Morphology of Flashing Feeds at Critical Fluid Properties in Larger Pipes

Tailored conditioning and control of flashing feeds in industrial applications requires knowledge of the evolving flow morphology and phase fractions along the feed pipe. Design methods obtained from reference systems (e.g. water-air) are hardly applicable for commercial scales and critical fluid properties (e.g. high vapor densities, low surface tension). In this study, the flow morphology of flashing feeds in a novel refrigerant test rig at critical fluid properties was analyzed using wire-mesh sensors at two locations along the feed pipe and experimental data from the water-air system.


Introduction
Flash evaporation (flash distillation) in a single equilibrium stage is one of the basic thermal separation operations [1]. Here, liquid multi-component mixtures (initially at saturation or subcooled) undergo a pressure reduction below the boiling point of the more volatile (lighter) component(s), while passing through a throttling valve or aperture. This way, the resulting vapor is enriched with the light key component(s).
Examples for industrial separations applying flash evaporation are water desalination via multistage flash evaporators [2] or multi-effect distillation [3], crude oil fractionation [4,5] or the separation of nitrogen from natural gas [6].
Flashing feeds can be encountered in pre-flash devices such as vessels, drums or columns [7] or in feed lines [8].
Since effective separation can only be achieved for high relative volatilities [1,9], flashing feeds are primarily used for feed preparation and conditioning such as -splitting into multiple feeds [10], -selective heating of vapor or liquid phases [11], -variation of the column vapor load [12], -adaptation to seasonal fluctuations of heat sources [13], -reduction of the overall process heat duties and improved hydraulic performance [7,14], -removal of impurities [14] and -coverage of existing heat losses or required reflux [15].
Flashing feeds feature complex morphologies caused by intrinsically tied thermal hydraulics, which can result in adverse two-phase flow patterns in feed lines [16], which cause detrimental material and enthalpy distribution at the column inlet [5] or undesired droplet carryover. Considering feed lines with diameters that can reach several meters and hardly exceed length-to-diameter ratios of L/D = 10 [17], resulting two-phase flows are way beyond fully developed. Upon pressure reduction the metastable liquid undergoes a transition towards thermal equilibrium of a saturated vaporliquid flow [18]. Relaxation time [19], flash evaporation rate [2] and cavitation number [18] have been used to estimate the distance required to reach thermal equilibrium. However, they are hardly applicable for flashing feeds of distillation columns since they were obtained from studies of either free jets or transient flashing processes using rupture discs in large vessels. Consequently, common tools such as flow regime maps (mainly derived for fully developed adiabatic gas-liquid flows in narrow pipes) can hardly be consulted for selection and design of proper inlet devices for flashing feeds. Moreover, available experimental data for flashing feeds are limited to steam-water systems [20,21] or 1,1,1,2tetrafluoroethane (also known as R-134a) [22] in comparably small conduits only and do not account for fluid properties relevant for distillations processes. Consequently, process design is made with costly safety margins [23].
In this study, we provide unique experimental data for flow morphology and phase fractions of flashing feeds obtained in a 200-mm feed pipe of a recently comissioned test rig with a refrigerant cycle [24]. The applied refrigerant 3MÔ Novec 649Ô (N649) features critical fluid properties for industrial separation tasks, where liquids with low surface tension and viscosity as well as variable density differences between vapor and liquid are present. In this work, the term 'critical' refers to operating conditions or fluid properties that are known to cause instable operating conditions, equipment failure or droplet carryover in separation units. The two-phase flow morphology is visualized via capacitance wire-mesh sensors (WMSs).

Test Section and Model Fluid
The refrigerant cycle is schematically shown in Fig. 1a. N649 (with a pressure corresponding to the subcooled or supercritical state) is pumped to the electric heater (2) using the circuit pump (1). While passing the flash valve (Flowserve, type V726 DKVBQ), (3) a certain fraction of the liquid is flashed (evaporated) into the modular feed section. Eventually, the two-phase mixture enters the column where a fraction of the incoming liquid is drained at the column bottom. The vapor stream from the column head is fully condensed against cooling water (4). Condensate and bottom drainage are recycled and subcooled (4). The condensation rate controls the pressure in the feed line and the test separator.
The modular horizontal feed section has a diameter of D = 200 mm and a length of 4 m (L = 20 D). Two WMSs are installed at L = 2.5 D and L = 17.5 D downstream the flash valve to visualize the flow morphology near the valve and the separator inlet, respectively (Fig. 1b). The vapor mass fraction _ x is adjusted controlling pressure p sc and temperature T sc (via heater and flash valve, respectively), while keeping pressure p F and temperature T F in the feed section (via condensation rate) and total fluid mass flow _ M (via circulation pump) constant. The vapor mass fraction is calculated according to assuming isenthalpic flashing. Here, h evap is the evaporation enthalpy of the fluid at p F and T F in the feed section. Fig. 2 shows the log p-h diagram for N649 and illustrates both feed section pressure levels presented in this study.
Tab. 1 summarizes the experimental conditions applied in this study. Considering experiment no. 11 as an example in the log p-h diagram, subcooled N649 (p sc = 9.5 bar, T sc = 118°C) was flashed into the feed section (p F = 2.4 bar, T F = 77°C) corresponding to _ x = 0.59. The experiments shown in Tab. 1 were performed adjusting total fluid mass flow _ M and vapor mass fraction _ x, which results in different vapor phase fractions and flow morphologies. The F-factors given in Tab. 1 correspond to the gas load of the separator (300 mm diameter). The terms 'vapor' and 'gas' denoted by the subscript 'G' are used interchangeably in this work. Two different p F -T F levels corresponding to experiment no. 1-3 and no. 4-11 were selected to investigate the flow morphology at different fluid parameters (see Tab. 2).

Capacitance-Based Wire-Mesh Sensor
In this study capacitance-based WMSs were employed [25,26]. Although N649 has low dielectric constants, preliminary tests confirmed that gas (e rel = 1) and liquid phase (e rel = 1.8) can reliably be distinguished for the whole range of operating conditions of the test rig (p F = 1.5-12 bar, T F = 16-143°C). The overall sensor design according to Schleicher et al. [27] features a base body with flanges and a slide-in sensor as shown in Fig. 3. Fig. 3a shows the base body element, which is internally equipped with a temperature and pressure measurement to account for the local operating conditions. The slide-in sensor (Fig. 3b) comprises a 32 32 sensor grid, which corresponds to a spatial resolution of 6.25 mm 6.25 mm. It features a chemically resistant suporting element and a compensation for the thermal expansion of the electrodes. The sensor grid consists of two planes of parallel electrode wires with a small axial distance of 3 mm. The wires of the transmitter plane are consecutively excited with a voltage signal. Measuring the voltage between the transmitter wires and the perpendicularly oriented receiver wires provides a signal that is proportional to the local capacitance of the fluid [25]. Subsequently, the phase fraction for each crossing point is obtained from the measured capacitance e rel,x at a sampling rate of 2.5 kHz using the parallel model [28] according to a G ¼ e rel;L À e rel;x e rel;L À e rel;G Here e rel,L/G refer to the permittivities obtained from reference measurements for the low permittivity (gas) and   high permittivity (liquid) phase, respectively. Reference measurements in the rig were performed for both p F -T F levels (see Tab. 2).

Flow Morphology and Phase Fraction of Flashing Feeds
Different peculiarities of separated (stratified) flows were observed via WMSs. Fig. 4 exemplarily shows 3D visualizations and corresponding probability density function estimates (PDF; obtained using Matlab ksdensity function) of characteristic phase fraction data obtained from measurements at both WMS locations. Such PDF of the stratified flows comprises characteristic features of the time series. Three main morphologies were observed (see Tab. 3). Stratified smooth flow (SS) is characterized by a calm gas-liquid interface with no or negligible waviness represented by a narrow PDF (e.g., experiment no. 2 in Fig. 4). Stratified wavy flow (SW) is represented by a broad PDF (e.g., experiment no. 1 in Fig. 4a and experiment no. 9 in Fig. 4b), which accounts for the presence of waves with various amplitudes or frequencies. Contrary to the continuous liquid bulk of the smooth and wavy patterns, stratified flow with fractured liquid phase (SF) was encountered close to the flash valve (e.g., experiment no. 9 in Fig. 4a). Stratified fractured flow is represented by a broad PDF and a mean gas fraction close to unity. The latter represents a high vaporization rate of the refrigerant up to total evaporation.
In their work, Min et al. [29] correlated the transition from the metastable liquid state of a flashing jet towards two-phase status depending on the overpressure by the Euler number, defined as Here, the Euler number Eu denotes the ratio of pressure forces to inertia and r sc and u sc are density and velocity of the subcooled liquid, respectively. Plotting F-factors vs. Eu, the morphologies encountered at both measurement locations can be clustered as shown in Fig. 5.
Zones of similar flow morphologies are indicated by the gray dashed lines in Fig. 5. Note that close to the flash nozzle in Fig. 5a, stratified smooth flow was observed for one experiment (experiment no. 2) only. However, experiments no. 4 and 8 comprised less pronounced features of the respective morphologies, namely small wave amplitudes and mostly continuous and swaying bulk. Thus, the dashed lines account for the possible morphology transitions. Fig. 5b illustrates that for typical column operating conditions (F = 0.7-2.7 Pa 0.5 [30] and at vapor densities between 18 and 29 kg m -3 ) at entrance lengths larger then normally recommended (L/D > 10 [17]), the flow morphol-  ogy in the feed section still depends on the overpressure (p scp F ). Most morphology changes (from flash valve to column inlet) occur at Eu < 4 at low F-factors. At stratified fractured flow conditions, decreasing the overheating (T sc -T F ) (i.e., increasing liquid fraction) changes the morphology from few rivulets formed at different circumferential positions at the lower pipe half (experiments no. 3, 10, 11) towards larger liquid bulks (experiments no. 6, 7) at L = 2.5 D. At constant _ M, the stratified fractured flow changes towards stratified wavy for increasing overpressure (p scp F ) (experiments no. 6 to 5 and experiments no. 9 to 8), while for decreasing overpressure stratified smooth flow evolves (experiments no. 3 to 2).
The gas fractions obtained from WMS data (L = 2.5 D) for experiments no. 6 and 5 are shown in Fig. 6. For the stratified fractured flow (experiment no. 6) partial dry-out can be observed (t = 0.1 and t = 5.1 s). At higher overheating (T sc -T F ), the stratified wavy morphology (experiment no. 5) comprised symmetric (e.g. t = 1.2 to t = 2.9 s) and swaying (compare t = 9.2 s and t = 9.4 s) waves.
Near the column inlet (L = 17.5 D), no stratified fractured flow was observed. Instead, stratified smooth morphologies were encountered at low liquid flow rates _ M L 5.2 t h -1 (corresponding to experiments no. 1-5 and 11) and wave formation was observed for higher liquid flow rates (corresponding to experiments no. [6][7][8][9][10]. For the stratified smooth morphologies, the average liquid level increased with _ M L , while the opposite trend was observed near the flash nozzle (experiments no. 4 and 2). For _ M L 5.2 t h -1 , increasing the total refrigerant flow rate _ M caused larger amplitude waves with lower frequency.
Such dynamics indicate that both thermodynamics and hydrodynamics have not reached the fully developed state. Comparing the data obtained at both WMSs, higher liquid levels (i.e., lower a V ) were observed for most experiments at L = 17.5 D as illustrated in the parity plot in Fig. 7a.
The scattering data in Fig. 7a further confirm the flow development downstream the flash valve. Plotting the gas fraction a G at the downstream position (L = 17.5 D) against the modified total mass flow density returns a linear trend (Fig. 7b). The modified mass flow density ( _ m tot;B ) is obtained applying the Baker scaling parameters l and Ψ [31] for gas and liquid, respectively. Both parameters relate fluid properties (Tab. 2) to the reference system water (W) and air (A) according to and

Comparison with the Water-Air System and Applicability of Flow Regime Maps
To analyze the effect of the fluid properties on flow morphology and phase fractions, post-processed WMS data at L = 17.5 D for flashing N649 are compared with complementary adiabatic experiments. In their study, Döß et al. [32] investigated phase fraction data and the flow morphology of water and air mixtures (W/A) with a wire-mesh sensor (64 64 sensor grid, 2.5 kHz sampling rate). The experi- . In the following, the applicability of established flow regime maps is assessed. It should be noted that currently no available flow regime map accounts for flashing feeds and particularly not in pipes of large diameters. Most of the flow maps were derived for water-air systems only. However, amongst them few account for deviating fluid properties (r, h, s). Fig. 8 illustrates exemplarily the flow regime predictions using the modified Baker map [31] and the map of Weisman [33] in comparison with the flow morphologies obtained for N649 near the column inlet.
Although the modified mass flow densities _ m G;B and _ m L;B account for the deviating fluid properties of N649, the modified Baker map (Fig. 8a) fails to predict any of the morphologies encountered. The Weisman map (Fig. 8b) was generated based on flow regime data obtained for various fluid systems. Each of the proposed transitional superficial gas and liquid velocities (u s;G=L ¼ _ m G;L r G;L ), respectively, account for fluid properties deviating from the air-water system. Since two temperature and pressure levels (p F , T F ) were considered, two sets of transition lines are predicted. However, the sensitivity of the calculated flow regime map with regard to the fluid properties is rather low. It should be noted that the Weisman map is known to hardly predict the transition between stratified smooth and wavy flow. How-ever, it reliably clusters the data points within the combined regime of separated flows (see gray area in Fig. 8b).
In order to compare the phase fractions of both fluid systems, the gas fractions derived from the WMS measurements are plotted against _ m tot;B in Fig. 9. Higher phase fractions a G are obtained for all experiments with N649 compared with those of the water-air system. The deviations decrease with increasing mass flow densities _ m L (indicated by the arrow in Fig. 9). For similar _ m L , the phase fractions for p F = 1.5 bar show a slightly higher deviation (e.g., compare experiment no. 2 with experiments no. 6 and 7). The time-averaged phase distribution images for experiments no. 2 and 3 exemplarily illustrate the lower liquid level of N649. The smoother interface is a result of the lower relative velocity between the phases.

Conclusions
The flow morphologies of flashing feeds at critical fluid properties in a large diameter feed section (D = 200 mm, L = 20 D) were studied using wire-mesh sensors for F-factors ranging from 0.2 to 4.1. The investigated flow morphologies involve dynamic phase interactions near the flash nozzle governed by the flash expansion. Downstream of the flash nozzle the two-phase flow evolves towards a region governed mainly by hydrodynamics and fluid properties.   The experimental data indicate that the morphological changes along the feed pipe are almost completed at L = 17.5 D. Moreover, the comparison with the reference system water-air highlights the deviations casued by the different fluid properties. Future work will focus on the determination and prediction of the transition between thermodynamically and hydrodynamically dominated flow morphologies. Moreover, the unique experimental data can be used to validate and extend existing predictive tools such as CFD or reduced order models for critical fluid properties and flashing feeds.
The authors gratefully acknowledge the German Federal Ministry of Economic Affairs and Energy (BMWI) for their financial support of the project ''TERESA'' (FKZ 03ET1395).