Materials Transactions, Vol. 48, No. 6 (27) pp. 1463 to 1468 #27 Japan Thermal Spraying Society Numerical Simulation on Impact Velocity of Ceramic Particles Propelled by Supersonic Nitrogen Gas Flow in Vacuum Chamber Hiroshi Katanoda, Minoru Fukuhara and Naoko Iino Department of Mechanical Engineering, Kagoshima University, Kagoshima 89-65, Japan A low-pressure cold spray, which is conducted in a vacuum chamber, is under development in Japan. In this paper, the gas flow-field as well as the particle velocity of the low-pressure cold spray is numerically solved. A special attention is paid to the effect of the pressure in the vacuum chamber (back pressure) on the particle velocity. The working gas is nitrogen, and its stagnation temperature upstream of the nozzle is set at 573 K. The back pressure is set at constant values ranging from 3 1 2 to 1 1 5 Pa. The stagnation pressure upstream of the nozzle is kept constant at 3 times as much as the back pressure. The numerical results show that the decrease in the back pressure causes the decrease in the particle velocity in front of the normal shock wave. On the contrary, the decrease in the back pressure eases the particle deceleration through the normal shock wave. As a whole, due to the balance of the effects of the back pressure and the normal shock wave, the optimum value of the back pressure to obtain the maximum impact velocity varies depending on the particle diameter. [doi:1.232/matertrans.t-mra27833] (Received November 16, 26; Accepted February 26, 27; Published May 25, 27) Keywords: low density supersonic jet, shock wave, gasdynamics, numerical simulation, particle velocity 1. Introduction A cold spray process 1) is normally conducted in an atmospheric pressure by using electrically heated nitrogen or helium gas to propel spray particles. The process can deposit particles of a wide variety of metals and alloys. However, light particles, such as ceramic particles smaller than 1 mm, can not be deposited by the process. This is because a supersonic gas flow directed at the substrate is abruptly decelerated to a subsonic speed through a normal shock wave (NSW) in front of the substrate. 2) The spray particle injected in the upstream part of the nozzle is monotonously accelerated by the gasdynamic drag force until the exit of the nozzle, because the gas velocity is generally higher than the particle velocity in the nozzle. However, when the particle passes through the NSW after exiting the nozzle, the gas velocity suddenly becomes smaller than the particle velocity, causing the decelerating gasdynamic drag force on the particle. As a result, light particles, such as ceramic particles smaller than 1 mm, largely suffer from the decelerating gasdynamic drag force causing an insufficient impact velocity to deposit on the substrate. Because of this reason ceramic coatings deposited by the cold spray process has not been reported in open literatures as far as the authors know. In recent years, a low-pressure cold spray 3) (LPCS) and an aerosol deposition method 4) have succeeded in depositing fine ceramic particles on a substrate in a vacuum chamber by using a high-speed nitrogen or helium jet. A schematic diagram of the LPCS is shown in Fig. 1 The inert gas of stagnation pressure p s and stagnation temperature T s is discharged into a vacuum chamber of pressure p b and temperature T b through a supersonic nozzle. The reason why the LPCS is conducted in the vacuum chamber is to obtain higher particle impact velocity by reportedly easing the deceleration effect of the particle in the region between the NSW and the substrate. That is, when the pressure in the vacuum chamber (back pressure, p b ) is lowered, the gas density between the NSW and the substrate also becomes Gas Particle Fig. 1 ps, Ts thin, probably resulting in a small amount of deceleration in the particle velocity. Therefore, the setting value of the back pressure is considered to be an important parameter for the LPCS to control the impact velocity of the particle and therefore quality of the coating. Presently, however, the research on the LPCS is restricted to experimental works, therefore the effect of the back pressure on the gas/particle velocity has not been clarified yet. In this paper numerical simulations of gas/particle flows in a vacuum chamber are performed ranging from standard cold spray to LPCS by varying the value of the back pressure. The stagnation pressure upstream of the nozzle against the back pressure is kept constant at 3 in all cases of the present simulation. The effect of the back pressure on the gas flow and the particle impact velocity are numerically clarified. 2. Numerical Method Nozzle Vacuum pump Substrate Jet Normal shock wave pb, Tb Schematic diagram of low-pressure cold spray. 2.1 Gas flow The gas flow is assumed to be two-dimensional axisymmetric in the computational fluid dynamics (CFD) model.
1464 H. Katanoda, M. Fukuhara and N. Iino φ 7 φ 2 5 1 1 7 φ the nozzle. This grid size was found to be enough to obtain an almost mesh-size-independent solution; the finer computational grids, 225 6 and 9 135 for inside and outside the nozzle, respectively, showed negligible change in the gas velocity along the center line. Fig. 2 Nozzle geometry (unit:mm). The governing equations of the gas flow are given by the conservation form of the two-dimensional axisymmetric, time-dependent Navier-Stokes equations along with the k-" turbulence model to include the effect of turbulence of the gas flow. The governing equations are solved sequentially in an implicit, iterative manner using a finite difference formulation. For the present calculations, the governing equations are solved with the Chakravarthy-Osher type thirdorder, upwind, total variation diminishing scheme. The working gas in the CFD model is nitrogen. The gas is assumed to follow the equation of state. The back pressure p b was set at constant ranging from 3 1 2 to 1 1 5 Pa. The pressure ratio PR, the stagnation pressure upstream of the nozzle p s against the back pressure p b, was kept constant at PR ¼ 3. The stagnation temperature upstream of the nozzle T s and that outside the nozzle T b were set at 573 and 3 K, respectively. Since this paper investigates the effect of the back pressure on the particle velocity of the cold spray, the commonly used gas conditions in cold spray application (PR ¼ 3 and T s ¼ 573 K) are used in the CFD. Figure 2 shows the geometry of an axisymmetric conical nozzle used in the CFD model. The nozzle has a throat diameter of 2 mm, exit diameter of 7 mm and a diverging-part length of 1 mm. The nozzle geometry like Fig. 2 is one of the typical ones which is often used in the cold spray. The Reynolds number R ed based on the gas density, velocity, viscosity at the nozzle exit, and the nozzle-exit diameter is in the range of 4:5 1 2 to 1:5 1 5 from the theory of the steady one-dimensional isentropic flow. 5) According to Troutt and McLaughlin 6) the jet flow is laminar for R ed < 1 1 3. Therefore, no turbulence model was used to simulate the laminar flow for the conditions of R ed < 1 1 3 in the CFD model. According to the theories of the steady one-dimensional isentropic flow and NSW, the nozzle shown in Fig. 1 and the numerical condition PR ¼ 3 are expected to generate an over-expanded flow including NSW or oblique shock wave in the nozzle. The size of the computational grid used in this simulation is 15 4 grids inside the nozzle, and 6 9 grids outside 2.2 Particle flow The following assumptions are used to simplify the calculations of particle velocity and temperature. (1) The particles are spherical in shape. (2) The interaction between particles can be ignored. (3) The only force acting on a particle is gasdynamic drag force. (4) The presence of particles has a negligible effect on the gas velocity and temperature. (5) The particles have a constant specific heat and a constant density. (6) The temperature in the particle is uniform. As a result of the assumptions made above, the gas-solid two-phase problem can then be independently solved. One can simulate the gas flow first, then use the resulting thermal and velocity fields to study the flow of different particles. The particle velocities were determined from a step-wise integration of their equations of motion under the influence of gasdynamic drag force. In this paper, only the particle motion along the center line is calculated. The governing equation for momentum transfer between a single particle of mass m p and the gas can be written as: du p m p ¼ 1 dt 2 c d g ðu g u p Þju g u p ja p ð1þ where, u p the particle velocity, u g the gas velocity in the axial direction, g the gas density, t the time, c d the drag coefficient of the particle and A p the projected area of the particle. The drag coefficient, c d, is a function of the particle Reynolds number, R ep, and the particle Mach number, M p, defined by the following equations: R ep ¼ gju g u p jd p ð2þ g M p ¼ ju g u p j ð3þ a g where, d p is the particle diameter, g the viscosity of the gas and a g the speed of sound of the gas. The value of c d for R ep 1 was obtained from a database made from the experimental data, 7) and that for R ep < 1 was calculated by the equations proposed by Henderson 8) given as: 2 8 3:65 1:53 T 19 3 1 p >< T >= g c d ¼ c d;sub ¼ 24 R ep þ S 4:33 þ B @ 1 þ :353 T C exp :247 R ep 6 7 4 pw A S 5 >: >; T g þ exp :5M!" p p 4:5 þ :38ð:3Rep þ :48 ffiffiffiffiffiffi # R epþ pffiffiffiffiffiffi p R ep 1 þ :3R ep þ :48 ffiffiffiffiffiffi þ :1Mp 2 R þ :2M8 p ep þ 1 exp M p :6S R ep for < M p < 1 ð4þ
Numerical Simulation on Impact Velocity of Ceramic Particles Propelled by Supersonic Nitrogen Gas Flow in Vacuum Chamber 1465 Fig. 3 Simulated Mach number contour (a) p b ¼ 1 1 5 Pa, (b) p b ¼ 3 1 2 Pa. c d ¼ c d;sub ¼ :9 þ :34 M 2 p þ 1:86 Mp R ep 2 2 þ 2 S þ 1:58 Tp 1S 2 S T 4 g sffiffiffiffiffiffiffi for M p > 1:75 ð5þ 1 þ 1:86 M p R ep c d ¼ c d;sub þ 4 3 ðm p 1Þðc d;sup c d;sub Þ for 1 < M p < 1:75 ð6þ p ffiffiffiffiffiffiffi where S is M p =2, Tg the gas temperature, T p the particle temperature. The particle temperature was calculated by numerically integrating the unsteady heat transfer equation of a sphere particle in a gas flow, as explained in the literature. 9) The powder material used in the simulation is Al 2 O 3, and its density was set at 3,9 kg/m 3. The range of the particle diameter was selected as.5 15 mm. The spray particle is injected at 15 mm upstream of the nozzle throat on the center line. The velocity and temperature of the particle at the injection position are set as 1 m/s and 3 K, respectively. The reliability of the present numerical method is verified in Ref. 1), in which the gas flow field is similar to the present cases, by comparing the simulated particle velocity of ceramic particle with the experimental results. 3. Results and Discussion 3.1 Gas flow The simulated Mach number contours are shown in Fig. 3. Figure 3(a), the back pressure p b ¼ 1 1 5 Pa, shows a typical cold spray flow. The axial distance measured from the nozzle exit, x, is shown in the bottom side of the figure. Figure 3(a) shows that the gas flow is accelerated to supersonic flow through the throat in the downstream direction. Then, the Mach number reaches the maximum value of M g ¼ 3:66 at x 1 mm, around where the flow separates from the nozzle wall. The reason why flow separation occurs is that according to the steady onedimensional isentropic theory, the static pressure at M g ¼ 3:66 becomes :31 1 5 Pa, which is one-third of the back pressure 1 1 5 Pa. Therefore, the higher back pressure prevents the gas flow in the nozzle from expanding further in the downstream direction, causing the gas flow to decelerate. When the gas flow is decelerated the gas velocity in the boundary layer next to the nozzle-wall directs upstream, resulting in the flow separation. The flow with separation is often seen in over-expanded nozzle flows. 11) After the flow separates at x 1 mm in Fig. 3(a), the cross-sectional area of the gas flow converges in the downstream direction, meaning the decrease in the Mach number. Figure 3(b) shows the Mach number contour of LPCS for the back pressure p b ¼ 3 1 2 Pa, which is the lowest value in the present calculation. The maximum Mach number in the nozzle is M g ¼ 2:4 at x 4 mm, and it is as small as 56% of M g ¼ 3:66 for Fig. 3(a). The reason why the gas flow in Fig. 3(b) is not so accelerated is due to the low Reynolds number (low density of the gas), the boundary layer develops thicker along the nozzle wall. The thick boundary layer
1466 H. Katanoda, M. Fukuhara and N. Iino 12 1 Throat p b = 1 1 5 Pa Normal shock wave 12 1 d p =.5µm, 1µm 8 8 u g /m s -1 6 4 p b = 1 1 4 Pa p b = 1 1 3 Pa u p /m s -1 6 4 Gas d p = 5µm, 1µm 2 Nozzle region 2 Nozzle region Normal shock wave -12-1 -8-6 -4-2 2 x/ mm -3-2 -1 1 x/ mm Fig. 4 Gas velocity along center line. Fig. 5 Gas/particle velocity on center line for p b ¼ 1 1 5 Pa. reduces the effective cross-sectional area of the gas flow. The small cross-sectional area of the flow results in a small amount of expansion of the gas in the nozzle. Then, small amount of expansion of the gas in the nozzle results in the static pressure to be above the back pressure at the nozzle exit, causing no flow separation. The gas velocity along the center line is shown in Fig. 4 for three values of back pressure. For p b ¼ 1 1 5 Pa of atmospheric pressure, the gas flow is sharply accelerated through the nozzle throat, then the gas is gradually accelerated in the downstream direction. However, the gas flow starts deceleration at x 1 mm due to the contraction of cross-sectional area of the gas flow induced by the flow separation, as can be seen in Fig. 3(a). After exiting the nozzle, the gas flow is suddenly decelerated by the NSW at x ¼ 9 mm, then the gas flow becomes gradually stagnant on the substrate at x ¼ 1 mm. For p b ¼ 1 1 4 Pa, the gas velocity is slightly smaller than that of p b ¼ 1 1 5 Pa in the range of 95 mm < x < 25 mm due to the thicker boundary layer along the nozzle wall. The separation point for p b ¼ 1 1 4 Pa is upstream compared to that of p b ¼ 1 1 5 Pa. This is caused by the reduction of turbulence in the boundary layer due to one-tenth of the gas density of p b ¼ 1 1 5 Pa case, resulting in smaller gas velocity in the boundary layer. The smaller the turbulence is, the smaller the gas velocity in the boundary layer becomes, then, the easier the flow separates in the upstream part of the nozzle for overexpanded flows. For p b ¼ 1 1 3 Pa, the gas velocity is much smaller than that for p b ¼ 1 1 5 Pa in the region 95 mm < x < 1 mm, due to one-hundredth of the gas density of p b ¼ 1 1 5 Pa case. There is no flow separation for p b ¼ 1 1 3 Pa. The locations of the NSW are not affected by the reduction of the back pressure. 3.2 Particle flow The particle velocity, as well as the gas velocity, along the center line in the downstream part of the nozzle is shown in Fig. 5 for p b ¼ 1 1 5 Pa. The smallest.5 mm particle shows highest sensitivity to the variation in the gas velocity. Therefore, although the.5 mm particle has a high velocity of 853 m/s right before the NSW at x ¼ 9 mm, the velocity decreases to 316 m/s on the substrate. The larger particle shows the smaller amount of deceleration through the NSW due to larger inertia. u pi /m s -1 1 8 6 4 2 d p =.5µm, 1µm d p = 5µm, 1µm 1 2 1 3 1 4 1 5 2 4 6 8 2 4 6 8 2 4 6 8 Fig. 6 p b /Pa Particle impact velocity. The particle impact velocity, u pi, is summarized in Fig. 6 as a function of the back pressure. The value of u pi is the particle velocity when it impinges on the substrate. As can be seen in the figure, the dependency of u pi on the back pressure is totally different for the diameter smaller than 1 mm and that larger than 5 mm. That is, by decreasing the back pressure p b, u pi of the particle smaller than 1 mm increases to the maximum value, then further decrease in the back pressure decreases u pi. On the other hand, u pi of the particles larger than 5 mm simply decreases by decreasing the back pressure. As can be seen in Figs. 5 and 6, the particle impact velocity depends on three factors; 1) particle diameter, 2) back pressure, and 3) deceleration after the NSW. In order to clarify the extent of contribution of the three factors independently, the authors consider dimension-free velocity ratio; VR u pi =u ge, instead of u pi itself. The theoretical nozzle-exit gas velocity, u ge, in the definition of VR is calculated as 961 m/s by the steady one-dimensional isentropic theory. The gas velocity u ge depends on the nozzle geometry, the type of working gas, stagnation temperature, but not the stagnation pressure (or the back pressure). Therefore u ge is a constant value for the present numerical conditions. Then, the velocity ratio, VR, can be written as follows: u pi ¼ VR ¼ VR CS VR pb VR NSW ð7þ u ge
Numerical Simulation on Impact Velocity of Ceramic Particles Propelled by Supersonic Nitrogen Gas Flow in Vacuum Chamber 1467 Velocity ratio Velocity ratio Velocity ratio 1.2 1..8.6.4.2 1.2 1..8.6.4.2 1.2 1..8.6.4.2 VR VR CS ðu psþ CS u ge ; VR pb u ps ðu ps Þ CS ; 5 5 5 VR VR pb d p / µm VR NSW u pi u ps where, u ps is the particle velocity on the center line right before the NSW, ðu ps Þ CS is u ps for the cold spray condition; p b ¼ 1 1 5 Pa. In Eqs. (7) and (8), VR CS means the particle velocity right before the NSW for the cold spray condition, divided by u ge. By definition, VR CS becomes larger for the smaller particle diameter (smaller mass), but is independent of the back pressure p b. Next, the value of VR pb means the particle velocity right before the NSW, divided by that of the cold-spray-condition value. That is, VR pb shows the effect of VR (a) d p / µm (b) VR pb d p / µm 1 1 1 VR NSW VR CS VR NSW VR CS VR pb VR NSW VR CS Fig. 7 Velocity ratios versus particle diameter (a) p b ¼ 1 1 5 Pa, (b) p b ¼ 1 1 4 Pa, (c) p b ¼ 1 1 3 Pa. (c) 15 15 15 ð8þ the back pressure on the particle velocity right before the NSW. Finally, the value of VR NSW means the particle impact velocity divided by the particle velocity right before the NSW. Therefore, VR NSW indicates the extent of the deceleration in the particle velocity by the NSW. The velocity ratios of VR CS, VR pb, fi VR NSW and fl VR are investigated and shown in Fig. 7, as a function of particle diameter d p. In Fig. 7(a), p b ¼ 1 1 5 Pa, the dotted line of VR CS increases for smaller d p, showing that the velocity of smaller particle approaches the gas velocity before the NSW. On the contrary, the solid line of fi VR NSW decreases for smaller d p, indicating that the particle velocity experiences larger amount of deceleration after the NSW for smaller d p. The dimension-free impact velocity, fl VR, takes the maximum value at d p ¼ 5 mm. Considering the curves of to fl in Fig. 7(a), u pi increases by decreasing d p for d p > 5 mm mainly by the reduction of the mass. Whereas, u pi decreases by decreasing d p for d p < 5 mm due to the deceleration effect by the NSW. The value of VR pb in Fig. 7(a) is unity regardless of d p because of the definition in Eq. (8). In Fig. 7(b), the velocity ratios for p b ¼ 1 1 4 Pa are shown. The dotted line of VR CS is the same one as plotted in Fig. 7(a) by the definition in Eq. (8). Figure 7(b) clearly shows that by reducing p b from (a) 1 1 5 Pa to (b) 1 1 4 Pa, the value of VR pb for d p > 1 mm decreases for larger d p. This means that the larger the particle diameter for d p > 1 mm, the larger the deceleration of the particle by the reduction of the back pressure. In addition, the deceleration effect by the NSW is apparently eased in Fig. 7(b) compared to Fig. 7(a). The curves to fl show that the decrease in u pi for d p < 2 mm is mainly caused by the NSW. The tendency of curves VR pb and fi VR NSW result in the maximum u pi at smaller d p ¼ 2 mm in Fig. 7(b) than d p ¼ 5 mm in Fig. 7(a). Figure 7(c) shows the velocity ratios for p b ¼ 1 1 3 Pa. The curves to fl in the figure show that the decrease in u pi for d p > 1 mm is mainly by the reduction of the back pressure. The value of u pi for d p < 1 mm increases in Fig. 7(c) than that in Fig. 7(b) due to the disappearance of the deceleration effect of the NSW in Fig. 7(c). The value of d p corresponding to the maximum u pi moves more left in Fig. 7(c) than that in Fig. 7(b). When the back pressure is reduced to 3 Pa, not shown here, the curves of to fl are similar to those in Fig. 7(c), however, the maximum u pi for d p < 1 mm decreases due to the effect of reduced back pressure. Finally, from Fig. 7 we can see the most important feature; the value of the back pressure to obtain the maximum u pi. For the particle diameter d p 2 mm, p b should be around 1 1 4 Pa to obtain the maximum u pi because of the balance of the effects of the NSW and the reduced back pressure. For the particle diameter d p < 1 mm, p b should be around 1 1 3 Pa because of the effect of the reduced back pressure and the absence of the effect of NSW. 4. Conclusions The numerical simulation was conducted to clarify the effect of the back pressure on the impact velocity of Al 3 O 3 particle in the low-pressure cold spray. The working gas was nitrogen, and its stagnation temperature upstream of the
1468 H. Katanoda, M. Fukuhara and N. Iino nozzle was set at 573 K. The back pressure was set at constant values ranging from 3 1 2 to 1 1 5 Pa. The stagnation pressure upstream of the nozzle was kept constant at 3 times as much as the back pressure. The results obtained by the present research are summarized as follows; (1) The decrease in the back pressure causes the decrease in the particle velocity in front of the normal shock wave. This effect is more serious for the larger diameter particle. However, when the back pressure is decreased below 1 1 3 Pa, the velocity of particles smaller than 1 mm is also decreased. (2) The decrease in the back pressure eases the particle deceleration through the normal shock wave in front of the substrate. (3) Due to the balance of the effects of the back pressure and normal shock wave, the impact velocity for the particle diameter of around 2 mm becomes maximum at the back pressure of around 1 1 4 Pa. For the particles of less than 1 mm, the back pressure of around 1 1 3 Pa provides the maximum impact velocity. REFERENCES 1) A. Papyrin: Advanced Materials & Processes 159 (21) 49 51. 2) D. L. Gilmore, R. C. Dykhuizen, R. A. Neiser, T. J. Roemer and M. F. Smith: J. Thermal Spray Technol. 8 (1999) 576 582. 3) N. Ohno and H. Fukanuma: Proc. Nat. Joint Conf. on Thermal Spray 25, ed. by Jpn. Thermal Spraying Soc. and High Temp. Soc. Jpn. (25) pp. 27 28. 4) J. Akedo and M. Lebedev: Jpn. J. Appl. Phys. 4 (21) 5528 5532. 5) M. J. Zucrow and J. D. Hoffman: Gas Dynamics, (John Wiley & Sons, NY, 1976) pp. 16 181. 6) T. R. Troutt and D. K. McLaughlin: J. Fluid Mech. 116 (1982) 123 156. 7) A. B. Bailey and J. Hiatt: AIAA J. 1 (1972) 1436 144. 8) C. B. Henderson: AIAA J. 14 (1976) 77 78. 9) H. Katanoda: Mater. Trans. 47 (26) 2791 2797. 1) H. Katanoda and K. Matsuo: Trans. Mater. Res. Soc. Jan. 31 (26) 13 16. 11) H. D. Kim, K. Matsuo and T. Setoguchi: Shock Waves 6 (1996) 275 286.