CN113739779A - Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network - Google Patents
Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network Download PDFInfo
- Publication number
- CN113739779A CN113739779A CN202111013325.XA CN202111013325A CN113739779A CN 113739779 A CN113739779 A CN 113739779A CN 202111013325 A CN202111013325 A CN 202111013325A CN 113739779 A CN113739779 A CN 113739779A
- Authority
- CN
- China
- Prior art keywords
- temperature
- module
- compensation
- signal
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 49
- 238000000605 extraction Methods 0.000 claims description 47
- 238000012549 training Methods 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 24
- 238000010606 normalization Methods 0.000 claims description 20
- 210000002569 neuron Anatomy 0.000 claims description 18
- 230000008859 change Effects 0.000 claims description 15
- 230000009467 reduction Effects 0.000 claims description 15
- 238000009529 body temperature measurement Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000002474 experimental method Methods 0.000 claims description 4
- 101100127891 Caenorhabditis elegans let-4 gene Proteins 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000003466 welding Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims description 2
- 230000007774 longterm Effects 0.000 abstract description 4
- 239000000284 extract Substances 0.000 description 3
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013016 damping Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 229910052751 metal Inorganic materials 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- BASFCYQUMIYNBI-UHFFFAOYSA-N platinum Chemical compound [Pt] BASFCYQUMIYNBI-UHFFFAOYSA-N 0.000 description 2
- 239000010453 quartz Substances 0.000 description 2
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 2
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 239000000919 ceramic Substances 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 229910052697 platinum Inorganic materials 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 229910052710 silicon Inorganic materials 0.000 description 1
- 239000010703 silicon Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C19/00—Gyroscopes; Turn-sensitive devices using vibrating masses; Turn-sensitive devices without moving masses; Measuring angular rate using gyroscopic effects
- G01C19/56—Turn-sensitive devices using vibrating masses, e.g. vibratory angular rate sensors based on Coriolis forces
- G01C19/567—Turn-sensitive devices using vibrating masses, e.g. vibratory angular rate sensors based on Coriolis forces using the phase shift of a vibration node or antinode
- G01C19/5691—Turn-sensitive devices using vibrating masses, e.g. vibratory angular rate sensors based on Coriolis forces using the phase shift of a vibration node or antinode of essentially three-dimensional vibrators, e.g. wine glass-type vibrators
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C19/00—Gyroscopes; Turn-sensitive devices using vibrating masses; Turn-sensitive devices without moving masses; Measuring angular rate using gyroscopic effects
- G01C19/56—Turn-sensitive devices using vibrating masses, e.g. vibratory angular rate sensors based on Coriolis forces
- G01C19/5776—Signal processing not specific to any of the devices covered by groups G01C19/5607 - G01C19/5719
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to a multi-element temperature compensation system and a method for a hemispherical resonance gyroscope based on a BP (back propagation) neural network. And embedding the compensation model into a gyro control program mode to realize online compensation of gyro output in a temperature-varying environment. The invention can effectively improve the temperature characteristic of the gyroscope and improve the zero offset stability of the gyroscope during long-term operation.
Description
Technical Field
The invention belongs to the technical field of inertial instrument control, and particularly relates to a multi-element temperature compensation system and method of a hemispherical resonant gyroscope based on a BP neural network.
Background
The resonance gyroscope comprises a quartz hemispherical resonance gyroscope, a metal cylindrical resonance gyroscope, a nested ring gyroscope, a micro hemispherical gyroscope and the like, is a solid wave principle gyroscope with long service life, high reliability and high precision, and has the tendency of replacing various optical gyroscopes. The harmonic oscillator of the core sensitive element has the non-ideal characteristics of uneven mass and rigidity distribution, defects and the like due to factors such as imperfect materials, processing and processes; meanwhile, the geometry and physical properties of the harmonic oscillator are affected by the external environment, long-term drift such as gain error and cross damping error is generated, and the working state and the gyro precision are affected.
In the force feedback working mode, the position of the standing wave is maintained to be constant by applying a driving signal in a sensitive mode. However, the two-mode coupling in-phase error component caused by cross damping in the loop response signal is the same as the phase of the signal generated by the external angular motion precession effect, so that the error is difficult to be removed from the gyro output signal. When the gyroscope is in a vibration environment or a temperature change environment, the in-phase error component drifts, so that the zero fluctuation of the gyroscope is caused, and the zero offset stability of the gyroscope is reduced. Usually, the error is compensated by adopting a modeling and calibrating mode, however, the zero offset of the gyroscope has correlation with parameters such as temperature and temperature gradient, and the coupling is usually represented as nonlinearity, and an ideal compensation effect is difficult to achieve by depending on a traditional polynomial modeling mode.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a system and a method for compensating the multi-element temperature of a hemispherical resonance gyroscope based on a BP neural network, and can effectively improve the temperature characteristic of the gyroscope and improve the zero offset stability of the gyroscope during long-term operation.
The technical problem to be solved by the invention is realized by adopting the following technical scheme:
a multi-element temperature compensation system of a hemispherical resonance gyroscope based on a BP neural network comprises a harmonic oscillator, an electrode, a base, a shell, a buffer amplifier, an analog-to-digital converter, a signal extraction module, a temperature measurement element, a temperature measurement circuit, a normalization module, a compensation model module, an error compensation module, an output estimation module, a signal noise reduction module, a feature extraction module and a model training module, wherein the electrode is connected with the harmonic oscillator and used for driving and checking the vibration of the harmonic oscillator, the harmonic oscillator is fixedly arranged on the base, the shell is fixed on the base through laser welding to form a vacuum closed space, the temperature measurement element is arranged on the base, the electrode, the buffer amplifier, the analog-to-digital converter, the signal extraction module, the error compensation module and the output estimation module are connected in series, the temperature measurement module, the temperature measurement circuit, the normalization module, the compensation model module and the error compensation module are connected in series, the signal extraction module, the normalization module, the feature extraction module, the model training module, the feature extraction module, the error compensation module and the output estimation module are connected in series, The signal noise reduction module, the feature extraction module and the model training module are connected in series;
the temperature measurement circuit scans the full temperature range, records the gyro frequency, the temperature gradient, the temperature change rate and the gyro output information by the temperature measurement circuit, the buffer amplifier, the analog-to-digital converter, the signal extraction module and the normalization module, and repeats the experiment to obtain a plurality of groups of data samples;
the signal denoising module adopts wavelet denoising smoothing or filtering means, for example, to denoise original data, improve the signal-to-noise ratio and improve the calculation precision of subsequent training links;
the characteristic extraction module analyzes the correlation degree of the model variable and the gyro drift by adopting a lifting dimension method, performs parameter optimization and eliminates weak correlation items;
the model training module sends the optimized data sample into a configured BP neural network training set, and calculates a model weight;
the signal extraction module acquires each error signal in real time and resolves and normalizes the error signal into a training set range;
the model training module substitutes the normalized error signal into a BP neural network compensation model, and calculates real-time compensation data of the gyroscope under the current working condition to form a compensation signal;
the error compensation module fuses the gyro original signal and the compensation signal, estimates the current external input angular velocity according to the scale factor of the gyro, realizes error compensation, and sends the compensated output to a user through the output estimation module.
The compensation method of the hemispherical resonant gyro multi-element temperature compensation system based on the BP neural network comprises the following steps:
step 1, acquiring temperature information of a gyroscope and temperature information of a base through a temperature measuring element and a temperature measuring circuit;
step 5, establishing a BP neural network approximation structure, and carrying out BP neural network training in a model training module by using the feature extraction data;
and 6, carrying out neural network compensation in a compensation model module through the established BP neural network approximation structure.
Moreover, the specific implementation method of the step 1 is as follows: under different temperatures, the analog-digital converter obtains the real-time frequency coefficient of the harmonic oscillator through the electrodes, calculates the temperature of the sensitive area of the harmonic oscillator according to the obtained frequency coefficient, detects the temperature of the base through the temperature measuring element, transmits a temperature signal to the temperature measuring circuit, and converts the temperature signal into an electric signal which is input to the normalization module.
Moreover, the specific implementation method of the step 2 is as follows: the analog-to-digital converter acquires detection information of two modes of an X axis and a Y axis of the harmonic oscillator through the electrodes, sequentially inputs the detection information into the signal extraction module and the normalization module for resolving to obtain a gyroscope sensitive output and a standing wave phase angle, integrates the gyroscope temperature and the base temperature acquired in the step 1, and calculates a temperature gradient and a temperature change rate.
Moreover, the specific implementation method of step 3 is as follows: the signal denoising module performs noise decomposition on the gyro signal by using the sym5 wavelet base.
Moreover, the specific implementation method of the step 4 is as follows: the characteristic extraction module adopts a 3 rd order polynomial model to extract the factors of temperature, temperature change rate and temperature gradient cross terms; weak correlation is eliminated by adopting a lifting dimension method, and the complexity of the model is reduced; analyzing the significance degree between each variable of the model and the gyro drift by adopting a multivariate linear stepwise regression method, and optimizing the model; and (3) finding an optimal classification surface in the two known types of linearly separable data samples by adopting a support vector machine, and obtaining maximum segmentation of the two types of data samples through the surface.
Furthermore, the 3 rd order polynomial model is:
wherein Δ T is the temperature gradient, T is the temperature, α1、α2、α3、α4、α5、α6、α7、α8、α9、α10、α11、α12、α13、α14、α15、α16、α17Is a polynomial model coefficient, and is a model coefficient,
the specific implementation method adopting the support vector machine comprises the following steps: h is a data classification line, and H is moved in parallel to obtain a boundary position H1And H2If { (x)i,yi),i＝1,……,n},xi∈R,yiE {1, -1} is two classes of linearly separable data samples, where xiAs input feature vectors, yiFor the class label of the input feature vector, f (x) ═ ω · x + b is a general form of the linear discriminant function, and the expression of the corresponding classification plane is:
ω·x+b＝0
by direct calculation, when H1And H2The distance between them is 2/| ω |, i.e. the classification interval is 2/| ω |, if the interval is maximized, it is the led to make | | ω | |2And (3) converting the optimal classification surface problem for solving the two types of data samples into the optimization problem for solving the variables omega and b：
yi[(ω·xi+b)]-1≥0，1＝1，……，n
Mapping data to a high-dimensional feature space through a kernel function, converting a complex nonlinear problem into a simple linear problem, and then obtaining a solution of an original problem by solving the linear problem:
s.t.yi(ω·φ(xi)+b)＜1-ξi,i＝1,2,...N
ξi≥0,i＝1,2,...N
wherein C is a penalty factor, ξiN is the sample capacity, the larger the value of C is, the better the regression effect is shown,
the method comprises the following steps of adopting a radial basis kernel function with strong local description capacity to carry out feature extraction on data, wherein the expression of the radial basis kernel function is as follows:
wherein, σ is the width of the kernel function, and the smaller the value of σ is, the stronger the local description capability is.
Moreover, the specific implementation method of step 5 is as follows: the BP neural network approximation structure is as follows: u (k) is input, u (k) is output by a controlled object y (k), k is iteration step number of the network, u (k) and y (k) are input by an approximator BP, y (k) is input by a controlled objectn(k) For the output of the approximator BP, y (k) and yn(k) As the adjustment signal of the approximator BP;
the BP algorithm learning process of the BP neural network consists of forward propagation and backward propagation, in the forward propagation, input information is processed layer by layer from an input layer through a hidden layer and is transmitted to an output layer, the state of each layer of neuron only influences the state of the next layer of neuron, if expected output cannot be obtained in the output layer, the process is switched to the backward propagation, error signals are calculated reversely according to a connecting channel, the weight of each layer of neuron is adjusted by a gradient descent method, and the error is reduced;
then, the weight between layers, the connection weight omega of the output layer and the hidden layer are adjusted by adopting a gradient descent methodjlThe learning algorithm is as follows:
wherein the content of the first and second substances,
is a partial derivative, EpTo predict the result error, elFor the output of the ith excited function via the hidden layer, xlIs an input of the l-th output layer, x'jIs the output of the jth hidden layer;
the weight of the network at the moment k +1 is:
ωjl(k+1)＝ωjl(k)+Δωjl
the hidden layer and the input layer are connected with a weight omegaijThe learning algorithm is as follows:
wherein the content of the first and second substances,
the weight of the network at the moment k +1 is:
ωij(k+1)＝ωij(k)+Δωij
if the influence of the previous weight value on the change of the current weight value is considered, a momentum factor alpha needs to be added, and the weight value at this time is as follows:
ωjl(k+1)＝ωjl(k)+Δωjl+α(ωjl(k)-ωjl(k-1))
ωij(k+1)＝ωij(k)+Δωij+α(ωij(k)-ωij(k-1))
wherein eta is the learning rate, alpha is the momentum factor, eta belongs to [0,1], and alpha belongs to [0,1 ].
Moreover, the specific implementation method of step 6 is as follows:
input x for hidden layer neuronsjIs the weighted sum of all inputs, i.e.:
output of hidden layer neuron x'jExciting x with S functionj：
Then the following results are obtained:
the output of the output layer neurons is:
net ith output and corresponding ideal output
Error e oflComprises the following steps:
error performance indicator function E for the p-th samplepComprises the following steps:
wherein N is the number of the neurons of the network output layer.
The invention has the advantages and positive effects that:
the invention extracts a plurality of groups of temperatures, temperature gradients and temperature change rates through a temperature measuring element and a temperature measuring circuit, an analog-digital converter extracts data information of a harmonic oscillator through an electrode, the data information is input into a normalization module to be calculated, a plurality of groups of gyroscope frequencies, standing wave azimuth angles, gyroscope output and other information are obtained, then a sample characteristic is extracted by adopting a radial basis kernel function, a BP neural network is trained, and a compensation model weight relation is established. And embedding the compensation model into a gyro control program mode to realize online compensation of gyro output in a temperature-varying environment. The invention can effectively improve the temperature characteristic of the gyroscope and improve the zero offset stability of the gyroscope during long-term operation.
Drawings
FIG. 1 is a system connection block diagram of the present invention;
FIG. 2 is a diagram of the optimal classification of the present invention;
FIG. 3 is a BP neural network approximation model of the present invention;
FIG. 4 shows a temperature compensated BP neural network structure of a resonator gyroscope according to the present invention.
Description of the drawings:
the device comprises a 1-harmonic oscillator, a 2-electrode, a 3-base, a 4-shell, a 5-buffer amplifier, a 6-analog-to-digital converter, a 7-signal extraction module, an 8-temperature measuring element, a 9-temperature measuring circuit, a 10-normalization module, an 11-compensation model module, a 12-error compensation module, a 13-output estimation module, a 14-signal noise reduction module, a 15-feature extraction module and a 16-model training module.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
A multi-element temperature compensation system of a hemispherical resonance gyroscope based on a BP neural network comprises a harmonic oscillator, electrodes, a base, a shell, a buffer amplifier, an analog-to-digital converter, a signal extraction module, a temperature measuring element, a temperature measuring circuit, a normalization module, a compensation model module, an error compensation module, an output estimation module, a signal noise reduction module, a characteristic extraction module and a model training module, wherein the electrodes are connected with the harmonic oscillator and used for driving and checking the vibration of the harmonic oscillator, the harmonic oscillator is fixedly arranged on the base, the shell is fixed on the base through laser welding and vacuumized to form a vacuum closed space, the temperature measuring element is arranged on the base, the electrodes, the buffer amplifier, the analog-to-digital converter, the signal extraction module, the error compensation module and the output estimation module are connected in series, the temperature measuring module, the temperature measuring circuit, the normalization module, the compensation model module and the error compensation module are connected in series, the signal extraction module, the normalization module, the signal noise reduction module, the feature extraction module and the model training module are connected in series.
The harmonic oscillator is a gyro core sensitive module, and can be made of quartz, silicon substrate, metal and the like according to different application requirements and precision grades. The electrodes are used for driving and checking harmonic oscillator vibration, and comprise contact type and non-contact type, such as piezoelectric ceramics, capacitors and the like. The buffer amplifier is used for extracting vibration information of the harmonic oscillator acquired on the electrode, and plays roles in signal conversion and isolation amplification, such as a charge amplifier and the like. And voltage signals containing gyro vibration information obtained by the buffer amplifier are acquired and converted into digital quantity through the analog-to-digital converter and are sent to the main control chip. And the signal extraction module extracts the gyro state information according to a set algorithm. The temperature measuring element measures a temperature field signal of the gyroscope base and the temperature field signal is converted into an electric signal by the temperature measuring circuit. And the normalization module receives the gyro information provided by the signal extraction module and the temperature information provided by the temperature measurement element and normalizes the information. And (3) a factory modeling link, preprocessing the normalized signal by a signal noise reduction module, acquiring modeling sample data by a characteristic extraction module, calculating by a model training module to obtain BP neural network weight information, and providing the BP neural network weight information to a compensation model. In the online compensation step, the compensation model calculates error compensation data according to the normalized signal and the training weight, and the error compensation module performs fusion compensation on the original signal. And the output estimation module calculates the external angular velocity information according to the compensated data and the scale factor and provides the external angular velocity information for the user.
The hemispherical resonant gyro multi-element temperature compensation system based on the BP neural network comprises a factory modeling and online compensation part, wherein the factory modeling comprises the following steps:
acquiring original data: a simulated environment test platform is built, full-temperature-range temperature scanning is achieved, information such as gyro frequency, temperature gradient, temperature change rate and gyro output is recorded, and multiple groups of data samples are obtained through repeated experiments.
Signal noise reduction: the original data is denoised by means such as wavelet denoising smoothing or filtering, so that the signal-to-noise ratio is improved, and the calculation precision of a subsequent training link is improved.
Feature extraction: and analyzing the correlation degree of the model variable and the gyro drift by adopting a lifting dimension method such as a multivariate linear stepwise regression method, independent component analysis and the like, optimizing parameters, eliminating weak correlation items and reducing the calculation complexity of subsequent links.
BP neural network training: and sending the optimized data sample into a configured BP neural network training set, and calculating the model weight.
Online compensation:
and (3) error signal extraction: and the gyro control circuit acquires each error signal in real time and resolves and normalizes the error signal into a training set range.
Calculating model compensation parameters: and substituting the normalized error signal into a BP neural network compensation model, and calculating real-time compensation data of the gyroscope under the current working condition.
Error compensation and output estimation: and fusing the gyro original signal and the compensation signal, estimating the current external input angular speed according to the scale factor of the gyro, realizing error compensation, and outputting and sending the compensated output to a user.
The compensation method of the hemispherical resonant gyro multi-element temperature compensation system based on the BP neural network comprises the following steps:
step 1, acquiring temperature information of the gyroscope and temperature information of the base through the temperature measuring element and the temperature measuring circuit.
The harmonic oscillator keeps a resonance state constantly under the frequency stabilization control action of the main control circuit, when the external temperature changes, the resonance frequency changes, the analog-to-digital converter obtains gyro frequency information in real time through the electrodes, and the temperature of the sensitive area of the harmonic oscillator is obtained through calculation according to the frequency temperature coefficient. The temperature of the base is obtained by a temperature measuring element, such as a platinum resistor, a temperature measuring chip and the like, which is arranged at the base, through a corresponding temperature measuring circuit. And the temperature signal is transmitted to the temperature measuring circuit to be changed into an electric signal which is input to the normalization module.
And 2, acquiring information of the gyroscope through the analog-to-digital converter.
The analog-to-digital converter acquires detection information of two modes of an X axis and a Y axis of the harmonic oscillator through the electrodes, sequentially inputs the detection information into the signal extraction module and the normalization module for resolving to obtain a gyroscope sensitive output and a standing wave phase angle, integrates the gyroscope temperature and the base temperature acquired in the step 1, and calculates a temperature gradient and a temperature change rate for model training and real-time compensation.
And 3, carrying out signal noise reduction on the gyro information obtained in the steps 1 and 2 through a signal noise reduction module.
Because the noise of the original data output by the gyroscope is large, the influence on the fitting precision of the model is large, and therefore the experimental data is preprocessed before modeling. The signal is noise decomposed, for example, using the sym5 wavelet basis, to reduce phase distortion when the wavelet is used to analyze and reconstruct the signal.
And 4, extracting the characteristics of the noise-reduced signal through a characteristic extraction module.
The multiple relevant variable combinations are of various types, so that the model complexity is high, and the training and compensation are not facilitated. The characteristic extraction module adopts a 3 rd order polynomial model to extract the factors of temperature, temperature change rate and temperature gradient cross terms:
wherein Δ T is the temperature gradient, T is the temperature, α1、α2、α3、α4、α5、α6、α7、α8、α9、α10、α11、α12、α13、α14、α15、α16、α17Is a polynomial model coefficient.
Weak correlation is eliminated by adopting a lifting dimension method, and the complexity of the model is reduced; and if the significance degree between each variable of the model and the gyro drift is analyzed by adopting a multiple linear stepwise regression method, optimizing the model.
In practical application, environment changes have randomness, and multiple groups of data are adopted for feature extraction, so that randomness errors are reduced. If a support vector machine is adopted, an optimal classification surface is found in the known two types of linearly separable data samples, and the two types of data samples are maximally divided through the surface. As shown in FIG. 2, assuming that the circle and the square in the graph are two kinds of data samples, respectively, and H is a classification line, H is shifted in parallel to obtain a boundary position H1And H2The optimal classification surface not only can correctly separate the two types of data samples to make the training error rate zero, but also can ensure that the classification interval between the two types of data samples reaches the maximum.
Is { (x)i,yi),i＝1,……,n},xi∈R,yiE {1, -1} is two classes of linearly separable data samples, where xiAs input feature vectors, yiFor the class label of the input feature vector, f (x) ═ ω · x + b is a general form of the linear discriminant function, and the expression of the corresponding classification plane is:
ω·x+b＝0
by direct calculation, when H1And H2The distance between them is 2/| ω |, i.e. the classification interval is 2/| ω |, if the interval is maximized, it is the led to make | | ω | |2And (3) converting the optimal classification surface problem for solving the two types of data samples into an optimization problem for solving variables omega and b:
yi[(ω·xi+b)]-1≥0，1＝1，……，n
extracting a gyro output trend term, which obviously belongs to the problem of solving nonlinear data optimization, mapping data to a high-dimensional feature space through a kernel function for linear indivisible data samples, converting a complex nonlinear problem into a simple linear problem, and then obtaining the solution of the original problem by solving the linear problem:
s.t.yi(ω·φ(xi)+b)＜1-ξi,i＝1,2,...N
ξi≥0,i＝1,2,...N
wherein C is a penalty factor, ξiFor the relaxation factor, N is the sample volume, and the larger the value of C, the better the regression.
With the continuous progress of the SVM, the influence of the kernel function on the data feature extraction is larger and larger. Therefore, when the characteristics of the gyro output are extracted, the selection of the kernel function is very critical. In order to obtain the characteristics of the micro change of the gyro output curve, the invention adopts a radial basis kernel function with stronger local description capability to carry out characteristic extraction on data obtained by experiments, and the expression of the radial basis kernel function is as follows:
wherein, σ is the width of the kernel function, and the smaller the value of σ is, the stronger the local description capability is.
And 5, establishing a BP neural network approximation structure, and using the feature extraction data to train the BP neural network in a model training module.
As shown in fig. 3, the BP neural network approximation structure is: u (k) is input, u (k) is output by a controlled object y (k), k is iteration step number of the network, u (k) and y (k) are input by an approximator BP, y (k) is input by a controlled objectn(k) For the output of the approximator BP, y (k) and yn(k) As the adjustment signal of the approximator BP;
fig. 4 shows a resonant gyro temperature compensation BP network. The BP algorithm learning process of the BP neural network consists of forward propagation and backward propagation, in the forward propagation, input information is processed layer by layer from an input layer through a hidden layer and is transmitted to an output layer, the state of each layer of neuron only influences the state of the next layer of neuron, if expected output cannot be obtained in the output layer, the process is switched to the backward propagation, error signals are calculated reversely according to a connecting channel, the weight of each layer of neuron is adjusted by a gradient descent method, and the error is reduced;
then, the weight between layers, the connection weight omega of the output layer and the hidden layer are adjusted by adopting a gradient descent methodjlThe learning algorithm is as follows:
wherein the content of the first and second substances,
is a partial derivative, EpTo predict the result error, elFor the output of the ith excited function via the hidden layer, xlIs an input of the l-th output layer, x'jIs the output of the jth hidden layer;
the weight of the network at the moment k +1 is:
ωjl(k+1)＝ωjl(k)+Δωjl
the hidden layer and the input layer are connected with a weight omegaijThe learning algorithm is as follows:
wherein the content of the first and second substances,
the weight of the network at the moment k +1 is:
ωij(k+1)＝ωij(k)+Δωij
if the influence of the previous weight value on the change of the current weight value is considered, a momentum factor alpha needs to be added, and the weight value at this time is as follows:
ωjl(k+1)＝ωjl(k)+Δωjl+α(ωjl(k)-ωjl(k-1))
ωij(k+1)＝ωij(k)+Δωij+α(ωij(k)-ωij(k-1))
wherein eta is the learning rate, alpha is the momentum factor, eta belongs to [0,1], and alpha belongs to [0,1 ].
And 6, carrying out neural network compensation in a compensation model module through the established BP neural network approximation structure.
Input x for hidden layer neuronsjIs the weighted sum of all inputs, i.e.:
output of hidden layer neuron x'jExciting x with S functionj：
Then the following results are obtained:
the output of the output layer neurons is:
net ith output and corresponding ideal output
Error e oflComprises the following steps:
error performance indicator function E for the p-th samplepComprises the following steps:
wherein N is the number of the neurons of the network output layer.
It should be emphasized that the embodiments described herein are illustrative rather than restrictive, and thus the present invention is not limited to the embodiments described in the detailed description, but also includes other embodiments that can be derived from the technical solutions of the present invention by those skilled in the art.
Claims (9)
1. Based on many first temperature compensation systems of BP neural network hemisphere resonance top, its characterized in that: comprises a harmonic oscillator, an electrode, a base, a shell, a buffer amplifier, an analog-to-digital converter, a signal extraction module, a temperature measuring element, a temperature measuring circuit, a normalization module, a compensation model module, an error compensation module, an output estimation module, a signal noise reduction module, a characteristic extraction module and a model training module, the electrode is connected with the harmonic oscillator and used for driving and checking the vibration of the harmonic oscillator, the harmonic oscillator is fixedly arranged on the base, the shell is fixed on the base through laser welding to form a vacuum closed space, the temperature measuring element is arranged on the base, the electrode, the buffer amplifier, the analog-to-digital converter, the signal extraction module, the error compensation module and the output estimation module are connected in series, the temperature measuring module, the temperature measuring circuit, the normalization module, the compensation model module and the error compensation module are connected in series, and the signal extraction module, the normalization module, the signal noise reduction module, the characteristic extraction module and the model training module are connected in series;
the temperature measurement circuit scans the full temperature range, records the gyro frequency, the temperature gradient, the temperature change rate and the gyro output information by the temperature measurement circuit, the buffer amplifier, the analog-to-digital converter, the signal extraction module and the normalization module, and repeats the experiment to obtain a plurality of groups of data samples;
the signal denoising module adopts wavelet denoising smoothing or filtering means to denoise original data, improve the signal-to-noise ratio and improve the calculation precision of a subsequent training link;
the characteristic extraction module analyzes the correlation degree of the model variable and the gyro drift by adopting a lifting dimension method, performs parameter optimization and eliminates weak correlation items;
the model training module sends the optimized data sample into a configured BP neural network training set, and calculates a model weight;
the signal extraction module acquires each error signal in real time and resolves and normalizes the error signal into a training set range;
the model training module substitutes the normalized error signal into a BP neural network compensation model, and calculates real-time compensation data of the gyroscope under the current working condition to form a compensation signal;
the error compensation module fuses the gyro original signal and the compensation signal, estimates the current external input angular velocity according to the scale factor of the gyro, realizes error compensation, and sends the compensated output to a user through the output estimation module.
2. The compensation method of the multivariate temperature compensation system based on the BP neural network hemispherical resonator gyro as claimed in claim 1, characterized in that: the method comprises the following steps:
step 1, acquiring temperature information of a gyroscope and temperature information of a base through a temperature measuring element and a temperature measuring circuit;
step 2, acquiring information of the gyroscope through an analog-to-digital converter;
step 3, carrying out signal noise reduction on the gyro information obtained in the step 1 and the step 2 through a signal noise reduction module;
step 4, extracting the characteristics of the noise-reduced signal through a characteristic extraction module;
step 5, establishing a BP neural network approximation structure, and carrying out BP neural network training in a model training module by using the feature extraction data;
and 6, carrying out neural network compensation in a compensation model module through the established BP neural network approximation structure.
3. The compensation method of the BP neural network hemisphere resonant gyro multivariate temperature compensation system based on claim 2, wherein the compensation method comprises the following steps: the specific implementation method of the step 1 comprises the following steps: under different temperatures, the analog-digital converter obtains the real-time frequency coefficient of the harmonic oscillator through the electrodes, calculates the temperature of the sensitive area of the harmonic oscillator according to the obtained frequency coefficient, detects the temperature of the base through the temperature measuring element, transmits a temperature signal to the temperature measuring circuit, and converts the temperature signal into an electric signal which is input to the normalization module.
4. The compensation method of the BP neural network hemisphere resonant gyro multivariate temperature compensation system based on claim 2, wherein the compensation method comprises the following steps: the specific implementation method of the step 2 comprises the following steps: the analog-to-digital converter acquires detection information of two modes of an X axis and a Y axis of the harmonic oscillator through the electrodes, sequentially inputs the detection information into the signal extraction module and the normalization module for resolving to obtain a gyroscope sensitive output and a standing wave phase angle, integrates the gyroscope temperature and the base temperature acquired in the step 1, and calculates a temperature gradient and a temperature change rate.
5. The compensation method of the BP neural network hemisphere resonant gyro multivariate temperature compensation system based on claim 2, wherein the compensation method comprises the following steps: the specific implementation method of the step 3 is as follows: the signal denoising module performs noise decomposition on the gyro signal by using the sym5 wavelet base.
6. The compensation method of the BP neural network hemisphere resonant gyro multivariate temperature compensation system based on claim 2, wherein the compensation method comprises the following steps: the specific implementation method of the step 4 comprises the following steps: the characteristic extraction module adopts a 3 rd order polynomial model to extract the factors of temperature, temperature change rate and temperature gradient cross terms; weak correlation is eliminated by adopting a lifting dimension method, and the complexity of the model is reduced; analyzing the significance degree between each variable of the model and the gyro drift by adopting a multivariate linear stepwise regression method, and optimizing the model; and (3) finding an optimal classification surface in the two known types of linearly separable data samples by adopting a support vector machine, and obtaining maximum segmentation of the two types of data samples through the surface.
7. The compensation method of the BP neural network hemispherical resonator gyro multivariate temperature compensation system based on claim 6, wherein the compensation method comprises the following steps: the 3 rd order polynomial model is as follows:
wherein, Delta T is temperature gradient, T is temperature and alpha1、α2、α3、α4、α5、α6、α7、α8、α9、α10、α11、α12、α13、α14、α15、α16、α17Is a polynomial model coefficient, and is a model coefficient,
the specific implementation method adopting the support vector machine comprises the following steps: h is a data classification line, and H is moved in parallel to obtain a boundary position H1And H2If { (x)i，yi)，i＝1，……，n}，xi∈R，yiE {1, -1} is two classes of linearly separable data samples, where xiAs input feature vectors, yiFor the class label of the input feature vector, f (x) ═ ω · x + b is a general form of the linear discriminant function, and the expression of the corresponding classification plane is:
ω·x+b＝0
by direct calculation, when H1And H2The distance between them is 2/| ω |, i.e. the classification interval is 2/| ω |, if the interval is maximized, it is the led to make | | ω | |2And (3) converting the optimal classification surface problem for solving the two types of data samples into an optimization problem for solving variables omega and b:
yi[(ω·xi+b)]-1≥0，i＝1，……，n
mapping data to a high-dimensional feature space through a kernel function, converting a complex nonlinear problem into a simple linear problem, and then obtaining a solution of an original problem by solving the linear problem:
s.t.yi(ω·φ(xi)+b)＜1-ξi,i＝1,2,...N
ξi≥0,i＝1,2,...N
wherein C is a penalty factor, ξiN is the sample capacity, the larger the value of C is, the better the regression effect is shown,
the method comprises the following steps of adopting a radial basis kernel function with strong local description capacity to carry out feature extraction on data, wherein the expression of the radial basis kernel function is as follows:
wherein, σ is the width of the kernel function, and the smaller the value of σ is, the stronger the local description capability is.
8. The compensation method of the BP neural network hemisphere resonant gyro multivariate temperature compensation system based on claim 2, wherein the compensation method comprises the following steps: the specific implementation method of the step 5 is as follows: the BP neural network approximation structure is as follows: u (k) is input, u (k) is output by a controlled object y (k), k is iteration step number of the network, u (k) and y (k) are input by an approximator BP, y (k) is input by a controlled objectn(k) For the output of the approximator BP, y (k) and yn(k) As the adjustment signal of the approximator BP;
the BP algorithm learning process of the BP neural network consists of forward propagation and backward propagation, in the forward propagation, input information is processed layer by layer from an input layer through a hidden layer and is transmitted to an output layer, the state of each layer of neuron only influences the state of the next layer of neuron, if expected output cannot be obtained in the output layer, the process is switched to the backward propagation, error signals are calculated reversely according to a connecting channel, the weight of each layer of neuron is adjusted by a gradient descent method, and the error is reduced;
then, the weight between layers, the connection weight omega of the output layer and the hidden layer are adjusted by adopting a gradient descent methodjlThe learning algorithm is as follows:
wherein the content of the first and second substances,
is a partial derivative, EpTo predict the result error, elFor the output of the ith excited function via the hidden layer, xlIs an input of the l-th output layer, x'jIs the output of the jth hidden layer;
the weight of the network at the moment k +1 is:
ωjl(k+1)＝ωjl(k)+Δωjl
the hidden layer and the input layer are connected with a weight omegaijThe learning algorithm is as follows:
wherein the content of the first and second substances,
the weight of the network at the moment k +1 is:
ωij(k+1)＝ωij(k)+Δωij
if the influence of the previous weight value on the change of the current weight value is considered, a momentum factor alpha needs to be added, and the weight value at this time is as follows:
ωjl(k+1)＝ωjl(k)+Δωjl+α(ωjl(k)-ωjl(k-1))
ωij(k+1)＝ωij(k)+Δωij+α(ωij(k)-ωij(k-1))
wherein eta is the learning rate, alpha is the momentum factor, eta belongs to [0,1], and alpha belongs to [0,1 ].
9. The compensation method of the BP neural network hemispherical resonator gyro multivariate temperature compensation system based on claim 2 or 8, wherein the compensation method comprises the following steps: the specific implementation method of the step 6 comprises the following steps:
input x for hidden layer neuronsjIs the weighted sum of all inputs, i.e.:
output of hidden layer neuron x'jExciting x with S functionj：
Then the following results are obtained:
the output of the output layer neurons is:
net ith output and corresponding ideal output
Error e oflComprises the following steps:
error performance indicator function E for the p-th samplepComprises the following steps:
wherein N is the number of the neurons of the network output layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111013325.XA CN113739779A (en) | 2021-08-31 | 2021-08-31 | Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111013325.XA CN113739779A (en) | 2021-08-31 | 2021-08-31 | Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113739779A true CN113739779A (en) | 2021-12-03 |
Family
ID=78734394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111013325.XA Pending CN113739779A (en) | 2021-08-31 | 2021-08-31 | Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113739779A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114184212A (en) * | 2021-12-27 | 2022-03-15 | 北京计算机技术及应用研究所 | Zero-position temperature compensation method for inertial instrument |
| CN114413891A (en) * | 2022-01-18 | 2022-04-29 | 天津大学 | Solid fluctuation localization inertial navigation gyroscope for temperature sensing and defect suppression |
| CN114608612A (en) * | 2022-03-11 | 2022-06-10 | 中国船舶重工集团公司第七0七研究所 | Online compensation system and method for damping non-uniform drift of full-angle mode resonant gyroscope |
| CN114964195A (en) * | 2022-07-27 | 2022-08-30 | 中国船舶重工集团公司第七0七研究所 | Hemispherical resonant gyroscope angular velocity signal temperature compensation method |
| CN115855016A (en) * | 2023-02-27 | 2023-03-28 | 南开大学 | Low-temperature impact error compensation method for optical fiber gyroscope |
| CN116358545A (en) * | 2023-05-24 | 2023-06-30 | 融感科技(北京)有限公司 | Zero offset temperature compensation method for inertial sensor |
| CN116608890A (en) * | 2023-07-21 | 2023-08-18 | 中国船舶集团有限公司第七〇七研究所 | Scale error compensation method of full-angle mode hemispherical resonator gyroscope |
| CN117109637A (en) * | 2023-10-19 | 2023-11-24 | 四川图林科技有限责任公司 | Temperature drift error correction compensation method for hemispherical resonator gyroscope |
| CN117272695A (en) * | 2023-11-21 | 2023-12-22 | 中国船舶集团有限公司第七〇七研究所 | Rigidity design method of hemispherical resonant gyro insulator |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101013035A (en) * | 2007-02-08 | 2007-08-08 | 北京航空航天大学 | Neural net based temperature compensating optical fibre gyroscope |
| CN103499345A (en) * | 2013-10-15 | 2014-01-08 | 北京航空航天大学 | Fiber-optic gyro temperature drift compensating method based on wavelet analysis and BP (back propagation) neutral network |
| CN103699873A (en) * | 2013-09-22 | 2014-04-02 | 杭州电子科技大学 | Lower-limb flat ground walking gait recognition method based on GA-BP (Genetic Algorithm-Back Propagation) neural network |
| CN107621269A (en) * | 2016-07-15 | 2018-01-23 | 北京计算机技术及应用研究所 | Fiber Optic Gyroscope Temperature Drift error compensating method |
| CN108168577A (en) * | 2017-12-22 | 2018-06-15 | 清华大学 | MEMS gyro random error compensation method based on BP neural network |
| CN109916388A (en) * | 2019-03-25 | 2019-06-21 | 西安工业大学 | Fiber Optic Gyroscope Temperature Drift compensation method based on wavelet de-noising and neural network |
| CN112880705A (en) * | 2021-01-22 | 2021-06-01 | 重庆邮电大学 | MEMS gyroscope temperature compensation method based on particle swarm optimization radial basis function neural network |
-
2021
- 2021-08-31 CN CN202111013325.XA patent/CN113739779A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101013035A (en) * | 2007-02-08 | 2007-08-08 | 北京航空航天大学 | Neural net based temperature compensating optical fibre gyroscope |
| CN103699873A (en) * | 2013-09-22 | 2014-04-02 | 杭州电子科技大学 | Lower-limb flat ground walking gait recognition method based on GA-BP (Genetic Algorithm-Back Propagation) neural network |
| CN103499345A (en) * | 2013-10-15 | 2014-01-08 | 北京航空航天大学 | Fiber-optic gyro temperature drift compensating method based on wavelet analysis and BP (back propagation) neutral network |
| CN107621269A (en) * | 2016-07-15 | 2018-01-23 | 北京计算机技术及应用研究所 | Fiber Optic Gyroscope Temperature Drift error compensating method |
| CN108168577A (en) * | 2017-12-22 | 2018-06-15 | 清华大学 | MEMS gyro random error compensation method based on BP neural network |
| CN109916388A (en) * | 2019-03-25 | 2019-06-21 | 西安工业大学 | Fiber Optic Gyroscope Temperature Drift compensation method based on wavelet de-noising and neural network |
| CN112880705A (en) * | 2021-01-22 | 2021-06-01 | 重庆邮电大学 | MEMS gyroscope temperature compensation method based on particle swarm optimization radial basis function neural network |
Non-Patent Citations (3)
| Title |
|---|
| 张勇猛; 吴宇列; 席翔; 崔红娟; 陶溢: "杯形陀螺的温度性能与补偿方法研究", 《传感技术学报》 * |
| 李志杰; 王卿; 党建军; 康文芳: "BP神经网络在半球谐振陀螺仪零偏温度补偿中的应用", 《微纳电子技术》 * |
| 韦俊新; 王泽涛; 丛正; 魏艳勇; 齐国华: "基于支持向量回归的金属谐振陀螺多元温度补偿方法", 《中国惯性技术学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114184212A (en) * | 2021-12-27 | 2022-03-15 | 北京计算机技术及应用研究所 | Zero-position temperature compensation method for inertial instrument |
| CN114184212B (en) * | 2021-12-27 | 2023-09-26 | 北京计算机技术及应用研究所 | Zero temperature compensation method for inertial instrument |
| CN114413891B (en) * | 2022-01-18 | 2024-01-23 | 天津大学 | Solid fluctuation localization inertial navigation gyroscope for temperature sensing and defect suppression |
| CN114413891A (en) * | 2022-01-18 | 2022-04-29 | 天津大学 | Solid fluctuation localization inertial navigation gyroscope for temperature sensing and defect suppression |
| CN114608612A (en) * | 2022-03-11 | 2022-06-10 | 中国船舶重工集团公司第七0七研究所 | Online compensation system and method for damping non-uniform drift of full-angle mode resonant gyroscope |
| CN114964195A (en) * | 2022-07-27 | 2022-08-30 | 中国船舶重工集团公司第七0七研究所 | Hemispherical resonant gyroscope angular velocity signal temperature compensation method |
| CN114964195B (en) * | 2022-07-27 | 2022-10-11 | 中国船舶重工集团公司第七0七研究所 | Hemispherical resonant gyroscope angular velocity signal temperature compensation method |
| CN115855016A (en) * | 2023-02-27 | 2023-03-28 | 南开大学 | Low-temperature impact error compensation method for optical fiber gyroscope |
| CN116358545A (en) * | 2023-05-24 | 2023-06-30 | 融感科技(北京)有限公司 | Zero offset temperature compensation method for inertial sensor |
| CN116358545B (en) * | 2023-05-24 | 2023-08-18 | 融感科技(北京)有限公司 | Zero offset temperature compensation method for inertial sensor |
| CN116608890A (en) * | 2023-07-21 | 2023-08-18 | 中国船舶集团有限公司第七〇七研究所 | Scale error compensation method of full-angle mode hemispherical resonator gyroscope |
| CN116608890B (en) * | 2023-07-21 | 2023-10-13 | 中国船舶集团有限公司第七〇七研究所 | Scale error compensation method of full-angle mode hemispherical resonator gyroscope |
| CN117109637A (en) * | 2023-10-19 | 2023-11-24 | 四川图林科技有限责任公司 | Temperature drift error correction compensation method for hemispherical resonator gyroscope |
| CN117109637B (en) * | 2023-10-19 | 2023-12-19 | 四川图林科技有限责任公司 | Temperature drift error correction compensation method for hemispherical resonator gyroscope |
| CN117272695A (en) * | 2023-11-21 | 2023-12-22 | 中国船舶集团有限公司第七〇七研究所 | Rigidity design method of hemispherical resonant gyro insulator |
| CN117272695B (en) * | 2023-11-21 | 2024-01-30 | 中国船舶集团有限公司第七〇七研究所 | Rigidity design method of hemispherical resonant gyro insulator |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113739779A (en) | Hemispherical resonance gyro multi-element temperature compensation system and method based on BP neural network | |
| Blackman et al. | Numerical relativity waveform surrogate model for generically precessing binary black hole mergers | |
| Kishore et al. | Evolution of lock-in amplifier as portable sensor interface platform: A review | |
| Wilson et al. | Array optimization and preprocessing techniques for chemical sensing microsystems | |
| US6687631B2 (en) | Laplace transform impedance spectrometer and its measurement method | |
| CN111578966B (en) | Hemisphere harmonic oscillator characteristic parameter identification method based on LMS algorithm | |
| CN113686356A (en) | Resonance gyro zero-bias online self-compensation system and method based on RBF network | |
| Cao et al. | Tree tensor-network real-time multiorbital impurity solver: Spin-orbit coupling and correlation functions in Sr 2 RuO 4 | |
| CN112329142B (en) | Method for identifying uncertain dynamic loads of wing structure based on support vector regression | |
| CN114623848A (en) | Hemispherical resonant gyroscope random error compensation method based on variational modal decomposition and FLP | |
| CN106289212A (en) | Integrated measurement and control unit for silicon microphony fork gyroscope | |
| CN112114164B (en) | Method for determining nonlinear vibration of silicon micro-resonance type accelerometer | |
| CN112067844B (en) | MEMS sensor array type high-precision output control method | |
| Yan et al. | An intelligent adaptive Kalman filter for integrated navigation systems | |
| Legendre et al. | High-resolution micro-pirani pressure sensor with transient response processing and time-constant evaluation | |
| CN115455346A (en) | Nonlinear error compensation method for MEMS accelerometer | |
| CN212341096U (en) | Nano cantilever beam array and resonant array gas sensor | |
| Yu et al. | Nonlinear dynamic compensation of sensors using inverse-model-based neural network | |
| KR101103578B1 (en) | Cantilever sensor | |
| Melnyk et al. | INCREASING THE SENSITIVITY AND METROLOGICAL RELIABILITY OF A DIFFERENTIAL CONDUCTOMETRIC BIOSENSOR SYSTEM. | |
| CN115600480B (en) | Global linearization frequency dispersion hysteresis modeling method and device for piezoelectric transducer, computer equipment and storage medium | |
| Park et al. | Mass measurement method based on resonance frequency of the capacitive thin membrane sensor | |
| Błachowski et al. | Early design stage of the MsAa-4 Mössbauer spectrometer | |
| Guan et al. | Rapid scan Fourier transform detection of a frequency encoded quartz crystal microbalance array | |
| JP4216692B2 (en) | Mass sensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211203 |
|
| RJ01 | Rejection of invention patent application after publication |