Error Analysis And Compensation Of Fiber-Optic Gyro In Strapdown Inertial Navigation System

Fiber-optic gyro(FOG)was widely used in strap-down inertial navigation system(SINS).As a sensor which is sensitive to the angular velocity of moving objects,its precision directly determines precision of the whole navigation system.Because FOG involve in military affairs and national defense,foreign countries keeping secret for achievements about FOG,and our country's analysis of FOG was left behind.In the case of the limited precision of FOG,Filtering process is used on the output of FOG before it enter the navigation computer through signal processing techniques.This dissertation make the interferometric closed-loop FOG as the research object,introduce the theory of FOG and SINS,analyse the random noise in FOG,simulate it and establish Allan variance model.,evaluate the performance of FOG through Allan variance method.In the beginning this paper filter the actual,static output of FOG based on Kalman filter to filter.An ARMA model was built and the Kalman filter was used to filter the output sequence.Then the wavelet soft-threshold method was used to filter the sequence.It's necessary for FOG signal filter to selected an appropriate wavelet fundamental function,resolving scale and threshold.Allan variance method was used to evaluate the results of two filter method.For high-frequency noise,the wavelet soft-threshold method is better than Kalman filter method,but the latter cannot be ignored in filtering middle and low frequency noise.This paper introduces the empirical mode decomposition(EMD)method and improves it.Because of the end problem in EMD method,the points in the beginning and the end cannot be judged as the maximal or minimal value,which can lead to errors in subsequent process.Symmetric extension method was used in this paper to deal with the end problem.After symmetric extension,the treated data was decomposed by EMD method,and the ort was further less than the given value,which means that the end problem is well solved.In the meanwhile,threshold chosen method was improved in EMD.Thresholds are chosen according to the distribution of the noise energy in different IMFs components,then compared with absolute value of each extreme point between each two zero-crossing points.All the data between two zero-crossing points is treated based on the result of comparison.After evaluating the three filter methods,the result shows that improved EMD filter method has advantages of the two previous methods.The Allan variance and random error coefficients are reduced evidently in long or short correlationtime,which shows that any sort of error has been effectively filtered by the improved EMD filter method.