The invention relates to an inlet chamber for a spray dryer, a spray dryer including said inlet chamber, the use of said spray dryer, and methods of forming a spray dried product using said spray dryer.

Background of the invention

In spray dryers, wall deposition of particles or droplets may lead to problems, such as reduced yield; degradation of the products due to long drying time; frequent down-time of the equipment for cleaning, and possible fire and explosion hazards. As spray drying is normally the final stage of a process, wall deposition problems may directly affect the quality of the final products. Many studies have been conducted to understand the factors affecting wall deposition.

Wall deposition is caused by the combined effects of adhesion and cohesion of the particles. In the initial stage of spray drying, particles may adhere to the chamber wall when they collide with it, forming a fine layer, and the strength of this adhesion is affected by the surface energy of the wall materials. Cohesion of other particles to the wall layer particles then occurs on the particle layer as drying proceeds, leading to wall deposition. Numerous studies have been conducted to investigate addressing the problem of wall deposition. A number of studies have looked at altering the material from which the wall is formed. By way of example, forming the wall from Teflon has been shown to result in a lower degree of product adhesion than compared with walls formed from stainless steel or glass. Further studies have suggested that walls formed from a material having a low surface energy (for example Teflon) result in a reduction in wall deposition. Previous studies have also looked at the effect of the surface roughness of the materials and the influence of this property on the likelihood of wall deposition. Both of these properties are particularly relevant in sections of the pray dryer where the impinging velocity and moisture content of the particles are high. In such cases, it is thought that the effect of wall roughness may be even greater than that of surface energy. Further studies have suggested that the reduction in wall deposition by using materials with a lower surface energy was also affected by the nature of the feed materials. In particular, the advantage conferred by forming the wall from a material with low surface energy was less significant in addressing the problem of wall deposition in spray drying processes used to form fat-containing or lactose-rich particles. A further important problem with many low-surface-energy wall materials, such as Teflon and Nylon, is that they become electrostatically charged very readily (having high dielectric constants), resulting in significant deposition of dry particles. These considerations, together with the relatively higher cost of using Teflon in industrial-scale spray dryers, limit the application of low surface-energy wall materials and mean that these low surface-energy materials do not reduce the overall rate of wall deposition for particles, because the high dielectric constants of these materials results in high amounts of dry particle deposition.

Apart from the chamber wall materials, other studies have investigated the effects of operating conditions on the degree of wall deposition, including the temperatures, the feed rates, and the feed compositions. In many studies, it has been suggested that particle stickiness is a function of the particle temperature and the glass- transition temperature of the material. For instance, researchers have investigated the relationship between the "sticky point", the critical temperature at which particles become sticky to form deposition, and the glass transition temperature Tg as well as the moisture content. These studies concluded that the sticky point is around 10-20 °C higher than the Tg, and both the temperatures decrease with increasing moisture content. Such reduction in the transition temperature becomes problematic for amorphous particles if the value of Tg falls below the particle temperature. It should be noted, however, that there is a limitation when only using Tg as the index to control wall deposition. Since correlations that consider only the glass-transition temperature do not consider the rigidity and collision dynamics of particles, these models normally over- predict deposition rates. Apart from the temperature, the feed flowrates may also influence the adhesion tendency. Several experiments performed using pilot-scale spray dryers showed that increasing the feed flowrates causes higher deposition fluxes or lower product yields. This situation arises because, at the same air temperature, increasing the feed rates introduces more moisture into the chamber. Increasing the liquid flowrate through the atomiser also increases the droplet sizes, leading to shorter residence times, lower drying rates, and less evaporation. Another factor in the feed stream affecting wall deposition is the feed composition, and particularly the presence of carriers in the feed. Altering the feed composition with macromolecules, such as maltodextrin, has also been trialed to increase the glass-transition temperature and reduce the stickiness of the product. The relatively high glass-transition temperature of maltodextrin raises the overall glass-transition temperatures of the particles. In addition, some carriers, such as proteins, can form a layer on the surface of the particles, reducing their stickiness to the chamber wall. Despite the effectiveness of using spray carriers, the required amount of these additives varies depending on the type of core material. Therefore, specific correlations or trial-and-error processes are normally required for different products. Still further, the use of cooled walls has been investigated. The concept here is that cooling the walls may make the particles less sticky since stickiness is a function of temperature. The disadvantages with this approach include additional heat loss from the dryer and the formation of condensation on the walls which in turn leads to increased particle deposition. For at least these reasons, this approach is rarely used. Another important factor influencing the wall-deposition tendency is the design of the drying chamber. Studies of the chamber design have mainly focused on two areas: the inlet plenum chamber, and the main chamber geometries. Both factors can directly affect the flow pattern of drying air, which also influences the impacts of particles on the walls. Several studies have been conducted on the relation between the angle swirl vanes and stability of flow. Experimental and numerical studies suggested that a 25°swirl angle is optimal for adequate gas-spray mixing without increasing wall deposition.

Apart from the inlet vane angle, the geometries of the drying chambers are also very important in controlling the amount and patterns of particle wall deposition. Various researchers have investigated different chamber geometries (e.g. cylinderon-cone, conical, hour-glass-shaped, and lantern-shaped). Simulation results indicate that different chamber geometries cause variations in the droplet trajectories and the particle residence times. Compared with the previously discussed approaches, changing the chamber design essentially controls the flow pattern of drying air rather than the feed liquid stream, thus this factor is less influenced by the properties of different feed materials. Such an advantage makes the approach of changing the chamber design a promising solution to wall deposition problems. However, an important consideration, when changing the chamber design to control the amount of wall deposition, is the difficulty in the measurement of, and the poor understanding of, the complex air flow patterns inside the chamber.

Still, another approach has involved modifying the dryer to include air sweeps. Air sweeps are nozzles located near the walls that blow air into the dryer that aim to keep the walls clean. However, the effect of air sweeps is localised around the sweeps themselves, and the swept area is only a small part of the dryer. This approach has been adopted by industry, but is only a partial solution.

A further approach is the use of hammers on the walls. Hammers are often used on the walls of spray dryers (from the outside) to vibrate the walls and dislodge deposits of particles on the inside walls of the dryers. The area of application tends to be localised to each hammer location, so deposits are not removed from the walls across the entire dryer, and this solution is noisy, damages the dryer walls, and is a respective patch-up rather than addressing the source of the problem. In view of the above, an object of the present invention is to address at least one of the aforementioned shortcomings of the prior art.

Reference to any prior art in the specification is not an acknowledgment or suggestion that this prior art forms part of the common general knowledge in any jurisdiction or that this prior art could reasonably be expected to be understood, regarded as relevant, and/or combined with other pieces of prior art by a skilled person in the art.

Summary of the invention

In one aspect of the invention, there is provided a spray dryer comprising: a cylindrical drying chamber having an internal diameter; an inlet region configured to provide a radially inward expanding flow of drying gas to the drying chamber, the inlet region including: an annular chamber axially aligned with the cylindrical drying chamber and formed between an inner concentric cylindrical wall having an external diameter and an outer concentric cylindrical wall having an internal diameter; wherein the internal diameter of the outer concentric cylindrical wall is equal to the internal diameter of the cylindrical drying chamber; and an atomiser to disperse atomised liquid into the drying gas.

In an embodiment, the inlet region is configured to provide the flow of drying gas to the drying chamber with no radially outward expansion.

In an embodiment, a ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is from 1 : 10 to 3:4. Preferably, the ratio is 1 :4 to 3:4. More preferably, the ratio is 1 :3 to 2:3. Even more preferably, the ratio is 2:5 to 3:5. Most preferably, the ratio is about 1 :2. In an embodiment, the ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is at least 1 : 10, preferably at least 1 :4, more preferably at least 1 :3, and most preferably 2:5. Additionally, or alternatively, the ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is at most 3:4, preferably at most 2:3, and more preferably at most 3:5.

In an embodiment, the inner concentric wall is a wall of the atomiser.

In an embodiment, the annular chamber has an axial length, and a ratio of the internal diameter of the outer concentric wall to the axial length is from 2:5 to 4:5.

In an embodiment, the annular chamber further includes a swirl module at the downstream end of the annular chamber. In one or more forms of this embodiment, the swirl module comprises a plurality of vanes arranged to provide an inlet swirl angle of from 5 to 15°. Preferably, the vanes are arranged to provide an inlet swirl angle of from 7° to 9°.

In an embodiment, a ratio of an axial length of the annular chamber to the drying chamber is from about 3: 17 to 7: 17. Preferably, the ratio of the axial length of the annular chamber to the drying chamber is from about 4: 17 to 6: 17.

In an embodiment, the spray dryer is a modular spray dryer, and wherein the inlet region comprises at least one inlet module and the drying chamber comprises at least one drying chamber module, wherein the inlet module and the drying chamber module are coupled and are decouplable via a coupling mechanism.

In an embodiment, the spray dryer further comprises a post processing region connected to an outlet of the drying chamber. In one form of this embodiment, the post processing region comprises at least one post processing module, wherein the post processing module is coupled to the drying chamber, and is decouplable from the drying chamber via a coupling mechanism.

In another aspect of the invention, there is provided an inlet module for a spray dryer, the inlet module configured to provide a radially inward expanding flow of drying gas to a cylindrical drying chamber of the spray dryer, the inlet module including: an annular chamber configured to axially aligned with the cylindrical drying chamber, the annular chamber formed between an inner concentric cylindrical wall having an external diameter and an outer concentric cylindrical wall having an internal diameter, wherein the internal diameter of the outer concentric cylindrical wall is equal to an internal diameter of the cylindrical drying chamber; an atomiser to disperse atomised liquid into the drying gas; and a coupling mechanism for coupling the inlet module to the drying chamber of a spray dryer.

In an embodiment, a ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is from 1 : 10 to 3:4. Preferably, the ratio is 1 :4 to 3:4. More preferably, the ratio is 1 :3 to 2:3. Even more preferably, the ratio is 2:5 to 3:5. Most preferably, the ratio is about 1 :2.

In an embodiment, the ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is at least 1 : 10, preferably at least 1 :4, more preferably at least 1 :3, and most preferably 2:5. Additionally, or alternatively, the ratio of the external diameter of the inner concentric wall to the internal diameter of the outer concentric walls is at most 3:4, preferably at most 2:3, and more preferably at most 3:5. In an embodiment, the annular chamber has an axial length, and a ratio of the internal diameter of the outer concentric wall to the axial length is from 2:5 to 4:5.

In an embodiment, the annular chamber further includes a swirl module at the downstream end of the annular chamber. In one or more forms of this embodiment, the swirl module comprises a plurality of vanes arranged to provide an inlet swirl angle of from 5 to 15°. Preferably, the vanes are arranged to provide an inlet swirl angle of from 7° to 9°.

In a further aspect of the invention, there is provided the use of a spray dryer as previously described. In still further aspects of the invention, there is provided a method of forming a spray dried product, the method including operating a spray dryer as previously described to prepare a spray dried product.

In yet a further aspect of the invention there is provided a method of forming a spray dried product, the method including: supplying a drying gas through an inlet region of the spray dryer to a cylindrical drying chamber of the spray dryer, the inlet region including: an annular chamber axially aligned with the cylindrical drying chamber and formed between an inner concentric cylindrical wall having an external diameter and an outer concentric cylindrical wall having an internal diameter; wherein the internal diameter of the outer concentric cylindrical wall is equal to an internal diameter of the cylindrical drying chamber; injecting atomised liquid into the inward converging flow of drying gas in the cylindrical drying chamber to dry the atomised liquid; and forming a spray dried product. In yet another aspect of the invention, there is provided a spray dried product formed using the spray dryer and/or the methods described above. As used herein, except where the context requires otherwise, the term "comprise" and variations of the term, such as "comprising", "comprises" and "comprised", are not intended to exclude further additives, components, integers or steps. Further aspects of the present invention and further embodiments of the aspects described in the preceding paragraphs will become apparent from the following description, given by way of example and with reference to the accompanying drawings.

Brief description of the drawings

Figure 1 : A schematic diagram of the outward and inward expansion geometries.

Figure 2: Typical inlet geometry for laboratory-scale spray dryers, including Buchi designs.

Figure 3: Typical inlet geometry for Niro and Anhydro spray dryers.

Figure 4: Typical inlet geometry for simple sudden expansions. Figure 5: Schematic diagram of the inlet geometry according to an embodiment of the invention.

Figure 6: An illustrative schematic showing a modular spray drying apparatus, spray drying unit, and post-processing region.

Figure 7: Illustration of the setup of the experimental flow apparatus used. Figure 8: Schematic illustrating the the experimental spray drying apparatus.

Figure 9: Nomenclature for the axial and radial locations of the pressure measurement points.

Figure 10: Pressure measurements and FFT analysis results of the outward expansion geometry. The data were obtained at point 1 a (15 cm axial and 15 cm radial distances). Operating condition: high flow case (480 m ³ /h, Re = 36 000). Figure 11 : Comparison of the pressure measurements and FFT analysis results of the measurement points 1 a, 2a and 3a (15 cm radial, 15, 20 and 25 cm axial distances). The operating conditions are the same as in Figure 10.

Figure 12: Comparison of the pressure measurements and FFT analysis results of the measurement points 1 b, 2b and 3b (7.5 cm radial, 15, 20 and 25 cm axial distances). The operating conditions are the same as in Figure 10.

Figure 13: Comparison of the pressure measurements and FFT analysis results of the measurement points 1 c, 2c and 3c (5 cm radial, 15, 20 and 25 cm axial distances). The operating conditions are the same as in Figure 10. Figure 14: Original (top) and enhanced (bottom) images of the smoke visualisation of the flow produced from the outward expansion. The figures show the asymmetric oscillation of the smoke stream.

Figure 15: Comparison of the pressure measurements and FFT analysis results of the measurement points 1 a (15 cm radial and axial distances). The operating conditions are the same as in Figure 10.

Figure 16: Comparison of the pressure measurements and FFT analysis results of the measurement points 1 a, 1 b and 1 c (5, 7.5 and 15 cm radial, 15 cm axial distances). The operating conditions are the same as in Figure 10.

Figure 17: Comparison of the pressure measurements and FFT analysis results of the measurement points 1 c, 2c and 3c (5 cm radial, 15, 20 and 25 cm axial distances). The operating conditions are the same as in Figure 10. The arrows with different colours indicate the peaks that appear to "shift" in frequency as the measurement point changes.

Figure 18: Smoke visualisation results in two flow cases: (a) inlet fan speed of 600 rpm, outlet low flow (240 m ³ /h, Re = 18 000), (b) high flow (480 m ³ /h, Re = 36 000). Boxes in the figures show the increased length of a complete eddy.

Figure 19: Lightness (a) and FFT (b) analysis of the visualisation results at regions approximately 15 cm, 20 cm and 25 cm from the expansion point. The operating conditions are the same as in Figure 10. The FFT analysis results of the pressure measured at different vertical positions (c) are shown as a comparison with the lightness measurements.

Figure 20: Comparison of the pressure measurements and FFT analysis results for the swirl and non-swirl cases. The data were obtained at the position 1 c (5 cm radial and 15 cm axial distances). The operating conditions are the same as in Figure 10. The boxes highlight the frequency range for the bands of peaks in both cases.

Figure 21 : Comparison of the pressure measurements and FFT analysis results for the swirl and non-swirl cases. The data were obtained at the position 1 b (7.5 cm radial and 15 cm axial distances). The operating conditions are the same as in Figure 10.

Figure 22: Comparison of the pressure measurements and FFT analysis results for the swirl and non-swirl cases. The data were obtained at the position 1 a (15 cm radial and axial distances). The operating conditions are the same as in Figure 10. Figure 23: Comparison of the eddy patterns observed in the non-swirl and swirl cases, (a) and (b) were captured at low flowrates (240 m ³ /h, Re = 18 000), (c) and (d) were captured at high flowrates (480 m ³ /h, Re = 36 000).

Figure 24: Comparison of the dissipation behaviour of the smoke stream in the non-swirl (top) and swirl (bottom) cases. Operating condition are the same as in Figure 10. Boxes in the figures highlight the area where the eddies dissipated.

Figure 25: Lightness and FFT analysis for the visualisation results at regions approximately 15 cm, 20 cm and 25 cm from the expansion point with swirl. The operating conditions are the same as in Figure 10.

Detailed description of the embodiments In spray dryers, one of the major effects of the unsteady flow patterns is that the droplets entrained in the flow can be carried to the chamber wall, thereby causing wall deposition. Experimental and simulation studies conducted by Oakley et al. (D. E. Oakley, R. E. Bahu, and D. Reay, "The Aerodynamics of Co-Current Spray Dryers," in Sixth International Drying Symposium IDS, 1988, vol. 88, pp. 5-8 - the entire disclosure of which is hereby incorporated by reference) and Oakley & Bahu (D. E. Oakley and R. E. Bahu, "Spray/gas Mixing Behaviour within Spray Dryers," Drying, vol. 91 , pp. SOS- SI S, 1991 - the entire disclosure of which is hereby incorporated by reference) on a co- current short-form spray dryer showed that recirculation zones were formed near the chamber wall. Whilst the droplets of small to moderate sizes were carried through the dryer by the central jet and were evaporated quickly, the larger droplets were likely to travel into the recirculation zone. Simulation work by Frydman et al. (A. Frydman, J. Vasseur, F. Ducept, M. Sionneau, and J. Moureh, "Simulation of Spray Drying in Superheated Steam Using Computational Fluid Dynamics," Dry. Technol., vol. 17, no. 7-8, pp. 1313-1326, 1999; and A. Frydman, J. Vasseur, J. Moureh, M. Sionneau, and P. Tharrault, "Comparison of Superheated Steam and Air Operated Spray Dryers Using Computational Fluid Dynamics," Dry. Technol., vol. 3937, no. March, 2007 - the entire disclosures of each of these documents are hereby incorporated by reference) on a co- current tall-form spray dryer also showed similar results. The simulation results indicated that larger droplets of around 80 pm tended to penetrate through the central jet and were carried into the recirculation zones. In the recirculation zones, a significant portion of droplets then adhered to the chamber wall and formed deposits. Besides the recirculation zones, excessive swirling of the inlet air also increases the amount of wall deposition, as observed in the flow patterns by Southwell & Langrish (D. B. Southwell and T. A. G. Langrish, "The Effect of Swirl on Flow Stability in Spray Dryers," IChemE, vol. 79, no. April, 2001 - the entire disclosure of which is hereby incorporated by reference). In the case where high swirl was produced, the droplets approached the chamber wall closely, leading to excessive wall deposits. A subsequent study by Ozmen & Langrish (L. Ozmen and T. A. G. Langrish, "An Experimental Investigation of the Wall Deposition of Milk Powder in a Pilot-Scale Spray Dryer," Dry. Technol., vol. 21 , no. 7, pp. 1253-1272, 2003 - the entire disclosure of which is hereby incorporated by reference) on the deposition fluxes of skim milk also found that changes in inlet swirl angle had a significant effect on the amount of wall deposition. The study showed an increase of 44% in the deposition flux by increasing the swirl angle from 0° to 30°, and the most significant increase occurred when the angle was increased from 25° to 30°.

Many techniques have been employed to analyse the complex and transient fluid behaviours in equipment with complicated geometries, including sudden expansions. The techniques can be divided into two categories: direct analysis using flow visualisation and indirect interpretation of other measurements. Visualisation techniques have been one of the fundamental methods used to interpret and analyse the flow patterns. Particle Image Velocimetry (PIV) is an effective way to capture the characteristics of the flow field. By seeding particles into the air flow, the technique analyses the area of interest illuminated by a stroboscopic laser sheet. By taking images of the illuminated area, the technique is capable of producing an instantaneous velocity map of the full flow field. This technique has been used to investigate the relationship between the Reynolds number and the reattachment of laminar flow through annular outward and inward expansions. This technique has also been applied on a pilot-scale co-current spray dryer to successfully capture the deflected central jet and recirculation zones near the chamber wall. Another visualisation technique that has been widely applied in different fluid fields is Laser Doppler Velocimetry (LDV), which was used to make comprehensive observations of the flow patterns in a pilot-scale spray dryer (see Southwell & Langrish (D. B. Southwell and T. A. G. Langrish, "Observations of Flow Patterns in a Spray Dryer," Dry. Technol., vol. 18, no. 3, pp. 661- 685, 2000 - the entire disclosure of which is hereby incorporated by reference)). The complimentary results of direct flow visualisation using cotton turfs and LDV showed several distinctive features of the air flow through the axisymmetric expansion after the inlet of the dryer. There are also some techniques that capture the flow patterns as images for quantitative analysis. For instance, Lebarbier et al. (C. Lebarbier, T. K. Kockel, D. F. Fletcher, and T. A. G. Langrish, "Experimental Measurement and Numerical Simulation of the Effect of Swirl on Flow Stability in Spray Dryers," Chem. Eng. Res. Des., vol. 79, no. 3, pp. 260-268, 2001 - the entire disclosure of which is hereby incorporated by reference) applied the flow visualisation technique in a Perspex model of a co-current, short-form pilot-scale spray dryer. The experiment used water seeded with polyethylene flakes as the working fluid. The movement of the flakes through a light sheet was recorded as video images. The variation of lightness in the area of interest was extracted from the video frames, from which the time series of lightness was established. The time series was further analysed using Fast Fourier Transforms (FFT) in order to determine the frequency of the flow precession. A similar study was conducted by Narayanan et al. (V. Narayanan, M. D. Lightfoot, S. A. Schumaker, S. A. Danczyk, and B. Eilers, "Use of Proper Orthogonal Decomposition towards Time- Resolved Image Analysis of Sprays," DTIC Document, 201 1 - the entire disclosure of which is hereby incorporated by reference) on a precessing spray. The experiment captured high-speed videos of the oscillating spray produced by a swirl coaxial injector. By extracting the variation of lightness between frames, the authors were able to conduct Proper Orthogonal Decomposition (POD) on the area of interest. The analyses showed the most dominant flow pattern in the spray, allowing the main frequency of the oscillation to be determined.

Visualisation results have also been combined with velocity and pressure measurements. Kieviet et al. (F. G. G. Kieviet, J. Van Raaij, P. P. E. A. De Moor, and P. J. A. M. Kerkhof, "Measurement and Modelling of the Air Flow Pattern in a Pilot-Plant Spray Dryer," Chem. Eng. Res. Des., vol. 75, no. 3, pp. 321-328, 1997 - the entire disclosure of which is hereby incorporated by reference) and Woo et al. (M. W. Woo, W. R. W. Daud, A. S. Mujumdar, Z. Wu, M. Z. M. Talib, and S. M. Tasirin, "Non-Swirling Steady and Transient Flow Simulations in Short-Form Spray Dryers," Chem. Prod. Process Model., vol. 4, no. 1 , 2009 - the entire disclosure of which is hereby incorporated by reference) employed a hot-wire anemometer system to obtain the distribution of velocities in a flow pattern as a function of time. An anemometer was placed on arms that could move axially and radially in order to reach different positions in the chamber. Although the system could not determine an accurate mean velocity for unsteady flow, different flow characteristics, such as reversed flow and unsteady flow, could be interpreted from the velocity distributions obtained at various locations. Nathan et al. (G. J. Nathan, S. J. Hill, and R. E. Luxton, "An Axisymmetric 'fluidic' Nozzle to Generate Jet Precession," J. Fluid Mech., vol. 370, pp. 347-380, 1998 - the entire disclosure of which is hereby incorporated by reference) also employed pressure measurements as a means to interpret the flow field. Pressure measurements were obtained at four different locations in the expansion chamber (expansion point, chamber wall upstream, chamber wall downstream, and inside the downstream chamber, respectively). The pressure measurements agreed with the two types of jet behaviours that were observed using flow visualisation techniques. This experimental work also showed the ability to combine flow visualisation data and pressure measurements in order to provide a more comprehensive understanding of complex flow patterns. All of the studies mentioned above relate to flow patterns produced from outward expansion geometries. In contrast, the present invention relates to inward expansion geometries (also referred to as annular backsteps), which have not been well studied. The differences between outward and inward expansions are depicted in Figure 1. In the outward expansion, fluid is introduced into the radial centre of the expansion chamber, from which the flow expands outwards to the chamber wall. In the inward expansion, however, the fluid is introduced from the perimeter of the chamber and expands towards the centre.

Figure 2, Figure 3, and Figure 4 illustrate inlet arrangements for known spray dryers.

Figure 2 illustrates the typical inlet geometry for laboratory-scale spray dryers, including Buchi designs. As can be seen, this inlet geometry includes an annular inlet that introduces a drying gas into a drying chamber. The internal diameter of the drying chamber is greater than the internal diameter of the annular inlet. Figure 2 shows that as the drying gas is introduced into the drying chamber, there are regions of outward expansions and inward expansions. Figure 3 illustrates the typical inlet geometry for Niro and Anhydro spray dryers.

These spray dryers include a conical annular air inlet which introduces a flow of drying gas into the drying chamber toward the centre of the drying chamber. This manner of inlet results in large outward expansions as illustrated in Figure 3.

Figure 4 illustrates the typical inlet geometry for simple sudden expansions (atomiser not shown). The arrows show the flow, including the outward expansions, which are also unstable with respect to time.

In view of the above, in one form the invention relates to a spray dryer comprising: a cylindrical drying chamber having an internal diameter; an inlet region configured to provide a radially inward expanding flow of drying gas to the drying chamber, the inlet region including: an annular chamber axially aligned with the cylindrical drying chamber and formed between an inner concentric cylindrical wall having an external diameter and an outer concentric cylindrical wall having an internal diameter; wherein the internal diameter of the outer concentric cylindrical wall is equal to the internal diameter of the cylindrical drying chamber; and an atomiser to disperse atomised liquid into the drying gas. In another form, the invention relates to an inlet module for a spray dryer, the inlet module configured to provide a radially inward expanding flow of drying gas to a cylindrical drying chamber of the spray dryer, the inlet module including: an annular chamber axially aligned with the cylindrical drying chamber and formed between an inner concentric cylindrical wall having an external diameter and an outer concentric cylindrical wall having an internal diameter, wherein the internal diameter of the outer concentric cylindrical wall is equal to an internal diameter of the cylindrical drying chamber. The invention also relates to a spray dryer comprising this inlet module.

Figure 5 illustrates an annular inlet 500 and cylindrical drying chamber 502 of a spray dryer according to one form of the invention. As can be seen, the annular inlet is formed between two concentric cylindrical walls 504 and 506. The inner concentric wall 504 is an external wall of atomiser 508 and has an external diameter D1 . The outer concentric wall 506 is an internal wall of the annular inlet 500 region of the spray dryer and has an internal diameter D2. The drying chamber 502 has a cylindrical wall 510 of internal diameter D3. In the depicted embodiment, the outer concentric wall 506 of the annular inlet 500 and the cylindrical wall 510 of the drying chamber 502 are commonly formed and diameters D2 and D3 are the equal. In such embodiments, where D2 and D3 are equal, the annular flow of drying gas flows straight down the outside wall and is confined to inward expansion. This arrangement advantageously prevents outward expansion and additionally shields the walls from any instability of the inward expansions. Thus, the invention mitigates and/or avoids issues of wall deposition that are prevalent in prior art spray dryers where outward expansion does occur.

The inventors have found that adopting an annular chamber geometry with a ratio of a diameter of the inner and outer concentric walls of the annular chamber of from 1 : 10 to 3:4 is particularly beneficial as this further assists in stabilising the inward expansions thus preventing or further reducing wall deposition. Alternatively, or additionally, embodiments of the invention include a swirl module which also assists in stabilising the inward expansions.

Figure 6 (a) is an illustrative schematic showing a modular spray drying apparatus 200. The spray drying apparatus 200 includes a spray drying unit 202 comprising an inlet region 204 in the form of a single module and a drying chamber 206 formed from two modules 206a and 206b. Module 206b of the drying chamber 206 is connected to a post-processing chamber 208 via a pipe 210 that connects an outlet of module 206b to an inlet of the post-processing region 208.

Figure 6 (b) provides a more detailed illustration of the spray drying unit 202. The inlet region 204 of the spray drying unit 202 includes an annular chamber 212. The annular region is sized to provide drying gas into the drying chamber 206 in a radially inward converging manner. In use, drying gas is provided through an inlet 216 located at an upstream location of the inlet region 204, the drying gas flows through the inlet region 204 within the annular chamber 204 downstream towards the drying chamber 206. This inlet region structure provides a flow of drying gas into the drying chamber in a manner such that the flow of drying gas converges towards a central axis 214 of the spray drying unit 202. Atomiser 216 provides a spray of liquid droplets into the drying gas, which droplets are then dried to form the spray dried product.

Figure 6 (c) is a more detailed illustration of the post processing region 208. As can be seen, the post processing region 208 is formed from three separate modules Module 218 is an inlet module for receiving the drying gas with the spray dried product entrained therein. The drying gas with entrained spray dried product is further treated in module 220 of the post processing region 208, where additional processes (subject to the nature of the spray dried product), such as solid-phase crystallization, can be enhanced. Module 222 is an outlet module of the post processing region 208 where the drying gas with or without post processed spray dried product is withdrawn from the post processing region 208.

The post processing region 208 depicted in Figure 6 is an upward air flow post processing region. This may be operated in a manner which allows for separation of the post processed spray dried product from the drying gas, for example, by allowing the downward settling of particles with a sufficiently low drying gas flow rate. Similarly, the post processing region 208 may be operated with a drying gas flow rate that promotes fluidisation of the spray dried product for treatment post processing in a particular zone or module of the post processing region 208. The post processing region 208 may be controlled to vary one or more of: temperature, humidity (such as further drying and/or the addition of moisture, for example by steam injection), and/or the residence time of particles in the post processing region. The post processing region 208 may include extra particle separation devices, extra drying and mass-transfer devices, such as fluidized beds, and combinations of these devices.

The above description relates to a modular spray drying arrangement. The modular arrangement provides flexibility and interchangeability. This allows each section of the dryer (inlet region, drying chamber, and the post-processing region) to be modified or arranged into a different configuration (such as change in size or changing between co-flow and counter-flow) for specific products.

Notwithstanding the above, in other forms, the invention relates to an integrally formed spray dryer including an inlet region including an annular chamber axially aligned with the drying chamber and formed between inner and outer concentric cylindrical walls, the inlet region for receiving a flow of drying gas from an inlet and to provide a radially inward converging flow of drying gas as it leaves the annular chamber. As above, the inventors have found that adopting an annular chamber geometry with a ratio of a diameter of the inner and outer concentric walls of the annular chamber of from 1 : 10 to 3:4 is particularly beneficial.

Examples

These examples investigate the flow patterns produced by inward expansion geometry by means of pressure measurements and flow visualisation, and compare this geometry with normal outward expansion geometry in terms of flow stability. Experimental setup

Experiments were conducted using a sudden expansion apparatus as shown in Figure 7 and Figure 8. The apparatus consisted of two main sections: the pre- expansion chamber, and the expansion chamber. Both of the chambers were made of Perspex. Four sets of pressure measurement ports were placed along the vertical axis of the expansion chamber. The four sets were equally spaced around the perimeter of the chamber. Two pre-expansion sections were made for outward and inward expansions, as shown in the figure. For the inward expansion, air flow was introduced into the plenum chamber, where the flow was made more uniform by a perforated plate. The centres of both the plenum and the pre-expansion sections were occupied coaxially by a housing, so air flow could only travel through the outer periphery of the chambers. A 3D-printed swirl module may be placed at the entrance of the expansion chamber. The ring-shaped module contained 16 vanes angled at 10°. The module produced an inlet swirl angle of around 7° to 9° in the inlet air flow within the air velocity range of 0.9 to 1.9 m/s. The pre-expansion section could be easily changed between inward and outward geometries by replacing the whole section. The dimensions of both the pre- expansion sections are summarised in Table 1 below.

Table 1: Dimensions of the pre-expansion and expansion sections for both inward and outward expansion geometries.

Two flow scenarios were used, which are referred to as 'high flow' and 'low flow' cases respectively. All the quantitative analysis has been conducted using the high flow case, with the low flow case only used as a comparison in the qualitative analysis of the visualisation results. The basic characteristics of both cases are summarised in Table 2.

Table 2 Flow characteristics of the "low flow" and "high flow" cases used in the experiment.

Pressure Measurement

The pressure measurements were obtained using low range differential pressure transducers (Validyne Engineering, Northridge, USA, PC17), with a measurement range of + 25 Pa and accuracy of 0.125 Pa. For each measurement, two adjacent measurement ports at the same distance from the expansion point were used. The variation of pressure difference between the two ports was measured. The sampling frequency was set to be 500 Hz, and the measurement time for each run was at least 60 s to ensure sufficient signal capture for a range of frequencies to be analysed from 0.03 s ^"1 to 250 s ^~1 . To investigate the variation of pressure inside the expansion chamber, extension tubes of difference lengths (7.5 cm and 10 cm, respectively) were placed at one of the two measurement ports, through which the pressure differences between the tubes and the other port were measured. Pressure differences were measured at nine positions in total. The axial and radial locations of the nine measurement points are shown in Figure 9. The time series of pressure measurements were normalised by subtracting the mean pressure. The data were then analysed using Fast Fourier Transforms, in order to determine the dominant frequencies in the flow patterns.

Flow Visualisation

The flow visualisation was performed by introducing a smoke stream into the chamber. For the outward expansion case, a smoke generator (Antari Lightning and Effects Ltd., Taiwan, Rave AF-1214) was connected to the inlet of the pre-expansion section by a flexible pipe of 100 mm I.D. Once the generator was fully heated, the smoke was introduced through the pipe over five minutes. During this time, the movement of smoke in the expansion section was filmed at a frame rate of 30 FPS. To enhance the video quality, the expansion section was mostly covered by black paper, with a torch light placed at the bottom of the chamber. As the generator did not have any pumping system, a suction flow was required in the chamber. Therefore, the inlet fan speed was slightly reduced in both low flow and high flow cases. Pressure measurements showed that no distinctive difference was caused by such a reduction in inlet fan speed, therefore the changes to the flow patterns were negligible. Quantitative analyses were then performed on the captured videos by a similar approach to Narayanan et al. (V. Narayanan, M. D. Lightfoot, S. A. Schumaker, S. A. Danczyk, and B. Eilers, "Use of Proper Orthogonal Decomposition towards Time- Resolved Image Analysis of Sprays," DTIC Document, 201 1 - the entire disclosure of which is hereby incorporated by reference). Frames of 1920 by 1080 pixels were extracted from the captured video using motion analysis software (PhysMo V2.0). The frames were converted from RGB space into CIE 1976 l_Vb ^* space, in which the lightness indices of each pixel were obtained. For each frame, several areas were selected for quantitative analysis. The vertical positions of the analysed areas (namely the distances from the expansion point) were selected in accordance with the measurement points for pressure differences. The radial positions of the areas were chosen to provide distinctive variations of lightness over time. The lightness of the areas of interest were stored in the following matrix forms:

^■ · Li _j

Equation (1 )

Li _! · ^• L,

where / and j represent the two-dimensional coordinates of individual pixels in the selected area. The average lightness was calculated for each area in every frame, and the variation of the averaged lightness over time was established. FFT analysis was also conducted on the time series of lightness in order to determine the most distinctive frequencies with the highest signal amplitudes.

Results and Discussion The Flow Pattern Produced from the Outward Expansion

Pressure Measurements

The pressure differences were measured at different positions of the expansion chamber, in order to capture the flow behaviours across the flow field. The present work focused more on the upstream section of the expansion chamber, since the flow patterns in this region were influenced by the expansion geometries more strongly. Measurements have been first obtained for the outward expansion geometry. The measurement data at position 1 a are shown in Figure 10 as a typical example. Figure 10 shows that there was a variation of the pressure difference from -5 to 5 Pa. Two types of variation can be observed in the pressure measurements: fluctuations of small magnitudes but high frequencies, and large periodic variations of low frequencies. The oscillation of a higher frequency was found throughout the whole period as a 'baseline', on which was superimposed the low frequency variation. Such superimposition between two types of flow oscillations was found to be similar to previous simulation work.

The two types of variation can also be seen in the frequency spectrum obtained by FFT. A frequency range of 0.1 to 100 Hz was chosen as the range of interest, so that the 'DC-like' component of very high magnitudes at lower frequency could be excluded while keeping the other meaningful signals. The two strongest peaks were observed at the frequencies of 0.5 and 0.7 Hz, whilst a band of lower peaks were found around a frequency of 10 Hz. Instead of individual sharp peaks, the band consisted of a spectrum of signals with lower magnitudes, with a few distinctive peaks. The higher magnitudes of the peaks at low frequencies indicated that the flow pattern in the expansion chamber was dominated by the precession of the air flow, and the precession was coupled with a high frequency oscillation. Another point worth noting is that multiple signals were found to have high magnitudes (0.5 and 0.7 Hz). This phenomenon showed the complexity of the flow pattern, with a mixture of various frequencies, which was also reported by Woo et al. [10] when simulating the radial velocity of the air flow in a spray dryer.

In order to further understand the flow pattern inside the expansion chamber, pressure differences were measured at different axial and radial positions. For ease comparison, the results were divided into three groups based on the radial positions, as shown in Figure 11 , Figure 12, and Figure 13. Figure 11 shows that, for the same radial distance, the pressure variations obtained at different axial positions appeared to have a similar trend. Data from all three positions showed the strongest pattern within the range of 0.6 to 1 Hz. No significant difference was observed in the magnitudes of the peaks as the measurement point became further away from the expansion. This phenomenon indicated that there were no significant changes in the flow pattern along the chamber wall, and that the precession behaviour of low frequency remained dominant in the upstream region of the expansion chamber. It should be noted that, although within a close range, there were slight 'shifts' in the frequencies between the data obtained at different axial distances. Since measurements were taken separately for different measurement points, such shifts were possibly caused by the variation between each run. The shifting of peaks indicated that, although the flow pattern showed a characteristic frequency range, the actual frequencies of the oscillations could still vary slightly. As such a variation in frequency was relatively small (no greater than 0.1 Hz), these peaks were considered to belong to the same group.

Significant differences could be observed in the data as the measurement position changed to 7.5 cm from the radial centre of the expansion chamber (see Figure 12). The magnitudes of the low frequency oscillations were reduced by around 0.1 Pa, whilst the magnitudes of the signals at higher frequencies were increased by around 0.2 Pa. Such changes in magnitudes indicated the different regions in which the flow patterns of low and high frequencies existed. The reduction in the magnitudes of the low frequency signals showed that the precession of the flow possibly approached the chamber wall closely. In spray-drying applications, such a precession of particle- entrained flow may lead to a higher likelihood of collisions with the walls of the spray dryers. The higher magnitudes of the signals at around 10 Hz indicated that the high frequency oscillation became more dominant at this radial distance. Apart from the peaks at the frequency range of 0.6 to 1 Hz, which was also observed at the periphery of the chamber, a few peaks were also found at even lower frequencies of 0.1 to 0.3 Hz.

As the measurement points became closer to the radial centre of the expansion chamber (5 cm), the low frequency oscillation became dominant again with even higher magnitudes. It can be seen from Figure 13 that several distinctive peaks appear in the frequency range of 0.1 to 1 Hz, with magnitudes of 1 to 2 Pa. Such phenomenon probably occurred because the measurement points are placed along the radius of the pre-expansion chamber (5 cm from the centre of the chamber), where the interaction of air flows was expected to be strong. Another point worth noting is that, at the position 15 cm from the expansion point, the highest peak captured was at a frequency of around 0.4 Hz, whilst at positions farther away, higher frequencies appear to be more dominant (0.7 Hz at 20 cm and 0.9 Hz at 25 cm). This phenomenon indicated that, even within the same characteristic range of frequency, the dominance of a specific oscillation may still change along the axis of the chamber. Flow Visualisation

A smoke stream was introduced, together with the inlet air from the pre- expansion chamber, in order to conduct visualisation tests for the flow through an outward expansion. Due to the high flow velocity in the pre-expansion chamber, the smoke stream dissipated quickly after entering the chamber. However, some typical features can still be recognised in the visualisation results, as shown in Figure 14. It can be seen from the Figures that the smoke stream diverged from the chamber axis for a significant distance. Moreover, the position where the stream diverged was found to be changing around the axis during the experiment, which may correspond to the low frequency oscillation observed in the pressure measurement. Although the rapid dissipation of the smoke stream made it difficult to capture any flow patterns that were clear enough for quantitative analysis, the enhanced images still showed the oscillating behaviours of the air flow.

Flow Pattern Produced from the Inward Expansion Pressure Measurements

The pressure differences were then measured for the inward expansion. The measurements followed the same procedures for the outward expansion using the same operating conditions (480 m ³ /h, Re = 36 000). As a comparison, the pressure measurements obtained at position 1 a (15 cm radial and axial distances) for both inward and outward expansion are shown in Figure 15.

Figure 15 shows that that the flow pattern near the chamber wall was relatively steady for the inward expansion, with little variation in the pressure measurements compared with the outward expansion. From the FFT analysis, it also clear that the outward expansion showed a stronger periodic pattern than the inward expansion, with the strongest peak found to be 0.9 Pa s for the outward expansion, and 0.02 Pa s for the inward expansion. This result has illustrated that the outward expansion produced stronger oscillations at both low and high frequencies. Such a difference is of great importance, since it indicates that, at the same flowrate, the inward expansion significantly improved the stability of the flow field near the chamber wall. The stable flow along the chamber walls may reduce the likelihood of collisions with the chamber walls in spray-drying applications, thereby reducing the formation of wall deposits. To further investigate the characteristics of the flow field in the inward expansion, the pressure measurements were then obtained at the same positions as the outward expansion. Due to the stability of the flow field, variations in the pressure readings are very small. Therefore only two groups of data with the most distinctive differences have been discussed: measurements at points 1 a, 1 b and 1 c (15 cm axial distance, 5, 7.5 and 15 cm radial distances) are shown in Figure 16; measurements at points 1 c, 2c and 3c (5, 7.5 and 15 cm axial distance, 5 cm radial distances) are shown in Figure 17.

It can be observed in Figure 16 that the pressures in the central area of the chamber (5 cm and 7.5 cm from the centre) showed much stronger fluctuations than those at the periphery (15 cm from the centre, i.e. the chamber wall). It can also be seen in the frequency spectrum that multiple peaks of relatively high magnitudes are found in the frequency range of 0.1 to 1 Hz for readings in the central area (5 and 7.5 cm radial distances), whilst these patterns were not found in the readings obtained at the periphery. This range was the same as the low frequency oscillation for the outward expansion. This phenomenon indicated that the precessing flow that was previously observed near the wall of the outward expansion could only be found close to the radial centre in the inward expansion. The magnitude of the band at the frequencies of 1 to 20 Hz also increased with decreasing radius, indicating that the high frequency oscillations were also confined to the central area of the expansion chamber. Without wishing to be bound by theory, the inventors believe that the pressure gradient produced by the inward back-step led to flow separation. Although the interaction between the air flows in the central area of the chamber still produced a precessing flow, it had a weaker influence on the flow field near the chamber wall. Hence the flow along the chamber wall remained stable. Figure 17 shows that there is less difference among the vertical positions, compared with changing the measurement ports horizontally. The pressure readings at different positions fluctuated within a similar range of ± 1 Pa except for a few peaks in the measurements closer to the expansion point (15 cm). The power spectra for all three positions still consisted of two types of oscillations, as with the outward expansion, but with much lower magnitudes. The 'shifts' of specific signals at low frequencies can also be observed for the inward expansion (noted by arrows), indicating that the flow patterns still varied within the same characteristic range. Flow Visualisation

(a) Qualitative analysis

To probe the flow pattern produced in the inward expansion more directly, flow visualisation was conducted on the apparatus. Since it had already been found from the pressure measurements that the fluctuation of air flow was stronger in the central area of the chamber (5 to 7.5 cm from the radial centre), this area was therefore considered to be the main focus for the test. To avoid the dissipation that occurred in the visualisation test for the outward expansion, the smoke was introduced independently from the coaxial housing in the pre-expansion section. The captured frames were first analysed qualitatively to understand the characteristics of the flow field. Typical frames captured with two different flowrates (240 and 480 m ³ /h, respectively) are shown in Figure 18.

Figure 18 shows that the smoke stream formed a 'swirling' pattern in the central area of the chamber, with the centre of the stream diverging away from the chamber axis. Such a pattern has been observed in both flow cases, but the pattern appeared to form at a more distinct position from the expansion in the high flow case (around 5 cm) compared with the low flow case (around 2.5 cm). In addition, the length of one complete eddy also increased at higher flowrates. Such differences in the swirl patterns between low and high flowrates may be related to the reattachment length of the flow. It was observed that the reattachment length of the flow after the expansion increased with the flowrate, so the formation of the swirling pattern occurred at a greater distance from the expansion. It has also been observed that dissipation of the smoke stream occurred after 30 to 40 cm from the expansion point (around one diameter of the expansion chamber). Although this dissipation led to a slight increase in the radius of the swirling pattern, the pattern was still confined to the central area of the chamber. This observation agreed well with the pressure measurements, showing that the precession behaviours of the flow mainly occurred in the central area of the chamber, leaving the flow field near the wall relatively more stable than that in the outward expansion. (b) Quantitative analysis Quantitative analysis of the visualisation results has been conducted to further understand the swirling pattern. Three regions have been selected to correspond to the axial positions of the pressure measurements (15, 20 and 15 cm, respectively). The lightness for each frame was treated as a similar indicator to pressure, and the variations in the lightness over time in different regions of interest have been converted to time-series data. The advantage of using lightness as the indicator is that it is directly related to the flow behaviour that has been visualised. The time series of the mean- subtracted lightness and FFT analysis of the data have been shown in Figure 19. As a comparison, the FFT analysis of the pressure measurements at the same axial positions is also shown in Figure 19. The maximum frequency for lightness power spectrum was limited to 15 Hz, since the video was captured at 30 FPS.

Figure 19 shows that the lightness readings showed a more distinct low- frequency variation compared with the pressure measurements (see Figure 16 and Figure 17), indicating that the low frequency oscillations contributed strongly to the swirling pattern that was observed in the visualisation tests. In addition, the frequency domain showed similar patterns as the pressure measurements, but with higher peaks at low frequencies. As with the pressure measurements, the spectrum can also be divided into two regions: multiple peaks from 0.1 to 1 Hz, and a broad band from 1 to 20 Hz. However, large-magnitude peaks were found at the low frequency range (around 1 Hz). The results supported the suggestion that the swirling pattern in the visualisation tests corresponded to the low-frequency oscillations in the pressure measurements. It is possible that, due to the higher stability of the overall flow field, the low-frequency oscillations became weaker in the case of the inward expansion than was the case in the outward expansion. More importantly, the comparison between lightness and pressure measurements showed the use of the visualisation technique as a complimentary tool to assess complex flow behaviour. Swirl and Non-swirl Cases for the Inward Expansions

Pressure Measurements

To further investigate the influence of the inward expansion geometry on the flow patterns, a swirl module was placed in the pre-expansion. The module produced an inlet swirl angle of around 9° at the operating condition (inlet velocity of 1.9 m/s), which corresponded to a swirl number of 0.1 1 according to following approximation:

S = 2/3 tan 0 Equation (2)

where S is the swirl number, d; and d ₀ are the internal and outer diameters of the annular module, and Θ is the vane angle.

Pressure measurements and flow visualisation tests were conducted, following the same procedures as before. From the previous results, the pressure measurements showed some distinctive results at points 1 a, 1 b and 1 c (15 cm axial distance). Therefore, this axial distance was selected for comparison between the non-swirl and the swirl cases. The comparative results at these three positions are shown in Figure 20, Figure 21 , and Figure 22 Figure 20 shows that the pressure measurements in both the swirl and non-swirl cases did not show significant variations (with the variation range being within ± 1 Pa), indicating that the air flow patterns in both cases were stable. However, some differences can be seen when comparing the FFT analysis results between the two scenarios. Firstly, there were several stronger peaks for the non-swirl case over the frequency range of 0.1 to 3 Hz, but peaks of similar frequencies were found to be smaller in the swirl case. The results indicated that placing the swirl module slightly influenced the low-frequency oscillations by reducing their amplitudes. Secondly, as highlighted in the figure, the frequency range of the high-frequency band of peaks appeared to shift to a higher range in the swirl case. When comparing the data obtained further away from the centre of the chamber

(7.5 cm and 15 cm, respectively, see Figure 21 and Figure 22), the differences between the non-swirl and swirl cases became smaller for both pressure measurements and FFT analysis results. This observation indicated that the effects of introducing the swirl had stronger effects on the area closer to the centre of the chamber. This situation is expected, as the previous observation in the non-swirl case already showed that the inward expansion geometry confined the precession of air streams to the central area of the chamber. Flow Visualisation

(a) Qualitative analysis

A smoke visualisation test was conducted for the swirl case, following the same procedures as for the non-swirl case. For ease of comparison, the camera was set up as closely as possible to the location in the previous experiment. Typical frames of the smoke streams for both swirl and non-swirl cases have been shown in Figure 23.

Figure 23 (a), (b) and (c), (d) show that similar swirling patterns were formed in both swirl and non-swirl cases, but there are several differences between the two situations. Firstly, the eddies in the swirl cases dissipated more quickly, which made the edge of the stream less distinct from the background. This phenomenon was more obvious in the high flow scenario (see Figure 23 (c) and (d)). In addition, the length of each eddy appeared to be greater in the swirl case. With low flowrates (240 m ³ /h, see Figure 23 (a) and (b)) particularly, the smoke stream deviated further from the centre of the chamber compared with the non-swirl case.

Besides the differences in the overall patterns of the smoke streams, differences were also observed in the way that eddies dissipated. Consecutive frames of the smoke stream in both cases are shown in Figure 24. Only the figures from the low flow case have been shown because of their better quality, but similar phenomena were also observed in the high flow case. Figure 24 shows that for the non-swirl case (top row), each eddy gradually deformed and dissipated as it travelled downward. In the swirl case, however, the whole smoke stream occasionally broke from the centre of the stream. Such a difference indicated that, by introducing swirl at the entrance, additional angular momentum could be given to the flow, making the flow pattern dissipate more quickly. (b) Qualitative analysis

From the qualitative observations of the visualisation results, frame-by-frame analysis was conducted on the videos. For ease comparison, the frame-by-frame analyses were conducted at the same axial positions as those for the pressure measurements, with both swirl and non-swirl cases (15, 20 and 25 cm from the expansion point). The visualisation results are shown in Figure 25. When comparing the non-swirl (see Figure 19) and swirl cases (see Figure 25), it can be observed that the difference in the magnitudes of the lightness variations between each axial position are smaller in the swirl case. This situation is possibly caused by the dissipation of the smoke stream. As discussed in the qualitative analysis, the smoke steam in the swirl case appeared to dissipate more quickly, making the overall shape of the steam fainter. This situation might have reduced the variation in the lightness when the eddy passed the region being analysed. FFT analysis showed that the data obtained closest to the expansion point (15 cm) appeared to have the highest magnitude in the swirl case, whilst in the non-swirl case, the strongest peaks were observed 25 cm from the expansion point. This difference indicated that, in the swirl case, the flow pattern started to dissipate at a shorter distance from the entrance. The inlet swirl angle appeared to give more cross-stream mixing in the central area of the chamber, causing the swirl pattern to dissipate more rapidly. Conclusion

Pressure measurements and visualisation tests have been conducted on a flow apparatus to investigate different flow behaviours produced from inward sudden and outward expansion geometries. The pressure measurements of the outward expansion geometry have shown a superimposition of two types of oscillations: high-magnitude peaks at low frequencies (around 1 Hz) and a low-magnitude band at higher frequencies (around 10 Hz). The stronger oscillations at low frequencies have been found to be related to the precession of the air flow. The pressure measurements obtained at different radial and axial positions have shown that, for the outward expansion, the precessing behaviour of the flow approached the chamber closely, which has been supported by smoke visualisation tests. Compared with the outward expansion, the pressure measurements in the inward expansion show greater flow stability. The power spectrum shows peaks with two similar ranges of frequencies (around 1 Hz and 10 Hz) that were observed in the outward expansion, but with significantly lower magnitudes. Further measurements at different radial distances showed that the low-frequency oscillations appear to be stronger in the central area of the chamber (5 to 7.5 cm from the radial centre). Qualitative and quantitative analyses of the visualisation results have also shown that the low-frequency precession of the flow is confined to the central area of the expansion chamber. It is possible that the pressure gradients produced by the inward expansion have caused flow separation, which has then led to the precession in the central area of the chamber. Such precession in the inward expansion had smaller effects on the periphery of the chamber, therefore giving higher stability in the flow field near the chamber wall.

A swirl module was placed at the expansion in order to investigate its effect on the air flow. The pressure measurement results indicated that, although the air flow appeared to be steady in both swirl and non-swirl cases, the oscillations at low frequencies became less distinctive once swirl was added. Further flow visualisation has shown that the eddies dissipated more quickly in the swirl case, and the length of each eddy was found to be longer. Visualisation results have also shown that, in the swirl case, the smoke stream has occasionally broken away from the centre, which has not been observed in the non-swirl case. Quantitative analysis of the visualisation results showed a reduction in the magnitudes of the peaks at low frequencies of 0.1 to 10 Hz, which similar to the pressure measurements. The FFT results also indicated that, for the swirl case, the position of the strongest peaks became closer to the expansion point (15 cm and 20 cm from the entrance for the swirl case, compared with 25 cm in the non-swirl case). This change supports the observation in the pressure measurements that the swirling pattern dissipated more quickly with the introduction of swirl.

It will be understood that the invention disclosed and defined in this specification extends to all alternative combinations of two or more of the individual features mentioned or evident from the text or drawings. All of these different combinations constitute various alternative aspects of the invention.