Research On Information Fusion Method Of Multiple MEMS Gyroscopes

In recent years, small satellites have been promoted to be miniaturization,lightweight and commercial by MEMS technology, which makes the appearance ofnano-satellite and pico-satellite possible and broadens the field of spaceapplications for small satellites. However, as a inertial measurement device of asmall satellite, the measurement accuracy of the MEMS gyroscope under COTStechnology is low, which greatly limits the application on small satellites or otherproducts. Aiming at this situation, the thesis will compose a lot of MEMSgyroscopes of the same precision into the multiple MEMS gyroscopes system.Combining the technology of information fusion, the design of information fusionsystem based on multiple MEMS gyroscopes has been made. The contents of thisthesis include:Firstly, the MEMS gyroscope error, which has been analyzed in detail, showsthat the random error would be paid more attention. A data acquisition systemhardware for the multiple MEMS gyroscopes data collection is described. Thestructural and modeling method of ARMA model has been introduced and theARMA model is established by pretreating the random error data of the true MEMSgyroscope.The thesis introduces the Allan variance method to identify the MEMSgyroscope random error, and the dynamics of MEMS gyroscope error characteristicsis analyzed using the DAVAR method. The error model of the MEMS gyroscope isestablished by the Allan variance method，and the performance index to scale theefficiency of the information fusion algorithm is given.Then, Empirical Mode Decomposition is used to study the data of the MEMSgyroscopes which has weak nonlinearity. As ordinary EMD de-noising method maycause loss of useful information, the thesis combines EMD method with waveletthreshold de-noising method based on wavelet entropy. And according to thecharacteristics of intrinsic mode functions, an algorithm that hierarchical de-noseand fusion IMFs is presented, which improves the accuracy of the fusion data.Toavoid the effect of malfunction signal, the thesis proposes a fault-tolerant fusionalgorithm that based on the support degree. The efficiency of the algorithm istestified by artificially adding fault signal caused by the hardware or softwarefaulty.Finally, a fusion algorithm with sequence disposal based on Kalman filter isimplemented with the previous research to construct a kind of multi-levelfault-tolerant sequential fusion algorithm which is suitable for the large number ofMEMS gyroscopes. The result shows that the multi-level fault-tolerant sequential fusion algorithm could improve the accuracy of the fusion data, and effectivelytroubleshoot the fault MEMS gyroscope.