Abstract: Modeling and compensation for the temperature drift is an effective method to improve the measurement precision of the fiber optic gyroscope(FOG). Due to the coupling effect of temperature and input rotation rate, the complicated nonlinearity of FOG output is aggravated. To describe the nonlinearity exactly and compensate the drift accurately, a new structure of the temperature drift compensation model is designed based on neural network and a novel identification approach based on improved fuzzy inference is proposed. According to the new determinate model structure, the fuzzy regulations of the fuzzy system are designed. The parameters of the member functions are adjusted adaptively using the training data, which is constructed by the actual measurement data. Then the compensation model is identified by fuzzy inference based on the least root square mean error criterion. The results show that the established model can obtain high accuracy, good applicability and attractive prophecy.

Key words: fiber optic gyroscope(FOG);temperature drift;adaptive neural network-fuzzy inference;comprehensive compensation;nonlinearity analysis;model identification