Aiming at problem of drift compensation of quartz elastic accelerometer bias during storage caused by effect of external environment

a predictive method for accelerometer zero-bias

which based on fast wavelet transform

is studied. A stationary detail sequence and non-linear trend sequence are extracted from non-stationary zero-bias series by Mallat algorithm

according to characteristics of sequences

auto regressive moving average（ ARMA） model and radial basis function（ RBF） model are used for prediction and modeling; using wavelet reconstruction formula

predicted value of zero-bias is got. In order to verify the effectiveness of the proposed method

a 2-year natural storage environmental scale value of accelerometer zero-bias is modeling and simulated with the combined model. Experimental results show that the WT-ARMA-RBF model

compared with the single ARIMA model and RBF model

enhances the prediction precision about 45. 5 % and 47. 4 %.