Department of Chemical Engineering 1, Lund Institute of Technology, Lund, Sweden

Transcription

1 Atomization by Nozzles an Empirical Study Simon Smrtnik Department of Chemical Engineering 1, Lund Institute of Technology, Lund, Sweden Abstract The formation of droplets by nozzles is studied with the Newtonian fluids water and glycerin. The survey is carried out by use of laser diffraction (Malvern Series 2) where the algorithm is based on Fraunhofer diffraction theory. The influence of the parameters viscosity, temperature, air flow (mass) or pressure, liquid flow (mass), the distance from the orifice along the spray axis and six two-fluid nozzles with external mixture are investigated. It is proved that the viscosity has a significant effect on the droplet formation. With higher air flows (pressure) the effect of viscosity diminishes. An increased quotient between air and liquid flow (mass) will give smaller droplets as an increased liquid flow for a specific quotient gives, i.e. keeping the ratio between air and liquid flow constant. The spray consists of a thin thread with a diameter about 1 mm after the first -8 mm from the orifice, which consists of ligaments to be atomized to droplets forming a spray. It s characterized by low positively skewness with low span, because the droplet formation is not yet complete. Further away from the orifice the smaller droplets are found and the wider and more skewed distribution will be found in the cross section to a state around 3- cm after the orifice where the system is stabilized. The distributions from the nozzles have found to be Rosin-Rammler distributed. Key words: nozzle; spray; distribution; ligament; droplet; quotient Introduction Nozzles are used in a variety of fields. They are used for dosage of a specific amount of liquid under a certain time in a combustion engine, to create a specific spray pattern and to distribute the liquid in a special way, to give droplets a specific momentum and thereby creating a specific impact. [1] In the pharmaceutical industry applications such as spray drying and coating are common where the spray process is significant. The process where a volume of liquid is transformed to droplets is called atomization, where the specific surface of the liquid is dramatically increased. Atomization is initiated at the orifice of the nozzle and is principally determined by the air flow (pressure), the viscosity and velocity of the liquid and by the properties of the nozzle design. [2] Materials Fluids used in this work are the Newtonian fluids water and glycerin. Glycerin, C 3H 8O 3, is a clear, colourless and viscous liquid whose viscosity is strongly temperature dependent. Glycerin is also of a strongly hygroscopic nature. Experimental Equipment and analysis The experiments are carried out through measurement of the droplet distribution in sprays with laser diffraction (Malvern Series 2). A 2 mw collimated He-Ne laser is used with the wavelength 32.8 nm and the evaluation method is based on Fraunhofer diffraction theory. A flow meter, Alicat Scientific 1 Series, was used to determine the air flow and pressure. The liquid flow was verified through weighing a beaker and noting the time for a specific weight decrease. The beaker was held on a temperate plate to keep a certain temperature concerning glycerine, because of the strong temperature dependency of the viscosity. Bilaga XIV:1()

2 Analysis of the results was done by using statistical methods mainly by the programme Modde version. and by using ANOVA (Analysis of Variance). Nozzles Two spray nozzles were used during the experiments; 1/8 JJAU 1 and one from the spray dryer Büchi Mini Spray Dryer B-191. The 1/8 JJAU nozzle was used together with two needles; diameters.7 mm and 1.3 mm, which close the orifice when necessary and five air nozzles. The thin needle was used with the round little nozzle a in Figure 1. The other nozzles in Figure 1. were used with the coarser needle. Altogether, with the nozzle spray dryer, six nozzles were used. When nothing else is said the nozzle thin (see Figure 1.) is evaluated ahead. a c d Figure 1. Nozzles 2 a Round little and round big with conical spray pattern with designation J-7-SS respectively J-12- SS which are called thin respectively coarse ahead. b Flat spray pattern with designation J and is called flat ahead. c Round wide spray pattern with the designation J and is called wide ahead. d Round conical with the designation J-1111 and is called cone ahead. Parameters and responses The parameters which mainly are investigated are the distance from the axis along the spray axis, liquid flow, viscosity, air flow (pressure) which is measured in Nl/min (bar), temperature and finally six nozzles. The liquid flow was studied in three areas; - g/min, 1-2 g/min and 3- g/min. The viscosity values were; 1, and 1 cp which correspond to water (2-22 C) and glycerin (8 and 8 % at 2 C). The pressure was varied between 1 to 2 bar (overpressure) The temperature was varied between the two b values 2 and C both for water and glycerin (8 %). The responses mainly used in the work are De Broucker mean, D[,3], the distribution measure span and the skewness measure IQCS [] (interquartile coefficient of skewness). Results and discussion Liquid flow, viscosity and temperature Calculations with the programme Modde version. gives information about the significance about the parameters studied. Regarding liquid flows around - g/min the viscosity and the air flow have the greatest impact on the droplet size. The correlation between viscosity and droplet size is positive and between air flow negative. With increased liquid flow the parameter becomes significant. The viscosity has a significant effect on the droplet distribution and therefore also the temperature, because the two factors are directly related to each other. When heating the air flow the nozzle will, after a while, hold the same temperature and the nozzle will therefore act as a heat exchanger. Heat will be transferred to the colder stream, the liquid, from the warmer, air flow. Heating a liquid decreases its viscosity and consequently decreases the mean diameter, D[,3]. For the viscosity 1 cp the change in temperature from 2 C to C gives a change in D[,3] from approximately 2 to 1 µm (for 1 bar overpressure). Quotient An increase of the quotient between air and liquid flow decreases the mean droplet size as shown in Figure 2. The box in the figure gives the values for liquid flows in g/min quotient between air- and liquidflow [-] Figure 2. D[,3] plotted against the quotient for varied series of liquid flow; viscosity 1 cp Spraying Systems Co. 2 Ibid. Bilaga XIV:2()

3 The bigger the liquid flow the smaller the droplets will be for every quotient in Figure 2. The differences for every quotient, concerning the mean droplet diameter, are statistically significant on the level of %. A possible explanation to the trend is as follows. Atomization of a liquid bulk consumes a certain amount of energy supplied which is very little. According to Masters [] just a few hundreds of a percentage of supplied energy is consumed. Thus most energy is wasted. Increasing the liquid flow the air flow must be compensated by the same amount to keep the ratio for a specific quotient. This means that the total energy supplied to the nozzle is increased and thus the absolute energy uptake is higher. It s possible that the system has a threshold energy which must be overcome to initiate atomization of the liquid bulk, which also is verified by the need of a minimum pressure of about approximately. bar in forming a homogenous spray. With increased air flow the pressure is increased for a specific nozzle and the specific surface increases and thus smaller droplets are formed. The system however seems to have an inherent synergy effect between increased air flow and liquid flow, because an increase in energy uptake proportional to increased liquid flow gives a higher grade of atomization. This is true for a limited area. For bigger quotients the mean diameter approaches an asymptotical value with further increase of the quotient. The trend can clearly be seen in Figure 2. The spread (standard deviation) of the values on D[,3] is calculated for each quotient which diminishes with the quotient (see Figure 3.) The figure shows two series: one which include the liquid flow g/min and one which has excluded it for the two highest quotients. The series where g/min is excluded shows a clearly decreasing spread (standard deviation). The pressure is. respectively.8 bar for the quotients and 7. The preceding value is the threshold for the atomization and the later is low compared with the others where the pressure is higher keeping a specific quotient, which explains the deviating trend for g/min in the series. Thus the pressure is the most important parameter controlling the atomization not the air flow standard deviation [-] quotient between air- and liquidflow [-] inc. exc. Figure 3. Standard deviation plotted against the quotient. The span increases among the series with the liquid flow and quotient. This can be explained by the fact that smaller droplets are formed with increased pressure. The big droplets will also become smaller which cause a long tail forming from the most frequent value in the distribution to the bigger droplets. The phenomena can be compared with milling, where often bimodal distributions are formed. As the milling proceeds and more energy is supplied to the system the most frequent value is decreased for particles in the bigger areas of the distribution and a tail is being formed. The distributions are positively skewed. IQCS is about.2 with low liquid flows and quotients and rises to about. for high liquid flows and quotients. The differential distributions for viscosities 1 and 1 cp are shown for different pressures in Figure, which follow the Rosin- Rammler distribution []. The box in the figure gives the viscosity values ( vi ) and the pressure ( P ). Frequency [%] Volume diameter [µm] vi=1; P=,8 vi=1, P=1,8 vi=1; P=,8 vi=1; P=1,7 Figure. Differential distributions for liquid flow g/min. The least skewed (IQCS:.19) distribution is valid for viscosity 1 cp and pressure.8 bar. However the distribution for 1 cp and 1.7 bar gives the highest skewness (IQCS:.). Increased pressure thus gives Bilaga XIV:3()

4 increased skewness but the volume diametercorresponding to the mode value decreases. The mean diameter, D[,3], increases with the viscosity. The differential distribution does not change evidently with the liquid flow. The skewness, IQCS, increases from.19 to.2 when changing the liquid flow from to 12 g/min. The mean diameter, D[,3], (22. µm) changes hardly as the span from 1.3 to 1.9 (keeping the pressure,.8 bar, and the viscosity, 1 cp, constant). Droplet formation Initiation of the atomizationprocess requires high air flows to create high friction forces over an instable liquid surface as the air flow hits the liquid bulk, which is torn apart to small ligaments and big droplets. In a second phase the big droplets become smaller. [] In Figure. the pattern, stated above, can be seen. First big droplets are formed in the near of the orifice which decrease fast to smaller droplets and are stabilized around cm and the size of the droplets are thereafter not significantly changed. For longer distances the droplets will coalesce forming bigger droplets [], because the energy, distributed on a bigger cross section, is not sufficient to both keep them apart and driving them forward distance to the orifice [cm] Figure. D[,3] along the spray axis; with nozzle coarse and viscosity cp. The droplet distributions are positively skewed along the spray axis, where the skewness, IQCS, is.33 at the orifice and increases up to 2 cm (.). The atomization is initiated at the orifice and is incomplete and a broad distribution is not yet formed. As the atomization proceeds smaller droplets are formed and the spread, span, is increased from 2.71 to 3.3 because there are still a lot of big droplets in the spray near the orifice. Thereafter the skewness decreases to.37 at 1 cm, which also cause the spread to decrease to under 3. The mean droplet diameter, D[,3] follows the same trend as described. At first the mean droplet size is µm and decreases at most to 23 µm at cm distance from the orifice. Then the size increases somewhat to lie between 2-3 µm. Spray density The spray density can be determined through the obscuration given by Malvern Series 2 during measurement, which gives the amount of laser energy that is scattered or absorbed by the droplets [3]. The more droplets in the laser ray the more light will be scattered and the bigger will the obscuration be. The spray density is analyzed for distances perpendicular to the spray axis. In this way a density profile is established in a cross section. Figure. shows how mean droplet size and obscuration vary with the perpendicular distance to the spray axis. The mean droplet size increases from the centre to the periphery (disregarding the distance. cm, which probably is due to failure measurement) which is due to coalescence. That can be interpreted as the energy is as densest in the centre where it s enough to keep the droplets separated. The same trend is shown by the obscuration which is decreasing showing bigger droplets forming with the perpendicular distance distance [cm],,,,3,2,1 obscuration [-] D[,3] obsc Figure. D[,3] and obscuration plotted against the perpendicular distance to the spray axis. The droplet distribution in the cross section changes with an increasing positive skewness up to approximately.. The spread also increases up to 2.9 and D[,3] is at least 2 µm and increases above 1 µm in the periphery. Comparison between nozzles Spray angles Masters [] describes that the spray angle can considerably not be varied with the air flow and that the spray angle commonly is small. However the spray angle decreases somewhat Bilaga XIV:()

5 with increased air flow or a decrease in the liquid flow. Up on variation of air flow, liquid flow and viscosity, the investigation shows that the spray angle marginally changes. The change is inside the margin of error. However there is a trend in decreasing liquid flow and viscosity and increasing air flow giving a lower spray angle. The widest spray pattern (approximately 9 ) is given by the nozzle wide. The nozzle flat gives approximately the same spray angle. The smallest spray angle is given by the nozzle thin (3 ), but the other nozzles, except nozzle spray dryer, give spray angles in the same area (3- ). The nozzle spray dryer gives approximately. The spray angles however decrease with the distance along the spray axis, which is due to that the spray doesn t expand with the initial spray angle, because the energy presented in the air flow is most concentrated in the centre and decreases towards the periphery. Distributions The variation of D[,3] with the pressure, for six nozzles, is described in Figure 7. Lower pressure gives a bigger mean droplet diameter and higher spread among the measurements. For larger pressure the diameter is smaller as the spread. Span is not changed significantly between the pressure areas pressure [bar] thin coarse spray dryer flat wide cone Figure 7. D[,3] plotted against the pressure, viscosity 1 cp. Conclusions All the studied parameters: viscosity, liquid- and air flow (pressure), nozzles, distance along the spray axis and temperature, have significant effect on the responses. The span increases with the liquid flow for a specific quotient as with an increase with the quotient and thereby even with the pressure. The distributions are positively skewed and follow the Rosin-Rammler distribution. IQCS is approximately.2 for low quotients and. for the highest, which have been studied. When comparing nozzles with each other it s shown that the span is approximately the same. With low pressure the nozzle flat give the highest mean diameter and the nozzles coarse and wide give slightly smaller mean diameters. The rest of the nozzles give the smallest diameters. D1 is at least 2 µm for high pressures independent of the viscosity. For viscosity 1 cp the D[,3] will be at least 8 µm and for viscosity 1 cp at least 2 µm. Generally D[,3] is varied between 1- µm in the study. The spray angle changes marginally with viscosity, air- and liquid flow and the changes are within the margin of error. Acknowledgements The author gratefully acknowledges Amarin Development AB for assisting knowledge and laboratory equipment making this work and outcome able and the supervisors Anders Axelsson at the Department of Chemical Engineering 1 and John Kendrup at Amarin Development AB. References [1] Allan Rehnström, Handboken goda råd om dysor, Technical Manual No. 3A, Spraying Systems Co., Wheaton, Illinois, 199. [2] Jan Schelling, Lothar Reh, Influence of atomiser design and coaxial gas velocity on gas entrainment into sprays, Chemical Engineering and Processing 38 (1999) [3] Malvern Instruments Ltd., Series 2-User Manual, Spring Lane South, Worcs. WR11AQ, England [] Terence Allen, Particle Size Measurement, fourth ed., Chapman & Hall, TJ Press (Padstow) Ltd, Cornwall 1993 [] John Staniforth, Particle-size analysis in Pharmaceutics The science of dosage form design (Edited by M.E. Aulton), second ed., Hartcourt Publishers Limited 1998 [] K. Masters, Spray drying handbook, fifth ed., Longman Scientific & Technical and John Wiley and Sons Inc., New York Bilaga XIV:()