Nonlinear System Identification for an Electromagnetic Groundwater Flowmeter

Abstract

The signals measured by electromagnetic flowmeters are corrupted by 1/f noise from the electrodes and interference from changing electric and magnetic fields when driving the electromagnet. The effect of this interference can be mitigated using system identification using a generalized linear least squares approach and a suitable model. However, this approach fails for linear models when the level of interference is strong compared to the desired flow signal. This is due to non-linearities in the behaviour of the measurement electrodes. In this paper we show that the non-linear behaviour can be mitigated using non-linear system identification procedures. Specifically, a variant of the NARMAX algorithm was employed to determine a non-linear system model. Experimental results show that this approach produces an improved and more consistent estimation of the flow signal.