Development of a High Transfer Efficiency Painting Technology Using Effervescent Atomization

Transcription

1 Development of a High Transfer Efficiency Painting Technology As presented at: ILASS Americas, 18th Annual Conference on Liquid Atomization and Spray Systems, Irvine, CA, May 005 Charles Tricou, Applied Research Laboratory: Penn State University University Park, PA Keith F. Knasiak Spraying Systems Co. Wheaton, IL USA Abstract The objective of this project is to develop a high-efficiency paint application system. The overall goal is to reduce paint usage and eliminate cleanup costs associated with paint overspray. In many industrial applications, including shipbuilding, overspray from painting operations generates substantial added cost associated with material waste, cleanup costs and environmental emissions. This work involves the investigation of a novel, effervescent spray atomization technique. An experimental design approach was employed to identify critical parameters. This methodology used a four (4) factor D-Optimal Response Surface Analysis. The design factors studied for the effervescent atomization process are the number of holes in the gas (N ) injector, the gas-to-liquid ratio (GLR), the total area of the holes in the gas injector and the pressure. The responses used in this analysis are: transfer efficiency, spray fan quality, D V0.5, D V0.1, D V0.9 and D 3. A prototype high-efficiency, effervescent paint application system was designed, built and optimized with the intention of maximizing the transfer efficiency. The results of this investigation show that the effervescent atomization technique is capable of achieving transfer efficiencies of 95% or higher. The Response Surface Analysis identified strong correlations between the factors studied and measured responses. These results also show that the pressures required to obtain a good spray pattern with high transfer efficiency are much lower for effervescent atomization than high pressure hydraulic atomization. Experts in Spray Technology Spray Nozzles Spray Control Spray Analysis Spray Fabrication

2 Introduction The application of anti-corrosive and anti-foulant coatings to the underwater hull of naval vessels is performed using the airless paint spray process with transfer efficiencies in the range of 40% 60% [1]. There are substantial economic and environmental reasons to increase transfer efficiency and reduce overspray. The Applied Research Lab (ARL) was tasked to investigate effervescent atomization for the spray application of marine coatings. Effervescent atomization is an atomization technique in which a compressed gas is injected into a fluid slightly upstream of the nozzle orifice, creating a bubbly-mix of compressed gas and fluid. Upon exiting the nozzle, the small bubbles of compressed gas expand rapidly, breaking up the fluid to be atomized. Effervescent atomization has been used successfully in the combustion of pulp and paper mill waste for energy recovery. The developmental process was focused on technologies compatible with marine coating operations and on technologies having a high likelihood of success in shipyard production environments. Such applications and technologies have the following characteristics: High material flow rate or production rate High-solids and highly viscous materials Robust operating regime Low-technology processes ARL utilized a design of experiments (DOE) approach to investigate the effect of gas-to-liquid ratio (GLR), pressure, and gas injector tube geometries on transfer efficiency, spray particle size and pattern quality. Control of paint droplet size and pattern quality was demonstrated for an effervescent atomization spray device using gas-to-liquid ratio (GLR), paint pressure, and to a lesser extent, total area of the gas injection tube holes. The application of marine coatings to naval and commercial ships is performed almost exclusively using the airless paint spray process. The airless paint spray process is well-suited for the application of marine coatings, as most marine coatings are highly viscous and high film thicknesses are desired. The transfer efficiency of the airless paint spray process is dependent upon a variety of factors, including orifice size, paint viscosity, fluid pressure, part size, operator skill level, distance from the substrate and environmental conditions such as wind speed or cross-flow air velocity. The transfer efficiency of the airless paint spray process varies substantially but is generally accepted as being within the range of 40% 60%. Figure 1 shows the approximate transfer efficiency obtained using various methods for coating application [1]. Figure 1. Approximate Transfer Efficiency of Various Paint Application Technologies.

3 EQUIPMENT & METHODS Background Transfer efficiency is defined as the ratio of the weight of coatings solids deposited on a substrate to the total weight of coatings solids used in a coating application step, expressed as a percentage []. Aside from cleaning of lines and pumps, the primary source of material loss during the airless paint spray process is due to overspray. A single medium-sized shipyard will use approximately 30,000 gallons of paint per year in the building and repair of ships. Large shipyards will use ten times this amount annually. The effect of increasing transfer efficiency from 70% to 90% can be seen in Table 1. In the shipbuilding industry, the cost of overspray is conservatively estimated to approach $50 Billion. Aside from economic considerations, environmental reasons also exist to increase transfer efficiency and reduce overspray. Increasingly stringent government regulations regarding the release of toxic or hazardous materials to the environment during ship building and repair operations have increased the burden on shipyards to improve manufacturing processes. In the application of copper-ablative antifoulant coatings to the underwater hull of ships in drydock, release of overspray from the paint application process results in the discharge of copper to inter-coastal waterways, which is in violation of shipyard NPDES permits. Conventional overspray remediation techniques involve the construction of total negative-pressure containment around the structure being painted. For large surface ships in drydock, such an approach is impractical from both economic and production process perspectives as such a containment strategy prevents overhead crane access necessary to perform other repair operations during the drydock period. Assumptions: Current Transfer Efficiency: 70% Transfer Efficiency of Improved Process: 90% Data: Annual Avg. Paint Usage (Medium to Large Shipyards): 30,000 gal Gallons of paint delivered (0.7 * 30,000 gallons) 1,000 gal Paint cost: (per gallon) $30.00P Average Annual Overspray Cleanup cost (Detailing): $5,000 Paint VOC content 340 g / liter Quantitative Benefits: Gallons of paint required (@ 90% TE) to deliver 1,000 gallons 3,333 gal Estimated reduction in paint usage (@ constant production volume) 6,667 gal Estimated cost avoidance (materials) $00 K Estimated cost avoidance (car detailing) $0 K Total annual estimated cost avoidance $0 K Table 1. Economic Impact of Improving Transfer Efficiency. 3

4 EQUIPMENT & METHODS Objective Investigate effervescent atomization with the goal of increasing transfer efficiency and reducing overspray while producing an acceptable-quality spray pattern for marine coating applications. Literature Review The vast majority of research in spray atomization deals with low-viscosity fluids atomized at low flow rates and pressures compared to marine coatings. In the atomization of fuels, the goal is to produce very small, uniformly-sized droplets in order to maximize combustion efficiency and minimize pollutant production. The fluid characteristics, flow rates, atomization techniques and production issues associated with combustion applications bear little resemblance to shipyard airless paint spray application processes. A potentially useful area of research is in the atomization and spray application of agricultural chemicals, which has similar concerns with median particle size and overspray, termed agricultural drift, as the airless paint spray process. While the pernozzle flow rate in these processes are similar to that of airless paint spray, the viscosity of the agricultural chemicals are an order of magnitude lower than that of marine coatings and unlike marine coatings, behave in a Newtonian fashion. The area of research having the most similarity to the airless paint spray process is in the atomization of high-solids black liquor waste. Black liquor is a waste product produced during the manufacture of paper. The energy content of the black liquor is recovered through a combustion process in a recovery boiler. Particle size issues in the black liquor spray are similar to that of the airless paint spray process. While very small droplet sizes are desirable from a heat and mass transfer standpoint, entrainment and carryover of these small particles have a potential negative impact through fouling of boiler tubes and other undesirable effects [3]. Overspray is comprised of very small droplets generated during the atomization process which do not deposit on the surface being painted and are carried away by air currents. Through a mechanism known as entrainment, the fast-moving sheet of paint accelerates air from the surrounding atmosphere towards the target being painted. The collateral air stream cannot penetrate the surface being coated, and instead forms an energetic and stable wall-jet. Paint particles that do not have sufficient momentum (mass velocity) to penetrate and cross the wall-jet are carried away as overspray [4]. See Figure. Figure. Diagram of the Airless Paint Spray Process. Kwok and Liu [5] and Hicks and Senser [6] suggested overspray could be reduced or eliminated by controlling the size distribution of atomized droplets. The results of Hicks and Senser suggest that transfer efficiency should approach 100% if the particle fraction finer than 80 μm could be eliminated. There has been research into associated phenomena in the application of agricultural chemicals. In the application of agricultural chemicals a narrow distribution of droplet sizes is desired. Droplets should have sufficient mass to penetrate the foliage canopy, yet should not be so large that crops can be damaged by fertilizer burn. If droplet sizes are too small, agricultural chemicals can be carried great distances 4

5 EQUIPMENT & METHODS and threaten nearby crops, livestock or human habitations. Known as agricultural drift, this overspray has been attributed to particle sizes below 105 цm [7]. The airless spray process is a form of pressure atomization. In airless paint spray atomization process, paint is discharged through an elliptical orifice under high pressure to produce a thin, fastmoving liquid sheet oriented parallel to the major axis of the orifice that disintegrates into a narrow elliptical spray. This process is referred to as sheetatomization. Fraser and Eisenklam identified three primary modes of sheet disintegration, rim, wave, and perforated sheet disintegration [8]. In the rim mode, forces created by surface tension cause the free edge of a liquid sheet to contract into a thick rim that breaks up in a manner similar to that of a free jet. The resulting drops continue to move in the original flow direction but remain attached to the receding surface by thin threads that also break up into rows of small drops. This mode is prominent where the viscosity and surface tension of the liquid is high and tends to produce large drops with numerous small satellites. In the airless paint spray process this mode is evident at low atomization pressures and can be observed as tails in the under-pressurized paint sheet. In perforated sheet disintegration, holes appear in the liquid sheet, expand towards each other and coalesce to form threads and ligaments of irregular shape [9]. These ligaments break up according to the Rayleigh instability criteria [10]. Variation in the diameter of these ligaments will result in a wide variation in drop sizes. Another mechanism of sheet disintegration is caused by the generation of a wave motion on the sheet. In wavy-sheet atomization the crests of the waves created by aerodynamic interaction with the surrounding gas are torn away in patches. At very high liquid velocities corresponding to high injection pressures, sheet disintegration occurs close to the nozzle exit. The orderliness of the disintegration process and the uniformity of the production of threads and ligaments have a significant impact on drop size distribution [11]. Perforations occurring in the sheet at the same distance from the orifice have a similar history, which produces thread diameters and drop sizes of a uniform size. Wavy sheet atomization, on the other hand, is highly irregular and produces much more varied drop sizes. Atomizers that discharge liquids in the form of a sheet can exhibit all three modes of sheet disintegration and sometimes two different modes can appear simultaneously. Although several modes of sheet disintegration have been identified, in all cases the final atomization process is one in which ligaments break up into drops according to the Rayleigh criteria [1]. In pressure atomization processes, the physical properties of the fluid to be atomized have a significant impact on atomization and subsequent drop size distribution of the spray. Marine coatings are dense and highly viscous and have surface tension characteristics that enable flow and leveling of the coating. In addition, epoxy anti-corrosive coatings and antifoulant coatings exhibit strong non- Newtonian shear-thinning behavior. The room-temperature viscosity of marine highsolids epoxy paint can vary from approximately 3000 mpa-sec at low shear rates (<500 s -1 ) to a shearthinning limit of approximately 750 mpa-s at high shear rates (>10,000 s -1 ) [4]. Si-Alkyd paint used on the freeboard of the hull behaves in near-newtonian fashion and is relatively insensitive to shear rate. The room-temperature viscosity of Si-Alkyd paint at representative shear rates is on the order of 450 mpa-s. In conventional airless paint spray, the shear rate at the nozzle is on the order of 5,000 s -1. As described above, in the airless paint spray process paint is forced at high pressure through a small elliptical orifice. The paint is formed into a thin sheet moving at high-velocity. Through a mechanism known as entrainment, the fast-moving sheet of paint accelerates air from the surrounding atmosphere towards the target being painted. The collateral air stream cannot penetrate the surface being coated, and instead forms an energetic and stable wall-jet. 5

6 EQUIPMENT & METHODS Paint particles that do not have sufficient momentum to penetrate and cross the wall-jet are carried away as overspray [4]. Like the related phenomenon of agricultural drift, the particles that comprise paint overspray will travel great distances before settling out of the air. In the airless paint spray process it is the combination of the production of too-small particles in combination with air entrainment and subsequent formation of a wall-jet that contributes to poor transfer efficiency. The primary parameters affecting whether a particle will reach the intended target is the mass of the particle, and the strength and extent of the air currents the particle must contend with in order to reach the target. Effervescent Atomization In effervescent atomization, air or gas is injected into the fluid at some point upstream of the exit orifice. Upon exiting the nozzle the gas bubbles contained in the liquid stream expand rapidly (explode) shattering the fluid into small droplets. The separate injection of gas allows the number, size, and spatial distribution of bubbles to be controlled in effervescent designs. As the gas-to-liquid ratio (GLR) is increased across the operating range, three regimes of atomizer operation can be identified: bubbly-flow regime, transition regime and annular flow regime. The bubblyflow regime is characterized by a well-dispersed mixture of liquid and fine gas bubbles. As the GLR is increased, the bubbles begin to coalesce and the atomizer begins to exhibit instabilities. This is the transition regime between the stable bubbly-flow regime and a stable annular flow regime. There may be several commonly observed two-phase flow patterns in the transition regime, including slug flow, plug flow, and churn flow. The annular flow regime occurs at high GLR and is characterized by large streams of gas surrounded by liquid jets [13]. Loebker and Empie [14] studied the effect of GLR on droplet diameter on model liquids spanning a viscosity range from 100 mpa s to 10,000 mpa s. At the low end of this viscosity range they found a hump in the MMD response in the very low GLR range, see Figure 3. Figure 3. Drop Size MMD as a function of GLR and Liquid Flow rate at 100 mpa s viscosity. Empie and Loebker proposed the following explanation for the increase in MMD: In conventional liquid-only spraying strands are formed during liquid sheet disintegration. With a small amount of gas added, the liquid sheet disappears and is replaced by somewhat uniformly distributed strands exiting directly from the nozzle orifice oriented perpendicular to the direction of the spray flow. The increase in drop size at low GLR may be due to the formation of thicker strands compared to those produced as a result of liquid sheet disintegration. As GLR increases strand diameter and spacing decrease, subsequently producing smaller drops. At higher GLR, liquid strand disintegration occurs closer to the nozzle and the strands become oriented in a direction parallel to the spray flow. Increasing viscosity results in the formation of thicker strands that require a higher gas flow rates to disintegrate into drops. Another interesting result of the work performed by Loebker and Empie was that the droplet size distribution did not increase with the increase of MMD at low GLR, and in some cases actually narrowed. Santangelo and Sojka investigated the near-nozzle structure of an effervescent atomizer-produced spray using focused-image holography. They found liquid breakup in the bubbly-flow regime to be governed by individual bubble expansion. The liquid exits the nozzle as a solid jet or trunk containing the train of bubbles. Shortly after exiting the nozzle, the trunk breaks up rather abruptly into ligaments and droplets 6

7 EQUIPMENT & METHODS via single bubble expansions. In the transition regime the trunk becomes distorted, wavy, and begins to shred. The shredding trunk breaks up into large limb-like structures which subsequently break up into smaller branches, ligaments and drops. In the annular flow regime the trunk is greatly reduced in length extending only a fraction of a millimeter from the nozzle exit. The limbs are wavy and vary in diameter along their length. These limbs branch off into smaller and smaller ligaments and eventually form drops [15]. This process is shown graphically in Figure 4. Figure 4. Bubbly-Flow Regime Near-Nozzle Structure. Roesler and Lefebvre studied the effervescent atomization of low-viscosity fluids in the bubbly-flow regime and found Sauter Mean Diameter (SMD) to have a strong dependence on GLR and fluid injection pressure while being nearly independent of mass flow-rate and nozzle orifice diameter [16]. Buckner and Sojka investigated effervescent atomization in the annular flow regime and concluded that mean droplet diameter was a strong function of GLR while being nearly independent of liquid viscosity, fluid delivery pressure and total mass flow rate [17]. Summary of Effervescent Spray Technology Droplet mass mean diameter (MMD) is relatively insensitive to viscosity and orifice geometry in effervescent atomization. There are distinctly different mechanisms of liquid jet disintegration in the bubblyflow and annular flow regime which allow for the possibility of independent drop size control using GLR (and to a lesser extent, fluid injection pressure) over a wide range of viscosity and mass flow rates. The advantages offered by effervescent atomization are the following [1]: Atomization is very good even at very low injection pressures and low gas flow rates. Mean drop sizes are comparable to those obtained with air-assist atomizers for the same gas/liquid ratio. The system has large holes and passages so problems of plugging are greatly reduced or eliminated. The basic simplicity of the device lends itself to good reliability, easy maintenance, and low cost. Effervescent atomization is effective over an extremely broad range of flow rates on the order of 0.05 LPM to over 50 LPM. Effervescent atomization can be effectively used to atomize Newtonian and non-newtonian liquids spanning a viscosity range from 1 to 10,000 mpa s [14, 17-19]. Atomization quality is nearly independent of orifice geometry. Technical Approach An experimental design was created to study those parameters that can be most easily controlled to produce a consistent, high-quality spray in a production environment. Due to the fillers used in paint, shear and extensional viscosity may vary substantially from one type of paint to another, from one color of paint to another of the same type, and even from one batch of paint to another. Different paints with the same terminal viscosity might not atomize the same way because of the effect that the fillers used will have on extensional viscosity. Numerous methods exist to measure shear 7

8 EQUIPMENT & METHODS viscosity, but no such methods exist to measure extensional viscosity. The difficulty associated with measurement of extensional viscosity and the inherent variability of paint makes viscosity a poor control variable. From a production standpoint, paint viscosity may be controlled by controlling paint temperature. On-the-fly adjustments of paint temperature can be made to achieve good atomization and pattern control. For these reasons, and to reduce the size of the experiment, paint viscosity was not included as a control variable and paint from a single batch was used to minimize batchto-batch variations. Limitations on the pressure capacity of the gas flow controller dictated that the paint pressure be kept below 1000 psi. Given these limitations on paint pressure, two spray tip / pressure combinations were chosen to achieve the target production rate of approximately 0.35 gpm. To minimize pressure pulse fluctuations caused by standard industrial paint pumps, a long-stroke paint pump capable of delivering 0.35 gpm for 30 seconds on a single stroke was designed and fabricated by AST Corp. of New Hope, MN. A modified Graco plural-component mixer manifold was used to inject Nitrogen into the paint. The number and size of injector tube holes was varied over a wide range to ascertain if these parameters had a significant effect on atomization and pattern quality. Based on previous work of Loebker and Empie, gasto-liquid mass ratios of to were studied. Deg C Deg F Brookfield (CPS units) Brookfield Visc (Test Method 07) Spindle #4 & RPM 0 (all measurements) Table. Viscosity / Temperature profile of INTERLAC 537 Soft White (5347). central block of the mixing manifold has been modified to allow nitrogen to flow into the mixing chamber through a central passage, while the paint flows into the chamber from the side. The exit of the nitrogen passage is threaded to allow insertion of a variety of gas injection tubes. Variations in the number of holes and total area of the gas injection tube are accomplished by insertion of the appropriate tube into the plural component mixer. The tip of one of the gas injection tubes may be seen protruding from the plural component mixer material outlet in Figure 5. Experimental Procedure The paint chosen for this evaluation was International Coatings INTERLAC 537 Soft White (5347), Batch# CL318UHC5, with specific gravity of All experiments were performed at room temperature. The temperature / viscosity profile is provided in Table. The paint and nitrogen were mixed within a modified Graco Plural Component Mixer Manifold, p/n: In this arrangement, paint is provided to the resin side of the mix manifold, while nitrogen is provided to the hardener side of the mixer manifold. The Figure 5. Modified Graco plural component mixer with N injection tube visible at the material outlet port. 8

9 Development of a High Transfer Efficiency Painting Technology EQUIPMENT & METHODS The paint/nitrogen mixture exits the mixer manifold at the material outlet, which is directly connected to a standard Graco Plus spray gun. Two shutoff valves were utilized prevent backflow of paint or solvent into the mass flow controller. The entire system is shown in Figure 8. Two of the control parameters (number of holes in the injection tube and total injection tube area) were controlled by using interchangeable injection tube inserts mounted within the mixing chamber of the mixer manifold. Some of these gas injection tubes are shown in Figure 6. Figure 8. Effervescent atomization device with mass flow controller and spray gun. Two spray tips (p/n: 4001 and 40015) manufactured by Spraying Systems Co. of Wheaton, Illinois were chosen to deliver approximately 0.35 gpm at 400 and 900 psi respectively. Figure 6. Gas injection tube variations with spiral hole pattern arrangement. Flow Rate Calculations A Porter Model 01 gas flow mass controller, p/n: 01-DKASVCAA (s/n: ) and Model CM control module (s/n: 6306) were used to meter nitrogen from a standard nitrogen tank into the paint line. (Figure 7) The Model 01 gas flow controller was mounted directly to the plural component nitrogen inlet to minimize the distance between the mass flow controller and the injection tube. Flow through a nozzle is affected by pressure and by variations in the specific gravity and viscosity of the fluid. Typically an increase in viscosity or specific gravity (or both) will reduce the flow rate of a nozzle. Conversely an increase in the pressure difference across the orifice increases the flow rate of the nozzle. The liquid being sprayed in this study is a latex-based paint and can exhibit non-newtonian shear characteristics such that the material becomes less viscous at high shear rates typical of these and other paint spraying operations. Because of this shear-thinning behavior, the viscosity was assumed to have a small to minimal effect on the overall nozzle flow rate. The specific gravity, however, is quite large as compared to water with a value of SGPaint = The pressures are also considered to be large relative to the way the nozzles used in this study are rated. A definition of the and 4001 nozzles provides that at 40 psi across the orifice the nozzle will have a 40º spray angle and will have flow rate of 0.10 gpm and 0.15 gpm using water. In this study the sprays were to be operated in the range of 400 psi to 900 psi. Figure 7. Porter Model 01 gas flow device. 9

10 EQUIPMENT & METHODS Specific Gravity Specific gravity (SG) is the ratio of the mass of a given volume of liquid to the mass of the same volume of water. In spraying, the main effect of the specific gravity of a liquid (other than water) is on the capacity of the spray nozzle. Since the spray angle and flow values of the nozzles used in this study are based on spraying water, a conversion factor can be applied to determine the nozzle capacity when using a liquid other than water. The following formula was used to estimate the change in flow rate due to specific gravity: Q Liq = Q H 0 (SG) -0.5 where Q volumetric flowrate This conversion factor [(SG) -0.5 ] accounts only for the effect of specific gravity on capacity and does not account for other factors affecting capacity. Pressure Effect The effect of pressure on flow rate is that as the pressure differential across a nozzle orifice increases the volumetric flow rate through that orifice also increases. This can also have the effect of increasing the shear rate on the fluid. This means that increasing pressure will have a greater effect on Thixotropic (shear-thickening) liquids than shear-thinning or Newtonian liquids. A simple derivation using the continuity equation and Bernoulli s equation can be made to estimate the change in flow rate with respect to pressure. The flow is assumed to be incompressible, inviscid, and irrotational. Q = VA Where V- Velocity A- Orifice area 1 1 p 1 + ρv1 = p + ρv = C Where p - Pressure ρ - Density V - Velocity For small nozzles, the assumption that the internal velocity is much, much smaller than the external velocity is reasonable. External pressure is atmospheric and assumed to be negligible compared to the pressures upstream of the nozzle exit. These assumptions are shown as: V << V and P 0 This leads to: p = ρv Q = V Q = A A p1 ρ Q A V = = πd 4 p1 ρ p1 ρ p1 ρ With incompressible flow Q1 = Q. Using this relation and dividing we obtain a pressure flow ratio as shown in the following: Q 1 P = 1 Q P Using this equation and the known flow rates of the nozzles selected the approximate flow rates for these nozzles may be calculated at the working pressures of 400 psi and 900 psi including compensation for the specific gravity of the paint (SGPaint = 1.79.) Nozzle Pressure (psi) Flow (gpm) SG Comp. Flow (gpm) Table 3. Nozzle Flow as Function of Pressure with Specific Gravity Compensation. Typical operating pressures for this paint in a shipyard production situation will depend upon the paint vendor and the ambient temperature. A tip is normally used at pressures ranging from psi. 10

11 EQUIPMENT & METHODS Paint was supplied to the atomizer using a large hydraulic cylinder capable of delivering 0.35 gpm for 30 seconds on a single stroke. This pumping arrangement eliminates pressure fluctuations common with standard industrial paint pumps. A linear displacement meter was attached to the hydraulic piston to allow direct measurement of piston displacement, and hence, total paint volume sprayed during each run. F(D) = 1- exp - D X The X and N parameters for this equation were generated from the actual raw data. A photograph of the instrument with a typical test setup is shown in Figure 9. N During each trial, paint was sprayed onto thick by 36 wide by 6 long pre-weighed aluminum foil sheets. The foil was obtained from All Foils, Inc. of Brooklyn Heights, Ohio. The foil sheets were pre-weighed using an OHAUS 3100 x 0.01 gram Adventurer scale with the weighbelow option. The weigh-below option allows parts to be suspended from the underside of the scale. The scale was mounted atop a 3-sided plywood cabinet measuring 8 tall x 4 wide x deep. Following each trial, the foil sheets were immediately weighed and laid flat to dry. Transfer efficiency was calculated directly from the pre-spray foil weight and post-spray foil weight. After applying paint to the foil sheet, the particle size distribution of the spray was characterized using a Malvern 600 particle analyzer. The Malvern analyzer is a laser diffraction instrument that measures drop size based on the energy of the diffracted light caused by drops passing through the analyzer s sampling area. The scattered light intensity is measured using a series of semicircular photo diodes housed in the receiver unit. The Rosin-Rammler distribution function is used to convert the light intensity distribution into a drop size distribution function. Rosin-Rammler Parameters The Rosin-Rammler distribution function is a representation of the drop population and size in a spray. The exact size for every volume fraction F(D) in the spray can be calculated using the X and N parameters. Figure 9. Malvern 600. The D V0.5, D V0.1, and D V0.9 diameters were used to evaluate the drop size data. The drop size terminology is as follows: D V0.5: Volume Median Diameter (also known as VMD or MVD). A means of expressing drop size in terms of the volume of liquid sprayed. The VMD is a value where 50% of the total volume (or mass) of liquid sprayed is made up of drops with diameters larger than the median value and 50% smaller than the median value. This diameter is used to compare the change in (or mass) of liquid sprayed is made up of drops with diameters smaller or equal to this value. D V0.1: Is a value where 10 percent of the total volume (or mass) of liquid sprayed is made up of drops with diameters smaller or equal to this value. D V0.9: Is a value where 90 percent of the total volume (or mass) of liquid sprayed is made up of drops with diameters smaller or equal to this value. 11

12 EQUIPMENT & METHODS DX6 Run GLR / Area / 1 Hole / 900 psi Percent in Dv µm Dv Dv D (3,) Percent under Particle size (µm) Figure 10. Drop Size Distribution The particle size distribution was measured three times for each set of conditions. D V0.5, D V0.1, and D V0.9 and Sauter Mean Diameter were calculated for each measurement and averaged for each trial. A typical drop-size distribution is shown in Figure 10. Once the foils sheets were completely dried, an approximately 4 wide section was cut from the middle section of each foil sheet. The total thickness of the paint and foil was measured at 1 cm intervals across the entire width of the spray pattern, at each end of the 4 section of foil. The pattern trace for Run 1 is shown in Figure 11. closely approximate an inverse quadratic polynomial. The shape factor is a positive number between 0 and 1, with the better quality patterns having a shape factor approaching 1. The distance from the pattern center (cm) and whether or not the trace was measured at the top or bottom end of the foil sheet were the two parameters included in the model. The use of these two variables provides information on both the symmetry of the spray about the mid-line as well as the stability of the spray from one point in time to the next. DX Run GLR / Area / 1 Hole / 900 psi 0.0 From the pattern traces, the volume of paint contained in the tails was estimated by calculating the area under the tails, and dividing this quantity by the total area under the paint-trace curve to arrive at an approximate percentage of the volume of paint contained in the tails. The paint traces were also used to calculate a pattern shape-factor to assess the quality of the spray distribution. The best quality patterns Total Thickness ( 0.001" ) Distance from Pattern Center (cm) Figure 11. Pattern Trace, Run 1. Top Bottom 1

13 EQUIPMENT & METHODS Experimental Design Four factors were studied: Number of Holes, Gas-to- Liquid Ratio (GLR), Total Area of Holes in the Injection Tube, and Fluid Pressure. A quadratic D-Optimal response surface model was chosen to resolve nd order and interaction effects. Gas injection tubes having 1, 5 and 9 holes were fabricated. The holes were arranged in a spiral pattern to maximize the probability of achieving a welldispersed bubbly mix. Total area of the injection tube holes was controlled at the levels of in, in and in. GLR was varied between the levels of (0.01% by weight) and (0.1% by weight). Two pressure levels were chosen, 400 psi and 900 psi; and two nozzles were chosen to deliver approximately 0.35 gpm as described earlier in this document. The D-Optimal is a highly efficient design. A full-factorial nd order experiment with three (3) continuous and one (1) categorical variable would require a minimum of 54 unique trials. The quadratic D-optimal design chosen required a total of 4 trials, including pure replicates. The quadratic D-optimal design is shown in Table 4. The results are shown in Table 5. Standard Order Run Order Number of Holes Target GLR (%) Actual GLR (%) Area of Holes (0.001 in) Pressure (psi) / Nozzle / / / / / / / / / / / / / / / / / / / / / / / / 4001 Table 4. Experimental Design. 13

14 RESULTS & DISCUSSION Results Transfer Efficiency was calculated assuming a specific gravity of No correction for paint compressibility has been made, even though the paint is estimated to compress approximately 5% per 1000 psi. Drop size data was measured directly using the Malvern Analyzer. Tail volume was estimated using a relatively coarse integration technique and measuring the area under the tails and dividing by the total area under the curve in order to arrive at an approximate percentage of the volume of paint comprising the tails. The pattern shape factor was calculated by fitting the paint trace data to an inverse quadratic polynomial and calculating the Adjusted R value. The Adjusted R value represents the extent to which the data will fit the polynomial. An Adjusted R value of 1 represents a perfect fit to the polynomial with all data points falling on the curve. A low Adjusted R value represents a poor fit to the polynomial, with the data falling in a random fashion. This shape factor will take into account the presence of tails, but will also take into consideration asymmetry of the spray about the mid-line, smoothness of the curve, and flatness of the curve near the center of the pattern. The experimental results appear in Table 5. Each of the response variables is discussed individually in the following sections. Run Order Transfer Efficiency (%) Tail Volume (%) Shape Factor Dv 0.5 (µm) Dv 0.1 (µm)ee Dv 0.9 (µm) Sauter (µm) e 4.75e 0.77e e e e Table 5. Experimental Results. 14

15 RESULTS & DISCUSSION Transfer Efficiency When the transfer efficiency data is analyzed according to a linear model, the pressure and total area of the gas injection tube holes were found to be significant factors affecting the transfer efficiency. These factors explained approximately 30% of the variation observed. The Analysis of Variance (ANOVA) chart for this response value is shown in Table 6. The model is shown graphically in Figure 1. The Model F-value of 5.86 implies the model is significant. There is less than a 0.95% chance that a Model F-Value this large could occur due to noise. The Adjusted R value is in reasonable agreement with the Predicted R value, indicating robustness of the model. The transfer efficiency was slightly higher for the 900 psi / nozzle configuration than it was for the 400 psi / 4001 nozzle. This result is likely due to compression of the paint. Assuming that the paint is compressible by approximately 5% per 1000 psi, the transfer efficiency will be overestimated by approximately % at 400 psi and 4.5% at 900 psi. The difference of.5% is consistent the observed difference in transfer efficiency between these levels. Transfer Efficiency (%) Transfer Efficiency - Linear Model D: Pressure 400 psi 900 psi Total Hole Area (in ) Figure 1. Transfer Efficiency as a function of Pressure and Area of Injection Tube Holes. Source Sum of Mean DF Squares Square F Value Prob > F Comment Model Significant Area Pressure Residual Cor Total Std. Dev R-Squared Mean Adj R-Squared C.V. 1.7 Pred R-Squared PRESS Adeq Precision Table 6. ANOVA Results for Transfer Efficiency Linear Model. 15

16 RESULTS & DISCUSSION When higher order terms are included in the analysis, the number of holes was also found to be a significant factor affecting transfer efficiency. Inclusion of the nd order term (Holes) increases the amount of variation explained by the model to ~36%. The ANOVA results for the quadratic model are given in Table 7. The results are displayed graphically in Figure 13. The F-Value of 3.98 indicates that the model is significant. There is only a 1.85% chance that a Model F Value this large could occur due to chance. The Predicted R value of was not in as close agreement to the Adjusted R value of as would be desired. This is an indication that there may have been a problem with the model or data (outliers). In this case, the model was not as robust as it could have been. Due to limitations in the available gas injection tubes available at the time of the test, there were two modifications to the original experimental design. Run 1 should have been run with an injection tube having 7 holes as opposed to the 5-hole tube used and Run 16 was to have used a gas injection tube having a total gas injection hole area of in instead of the in tube used. It is likely that the discrepancy between Predicted R and Adjusted R is due to the liberties taken with the experimental design. Transfer Efficiency (%) Transfer Efficiency 400 psi 900 psi Number of Holes Figure 13. Quadratic Model of Transfer Efficiency as a Function of Number of Injection Tube Holes at Total Hole Area of in. Source Sum of Mean DF Squares Square F-Value Prob > F Comment Model Significant Holes 1.506E E E Area Pressure Holes Residual Cor Total Std. Dev R-Squared Mean Adj R-Squared C.V Pred R-Squared PRESS 7.79 Adeq Precision Table 7. ANOVA Results for Transfer Efficiency Quadratic Model. 16

17 RESULTS & DISCUSSION Sources of Error: Transfer Efficiency For engineering purposes, these models explain a relatively small portion of the variation observed. This is due in large part to the difficulties associated with accurately measuring the transfer efficiency on a trial-by-trial basis. Factors that will affect measurement of transfer efficiency are listed below. Factors affecting the measured values of transfer efficiency: Accuracy of linear displacement counter (± cc) Volatile evaporation between end of trial and measurement of foil weight. This will be affected by time delays between the end of spraying and weight measurement Compressibility of paint and non-linearity of compression / pressure curve Expansion of paint lines The volume of paint sprayed during each trial was measured using a linear displacement meter. The accuracy of measurement was ± cc. With specific gravity of 1.79, the accuracy of the volume measurement will result in an error of ± 1.% in the measured transfer efficiency. In addition to the uncertainty in measurement of the piston displacement, paint compressibility and expansion of the paint lines will also artificially inflate the volume of paint used. This effect will be greater for the 900 psi trials. The total contribution of these factors to the transfer efficiency is conservatively estimated to be less than 5%. D V0.1 GLR, Pressure and the GLR* Pressure interaction term were significant factors affecting D V0.1. These factors explained ~94% of the variation observed in the data. The ANOVA results for D V0.1 are shown in Table 8. The model is shown graphically in Figure 14. The Model F-value of 119. implies the model is significant. There is less than a 0.01% chance that a Model F-Value this large could occur due to noise. The Adjusted R value is in good agreement with the Predicted R value. The interaction effect between pressure and GLR is shown by the differing slopes of the lines for the 400 psi / nozzle configuration and the 900 psi / 4001 nozzle configuration. GLR has little effect on D V0.1. for the 900 psi / nozzle configuration, but GLR has a substantial effect on D V0.1 for the 400 psi / nozzle configuration. For the 400 psi configuration, D V0.1 becomes smaller with increasing GLR. Dv 0.1 (um) Interaction Graph GLR (%) 400 psi 900 psi Figure 14. Graphic representation of GLR and Pressure Effect on DV0.1. Solid round dots represent design points at which data was collected. 17

18 RESULTS & DISCUSSION Source Sum of Mean DF Squares Square F-Value Prob > F Comment Model < Significant GLR Pressure < GLR * Pressure Residual Cor Total Std. Dev..0 R-Squared Mean Adj R-Squared C.V. 3.7 Pred R-Squared PRESS Adeq Precision.193 Table 8. ANOVA Results for DV0.1. D V0.5 Pressure and the Pressure*Area interaction term were significant factors affecting D V0.5. These factors explained ~94% of the variation observed in the data. The ANOVA results are shown in Table 9. The model is shown graphically in Figure 15. The Model F-value of implies the model is significant. There is less than a 0.01% chance that a Model F-Value this large could occur due to noise. The Adjusted R value of is in good agreement with the Predicted R value of The separation of the 900 psi and 400 psi lines in the D V0.5 model shows that pressure has the most substantial effect on D V0.1. The interaction effect between pressure and injector hole area is shown by the different slope of the lines 400 psi (upper) and 900 psi (lower) lines. The pressure effect behaves as would be expected, with increasing pressure resulting in smaller droplet size. Dv 0.5 (um) psi 900 psi Dv Gas Injection Area (in ) Figure 15. Graphic Representation of DV0.5 as a Function of Injection Tube Area and Pressure. Source Sum of DF Mean F-Value Prob > F Squares Square Model < significant Area Pressure < Area * Pressure Residual Cor Total Std. Dev. 3.7 R-Squared Mean Adj R-Squared C.V Pred R-Squared PRESS Adeq Precision 0.16 Table 9. ANOVA Results for DV

19 RESULTS & DISCUSSION D V0.9 Pressure and the Pressure*Area interaction term were significant factors affecting D V0.9. The model explained ~89% of the variation observed in the data. The ANOVA results are shown in Table 10. The model is shown graphically in Figure 16. The Model F-value of implies the model is significant. There is only a 0.01% chance that a Model F-Value this large could occur due to noise. The Pred R-Squared of is in good agreement with the Adj R-Squared of D V0.9 shows a similar trend to that of the D V0.5 model. The separation of the 900 psi and 400 psi lines shows that pressure has the most substantial effect on D V0.9, while the Pressure * Area interaction effect is shown by the differing slope of the lines at 400 psi (upper) and 900 psi (lower). Dv 0.9 (um) psi 900 psi Dv Area (in ) Figure 16. Graphic Representation of DV0.5 as a Function of Injection Tube Area and Pressure. Source Sum of Squares DF Mean Square F Value Prob > F Comment Model < significant Area E Pressure < Area * Pressure Residual Cor Total Std. Dev R-Squared Mean Adj R-Squared C.V Pred R-Squared PRESS Adeq Precision Table 10. ANOVA Results for DV

20 RESULTS & DISCUSSION Source Sum of Squares DF Mean Square F Value Prob > F Comment Model < Significant GLR Pressure < GLR* Pressure Residual Cor Total Std. Dev. 3.8 R-Squared Mean Adj R-Squared C.V Pred R-Squared PRESS 9.90 Adeq Precision Table 11. ANOVA Results for Sauter Mean Diameter. Sauter Mean Diameter GLR, Pressure and the GLR*Pressure interaction term were significant factors affecting Sauter Mean Diameter. These factors explained ~94% of the variation observed in the data. The ANOVA results for Sauter Mean Diameter are shown in Table 11. The model is shown graphically in Figure 17. The Model F-value of implies the model is significant. There is less than a 0.01% chance that a Model F-Value this large could occur due to noise. The Pred R-Squared of is in good agreement with the Adj R-Squared of Sauter Mean Diameter behaves similarly to D V0.1. The interaction effect between pressure and GLR on Sauter Mean Diameter is shown by the differing slopes of the lines for the 400 psi / nozzle configuration and the 900 psi / 4001 nozzle configuration. As with the D V0.1 model, GLR has little effect on D V0.1 for the 900 psi / nozzle configuration, but GLR has a substantial effect on D V0.1 for the 400 psi / nozzle configuration. Sauter Mean Diameter (um) Sauter Mean Diameter 400 psi 900 psi GLR (%) Design Points Figure 17. Effect of GLR and Pressure on Sauter Mean Diameter. Pattern Tail Volume GLR, Pressure and the GLR*Pressure interaction term were significant factors affecting the size of the tails. These factors explained ~94% of the variation observed in the data. The ANOVA results for the Pattern Tail Volume are shown in Table 13. The model is shown graphically in Figure 18. The Model F-value of 7.11 implies the model is significant. There is less than a 0.01% chance that a Model F-Value this large could occur due to noise. The Pred R-Squared of is in reasonable agreement with the Adj R-Squared of As expected, the tail volume for the 900 psi / 4001 nozzle condition is consistently and substantially 0