Statistical Signal Processing for the Detection of Gas Bubbles in a Magnetic Flowmeter

Inhalt:
The detection of gas bubbles in a liquid flow is an important diagnostic feature as gas bubbles falsify the measurement of the flow rate. But often even more important is the fact that the flowmeter becomes a monitor of the process quality. Gas bubbles can show up in the case of (unwanted) conditions like boiling or cavitation at edges of the flow path, or due to leakages in the system. Gas bubble detection functions are provided in some flowmeter systems, e.g., in Coriolis flowmeter, through a measurement of the energy dissipation. We show the possibility to detect gas-bubbles or solid partides in a magnetic flowmeter based on statistical signal processing. For this the time-dependent response in one electrode due to the application of a small AC voltage signal on a second electrode is used. In order to analyze the form of the signal we have made electric field simulations of the passage of a single bubble in a realistic flowmeter geometry. A superposition of gas bubbles of different sizes and positions and assuming a Poisson process for their distribution allows simulating a signal, which is compared to the lab measurements. With this knowledge one can design different detection algorithins being sensitive to different aspects of the bubble flow signal inside the magnetic flowmeter. Several algorithms have been compared in tenns of their performance but also their computing power requirements. The algorithins have been transferred from a Labview/Matlab prototype to an embedded microcontroller and were tested both in a lab setup and in the field. They are currently implemented and tested for the next generation of the ABB EMF magnetic flowmeter system.