Spraying of Viscous Liquids: Inﬂuence of Fluid-Mixing Mechanism on the Performance of Internal-Mixing Twin-Fluid Atomizers

: The thermal usage of liquid fuels implies their combustion, which is a process strongly inﬂuenced by the performance of the atomizer, which disrupts the fuel into drops of the required sizes. The spray quality of the twin-ﬂuid atomizers with internal mixing (IM-TFA) is primarily inﬂuenced by the two-phase ﬂow pattern inside the mixing chamber. We studied the performance of the four types of the IM-TFA nozzles by the optical diffraction system (Malvern Spraytec) to answer the question of how the mixing chamber design inﬂuences the spray quality at low atomizing gas consumption. We tested the effervescent atomizer in outside-in-liquid (OIL) and outside-in-gas (OIG) conﬁgurations, the Y-jet nozzle and new nozzle design, and the CFT atomizer when spraying model liquids with the viscosities comparable to the common fuels ( µ = 60 and 143 mPa · s). We found that the effervescent atomizer performance was strongly inﬂuenced by the conﬁguration of the inlet ports. Although the OIL conﬁguration provided the best spray quality ( D 32 = 72 µ m), with the highest efﬁciency (0.16%), the OIG nozzle was characterized by unstable work and poor spray quality. Both the devices were sensitive to liquid viscosity. The Y-jet nozzle provided a stable performance over the liquid viscosity spectrum, but the spray quality and efﬁciency were lower than for the OIL nozzle. Our ﬁndings can be used to improve the performance of the common IM-TFA types or to design new atomizers. The results also provide an overview of the tested atomizers’ performances over the wide range of working conditions and, thus, help to deﬁne the application potential of the tested nozzle designs.


Introduction
The combustion of liquid fuels is a process of high industrial importance. Even when the number of oil-burning furnaces is continuously reduced due to environment protection reasons, it still takes place in furnaces in central heating systems or industrial furnaces. It is also the main energy source for gas turbines and internal combustion engines.
The term heating oil is a general designation of, mostly, petroleum products used for thermal usage, with viscosities within the range from approximately 2.5 mPa·s up to 360 mPa·s (at 40 • C). effervescent atomizer and the new type of internal-mixing atomizer, the ACLR nozzle, is provided in [27,28]. Here, the authors used statistical analysis to compare the spray quality of the two nozzle types.
The lack of published results oriented to the testing and comparison of different nozzle designs motivated us to perform experiments to answer the following questions: (1) What way nozzle design influences the fluid-mixing mechanism? (2) What is the effect of the fluid-mixing mechanism on the spray quality?
It is a known fact that the annular internal flow leads to good atomizer performance [15]. The spray quality worsens rapidly when internal flow changes to the plug or slug regime [10]. This fact contradicts to the requirement for efficient work, as the annular flow is usually linked with high gas content in the mixture.
Our present results document that even simple design difference between the OIG and OIL devices influences the internal flow significantly. We were able to achieve annular internal flow with the OIL device and good spray quality even with lowest gas consumption, while the performance of the OIG nozzle was poor at the same working regimes due to the plug/slug internal flows.
The influence of the mixing chamber design on internal flow and the spray quality is shown by the comparison of four atomizer types, some well-established and some experimental concepts. This research is a continuation of our previous work [21], where the liquid disintegration process of the same nozzles was compared using high-speed camera.
The Y-jet atomizer (Figure 1a), designed according to Mullinger and Chigier [29], is a well-known nozzle design, used in several industrial applications Ferreira et al. [3]. The internal flow of this device is created by injection of liquid into a high-speed gas stream. In contrast to this atomizer, the OIG and the OIL nozzles ( Figure 1b) use fluid-mixing at low velocities. Both atomizers have the same design of their mechanical parts. The different atomizer performances were achieved by switching of the inlet ports into an "outside-in-gas" (OIG) or "outside-in-liquid" (OIL) configuration. The effervescent atomizer designed by the Lefebvre [30] represents the OIG configuration. As our previous work Mlkvik et al. [21] revealed, the investigated OIL configuration proved that the injection port configuration has a significant influence on the performance of the atomizer. This was later confirmed by Zaremba et al. [22]. The CFT (Figure 1c) nozzle has a design inspired by several previous works Chin [31], Ferreira M. Teixeira [32], Tamaki N.and Shimizu [33]. As opposed to the Y-jet nozzle, it is characterized by a low velocity of the internal flow. The mixing chamber is also significantly shorter than the mixing chamber of the OIG nozzle. This makes its design more compact and suitable for applications where a small atomizer body is required.

Experiment and Methods
The experiments were conducted on the test bench schematically drawn in Figure 2. The pressurized vessel was used to drive the liquid into the atomizer and the liquid mass flow was controlled by a throttle valve. This approach was chosen to remove liquid flow variations caused by a pump. The air pressure in the mixing chamber was maintained by a pressure reduction valve. All the experiments were performed at air-conditioned room at temperature within the range 22-24 • C with 40% (liq. 1) and 45% (liq. 2) aqueous maltodextrin solutions (Table 1). Liquids were prepared at least 12 h before the tests and were stored in the same room, where the experiments were performed. Therefore, the room and the liquid temperatures were equalized. The physical properties of the maltodextrin solutions (Table 1) were taken from [10] and were measured at 25 • C (Table 1). The atomizer working regimes were defined by ∆p = 0.14 MPa and with GLR within the range from 2.5 to 20% (Tables 2-5). We chose the inlet air pressure as an independent parameter because its contribution to the total energy balance dominates over the liquid inlet pressure [1]. Therefore, ∆p can be related to the potential energy of the atomization process. The GLR was chosen because it is commonly used, and easy to measure, dimensionless parameter. As we did not experimentally investigate the internal flows, we used three different two-phase flow maps to estimate the internal flow structure: [17][18][19]34].
The internal flows were further described by a set of dimensional (gas and liquid superficial velocities (v gas and v liq ), characteristic length scale of the internal flow structures (L)) and dimensionless criteria (Weber (We mix ), Bond (Bo mix ) and Froude (Fr mix )). The void fraction at the discharge orifice (α dis ) was estimated by a two-phase flow calculator [35].  Instantaneous drop sizes (D 32,t.d. ) were measured by the laser diffraction system (Malvern Spraytec) at a distance 100 mm downstream to the discharge orifice. The measurement distance was theoretically estimated according to [36]. The calculations were done by presumption that the liquid exits the atomizers in the form of a thin film. The input data of the calculations were taken from our previously published results [21], which dealt with the observations of the liquid discharge in the close distance from the exit orifice. The longest theoretically estimated breakup distance was 40 mm for the Y-jet nozzle working with the more viscous liquid at GLR = 2.5%. The liquid breakup distance for the other nozzles was estimated within the range from 7 to 8 mm. The distance 100 mm between the nozzle and the drop size measurement site was, therefore, considered to be sufficient to avoid the presence of the ligaments. This consideration was further verified by direct observation of the spray pattern at the drop sizes measurement site by high-speed camera. The presence of the ligaments was not observed for any of the tested operating regimes of the nozzles, except for the ones where the nozzles did not atomize the liquid. The sample image of the spray pattern, as observed in the drop sizes measurement site, is shown in Figure 3. The measurement frequency was 500 Hz and the measurement period was 25 s. The time-averaged Sauter mean diameters (D 32 ) were then estimated as the arithmetic means of the recorded 12,500 samples of D 32,t.d. values. The standard deviation (STD D32 ) was used to measure the drop sizes variations of the measured dataset. As previously shown in Kleinhans et al. [27], this method can identify the pulsations, but can lead to incorrect assumptions about the atomization process. To study the atomizers work, were involved two methods of the time-resolved analysis of the drop sizes. The first method was the conversion of the time-dependent data (D 32,t.d. ) into the Cumulative Distribution Function (CDF). This approach allowed us to analyze the whole range of the measured D 32,t.d. and in this way to identify the spray pulsations. The second method, the FFT (Fast Fourier Transform) analysis, was used to identify the periodic behavior of the spray pulsations and to indirectly gain information about the two-phase flow in the atomizer mixing chamber. Similar analysis was, for the numerical simulation results, used in [37].
The economy of the atomizer work was estimated in two ways. The atomization efficiency (η) and its sensitivity to the liquid viscosity are related to the process of liquid breakup and drops formation. The efficiency was calculated according to Jedelsky and Jicha [1] as a ratio of the surface tension energy of drops in the spray (E a ), to the total energy (E 1 ) required to produce the spray (η = E a E 1 ).

Time-Averaged Spray Analysis
The unstable atomizer performance was characterized by the spray cone pulsations, followed by the formation of the various drop sizes. In this section, we analyze the uniformity of the size distributions of the generated drops by the standard deviation of the time-averaged D 32 (denoted in Figure 3 by the error bars).
The expected trend, an increase in drop sizes with reducing GLR, was observed for all the investigated cases. Also, D 32 was comparable for all the tested devices for GLR ≥ 10%. As atomization with low gas consumption is the key advantage of internal-mixing devices in comparison to their external mixing counterparts, we focused our further analysis on regimes with GLR ≤ 5%.
The OIL atomizer generally produced the smallest drops from among all the tested devices. It also atomized both liquids under the whole range of investigated GLRs. The drops with D 32 = 130 µm were produced with the more viscous liquid at GLR = 2.5%. For liquid with lower viscosity the drops were smaller (75 µm). Similar sensitivity of the D 32 to the liquid viscosity was, observed for all the GLRs with this atomizer. On the other hand, low pulsations for almost all the investigated regimes were typical for the OIL device. The only significant error bar (±26 µm) is present for the GLR = 2.5% and liquid viscosity µ l = 143 mPa·s.
The OIG atomizer, however, has the same design as the previously mentioned device but provided considerably worse performance. It did not atomize the liq. 2 at GLR = 2.5%, and it produced D 32 larger than 300 µm at GLR = 5% with this liquid. For some applications, this can be considered, as unacceptable result. This device also shows sensitivity to µ l . It atomized the second, less viscous, liquid across the whole GLR range with acceptable drop sizes of 102 and 70 µm at the GLR = 2.5 and 5% respectively. The D 32 was not the only qualitative parameter which got worse when the OIG atomizer was used instead of its OIL counterpart. The error bars in Figure 4, indicate that at low GLRs, the OIG device produced a spray with varying drop sizes, which can be related to the plug/slug internal two-phase flow pattern previously estimated in [21]. The most intense pulsations were present when liquid liq. 2 was atomized, causing a very broad error bar at GLR = 5%.
The low sensitivity of the generated spray quality to the liquid viscosity (for the corresponding GLRs) was characteristic of the Y-jet atomizer. For GLR = 2.5%, the drops had sizes of 130 and 150 µm for liq. 1 and liq. 2, respectively. At GLR = 5% the average drop size for both liquids was 80 µm. The spray pulsations (characterized by the D 32 error bars) were most intense at lowest GLR and were comparable for both liquids (STD D32 ≈ 15 µm). This is an important difference in comparison to the previously discussed devices, where spray pulsations intensity, as well as drop size, depended on liquid viscosity.
The last of the tested devices, the CFT atomizer, produced at low GLRs the spray with the largest drops among the set of the tested devices. As with the OIG device, the CFD nozzle did not atomize the more viscous liquid at GLR = 2.5%. The error bars of the D 32 indicates that the drops size spectrum was comparable for the two liquids, except for GLR = 2.5%, where the comparison could not be done.
The time-averaged analysis provided us with some spray quality comparisons. Increasing STD D32 for low GLRs indicated unstable atomizer operation, but did not allow us to study the spray dynamic behavior. Accordingly, time-dependent analyses were used to study in detail drop size distributions and spray pulsations.    The stable regimes led to a narrow CDF, while CDF for the pulsating regimes was broad due to the wide spectrum of the drop sizes in the measured sample. Depicted values (D S,0.05 , D S,0.5 , D S,0.95 ) represent the 5th, 50th (median) and 95th percentile of the CDF of D 32,t.d. . The time-dependent data were further converted into power spectra (using the Fast Fourier Transform-FFT) to investigate the periodic spray behavior ( Figure 6).

Spraying at Low GLR
With regard to the values of We mix , Bo mix and the Fr mix (Table 2), the OIL device was characterized by the dominant influence of fluid inertia and gravity on the internal flow. The internal flow pattern was identified as the wall-attached [16,21,22], Figure 7a). The internal two-phase flow pattern led to the continuous passage of the liquid film through the discharge orifice without significant GLR fluctuations [15]. The resultant spray was produced with low drop sizes variations: 95% of all the drops were smaller than 84.8 µm. The overall spray quality was underlined by the maximum measured drop size D max = 97.1 µm (Table 6), which was just 14% larger than the D S,0.95 value. Therefore, even the largest drops were of acceptable size. The increase in liquid viscosity to 143 mPa·s caused, under the same working conditions, a decrease in spray quality. Although most of the drops were of the sizes 98.7 to 166.5 µm, the spray also contained drops as large as 656.9 µm. It was observed that the change in viscosity led to an increase of the Oh mix from 0.11 to 0.27 (Table 2). Both values are very low and point to a low influence of the viscous forces on the internal flow, but the relative Oh mix increase is high (150%). As the Ohnesorge number influences liquid deformation time [38], we can assume that the internal flow did not fully develop with the more viscous liquid.
The dimensionless parameters of the Y-jet nozzle internal flow indicate the low influence of gravitational forces on the two-phase flow pattern (the low value of Bo mix ) and the importance of the fluid inertia, documented by the We mix = 8, which resulted from small mixing chamber dimensions and the large gas-to-liquid velocity difference (Table 3). It was previously determined [21] that most of the injected liquid penetrates the high-velocity gas stream and reaches the opposite wall, continuing to flow downstream to the discharge, while a portion of the liquid detaches from the liquid stream and creates liquid drops of random sizes inside the mixing chamber Figure 7). The result of the above-described mixture discharge was a spray with D S,0.05 to D S,0.95 range within 99.5 and 150 µm and a drop median size 122 µm (Figure 8, Table 6). The largest drops were of 213 µm, which is acceptable in applications where the drop size is not the main qualitative criterion. According to the relatively high value of the Oh mix (0.7 (liq. 1) and 1.28 (liq. 2)), compared to the other investigated devices, we could expect sensitivity of the Y-jet nozzle performance to the liquid viscosity. However, the opposite was true. We observed that while the absolute values of the Oh mix are high, the relative change of the Ohnesorge number is small (82%), in comparison to the other tested devices (typically about 150-250%). Therefore, the work of the Y-jet atomizer was less influenced by the liquid used.  The OIG (effervescent) atomizer shows the influence of the mixing of the fluids on the internal flow. All the internal flow dimensionless parameters of this device were comparable to its OIL counterpart, and yet, the OIL and the OIG performance differed sharply. The switch of the gas and liquid injection ports led to an internal flow different from those observed for the OIL nozzle [16,22]. Two-phase flow maps are often used to determine the internal flow pattern for the OIG type of atomizers [1,39]. We applied three flow maps to estimate the internal flow pattern for the OIG device ( Table 7). The results show a disagreement between the flow maps in the two-phase flow pattern estimation. In particular, the map provided by [18] estimated an annular flow which is in contradiction with the other two-phase flow maps as well as with the experimental results. Figure 8 shows that most of the produced drops were of the size within the D S,0.05 to D S,0.95 interval from 71.1 to 122.1 µm. The spray quality was lowered by the presence of a drops with D max as large as 1400 µm, which points to unstable conditions at discharge, typical for the plug flow [15]. The shape of the power spectrum ( Figure 6) was characteristic for the stochastic signal. When the plug flow occurs in the mixing chamber (Figure 9), the ratio of the gas and liquid mass flows in the discharge orifice randomly varies, due to the passage of the large bubbles (plugs) alternated by the flow of the liquid without gas content. The variation of the discharged mixture composition is the source of the variations of the liquid breakup conditions in the near nozzle area of the atomizer which leads to the production of the wide spectrum of the drop sizes. We presume that the largest drops originated from the occasional discharge of a large amount of liquid not well mixed with the atomizing gas as the result of the internal plug/slug two-phase flow pattern. The above-described behavior should be considered when using the OIG atomizer, especially in applications such as spray-drying, where the large drops cannot be dried quickly enough. These drops can hit the hot wall, burn and contaminate the spray dryer with smoke. When liquid viscosity raised to 143 mPa·s, the OIG device did not atomize the liquid. The non-atomized liquid was ejected from the discharge orifice (Figure 10a) followed by the expansion of a pressurized gas (Figure 10b) with low liquid content. This indicates an internal flow with insufficient mixing, leading to separate gas and liquid volumes being discharged without proper liquid atomization. Similar observations were previously provided by [27]. Table 7. Estimation of the OIG atomizer internal flows by different two-phase flow maps at GLR = 2.5%. At the GLR = 2.5%, the CFT nozzle provided the worst overall results, despite the fact that the dimensionless parameters (Table 5), as well as the two-phase flow maps [16,21], indicated internal flow conditions similar to the OIG nozzle. The only geometrical parameter not involved in our dimensionless analysis was the mixing chamber length. We can, therefore, presume that the CFT nozzle mixing chamber was not long enough for a two-phase flow to develop. As a result, the spray is characterized by drops with non-uniform drop sizes, with the largest drops reaching the size of 1080 µm in diameter. Moreover, the more viscous liquid was not atomized at all and the CFT atomizer provided behavior similar to the OIG device (Figure 10a,b).

Spraying at High GLRs
The increased gas consumption at GLR = 5% consequently led to a discharge void fraction increase of up to 95% for all the devices. As the atomizing energy increased ( [1]), all the devices produced a spray of higher quality in comparison to the low-GLR regime.
The internal flows of all the atomizers were mainly influenced by the fluid inertial forces (Tables 2-5).
The best performance with both liquids is provided by the OIL atomizer. This device's work was characterized by low drop size oscillations, as documented by the flat power spectrum ( Figure 11) and the narrow CDF curve ( Figure 12).
The Y-jet nozzle provided a spray with median drop sizes D S,0.5 = 79.8 and 84.2 µm with liq. 1 and liq. 2, respectively. The internal flow pattern did not change with the increased gas consumption [21], so the CDFs for both liquids indicate low spray pulsations (Table 8, Figures 11 and 12 ).
OIG device spraying with the less viscous liquid is characterized by a narrow drop size spectrum with a median drop size of 66.9 µm. In comparison to the low-GLR regime, the spray quality improvement lies in the fact that no extremely large drops were detected. This implies a change in the internal flow pattern. The increase of the We mix led to a reduction in the characteristic length (L) of the gas structures in the mixing chamber. It is important to mention that none of the flow maps used (Table 7) predicted this fact. Except for the Ohnesorge number, the liquid viscosity increase did not influence the dimensionless parameters of the OIG nozzle internal flow.
As the Oh mix raised, the influence of viscosity on the flow become significant. Under these conditions, intense random gas and liquid bursts occurred (Figure 11b), followed by the production of a wide drop sizes spectrum (Figure 12b).

Atomization Efficiency
Although drop size is an important marker of spray quality, atomization efficiency is relevant for an evaluation of the energy consumption of the entire technological process (combustion, spray-drying and others). The relation of the atomization efficiency and spray quality is plotted in Figure 13. It shows the known dependency of the D 32 : smaller drops are produced with lower efficiency. The physical explanation of this trend can be found in previous works [1,10]. We will, therefore, focus only on the comparison of the nozzles.
Small drops (<60 µm) were produced with efficiency from 0.03 to 0.06% by all the atomizers, regardless of the liquid used. This drop size range was related to high GLR regimes (GLR > 5%) where the spraying process was characterized by stable spray production. The significant difference in the atomization efficiency was found at lower GLRs, characterized by the production of larger drops. The differences among the atomizers is shown on two parameters: atomization efficiency and its sensitivity to liquid viscosity at low GLRs.
When the less viscous liquid was sprayed, the best overall results were observed for the OIL nozzle ( Figure 13). It reached 0.16% efficiency for the D 32 = 75 µm. The other atomizing devices reached this spray quality with much lower efficiencies-less than 0.1%.
The Y-jet atomizer was typified by the stable spray production, with efficiency comparable to the OIG and the CFT nozzles.
At low GLRs, the viscosity increase to 143 mPa·s led to a less efficient spraying process for all the tested devices. The lowest efficiency was recorded with the OIG and the CFT nozzles. Both devices produced averaged drop sizes 100 µm with η ≈ 0.05%. The highest sensitivity of η to liquid viscosity was shown by the OIL atomizer. However, at the lowest GLR with liq. 2, this device provided the smallest drops (135 µm) with the highest efficiency (0.08%) among the tested atomizers; the relative decrease of the η with liquid change was the most significant. On the other hand, when spraying the more viscous solution the Y-jet atomizer showed the lowest relative decrease of efficiency, which is in accordance with the fact that the spray stability of this device (for corresponding GLRs) was not significantly influenced by liquid physical properties.

Overall Assessment of the Atomizers
As a variety of analyses were provided to show the differences between the work of the atomizers, we summarize some of the most significant observations in the following table to provide the reader with a complex view of our results ( Table 9). The nozzles were evaluated in 5 categories by grades within a range from 1 (best) to 4 (worst). The "X" indicates where a comparison could not be performed.
The presented comparison shows the qualitative dependency of the design parameters of the nozzles (mixing chamber and discharge orifice dimensions, liquid injection), spray quality and spraying efficiency. It can be stated that the best performance will be achieved by a well-designed nozzle with a wall-attached internal flow pattern and sudden acceleration at the discharge orifice (OIL). On the other hand, however, this design shows a sensitivity on liquid viscosity. In applications where liquid viscosity cannot be constantly maintained, a nozzle with fast internal flow (Y-jet) provides stable spray quality. Table 9. Overall rating of our atomizers for spraying at low GLRs.

Conclusions
We examined four atomizers, spraying two highly viscous liquids with a set of operating conditions to provide a comparison of their spraying abilities in order to define the advantages, disadvantages and potential applications for the spraying of liquid fuels. The comparison criteria were chosen to highlight the atomizers' differences in two areas: spray quality and the efficiency of work.
The time-averaged D 32 was comparable for all the tested devices for GLR ≥ 10%. Also, the spray pulsations were weak, which led to small variations in drop size. Good spray quality was provided by high pressurized gas consumption, which led to low atomization efficiency for all the nozzles (η ≤ 0.06%).
An analysis of the time-resolved spray behavior documents that the nozzle designs based on the liquid injected to the gas stream (OIL and Y-jet atomizers) featured low spray pulsations, narrow drop size range and acceptable spraying efficiency. Their internal flows were characterized with wall-attached liquid stream even at low-GLR regimes.
Switching of the fluid injection ports from OIL to the OIG nozzle configuration led to internal flow with the presence of large gas and liquid structures, which caused local GLR fluctuations in the discharge orifice, along with spray pulsations. It resulted in large drops present in the spray, generated by the random discharge of the liquid that was not properly mixed with the pressurized air.
The OIL nozzle has shown to have a very good spraying ability in comparison to its OIG counterpart. It was caused by the fact that the injected liquid was immediately attached to the mixing chamber wall without a chance to interact with the gas core (as it did in the OIG atomizer). The simple design change influenced fluid-mixing mechanism so the annular flow was able to form even with low gas consumption. The working principle of the OIL nozzle differs from the original effervescent atomizer design (OIG) so much that it can be considered to be a separated atomizer type. This device is characteristic by the stable operation and high efficiency when working with the liquid of constant viscosity. It is a good alternative of the currently used OIG nozzles which are often used for injection of fuel to the gas turbines or industrial furnaces.
The main advantage of the Y-jet atomizer was relative low sensitivity of the spray quality on the liquid viscosity. It provided stable performance with a narrow drop sizes spectrum and acceptable mean drop size. It is currently widely used in the industrial furnaces, but the observed behavior allows use of this nozzle in applications, where the liquid viscosity oscillates, such as spray-drying of the liquids in the food-processing industry.
The CFT nozzle, however, although giving the finest spray with the less viscous liquid at GLR = 20%, performed poorly in low gas consumption regimes and its usage potential at the atomization of the viscous fuel is low. Its potential lies in the other applications, where low gas consumption and high viscosity of the liquid are not required but the very fine spray is needed.
One interesting secondary finding is that for the OIG device the Golan and Stenning two-phase flow map predicted an annular flow for all of the tested working regimes which was not in accordance with the observed spray behavior, especially at low GLRS. The other tested flow map (Baker, Barnea) provided acceptable results, although the internal flow pattern estimation had to be supported by the additional dimensionless parameters to satisfactorily explain the OIG nozzle spray behavior.