Flow Pattern Recognition Based on the Received Waveforms of a Clamp-on Ultrasonic Flowmeter for Wet Steam Flows

Abstract

Clamp-on ultrasonic flowmeter is a powerful tool for measuring flow rate in an existing pipe. However, measuring flow rate of wet steam flow has not been established yet. It is important to investigate the various error factors caused by the wetness fraction. Based on our previous investigations, it was clarified that the error of the flow rates tends to increase with the wetness fraction because of the changes in the velocity distribution in the wet steam and the effect of the liquid holdup. Therefore, it is important to estimate the flow regime of the wet-steam flow in the pipe. This study proposed a flow pattern recognition method using a machine learning based on the guided wave that is the propagated ultrasound through the pipe wall. The guided wave changes depending on the flow regime due to the fluctuations at the liquid-gas interface. Using the received ultrasonic signals of the guided wave in stratified flow, annular mist flow, and transition between them were used as the teaching data, and the other conditions were tested for the flow pattern recognition. Flow patterns were accurately predicted in adiabatic air-liquid two-phase and wet-steam flows in horizontal pipes. Furthermore, it was shown that the selection of the guided wave region is important to predict the flow pattern in various pressure conditions using the same teaching data.