SRUKF for the Calibration of the Sensitivity in MEMS Accelerometers
Please login to view abstract download link
Micro Electro-Mechanical Systems (MEMS) are prone to fabrication uncertainties that directly affect their sensitivity. Specifically, the usual approach in order to calibrate sensitivity in capacitive accelerometers is to apply a direct mechanical acceleration stimulus. Recently, a new purely electrical procedure has been proposed that relies on accurate surrogates of the device and Markov-Chain Monte Carlo (MCMC) methods [1]. Here we propose as an alternative the application of Square-Root Unscented Kalman Filter (SRUKF) [2] that proves to be highly stable and accurate. We investigate increasingly complex SRUKF versions adding adapting capabilities [3, 4] and constraints on the Kalman gain [5]. By comparing with MCMC results, we show that an excellent calibration of the sensitivity is achieved with the advantage of considerably reducing the computational time.
