Characterization of Flow Regimes in Pulps-Water-Gas Three Phase Flows

摘要

Hydrodynamic characteristics of air-water-Kraft softwood fiber mixtures flowing in short vertical column (5.08cmrndiameter, 1.8 in long) were experimentally studied, using an instrumented flow loop. The pulp consistency wasrnvaried in the 0.0 ∼1.5% range. Five distinct flow regimes could be visually identified: dispersed bubbly, layeredrnbubbly, slug, churn-turbulent, and slug. The flow regime transitions were sensitive to consistency. Cross sectionaveragernvoid fractions (gas hold-ups) were also measured using Gamma-ray densitometry. They agreed withrnhomogeneous flow model predictions in dispersed bubbly and layered bubbly regimes, and their trends werernconsistent with Drift Flux Model in other regimes.rnThe feasibility of flow regime identification using the statistical characteristics of pressure fluctuations measured byrna single non-intrusive pressure sensor and artificial neural networks (ANNs) was demonstrated. Three-layered,rnfeed-forward and back-propagation ANNs with two different input schemes were successfully trained to recognizernthe major flow regimes. In one scheme, the standard deviation, coefficients of skewness and kurtosis, and severalrntime shift auto-correlations of normalized pressure signals were used. In another scheme, the characteristics of thernspectral power density distributions of normalized pressure fluctuations were utilized. A method was developed tornrender the latter ANN scheme transportable from one sensor to other similar sensors.rnThe effects of fiber consistency and several other hydrodynamic parameters on the column average interfacial arearnconcentration were investigated using the gas absorption technique, with CO_2 as the transferred species and sodiumrnhydroxide as the reactant agent in the liquid. The interfacial surface area concentration was found to dependrnstrongly on gas superficial velocity and void fraction, and relatively weakly on liquid superficial velocity and pulprnconsistency. The data were empirically correlated.