X-Ray Analysis of Fluidized Beds and Other Multiphase Systems

The attenuation of a beam of X-rays as it passes through a body has been used for over 100 years in diagnostic medicine. A more recent application of the technique has been in the study of industrial-type units such as gas-solid f luidized beds, and this paper reviews this application in the analysis of the hydrodynamic features of bubbling and circulating f luidized beds. The measurement of jet penetration into bubbling beds and the study of bubble slurry columns are also considered. The equipment used and its methods of operation are described and the salient results of the more revealing investigations are summarised both from a qualitative and a quantitative viewpoint. * Torrington Place, London WC1E 7JE, UK ** Chertsey Toad, Sunbury-on-Thames, Middlesex TW16 7LL, UK † Accepted: June 17, 2002 powder at gas velocities between zero and the minimum bubbling point and accurately measuring the expanded bed height; then under knowledge of the mass of powder in the bed, an average bed density could be found for each gas velocity. The value of lB was the thickness of pure aluminium that gave the same beam attenuation as a bed of solids of density rA. The work clearly demonstrated the two-phase nature of f luidized beds but gave little in the way of detailed structure. It did, however, make the first mention of the “bed collapse” technique which has been applied more recently to investigate the effect of operating parameters on dense-phase voidage (see below). Romero and Smith [2] used f lash X-ray radiography to study the internal structure of f luidized beds and obtained data not only on bed density distribution but also on the shape, size and velocity of gas bubbles in an air-f luidized bed of sand. Their unit consisted of two f lash tubes, one operating at 300 kV and the other at 600 kV, each producing a square pulse of energy at 1,000 amps for a duration of about 0.2 μs. The tubes were mounted on opposite walls of a lead-lined room which housed a 7.6-cm square plexiglass column containing a f luidized bed of sand. After passing through the bed, the X-ray beam was recorded on 20 25-cm sheets of film, the 300-kV tube illuminating the lower part of the bed and the 600-kV tube the upper part. To measure bubble velocity, the two units were fired in sequence a few tenths of a second apart. To determine bed density, Romero and Smith measured the darkness of their films using an optical densitometer, but in order to compensate for variations in X-ray operating voltages and variations in the quality and processing of the films, each photograph was compared with a standard consisting of sand-filled wedges made from plexiglass. These were mounted on the side of the f luidization column so that an X-ray photograph of them was obtained each time the X-ray was fired. The wedge gave a film darkness gradation that could be related to sand thickness, and by comparing film darkness in the wedge with darkness in the bed, the local bed density could be obtained. Rowe and Everett [3,4,5] described the design and operation of a f luidized bed X-ray system in which the attenuated beam on passing through the bed was registered on a 0.22-m diameter Mullard image intensifier. The X-ray tube used was a Machlett Super Dynamax 125 with a rotating anode and a fine or broad focal spot of 1 mm or 2 mm diameter. It had a rating of 200 mA at 120 kVp for 10 2 s on fine focus or 500 mA at the same voltage and duration on broad focus. The image intensifier augmented the beam intensity by a factor of 3 103 and was filmed by a 70mm cine camera at a rate of 6 or 8 frames/s, the X-ray beam being synchronised with the camera and pulsed for 2.5 ms. A series of rectangular section f luidized beds were studied with thicknesses from 1.4 cm to 29.5 cm. Bubble sizes, their number and velocity were measured at various heights up to 70 cm in beds of alumina, carbon, quartz, glass ballotini and crushed glass powder f luidized by ambient air; Fig. 1 shows a typical image. Fig. 2 shows a sketch of the X-ray imaging technique. 134 KONA No.20 (2002) Fig. 1 X-ray image of a freely bubbling f luidized bed [3] Fig. 2 X-ray imaging technique Image intensifier field of view Internal f low patterns Gas bubbles Image intensifier CCD video camera HT Generator X-ray source One result to come from these three seminal papers related to the effect of the bed thickness on the minimum f luidization velocity, Umf. It was found that Umf remained constant for beds thicker than 15 cm but decreased by up to 50% as the bed size decreased to the smallest dimension of 1.4 cm. A similar system to that of Rowe and Everett [3,4,5] was used by Rowe et al. [6] to make detailed measurements of the emulsion-phase voidage in f luidized beds and its variation with the particle size distribution of the bed material. In this study the cine camera was operated at 50 frames/s. As described above for Romero and Smith [4], a calibrating wedge filled with bed material was used to quantify the observations. By comparing the optical density of a region on the film with that at a given level of the wedge of known dimensions, the powder voidage in that region could be determined from: ee 1 (1 ew)(DW/DB) (4) where ee and ew are the powder voidages in the emulsion phase and the wedge, respectively, DW is the wedge thickness with the same optical density as the bed, and DB is the bed thickness. An example of such a measurement is shown in Fig. 3. A series of powders with different contents of “fines” was studied, fines being defined as the weight percentage of particles with sieve sizes less than 45 μm. The experimental results are summarised in Figs. 4 and 5 and consisted of measurements of the dense-phase voidage in bubble-free regions of the bed and measurements of the bed height. No variation with height of the dense-phase voidage was observed. Settled-bed voidages refer to the tapped, unf luidized bed. Theoretical considerations by Rowe et al. [6] led to the following expression for the emulsion-phase gas KONA No.20 (2002) 135 Fig. 4 Variation of f luidized bed height with superficial gas velocity [6]


Introduction
Non-invasive experimental techniques are invaluable for providing a detailed insight into the flow patterns and general hydrodynamic characteristics of multiphase systems.One such technique involves the use of X-rays and is based on the same principles as are used in diagnostic medicine to record the structures of body parts that are opaque to visible light.The first recorded use of X-rays to study gas-solid f luidized beds dates from the mid-1950s, and over the subsequent fifty or so years, the technique has been applied to a wide range of two-and three-phase f luidized systems.The technique has enabled gas and solids f low patterns to be observed and quantified and has led to the improved design of internal structures such as gas distributors and heat transfer surfaces in industrial units.

Bubbling fluidized beds
Grohse [1] first reported the use of X-rays to determine the instantaneous and time-averaged bulk density of a finely divided powder fluidized at a range of gas velocities and with three different designs of gas distributor, namely a perforated plate, a mesh screen and a porous plate.The X-ray source consisted of standard components of an X-ray diffraction unit with a tungsten target X-ray tube.The detection equipment consisted of a modification of the rate meter in a standard X-ray detector fed by a scintillation probe.The rate meter response was logarithmic over most of its range of sensitivity and in accordance with Eq. (1) below, the reading varied approximately linearly with the local bed density being measured.The f luidized bed material was silicon powder with a mean particle size of about 50 µm and was contained in a cylindrical borosilicate glass column of diameter 8.5 cm.The results were analysed in terms of the Beer-Lambert law which for the attenuation of essentially parallel, monochromatic X-rays may be written: where m m is the mass attenuation coefficient of the bed material, r is its density and l is the bed thickness.In practice, polychromatic radiation is normally employed but Eq. ( 1) is generally considered an adequate approximation.The term m m rl is called the optical density of the absorbing material and for two different materials A and B producing identical X-ray attenuation, the two optical densities are equal: then the bulk density of A may be written: powder at gas velocities between zero and the minimum bubbling point and accurately measuring the expanded bed height; then under knowledge of the mass of powder in the bed, an average bed density could be found for each gas velocity.The value of l B was the thickness of pure aluminium that gave the same beam attenuation as a bed of solids of density r A .The work clearly demonstrated the two-phase nature of f luidized beds but gave little in the way of detailed structure.It did, however, make the first mention of the "bed collapse" technique which has been applied more recently to investigate the effect of operating parameters on dense-phase voidage (see below).Romero and Smith [2] used f lash X-ray radiography to study the internal structure of f luidized beds and obtained data not only on bed density distribution but also on the shape, size and velocity of gas bubbles in an air-f luidized bed of sand.Their unit consisted of two f lash tubes, one operating at 300 kV and the other at 600 kV, each producing a square pulse of energy at 1,000 amps for a duration of about 0.2 µs.The tubes were mounted on opposite walls of a lead-lined room which housed a 7.6-cm square plexiglass column containing a f luidized bed of sand.After passing through the bed, the X-ray beam was recorded on 20҂25-cm sheets of film, the 300-kV tube illuminating the lower part of the bed and the 600-kV tube the upper part.To measure bubble velocity, the two units were fired in sequence a few tenths of a second apart.
To determine bed density, Romero and Smith measured the darkness of their films using an optical densitometer, but in order to compensate for variations in X-ray operating voltages and variations in the quality and processing of the films, each photograph was compared with a standard consisting of sand-filled wedges made from plexiglass.These were mounted on the side of the fluidization column so that an X-ray photograph of them was obtained each time the X-ray was fired.The wedge gave a film darkness gradation that could be related to sand thickness, and by comparing film darkness in the wedge with darkness in the bed, the local bed density could be obtained.
Rowe and Everett [3,4,5] described the design and operation of a f luidized bed X-ray system in which the attenuated beam on passing through the bed was registered on a 0.22-m diameter Mullard image intensifier.The X-ray tube used was a Machlett Super Dynamax 125 with a rotating anode and a fine or broad focal spot of 1 mm or 2 mm diameter.It had a rating of 200 mA at 120 kV p for 10 Ҁ2 s on fine focus or 500 mA at the same voltage and duration on broad focus.The image intensifier augmented the beam intensity by a factor of 3҂10 3 and was filmed by a 70mm cine camera at a rate of 6 or 8 frames/s, the X-ray beam being synchronised with the camera and pulsed for 2.5 ms.A series of rectangular section fluidized beds were studied with thicknesses from 1.4 cm to 29.5 cm.Bubble sizes, their number and velocity were measured at various heights up to 70 cm in beds of alumina, carbon, quartz, glass ballotini and crushed glass powder fluidized by ambient air; Fig. 1 shows a typical image.Fig. 2 shows a sketch of the X-ray imaging technique.One result to come from these three seminal papers related to the effect of the bed thickness on the minimum f luidization velocity, U mf .It was found that U mf remained constant for beds thicker than 15 cm but decreased by up to 50% as the bed size decreased to the smallest dimension of 1.4 cm.
A similar system to that of Rowe and Everett [3,4,5] was used by Rowe et al. [6] to make detailed measurements of the emulsion-phase voidage in fluidized beds and its variation with the particle size distribution of the bed material.In this study the cine camera was operated at 50 frames/s.As described above for Romero and Smith [4], a calibrating wedge filled with bed material was used to quantify the observations.By comparing the optical density of a region on the film with that at a given level of the wedge of known dimensions, the powder voidage in that region could be determined from: where e e and e w are the powder voidages in the emulsion phase and the wedge, respectively, DW is the wedge thickness with the same optical density as the bed, and DB is the bed thickness.An example of such a measurement is shown in Fig. 3.
A series of powders with different contents of "fines" was studied, fines being defined as the weight percentage of particles with sieve sizes less than 45 µm.The experimental results are summarised in Figs. 4 and 5 and consisted of measurements of the dense-phase voidage in bubble-free regions of the bed and measurements of the bed height.No variation with height of the dense-phase voidage was observed.Settled-bed voidages refer to the tapped, unf luidized bed.
Theoretical considerations by Rowe et al. [6] led to the following expression for the emulsion-phase gas Fig. 5 Variation of emulsion-phase voidage with superficial gas velocity [6] f low rate: Q e /A҃U ge (3Ҁ4f b )e e /3 (5) where all the quantities on the right-hand side of the equation could be calculated from the experimental results.Figure 6 shows the results of plotting Eq. ( 5), and it is clear that the emulsion gas flow rate is very much greater than the flow corresponding to that at minimum f luidization of the powder containing 27.6% fines, 0.207 cm/s, and that it increases with overall gas velocity and fines content.This is favourable from the point of view of fluidized bed reactor operation, a conclusion that was verified in a subsequent study by Yates and Newton [7].
The X-ray technique was used by Hoffmann and Yates [8] to study the effect of pressure on f luidized beds of powders in Groups A and B of the Geldart [9] classification.They found in the case of Group A powders that while the minimum fluidization velocity, U mf , was unaffected, the region of bubble-free expansion between U mf and U mb , the minimum bubbling velocity, increased with increasing pressure.This observation is in accordance with the theory of f luidbed stability developed by Foscolo and Gibilaro and recently reviewed by Gibilaro [10].For Group B powders, Hoffmann and Yates found that mean bubble diameters increased slightly up to 16 bar and decreased thereafter up to 60 bar pressure.In addition, the bubble velocity coefficient k in: was found to decrease up to 20 bar pressure but then to increase quite markedly above 20 bar up to the highest pressures at which observations were made.This latter increase coupled with the observed decrease in bubble size meant that although bubbles were becoming smaller, their rise velocities were increasing, the reverse of the tendency observed at ambient pressure.These results were subsequently corroborated by Olowson and Almstedt [11] using a combined capacitance and pressure probe to measure bubble characteristics in a bed of Group B powder.Other X-ray studies of the effects of system pressure on f luidization were reported by Rowe et al. [12] and Barreto et al. [13].
X-ray images of f luidized beds may be enhanced so as to investigate solids volume fractions in the regions close to the boundaries of bubbles.In one such study [14], the X-ray source was pulsed for a period of 1 ms and the pulses synchronised with a video recorder at an equivalent speed of 25 frames/s.Both the radiation source and the detector could be moved vertically relative to the fluidized bed, thus allowing different positions to be examined.The recorded images were transferred off-line for processing and analysis using a PC.The video signal was relayed to a framestore board (Imaging Technology VP 1100) via a timebase corrector, and the images processed using Bioscan Optimas software operating in Windows 3.0.
As discussed above, when X-rays pass through a solid material they are attenuated by processes of absorption, ref lection and scattering, and the extent of the attenuation is a function of the chemical nature of the solid and of the quantity of material in the path of the beam.The Beer-Lambert law (Eq.1) may be written: In many cases, the second order and higher terms in the exponential may be neglected and Eq. ( 7) may be used to an acceptable degree of accuracy.
The attenuation of X-rays by ambient air is negligible and so for an air-f luidized bed we can express the bulk density r in terms of the solids fraction of the bed through which the beam passes: where r s is the solids density and e is the voidage of the bed material.
In order to relate the measured intensity of the Variation of emulsion-phase gas flow with superficial gas velocity [6] transmitted beam to the voidage of the material through which it passes, it is necessary to calibrate the system according to Eq. ( 8) and this may be done as before using a wedge filled with the powder to be studied.The wedge is set up vertically in the X-ray beam and the transmitted intensity measured on a range of grey scales from 0 to 225 as a function of the wedge thickness l w .The calibration for the Group B alumina powder used in this study [14] gave an explicit relationship between bed voidage, transmitted intensity and path length l of: The recorded X-ray images of single bubbles were computer-enhanced to show coloured zones of varying voidage; such an image is shown in Fig. 7, where the regions of expanded voidage surrounding the bubble, the so-called "shell" region, are clearly visible.The black area is the emulsion phase with an average voidage of 0.445, the green with 0.495, the blue with 0.526, the cyan with 0.627 and the red area is the bubble itself with a voidage of 1.0.
Two points must be emphasised about this image.Firstly, the computer is programmed to calculate average values of voidage between any predetermined range of intensity values and to colour-code that range; there is in fact a continuous gradation in voidage between one zone and the next.Secondly, the image is a representation in two dimensions of a truly three-dimensional object, in other words a silhouette.This means that the voidage indicated is an average value across a chord of the more or less spherical region of the bed around the bubble.The deconvolu-tion of these averages to give point values of porosity will now be considered.
The X-ray intensity measured in the x-direction across a section of such a sphere may be expressed as: where b is a calibration factor, ε(x,z) is the space-averaged voidage across a chord within the shell, and ε mf is the voidage of the background emulsion phase.Now: z҃(r 2 Ҁx 2 ) 1/2 (11) and hence: dz҃ (12) I(x) may be expressed in terms of the variation of voidage in the radial direction, ε(r): This is the Abel transform of [ε(r)Ҁε mf ] [15], and a solution for the inverse transform of the function I(x) will enable the radial variation in voidage to be found.The Abel transform is useful for converting threedimensional images into two dimensions and vice versa, and is based on the assumption that the threedimensional object being considered is axially symmetrical, e.g., a perfect sphere.Clearly, bubbles in f luidized beds are not perfect spheres but they can be considered as having to a first approximation a sufficient degree of axial symmetry for the transform to be applicable, i.e. it can be assumed that the voidage around a bubble is a function of the radius of the sphere centred on the bubble.The transform may be reduced to a convolution integral and solved numerically to obtain ε(r) [15].The method was applied to a single bubble in the Group B powder observed 30 cm above the distributor with the results as shown in Fig. 8, and it is clear that no very sharply defined boundary exists between the shell and the emulsion phase.
Lettieri et al. [16] used the X-ray technique to study changes in the behaviour of the dense phase of fluidized beds with increasing temperature.The bed container was constructed from Inconel with a wall thickness of 3 mm, and was capable of being heated to 650°C.Fluidized beds of a range of materials were subjected to the "bed collapse test" in which the fluidizing gas supply was abruptly cut off and the resulting collapse of the bed surface observed using X-rays.Two solenoid valves were fitted to the equipment, one to cut the gas supply to the bed, the other to vent the Fig. 7 X-ray image of a bubble in a fluidized bed of a Group B powder [14] gas trapped in the windbox below the gas distributor.From these observations, a number of quantities may be derived for Group A powders, including the emulsion-phase voidage, ε e , and the "standardised collapse time" (SCT s/m) which gives a measure of the aeration capacity of the powder [17].The SCT is defined as follows: where t 0 and t c are the initial cut-off time and the time for the bed to reach total collapse, respectively, and H s is the settled bed height.The longer the SCT, the higher the aeration capacity of the powder and hence the larger the void fraction of the emulsion phase.Lettieri et al. [16] showed that for powders such as f luidized cracking catalyst (FCC) that are free from inter-particle forces, the SCT and hence the emulsionphase voidage increase markedly with increasing temperature, whereas powders that were coated so as to induce inter-particle adhesion showed the opposite effect.Table 1 summarises these results for three FCC powders and one silica powder coated with 10 wt % of potassium acetate.
Bed collapse experiments using X-rays to follow the process were also described by Chen and Weinstein [18].The X-ray image of the bed was projected onto a 0.23-m diameter image intensifier screen, the visible light from which was focused onto another screen equipped with 48 phototransistors.These were positioned such that three rows of 16 transistors scanned the projected image of the bed across three elevations at 1 cm intervals.The system is described in detail by Feindt [19].In this work, which also employed cracking catalyst, a single solenoid valve was used to shut off the gas supply to the bed.The bed collapse data were used to calculate the solids stress modulus and the drag coefficient for the catalyst powder in the solid fraction range between minimum fluidization and loose packing.The sedimentation wave velocity for the bed material was 0.027 m/s, a figure in good agreement with the SCT measurements of Lettieri et al. [16].

Jet penetration
When gas first enters a f luidized bed from a submerged orifice, it does so either in the form of discrete bubbles or as a flame-like jet which decays at some point above the grid into a bubble stream.Whether one or the other form seems to depend on the properties of the bed material and the size of the inlet nozzle.Rowe et al. [20] observed the point of entry of gas into a bed of a Group B powder using Xrays and saw only bubbles forming with a welldefined frequency.
Chen and Weinstein [21] carried out an X-ray study of a horizontal jet into a 15҂38-cm cross-section f luidized bed of cracking catalyst.Instantaneous solids volume fractions averaged along 15-cm chords were measured for two jet diameters and three initial jet velocities, and it was shown that there are three discernible regions in the area of the bed influenced by the jet: a coherent void; bubble trains; and a surrounding compaction zone.The jet penetration length was found to agree well with the correlations developed by a number of earlier workers.Weinstein et al. [22] have more recently studied the sudden injection of a horizontal gas jet into a bed at just above the min-  Yates and Cheesman [23] and Yates [24] measured jet penetrations, L max , in three-dimensional beds of two Group B powders at pressures of up to 20 bar at ambient temperature, and at temperatures of up to 800°C at ambient pressure.The results were correlated with: U cf being the velocity of complete f luidization of the powder.
Newton et al. [25] reported results from an X-ray study of upshot, downshot, horizontal and 30°-angled upshot jet penetration lengths from multiple orifice grid plates and nozzle spargers.Bed materials were Geldart Group A. The study concluded that none of the correlations for predicting vertical and horizontal jet penetrations adequately accounted for the measured values, but that the correlation of Zenz [26] was well suited for downshot jet penetrations over the range of conditions studied.

Circulating fluidized beds
Fluidized beds undergo a number of state transitions as the superficial gas f low rate increases.Thus a bubbling bed exists at gas velocities somewhat in excess of the minimum fluidization f low rate but at higher f low rates, bubbles become unstable and the bed passes into the "turbulent" regime.At sufficiently high f low rates, the bed particles become entrained and carried out of the containing vessel, and in order to maintain a constant solids inventory, the entrained material must be captured in a cyclone and returned to the bed normally at a point near its base.Such a system is called a "circulating" f luidized bed, and industrial interest in this type of system goes back to the earliest days of f luidization when they were employed in the catalytic cracking process [27].Serious scientific study of the area dates from the work of the group at City College, New York, in the mid-1970s [28], although Kehoe and Davidson [29] had identified the transition from bubbling to turbulent f luidization somewhat earlier.One of the earliest reports of the application of the X-ray technique to study gas and solids f low in the riser section (the so-called "fast" bed) of a circulating bed was that of Weinstein et al. [30].Their X-ray tubehead and the image recording devices (plate film and TV camera) were mounted on opposite sides of a carriage which could be moved up and down a 3-m long centre section of the fluidized bed.Data were obtained on the radial distribution of void fraction in the lower dense-phase region of the fast bed and demonstrated its dilute core-dense annulus structure.Variations in radial voidage were studied in response to three types of f low variation, (i) increasing superficial gas velocity from 1.1 m/s (a turbulent bed) to 5 m/s with varying solids circulation f lux, (ii) increasing gas velocity with a constant solids f lux, and (iii) holding the gas velocity constant at 3.1 m/s and varying the solids f lux between 50 and 121 kgm Ҁ2 s Ҁ1 .The results are shown in Figs.9-11, from which it was concluded that (a) at a constant gas velocity changes in solids flux have a small effect on radial voidage, (b) decreasing solids f lux moves the point of inf lection between the lower fast bed and the upper dilute-flow region in a downward direction until at a sufficiently low value of the flux, the dense-phase region disappears.
Subsequent X-ray studies by the City College group have investigated internal scales of turbulence in high-velocity beds [31], the effect of gas nozzle configuration on the acceleration of solids and the distribution of solids and gases in a riser [32][33][34], and the f low characteristics of downflowing high-velocity f luidized beds [35,36].

Slurr y bubble columns
By virtue of their excellent heat transfer characteristics, three-phase reactors Ҁ in which a gas is sparged into a solid-liquid slurry Ҁ are favoured for processes involving highly exothermic reactions such as the Fischer-Tropsch process [37,38].Smith et al. [39] used an X-ray system to study the hydrodynamics of a system in which an inert gas (nitrogen or helium) was passed into a slurry of zirconia particles in water.The solids loading was varied in the range 10-30%, and three 2-m long "reactors" with diameters of 50, 127 and 178 mm were used, each being fitted with a grade C sintered brass distributor plate.For comparison, a holed plate distributor was designed for the 127-mm column and consisted of 37҂3-mm diameter holes arranged on a triangular pitch.System pressures were up to 8 bar and gas velocities up to 0.09 m/s.Gas bubbles were generally small, of the order of a few mm in diameter, and since the gases used were insoluble, the mean bubble diameter could be estimated from the extent of expansion of the slurry column: where n b is the number of bubbles of mean diameter d b , n b being obtained by direct counting from the video image.Figure 12 shows the effect of gas velocity on mean bubble size, and it is clear that system pressure has a significant effect.Increasing the solids loading was found to increase both the hold-up and the mean size of bubbles, the latter increasing from 8 mm to 12.5 mm on increasing the loading from 10% to  [30] 20% in the 127-mm column at 8 bar and 0.04 m/s gas velocity.

Conclusions
The X-ray technique is one of an increasing number of non-invasive methods currently being employed to study multiphase systems of industrial interest [40].
The above examples demonstrate the versatility of the technique for real-time observation and analysis of the many hydrodynamic features of f luidized beds, and it will continue to be used for this purpose.An important application which is currently at an early stage is as a tool to validate theoretical models of f luidized systems based on computational fluid dynamic codes such as Fluent and CFX [41].The high degree of discrimination possible with X-rays makes it ideal for this validation process, and much progress in this direction can be expected in the near future.

Fig. 8
Fig.8 Scan of X-ray intensity across a horizontal plane through the equator of a single bubble in a Group B powder[14]

Table 1
Experimental SCT for FCC powders and a silica powder coated with KOAc imum f luidization conditions, but no quantitative correlation from the observations was given.