A Compensation Method for Long-term Zero Bias Drift of MEMS Gyroscope Based on Improved CEEMD and ELM

Abstract

In order to eliminating the long-term zero bias drift of MEMS gyroscope efficiently, a multi-scale processing method is proposed by utilizing signal decomposition. At first, an improved complete ensemble empirical mode decomposition (Improved CEEMD) is used to decompose the original signal into a series of stationary modes; then the distinct sub-series are clustered based on the sample entropy, and extreme learning machine (ELM) based model is used to train the sub-series; finally, the desired results can be obtained after de-noise and compensation. To verify the method, MEMS gyroscope CRG20 has been chosen for an hour test, and the experiment shows that zero bias drift reduced from 0.0706°/s to 0.0706°/s($1-\sigma )$ within temperature range of − 40° C to 70° C.