Efficient ensemble modeling method of FOG thermal-induced errors based on EEMD and extreme learning machine

In order to compensate the drift of fiber optic gyroscope (FOG) under intense ambient temperature variation, a novel ensemble modeling method named MS-ELM based on improved ensemble empirical mode decomposition (EEMD) and extreme learning machine (ELM) is addressed in this paper. Firstly, the thermal-induced error is analyzed analytically. Then, a bounded EEMD is used to extract the features of FOG drift signal. Sample entropy (SE) and probability density function (pdf) are used to analyze the relativity between obtained modes and original signal. Degree of relativity is taken as the criteria of variables selection for ELM modeling. Finally, an ensemble model of FOG drift can be obtained by adding up all the submodels. Semi-physical experiment illustrates that MS-ELM outperforms the modeling methods based on OP-ELM or ELM.