vibration frequency from a reasonable range, and allocates weights based on the power ratio, so that frequency bands with greater influence are given more attention during adjustment; the direction and amplitude of the magnetic field adjustment are determined by calculating the deviation value, thereby achieving precise adjustment of the magnetic field strength, making the adjusted magnetic field more adapted to the actual vibration conditions, improving the device's adaptability to different vibrations, and providing better magnetic field conditions for subsequent energy conversion.

[0116] In one possible embodiment, step 124, based on the adjusted magnetic field strength, controls the vibration sensing component inside the fiber optic gyroscope of the inertial measurement unit to generate deformation, converts the deformation into a corresponding initial voltage signal according to a predetermined conversion rule, and generates current data based on the initial voltage signal through processing by the signal conversion component, including: c1. Based on the deformation, determine the change in magnetic flux in the vibration sensing component, and the change in magnetic flux and the deformation amount satisfy a preset functional relationship.

[0117] Among them, deformation is the change in shape of the vibration sensing component caused by vibration; the deformation variable is the specific size of the deformation; the change in magnetic flux is the change in the amount of magnetic field passing through the vibration sensing component; the preset functional relationship is a formula that describes the relationship between the change in magnetic flux and the deformation variable, which is usually a linear relationship, expressed as ΔΦ=k×x, where ΔΦ represents the change in magnetic flux, k is the proportional coefficient, and x is the deformation variable. The larger the deformation variable, the greater the change in magnetic flux.

[0118] In an embodiment of the present application, the deformation of the vibration sensing component is determined. For example, the deformation in device A is 0.5 mm. The change in magnetic flux is calculated according to the preset functional relationship ΔΦ=k×x. If k=2, then ΔΦ=2×0.5=1, that is, the change in magnetic flux is 1 unit.

[0119] c2. Based on the change in magnetic flux, an induced electromotive force is generated in the coil.

[0120] Among them, the change in magnetic flux is the change in the amount of magnetic field passing through the component; the coil is a ring-shaped component wound with wire; the induced electromotive force is the electrical signal generated in the coil when the magnetic flux changes, and the expression is E=N×ΔΦ / Δt, where E is the induced electromotive force, N is the number of coil turns, ΔΦ is the change in magnetic flux, and Δt is the time taken for the change. The more turns there are and the faster the magnetic flux changes, the greater the induced electromotive force.

[0121] In an embodiment of the present application, the induced electromotive force is calculated based on the change in magnetic flux, the number of coil turns and the change time. For example, in device A, the number of coil turns N = 100, the change in magnetic flux ΔΦ = 1, and the change time Δt = 0.1 second. According to E = 100 × 1 / 0.1 = 1000, the induced electromotive force is 1000 units.

[0122] c3. The induced electromotive force is converted into an initial voltage signal by a rectifier according to a predetermined conversion rule. Based on the initial voltage signal, the high-frequency fluctuation component is filtered out by a filter device to obtain a stable target voltage signal.

[0123] Among them, the rectifier is a device that converts the induced electromotive force into a stable voltage signal; the initial voltage signal is the preliminary electrical signal obtained after rectification; high frequency refers to a signal with a frequency higher than a preset value (such as 1000Hz); the high-frequency fluctuation component is the high-frequency unstable part contained in the initial voltage signal; the filtering device is a device that removes the high-frequency fluctuation component; the target voltage signal is the stable voltage signal obtained after filtering.

[0124] In an embodiment of the present application, the induced electromotive force is converted into an initial voltage signal by a rectifier device. For example, the induced electromotive force of 1800 units in device A is rectified to obtain a 10V initial voltage signal, which contains a high-frequency fluctuation component of 2000Hz. These components are then removed by a filtering device to obtain a stable target voltage signal of 9V.

[0125] c4. Convert the target voltage signal into current data through a current conversion device.

[0126] Among them, the current conversion device is a device that converts a voltage signal into a current signal; the target voltage signal is a stabilized voltage signal after filtering; and the current data is a record output by the current conversion device that reflects the current size.

[0127] In an embodiment of the present application, the target voltage signal is input into the current conversion device and converted according to a preset voltage-current correspondence. For example, if the target voltage signal in device A is 11V, the conversion relationship is 1V corresponding to 0.1A, and 11×0.1=1.1A is calculated to generate current data 1.1A.

[0128] For example, in device A, the vibration-sensing component experiences a 0.3mm deformation due to vibration. According to the preset function relationship ΔΦ = 3 × x, the change in magnetic flux is 3 × 0.3 = 0.9 units. This change generates an induced electromotive force in a 200-turn coil over 0.1 seconds. According to E = 200 × 0.9 / 0.1 = 1800 units, this is converted by a rectifier into a 12V initial voltage signal, which contains a 2000Hz high-frequency fluctuation component. After processing by a filter, a stable 11V target voltage signal is obtained. This is then converted by a current converter using the relationship of 1V to 0.2A: 11 × 0.2 = 2.2A, resulting in a final current of 2.2A.

[0129] By executing c1 to c4, the embodiment of the present application realizes the conversion of vibration energy into electrical signals by converting the deformation caused by vibration into changes in magnetic flux, and then further converting it into electrical signals; after rectification and filtering, the unstable components in the signal are removed to obtain a stable voltage signal; finally, the voltage signal is converted into current data, providing a stable and reliable basis for subsequent noise analysis. The entire process realizes the effective capture and conversion of vibration information.

[0130] Figure 2 A schematic diagram of the structure of a filter optimization system for an inertial measurement unit fiber optic gyroscope provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes: The acquisition module 21 is used to acquire the three-dimensional vibration signal of the fiber optic gyroscope of the inertial measurement unit, perform short-time Fourier transform on the three-dimensional vibration signal, and generate vibration modal characteristic data.

[0131] The adjustment module 22 is used to dynamically adjust the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration mode characteristic data, and convert the environmental vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength.

[0132] The fusion module 23 is used to extract the first noise feature and the second noise feature from the current data and the raw angular velocity data output by the inertial measurement unit fiber optic gyroscope based on the current data by adopting a dual-channel convolutional neural network, and fuse the first noise feature and the second noise feature to output preliminary noise reduction data.

[0133] The optimization module 24 is used to perform variational modal decomposition on the preliminary denoised data, and to filter effective modal components from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data. The target angular velocity data is used to implement filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0134] Figure 2 The filter optimization system of the inertial measurement unit fiber optic gyroscope can perform Figure 1 The implementation principles and technical effects of the filter optimization method for an inertial measurement unit fiber optic gyroscope described in the illustrated embodiment will not be elaborated on here. The specific manner in which the various modules and units of the filter optimization system for an inertial measurement unit fiber optic gyroscope in the above-mentioned embodiment perform operations has been described in detail in the embodiments of the method and will not be elaborated on here.

[0135] In one possible design, Figure 2 The filter optimization system for the fiber optic gyroscope of the inertial measurement unit of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0136] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0137] The processing component 32 is used to execute the following process: collecting three-dimensional vibration signals from the fiber optic gyroscope of the inertial measurement unit, performing short-time Fourier transform on the three-dimensional vibration signals, and generating vibration modal characteristic data; dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the environmental vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength; based on the current data and combined with the original angular velocity data output by the fiber optic gyroscope of the inertial measurement unit, using a dual-channel convolutional neural network, extracting a first noise feature and a second noise feature from the current data and the original angular velocity data respectively, and fusing the first noise feature and the second noise feature to output preliminary noise reduction data; performing variational mode decomposition on the preliminary noise reduction data, and screening effective modal components from the preliminary noise reduction data after the variational mode decomposition according to a preset kurtosis rule to obtain target angular velocity data, and the target angular velocity data is used to realize filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

[0138] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0139] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0140] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0141] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0142] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0143] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0144] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The filter optimization method of the fiber optic gyroscope of the inertial measurement unit of the embodiment shown is shown.

[0145] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0147] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A filtering optimization method for an inertial measurement unit fiber optic gyroscope, characterized in that: include: Collecting three-dimensional vibration signals from an inertial measurement unit fiber optic gyroscope, performing short-time Fourier transform on the three-dimensional vibration signals, and generating vibration modal characteristic data; Dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength; Based on the current data and raw angular velocity data output by an inertial measurement unit fiber optic gyroscope, a dual-channel convolutional neural network is used to extract a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, and the first noise feature and the second noise feature are fused to output preliminary noise reduction data; Variational modal decomposition is performed on the preliminary denoised data, and effective modal components are screened from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data. The target angular velocity data is used to achieve filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

2. The method according to claim 1, characterized in that The method comprises: extracting a first noise feature and a second noise feature from the current data and the raw angular velocity data output by the fiber optic gyroscope of the inertial measurement unit based on the current data and combined with the raw angular velocity data output by the fiber optic gyroscope of the inertial measurement unit by adopting a dual-channel convolutional neural network, fusing the first noise feature and the second noise feature to output preliminary noise reduction data, including: Transmitting the current data into a first processing path in the dual-channel convolutional neural network, and transmitting the raw angular velocity data into a second processing path in the dual-channel convolutional neural network, wherein the first processing path and the second processing path are run simultaneously; Based on the first processing path, performing multiple feature extractions on the current data to obtain multiple first candidate features; Based on the second processing path, performing multiple feature extractions on the original angular velocity data to obtain multiple second candidate features; Extracting a first noise feature with the highest correlation with the noise-related feature from the plurality of first candidate features, and extracting a second noise feature with the highest correlation with the noise-related feature from the plurality of second candidate features; The first noise feature and the second noise feature are weighted according to their noise characterization capabilities and combined to generate a comprehensive noise feature. Based on the comprehensive noise feature, a noise transfer function including environmental parameters is constructed. Based on the noise transfer function, preliminary noise reduction data is generated to achieve dynamic suppression of multi-band noise.

3. The method according to claim 2, characterized in that The first noise feature and the second noise feature are weighted according to their noise characterization capabilities and combined to generate a comprehensive noise feature; a noise transfer function including environmental parameters is constructed based on the comprehensive noise feature; and preliminary noise reduction data is generated based on the noise transfer function to achieve dynamic multi-band noise suppression, including: generating a first noise characteristic value corresponding to the first noise characteristic and a second noise characteristic value corresponding to the second noise characteristic, and assigning weights to the first noise characteristic value and the second noise characteristic value respectively based on a noise characterization capability; Calculating the first noise characteristic value and the second noise characteristic value with corresponding weights respectively to obtain corresponding comprehensive noise characteristic values, and combining the comprehensive noise characteristic values ​​of all preset dimensions to form a comprehensive noise feature; collecting temperature and humidity parameters of the current environment as environmental parameters, and constructing a noise transfer function including noise suppression coefficients corresponding to different frequency bands based on the environmental parameters and the comprehensive noise characteristics; Based on the noise transfer function, a corresponding noise suppression coefficient is applied to each frequency band to generate preliminary noise reduction data.

4. The method according to claim 1, wherein The method further comprises performing variational modal decomposition on the preliminary denoised data, screening effective modal components from the preliminary denoised data after variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, wherein the target angular velocity data is used to implement filtering optimization of an inertial measurement unit fiber optic gyroscope, and comprising: Perform variational mode decomposition on the preliminary denoised data to obtain multiple data segments; Calculating a waveform distribution characteristic value of each of the data segments; Matching each waveform distribution characteristic value with a characteristic threshold range in a preset kurtosis rule, and screening data segments whose waveform distribution characteristic values ​​are within the characteristic threshold range as valid modal components; Each of the effective modal components is combined to obtain target angular velocity data.

5. The method according to claim 1, wherein The method of dynamically adjusting the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and converting the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength, includes: Dividing the vibration modal characteristic data into preset frequency bands, and extracting vibration frequency data and corresponding amplitude data based on each divided frequency band; Calculating, based on the corresponding amplitude data, a power ratio of the vibration frequency data in the vibration modal characteristic data; Determining an adjustment direction and a value of the magnetic field strength based on the power ratio, and adjusting the magnetic field strength based on the adjustment direction and the value; Based on the adjusted magnetic field strength, the vibration sensing component inside the fiber optic gyroscope of the inertial measurement unit is controlled to produce deformation. During the deformation process, it is converted into a corresponding initial voltage signal according to a predetermined conversion rule. Based on the initial voltage signal, current data is generated through processing by a signal conversion component.

6. The method according to claim 5, characterized in that The determining an adjustment direction and a value of the magnetic field strength based on the power ratio, and adjusting the magnetic field strength based on the adjustment direction and the value, includes: Presetting corresponding upper and lower limits for the vibration frequency data of each frequency band, and calculating, based on the magnitude of the vibration frequency data, a difference between the vibration frequency data and the upper or lower limit to obtain a first deviation value for each frequency band; Based on the power proportion of each frequency band, assigning a corresponding weight coefficient to each frequency band; Calculating a second deviation value by multiplying the first deviation value of each frequency band by the weight coefficient and adding the results together, and determining an adjustment direction according to the positive or negative value of the second deviation value; determining an adjustment value of the magnetic field strength based on a product of the second deviation value and a preset adjustment coefficient; Based on the adjustment direction and the adjustment value, the initial magnetic field strength value is calculated to obtain the adjusted magnetic field strength.

7. The method according to claim 5, characterized in that The converting into a corresponding initial voltage signal according to a predetermined conversion rule, and generating current data based on the initial voltage signal through processing by a signal conversion component, includes: Determining a change in magnetic flux within the vibration sensing component based on the deformation, wherein the change in magnetic flux and the deformation amount satisfy a preset functional relationship; Based on the change in the magnetic flux, an induced electromotive force is generated in the coil; The induced electromotive force is converted into an initial voltage signal by a rectifier according to a predetermined conversion rule, and a high-frequency fluctuation component is filtered out by a filter device based on the initial voltage signal to obtain a stable target voltage signal; The target voltage signal is converted into current data by a current conversion device.

8. A filter optimization system for an inertial measurement unit fiber optic gyroscope, characterized in that: include: An acquisition module is used to acquire a three-dimensional vibration signal from an inertial measurement unit fiber optic gyroscope, perform short-time Fourier transform on the three-dimensional vibration signal, and generate vibration modal characteristic data; an adjustment module, configured to dynamically adjust the magnetic field strength of the fiber optic gyroscope of the inertial measurement unit according to the vibration modal characteristic data, and convert the ambient vibration energy of the fiber optic gyroscope of the inertial measurement unit into current data based on the adjusted magnetic field strength; a fusion module, configured to extract, based on the current data and raw angular velocity data output by an inertial measurement unit fiber optic gyroscope, a first noise feature and a second noise feature from the current data and the raw angular velocity data, respectively, by employing a dual-channel convolutional neural network, and fuse the first noise feature with the second noise feature to output preliminary noise reduction data; An optimization module is configured to perform variational modal decomposition on the preliminary denoised data, and to filter effective modal components from the preliminary denoised data after the variational modal decomposition according to a preset kurtosis rule to obtain target angular velocity data, wherein the target angular velocity data is used to implement filtering optimization of the fiber optic gyroscope of the inertial measurement unit.

9. A computing device, characterized in that The invention comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the filtering optimization method of the fiber optic gyroscope of the inertial measurement unit according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the filtering optimization method of the fiber optic gyroscope of the inertial measurement unit according to any one of claims 1 to 7 is implemented.

Images