Acceleration Sensor Parameters Nonlinear Time Series Model To Predict And Achieve

Accelerometer is widely used in inertial navigation systems, and the improvement of accelerometer accuracy is very important to improve the accuracy of inertial navigation system. To improve the accuracy of the accelerometer, the design and production technology of accelerometer can be further improved, the parameters on the other hand can be compensated. The research indicates that the time series of accelerometer parameter has a strong non-linear characteristics, If we use traditional linear model, it is difficult to achieve a high modeling accuracy. Therefore, this paper proposes the accelerometer parameter nonlinear time series modeling to improve the modeling and prediction accuracy.This paper, has researched and designed calibration method of the accelerometer static parameters,and has studied the flexible quartz accelerometers. In order to obtain the bias and scale factor of the accelerometer, the accelerometer data acquisition software has been written. Accelerometer time series parameters have been proposed the use of nonlinear time series modeling, because they have a strong non-linear characteristic. The paper has been used the traditional BP neural network that increased delay loop so that it has a memory capacity of historical data, and has established the NAR neural network modeling. The paper has used AR model in which has joine d wavelet neural, and has established the model of the wavelet neural network in combination with the AR model, and the model has used AR model fitting the linear part, and used wavelet neural network fitting the nonlinear part. And the wavelet neural netw ork algorithm has been improved, so this model has the convergence rate fast, good training effect. Since the prediction of each model has its own unique characteristics and conditions, this paper use the theory of combination forecasting, which can absorb the advantages of each model. The paper has proposed to use a combination of improved Bayesian prediction method, This method has used the resulet of NAR network and combined model to predict. Compare predicted results using this method and the tradition al ARMA model shows that the combined forecast has a prediction effect and higher accuracy.