Analysis of fiber optic gyroscope vibration error based on improved local mean decomposition and kernel principal component analysis.

The fiber optic gyroscope (FOG), one version of an all solid-state rotation sensor, has been widely used in navigation and position applications. However, the elastic-optic effect of fiber will introduce a non-negligible error in the output of FOG in a vibration and shock environment. To overcome the limitations of mechanism structure improvement methods and the traditional nonlinear analysis approaches, a hybrid algorithm of an optimized local mean decomposition-kernel principal component analysis (OLMD-KPCA) method is proposed in this paper. The vibration signal features of higher frequency components are analyzed by OLMD and their energy is calculated to take shape as the input vector of KPCA. In addition, the output data of three axis gyroscopes in an inertial measurement unit (IMU) under vibration experiment are used to validate the effectiveness and generalization ability of the proposed approach. When compared to the wavelet transform (WT), experimental results demonstrate that the OLMD-KPCA method greatly reduces the vibration noise in the FOG output. Besides, the Allan variance analysis results indicate the error coefficients could be decreased by one order of magnitude and the algorithm stability of OLMD-KPCA is proven by another two sets of data under different vibration conditions.