Multi-scale modeling method based on EMD-LSSVM and its application

Due to the non-linear and non-stationary characteristics of laser gyro drift time series, it can not be predicted precisely by single forecasting model. A hybrid multi-scale modeling method based on empirical mode decomposition (EMD) and least squares support vector machines(LSSVM) was proposed, and its application in drift forecasting of laser gyro was also studied. Firstly, the drift data was decomposed into a series of intrinsic mode function via empirical mode decomposition. Secondly, Least Squares Support Vector Machines predicting models with appropriate kernel functions were constructed to predict each intrinsic mode function respectively. Thirdly, output of each predicting model were equally weighted and integrated into one output. In the end, the proposed method was used for laser gyro drift prediction. The experimental results show that the proposed prediction method which is capable of forecasting drift data precisely outperforms single Least Squares Support Vector Machines method, and can provide reference for drift compensation, fault prediction and reliability diagnoses of laser gyro.