Compensation of FOG temperature drift based on LS-SVM modeling

Abstract

A nonlinear prediction model based on least-square support vector machine (LS-SVM) is proposed for the fiber optic gyro (FOG) temperature drift. LS-SVM is an intelligent learning machine, and is good at solving nonlinear, small samples learning problem. In the proposed LS-SVM model, the environment temperature, temperature change rate and the temperature gradient was set to be three inputs, the FOG bias drift is the expectation output. A simulated annealing algorithm (SA) is introduced to determine two important parameters in the LS-SVM model. SA is a universal random search algorithm; it provides the LS-SVM model a best prediction accuracy. Two groups of simulation with different temperature rate were carried out to evaluate the proposed algorithm. The results proved that the proposed LS-SVM model is more efficient and accuracy than the traditional BP neural network in reducing the FOG temperature drift.