Correlative Mapping Method for Measuring Individual Phase Flow Rates in Air-Water Two-Phase Flow Based on Stochastic Features.

Minemura, Kiyoshi; Takeoka, T.; Syoda, Shinji; Egasira, Kazuyuki; Ogawa, Yutaka

doi:10.1299/jsmeb.41.863

Cited by 2 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A number of MFM methods employ machine-learning algorithms to correlate the phase flow rates to certain features of a measured signal. Earlier examples of such techniques include ones that used machine learning algorithms with measurements of differential pressure across orifices (Beg and Toral 1993) or Venturi tubes (Minemura et al 1998). The signals from other sensors, such as turbine flowmeters (Minemura et al 1996), electrical resistance tomographs (Meng et al 2010) and conductance probes (Fan and Yan 2013), have also been used as inputs to machine learning algorithms for the measurement of the phase flow rates.…”

Section: Introductionmentioning

confidence: 99%

The wire-mesh sensor as a two-phase flow meter

Shaban¹,

Tavoularis²

2014

Meas. Sci. Technol.

View full text Add to dashboard Cite

A novel gas and liquid flow rate measurement method is proposed for use in vertical upward and downward gas-liquid pipe flows. This method is based on the analysis of the time history of area-averaged void fraction that is measured using a conductivity wire-mesh sensor (WMS). WMS measurements were collected in vertical upward and downward air-water flows in a pipe with an internal diameter of 32.5 mm at nearly atmospheric pressure. The relative frequencies and the power spectral density of area-averaged void fraction were calculated and used as representative properties. Independent features, extracted from these properties using Principal Component Analysis and Independent Component Analysis, were used as inputs to artificial neural networks, which were trained to give the gas and liquid flow rates as outputs. The present method was shown to be accurate for all four encountered flow regimes and for a wide range of flow conditions. Besides providing accurate predictions for steady flows, the method was also tested successfully in three flows with transient liquid flow rates. The method was augmented by the use of the cross-correlation function of area-averaged void fraction determined from the output of a dual WMS unit as an additional representative property, which was found to improve the accuracy of flow rate prediction.

show abstract