Fault detection and flow regime identification based on analysis of signal noise from electromagnetic flowmeters

Abstract

The signal conditioning and processing circuits of conventional electromagnetic flowmeters have been designed to provide an accurate average flowrate measurement signal, principally for the purpose of process control. One consequence of this is that the ‘noise’ and other low frequency components of the electrode signal are suppressed. Hitherto, the possibility that they may carry potentially useful information has been overlooked, but there are studies which show that information regarding the flow regime can be identified by spectral analysis of the higher frequency or ‘noise’ components of the sensor signals from some other types of flowmeters. If the entire electrode signal is analysed, using well-established signal analysis methods, diagnostic information regarding the flow regime in which the flowmeter is operating can be recovered and several distinctly different flow regimes can be identified, such as increased turbulence, swirling flow, flow pulsations and two-phase flow, all of which adversely affect the performance of the flowmeter. This paper presents the results of laboratory simulations of these flow regimes and describes a fuzzy logic method for identifying them. It also suggests a change to the conventional mode of operation and signal processing which would enable the additional functions to be implemented without involving any modification to the design or construction of the conventional flowtube. The potential benefits which arise from its application include the identification of flow regimes which adversely affect the performance of the flowmeter in its installed position and the ability to verify, on line, that the flowmeter is functioning correctly.