Based on multi-scale analysis capabilities of wavelet transform and good nonlinear prediction and integrated capabilities of RBF neural network

a nonlinear integrate prediction method is studied. Aiming at problem of zero-bias drift inhibition during storage

of quartz elastic accelerometer

a zero-bias nonlinear integeration predictive method for accelerometer

which based on WT and RBF

is proposed. In order to verify the effectiveness of the proposed method

a gravitational field calibration test of accelerometer parameters is designed.The two-year calibration experiment of some type quartz accelerometer is carried out. Prediction analysis on zerobias calibration sequence by WT—RBF integrate model and RBF model

simulation results show that the WT—RBF integrate model

compared with the single RBF model

has better prediction performance.