You are on page 1of 13

Materials Science and Engineering A 379 (2004) 264276

A computational uid dynamics (CFD) investigation of the wake closure phenomenon


Jason Ting a, , Iver E. Anderson b
b a Alcoa Inc., Specialty Metals Division, Alcoa Technical Center, Alcoa Center, PA 15069-0001, USA Ames Laboratory (US DOE), Iowa State University, 222 Metals Development Building, Ames, IA 50011, USA

Received 5 September 2003; received in revised form 10 February 2004

Abstract Using a computational uid dynamics (CFD) software, the gas dynamics of the open-wake and closed-wake conditions of an annular-slit high-pressure gas atomization (AS-HPGA) nozzle were investigated to validate the predictions of a pulsatile atomization model that was recently proposed. The location of the recirculation zones, the oblique shocks and the Mach disks were analyzed for this type of closed-coupled gas atomization nozzle. The stagnation pressures located downstream of the Mach disk, in closed-wake condition, were found to be approximately twice as high as the stagnation pressure in an open-wake condition at a slightly lower atomization gas pressure. The turbulence model utilized within the CFD calculation scheme appeared to be inadequate for calculating aspiration pressure just below wake-closure pressure when the recirculation zone is extremely long and narrow. However, overall, the CFD calculation correlated well with the experimental results, showing that the aspiration pressure progressively lowers as operation pressure increases in open-wake condition, and rises as operation pressure increases in closed-wake condition. 2004 Elsevier B.V. All rights reserved.
Keywords: Gas atomization; CFD; Wake-closure; HPGA nozzle; Recirculation zone

1. Introduction One of the many methods of producing spherical metallic powder is by a gas atomization processa process that is not yet thoroughly understood. The practice of gas atomization in the metal powder manufacturing industry is widespread. Spherical metal alloy powders produced by the gas atomization process have very attractive material properties because of the high solidication rate the powders experience. This high solidication rate promotes microstructure renement in the resulting powders that is unattainable by conventional ingot casting metallurgy. This is particularly benecial for reducing elemental segregation in highly alloyed materials, where microstructure renement and alloy homogeneity are desirable. Furthermore, the spherical morphology of the powder is ideal for the growing metal injection molding and thermal spray coating industries. In general, two gas atomization processes exist in the powder manufacturing industry: close-coupled and free-fall

Corrresponding author. E-mail address: jason.ting@alcoa.com (J. Ting).

atomization. In close-coupled atomizers, the molten metal stream is delivered into the atomizing gas via a ceramic melt delivery tube immediately adjacent to the high pressure atomizing gas. In free-fall atomizers, the molten metal stream is allowed to free-fall a distance before the atomizing gas impinges upon the stream. The close-coupled atomization process is preferred over the free-fall process in the production of ne powder because the close proximity of the atomizing gas to the melt delivery tube, enhancing the molten metal breakup and making the formation of ner powder particles more efcient. However, close-coupled atomization can be difcult to practice and it is considered by many to be more an art than a science. The difculty mainly lies in the general lack of understanding of the gas recirculation effect within the atomization zone and how it relates to initial disruption and distribution of the melt ow, prior to the melt disintegration. The inherent gas recirculation effect is created by the global ow and pressure patterns of the atomizing gas, at the base of the ceramic nozzle. This attached recirculation zone is an extension of the ceramic melt delivery tube, creating a local environment at the melt orice that affects the

0921-5093/$ see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2004.02.065

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

265

ambient gas pressure. The gas pressure at the melt orice, also commonly known as the aspiration pressure, is known to inuence the melt ow rate. It was previously hypothesized [1] that the recirculation zone can be viewed as a control volume, and that the gas enters the recirculation zone through the wake front (also known as stagnation front) of the recirculation zone, and exits along the circumferential edge of the melt tip base. The aspiration pressure is a measure of the force of the total recirculating gas ow acting against the melt orice area. The control volume hypothesis [2] states that the aspiration pressure is a consequence of the mass balance between the gas entering and leaving the recirculation zone. Since pressure is proportional to the density of the gas, the stagnation pressure in the primary recirculation zone has a direct contribution to the mass of gas entering the recirculation zone and the resulting aspiration pressure measured at the melt orice. Under high-pressure atomization conditions, the melt feeding stability at the melt orice can be better controlled if the melt orice experiences a subambient pressure. Operating in this preferred aspiration condition, a subambient pressure at the melt orice should provide for an intrinsically more stable melt feed by drawing the melt towards the atomization zone, resulting in a consistent powder yield. The relationship between the melt ow rate and the ne metal powder yield, however, is mostly empirical. A literature review of gas-atomized metal powders indicates that the best-known and most commonly quoted correlation for gas atomized metal droplet diameter (or mass median diameter) comes from analysis of a free-fall atomization process by Lubanska [2]. In Lubanskas proposed correlation, the important process parameters and material properties are related to the mass median diameter of droplets (MMD) by the following equation, MMD = kD d0 with We =
2d L UG 0

1+

m L m G

vL vG We

12

(1)

(Weber equation),

(2)

where d0 is the melt delivery tube orice diameter, having values between 6.35 and 4.725 mm; kD is a constant whose value was found to be between 40 and 50 for various metals and inorganics, m G and m L are the mass ow rates of atomization gas and liquid metal, respectively; vG and vL are the kinematic viscosities of the gas and the liquid, respectively; and We is the dimensionless Weber number. In the Weber equation, L and represent the density and surface tension of the liquid, respectively, and UG is the gas velocity at impact with the liquid. It is obvious from the above empirical analysis that material properties such as viscosity, density, and surface tension are invariant for a predetermined melt temperature, and that the dimension d0 of the atomizer is constant for a given atomizer geometry. Thus, the mass median diameter is a strong function of the gas velocity and the

gas and liquid ow rates, m G and m L . This ow rate ratio L ) in Lubanskas equation is commonly referred to (m G /m as the gas-to-metal ratio (GMR) in melt atomization. Intuitively, one would assume that, at a xed d0 , the GMR is related to the operating pressure of the atomizer gas supply, such that the high and low GMRs are directly associated with the high and low atomization gas pressures, respectively. This, however, is not always true in a close-coupled atomization process. It was recently shown that the GMR can change dramatically (by 37.2%) at two atomization pressures that differ by less than 1.5% (or 130 kPa) [3,4]; these two pressures were on either side of the wake-closure pressure (WCP). The wake-closure phenomenon, which occurs at the WCP, is a gas dynamic condition where the recirculation zone is truncated by the sudden appearance of a Mach disk [5]. Simultaneously, the subambient aspiration pressure, measured at the melt orice, suddenly becomes deeper (i.e., lower below atmospheric pressure). The abrupt drop in aspiration pressure with the appearance of the Mach disk is due to the restrictive ow the Mach disk imposes on the truncated recirculation zone [3,4]. As mentioned before, with less mass of gas owing into the recirculation zone, the aspiration pressure measured at the melt orice deepens, i.e., there is a greater suction that acts on the exit of the pour tube. The gas ow conditions before and after the aspiration drop are known, respectively, as open and closed wakes [5]. The previous atomization studies [3,4], indicate that the deep aspiration observed in the closed-wake condition causes the melt ow to slow, which is contrary to what one would assume intuitively from increased suction. In the same studies [3,4,6], a pulsatile atomization model was proposed to relate the abrupt melt ow rate decrease to the gas dynamics of the atomization process. The model revealed that the gas dynamics around the melt delivery tube orice, which inuence the aspiration pressure, also controlled the melt ow rate and the GMR during atomization. The proposed pulsatile atomization model [3,4] was able to explain how the atomization of a nickel-based alloy immediately above the wake-closure pressure could produce a median powder size that is 42% smaller than that atomized immediately below the WCP; given that the increase in the atomization gas pressure was less than 1.5%. In addition, the melt ow rate was observed to slow when operating above the wake closure pressure, where the aspiration pressure is deeply subambient. The premise of the proposed pulsatile atomization model requires the existence of a high-pressure stagnation point located immediately downstream of the Mach disk, in the closed-wake operating condition [3,4,6]. In the model, when a dense melt ows into the recirculation zone, it disrupts the gas dynamics of the Mach disk (the Mach disk disappears), turning the recirculation zone into an open-wake condition. The stagnation pressure that was located behind the Mach disk (now absent) promptly enters the recirculation zone, pushes upwards against the melt orice, and disrupts temporarily the melt ow. The disrupted melt ow allows for

266

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

the atomizing gas to re-establish the gas dynamics of the closed-wake condition, with a deep subambient aspiration, and to resume the melt ow. This oscillation effect between the two states, closed- and open-wakes, can also be used to explain the pulsatile effect actually observed in liquid spray atomization [79]. In the open-wake condition, a weaker stagnation pressure exists and the Mach disk is absent, i.e., the pulsatile effect is expected to be absent in the open-wake condition [3,4]. It seems reasonable that the dynamic balance between the strong aerodynamic forces of the supersonic atomization gas and the high momentum and heat content of the melt ow could produce an oscillation between closed-wake and open-wake atomization behavior. However, the magnitude of the stagnation pressures at open and closed-wake conditions, and the relative location of the stagnation pressure fronts have not been thoroughly investigated in the previous effort at modeling the gas-only ow in close-coupled gas atomizers using CFD [1012]. The prior CFD results that were obtained provided a good correlation to actual aspiration measurements and matched (reasonably well) the Schlieren images of the atomization nozzles, which showed Mach disk and stagnation fronts features. Unfortunately, none of the models focused on the stagnation pressure magnitudes in either the open-wake or closed-wake conditions, making it difcult to support or refute the proposed model. Therefore, a CFD method was used in this study to qualitatively locate and quantitatively measure the stagnation pressure fronts in open-wake and closed-wake conditions. The velocity and pressure proles along the central axis from the CFD model were also examined to gain understanding of the gradients in these properties that occur in both conditions.

eter, located concentrically around the delivery tube measuring 10.36 mm in diameter. The AS-HPGA nozzle was designed to have the same total jet exit area as the DJ-HPGA, thus the gap distance between the inner and outer walls of the AS-HPGA is 0.254 mm, i.e., smaller than the jet diameter. Pressure transducers were located in the melt delivery tube and the atomizer manifold to measure the aspiration pressure at the melt delivery tube orice and the actual atomization gas delivery pressure to the annular slit, respectively. This procedure has been discussed previously in detail [1].

3. Computational modeling procedure A CFD program called Rampant (manufactured by Fluent Inc., Lebanon, New Hampshire) was used in this study. An unstructured triangular grid mesh density was created to delineate reasonable (non-interacting) gas ow eld boundaries for the atomizer nozzle. The coordinates of the mesh boundaries and overall calculation space are depicted in Fig. 1. Since the modeling involved an annular-slit HPGA type atomizer, possessing a rotational symmetry, the computational domain was simplied into modeling one half of a central cross-section of the atomizer and the melt feed tube (e.g., see Fig. 2). The atomization ow eld is oriented horizontally with the atomizer melt feed and gas supply on the left and the axis of symmetry boundary on the bottom of the computational domain, respectively, as shown in Fig. 1. A grid renement was applied during the CFD calculations to regions containing large pressure gradients and close to solid walls. An additional conservation equation for the turbulent kinetic energy and its dissipation rate was also solved to account for the effects of turbulence. The compressibility effect, included in the turbulence model, used Sarkars dilatation dissipation method [14]. The turbulent model closure and its specic handling are discussed in detail in another article [11]. Atomization gas pressures of 0.69, 2.07, 3.45, 4.82, 6.20, and 7.58 MPa were used in the CFD model to initialize the pressure in the manifold, upstream of the annular slit exit. The thermodynamic constants of argon gas were used.

2. Experimental procedure An annular-slit HPGA (AS-HPGA) nozzle, based upon a discrete-jet HPGA (DJ-HPGA [1]) geometry [13], was constructed. The AS-HPGA nozzle has parallel surfaces for the inner and outer walls of the annular-jet, and a gas ow apex angle of 45 , or 22.5 from the central axis. The DJ-HPGA nozzle is an ensemble of 20 discrete jets, 0.737 mm in diam-

Fig. 1. Schematic drawing of the computational eld showing the geometry of the boundaries. The dimensions of the boundaries are in mm, given in parentheses.

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

267

Fig. 2. CFD model of the AS-HPGA nozzle at gas atomization pressure of 0.69 MPa in the open-wake condition: (a) the global velocity output and the mirroring schematics of the gas ow structures; (b and c) the velocity, and static pressure proles, respectively, along the geometric centerline of the AS-HPGA nozzle.

4. Results 4.1. CFD modeling results Since the AS-HPGA atomizer nozzle was modeled axis-symmetrically, the calculated images of the velocity proles can be reected about the axis of symmetry to give the full ow eld of the AS-HPGA nozzle. The resulting analyses of the dominant shock structures in the modeled pressure conditions are presented as schematic drawings mirroring (below) the CFD outputs. It must be noted that

the velocity magnitude increases with decreasing shades of darkness in the CFD gures. The velocity ow eld modeled at an atomization gas pressure of 0.69 MPa is given in Fig. 2a. The rapidly expanding gas exiting the annular gas opening forms a small oblique shock near the edge of the ceramic melt delivery tube tip, slowing down the gas velocity. The velocity re-accelerates then decelerates across a set of PrandtlMeyer waves [15] before combining with the other gas streams to form a unied ow eld of enhanced PrandtlMeyer expansion waves along the central axis. A diminutive recirculation zone is

268

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

located adjacent to the melt orice surrounded by the gas ow eld mentioned above. Located 0.6 nozzle diameter length (NDL) units downstream, at the terminus of the recirculation zone, is the stagnation front where the gas velocity is zero. Delineating the recirculation zone is the sonic line [17], separating it from the surrounding atomizing gas. The turbulent layer (or viscous layer) [17] lies in the recirculation zone near the sonic boundary. In the recirculation zone, the recirculation gas is observed to ow upstream from the stagnation front toward the melt orice, along the central axis. At the melt orice, the recirculating gas turns laterally (radially outward) toward the circumferential edge of the melt

tip. At the edge, it encounters the sonic boundary, forcing it to turn and ow downstream, contained within the sonic boundary. In this case, a turbulent layer is found separating the gas that ows upstream within the core of the recirculation zone, from the gas that ows downstream along the inner perimeter of the sonic line. This turbulent layer is analogous to the recirculation eddies described by a proposed control volume model [2] (this is discussed in Section 5.1). As the modeled pressure increases to 2.07 MPa, the stagnation front extends downstream to about 1.8 NDL units from the melt orice, lengthening the recirculation zone (Fig. 3a). With the gas ow expansion and acceleration

Fig. 3. CFD model of the AS-HPGA nozzle at gas atomization pressure of 2.07 MPa in the open-wake condition: (a) the global velocity output and the mirroring schematics of the gas ow structures; (b and c) the velocity, and static pressure proles, respectively, along the geometric centerline of the AS-HPGA nozzle.

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

269

process, the recirculation zone forms a projected hourglass contour, due to the inuence of a recompression shock that gains strength. At a pressure of 3.45 MPa, the gas ow expansion downstream of the melt orice begins to contract the elongated recirculation zone, exaggerating the hourglass contour (Fig. 4a). The stagnation front is at 2.75 NDL units, and the recirculation zone has a total length of 3.4 NDL units. A closed wake condition is observed at 4.82 MPa (Fig. 5a). At this pressure, a Mach disk forms and truncates the exag-

gerated hourglass recirculation zone, giving rise to a shortened primary recirculation zone. The internal shock is seen to terminate at the edges of the bowed Mach disk; while, a recompression shock and a trailing shock are formed off the edge of the primary recirculation zone and off the edge of the Mach disk, respectively. The stagnation front associated with the primary recirculation zone is seen at only 1.0 NDL unit downstream, and the bowed Mach disk is located 2.5 NDL units downstream. There are two other stagnation points observed immediately downstream of the Mach disk

Fig. 4. CFD model of the AS-HPGA nozzle at gas atomization pressure of 3.45 MPa in the open-wake condition, near to wake-closure pressure: (a) the global velocity output and the mirroring schematics of the gas ow structures; (b and c) the velocity, and static pressure proles, respectively, along the geometric centerline of the AS-HPGA nozzle.

270

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

Fig. 5. CFD model of the AS-HPGA nozzle at gas atomization pressure of 4.82 MPa in the closed-wake condition: (a) the global velocity output and the mirroring schematics of the gas ow structures; (b and c) the velocity, and static pressure proles, respectively, along the geometric centerline of the AS-HPGA nozzle.

within the secondary recirculation zone (Fig. 5a). The rst of the two stagnation points, that is smaller, is close to the Mach disk, only 2.75 NDL units downstream. The larger (in static pressure) stagnation point is located 3.4 NDL units downstream and is very visible in the CFD velocity ow eld (Fig. 5a). Since the CFD models of the AS-HPGA for a moderate pressure range above 4.82 MPa have gas ow features similar to that at 4.82 MPa, at wake-closure, only a CFD model of 4.82 MPa will be presented here.

4.2. Velocity prole along the axis of symmetry The calculated centerline velocity at each modeled atomization gas pressure, starting from the melt orice position, is plotted below the CFD gures to help quantify the gas velocity magnitude, the ow direction, the length of the recirculation zones, and the location of the stagnation fronts and stagnation points. A negative magnitude in the gas velocity in the gures indicates that the gas is owing towards the melt orice (i.e., upstream), while a positive velocity

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

271

magnitude indicates that the gas is owing away from the melt orice (i.e., downstream). Stagnation fronts (locations of peaked stagnation pressure) are located where centerline velocities are equal to zero. The velocity prole at 0.69 MPa is depicted in Fig. 2b. A stagnation front that bounds the recirculation zone is located 0.6 NDL units from the orice. Within the recirculation zone, all the centerline ow is upstream, except for the immediate orice area where it goes to zero. Downstream of the stagnation front, the gas accelerates along the centerline to 400 m/s at 1.8 NDL and, then oscillates, reaching velocities as high as 390 m/s. The velocity oscillation, however, is less apparent for the downstream velocity prole at 2.07 MPa (Fig. 3b) because the (PrandtlMeyer) wavelength is too large for the computational eld of view. The maximum negative velocity within the recirculation zone reaches 250 m/s and the maximum positive velocity downstream of the recirculation zone is 400 m/s at 4.5 NDL. At a pressure of 3.45 MPa, the stagnation front is 2.75 NDL units downstream from the orice. A negative velocity oscillation is noted within the recirculation zone (Fig. 4b), possibly associated with the exaggerated hourglass shape of the recirculation zone. Beyond the stagnation front, the velocity increases, reaching 425 m/s at 4.8 NDL. At the pressure of 4.82 MPa, the length of the primary recirculation zone is truncated to 1.0 NDL unit (Fig. 5b). The velocity prole curve is seen to cross the zero velocity axis in three instances, indicating the occurrence of three stagnation fronts within the wake region. The rst stagnation front is just downstream of the melt orice in the primary recirculation zone, as mentioned above. The downstream gas velocity immediately after this reaches close to 400 m/s over a very short distance before it abruptly drops to zero velocity (at 2.7 NDL). A precipitous drop in velocity from 400 m/s to zero occurs across the bowed Mach disk (Fig. 5b). A secondary recirculation zone is established, located between 2.5 and 4.5 NDL downstream. Bounding the secondary recirculation zone are the two other stagnation points, located 2.8 and 3.4 NDL, respectively. Between these two stagnation points, the gas ow is reversed, owing upstream. Beyond the third stagnation point, located 3.4 NDL, the gas realigns its ow in the downstream direction and reaches a top velocity greater than 450 m/s before exiting the computational eld. 4.3. Pressure prole along the axis-symmetry At an atomization pressure of 0.69 MPa (Fig. 2c), the static pressure within the recirculation zone ranges between 130 and 255 kPa, with the highest pressure being at the stagnation front located at 0.6 NDL units. Beyond this recirculation zone, the static pressure oscillates above and below the atmospheric pressure, 101 kPa. The pressure oscillation is anti-phased to the velocity oscillation prole (Fig. 2b). The PrandtlMeyer wave oscillations compress and expand the gas, causing it to decelerate and accelerate, respectively.

At an atomization pressure of 2.07 MPa (Fig. 3c), the initial positive pressure peak is located at the stagnation front, i.e., zero velocity point, at 1.8 NDL downstream. A second positive static pressure peak is also found 1 NDL downstream from the rst. Beyond this point, the static pressure, again, oscillates above and below atmospheric pressure. When the modeled atomization pressure reaches 3.45 MPa (Fig. 4c), the recirculation zone is 2.7 NDL units long, but has a minor static pressure peak before reaching the stagnation pressure peak at the stagnation front. At this minor peak the static pressure is 120 kPa, while the pressure at the stagnation front is 250 kPa. Downstream from the stagnation front, the static pressure is again found to oscillate, similar to the 2.07 MPa case. When the modeling pressure is 4.82 MPa (Fig. 5c), two recirculation zones are apparent. The stagnation pressure of the primary recirculation zone is 120 kPa. Within the gap between the zones, the static pressure gradually falls until the 2.5 NDL point. Over a very short distance, the static pressure abruptly rises to 225 kPa, as the atomizing gas crosses the bowed Mach disk, reaching the rst of two downstream stagnation points at 2.8 NDL. Further downstream, the static pressure continues to rise, reaching a second peak static pressure of 310 kPa at 3.6 NDL downstream. After reaching the second peak pressure, the static pressure drops below the atmospheric pressure. Due to the length of the computational domain, one cannot observe the subsequent oscillation of the pressure caused by the PrandtlMeyer waves. 4.4. Computational aspiration prole The calculated aspiration pressures are determined from averaging the static pressures located across the melt orice divided by the concentric areas of the melt orice at each pressure location (Fig. 6). The aspiration pressure at 0.69 is over-ambient, and at 2.07 MPa the aspiration is approximately ambient. The calculated aspiration pressures above 2.07 MPa are all sub-ambient. The deepest subambient

Fig. 6. Computational and experimental aspiration pressure measured with respect to manifold pressure of the ASJ-HPGA nozzle.

272

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

aspiration pressure is achieved at 3.45 MPa, followed by a gradual rising trend. 4.5. Experimental aspiration prole The aspiration curve of the annular-slit HPGA nozzle shows that the aspiration pressure is subambient for all the atomization pressure values. The aspiration pressure initially decreases as the manifold pressure increases (Fig. 6), until the wake-closure pressure of 5.5 MPa is reached, and the aspiration pressure drops abruptly. Above 5.5 MPa in the closed-wake condition, the aspiration curve gradually increases as manifold pressure increases, as observed in the CFD modeled pressures.

5. Discussion 5.1. CFD images Typically, the gas exiting a solid (full bore) cylindrical (or convergent) jet under high pressure (greater than 0.15 MPa) in a free expansion condition forms a normal shock (Mach disc) near its exit [16,17]. The shock assists the rapidly expanding gas in adjusting to the ambient surrounding pressure. The AS-HPGA model shows a similar behavior, but because of the presence of the melt feed tube, a wake region is established in addition to the typical gas jet dynamics. At 0.69 MPa, the gas expands downstream from the nozzle, and the normal Mach disk, existing in the free stream condition, becomes an attached oblique shock that extends downstream from the feed tube tip (Fig. 2a). The location of the attached oblique shock can be related to the location at which boundary layer separation begins on the exterior surface of the melt feed tube. This was investigated by Espina and Piomelli [10], who showed that the boundary layer separation can be retarded at a higher operating pressure, thus pushing the oblique shock further down the exterior surface of the feed tube. As seen in the cylindrical jet studies [16], the PrandtlMeyer expansion waves in the model are found downstream of the oblique shock. After the second series of PrandtlMeyer waves, the individual gas jets from the opposing sides of the annular slit-jet coalesce to form a unied gas ow eld with enlarged PrandtlMeyer wave structures, resembling the diamond patterns often seen in Schlieren imagining [1,13,16,17]. At a higher pressure (2.07 MPa), the contour of the recirculation zone lengthens into an hourglass shape (Fig. 3a). Here, the atomizing gas expands into the recirculation zone, constricting it. As the modeling pressure increases to 3.45 MPa (Fig. 4a), the recompression shock composed of PrandtlMeyer compression shocks becomes fully developed, emanating from the elongated waist of the recirculation zone. Due to expansion of the atomizing gas pushing laterally into the recirculation zone, an exaggerated hour-

glass contour results. The ow patterns generated in the CFD models of these open wake cases are similar to those observed in Schlieren imaging [5]. As the modeling pressure increases to 4.82 MPa (Fig. 5a), a closed-wake condition is observed, revealing a Mach disk structure. However, unlike the at Mach disk seen in Schlieren images [18], this computed Mach disk has a bowed geometry. Such a shape results from the fact that the gas velocity is unlikely to be uniform and normal across the Mach disk surface when an uneven gas stream exists immediately upstream from the Mach disk. The recirculation zone would contour the streamlines of the atomizing gas, such that the velocity vectors immediately in front of the Mach disk (on the upstream side) would have radial velocity components to them. This would give rise to the bowed shaped Mach disk. A bowed shaped Mach disk has been observed by Giel and Mueller [19] in a truncated plug nozzle that had a gas geometry similar to the ASJ-HPGA nozzle investigated. In fact, the bowed shape of the Mach disk implies that there is a velocity prole with magnitudes that are greater along the centerline than at the edge of the Mach disk. Previously, the gas ow dynamics of the recirculation zone of an HPGA nozzle, in gas-only operation, had been represented in the form of a simple control volume [1]. Although the proposed two-dimensional triangular control volume is an over-simplication of a real three-dimensional recirculation zone, the simple control volume model adequately serves to represent the general events in the recirculation zone; but, in light of the results from this CFD study, the model needs to be rened. According to the model [1], the gas enters the wake from the stagnation front and exits the wake along the circumferential edge of the melt tip base, and a series of turbulent eddies line the boundary of the recirculation zone (see Fig. 7a). From the results of this CFD study, however, the recirculation eddies are located within the sonic boundary (Fig. 7b), and are represented as a turbulent layer. Giel and Mueller also observed this turbulent layer within the sonic boundary [19]. The turbulent layer is essentially a series of adjoining recirculation eddies that form the lateral perimeter of the recirculation zone within the sonic boundary. The turbulent layer delineates the upstream owing gas in the core of the recirculation zone from the downstream owing gas that is found external to the recirculation zone. The gas velocity magnitude in the turbulent layer is low because the turbulent layer sees reversal in gas ow direction across its prole (see Figs. 25, and relative velocity magnitude in Fig. 8); however, the mass exchanges across the layer could be signicant. In most cases, though, the mass exchanges across the turbulent layer may be equal, i.e., the gas mass entering and exiting across the turbulent layer are comparable. Compared to the gas mass entering the stagnation front and its inuence on aspiration pressure, the gas mass exchange across the turbulent layer boundary may not have a signicant effect on aspiration pressure. The sonic boundary in this rened control volume

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

273

Fig. 7. (a) Control volume model reproduced from Ting et al. [2]; (b) rened control volume showing the recirculation zone within the sonic boundary (recirculation wake).

model (Fig. 7b) separates the supersonic gas ow external to the boundary and the subsonic ow gas within it. In this study, the subsonic gas owing along the circumferential edge was abducted to ow downstream between the turbulent layer and the sonic boundary (see Fig. 7b). Most of the gas does not cross the sonic boundary from the subsonic-side into the supersonic-side of the atomizing free-stream. This is because it is difcult for the abducted subsonic gas to accelerate itself to supersonic speed in order to enter the supersonic free-stream gas on the other side of the sonic boundary. However, the atomizing supersonic free-stream gas can more readily lose its kinetic energy, decelerating to subsonic velocity, in order to cross the sonic boundary and to enter into the subsonic recirculation wake. Some of this decelerated, entrained gas from the supersonic free-stream ends up at the stagnation front, feeding into the recirculation zone via the stagnation front. Subsequently,

this gas addition is pushed upstream along the central axis of the wake until it eventually affects the aspiration pressure at the melt orice. At the stagnation front, the remaining balance of the entrained gas exits the wake, accelerating to supersonic velocity to cross again the sonic boundary into the free-stream. The gasmass ux across the sonic boundary occurs in a direction nearly parallel to this boundary, analogous to vehicles that smoothly exit and enter a high-speed highway. The velocity gradient across the sonic boundary resembles the shear layer prole of a supersonic ow across a parallel stationary boundary (see Fig. 8), with the stationary boundary being the turbulent layer. Meanwhile the mass-ux across the turbulent layer occurs essentially in a direction perpendicular to the turbulent layer (see Fig. 8). This is a result of the numerous toroidal vortices (or, turbulent eddies) that give the gas its circular momentum as each vortex rotates.

Fig. 8. A schematic depiction of the gas velocity prole across the turbulent layer (M 0) and the sonic boundary (M = 1). V denotes velocity and M denotes Mach number of the gas velocity.

274

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

Since the gas behaves as a vortex within the turbulent layer, the net mass-ux across the turbulent layer is negligible. 5.2. Stagnation pressures and aspiration pressures Theoretically based turbulence models have been successful in predicting features in incompressible, free shear layers [14]. However, attempts to extend these models to supersonic speeds and compressible shear layers have been less effective. Although the compressible dissipation model proposed by Sarkar et al. demonstrated relative agreement between normalized turbulence intensities and shear stresses with convective gas ows in a simple shear layer geometry, this is still only qualitative agreement and for a simple free shear layer geometry [14]. An accurate turbulence model is difcult to formulate, and the computations are made more complicated by the fact that a turbulence model is most accurately developed when using a highly rened computational mesh. Therefore, a trade-off between accuracy and computational time associated with a rened computational mesh must be reached. In the present study, a reduced stagnation pressure at the stagnation front, specically in the primary recirculation zone, reduces the aspiration pressure at the melt orice. For example, at an atomization gas pressure of 4.82 MPa the reduced static pressure at the primary stagnation front (about 1.0 NDL downstream in Fig. 5c) gives rise to the deep subambient aspiration pressure shown in Fig. 6. On the other hand, a higher primary stagnation front pressure (at about 1.8 NDL in Fig. 3c) for 2.07 MPa, gives a slightly subambient aspiration in Fig. 6. The stagnation front pressure is 115 kPa after wake-closure (at 4.82 MPa), while before wake-closure, it is 195 kPa, at 2.07 MPa. This stagnation and aspiration pressure relationship agrees with a previous investigation; whereby it was found, in general, that a reduction in the static pressure (mass of gas) at the wake front correlated to a reduced aspiration pressure at the melt orice [1]. Contrary to expectations, the modeled aspiration pressures, given in Fig. 6, show a rising trend from a manifold pressure of 3.45 to 4.82 MPa, even though the CFD image in Fig. 4a shows no Mach disk formed at 3.45 MPa, i.e., the CFD ow image shows an open-wake condition. Not until the manifold pressure is raised to 4.82 MPa does the CFD model reveal a Mach disk, or closed-wake condition, as indicated in Fig. 5a. This ambiguity may be due to the inherent approximations in the turbulent model used, resulting in an inaccuracy in modeled aspiration pressures in these highly energetic gas ow conditions. The highly elongated hourglass shaped recirculation zone at 3.45 MPa (see Fig. 4a) adds to the computational aspiration inaccuracy within the recirculation zone, particularly at this manifold pressure. The narrow section of the recirculation zone inhibits the recirculating gas from reaching the melt orice and inuencing the aspiration pressure. This inhibition can be seen in the static pressure rise within the recirculation zone between 0.8 and 1.8 NDL downstream from

the orice (Fig. 4c), and in the slowing of the gas velocity magnitude within this same narrow section (Fig. 4b). The constriction, compounding the inaccuracy of the turbulent model, diminishes the inuence of the stagnation front pressure on the aspiration pressure at the melt orice. Because the mass of gas is constricted at the neck, less recirculated gas is able to reach the melt orice. This gives rise to an increase in static pressure in the recirculation zone (Fig. 4c). Thus, the aspiration at 3.45 MPa is deeply subambient, articially resembling that of a closed-wake condition. As Fig. 6 illustrates, the total variation in computed aspiration pressures is greater than those experimentally measured for the open-wake and closed-wake conditions, which also suggests the inaccuracy of the turbulence model used in the CFD calculations. Espina and Piomelli [10] have also observed similar correlation deciencies between calculated and experimental aspiration pressure values. The atomization gas (manifold) pressure for CFD-computed wake-closure conditions appears to be lower (4.82 MPa, or less) than the experimentally measured value of about 5.5 MPa. Although both computational models and experimental measurements are based on the DJ-HPGA design, the effective annular-slit gas exit areas are different between the computed and experimental congurations. The computed AS-HPGA model has an effective jet area of 25.7 mm2 that is 300% larger than the experimental AS-HPGA jet area: 8.52 mm2 . The larger annular slit area may contribute, along with a sub-optimal turbulence model, to the production of a lower wake closure pressure in the CFD model. However, experimental verication is needed to conrm this possible explanation. 5.3. Closed wake and mach disk instability In the closed-wake condition, at 4.82 MPa, the maximum stagnation front static pressure downstream of the Mach disk (see Fig. 5c) is 28% greater (320 kPa versus 250 kPa) than the maximum stagnation front static pressure (see Fig. 4c) in the open-wake condition at 3.45 MPa. Furthermore, the maximum (third) stagnation point observed in the closed-wake condition (at about 3.5 NDL downstream in Fig. 5c) is considerably larger than the stagnation front pressure in all the other open-wake conditions modeled. Interestingly, at 4.82 MPa, the stagnation front pressure (120 kPa) in the primary recirculation zone (at about 1.0 NDL in Fig. 5c) is approximately 4050% lower than the stagnation front pressures at all the open-wake conditions modeled. Although this is not an average of the pressure in the recirculation zone, it however, as mentioned before, represents the gas mass entering the recirculation zone and inuencing the aspiration pressure. In other words, in the closed-wake condition (Fig. 5c), there are stagnation pressures upstream and downstream of the Mach disk that are lower and higher than all the computed stagnation pressures in the open-wake conditions (Figs. 2c, 3c and 4c). This dichotomy in stagnation front pressures lends support for

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276

275

the existence of a high instability in the Mach disk that was proposed in the pulsatile atomization model [3,4,6]. An intense recirculation zone is observed immediately downstream of the Mach disk, as indicated in Fig. 5a. It is formed as a result of having a strong gas velocity disparity (Fig. 5b) between the gas immediately upstream and downstream of the bowed Mach disk. The retarded gas velocity downstream of the Mach disk causes the formation of this secondary recirculation zone and the two stagnation points within this recirculation zone (Fig. 5). In addition, at the closed-wake condition the primary recirculation zone is isolated from the secondary recirculation zone, and the bowed Mach disk blocks the ow of gas upstream from the secondary recirculation zone to the primary recirculation zone. One notices that the upstream owing gas in the secondary recirculation zone is turned downstream as it nears the Mach disk (Fig. 5). Therefore, one observes a deep subambient aspiration pressure in the closed-wake due to the gas-starved stagnation front in the primary recirculation zone. According to a previous study, an aspiration discontinuity was observed across the wake-closure pressure [6]. It was hypothesized that the state of entropy of the atomizing gas increases with the formation of the Mach disk when the atomizing gas transitions from an open-wake to a closed-wake condition. However, the increased entropy is accompanied by a decrease in total energy of the gas ow pattern which prevents reversion back to an open wake condition. Thus, it is reasonable to claim that the physical continuity of the shock structures, between the internal shock and the bowed Mach disk, is necessary for the stability of the closed-wake condition. In other words, the Mach disk, in the closed-wake condition, would be destroyed if the internal shock structures were displaced from their equilibrium position by a strong disruptive force, i.e., when a high-density melt is introduced into the primary recirculation zone during the melt atomization process. 5.4. Atomization implications As proposed in the pulsatile atomization model, the melt owing from the delivery tube orice acts as a moving boundary that could disrupt the stability of the closed-wake structure by displacing the internal shock. In this instance, the Mach disk vanishes, the stagnation pressure from the secondary recirculation zone rushes into the opened primary recirculation zone, pushes upwards against the melt orice, and temporarily halts the melt ow. This scenario is supported by the fact that the magnitude of the stagnation front pressure in the secondary recirculation zone of the closed-wake condition is considerably higher than the stagnation front pressure in the primary recirculation zone. This pressure gradient would drive the gas in the secondary recirculation zone to ow towards the melt orice when the bowed Mach disk is disrupted by the melt ow. Subsequently, the melt ow is briey halted by the strong upstream gas ow. Next, as described in the pulsatile at-

omization model [3,4,6], the disrupted melt ow enables the reestablishment of the Mach shock structures. The deep aspiration pressure of the reestablished closed-wake condition vigorously draws the melt down the ceramic pour tube, reinitiating melt ow once again. The whole intermittent melt atomization process would repeat itself, pulsating, until the melt is exhausted.

6. Conclusions This study provided additional insight into the relationship between aspiration pressure at the melt orice and stagnation pressure in the primary recirculation zone under both open-wake and closed-wake conditions. This study also described the anatomical gas prole within the conical shear layer arrangement in the sonic boundary and also explained the different methods of gas transport occurring across the sonic boundary and the turbulent layer. In the open-wake condition, the calculated aspiration pressure reveals that the stagnation pressure is decreased as manifold pressure is increased to the onset point of stable atomization, i.e., when aspiration pressure drops to one atmosphere. Over all three open wake calculations, the stagnation front position moves downstream as the manifold pressure is increased. The deep subambient aspiration pressure at the wakeclosure condition is due to the existence of a relatively low-pressure stagnation front in the primary recirculation zone. The study also reveals that there exists a secondary recirculation zone downstream of the Mach disk structure, in the closed-wake condition, bounded by two relatively high-pressured stagnation points. These pressures are signicantly greater than the static pressure in the primary recirculation zone of the closed-wake condition, and the stagnation pressures in all the open-wake conditions modeled. The high-pressured recirculating gas behind the Mach disk ows rapidly upstream towards the melt orice if the Mach disk is destabilized during melt atomization. This calculated nding lends support to the governing effects of the proposed pulsatile atomization model [3,4,6].

Acknowledgements The CFD work was funded by the Materials Science Division of DOE-BES at Ames Laboratory under contract number W-7405-Eng-82. The authors would like to thank Jia Mi for developing and performing the CFD model calculations and Murielle Ting for reviewing the manuscript.

References
[1] J. Ting, I.E. Anderson, R. Terpstra, J. Mi, Design and testing of an improved convergentdivergent discrete-jet high pressure gas atomization nozzle, in: J.J. Oakes, J.H. Reinshagen (Eds.), Advances in

276

J. Ting, I.E. Anderson / Materials Science and Engineering A 379 (2004) 264276 powder metallurgy & particulate materials1998, APMI-MPIF, vol. 3, Part 10, Princeton, NJ, 1998, pp. 2939. H. Lubanska, J. Met. 2 (22) (1970) 4549. J. Ting, M.W. Peretti, W.B. Eisen, Control of ne powder production and melt ow rate using gas dynamics, in: H. Ferguson, D.T. Whychell (Eds.), Advances in Powder Metallurgy & Particulate Materials2000, APMI-MPIF, vol. 2, Princeton, NJ, 2000, pp. 2740. J. Ting, M.W. Peretti, W.B. Eisen, Mat. Sci. Eng. A326 (2002) 110 121. T.J. Mueller, W.P. Sule, A.F. Fanning, T.V. Giel, F.L. Galanga, Analytical and Experimental Study of Axisymmetric Truncated Plug Nozzle Flow Fields, UNDAS TN-601-FR-10, 1972. J. Ting, M.W. Peretti, W.B. Eisen, Control of ne powder production and melt ow rate using gas dynamics, in: W.B. Eisen, S. Kassam (Eds.), Advances in Powder Metallurgy & Particulate Materials2001, APMI-MPIF, vol. 2, Princeton, NJ, 2001, pp. 116. M. Glogowski, et al. Shear Coaxial Injector Instability Mechanisms, AIAA, 1994, pp. 942774. M. Zhu, A.P. Dowling, K.N.C. Bray, Self-excited Oscillations in Combustors with Spray Atomizers, International Gas Turbine & Aeroengine Congress & Exhibition, Munich, Germany, May 811, 2000-GT-108, 2000. Z. Fargo, N. Chigier, Atomization and Sprays 2 (1992) 137153. P.I. Espina, U. Piomelli, Numerical Simulation of the Gas Flow in GasMetal Atomizers, ASME Paper, FEDSM98-4901, 1998. [11] J. Mi, R.S. Figliola, I.E. Anderson, Metall. Mat. Trans. B 28B (10) (1997) 935941. [12] D.W. Kuntz, J.L. Payne, Simulation of powder metal fabrication with high pressure gas atomization, in: M. Philips, J. Porter (Eds.), Advances in Powder Metallurgy & Particulate Materials1995, APMI-MPIF, vol. 1, Princeton, NJ, 1995, pp. 6378. [13] I.E. Anderson, R.S. Figliola, H. Morton, Mat. Sci. Eng. A148 (1991) 101114. [14] S. Sarkar, L. Balakrishnan, Applications of a Reynold-Stress Turbulence Model to the Compressible Shear Layer, ICASE Report 90-18, NASA CR 18002, 1990. [15] R.D. Zucker, Fundamentals of Gas Dynamics, Matrix Publishers Inc., Chesterland, OH, 1977. [16] J. Ting, R. Terpstra, I.E. Anderson, R.S. Figliola, J. Mi, A novel high pressure gas atomizing nozzle for liquid metal atomization, in: Marcia Philips, John Porter (Eds.), Advances in Powder Metallurgy & Particulate Materials1996, APMI-MPIF, vol. 1, Princeton, NJ, 1996, pp. 97108. [17] D. Bergmann, U. Heck, U. Fritsching, K. Bauckhage, Particle size distribution in gas atomization of molten metals, in: M. Philips, J. Porter (Eds.), Advances in Powder Metallurgy & Particulate Materials1999, APMI-MPIF, vol. 1, Princeton, NJ, 1999, p. 57. [18] I.E. Anderson, R.S. Figliola, H. Morton, Mat. Sci. Eng. A148 (1991) 101104. [19] T.V. Giel Jr., T.J. Mueller, The Mach Disk in Truncated Plug Nozzle Flow, AIAA Paper No. 75886, 1975.

[2] [3]

[4] [5]

[6]

[7] [8]

[9] [10]

You might also like