Phase-split of Liquid-liquid Two-phase Flow in T-junction: Numerical, Experimental, and Computational Intelligence-Based Investigations

Abstract

By understanding the importance of the study of liquid-liquid two-phase flow (LLTPF) split in T-junctions and identifying the gaps in the literature through thorough literature surveys, it is found that investigations on the phase-split of LLTPF in T-junction employing unrestricted conditions at the outlets, identification of the influencing liquid properties and input parameters with their influence on the LLTPF-split phenomena, capturing the LLTPF-split characteristics, apprehending the influence of T-junction plane’s inclination on the LLTPF-split dynamics, characterization of the LLTPF-splits in T-junction employing LLTPF’s objective signatures, and modeling of the relationship between LLTPF-split performance in T-junction and the input flow situations under diverse working conditions by developing computational intelligence (CI)-based schemes are of prime importance, and present dissertation focuses on those investigations. At first, the split dynamics of uniform homogeneous liquid-liquid two-phase mixtures (LLTPM) entering through the inlet section of a horizontal T-junction is numerically investigated. 3D steady-state numerical solutions are achieved by adopting a Finite Volume-based Eulerian Multi-Fluid Volume of Fluid (VOF) model. The Shear-Stress Transport (SST) k-ω model is used to simulate the turbulence. Before producing the results, the adopted numerical methodology is successfully validated with the results available in the literature. Extensive studies are made by varying the inlet-volume-fraction, inlet-mixture-velocity, liquid pair (LP) of LLTPM, and the conduit-diameter of T-junction in a wide range, and the role of each of these parameters on the said phase-split phenomena is elaborately explored. Inertia imbalance is found to be the most vital factor in influencing the phase-split dynamics. More splitting is obtained for the mixture with a high density- difference between the component liquids. Depending on the conduit-diameter, liquid pair, and inlet-mixture- velocity, a critical inlet-volume-fraction (> 0.5) is found where no split occurs, and the phenomenon of phase- split is reversed at the critical inlet-volume-fraction. As the inlet-mixture-velocity increases, critical inlet- volume-fraction increases for the T-junction with a smaller conduit-diameter, but it increases very slowly for a larger conduit-diameter. The intensity of the split decreases with the increment in conduit-diameter. A larger critical inlet-volume-fraction is found for a larger conduit-diameter. Experimental and computational intelligence (CI)-based investigations on the phase-separation of nonhomogeneous LLTPFs entering at the inlet section of horizontal and vertical T-junctions are separately made considering five different liquid pairs of five different fluid property pairs and varying the constituent (heavier and lighter) phases’ superficial velocities (V sh, V sl) for each liquid pair (LP). The experimentally generated results are extensively analyzed using multiple strategies to identify, measure, quantify, parameterize, and represent the said LLTPF-split under different flow situations for each type of T junction. The role of the constituent phases’ superficial velocities is extensively investigated. The influences of fluid properties are elaborately studied, and the main influencing fluid properties are marked. Important intermediate parameters, namely fractional diversions of the heavier and lighter liquids through the branch arm (D hb and D lb), are defined and introduced to quantify the phase-split. The phase-split performance is quantified and parametrized by introducing the term phase-split capacity (ξ), which is defined as the absolute difference between the D hb and D lb. The phase-split performance of the horizontal T-junction is evaluated in another way also to incorporate some diversity in the process of its evaluation by introducing phase-split capability (ε), defined as the absolute difference between the lighter liquid’s volume fractions at the inlet section and branch outlet (VF li and VF lb). The experimental outcomes against different combinations of the liquid pair, V sh, and V sl, are embedded in some experimental data points. The position of an experimental data point in the proposed 2D coordinate system for D hb versus D lb plot or VF lb versus VF li plot perfectly reflects the phase-split intensity against that experimental data. The position of an experimental data point in the ξ versus fractional volume take-off (φv) plots or ε versus φv plots, as proposed here, represents the phase-split performance with respect to the total fractional diversion of the mixture via the branch arm. The split dynamics study shows that the inertia imbalance factor tries to increase the scope of diversion of liquid with higher flow inertia through the run arm and indirectly forces the liquid with lesser inertia to move through the branch arm. The viscous imbalance factor helps the lower viscous liquid to reach the junction point faster than the higher viscous liquid flow, increases the scope of lower viscous liquid diversion through the run arm, and indirectly helps to have a higher fractional diversion of the greater viscous liquid through the branch arm. The gravity force factor will promote a higher fractional diversion of the heavier phase through the branch arm for the vertical T-junction with a vertically downward-oriented branch arm. The gravity force factor is the most influential factor for vertical T-junctions and dominates other factors. For each of the two cases (for horizontal and vertical T-junctions), the highly complex and nonlinear relationship between phase-split performance and the input flow situations is successfully modeled by indigenously developed biologically inspired CI-based hybrid methodologies that are able to predict the phase-split performance with high accuracy by knowing the working conditions (against diverse combinations of LP, V sl, and V sh). The phase-split of nonhomogeneous kerosene-water LLTPF entering through the inlet section of a T- junction with a 45° branch angle is investigated through experimental and CI-based studies by placing the T- junction’s plane (the plane passing through the axes of the three arms) at three different inclined positions (0°, 30°, and 60° inclined locations). Extensive investigations are conducted by varying the constituent phases’ superficial velocities (V sk and V sw) at each inclination of the T-junction plane (θ). The phase-separations in the said T-junction under different input flow situations are analyzed, identified, quantified, parameterized, measured, and represented by developing different schemes. The effects of the V sk, V sw, and θ on ξ of the said T-junction are explored, and the main influencing factors are identified. The experimental consequences against different combinations of the θ, V sk, and V sw are encoded in some experimental data points. The positions of the experimental data points in the proposed 2D coordinate system for the plot between fractional diversions of water and kerosene through the branch arm (D wb and D kb) perfectly represent the respective phase-split intensity. The positions of the experimental data points in the proposed ξ versus φv plots represent the phase-split performance with respect to the total fractional flow diversion via the branch arm. The gravity force factor is found to remain silent at horizontal inclination. The degree of involvement of the gravity force factor and its influence in the split dynamics is found to increase with the increment in the inclination (θ) of the T-junction’s plane. At higher θ, this factor tries to become the most influential and dominate the other factors. At the very high θ, the gravity force factor’s contribution to the split dynamics is found to be very high, and the gravity force factor’s influence always (throughout the whole range of V sh variation) prevails over the combined impact of the inertia and viscous imbalances. The relationship between ξ and input flow situations is successfully caught by an indigenously developed CI-based hybrid methodology that can predict ξ by knowing the input combination of θ, V sk, and V sw. The said phase-split is characterized and identified using objective descriptions (in terms of time series voltage signals (TSVS) and time series current signals (TSCS) converted to time series pressure drop signals (TSPDS)) of the LLTPF by employing objective sensors like in-house fabricated conductivity probes and differential pressure transmitters to avoid possible confusion arising from subjective descriptions. The relationship of ξ with the non-parametric probability (NPP) distribution and statistical parameters extracted from the objective signatures is fruitfully captured using an indigenously developed CI-based hybrid scheme that can predict ξ with high accuracy by knowing the statistical features of the objective signatures under different flow situations.