The drift of the silicon micro-machined gyroscope features chaotic property

which can be predicted by phase space reconstruction.The drift of a certain gyroscope was calculated.Its Lyapunov exponent was 0.000 529 and the predictable step number of the drift was 1 890.The delaying time

embedding dimension and correlation dimension were also calculated by 57

7.042 and 15 respectively.Based on the phase space reconstruction

RBF neural network and the gyroscope array technology were applied to predict the static and dynamic drifts of the gyroscope.The accuracy of the static drift prediction and the dynamic drift prediction were improved by 5.39 and 2.65 times.The prediction accuracy was superior to the time series method and traditional RBF neural network.