84
19 (2002) 84 ± 95
Estimating Average Particle Size by Focused Beam Reflectance
Measurement (FBRM)
Alex R. Heath*, Phillip D. Fawell**, Parisa A. Bahri*, Jean D. Swift**
(Received: 5 April 2001; resubmitted: 11 January 2002)
Abstract
The Lasentec focused beam reflectance measurement
(FBRM) probe provides in situ particle characterisa-
tion over a wide range of suspension concentrations.
This is a significant advantage over conventional
instruments that require sampling and dilution. How-
ever, FBRM gives a chord distribution, rather than a
conventional diameter distribution. Both theoretical
and empirical methods for converting from chord to
diameter data are available, but the empirical method
was found to be more successful.
Keywords: FBRM, on-line, particle size analysis
1 Introduction
Various techniques are available for determining particle
size [1 ± 5]. However, few can be applied in situ/in-line, or
over a wide range of solid concentrations. Ultrasonic
techniques have been developed [6], but require lengthy
measurement times, making them unsuitable for dynam-
ic systems. Standard laboratory laser diffraction instru-
ments have also been modified [7 ± 10] to allow on-line
sizing of a side stream, but are inconvenient for field or
plant use and typically require some degree of dilution.
Alternatively, in-process measurements can be made by
focused beam reflectance measurement (FBRM). The
FBRM probe has a wide range of applications, frequently
in the process environment (Table 1). The instrument
gives a particle chord length distribution, which is a
function of the true particle diameter distribution. For
* A. R. Heath Ph.D student, Dr. P. A. Bahri, A. J. Parker
Cooperative Research Centre for Hydrometallurgy, School of
Engineering, Murdoch University, (Rockingham Campus),
Murdoch, WA 6150 (Australia).
** Dr. P. D. Fawell (to whom correspondence should be ad-
dressed), J. D. Swift, A. J. Parker Cooperative Research Centre
for Hydrometallurgy, CSIRO Minerals, Box 90, Bentley, WA
6982 (Australia).
E-mail: phillip.fawell@csiro.au
¹ WILEY-VCH Verlag GmbH, 69469 Weinheim, Germany, 2002
Part. Part. Syst. Charact.
The FBRM response has been compared to conven-
tional particle sizing techniques (laser diffraction and
electrical sensing zone) for a range of sieved aluminium
or calcite suspensions. The mode average of the square-
weighted chord length was found to be comparable to
other sizing techniques over the range from 50 to
400 mm. The square-weighted FBRM results were
essentially unaffected by changes to the instrument
focal position, the suspension fluid flow velocity, or the
solid fraction in the range 0.1 ± 20 % w/v.
many applications this distinction is not important
because the relevant process variable can be correlated
with some aspect of the chord data, and accurate particle
sizing is not specifically required.
The instrument×s rapid in-line analysis make it attractive
to workers modelling the kinetics of flocculation [29, 35]
or crystallisation [13, 18, 20] by population balance.
However, in dynamic modelling, both accurate sizing
and counting of particles are desirable for the estimation
of model parameters, and the raw chord length data may
not be suited to this.
In this paper both theoretical and empirical methods for
estimating particle diameters from chord data are
reviewed. These methods are then applied to chord
data collected from a range of sieved particles, and the
calculated diameter data compared with conventional
sizing for the same samples. The effect of length-
weighting has been examined, and different averaging
methods (mean, median, and mode) compared. Factors
affecting the FBRM data were also examined, in
particular the influence of the solid volume fraction.
0934-0866/02/0205-0084 $ 17.50+.50/0
Part. Part. Syst. Charact. 19 (2002) 84 ± 95 85

Table 1: Summary of common FBRM process applications.

Application Reference Application Reference
Control [11 ± 16] Grinding [16, 37 ± 41]
Crystallisation/precipitation [13, 15, 17 ± 23] Paper manufacture [42 ± 45]
Estuarine solids [9, 10, 24 ± 27] Petroleum [46 ± 48]
Flocculation [28 ± 40] Pharmaceutical [49, 50]

Several descriptions are given of previous instruments

and data processing regimes [16, 17, 37]. Data processing
of the detector response uses both the signal strength and
slope. Current instrument data processing is adjustable
through the use of either F (fine) or C (coarse)
electronics modules, with the main difference being in
the faster preamplification and greater edge definition
with the former. Although rarely identified in publica-
tions, it can be assumed the F-electronics have been
preferred for most particle studies, due to sharper
resolution and higher sensitivity to fine chords. The C-
electronics is required for studies of aggregated particles,
i.e. for flocculation applications [39].

2.2 Estimating Particle Size from FBRM Data

Fig. 1: Schematic of the FBRM probe (not to scale).
2.2.1 Theoretical Methods
Several papers deal with the transformation of chord to
diameter distributions from first principles [8, 22, 28, 48,
2 Background
52 ± 56]. Most workers assumed that the particles are
2.1 Principle of FBRM
The Lasentec FBRM instrument directs a 780 nm laser
through a lens rotating at 4500 r.p.m., giving a beam of
light highly focused at a point near the instrument×s
window (Figure 1). The position of the focal point is
adjustable, depending on the particle properties [51].
The diameter of the circular path of the laser beam is
8.5 mm, focused to a beam waist of approximately 0.7 
2 mm and a focal depth of approximately 10 mm [37].
However, the exact width of the beam appears to be
dependent on both the focal position and the size and
distance of the particle from the focal position [16, 17].
When the moving laser beam intersects the path of a
particle some of the light is reflected back to the
instrument×s detector. Since the tangential velocity of
the beam is known ( 1.9 m s 1), the duration of the
reflected light pulse is directly proportional to the width
of the particle intersected. The tangential beam velocity
is assumed to be much larger than the particle velocity
relative to the instrument window. However, since it is
unlikely that the laser beam will pass directly across the
centre of the particle (i.e. its diameter), a chord length is Fig. 2: Method of determining the chord distribution expected
measured. from a spherical particle of diameter Dp.
86 Part. Part. Syst. Charact. 19 (2002) 84 ± 95

spherical to simplify the mathematics, and most analysis The probability of a particle being detected is propor-
of non-spherical particles is limited to two dimensional tional to its diameter, introducing a bias. Therefore, for
ellipses [52 ± 56]. Assuming the particles are spherical modelling purposes the FBRM is assumed to measure
simplifies the mathematics, because regardless of the the first diameter moment of the chord distribution [48,
particle×s orientation it always presents a circular profile 56]. Calculating a chord distribution from a given
(Figure 2). spherical particle distribution is relatively straightfor-
This assumption may be valid for suspensions of ward, however the reverse algorithm is more difficult,
spherical latex or glass beads, or emulsions of immiscible usually requiring a numerical solution in the form:
liquids. Assuming the particles are spherical may also be M
X
2
reasonable for suspensions of particles with an aspect min

ˆ Ni Ni : …1†
Np
ratio (ratio major:minor axis) approaching 1 : 1. As the iˆ1

aspect ratio increases, and the particles or crystals tend

M  12  12
towards needles/plates, the chord distribution broadens  
X
2 2 2 2
N ˆN
i p D p C i;1 D p C
i;u : …2†
[22, 28]. iˆ1
Most descriptions assume that the focused laser beam
has no width, and that light is reflected immediately the The resulting calculated diameter distribution is a
beam reaches the edge of the particle. However, Hobbel number distribution, and is converted to the conven-
et al. [8] proposed a parameter (k) to account for the tional particle volume distribution by applying a length-
™arch of rejected reflection∫. Sparks and Dobbs [17] cube weighting (as discussed in Section 4.1).
considered the physics of the laser/lens interaction and
suggest that the beam may be broadened in some
2.2.2 Empirical Methods
conditions. Spreading of the beam away from the optimal
focal point may also affect the chord measurement [16], FBRM results have been compared with alternative
and beam spreading is likely to be a function of solids particle sizing techniques (Table 2) for a variety of
loading, fines, solution and particle surface character- particle sizes and materials. Early Lasentec FBRM
istics. software included a spherical-equivalent size algorithm

Table 2: FBRM comparisons with other sizing techniques.

Reference Particle material Alternative sizing method (size) FBRM size or size relationship (mm)
Hobbel et al. [8] Glass beads Helos laser diffraction 170
(66 mm)
Sparks and Duke standards, carbon, sand, Microtrac laser diffraction, certified Spherical equivalent,
Dobbs [17] alumina standards (0.8 ± 410 mm) FBRM ˆ 52.35 ‡ 1.036Duke
over the range 10 ± 162 mm
Murphy et al. Ground silica BET, SEM, Micromeritics sedigraph Oversized, could not be calibrated in
[33] (2.1 mm) this range
Monnier et al. Glass beads, latex, micronized phar- Cilas laser diffraction, Coulter Undersizes > 100 mm,
[51] maceutical product. Various support- Counter -electrical sensing zone, oversizes < 100 mm
ing fluids microscopy (4.1 ± 396 mm)
Dijkstra et al. Sugar Sieved (30 ± 250 mm) Sieve ˆ 2.04  10 3( FBRM )2.28
[15]
El-Hamouz and Oil/water emulsions Malvern laser diffraction Sauter mean (d32 ˆ 130 ± 210 mm),
Stewart [47] (d32 ˆ 60 ± 350 mm) undersizes > 180 mm,
oversizes < 180 mm
Law et al. [27] Pollen, glass and latex Duke stan- Malvern laser diffraction, Coulter FBRM Par-Tec spherical equivalent,
dards, fractionally settled sediment, Counter, certified standards undersizes > 500 mm, oversizes
sieved sand (6.5 ± 766 mm) < 150 mm
Phillips and Sieved river sediment Coulter laser diffraction FBRM Par-Tec spherical equivalent
Walling [25] (1 ± 355 mm) Coulter d50 ˆ 19.24 ‡ 1.064FBRM
Tadayyon and Sieved ion-exchange resin Sieve (212 ± 850 mm) FBRM undersized, better with cube-
Rohani [56] weighting
Daymo et al. [57] Bentonite, graphite, gibbsite, kaolin, Sieve Horiba sedigraph (0.3 ± 663 mm) Kaolin (5% w/v) 53 mm by FBRM
mica, silica, plastic beads (cube-weighted), 57 mm by sedigraph.
Weak correlation for other samples
Alfano et al. [40] Aggregated silica Malvern laser diffraction FBRM ˆ 10.09 ‡ 0.586 Malvern
(20 ± 180 mm)
Part. Part. Syst. Charact. 19 (2002) 84 ± 95 87

[14, 17, 25, 27, 29, 32]. Comparisons between the resulting
average size and alternative sizing methods showed that
the FBRM oversized small particles (< 150 mm [27],
< 300 mm [25]) and undersized larger ones (> 500 mm [25]).
Current software computes various chord statistics:
median (d50), mean and mode with various chord
weightings. Daymo et al. [57] found good agreement
between the cube-weighted mean size and sieve analysis
of kaolin suspensions with a mean size of around 55 mm.
Alfano et al. [40] show a good correlation between laser
diffraction volume-based mean size and unweighted
FBRM chord lengths for suspensions of aggregated silica
20 ± 200 mm. However, the unweighted FBRM results
again underestimated the size of the larger particles. El-
Hamouz and Stewart [47] compared FBRM mean
square-weighted chord lengths and laser diffraction
mean sizes for droplets of oil/water emulsions and found
the FBRM over-estimated smaller (< 180 mm) droplets,
but underestimated larger (> 180 mm) droplets.

3 Experimental

3.1 Samples
Most samples were either calcite (Omya Southern,
Commercial Minerals) or aluminium (Valimet, Stockton
California). SEM micrographs were obtained using a
JEOL JSM-5800LV, with the calcite particles gold-
coated for imaging purposes (Figure 3). The calcite
particles were blocky in shape, while the aluminium
particles were more spheroidal. In general, most par-
ticles had a low aspect ratio. A single suspension of
mono-sized latex spheres (Dynaspheres, Duke Scientif-
ic) with a specified size of 19.9 mm was also measured.
The particles were separated into narrow size fractions by
screening or cyclosizing. Dry sieving used conventional
Endecotts screens (38, 45, 53, 63, 75, 106, 125, 150, 180, 212,
250, 300, 425, 500, 600, 710, 850, 1000, 1180 mm) in a Rotap
RX ±29 ±10 mechanical shaker, while smaller size fractions
obtained by cyclosizing with a Warman M-4 cyclosizer.
The solids were weighed to  0.1 mg and made up to
concentrated aqueous stock suspensions in a volumetric
flask. A series of more dilute suspensions were obtained
by dilution of the stock suspension from a stirred tank,
baffled to ensure uniform solids distribution. The dry
sieved calcite samples were found to contain some fines
(possibly from surface abrasion or aggregation) and
where possible were washed and decanted. Suspensions
of aluminium particles were reactive, producing a gas
(presumably hydrogen), and were kept on ice to reduce
the reaction rate. Solids concentrations are expressed in
terms of solid weight as a percentage of suspension
volume (% w/v).
Fig. 3: SEM images of unsieved (a) calcite and (b) aluminium
particles (white bar represents 200 mm).
3.2 Equipment
FBRM measurements were made with a M500 field unit
(Lasentec, Redmond, Washington, USA). The probe has
a 12 mm diameter flat sapphire window set in a 25 mm
diameter stainless steel casing. As this study was
confined to non ± aggregated systems, only the F (fine)
electronics data are presented. Unless otherwise stated,
the focal point was set to 0 mm (external face of the
window) as calibrated by the manufacturer. This position
was also confirmed experimentally by placing a film of
marking pen ink on the window and adjusting for
maximum counts.
Suspensions of various particle size and solid fraction
were presented to the FBRM probe using a dedicated
stand (Lasentec¾). The impeller (four angled blades,
diameter 50 mm) was set to 650 r.p.m. unless otherwise
stated. The probe acts as a baffle in the 500 mL tall-form
beaker (internal diameter 75 mm), and is positioned with
the probe window 3 mm above the impeller tip to ensure
optimal sample presentation. A 30 s period of stirring
88 Part. Part. Syst. Charact. 19 (2002) 84 ± 95

was found to be sufficient to disperse the solid. The

measurements were then taken rapidly, as some fines
were released from calcite particles on extended stirring.
The use of dispersants can be advantageous in some
systems where aggregation is a factor [51], but were
found to give no significant improvement for the samples
studied here.
The FBRM control software (version 6) allows the
acquired chord length data to be discretised into a series
of user-defined intervals (channels). Data was collected
using 90 log ± channels over the range 1 to 1000 mm unless
otherwise stated. While the chord length distributions
should technically be viewed as histograms, for ease of
comparison they are presented as line graphs. A meas- Fig. 4: Unweighted chord length distributions for 10% w/v
urement duration of 10 seconds was used, with each sieved aluminium fractions produced by sieving.
distribution the average of five consecutive measure-
ments. For simplicity, the counts axis on all distributions
represents the number of counts per second.
Fluid flow past the probe window generally ensures that
the window is kept clean. However, fine particles are
prone to adhere to the window, and can increase the
count rate and affect the measured size. Window
cleaning was therefore performed regularly (  5 min.
measurement time) to give a total background count of
less than 50 counts per second in air.
Laser diffraction and electrical sensing zone instruments
(Malvern Mastersizer MS10 and Coulter Multisizer II,
respectively) were used as alternative sizing techniques,
with concentrated samples typically diluted to less than Fig. 5: Effect of applying various length-weightings (moments)
2% w/v for analysis. to the FBRM chord distribution for aluminium particles (sieved
45 ± 53 mm). Laser diffraction volume-weighted distribution also
shown.

4 Results and Discussion

4.1 Effect of Applying Weightings to FBRM Chord
Distributions
The raw (unweighted) chord length distributions meas-
ured by FBRM for a series of sieved aluminium fractions
at a constant solid concentration are shown in Figure 4.
As expected, the larger particles gave longer chord
lengths, and a reduced count rate due to the decreased
number of particles.
Figure 5 shows the effect of applying length weightings
(moments) to the chord distribution given by FBRM, i.e.:
Ni;n ˆ Ni;o Cni;A : …3†
It can be seen from Figure 5 that the square-weighted
(n ˆ 2) chord distribution gave the best agreement with
the conventional laser diffraction distribution, although
the FBRM distributions remained broader. This is in
general agreement with the theoretical analysis (Section
2.2.1) which suggests that FBRM measures the first
diameter weighting (moment) of the chord distribution.
On this basis, applying the square-weighting is then
effectively a cube (volume) weighting, comparable to the
volume-based distribution from laser diffraction.
4.2 Sample Presentation Effects
4.2.1 Focal Position
The position of the focal point of the laser is manually
adjustable in relation to the instrument window. This has
the effect of moving the position of the viewing region
further into the suspension, or, as negative settings
indicate, behind the window/suspension interface (Fig-
ure 1). The effect of focal position on an early version of
FBRM was examined by Monnier et al. [51] and Law
et al. [27], who varied the positioning into the solution by
up to 6 or 4 mm, respectively. Law et al. [27] made use of
a spherical equivalent diameter algorithm, while Mon-
nier et al. [51] monitored total counts and unweighted
mean chord lengths.
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
Both groups observed substantial changes with focal
position, suggesting that setting the position further into
solution (in the range 0.8 to 2 mm) gave better results for
larger particles. However, they did not report settings
closer to the probe window. While large focal positions
may give some improvement for very dilute suspensions,
in this work optimal performance was achieved when
focused close to the window (Figure 6). For large
particles it is suggested [58] that the focal position be
set to approximately 20 mm inside the window.
Figure 6 shows the effect of varying the focal position
between 100 and ‡ 100 mm from the window. As
expected, setting the position into the suspension
increases the number of longer chords, but by 50 mm
there was a significant loss of total counts. At 100 mm into
the suspension the counts dropped below a usable level.
This is most likely a consequence of the high solids
concentration, with such focal positions only applicable
in dilute solutions.
Larger particles may have difficulty entering the viewing
region as the focal position is brought back towards the
window, leading to an increased sensitivity of the
unweighted distribution to fine chords, which peak at
between 20 and 10 mm (Figure 6a). In contrast, the
square-weighted chord distribution (Figure 6b) remains
insensitive to the presence of the fine chords, and while
the peak in this distribution varied in intensity as the
Fig. 6: Effect of laser focus position on (a) unweighted and (b)
square-weighted chord length distributions for 20% w/v calcite,
(sieved 106 ± 125 mm, 500 r.p.m. impeller speed).
89
focal position was changed, its position did not shift. On
the basis of these results, a focal position setting of 0 mm
was chosen for all subsequent measurements.
4.2.2 Suspension Agitation
Accurate sizing by FBRM requires a representative
sample of particles to pass through the measurement
zone near the window, consequently the effectiveness of
mixing becomes an issue for suspensions containing
larger particles. The effect of the suspension agitation on
the measured size has been previously investigated, with
contradictory results reported. No significant effect of
stirring speed on the indicated size in the range 500 ±
1000 r.p.m. was observed by Law et al. [27]. Monnier
et al. [51] found that increasing the impeller speed from
500 ± 1000 r.p.m. dramatically increased both the meas-
ured size and total counts of glass beads, although
transparent particles are generally reported to be poorly
sized by FBRM [17]. Daymo et al. [57] reported that
increasing the flow velocity of a kaolin suspension from
1.3 to 2.4 m s 1 in a 76 mm ID pipe reduced the indicated
size by 13%. However, these changes may have been due
to particle setting at low flow rates, or particle breakage
at high flow rates.
The tangential velocity of the focused laser beam is
approximately 1.9 m s 1, and is assumed (see Section
2.2.1) to be far greater than the particle velocity with
respect to the instrument window. At approximately 750
r.p.m. the tip velocity of the 50 mm diameter pitched
blade impeller reaches 2 m s 1. However, since the
viewing region is adjacent to the window, the flow
velocities in the viscous sub-layer are likely to be
considerably lower than for the bulk of the fluid flow.
In practice, changing the impeller speed had little effect
(Figure 7) on the observed square ± weighted chord
distribution. Larger particles (> 500 mm) tended to
Fig. 7: Effect of impeller speed on square-weighted chord length
distributions for 20% w/v calcite particles (sieved 106 ± 125 mm).
90
visibly settle at low stirring speeds (< 500 r.p.m.), leading
to a reduction in the average measured size.
4.3 Comparison of Average Particle Size by Various
Techniques
Particle size distributions can be weighted and averaged
in a variety of ways [5, 59 ± 61]. Figure 8 plots various
length-weighted median FBRM average sizes with d50
values from laser diffraction for a range of sieved or
cyclosized fractions. Calculated geometric average val-
ues from sieving are also included for comparison. Laser
diffraction data is presented on the x-axis because it was
the only alternative technique that covered the entire
size range, but this is not meant to imply that the
technique provides a more accurate particle size.
While all the relationships shown in Figure 8 exhibited
good linearity over the size range studied, applying
length-weightings served to improve the measured slope,
increasing from 0.530 for the median unweighted chord
length to 0.898 for the median length square-weighted
chord length. The latter slope was comparable to that
achieved for sieve analysis (0.885).
For the calcite particles (which covered a larger range of
sizes) the length square-weighting was also found to give
the best correlation up to 400 mm. Above this size the
Fig. 8: Comparison of median average sizes given by various
techniques and chord weightings for sieved/cyclosized aluminium
fractions, showing the square-weighted chord gives the best
agreement with laser diffraction and sieve analysis. FBRM
suspensions 5% w/v.
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
application of a cube-weighting may improve the esti-
mation of larger particle sizes.
The mean, and to a lesser extent also the median,
averages were affected by the instrument×s sensitivity to
fine particles. However, the modes (distribution max-
ima) were less affected. Figure 9 shows direct compar-
ison of average sizes by the techniques for a series of
sieved or cyclosized suspensions. Linear correlation
equations (Figure 9, see labels) could be re-arranged to
give a 2-parameter calibration, to give reasonable agree-
ment over the range 20 to 500 mm.
4.4 Effect of Solid Volume Fraction on Indicated Size
Early studies with FBRM showed that the spherical-
equivalent size was essentially unaffected by solid
fraction [17, 27]. More recently, Daymo et al. [57] used
the chord cube-weighting and also found little variation
in the indicated size as the solid fraction was varied.
Figure 10a shows that increasing the solid volume
fraction has the effect of increasing the sensitivity of
the unweighted chord length distribution to fine par-
ticles, probably due to fines crowding the measurement
zone near the window. However, this effect is largely
overcome by applying a chord square-weighting to the
distribution (Figure 10b). Over most of the volume
fraction range the position of the mode for the square-
weighted distribution remains constant at 140 mm, dis-
playing a shift only at 20% w/v (to 110 mm).
Fig. 9: Comparison of mode average sizes from the FBRM
square-weighted distributions, with median values from laser
diffraction and electrical sensing zone analysis for sieved/cyclo-
sized calcite particles (5% w/v).
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
Fig. 10: Effect of solid fraction on the unweighted and square-
weighted chord length distributions for aluminium particles sieved
125 ± 150 mm.
The effect of solid volume fraction was studied over a
wide range of sizes (Figure 11). The mode of the square-
weighted chord length remained essentially constant
over the entire range, except for some deviation at large
size fractions. This may be due to release of fines from
these larger particles as a consequence of stirring.
The chord-to-diameter models described in Section 2.2.1
(Equation 2) are based on the unweighted chord
distribution. As shown in Figure 10, the unweighted
distributions are dependent upon the solid fraction, and
this effect is extended to the application of Eq. (2)
(Figure 12). This limits the usefulness of such algorithms,
and further model parameters would be required to
compensate. The calculated distributions are also notice-
ably broader than the laser diffraction volume-weighted
distribution, and are likely to be worse for particles that
deviate further from the sphericity assumed by the
model.
4.5 Effect of Solid Fraction on Particle Counts
Estimation of total particle number is desirable for mass
balancing or estimation of aggregate porosity. Unfortu-
nately, total FBRM counts did not correlate well with
91
Fig. 11: Effect of solid fraction on the square-weighted mode of
the chord length for sieved calcite particles (5% w/v).
Fig. 12: Effect of solid concentration (% w/v) on transformed
chord data for aluminium particles (sieved 38-45 mm), showing
that the model (Equation 2) over-estimated the particle size in
dilute suspensions.
solid fraction, tapering off at high particle concentrations
(Figure 13a). Similar results are reported by Barrett and
Glennon [22] and Daymo et al. [57]. This effect is
probably due to an increase in the instrument dead-
time at higher solid fraction, because the instrument can
only count one particle at a time. At a low solid fraction
only a small portion of the 1.9 m s 1 scanned by the laser
intersects a particle, and the likelihood of particles
overlapping in the viewing zone is low. However, at a
higher solid fraction, the high number of counts becomes
a significant portion of the 1.9 m s 1, reducing the time
available to detect other particles in the measurement
zone. This dead-time will be slightly larger than the time
spent traversing the particle, because the reflected light
must return to zero for a short period between particles.
The signal processing algorithm also requires the re-
flected light pulse to rise and fall quickly before and after
each particle, and counts from overlapping particles (or
92
Fig. 13: (a) Total counts and (b) total corrected counts as a
function of solid fraction for various sieved calcite fractions.
particles outside the viewing zone) are rejected, further
increasing the dead-time.
It has been suggested that the measured counts can be
corrected for such effects [8, 58]. Figure 13b shows the
effect of correcting for the dead-time according to
Equation 4. A parameter (k) is included to account for
the additional dead-time required before (and after) the
laser intersects the particle:
1:9  106
Nt ˆ Nt  M  : …4†
P

1:9  106 Ni Ci;A ‡ 2k
iˆ1
The dead-time may then be simply derived from k and
the laser velocity. The optimum value for k was found
numerically by maximising the sum of the correlation
coefficients to produce the most linear relationships. For
calcite particles the dead-time was found to be 13.8 ms,
compared to 5.74 ms for the aluminium particles. Al-
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
though the dead-time varies with the substrate, the value
of Equation 4 is that it allows the estimation of solid
fraction from FBRM counts for a particular substrate.
4.6 Effect of Particle Material
Particle shape [22, 28] and reflectivity [16, 17] were
expected to affect the reflection of the laser beam to
some extent. However, Daymo et al. [57] added food
colouring to a suspension of 4% w/v bentonite and
reported no change in the average size. Sparks and
Dobbs [17] studied various particles and found that
transparent particles (oil droplets, latex particles) were
generally poorly sized by FBRM compared to opaque,
reflective particles. The suspension of transparent mono-
sized spherical latex spheres (19.9 mm, Duke Dyna-
spheres), used in this study was also poorly sized by
FBRM, giving unweighted and square-weighted mean
chord lengths of 26 and 51 mm, respectively.
The aluminium and calcite particles used in this study
have quite distinct morphologies (Figure 3), but similar
densities (ral. ˆ 2.702, rcal. ˆ 2.710), giving them similar
volume fractions when made up by weight, and similar
flow requirements to keep the particles suspended.
Figure 14 compares aluminium and calcite particles at
the same solid fraction and size. The increased fine
counts shown by the unweighted distribution for calcite
is largely caused by fine material abraded from surfaces
during suspension mixing. The count rate increased in
this region with time, but was reduced by washing and
decantation. In addition, the sharper edges of the calcite
may result in a greater proportion of fine chords. The
equivalent aluminium particles, with more spherical
shape, and no fines, gave a more symmetrical unweighted
chord distribution.
Fig. 14: Comparison between chord length distributions of
aluminium or calcite (sieved 45 ± 53 mm, 5% w/v).
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
Fig. 15: Comparison between square-weighted median FBRM
results given by various sieved or cyclosized calcite and aluminium
fractions. FBRM suspensions 5% w/v.
In contrast, the corresponding square-weighted chord
distributions for the aluminium and calcite fractions are
remarkably similar (Figure 14) despite the different
morphologies (and presumably different reflectivities)
of the two materials.
A comparison of the median square-weighted chord
lengths across all the sieved fractions (Figure 15) also
shows that the FBRM results were essentially unaffected
by particle material. This is certainly not expected to be
applicable to all materials, but is more a consequence of
the relatively low aspect ratio of the two materials
studied.
5 Conclusions
The FBRM instrument provides rapid on-line, in situ,
particle characterisation of concentrated pulps. While
FBRM gives a particle chord distribution, as opposed to
a conventional diameter-based distribution, mean or
mode averages of the square-weighted chord lengths are
comparable to conventional sizing techniques over the
range approximately 50 ± 400 mm.
This range may be extended by using the length-
weighted average for smaller particles, and by applying
a cube-weighting to larger particles. However, using a
lower weighting increases the instrument×s susceptibility
to changes in solid fraction, while large particles may be
difficult to size accurately due to obscuration by fines.
Although the latter effect also appears to result in a poor
correlation between total particle counts and solid
93
fraction, correction for the instrument dead time can
substantially improve this relationship. FBRM square-
weighed average sizes were found to be essentially
independent of suspension bulk flow velocity, focal
position, or solid fraction in the range 0.1 ± 20% w/v.
6 Acknowledgements
This research has been supported by the Australian
Government×s Cooperative Research Centre (CRC)
program, through the AJ Parker CRC for Hydrometal-
lurgy. This support is gratefully acknowledged. One of us
(A.R. Heath) is grateful for support from the AMIRA
P266C ™Improving Thickener Technology∫ project.
Thanks are also extended to Phan Khanh and Mark
Schibeci for providing alternative particle sizing, and to
Dean Ilievski, Peter Austin and Rich Becker for useful
discussions.
7 Symbols and Abbreviations
Ci, u length of the upper boundary of the ith channel
(mm)
Ci, l length of the lower boundary of the ith channel
(mm)
Ci,A geometric average length of ith channel (mm) ˆ
(Ci, u  Ci, l)1/2
Dp diameter of pth particle (mm)
k additional dead time before (and after) particle
(mm)
minf objective function
M maximum number for i
Ni number of i sized chord counts observed (s 1)
Ni* number of i sized chord counts predicted (s 1)
Ni, n number of n-weighted chord counts in the ith
channel (mmn s 1)
Np* number of pth sized particles predicted (s 1)
Ni, o raw unweighted chord counts in the ith channel (s 1)
Nt* corrected total chord counts (s 1)
Nt total observed chord counts (s 1)
n exponent (0: unweighted; 1: length-weighted; 2:
square-weighted; 3: cube-weighted)
Wp, i width of pth particle giving chords in the ith channel
(mm)
8 References
[1] T. Allen: Particle Size Measurement. 4th edition, Chapman
and Hall, London 1990.
[2] H. G. Barth, S.-T. Sun: Particle size analysis. Anal. Chem. 63
(1991) 1R-10R.
94
[3] N. G. Stanley-Wood, R. W. Lines: Particle Size Analysis.
Royal Soc. Chem., Cambridge 1992.
[4] J. B. Farrow, L. J. Warren: Measurement of the size of
aggregates, in B. Dobias (ed.): Coagulation and Flocculation.
Marcel Dekker, New York 1993, pp. 391 ± 426.
[5] R. H. Perry, D. W. Green: Perry×s Chemical Engineers Hand-
book. 7th Edition, McGraw-Hill, New York 1997.
[6] P. J. Coghill, M. J. Millen, S. Rainey, B. D. Sowerby: On-line
particle size analysis of fine mineral slurries. Sixth Mill
Operator×s Conference, AusIMM, Madang, 7 ± 9 October
1997.
[7] A. J Bale, A. W. Morris: In situ measurement of particle size
in estuarine waters. Estuar. Coast. Shelf Sci. 24 (1987) 253 ±
263.
[8] F. Hobbel, R. Davies, F.W. Rennie, T. Allen, L. E. Butler, E. R.
Waters, J. T. Smith, R. W. Sylvester: Modern methods of on-
line size analysis for particulate process streams. Part. Part.
Syst. Charact. 8 (1991) 29 ± 34.
[9] A. J. Bale: In situ laser optical particle sizing. J. Sea Res. 36
(1996) 31 ± 36.
[10] J. M. Phillips, D. E. Walling: Measurement in situ of the
effective particle-size characteristics of fluvial suspended
sediment by means of a field-portable laser backscatter
probe: some preliminary results. Mar. Freshwater Res. 46
(1995) 349 ± 357.
[11] C. Heffels, R. Polke, M. Radle, B. Sachweh, M. Sch‰fer, N.
Scholz: Control of particulate processes by optical measure-
ment techniques. Part. Part. Syst. Charact. 15 (1998) 211 ±
218.
[12] T. E. Fakatselis: On-line crystal size control using three
dimensional chord length characterization measurements.
Proc. AIChE Annual Meeting 1994, San Francisco, CA,
1994.
[13] R. J. Farrell, Y. Tsai: Nonlinear controller for batch crystal-
lization: development and experimental demonstration.
AIChE J. 41 (1995) 2318 ± 2321.
[14] S. J. Peng, R. A. Williams: Control and optimisation of
mineral flocculation and transport using on-line particle
size analysis. Minerals Eng. 6 (1993) 133 ± 153.
[15] J. J. F. M. Dijkstra, L. H. de Nie, A. Pot: In-line measurement
of crystal size. Campaign experience with the Lasentec M200
system. Int. Sugar J. 98(1174) (1996) 521 ± 523.
[16] K. J. Reid, M. S. Stachowicz: Evaluation of the Par-tec 200
particle size analyser, in R. K. Rajamani, J. A. Herbst (eds.):
Control ×90 ± Mineral and Metallurgical Processing. Soc.
Min., Metall., Expl. Littleton, Colorado 1990, pp. 143 ± 149.
[17] R. G. Sparks, C. L. Dobbs: The use of laser backscatter
instrumentation for the on-line measurement of the particle
size distribution of emulsions. Part. Part. Syst. Charact. 10
(1993) 279 ± 289.
[18] O. Monnier, G. Fevotte, C. Hoff, J. P. Klein: Model identi-
fication of batch cooling crystallizations through calorimetry
and image analysis. Chem. Eng. Sci. 52 (1997) 1125 ± 1139.
[19] B. K. Johnson, C. Szeto, O. Davidson, A. Andrews: Optimi-
zation of pharmaceutical batch crystallization for filtration
and scale-up. AIChE Annual Meeting, 16 ± 21 Nov. 1997, Los
Angeles.
[20] V. Z. Poilov, K. Kosvintsev, J. Gramlich, J. Kodura, R. Karvot:
Formation of magnesium chloride particles in bulk poly-
thermal crystallization. Russ. J. Appl. Chem. 70 (1997) 1036 ±
1040.
[21] D. Bongartz, A. Schneider, A. Hesse: Direct measurement of
calcium oxalate nucleation with a laser probe. Urol. Res. 27
(1999) 135 ± 140.
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
[22] P. Barrett, B. Glennon: In-line FBRM monitoring of particle
size in dilute agitated suspensions. Part. Part. Syst. Charact.
16 (1999) 207 ± 211.
[23] K. L. Hildred, P. S. Townson, G. V. Hutson, R. A. Williams:
Characterisation of particulates in the BNFL enhanced
actinide removal plant. Powder Technol. 108 (2000) 164 ±
172.
[24] J. M. Phillips, D. E. Walling: An assessment of the effects of
sample collection, storage and resuspension on the repre-
sentativeness of measurements of the effective particle size
distribution of fluvial suspended sediment. Wat. Res. 29
(1995) 2498 ± 2508.
[25] J. M. Phillips, D. E. Walling: Calibration of a Par-Tec 200
laser back-scatter probe for in situ sizing of fluvial suspended
sediment. Hydrol. Process. 12 (1998) 221 ± 231.
[26] J. M. Phillips, D. E. Walling: The particle size characteristics
of fine-grained channel deposits in the river Exe basin,
Devon, UK. Hydrol. Process. 13 (1999) 1 ± 9.
[27] D. J. Law, A. J. Bale, S. E. Jones: Adaptation of focused beam
reflectance measurement to in-situ particle sizing in estuaries
and coastal waters. Marine Geol. 140 (1997) 47 ± 59.
[28] R. A. Williams, S. J. Peng, A. Naylor: In situ measurement of
particle aggregation and breakage kinetics in a concentrated
suspension. Powder Technol. 73 (1992) 75 ± 83.
[29] S. J. Peng, R. A. Williams: Direct measurement of floc
breakage in flowing suspensions. J. Colloid Interface Sci.
166 (1994) 321 ± 332.
[30] D. R. Spears, D. A. Stanley: A study of shear-flocculation of
silica. Miner. Metall. Process. 11 (1994) 5 ± 11.
[31] D. K. Sengupta, J. Kan, A. M. Al Taweel, H. A. Hamza:
Dependence of separation properties on flocculation dynam-
ics of kaolinite flocculation. Int. J. Miner. Process. 49 (1997)
73 ± 85.
[32] M. Caron-Charles, J. P. Gozlan: Improvement of the floc
resistance to a centrifugal shear field by polymer adjunction.
Chem. Eng. Sci. 51 (1996) 4649 ± 4659.
[33] D. D. Murphy, M. G. Bakker, D. R. Spears: New insights into
the mechanism of shear-flocculation of silica with fatty acids.
Miner. Metall. Process. 11 (1994) 26 ± 30.
[34] D. D. Murphy, D. R. Spears, M. G. Bakker: The role of
micellar structures in shear flocculation. Part 1: Flocculation
of quartz with alkylamine surfactants. Colloids Surf. A:
Physicochem. Eng. Aspects 96 (1995) 143 ± 154.
[35] S. K. Sharma, D. A. Stanley, J. Harris: Determining the effect
of physicochemical parameters on floc size using a popula-
tion balance model. Miner. Metall. Process. 11 (1994) 168 ±
173.
[36] P. Fawell, W. Richmond, L. Jones, M. Collisson: Focussed
beam reflectance measurement in the study of mineral
suspensions. Chem. in Aust. 64 (2) (1997) 4 ± 6.
[37] J. Hokanson, E. Preikschat: In-line particle size measurement
for improved process control in the mineral processing
industry, in E.R. Cutshall (ed.): Light Metals 1992. Miner.
Met. Mater. Soc. 1992, pp. 39 ± 42.
[38] R. Q. Honaker, G. H. Luttrell, R. H. Yoon: In-situ measure-
ment of coagula size using an on-line particle size analyzer.
SME Annual Meeting, Albuquerque, New Mexico, 14 ± 17
Feb., 1994.
[39] R. Hecker, L. Kirwan, A. Jefferson, P. D. Fawell, J. B. Farrow,
J. D. Swift: Focused beam reflectance measurement for the
continuous assessment of flocculant performance, in J. S.
Laskowski (ed.): Polymers in Mineral Processing. Canad.
Inst. Min. Metall. Petrol., Montreal 1999, pp. 91 ± 105.
[40] J. C. Alfano, P. W. Carter, J. E. Whitten: Polyelectrolyte-
induced aggregation of microcrystalline cellulose: reversibil-
Part. Part. Syst. Charact. 19 (2002) 84 ± 95
ity and shear effects. J. Colloid Interface Sci. 223 (2000) 244 ±
254.
[41] D. N. Thomas, S. J. Judd, N. Fawcett: Flocculation modelling
of primary sewage effluent, in H. H. Hahn, E. Hofman, H.
Èdegaard (eds.): Chemical Water and Wastewater Treatment
V: Proceedings of 8th International Gothenburg Symposium
1998. Springer, Berlin 1998, pp. 83 ± 98.
[42] J. Hanseler, W. McKean: Laser technology offers new way to
measure furnish components. Pulp Paper Can. 89 (1988) 25 ±
32.
[43] A. Blanco, C. Negro, A. Hooimeijer, J. Tigero: Polymer
optimization in paper mills by means of a particle size
analyser: an alternative to zeta potential measurements.
Appita 49 (1996) 113 ± 116.
[44] J. E. Anderson: Evaluation of the Lasentec laser microscope
for monitoring refining. Papermakers Conference. TAPPI
Press, Atlanta 1996, pp. 225 ± 235.
[45] J. C. Alfano, P. W. Carter, J. E. Whitten: Use of scanning laser
microscopy to investigate microparticle flocculation per-
formance. J. Pulp Paper Res. 25(6) (1999) 189 ± 195.
[46] B. Cowie, J. Schmoll: Investigation of a novel facility design
concept for heavy oil North Sea development. Off-shore
Technical Conference, Houston, Society for Petroleum En-
gineering, 5 May, 1994.
[47] A. M. El-Hamouz, A. C. Stewart: On-line measurement of
oil-water dispersion using a Par-Tec M300 laser backscatter
instrument. SPE Annual Technical Conference, Denver 1996,
pp. 785 ± 796.
[48] M. J. H. Simmons, P. A. Langston, A. S. Burbidge: Particle
and droplet size analysis from chord distributions. Powder
Tech. 102 (1999) 75 ± 83.
[49] H. X. Vo: Use of an on-line droplet size analyzer to design
and scale-up a process for mixing silicone oil and water.
Society of Cosmetic Chemists, Chicago, 13 September, 1994.
[50] R. Daniels, A. Barta: Preparation, characterization and
stability assessment of oil-in-water emulsions with hydrox-
95
ypropylmethyl cellulose as emulsifier. Lasentec User Meet-
ing, Williamsburg, Virginia, 20 ± 21 October, 1992.
[51] O. Monnier, J. P. Klein, C. Hoff, B. Ratsimba: Particle size
determination by laser reflection: methodology and prob-
lems. Part. Part. Syst. Charact. 13 (1996) 10 ± 17.
[52] N. N. Clark, R. Turton: Chord length distributions related to
bubble size distributions in multiphase flow. Int. J. Multi-
phase Flow 14(4) (1988) 413 ± 424.
[53] A. J. Jakeman, R. S. Anderssen: Abel type integral equations
in stereology. I. General discussion. J. Microscopy 105(2)
(1975) 121 ± 133.
[54] W. Liu, N. N. Clark: Relationships between distributions of
chord lengths and distributions of bubble sizes including their
statistical parameters. Int. J. Multiphase Flow 21 (1995)
1073 ± 1089.
[55] W. Liu, N. N. Clark, A. I. Karamavruc: Relationship between
bubble size distributions and chord-length distribution in
heterogeneously bubbling systems. Chem. Eng. Sci. 52 (1998)
1267 ± 1276.
[56] A. Tadayyon, S. Rohani: Determination of particle size
distribution by Par-Tec 100: modeling and experimental
results. Part. Part. Syst. Charact. 15 (1998) 127 ± 135.
[57] E. A. Daymo, T. D. Hylton, T. H. May: Acceptance testing of
the Lasentec focussed beam reflectance measurement
(FBRM) monitor for slurry transfer applications at Hanford
and Oak Ridge, in D. E. Robertson (ed.): Nuclear Waste
Instrumentation Engineering. SPIE ± Int. Soc. Opt. Eng.,
1999, pp. 82 ± 92.
[58] R. Becker: Personal communication.
[59] A. D. Randolph, J. A. Larson: Theory of Particulate Proc-
esses. Academic Press, San Diego 1988.
[60] A. Rushton, A. S. Ward, R. G. Holdich: Solid-Liquid Filtra-
tion and Separation Technology. VCH, Weiheim 1996.
[61] J. P. K. Seville, U. T¸z¸n, R. Clift: Processing Particulate
Solids. Blackie Academic & Professional, Melbourne, Aus-
tralia 1997.