Research On Technologies Of Error Processing Of Fog And Initial Alignment

Fiber Optic Gyroscope(FOG)is the core component of Fiber Optic Strapdown Inertial Navigation System(FOSIS).The accuracy of FOG determines the accuracy limitation of FOSIS.FOG is susceptible to external vibration interference,which can produce the output error of FOG because of the elasto-optic effect.The methods to solve the problem of vibration can be divided into hardware methods and software methods.The hardware methods mainly include installing vibration absorber,strengthening the fiber optic coil with adhesive,improving mechanical structure and adopting multi-pole symmetrical winding method,which are at the cost of increasing volume and cost.Compared with the hardware method,the software method has better flexibility and effect.The core components of FOG are sensitive to temperature,non-reciprocal phase errors will occur when the ambient temperature changes.Eliminating the vibration error and temperature drift error of FOG are the keys to improving the accuracy of FOG.Initial alignment is the premise of strapdown algorithm,and the accuracy of initial alignment is an important part of navigation accuracy.The convergence efficiency and alignment accuracy of compass alignment are contradictory,it is necessary to set the parameters of alignment reasonably to coordinate the conflict between them.Because of the constant drift of FOG,the horizontal misalignment errors have obvious drifts in fine alignment of linear Kalman filter.Coarse alignments with large misalignment angles and moving base have strong nonlinearity,it is an effective way to find high-order nonlinear filtering algorithm and corresponding improved algorithm to solve the problem.In this paper,FOG vibration error processing,temperature drift error processing,fine alignment based on compass method and linear Kalman method,coarse alignment with large misalignment angles based on high-order non-linear filtering are studied.The main innovations are as follows:1.An improved Empirical Mode Decomposition with masking signal(M-EMD)algorithm is proposed to extract and compensate the angular and linear vibration disturbances of FOG so as to eliminate the vibration errors.Aiming at the deficiency of traditional M-EMD in eliminating mode mixing,the optimal frequency and range of the optimal mask signal are analyzed and given.The frequency and amplitude of the mask signal are optimized by using Particle Swarm Optimization(PSO)algorithm.On the basis of vibration signal decomposition,the vibration signal is extracted and compensated according to the correlation coefficient between the intrinsic mode function(IMF)and the original vibration signal and the mean value of IMF.By modeling and predicting the periodic vibration signal,the real-time compensation of it is realized.2.Aiming at the modeling of FOG temperature drift error,a new Support Vector Machine(SVM)modeling algorithm based on multi-parameter combined kernel functions is proposed.The improved Particle Swarm Optimization(PSO)algorithm is used to optimize the parameters,and then a high-precision temperature drift model is obtained.Aiming at the problem of low regression accuracy of single kernel function,a combined kernel function with multiple parameters is proposed.In order to improve the convergence efficiency and accuracy of PSO algorithm,the inertia weight of W is optimized in this paper.The experimental results show that the FOG temperature drift model based on the improved algorithms has higher accuracy than the model based on the traditional methods.3.To solve the problem of compensation of FOG temperature drift under multi-temperature variable rates,a multi-scale temperature drift compensation method based on improved EEMD(Ensemble Empirical Mode Decomposition)algorithm is proposed.Firstly,the improved EEMD is used to decompose the temperature drift data of FOG in multi-scale,and the noise-related IMFs are filtered by the permutation entropies of IMFs.Secondly,the number and partition method of multi-scale model are determined by analyzing the mean values of the Hilbert instantaneous frequency of IMFs.In order to improve the accuracy of FOG temperature drift model,besides temperature parameters,historical temperature drift data are added as the characteristic attribute of SVM model,and the number of historical data is optimized by Akaike Information Criterion(AIC).Finally,SVM is used to compensate the multi-scale temperature drift.4.Aiming at the problem that the both of convergence efficiency and accuracy of compass fine alignment cannot be taken into account,the exponential finite time-varying damping period is introduced to improve the speed and convergence accuracy of compass fine alignment.Aiming at the problem of the drift of horizontal misalignment angle errors in linear Kalman fine alignment,a feedback algorithm is introduced to correct the attitude matrix in real time with full feedback of the estimation value of misalignment angles so as to solve the problem and improve the accuracy of fine alignment.Aiming at the problem that the misalignment angles do not converge to small angles or the failure of the coarse alignment,an adaptive 5-th Cubature Kalman Filter(CKF)algorithm is introduced to improve the feedback efficiency of the innovation and improve the utilization efficiency of the innovation and the filtering accuracy by calculating the feedback coefficient using the current innovation.5.Aiming at the coarse alignment with large misalignment angles,an adaptive fading 5-th CKF algorithm is proposed,which calculates the estimated value of current innovation covariance matrix by fading memory exponent method so as to improve the utilization efficiency of innovation and filtering accuracy.In order to improve the feedback efficiency of innovation,the fading feedback coefficient is used to modify the error covariance matrix of the next filtering cycle,which can improve the convergence efficiency of alignment.Aiming at the problem of oscillation or divergence caused by over-feedback of the fading Kalman algorithm,after the filter entering the convergence stage,the feedback is stopped at the maximum gradient of the estimated azimuth misalignment angle to ensure the convergence of the algorithm.The error processing method of FOG vibration has been validated by the angular vibration and linear vibration data of the shaking table respectively.Experiments show that the improved algorithm can effectively extract and compensate the vibration interference.FOG temperature drift modeling method and multi-scale compensation method have been validated in the temperature box experiment,and the modeling accuracy and compensation effect have been improved.The improved fine alignment algorithms and coarse alignment algorithm are verified in the turntable and vehicle experiments respectively,and good results are obtained.