Segmental compensation of FOG temperature error based on ELM prediction model

Aiming at the complex nonlinear relationship between temperature and the zero bias of fiber optic gyro (FOG), combining the prediction model of extreme learning machine (ELM) with the idea of piecewise modeling. A segmented compensation method based on ELM prediction model is proposed, improving the temperature performance of FOG. Analyzing the influence of temperature on the optical fiber gyroscope zero bias. Studying the effect of the ELM model's parameters on the prediction precision, Giving the ELM neural network method for determining the number of hidden layer neurons. The simulation analysis results to the collected measured data of FOG show that compared with linear regression model and single ELM neural network model, the segmented compensation method based on ELM prediction model has more significant effects. And has good temperature applicability. After compensation, the RMSE of gyro offset data is reduced by more than 90%.